254 Commits
Author SHA1 Message Date
Tanguy Pruvot 6ff4e50987 v2.3.1 release 2019-01-30 16:08:42 +01:00
Tanguy Pruvot 9a1f20d455 Handle lyra2v3 algo, for VTC fork
mostly imported from opensourced vertcoin-miner with a few fixes
2019-01-30 15:27:43 +01:00
pyritepirate c59bc2438a sha256q (#70) 2019-01-27 08:24:53 +01:00
Tanguy Pruvot 01e632cf05 handle standard blake2b stratum algo
no weird protocol or reversed endian like sia...
2018-11-30 08:05:51 +01:00
opensourcerulez b36d174554 Add exosis algo (#69) 2018-10-22 21:11:33 +02:00
Tanguy Pruvot 1f5efa7d36 makefile: add new cuda arch and remove sm5.0 by default 2018-09-20 20:14:13 +02:00
Tanguy Pruvot 4a76ca5cb6 bench: handle cryptonight variants + V100 fix 2018-08-03 20:08:34 +02:00
Tanguy Pruvot 370684f743 cryptonight: some code finitions 2018-06-24 12:25:42 +02:00
Tanguy Pruvot 654e8a10ec fix g++ 7.3 warnings (ubuntu 18.04) 2018-06-24 12:12:55 +02:00
Tanguy Pruvot d9f242b8d1 add sonoa algo, heavy x17 hashes
seems to works, more or less correctly (a few validation errors)
2018-06-23 17:51:07 +02:00
Tanguy Pruvot 6dc1bbdd47 prepare the new release 2018-06-23 13:39:38 +02:00
Tanguy Pruvot 77c4b8724e handle new cryptonight variants, stellite, aeon
special thanks for klausT changes and ystarnaud who helped me to adapt my kernel variants...

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2018-06-23 13:30:23 +02:00
Tanguy Pruvot 968d2ba049 phi2: fix the double endian swap on roots 2018-06-21 10:12:18 +02:00
Tanguy Pruvot 9fd5b04af6 phi2: handle stratum sc hashes 2018-06-18 20:43:12 +02:00
Tanguy Pruvot 07859f93ce update windows version + 2.2.6 readme 2018-06-10 18:32:37 +02:00
Tanguy Pruvot 3d03a1b9fd phi2 algo 2018-05-28 15:21:54 +02:00
Tanguy Pruvot b8190e4aa7 allium: add missing device cpu flag for linux 2018-05-17 19:29:39 +02:00
Tanguy Pruvot a9357e1ec8 lyra2: remove unused nonce param 2018-05-10 06:35:28 +02:00
Tanguy Pruvot 57f8f776fb timetravel: cleanup, remove unused algos
+ cubehash 80 midstate
2018-05-09 16:18:23 +02:00
Tanguy Pruvot ffd6cf38bf update readme and win ver 2018-05-06 18:18:05 +02:00
Tanguy Pruvot b97567a451 allium algo 2018-05-06 18:09:41 +02:00
Tanguy Pruvot bfcf7a9e52 neoscrypt: add extra space for recent vstudio madness 2018-04-22 23:12:37 +02:00
Tanguy Pruvot 16ebe53b72 x12: apply echo512 optimised kernel on recent cards 2018-04-09 17:54:13 +02:00
Tanguy Pruvot 5d0c0a665d x17: apply echo512 improvement
add a tiny 1% on x17, better than nothing...
2018-04-07 12:31:18 +02:00
Tanguy Pruvot a81ab0f7a5 readme for 2.2.5 release 2018-04-02 08:24:59 +02:00
Tanguy Pruvot ff73bda584 scrypt stuff 2018-04-02 08:24:59 +02:00
Tanguy Pruvot 2fdc10a39c x16r/s: move to a common x16 folder 2018-04-02 08:24:54 +02:00
Tanguy Pruvot 13e333d3e2 x16: optimised echo512-64 kernel for maxwell+
which keep the SM 3.0 compat with x11 kernel
2018-04-02 08:15:33 +02:00
Tanguy Pruvot 27199e2a4d api: missing break on mcast argument 2018-03-31 19:34:22 +02:00
Tanguy Pruvot 96adaf986c x16s algo, compatible with SM3+ 2018-03-29 06:31:02 +02:00
Tanguy Pruvot 5b1b77ba95 api: fix possible issue on linux 2018-03-29 02:32:10 +02:00
Tanguy Pruvot 0961544c85 x12: reorder cpu code to notice its not x11 one 2018-03-26 06:52:50 +02:00
Tanguy Pruvot 51c02a7e9a x12: a bit more code cleanup 2018-03-26 06:10:49 +02:00
Tanguy Pruvot 84337459b8 x12 stuff + readme 2018-03-26 03:14:21 +02:00
Galaxy Cash Developers e21640545f x12 algorithm (#59) 2018-03-26 03:08:43 +02:00
Tanguy Pruvot fd406f02c2 equihash: error fix for the TITAN V 2018-03-15 22:03:35 +01:00
Tanguy Pruvot d8a64d4bd2 neoscrypt: improve perfs and fix the Titan default intensity 2018-03-15 21:08:44 +01:00
Tanguy Pruvot 78dad7dd65 x16r algo and new kernels
Was a very long work but finally working,
and unlike xevan these new kernels are reusable..

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2018-02-04 08:06:44 +01:00
Tanguy Pruvot b71f85ec31 stratum: show reason of auth failure 2018-01-16 10:35:37 +01:00
hakemimi 50781f00eb lyra2: Translate japanese comments into english (#40)
* Revert "lyra2: fix compilation on japanese windows (#38)"

This reverts commit 30db7d2433.

* Restore comments removed in #38, translate them into English

The original line says "allocate XXXX bytes to adjust for X Warp."
Remove the number from comment since they are trivial

* Translate Japanese comments
2018-01-10 12:24:50 +01:00
Guillaume George 30db7d2433 lyra2: fix compilation on japanese windows (#38)
Remove comment that prevent compilation on japanese windows (#38)
2018-01-09 18:32:12 +01:00
Tanguy Pruvot 3761774ccf cuda: get ride of cuda 9 mask warnings 2018-01-08 14:08:23 +01:00
Chris Spillane f1a7de4e9e Update README.txt (#37)
remove duplicate x17 line, add a few more coin examples.
2018-01-05 11:56:54 +01:00
Tanguy Pruvot b54be470e1 upgrade BIGNUM class for openssl 1.1 2018-01-04 18:14:57 +01:00
Tanguy Pruvot 69e12962fc update README for 2.2.4 2018-01-04 16:07:48 +01:00
Tanguy Pruvot 73dd6aac5c keccak: avoid to use twice cuda_default_throughput
and drop useless gpu hash alloc...
2018-01-04 15:58:01 +01:00
Tanguy Pruvot 11a512f2a7 change defaults to handle cuda 9+, disable heavy and SM 2.x
Heavy is the only algo using thrust which is generally broken on new cuda releases

mjollnir dropped too... never seen this coin anyway...
2018-01-04 15:37:01 +01:00
Tanguy Pruvot 91af8ab6e6 decred doesnt allow pow votes 2018-01-04 15:19:20 +01:00
Tanguy Pruvot 6165562839 equihash: fix for cuda 9.1 build 2018-01-04 14:59:31 +01:00
dementeb 72466ee476 api: update websocket sample (#31) 2018-01-04 14:58:36 +01:00
Tanguy Pruvot b70409ab5b lyra2RE: link the merged blake/keccak kernel into algos
old keccak256_gpu_hash_32 kernel commented to reduce binary size

compat. not yet tested on old cards
2017-12-14 18:21:40 +01:00
Myrinia 18d29914ec Improve Lyra2RE2 Performance
Improved Lyra2Re2 Performance by 1 %
2017-12-14 18:21:04 +01:00
Tanguy Pruvot 6c0e656030 keccak: fix issue with intensity 2017-12-09 16:54:48 +01:00
Tanguy Pruvot 015d129aa6 keccak second nonce, and higher intensity 2017-12-04 21:52:11 +01:00
Tanguy Pruvot cf886b5907 import and adapt alexis optimised keccak256 for SM 5+
and increase default intensity for these recent cards
2017-12-04 16:38:20 +01:00
Tanguy Pruvot d19e2a1e62 Add keccakc for creative coin
Same hash as keccak, just different pool settings (sha256d and diff factor 256)
2017-12-03 16:52:21 +01:00
Tanguy Pruvot df4fcbe3a4 jha: remove deprecated warning 2017-11-20 08:50:40 +01:00
Tanguy Pruvot bfa616a2a4 linux: common openssl lib function for 1.1+ 2017-11-20 07:58:55 +01:00
Tanguy Pruvot dfff3b3f98 windows: high precision timer before threads start 2017-11-20 07:30:14 +01:00
Tanguy Pruvot 2e0a977784 polytimos algo (6 chained algos with streebog) 2017-11-16 10:42:35 +01:00
Tanguy Pruvot e1575c5cb2 update README for v2.2.2 2017-10-09 20:11:22 +02:00
Tanguy Pruvot ed27598a7a import and clean hsr algo 2017-10-09 20:09:29 +02:00
Tanguy Pruvot cf18cb632c equihash: missing free on benchs 2017-10-09 15:54:54 +02:00
Tanguy Pruvot e505d1051d new --submit-stale parameter (#24) 2017-10-09 15:54:54 +02:00
Tanguy Pruvot 5a90db192a phi: maxwell opt (aes final + streebog)
+ fix the fugue leak..

Also update sib algo with this improvement
2017-10-09 15:54:48 +02:00
Tanguy Pruvot 3dbcc5dbbf Import phi algo (by anorganix)
only made a few changes to keep algos in the alpha order
2017-10-07 11:18:15 +02:00
Tanguy Pruvot 575aa1a0e4 ccminer: argument fix for 10th device selection 2017-10-05 13:18:24 +02:00
Tanguy Pruvot 4d09d856f6 readme: move ubuntu stuff in INSTALL file, no ads 2017-09-20 05:08:23 +02:00
cedric walter e3e848149c readme: add howto for ubuntu (#23)
tested on 17.10
2017-09-19 09:41:58 +02:00
Tanguy Pruvot ebf055d585 update c11 like tribus + 2.2.1 readme 2017-09-01 20:43:52 +02:00
Tanguy Pruvot d47dd9de39 tribus: optimised version for recent cards
main improvement is on echo, based on alexis sib kernel work

tested on SM 3.0 and more recent
2017-09-01 20:15:17 +02:00
Tanguy Pruvot 949061d223 xmr: disable huge pages if not defined #19 2017-08-27 19:17:57 +02:00
Tanguy Pruvot 1e71dc5782 streebog: apply skunk improvements to veltor 2017-08-16 13:43:10 +02:00
Tanguy Pruvot 6055af1fa9 skunk: proper indent and trim spaces 2017-08-14 04:06:03 +02:00
Tanguy Pruvot dfdf4b0691 now works with all CUDA versions 2017-08-13 21:04:56 +02:00
Tanguy Pruvot 5001a0cd1b update the README 2017-08-13 19:53:03 +02:00
Tanguy Pruvot 2e56177a2d small changes for the 2.2 release 2017-08-13 19:42:21 +02:00
Tanguy Pruvot 9a7d74eac4 skunk: merged kernel to use when possible
for now, only used for maxwell/pascal with cuda 7.5 or 8.0
2017-08-13 19:01:18 +02:00
Tanguy Pruvot 6e32ec2be0 wildkeccak: increase scratchpad buffer size + proxy 2017-08-07 12:31:08 +02:00
Tanguy Pruvot 9f113effd8 streebog: use TPB of 128 to reduce errors 2017-07-28 04:43:19 +02:00
Tanguy Pruvot 9c053042c9 cubehash: improve a bit and cleanup 2017-07-27 23:53:22 +02:00
Tanguy Pruvot 08a3f3402c readme: remove alexis addresses, on demand 2017-07-26 21:11:29 +02:00
Tanguy Pruvot 9c4b8caf1a nvapi: apply memory offsets like nvsettings 2017-07-24 14:16:14 +02:00
Tanguy Pruvot cbede12c47 Add skunk algo, initial version 2017-07-21 22:41:05 +02:00
Tanguy Pruvot 5aa50a4215 equihash: use client.show_message to get block height
current stratum is imcomplete and doesnt provide the coinbase data
2017-07-13 19:41:47 +02:00
Tanguy Pruvot 7ce3f2ccce tribus: handle jh80 midstate (+10%) 2017-07-07 18:15:08 +02:00
Tanguy Pruvot 807e141a4d equi: sm 2 test 2017-06-22 09:35:32 +02:00
Tanguy Pruvot 9dedec5ecd sysinfos: fix some file leaks 2017-06-22 07:38:40 +02:00
Tanguy Pruvot c120ecae1d handle the new tribus algo
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-06-15 07:32:04 +02:00
Tanguy Pruvot 1473a8fd6f nvs: fix fscanf return warning 2017-06-15 07:30:11 +02:00
Tanguy Pruvot 79f8f71fc3 fedora: note about the alternative method via rpms 2017-06-15 04:47:16 +02:00
Tanguy Pruvot 037bbe3ad3 sia: add missing cputest 2017-06-15 03:48:59 +02:00
Tanguy Pruvot 0f3fac9536 nvs: remove test with -n 2017-06-12 06:19:38 +00:00
Tanguy Pruvot bcadca2c45 linux: prepare nvidia-settings queries
Note: this method is slower than real apis and requires X opened
and configured correctly.

sample usage: -d 0,1 --mem-clock=+200,-200
2017-06-08 22:07:29 +02:00
Tanguy Pruvot 05abaf2b26 update readme and sample config 2017-05-24 00:11:27 +02:00
Tanguy Pruvot 21600491b8 update instructions for fedora 25
its a shame this distrib doesnt provide older gcc toolchains.. :/
2017-05-23 23:54:35 +02:00
Tanguy Pruvot 8cf21599d4 Interface nicehash djeZo equihash solver (squashed)
Todo:
    - send block height via stratum protocol (encoded in jobid?)
    - remove equi/blake2 cpu algorithm to use common one

the extranonce imcompatibility is related to the solver nonce data,
offsets may be reversed in nheqminer, to check...

The solver was adapted for SM 3.0+ support (no perf changes)

Note: The solver was not improved on purpose, to be able compare
the two miners performances (nheqminer 0.5c the last open sourced, and ccminer)

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>

stratum: code cleanup, move equi fns in equi folder
2017-05-23 23:54:35 +02:00
Tanguy Pruvot c44be3f81d api: various fixes
- prefer calloc to zero memory
- bind: keep the normal network read access with -b 0.0.0.0 like previous versions
- help: keep the write mode commands hidden in read-only mode
- close the socket on refused api calls to prevent client hangs
- small other changes
2017-05-23 23:53:58 +02:00
bitbandi 47472ed3f1 API restriction policy and multicast announces (#16)
same as sgminer, fixed to build (tpruvot), untested yet
2017-05-18 19:50:08 +02:00
Tanguy Pruvot b372729ce6 api: change api bind variable name 2017-05-16 12:09:32 +02:00
Tanguy Pruvot 857c1819c9 ccminer 2.0 final 2017-05-14 11:40:46 +02:00
Tanguy Pruvot e584f28441 fix quark and jackpot algos 2017-05-14 11:35:27 +02:00
Tanguy Pruvot 81d5f4e862 Improve project build compatibility (mac/vs2015)
This is incomplete, but is a first step...
2017-05-09 05:56:28 +02:00
Tanguy Pruvot 5dfeee45ec rewrite jha algo to be more compatible
old "german" implementation based on the quark method is kept
in the source tree, but is currently broken. to be continued.

This is a quick fix which should be compatible with all cards.
2017-05-08 08:03:55 +02:00
Tanguy Pruvot 762c7f8656 nvapi: fix plimit argument on x64
now use the same percents as the x86

nvml watts (like linux) could work but requires admin rights
2017-05-07 12:09:55 +02:00
Tanguy Pruvot e23b577e32 nvml: show product ids and pci bus in ccminer -D -n 2017-05-06 02:00:24 +02:00
Tanguy Pruvot 8e4d059a41 bitcore algo for linux users
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-05-03 11:12:50 +02:00
Tanguy Pruvot 063992b25b decred: handle new stake version field 2017-04-23 17:02:27 +02:00
Tanguy Pruvot d2f836119a vstudio: remove remains, move sha in tree
+ small code fixes
2017-04-23 17:02:27 +02:00
Tanguy Pruvot 6c9852c153 nvapi: fix weird code style 2017-04-21 13:20:31 +02:00
Tanguy Pruvot 80c755188d Add sha256d algo (bitcoin) btw...
even if it works on yiimp, the stratum diff is too high for gpus (for now)
2017-03-11 15:29:21 +01:00
Tanguy Pruvot 61231bc66c fix various memory leaks on algo switch 2017-03-11 11:19:20 +01:00
Tanguy Pruvot 0041bf47f1 wildkeccak: add credits and licence notice 2017-03-11 08:49:38 +01:00
Tanguy Pruvot 73f6720121 whirlpool: restore old source code for SM 3.0
SM 3.0 implementation need a manual define in whirlpool.cu...

alexis variant is 2x slower on SM3.0 (GT 740)
2017-03-08 15:59:51 +01:00
Tanguy Pruvot e388c11c02 blake2s fix and more missing cuda arch (for the benchmarks) 2017-03-08 13:13:52 +01:00
Tanguy Pruvot 2cdf2ddd43 Add missing real cuda arch checks 2017-03-08 09:19:10 +01:00
Tanguy Pruvot 18637503f2 whirlpool: upgrade with alexis variant (2x)
to check on older cards... maybe breaks some
2017-03-08 08:04:56 +01:00
Tanguy Pruvot de41bc9d21 update readme and use cuda 8 as default 2017-03-07 21:14:15 +01:00
Tanguy Pruvot 07ebcb544d timetravel algo
+ new kernels jh512-80 groestl-80 and cubehash-80

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-03-07 20:16:15 +01:00
Tanguy Pruvot 3ede61b0cc bastion algo
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-03-07 20:10:26 +01:00
Tanguy Pruvot 1b7c2fc296 lyra2z algo (temporary algo)
based on djm34 version, cleaned up and adapted to ccminer 2.0
2017-03-07 20:05:03 +01:00
Tanguy Pruvot ebef3c8656 sha256t algo (3x sha256, not sha256d)
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-03-07 19:32:59 +01:00
Tanguy Pruvot 3d70026ae6 hmq1725 algo
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-03-07 16:26:27 +01:00
Tanguy Pruvot 4d95ef4aec cuda: ensure the system can find the right lib
if you get this error on linux, its because libcudart.so.x.x was not found

ldconfig -p | grep libcudart.so.8.0
2017-02-17 19:00:51 +01:00
Tanguy Pruvot 1d5db48797 credits: show binary arch (32/64) in the header 2017-02-17 17:48:25 +01:00
Tanguy Pruvot b549bd54fe lyra2: clean the intelli-non-sense stuff 2017-02-08 12:21:36 +01:00
Tanguy Pruvot c66e8622b3 api: report per thread cpu hash checks (ACC/REJ)
+ update all algos for that...
2017-02-07 06:26:02 +01:00
Tanguy Pruvot 4da35e0bcf api: use MHz for all clocks and show plimit
plimit value is reported in mW like the power, or % on Windows x86 (via nvapi)

+ the monitor thread will no more report 0W if the device doesnt support it

also upgrade nvml and sample php api.

some more changes may come in this temporary API 1.9 (for the final 2.0)
2017-02-06 06:50:38 +01:00
Tanguy Pruvot 3fab2792fd lyra2: remove old debug traces 2017-02-06 04:38:43 +01:00
Tanguy Pruvot 916ec5047e logs: show solved count on solved shares 2017-02-05 14:06:51 +01:00
Tanguy Pruvot aaa24a6262 stats: submit real freqs and plimit to yiimp benchmarks 2017-02-05 13:20:38 +01:00
Tanguy Pruvot ef44e88ef9 ccminer: link --max-log-rate to the command line 2017-01-30 12:36:50 +00:00
Tanguy Pruvot b9da6c67f5 improve jh512 with vectors (nist5,quark,sib,x11+,zr5)
the main improvement is to reduce asm calls to read global mem

but, a few more regs are used (68 mini vs 64 on SM 5.2)
so reduce the forced launch bounds to allow 80 or 128 regs per thread

Note: cuda 6.5 seems not able to store with v4.u32... (7.5 is fine)
        st.global.v4.u32        [%rd2], {%r3783, %r3824, %r3823, %r3822};
        st.global.v2.u32        [%rd2+16], {%r3821, %r3820};
        st.global.u32   [%rd2+24], %r3819;
        st.global.u32   [%rd2+28], %r3818;
        st.global.u32   [%rd2+44], %r3814;
        st.global.u32   [%rd2+40], %r3815;
        ...

todo, check alexis variant.. but wanted to keep this code before in git...
2017-01-30 07:27:01 +01:00
Tanguy Pruvot 6440a9bf41 windows: some default intensity adjustments 2017-01-30 02:31:44 +01:00
Tanguy Pruvot 2152fd102d lbry cleanup, and proper error on cuda 6.5
both merged and unmerged implementations are broken with CUDA 6.5

No perf changes...
2017-01-30 02:31:28 +01:00
Tanguy Pruvot aaef92cab2 nvml: workaround for beta drivers 378.49 clocks
even nvidia-smi doesnt report the right pascal clocks
2017-01-29 22:23:05 +01:00
Tanguy Pruvot b47d9acaf5 readme + small warnings detected by vstudio 2017-01-29 22:23:05 +01:00
Tanguy Pruvot c8ff854456 sia was migrated too... 2017-01-29 06:10:08 +01:00
Tanguy Pruvot 0ff75791e5 migrate 2nd nonce storage of most algos
This allow to keep pdata[19] as cursor between scans, and later, to sort them..

remains... heavy, scrypt, sia...
2017-01-29 05:46:45 +01:00
Tanguy Pruvot 5a77d36635 groestl: explain code and improve perf on SM 2.x
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-01-29 00:51:42 +01:00
Tanguy Pruvot feb99d020f skein: merge the double implementations in one
based on alexis skein kernels, tested ok on SM 2.1 and 3.0

code is a bit hard to read but... well... users dont care :p
2017-01-29 00:01:06 +01:00
Tanguy Pruvot 16ac9b688f x13+: improve and clean a bit fugue512
reduce constant mem and load global data in bulk (vectors)
2017-01-29 00:00:43 +01:00
Tanguy Pruvot 013cda1cd2 ccminer: show first block diff even with -q 2017-01-29 00:00:28 +01:00
Tanguy Pruvot 496052e47d xmr: vstudio warning fix about mpcount linkage
and move ptr type cast defines to common cuda helper
2017-01-25 21:16:31 +00:00
Tanguy Pruvot dc816b4673 xmr: nicehash nonce prefix/hack support (v2) 2017-01-22 23:19:22 +01:00
Tanguy Pruvot def9888bd5 xmr: prefer 32bit uint4 and smaller offsets in core
also prefer ulong2 shared load to be closer to the ptx
2017-01-16 06:36:24 +01:00
Tanguy Pruvot 214f392778 xmr: default settings with card attributes 2017-01-15 01:42:25 +01:00
Tanguy Pruvot 94aa6b8e91 ccminer: allow 192 car. for the username 2017-01-14 18:55:36 +01:00
Tanguy Pruvot 588c7ba361 xmr: dont use shared mem hack, windows dont like 2017-01-14 15:13:31 +01:00
Tanguy Pruvot bd030db5d1 xmr: vectors rewrite, now the phase2 is using only 40 regs
no more constant memory used for aes.

tested only on linux cuda 8 for now... wip
2017-01-14 04:34:23 +01:00
Tanguy Pruvot 23be7f308d xmr: link the --bfactor setting (0-11) 2017-01-12 23:02:00 +01:00
Tanguy Pruvot e231343060 xmr: make it more smooth on windows with defaults
also improve a bit the 750 ti on linux...
2017-01-12 06:00:59 +01:00
Tanguy Pruvot 12ae185594 hwmonitor: efficiency unit and clean dead code 2017-01-10 14:20:33 +01:00
Tanguy Pruvot 0dd022779b power monitoring thread + some api changes
based on alexis monitoring thread idea, but which only use one thread

note: other api changes will come soon, related to that
2017-01-10 02:26:28 +00:00
Tanguy Pruvot 242aa4144b scanlog: remember sharediff for multiple nonces
rpc2: handle properly secondary nonce(s) + api ping time fix

be sure to fully recompile, structures are changed
2017-01-09 18:40:00 +01:00
Tanguy Pruvot 93adb56c8e handle cryptonight light variant
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2017-01-09 18:30:59 +01:00
Tanguy Pruvot 39aad5a003 xmr: allow to set intensity on command line 2017-01-08 22:36:03 +00:00
Tanguy Pruvot 804b5b5f53 xmr: be safe with diff divide 2017-01-08 21:28:54 +00:00
Tanguy Pruvot 2479ffaaa2 xmr: fix decimal diff + aes cleanup
change default launch config to -l 32x16 to handle the 750 Ti better
not definitive, doing tests..
2017-01-08 21:16:45 +00:00
Tanguy Pruvot c1f1ad9280 xmr: stabilize the final kernel 2017-01-08 14:39:13 +01:00
Tanguy Pruvot 066a569357 import xmr, to finish
todo: fix jh cuda and wrong decimal diff (0xffff problem ?)
2017-01-07 01:40:59 +01:00
Tanguy Pruvot 2bbccc5ff4 wildkeccak, basic stratum port of rpc 2.0
scratchpad delete fix and redownload, reduce rejects

(work in progress)
2017-01-06 08:55:15 +01:00
Tanguy Pruvot 099389f64f ccminer: be more quiet with -q, skip header noise 2016-12-30 15:18:35 +01:00
Tanguy Pruvot 50534789bc Release 1.8.4 2016-12-21 20:35:09 +01:00
Tanguy Pruvot c11901260a limit per gpu hashrate logs to 3 sec intervals
may be required for very fast algos, like blake2s
2016-12-21 20:16:47 +01:00
Tanguy Pruvot 44bd244fc4 blake2s improved
based on alexis work, with the new work->nonces
2016-12-21 19:44:20 +01:00
Tanguy Pruvot ce6a8da188 cuda: prevent ptxas crash with -n 2016-12-18 03:35:19 +01:00
Tanguy Pruvot 397472818d prepare 1.8.4 release 2016-11-03 18:55:42 +01:00
Tanguy Pruvot 36aedbb48e veltor update, 10x faster :p
From Alexis work, sib hash rate 200% also..
2016-11-03 18:54:29 +01:00
Tanguy Pruvot 3eba451d4c nvml: add Elsa vendor and workaround for Colorful pid
Colorful (and Inno3D) only set their vid, with an empty product id
2016-10-25 03:40:07 +02:00
Tanguy Pruvot c27f3139aa update startup credits 2016-10-06 13:08:41 +02:00
Tanguy Pruvot 056098dd86 update readme 2016-09-29 03:37:36 +02:00
Tanguy Pruvot 7b82915032 cuda 8 2016-09-29 00:40:06 +02:00
Tanguy Pruvot 225f25a6b9 uint2: remove the slower asm in operators funcs 2016-09-28 07:12:05 +02:00
Tanguy Pruvot 665de3a1f2 sia: use the new work share diff 2016-09-28 01:28:04 +02:00
Tanguy Pruvot 1a31d4d2d6 sia: move specific code in a new rpc unit
part 1: longpoll stuff (nanopool)
2016-09-28 01:27:57 +02:00
Tanguy Pruvot f84c83afe5 nvml: force 64bits types for mem sizes
size_t can be a bit... imprevisible on x86
2016-09-28 00:31:13 +02:00
Tanguy Pruvot 5a0b779434 api: use the new throughput2intensity func 2016-09-27 23:20:28 +02:00
Tanguy Pruvot a43205a84f decred: multiple nonces code cleanup
The double loop is not useful, and prefer the __thread attribute
to enhance the code readability (remove the 2D host arrays).

squashed: return to host 2D array to allow the free
2016-09-27 22:50:52 +02:00
Tanguy Pruvot 6f6cf966f8 lbry: new share diff and duplicate fix
when 2 nonces were found, the next scan was not at the right value

Doesn't really affect mining performance...
2016-09-27 22:50:51 +02:00
Tanguy Pruvot 9eead77027 diff: show by default, rework shares diff storage
This will allow later more gpu candidates.

Note: This is an unfinished work, we keep the previous behavior for now
To finish this, all algos solutions should be migrated and submitted nonces attributes stored.
Its required to handle the different share diff per nonce and fix the possible solved count error (if 1/2 nonces is solved).
2016-09-27 09:03:24 +02:00
Tanguy Pruvot 2f57ee9157 bench: skip the disabled whirlpoolx
+ veltor free
+ some missed/extra log things...
2016-09-27 01:41:49 +02:00
Tanguy Pruvot 009b013d25 nist5: rename and move source file
build tip: autoreconf && make -j
2016-09-27 00:57:31 +02:00
Tanguy Pruvot 0f8271f849 vstudio: return to CUDA 6.5 by default on x86
change was for the 1.8.2 tag (optimal for lbry)

and prepare next version...
2016-09-27 00:38:18 +02:00
Tanguy Pruvot c933ad901e sia: add missing algo in --help 2016-09-27 00:33:11 +02:00
Tanguy Pruvot 34e97bf3e6 Show intensity on init for all algos 2016-09-27 00:33:06 +02:00
Tanguy Pruvot 65cd430d76 cuda: throughput2intensity function to show default 2016-09-27 00:20:48 +02:00
Tanguy Pruvot 9f2ed5135b lbry maxwell and pascal update (up to 10% on pascal)
Based on alexis78 work and sponsored by LBRY.IO team (thanks)

Release 1.8.2, use cuda 8 for x86
2016-09-22 19:37:17 +02:00
Tanguy Pruvot f982fd70d5 lbry: small sha512 improvement 2016-09-04 18:11:42 +02:00
Tanguy Pruvot f8aa16f8d2 skein: cleanup, and precompute h8 2016-09-04 18:11:42 +02:00
Tanguy Pruvot 683dc0e149 VeltorCoin Streebog based algo (veltor)
also known as "Thor's Riddle"... yes sure ;)

Credits to ocminer who found and "implemented" it.

Note: tested "ok" on x64 and CUDA 6.5 x86, not on 7.5 and 8.0 x86

PS: Don't have the time for a more proper CUDA implementation of Streebog
2016-08-18 18:47:37 +02:00
Tanguy Pruvot 8f965c2cb2 windows: prevent system sleep while mining 2016-08-16 18:00:16 +02:00
Tanguy Pruvot 1a51a84a94 sia: small mutex fix on errors 2016-08-10 14:48:19 +02:00
Tanguy Pruvot 2520f9a388 lyra2: import latest nanashi code (v1) 2016-08-10 14:23:18 +02:00
Tanguy Pruvot 35bfe8d191 update version 2016-08-09 22:56:52 +02:00
Tanguy Pruvot d8e518ac3d sia: support for nanopool http protocol
-o http://xxxx:port -u <addr> -p <worker>

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2016-08-09 22:55:53 +02:00
Tanguy Pruvot b31fb5316b sia: cuda impl. and suprnova getwork over stratum
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2016-08-06 12:56:35 +02:00
Tanguy Pruvot 3d701944b2 lbry: exit on card failure, dont loop 2016-08-06 12:56:34 +02:00
Tanguy Pruvot 883fb19908 lbry: some changes from alexis, remove shared mem
105 LBC tipped ;)
2016-08-06 12:56:02 +02:00
Tanguy Pruvot de738ccc2b x11: secure groestl against possible cuda errors
big cleanup...
2016-08-06 12:56:02 +02:00
Tanguy Pruvot 0a0fd33cac attempt to reduce shared mem errors 2016-08-06 12:56:02 +02:00
Tanguy Pruvot 8985f7ee91 nvapi: asus rgb led
seems to be too slow to handle all wanted modes
2016-08-06 12:51:23 +02:00
Tanguy Pruvot 9123eff7dd Update Galax vendor id, used on 1070s 2016-07-25 16:37:00 +02:00
Tanguy Pruvot af1d5c9f3d nvapi: zotac rgb led support, to improve..
zotac cards doesnt allow a real RGB value, so i converted manually
the card seems to handle it, but seems to lock sometimes for some time...

colors can be tested with "ccminer -d 2 --led=0x1f1f00 -Dn"

but... unsure the card handle it fast enough to be fully usable
2016-07-22 01:20:30 +02:00
Tanguy Pruvot 78c3d25ccf release 1.8, as promised 2016-07-20 10:28:50 +02:00
Tanguy Pruvot be8be03eb8 neoscrypt: remove warnings and rename host funcs
also reduce the few errors on coins using shared mem
2016-07-20 10:26:42 +02:00
Tanguy Pruvot 6abee0659e update neoscrypt with Nanashi changes 2016-07-20 10:26:42 +02:00
Tanguy Pruvot bf17f34001 preview 3, with alexis78 touch 2016-07-20 10:26:13 +02:00
Tanguy Pruvot 223077d11a merged sha256d/ripemd
drop ripemd.cu and remove unused kernels
2016-07-16 22:29:12 +02:00
Tanguy Pruvot 940c1b3a2f lbry: small changes for second build
sha512/ripemd swab

this was preview 2
2016-07-16 22:28:48 +02:00
Tanguy Pruvot 5703a74e22 lbry algo (stratum only)
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2016-07-13 09:44:13 +02:00
Tanguy Pruvot dbb9507d2b api: change unit of device mem to MB
without that, no way to read sizes > 4GB on x86 binaries
2016-07-10 13:12:16 +02:00
Tanguy Pruvot 4ca7b5a404 neoscrypt: reduce spill load and increase pascal def intensity
1 MH/s reached on the 1070 ...
2016-07-09 23:39:15 +02:00
Tanguy Pruvot 26a862c7f6 neoscrypt: reduce a bit the cpu usage 2016-07-08 20:04:45 +02:00
Tanguy Pruvot a4196b341d neoscrypt: apply last VTC improvements
rewrote almost properly ;)
2016-07-08 16:30:26 +02:00
Tanguy Pruvot c0e9370ba2 quark: real hashrate was wrong, add a few kHs 2016-07-06 16:35:58 +02:00
Tanguy Pruvot 06e2485128 myr-gr: handle a second nonce & more cleanup 2016-07-05 13:06:51 +02:00
Tanguy Pruvot 374174c7c8 groestl: same, remove useless host alloc 2016-07-05 11:15:06 +02:00
Tanguy Pruvot 3fe4dda4c1 myr-gr: remove unused allocated memory + pascal tweak
+ cleanup...
2016-07-05 09:56:48 +02:00
Tanguy Pruvot 39d0e42b11 improve the --led values parsing
sample usage if GPU #0 is RVB and GPU #1 a generic led:

ccminer --led=0x00ff00,100,mining to put the led on while the gpus are scanning
2016-07-04 16:31:07 +02:00
Tanguy Pruvot fc1472b5f7 nvapi: allow different led modes (mining,shares) 2016-07-04 09:14:19 +02:00
Tanguy Pruvot 0e942bf60c nvapi: proper rgb color dimm function 2016-07-03 18:46:40 +02:00
Tanguy Pruvot d43084cec6 nvapi: improve rgb led stability + multi gpus 2016-07-03 18:25:38 +02:00
Tanguy Pruvot 8c4b25426f nvapi: device led flash on submit
disabled by default, require --led=100 on nvidia compatible devices

Gigabyte RGB led hack is not perfect for the moment, can fail and require a reboot.
2016-07-03 12:07:37 +02:00
Tanguy Pruvot 174e26164a nvapi: fix device mapping in -D -n
and rename RVB to RGB, french typo...
2016-07-02 20:21:33 +02:00
Tanguy Pruvot 036473347e nvapi: fan level, for future use... maybe 2016-07-01 20:13:47 +02:00
Tanguy Pruvot 24436febd6 led: proper log format 2016-07-01 20:13:17 +02:00
Tanguy Pruvot d3976cf38d nvapi: sample i2c rvb led color (gigabyte)
you can now use --led=0xFF00FF to change the color/level (windows only)

for non-rvb standard nvidia devices, you can use a level value (0 to 100)

ps: i really need to find how to turn them off on linux, not found yet...
2016-06-29 12:28:38 +02:00
Tanguy Pruvot 801db3d076 nvapi: be more secure with unsupported apis
also query the led level and fix OV delta output

note: the Gigabyte 1080 G1 seems not compatible with the current nvapi
2016-06-29 12:00:40 +02:00
Tanguy Pruvot eb84ef9c8a vendors: add Palit and the right PNY VIDs
seen on yiimp benchs, PNY duplicate could be for usb keys only
2016-06-28 01:39:53 +02:00
Tanguy Pruvot 095f25f9a8 nvapi: pascal core voltage boost + meminfo fix x86
cuda 6.5 one seems to crash on pascal or report invalid mem sizes
2016-06-26 22:30:54 +02:00
Tanguy Pruvot 0555201384 linux: return to default cuda, mistake in a nvapi commit 2016-06-26 20:31:31 +02:00
Tanguy Pruvot d19039fd7f nvapi: read pascal voltage + show meminfo in -Dn
also add perf stats api, unsure what it is exactly..
2016-06-26 08:28:44 +02:00
Tanguy Pruvot 7ff179abe9 nvapi: base memclock was wrong on maxwell 2016-06-25 16:18:39 +02:00
Tanguy Pruvot bc6ac3a3ab nvapi: link clocks and tlimit to command line
boost clocks and the thermal limit are shared with afterburner
beware with your settings, not as safe as application clocks!

Note: both nvapi and nvml are now used on windows x64
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2016-06-25 10:00:43 +02:00
Tanguy Pruvot 2ee8bc9791 nvapi: do not print that on normal -D 2016-06-24 10:14:58 +02:00
Tanguy Pruvot bdc441ebd0 nvapi: link some more apis, pascal boost table
these informations are shown with ccminer -D -n
2016-06-24 10:06:54 +02:00
Tanguy Pruvot eae4ede111 decred: return to previous implementation + second nonce
seems better on windows and a bit easier to read...
2016-06-23 03:54:33 +02:00
Tanguy Pruvot c643b3b900 decred: and even faster implementation by Alexis
optimized for the 9xx and more recent, same results on the 750 Ti
+ restore second nonce support not present in nicehash published version

Better on linux at least...
2016-06-23 00:36:28 +02:00
Tanguy Pruvot 7e490693e0 decred: nicehash/alexis improvement 2016-06-22 22:32:23 +02:00
Tanguy Pruvot 9296d5218e nvml: remove autoboost functions, unsupported
not for GTX pascal neither...
2016-06-21 05:16:21 +02:00
Tanguy Pruvot 0deb9a2aca win32: implement a nvapi.dll wrapper like nvml
Allow to get/set missing infos like the power limit on x86

squashed for a better min/max and device mapping

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2016-06-21 05:16:12 +02:00
Tanguy Pruvot 7a8ae1ab7a update static nvapi lib to R361 v2 (Jan 2016) 2016-06-21 03:00:44 +02:00
Tanguy Pruvot fd757f2180 nvml: add new getclock api (v8)
+ some new functions to test on pascal (if supported)
2016-06-20 07:32:26 +02:00
Tanguy Pruvot fe99de24a8 nvml: small changes to handle the 750ti app clocks
dynamically alloc the hardcoded gpu clocks list array size.

Unlike the 9xx, the 750ti api call requires to get them all (361.62)

Something was "fixed" recently for the 750 Ti in linux drivers...
2016-06-19 20:13:01 +02:00
Tanguy Pruvot 41a68f8f78 lyra2v2 fix define position 2016-06-11 13:37:51 +02:00
Tanguy Pruvot bbd3c6d5b9 lyra2v2: update credits, increase SM 5.0 default int to 19
+ small klausT cleanup..
2016-06-10 21:50:00 +02:00
Tanguy Pruvot 53275e3a00 Apply lyra2v2 boost published by NH
changes only tested/applied to SM 5+ with some cleanup
2016-06-09 21:12:09 +02:00
Tanguy Pruvot 8ecf0a9101 nist5: reported ccminer hashrate was wrong 2016-06-08 19:29:16 +02:00
Tanguy Pruvot 85c212eaad implement x11evo algo
Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2016-05-31 20:05:15 +02:00
Tanguy Pruvot f85e3772f9 api: sample command line api call in perl
tx to lilyth for the idea (works on cpuminer, sgminer and cgminer too)
2016-05-31 19:20:58 +02:00
Tanguy Pruvot a7905e6802 some changes to prepare 1.8 and VS2015 support
also move vstudio specific cpuminer-config.h in compat/
2016-05-28 06:27:41 +02:00
Tanguy Pruvot 1b124720fe fix vote default value for suprnova 2016-05-20 22:27:48 +02:00
Tanguy Pruvot 282129f584 linux: add temporary cuda files to gitignore 2016-05-20 22:27:48 +02:00
244 changed files with 59445 additions and 13171 deletions

No files matched your search

+18 -2
View File
@@ -15,7 +15,7 @@ depcomp
missing
install-sh
stamp-h1
cpuminer-config.h*
ccminer-config.h*
compile
config.log
config.status
@@ -25,7 +25,7 @@ config.sub
mingw32-config.cache
*/.dirstamp
.dirstamp
.DS_Store
Desktop.ini
Thumbs.db
@@ -41,6 +41,8 @@ x64/Release/
*.opensdf
*.sdf
*.pdb
*.VC.opendb
*.VC.db
.settings/
.project
@@ -52,3 +54,17 @@ x64/Release/
compat/curl-for-windows/
*.ptx
*.i
*.ii
*.gpu
*.cubin
*.stub.c
*.module_id
*.hash
*.fatbin.c
*.fatbin
*.cudafe1.cpp
*.cudafe1.c
*.cudafe2.c
+29 -19
View File
@@ -45,7 +45,7 @@ static uint32_t *h_resNonce[MAX_GPUS];
/* max count of found nonces in one call */
#define NBN 2
static uint32_t extra_results[NBN] = { UINT32_MAX };
static __thread uint32_t extra_results[NBN] = { UINT32_MAX };
#define GSPREC(a,b,c,d,x,y) { \
v[a] += (m[x] ^ c_u256[y]) + v[b]; \
@@ -504,6 +504,9 @@ extern "C" int scanhash_blake256(int thr_id, struct work* work, uint32_t max_non
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
@@ -518,46 +521,53 @@ extern "C" int scanhash_blake256(int thr_id, struct work* work, uint32_t max_non
do {
// GPU HASH (second block only, first is midstate)
uint32_t foundNonce = blake256_cpu_hash_16(thr_id, throughput, pdata[19], targetHigh, blakerounds);
work->nonces[0] = blake256_cpu_hash_16(thr_id, throughput, pdata[19], targetHigh, blakerounds);
if (foundNonce != UINT32_MAX)
*hashes_done = pdata[19] - first_nonce + throughput;
if (work->nonces[0] != UINT32_MAX)
{
uint32_t vhashcpu[8];
uint32_t Htarg = ptarget[6];
uint32_t _ALIGN(64) vhashcpu[8];
const uint32_t Htarg = ptarget[6];
for (int k=16; k < 19; k++)
be32enc(&endiandata[k], pdata[k]);
be32enc(&endiandata[19], foundNonce);
be32enc(&endiandata[19], work->nonces[0]);
blake256hash(vhashcpu, endiandata, blakerounds);
if (vhashcpu[6] <= Htarg && fulltest(vhashcpu, ptarget))
{
rc = 1;
work->valid_nonces = 1;
work_set_target_ratio(work, vhashcpu);
*hashes_done = pdata[19] - first_nonce + throughput;
pdata[19] = foundNonce;
#if NBN > 1
if (extra_results[0] != UINT32_MAX) {
be32enc(&endiandata[19], extra_results[0]);
work->nonces[1] = extra_results[0];
be32enc(&endiandata[19], work->nonces[1]);
blake256hash(vhashcpu, endiandata, blakerounds);
if (vhashcpu[6] <= Htarg && fulltest(vhashcpu, ptarget)) {
pdata[21] = extra_results[0];
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio) {
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio[0]) {
work_set_target_ratio(work, vhashcpu);
xchg(pdata[21], pdata[19]);
xchg(work->nonces[0], work->nonces[1]);
} else {
bn_set_target_ratio(work, vhashcpu, 1);
}
rc = 2;
work->valid_nonces = 2;
}
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
extra_results[0] = UINT32_MAX;
} else {
pdata[19] = work->nonces[0] + 1; // cursor
}
#endif
return rc;
return work->valid_nonces;
}
else if (opt_debug) {
applog_hash((uchar*)ptarget);
applog_compare_hash((uchar*)vhashcpu, (uchar*)ptarget);
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNonce);
else if (vhashcpu[6] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
pdata[19] = work->nonces[0] + 1;
continue;
}
}
+436 -386
View File
@@ -1,402 +1,429 @@
/**
* Blake2-S 256 CUDA implementation
* @author tpruvot@github March 2016
* Based on the SPH implementation of blake2s
* Provos Alexis - 2016
*/
#include <stdio.h>
#include <string.h>
#include <stdint.h>
#include <memory.h>
#include "miner.h"
extern "C" {
#define NATIVE_LITTLE_ENDIAN
#include <sph/blake2s.h>
}
#include <string.h>
#include <stdint.h>
//#define GPU_MIDSTATE
#define MIDLEN 76
#define A 64
static __thread blake2s_state ALIGN(A) s_midstate;
static __thread blake2s_state ALIGN(A) s_ctx;
#include "cuda_helper.h"
#include "sph/blake2s.h"
#include "sph/sph_types.h"
#ifdef __INTELLISENSE__
#define __byte_perm(x, y, b) x
#endif
#ifndef GPU_MIDSTATE
__constant__ uint2 d_data[10];
#include "cuda_helper.h"
#ifdef __CUDA_ARCH__
__device__ __forceinline__
uint32_t ROR8(const uint32_t a) {
return __byte_perm(a, 0, 0x0321);
}
__device__ __forceinline__
uint32_t ROL16(const uint32_t a) {
return __byte_perm(a, 0, 0x1032);
}
#else
__constant__ blake2s_state ALIGN(8) d_state[1];
#define ROR8(u) (u >> 8)
#define ROL16(u) (u << 16)
#endif
/* 16 adapters max */
static uint32_t *d_resNonce[MAX_GPUS];
static uint32_t *h_resNonce[MAX_GPUS];
/* threads per block */
#define TPB 512
/* max count of found nonces in one call */
#define NBN 2
#if NBN > 1
static uint32_t extra_results[NBN] = { UINT32_MAX };
__device__ __forceinline__
uint32_t xor3x(uint32_t a, uint32_t b, uint32_t c)
{
uint32_t result;
#if __CUDA_ARCH__ >= 500 && CUDA_VERSION >= 7050
asm ("lop3.b32 %0, %1, %2, %3, 0x96;" : "=r"(result) : "r"(a), "r"(b),"r"(c)); //0x96 = 0xF0 ^ 0xCC ^ 0xAA
#else
result = a^b^c;
#endif
return result;
}
static const uint32_t blake2s_IV[8] = {
0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
};
static const uint8_t blake2s_sigma[10][16] = {
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
{ 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
{ 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
{ 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
{ 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
{ 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
{ 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
{ 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
{ 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13 , 0 },
};
#define G(r,i,a,b,c,d) \
do { \
a = a + b + m[blake2s_sigma[r][2*i+0]]; \
d = SPH_ROTR32(d ^ a, 16); \
c = c + d; \
b = SPH_ROTR32(b ^ c, 12); \
a = a + b + m[blake2s_sigma[r][2*i+1]]; \
d = SPH_ROTR32(d ^ a, 8); \
c = c + d; \
b = SPH_ROTR32(b ^ c, 7); \
} while(0)
#define ROUND(r) \
do { \
G(r,0,v[0],v[4],v[ 8],v[12]); \
G(r,1,v[1],v[5],v[ 9],v[13]); \
G(r,2,v[2],v[6],v[10],v[14]); \
G(r,3,v[3],v[7],v[11],v[15]); \
G(r,4,v[0],v[5],v[10],v[15]); \
G(r,5,v[1],v[6],v[11],v[12]); \
G(r,6,v[2],v[7],v[ 8],v[13]); \
G(r,7,v[3],v[4],v[ 9],v[14]); \
} while(0)
extern "C" void blake2s_hash(void *output, const void *input)
{
uint8_t _ALIGN(A) hash[BLAKE2S_OUTBYTES];
blake2s_state blake2_ctx;
blake2s_init(&blake2_ctx, BLAKE2S_OUTBYTES);
blake2s_update(&blake2_ctx, (uint8_t*) input, 80);
blake2s_final(&blake2_ctx, hash, BLAKE2S_OUTBYTES);
memcpy(output, hash, 32);
}
__host__
inline void blake2s_hash_end(uint32_t *output, const uint32_t *input)
{
s_ctx.buflen = MIDLEN;
memcpy(&s_ctx, &s_midstate, 32 + 16 + MIDLEN);
blake2s_update(&s_ctx, (uint8_t*) &input[MIDLEN/4], 80-MIDLEN);
blake2s_final(&s_ctx, (uint8_t*) output, BLAKE2S_OUTBYTES);
}
__host__
void blake2s_setBlock(uint32_t *penddata, blake2s_state *pstate)
{
#ifndef GPU_MIDSTATE
CUDA_SAFE_CALL(cudaMemcpyToSymbol(d_data, penddata, 80, 0, cudaMemcpyHostToDevice));
#else
CUDA_SAFE_CALL(cudaMemcpyToSymbol(d_state, pstate, sizeof(blake2s_state), 0, cudaMemcpyHostToDevice));
#endif
}
__device__ __forceinline__
uint64_t gpu_load64(void *src) {
return *(uint64_t*)(src);
}
__device__ __forceinline__
void gpu_store32(void *dst, uint32_t dw) {
*(uint32_t*)(dst) = dw;
}
__device__ __forceinline__
void gpu_store64(void *dst, uint64_t lw) {
*(uint64_t*)(dst) = lw;
}
__device__ __forceinline__
void gpu_blake2s_set_lastnode(blake2s_state *S) {
S->f[1] = ~0U;
}
__device__ __forceinline__
void gpu_blake2s_clear_lastnode(blake2s_state *S) {
S->f[1] = 0U;
}
__device__ __forceinline__
void gpu_blake2s_increment_counter(blake2s_state *S, const uint32_t inc)
{
S->t[0] += inc;
S->t[1] += ( S->t[0] < inc );
}
__device__ __forceinline__
void gpu_blake2s_set_lastblock(blake2s_state *S)
{
if (S->last_node) gpu_blake2s_set_lastnode(S);
S->f[0] = ~0U;
}
__device__
void gpu_blake2s_compress(blake2s_state *S, const uint32_t *block)
{
uint32_t m[16];
uint32_t v[16];
uint32_t h[8];
const uint32_t blake2s_IV[8] = {
0x6A09E667UL, 0xBB67AE85UL, 0x3C6EF372UL, 0xA54FF53AUL,
0x510E527FUL, 0x9B05688CUL, 0x1F83D9ABUL, 0x5BE0CD19UL
};
uint32_t *in = (uint32_t*)input;
// COMPRESS
for(int i = 0; i < 16; ++i )
m[i] = in[i];
const uint8_t blake2s_sigma[10][16] = {
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
{ 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
{ 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
{ 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
{ 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
{ 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
{ 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
{ 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
{ 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13 , 0 },
};
h[0] = 0x01010020 ^ blake2s_IV[0];
h[1] = blake2s_IV[1];
h[2] = blake2s_IV[2];
h[3] = blake2s_IV[3];
h[4] = blake2s_IV[4];
h[5] = blake2s_IV[5];
h[6] = blake2s_IV[6];
h[7] = blake2s_IV[7];
#pragma unroll
for(int i = 0; i < 16; i++)
m[i] = block[i];
for(int i = 0; i < 8; ++i )
v[i] = h[i];
#pragma unroll
for(int i = 0; i < 8; i++)
v[i] = S->h[i];
v[ 8] = blake2s_IV[0]; v[ 9] = blake2s_IV[1];
v[10] = blake2s_IV[2]; v[11] = blake2s_IV[3];
v[12] = 64 ^ blake2s_IV[4]; v[13] = blake2s_IV[5];
v[14] = blake2s_IV[6]; v[15] = blake2s_IV[7];
v[ 8] = blake2s_IV[0];
v[ 9] = blake2s_IV[1];
v[10] = blake2s_IV[2];
v[11] = blake2s_IV[3];
v[12] = S->t[0] ^ blake2s_IV[4];
v[13] = S->t[1] ^ blake2s_IV[5];
v[14] = S->f[0] ^ blake2s_IV[6];
v[15] = S->f[1] ^ blake2s_IV[7];
ROUND( 0 ); ROUND( 1 );
ROUND( 2 ); ROUND( 3 );
ROUND( 4 ); ROUND( 5 );
ROUND( 6 ); ROUND( 7 );
ROUND( 8 ); ROUND( 9 );
#define G(r,i,a,b,c,d) { \
a += b + m[blake2s_sigma[r][2*i+0]]; \
d = __byte_perm(d ^ a, 0, 0x1032); /* d = ROTR32(d ^ a, 16); */ \
c = c + d; \
b = ROTR32(b ^ c, 12); \
a += b + m[blake2s_sigma[r][2*i+1]]; \
d = __byte_perm(d ^ a, 0, 0x0321); /* ROTR32(d ^ a, 8); */ \
c = c + d; \
b = ROTR32(b ^ c, 7); \
}
for(size_t i = 0; i < 8; ++i)
h[i] ^= v[i] ^ v[i + 8];
#define ROUND(r) { \
G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
G(r,2,v[ 2],v[ 6],v[10],v[14]); \
G(r,3,v[ 3],v[ 7],v[11],v[15]); \
G(r,4,v[ 0],v[ 5],v[10],v[15]); \
G(r,5,v[ 1],v[ 6],v[11],v[12]); \
G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
}
// COMPRESS
m[0] = in[16]; m[1] = in[17];
m[2] = in[18]; m[3] = in[19];
for(size_t i = 4; i < 16; ++i)
m[i] = 0;
ROUND( 0 );
ROUND( 1 );
ROUND( 2 );
ROUND( 3 );
ROUND( 4 );
ROUND( 5 );
ROUND( 6 );
ROUND( 7 );
ROUND( 8 );
ROUND( 9 );
for(size_t i = 0; i < 8; ++i)
v[i] = h[i];
#pragma unroll
for(int i = 0; i < 8; i++)
S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
v[ 8] = blake2s_IV[0]; v[ 9] = blake2s_IV[1];
v[10] = blake2s_IV[2]; v[11] = blake2s_IV[3];
v[12] = 0x50 ^ blake2s_IV[4]; v[13] = blake2s_IV[5];
v[14] = ~blake2s_IV[6]; v[15] = blake2s_IV[7];
#undef G
#undef ROUND
ROUND( 0 ); ROUND( 1 );
ROUND( 2 ); ROUND( 3 );
ROUND( 4 ); ROUND( 5 );
ROUND( 6 ); ROUND( 7 );
ROUND( 8 ); ROUND( 9 );
for(size_t i = 0; i < 8; ++i)
h[i] ^= v[i] ^ v[i + 8];
memcpy(output, h, 32);
}
#if 0
/* unused but kept as reference */
__device__ __forceinline__
void gpu_blake2s_update(blake2s_state *S, const uint8_t *in, uint64_t inlen)
{
while(inlen > 0)
{
const int left = S->buflen;
size_t fill = 2 * BLAKE2S_BLOCKBYTES - left;
if(inlen > fill)
{
memcpy(S->buf + left, in, fill); // Fill buffer
S->buflen += fill;
#define TPB 1024
#define NPT 256
#define maxResults 16
#define NBN 1
gpu_blake2s_increment_counter(S, BLAKE2S_BLOCKBYTES);
gpu_blake2s_compress(S, (uint32_t*) S->buf); // Compress
memcpy(S->buf, S->buf + BLAKE2S_BLOCKBYTES, BLAKE2S_BLOCKBYTES); // Shift buffer left
S->buflen -= BLAKE2S_BLOCKBYTES;
in += fill;
inlen -= fill;
__constant__ uint32_t _ALIGN(32) midstate[20];
static uint32_t *d_resNonce[MAX_GPUS];
static uint32_t *h_resNonce[MAX_GPUS];
#define GS4(a,b,c,d,e,f,a1,b1,c1,d1,e1,f1,a2,b2,c2,d2,e2,f2,a3,b3,c3,d3,e3,f3){ \
a += b + e; a1+= b1 + e1; a2+= b2 + e2; a3+= b3 + e3; \
d = ROL16( d ^ a); d1 = ROL16(d1 ^ a1); d2 = ROL16(d2 ^ a2); d3 = ROL16(d3 ^ a3); \
c +=d; c1+=d1; c2+=d2; c3+=d3;\
b = ROTR32(b ^ c, 12); b1 = ROTR32(b1^c1, 12); b2 = ROTR32(b2^c2, 12); b3 = ROTR32(b3^c3, 12); \
a += b + f; a1+= b1 + f1; a2+= b2 + f2; a3+= b3 + f3; \
d = ROR8(d ^ a); d1 = ROR8(d1^a1); d2 = ROR8(d2^a2); d3 = ROR8(d3^a3); \
c += d; c1 += d1; c2 += d2; c3 += d3;\
b = ROTR32(b ^ c, 7); b1 = ROTR32(b1^c1, 7); b2 = ROTR32(b2^c2, 7); b3 = ROTR32(b3^c3, 7); \
}
__global__ __launch_bounds__(TPB,1)
void blake2s_gpu_hash_nonce(const uint32_t threads, const uint32_t startNonce, uint32_t *resNonce, const uint32_t ptarget7)
{
const uint32_t step = gridDim.x * blockDim.x;
uint32_t m[ 3];
uint32_t v[16];
m[0] = midstate[16];
m[1] = midstate[17];
m[2] = midstate[18];
const uint32_t h7 = midstate[19];
for(uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x ; thread <threads; thread+=step){
#pragma unroll
for(int i=0;i<16;i++){
v[ i] = midstate[ i];
}
else // inlen <= fill
{
memcpy(S->buf + left, in, (size_t) inlen);
S->buflen += (size_t) inlen; // Be lazy, do not compress
in += inlen;
inlen -= inlen;
uint32_t nonce = cuda_swab32(startNonce + thread);
// Round( 0 );
v[ 1] += nonce;
v[13] = ROR8(v[13] ^ v[ 1]);
v[ 9] += v[13];
v[ 5] = ROTR32(v[ 5] ^ v[ 9], 7);
v[ 1]+= v[ 6];
v[ 0]+= v[ 5];
v[12] = ROL16(v[12] ^ v[ 1]);
v[13] = ROL16(v[13] ^ v[ 2]);
v[15] = ROL16(v[15] ^ v[ 0]);
v[11]+= v[12]; v[ 8]+= v[13]; v[ 9]+= v[14]; v[10]+= v[15];
v[ 6] = ROTR32(v[ 6] ^ v[11], 12); v[ 7] = ROTR32(v[ 7] ^ v[ 8], 12); v[ 4] = ROTR32(v[ 4] ^ v[ 9], 12); v[ 5] = ROTR32(v[ 5] ^ v[10], 12);
v[ 1]+= v[ 6]; v[ 2]+= v[ 7]; v[ 3]+= v[ 4]; v[ 0]+= v[ 5];
v[12] = ROR8(v[12] ^ v[ 1]); v[13] = ROR8(v[13] ^ v[ 2]); v[14] = ROR8(v[14] ^ v[ 3]); v[15] = ROR8(v[15] ^ v[ 0]);
v[11]+= v[12]; v[ 8]+= v[13]; v[ 9]+= v[14]; v[10]+= v[15];
v[ 6] = ROTR32(v[ 6] ^ v[11], 7); v[ 7] = ROTR32(v[ 7] ^ v[ 8], 7); v[ 4] = ROTR32(v[ 4] ^ v[ 9], 7); v[ 5] = ROTR32(v[ 5] ^ v[10], 7);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],m[ 1],0, v[ 1],v[ 6],v[11],v[12],m[ 0],m[ 2], v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],0,nonce);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,m[ 0], v[ 2],v[ 6],v[10],v[14],0,m[ 2], v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],0,0, v[ 1],v[ 6],v[11],v[12],nonce,0, v[ 2],v[ 7],v[ 8],v[13],0,m[ 1], v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],nonce,m[ 1], v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],m[ 2],0, v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,m[ 0], v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,m[ 0], v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],m[ 2],0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],0,m[ 1], v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],nonce,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],m[ 2],0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],m[ 0],0, v[ 3],v[ 7],v[11],v[15],0,nonce);
GS4(v[ 0],v[ 5],v[10],v[15],0,0, v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],m[ 1],0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],m[ 1],0, v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],m[ 0],0, v[ 1],v[ 6],v[11],v[12],0,nonce, v[ 2],v[ 7],v[ 8],v[13],0,m[ 2], v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,m[ 1], v[ 3],v[ 7],v[11],v[15],nonce,0);
GS4(v[ 0],v[ 5],v[10],v[15],0,m[ 0], v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],m[ 2],0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,nonce, v[ 3],v[ 7],v[11],v[15],m[ 0],0);
GS4(v[ 0],v[ 5],v[10],v[15],0,m[ 2], v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],m[ 1],0, v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,m[ 2], v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],m[ 1],0);
// GS(9,4,v[ 0],v[ 5],v[10],v[15]);
v[ 0] += v[ 5];
v[ 2] += v[ 7] + nonce;
v[15] = ROL16(v[15] ^ v[ 0]);
v[13] = ROL16(v[13] ^ v[ 2]);
v[10] += v[15];
v[ 8] += v[13];
v[ 5] = ROTR32(v[ 5] ^ v[10], 12);
v[ 7] = ROTR32(v[ 7] ^ v[ 8], 12);
v[ 0] += v[ 5];
v[ 2] += v[ 7];
v[15] = ROR8(v[15] ^ v[ 0]);
v[13] = ROR8(v[13] ^ v[ 2]);
v[ 8] += v[13];
v[ 7] = ROTR32(v[ 7] ^ v[ 8], 7);
if (xor3x(h7,v[7],v[15]) <= ptarget7){
uint32_t pos = atomicInc(&resNonce[0],0xffffffff)+1;
if(pos < maxResults)
resNonce[pos] = nonce;
return;
}
}
}
#endif
#ifndef GPU_MIDSTATE
__device__ __forceinline__
void gpu_blake2s_fill_data(blake2s_state *S, const uint32_t nonce)
__global__ __launch_bounds__(TPB,1)
void blake2s_gpu_hash_nonce(const uint32_t threads, const uint32_t startNonce, uint32_t *resNonce)
{
uint2 *b2 = (uint2*) S->buf;
#pragma unroll
for (int i=0; i < 9; i++)
b2[i] = d_data[i];
b2[9].x = d_data[9].x;
b2[9].y = nonce;
S->buflen = 80;
}
#endif
const uint32_t step = gridDim.x * blockDim.x;
__device__ __forceinline__
void gpu_blake2s_update_nonce(blake2s_state *S, const uint32_t nonce)
{
gpu_store32(&S->buf[76], nonce);
S->buflen = 80;
}
uint32_t m[ 3];
uint32_t v[16];
__device__ __forceinline__
uint2 gpu_blake2s_final(blake2s_state *S)
{
//if (S->buflen > BLAKE2S_BLOCKBYTES)
m[0] = midstate[16];
m[1] = midstate[17];
m[2] = midstate[18];
const uint32_t h7 = midstate[19];
for(uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x ; thread <threads; thread+=step)
{
gpu_blake2s_increment_counter(S, BLAKE2S_BLOCKBYTES);
gpu_blake2s_compress(S, (uint32_t*) S->buf);
S->buflen -= BLAKE2S_BLOCKBYTES;
//memcpy(S->buf, S->buf + BLAKE2S_BLOCKBYTES, S->buflen);
}
#pragma unroll
for(int i=0;i<16;i++){
v[ i] = midstate[ i];
}
gpu_blake2s_increment_counter(S, (uint32_t)S->buflen);
gpu_blake2s_set_lastblock(S);
//memset(&S->buf[S->buflen], 0, 2 * BLAKE2S_BLOCKBYTES - S->buflen); /* Padding */
gpu_blake2s_compress(S, (uint32_t*) (S->buf + BLAKE2S_BLOCKBYTES));
uint32_t nonce = cuda_swab32(startNonce+thread);
//#pragma unroll
//for (int i = 0; i < 8; i++)
// out[i] = S->h[i];
return make_uint2(S->h[6], S->h[7]);
}
// Round( 0 );
v[ 1] += nonce;
v[13] = ROR8(v[13] ^ v[ 1]);
v[ 9] += v[13];
v[ 5] = ROTR32(v[ 5] ^ v[ 9], 7);
/* init2 xors IV with input parameter block */
__device__ __forceinline__
void gpu_blake2s_init_param(blake2s_state *S, const blake2s_param *P)
{
//blake2s_IV
S->h[0] = 0x6A09E667UL;
S->h[1] = 0xBB67AE85UL;
S->h[2] = 0x3C6EF372UL;
S->h[3] = 0xA54FF53AUL;
S->h[4] = 0x510E527FUL;
S->h[5] = 0x9B05688CUL;
S->h[6] = 0x1F83D9ABUL;
S->h[7] = 0x5BE0CD19UL;
v[ 1]+= v[ 6];
v[ 0]+= v[ 5];
S->t[0] = 0; S->t[1] = 0;
S->f[0] = 0; S->f[1] = 0;
S->last_node = 0;
v[13] = ROL16(v[13] ^ v[ 2]); v[12] = ROL16(v[12] ^ v[ 1]); v[15] = ROL16(v[15] ^ v[ 0]);
S->buflen = 0;
v[ 8]+= v[13]; v[11]+= v[12]; v[ 9]+= v[14]; v[10]+= v[15];
v[ 7] = ROTR32(v[ 7] ^ v[ 8], 12); v[ 6] = ROTR32(v[ 6] ^ v[11], 12); v[ 4] = ROTR32(v[ 4] ^ v[ 9], 12); v[ 5] = ROTR32(v[ 5] ^ v[10], 12);
v[ 2]+= v[ 7]; v[ 1]+= v[ 6]; v[ 3]+= v[ 4]; v[ 0]+= v[ 5];
v[13] = ROR8(v[13] ^ v[ 2]); v[12] = ROR8(v[12] ^ v[ 1]); v[14] = ROR8(v[14] ^ v[ 3]); v[15] = ROR8(v[15] ^ v[ 0]);
v[ 8]+= v[13]; v[11]+= v[12]; v[ 9]+= v[14]; v[10]+= v[15];
v[ 6] = ROTR32(v[ 6] ^ v[11], 7); v[ 7] = ROTR32(v[ 7] ^ v[8], 7); v[ 4] = ROTR32(v[ 4] ^ v[ 9], 7); v[ 5] = ROTR32(v[ 5] ^ v[10], 7);
#pragma unroll
for (int i = 8; i < sizeof(S->buf)/8; i++)
gpu_store64(S->buf + (8*i), 0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],m[ 1],0, v[ 1],v[ 6],v[11],v[12],m[ 0],m[ 2], v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],0,nonce);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,m[ 0], v[ 2],v[ 6],v[10],v[14],0,m[ 2], v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],0,0, v[ 1],v[ 6],v[11],v[12],nonce,0, v[ 2],v[ 7],v[ 8],v[13],0,m[ 1], v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],nonce,m[ 1], v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],m[ 2],0, v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,m[ 0], v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,m[ 0], v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],m[ 2],0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],0,m[ 1], v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],nonce,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],m[ 2],0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],m[ 0],0, v[ 3],v[ 7],v[11],v[15],0,nonce);
GS4(v[ 0],v[ 5],v[10],v[15],0,0, v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],m[ 1],0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],m[ 1],0, v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],0,0);
GS4(v[ 0],v[ 5],v[10],v[15],m[ 0],0, v[ 1],v[ 6],v[11],v[12],0,nonce, v[ 2],v[ 7],v[ 8],v[13],0,m[ 2], v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,m[ 1], v[ 3],v[ 7],v[11],v[15],nonce,0);
GS4(v[ 0],v[ 5],v[10],v[15],0,m[ 0], v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],0,0, v[ 3],v[ 4],v[ 9],v[14],m[ 2],0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,0, v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,nonce, v[ 3],v[ 7],v[11],v[15],m[ 0],0);
GS4(v[ 0],v[ 5],v[10],v[15],0,m[ 2], v[ 1],v[ 6],v[11],v[12],0,0, v[ 2],v[ 7],v[ 8],v[13],m[ 1],0, v[ 3],v[ 4],v[ 9],v[14],0,0);
GS4(v[ 0],v[ 4],v[ 8],v[12],0,m[ 2], v[ 1],v[ 5],v[ 9],v[13],0,0, v[ 2],v[ 6],v[10],v[14],0,0, v[ 3],v[ 7],v[11],v[15],m[ 1],0);
uint64_t *p = (uint64_t*) P;
v[ 0] += v[ 5];
v[ 2] += v[ 7] + nonce;
v[15] = ROL16(v[15] ^ v[ 0]);
v[13] = ROL16(v[13] ^ v[ 2]);
v[10] += v[15];
v[ 8] += v[13];
v[ 5] = ROTR32(v[ 5] ^ v[10], 12);
v[ 7] = ROTR32(v[ 7] ^ v[ 8], 12);
v[ 0] += v[ 5];
v[ 2] += v[ 7];
v[15] = ROTR32(v[15] ^ v[ 0],1);
v[13] = ROR8(v[13] ^ v[ 2]);
/* IV XOR ParamBlock */
#pragma unroll
for (int i = 0; i < 4; i++)
S->h[i] ^= gpu_load64(&p[i]);
}
v[ 8] += v[13];
// Sequential blake2s initialization
__device__ __forceinline__
void gpu_blake2s_init(blake2s_state *S, const uint8_t outlen)
{
blake2s_param P[1];
// if (!outlen || outlen > BLAKE2S_OUTBYTES) return;
P->digest_length = outlen;
P->key_length = 0;
P->fanout = 1;
P->depth = 1;
P->leaf_length = 0;
gpu_store64(P->node_offset, 0);
//P->node_depth = 0;
//P->inner_length = 0;
gpu_store64(&P->salt, 0);
gpu_store64(&P->personal, 0);
gpu_blake2s_init_param(S, P);
}
__device__ __forceinline__
void gpu_copystate(blake2s_state *dst, blake2s_state *src)
{
uint64_t* d64 = (uint64_t*) dst;
uint64_t* s64 = (uint64_t*) src;
#pragma unroll
for (int i=0; i < (32 + 16 + 2 * BLAKE2S_BLOCKBYTES)/8; i++)
gpu_store64(&d64[i], s64[i]);
dst->buflen = src->buflen;
dst->last_node = src->last_node;
}
__global__
void blake2s_gpu_hash(const uint32_t threads, const uint32_t startNonce, uint32_t *resNonce, const uint2 target2, const int swap)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
const uint32_t nonce = swap ? cuda_swab32(startNonce + thread) : startNonce + thread;
blake2s_state ALIGN(8) blake2_ctx;
#ifndef GPU_MIDSTATE
gpu_blake2s_init(&blake2_ctx, BLAKE2S_OUTBYTES);
//gpu_blake2s_update(&blake2_ctx, (uint8_t*) d_data, 76);
gpu_blake2s_fill_data(&blake2_ctx, nonce);
#else
gpu_copystate(&blake2_ctx, &d_state[0]);
gpu_blake2s_update_nonce(&blake2_ctx, nonce);
#endif
uint2 h2 = gpu_blake2s_final(&blake2_ctx);
if (h2.y <= target2.y && h2.x <= target2.x) {
#if NBN == 2
if (resNonce[0] != UINT32_MAX)
resNonce[1] = nonce;
else
resNonce[0] = nonce;
#else
resNonce[0] = nonce;
#endif
if(xor3x(v[ 7],h7,v[ 8])==v[15]){
uint32_t pos = atomicInc(&resNonce[0],0xffffffff)+1;
if(pos < maxResults)
resNonce[pos]=nonce;
return;
}
}
}
static __inline uint32_t swab32_if(uint32_t val, bool iftrue) {
return iftrue ? swab32(val) : val;
}
__host__
uint32_t blake2s_hash_cuda(const int thr_id, const uint32_t threads, const uint32_t startNonce, const uint2 target2, const int swap)
static void blake2s_setBlock(const uint32_t* input,const uint32_t ptarget7)
{
uint32_t result = UINT32_MAX;
uint32_t _ALIGN(64) m[16];
uint32_t _ALIGN(64) v[16];
uint32_t _ALIGN(64) h[21];
dim3 grid((threads + TPB-1)/TPB);
dim3 block(TPB);
// COMPRESS
for(int i = 0; i < 16; ++i )
m[i] = input[i];
/* Check error on Ctrl+C or kill to prevent segfaults on exit */
if (cudaMemset(d_resNonce[thr_id], 0xff, NBN*sizeof(uint32_t)) != cudaSuccess)
return result;
h[0] = 0x01010020 ^ blake2s_IV[0];
h[1] = blake2s_IV[1];
h[2] = blake2s_IV[2]; h[3] = blake2s_IV[3];
h[4] = blake2s_IV[4]; h[5] = blake2s_IV[5];
h[6] = blake2s_IV[6]; h[7] = blake2s_IV[7];
blake2s_gpu_hash <<<grid, block>>> (threads, startNonce, d_resNonce[thr_id], target2, swap);
cudaThreadSynchronize();
for(int i = 0; i < 8; ++i )
v[i] = h[i];
if (cudaSuccess == cudaMemcpy(h_resNonce[thr_id], d_resNonce[thr_id], NBN*sizeof(uint32_t), cudaMemcpyDeviceToHost)) {
result = swab32_if(h_resNonce[thr_id][0], swap);
#if NBN > 1
for (int n=0; n < (NBN-1); n++)
extra_results[n] = swab32_if(h_resNonce[thr_id][n+1], swap);
#endif
v[ 8] = blake2s_IV[0]; v[ 9] = blake2s_IV[1];
v[10] = blake2s_IV[2]; v[11] = blake2s_IV[3];
v[12] = 64 ^ blake2s_IV[4]; v[13] = blake2s_IV[5];
v[14] = blake2s_IV[6]; v[15] = blake2s_IV[7];
ROUND( 0 ); ROUND( 1 );
ROUND( 2 ); ROUND( 3 );
ROUND( 4 ); ROUND( 5 );
ROUND( 6 ); ROUND( 7 );
ROUND( 8 ); ROUND( 9 );
for(int i = 0; i < 8; ++i )
h[i] ^= v[i] ^ v[i + 8];
h[16] = input[16];
h[17] = input[17];
h[18] = input[18];
h[ 8] = 0x6A09E667; h[ 9] = 0xBB67AE85;
h[10] = 0x3C6EF372; h[11] = 0xA54FF53A;
h[12] = 0x510E522F; h[13] = 0x9B05688C;
h[14] =~0x1F83D9AB; h[15] = 0x5BE0CD19;
h[ 0]+= h[ 4] + h[16];
h[12] = SPH_ROTR32(h[12] ^ h[ 0],16);
h[ 8]+= h[12];
h[ 4] = SPH_ROTR32(h[ 4] ^ h[ 8],12);
h[ 0]+= h[ 4] + h[17];
h[12] = SPH_ROTR32(h[12] ^ h[ 0],8);
h[ 8]+= h[12];
h[ 4] = SPH_ROTR32(h[ 4] ^ h[ 8],7);
h[ 1]+= h[ 5] + h[18];
h[13] = SPH_ROTR32(h[13] ^ h[ 1], 16);
h[ 9]+= h[13];
h[ 5] = ROTR32(h[ 5] ^ h[ 9], 12);
h[ 2]+= h[ 6];
h[14] = SPH_ROTR32(h[14] ^ h[ 2],16);
h[10]+= h[14];
h[ 6] = SPH_ROTR32(h[ 6] ^ h[10], 12);
h[ 2]+= h[ 6];
h[14] = SPH_ROTR32(h[14] ^ h[ 2],8);
h[10]+= h[14];
h[ 6] = SPH_ROTR32(h[ 6] ^ h[10], 7);
h[19] = h[7]; //constant h[7] for nonce check
h[ 3]+= h[ 7];
h[15] = SPH_ROTR32(h[15] ^ h[ 3],16);
h[11]+= h[15];
h[ 7] = SPH_ROTR32(h[ 7] ^ h[11], 12);
h[ 3]+= h[ 7];
h[15] = SPH_ROTR32(h[15] ^ h[ 3],8);
h[11]+= h[15];
h[ 7] = SPH_ROTR32(h[ 7] ^ h[11], 7);
h[ 1]+= h[ 5];
h[ 3]+= h[ 4];
h[14] = SPH_ROTR32(h[14] ^ h[ 3],16);
h[ 2]+= h[ 7];
if(ptarget7==0){
h[19] = SPH_ROTL32(h[19],7); //align the rotation with v[7] v[15];
}
return result;
cudaMemcpyToSymbol(midstate, h, 20*sizeof(uint32_t), 0, cudaMemcpyHostToDevice);
}
static bool init[MAX_GPUS] = { 0 };
@@ -404,23 +431,21 @@ static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_blake2s(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const int swap = 1; // to toggle nonce endian
uint32_t *resNonces;
const uint32_t first_nonce = pdata[19];
int dev_id = device_map[thr_id];
int intensity = (device_sm[dev_id] >= 500 && !is_windows()) ? 28 : 25;
if (device_sm[dev_id] < 350) intensity = 22;
const int dev_id = device_map[thr_id];
int rc = 0;
int intensity = is_windows() ? 25 : 28;
uint32_t throughput = cuda_default_throughput(thr_id, 1U << intensity);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark) {
ptarget[6] = swab32(0xFFFF0);
ptarget[7] = 0;
}
const dim3 grid((throughput + (NPT*TPB)-1)/(NPT*TPB));
const dim3 block(TPB);
if (!init[thr_id])
{
@@ -429,73 +454,97 @@ extern "C" int scanhash_blake2s(int thr_id, struct work *work, uint32_t max_nonc
cudaDeviceReset();
// reduce cpu usage (linux)
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], maxResults * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], maxResults * sizeof(uint32_t)), -1);
init[thr_id] = true;
}
resNonces = h_resNonce[thr_id];
for (int i=0; i < 19; i++) {
be32enc(&endiandata[i], pdata[i]);
}
blake2s_setBlock(endiandata,ptarget[7]);
// midstate
memset(s_midstate.buf, 0, sizeof(s_midstate.buf));
blake2s_init(&s_midstate, BLAKE2S_OUTBYTES);
blake2s_update(&s_midstate, (uint8_t*) endiandata, MIDLEN);
memcpy(&s_ctx, &s_midstate, sizeof(blake2s_state));
blake2s_setBlock(endiandata, &s_midstate);
const uint2 target = make_uint2(ptarget[6], ptarget[7]);
cudaMemset(d_resNonce[thr_id], 0x00, maxResults*sizeof(uint32_t));
do {
uint32_t foundNonce = blake2s_hash_cuda(thr_id, throughput, pdata[19], target, swap);
if(ptarget[7]) {
blake2s_gpu_hash_nonce<<<grid, block>>>(throughput,pdata[19],d_resNonce[thr_id],ptarget[7]);
} else {
blake2s_gpu_hash_nonce<<<grid, block>>>(throughput,pdata[19],d_resNonce[thr_id]);
}
cudaMemcpy(resNonces, d_resNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
*hashes_done = pdata[19] - first_nonce + throughput;
if (foundNonce != UINT32_MAX)
if(resNonces[0])
{
uint32_t _ALIGN(A) vhashcpu[8];
cudaMemcpy(resNonces, d_resNonce[thr_id], maxResults*sizeof(uint32_t), cudaMemcpyDeviceToHost);
cudaMemset(d_resNonce[thr_id], 0x00, sizeof(uint32_t));
//blake2s_hash(vhashcpu, endiandata);
endiandata[19] = swab32_if(foundNonce, swap);
blake2s_hash_end(vhashcpu, endiandata);
if(resNonces[0] >= maxResults) {
gpulog(LOG_WARNING, thr_id, "candidates flood: %u", resNonces[0]);
resNonces[0] = maxResults-1;
}
if (vhashcpu[7] <= target.y && fulltest(vhashcpu, ptarget)) {
uint32_t vhashcpu[8];
uint32_t nonce = sph_bswap32(resNonces[1]);
be32enc(&endiandata[19], nonce);
blake2s_hash(vhashcpu, endiandata);
*hashes_done = pdata[19] - first_nonce + throughput;
if(vhashcpu[6] <= ptarget[6] && fulltest(vhashcpu, ptarget))
{
work_set_target_ratio(work, vhashcpu);
pdata[19] = work->nonces[0] = swab32_if(foundNonce, !swap);
#if NBN > 1
if (extra_results[0] != UINT32_MAX) {
endiandata[19] = swab32_if(extra_results[0], swap);
blake2s_hash_end(vhashcpu, endiandata);
if (vhashcpu[7] <= target.y && fulltest(vhashcpu, ptarget)) {
work->nonces[1] = swab32_if(extra_results[0], !swap);
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio) {
work->nonces[0] = nonce;
rc = work->valid_nonces = 1;
// search for 2nd best nonce
for(uint32_t j=2; j <= resNonces[0]; j++)
{
nonce = sph_bswap32(resNonces[j]);
be32enc(&endiandata[19], nonce);
blake2s_hash(vhashcpu, endiandata);
if(vhashcpu[6] <= ptarget[6] && fulltest(vhashcpu, ptarget))
{
gpulog(LOG_DEBUG, thr_id, "Multiple nonces: 1/%08x - %u/%08x", work->nonces[0], j, nonce);
work->nonces[1] = nonce;
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio[0]) {
work->shareratio[1] = work->shareratio[0];
work->sharediff[1] = work->sharediff[0];
xchg(work->nonces[1], work->nonces[0]);
work_set_target_ratio(work, vhashcpu);
xchg(work->nonces[1], pdata[19]);
} else if (work->valid_nonces == 1) {
bn_set_target_ratio(work, vhashcpu, 1);
}
return 2;
work->valid_nonces++;
rc = 2;
break;
}
extra_results[0] = UINT32_MAX;
}
#endif
return 1;
} else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNonce);
pdata[19] = max(work->nonces[0], work->nonces[1]); // next scan start
return rc;
} else if (vhashcpu[7] > ptarget[7]) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", resNonces[0]);
}
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart && max_nonce > (uint64_t)throughput + pdata[19]);
} while (!work_restart[thr_id].restart && (uint64_t)max_nonce > (uint64_t)throughput + pdata[19]);
*hashes_done = pdata[19] - first_nonce;
return 0;
return rc;
}
// cleanup
@@ -513,3 +562,4 @@ extern "C" void free_blake2s(int thr_id)
cudaDeviceSynchronize();
}
+39 -22
View File
@@ -6,9 +6,8 @@ extern "C" {
#include "sph/sph_bmw.h"
}
#include "miner.h"
#include "cuda_helper.h"
#include <miner.h>
#include <cuda_helper.h>
static uint32_t *d_hash[MAX_GPUS];
@@ -34,17 +33,12 @@ extern "C" void bmw_hash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
static __inline uint32_t swab32_if(uint32_t val, bool iftrue) {
return iftrue ? swab32(val) : val;
}
extern "C" int scanhash_bmw(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
bool swapnonce = true;
uint32_t throughput = cuda_default_throughput(thr_id, 1U << 21);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
@@ -53,6 +47,12 @@ extern "C" int scanhash_bmw(int thr_id, struct work* work, uint32_t max_nonce, u
if (!init[thr_id]) {
cudaSetDevice(device_map[thr_id]);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_check_cpu_init(thr_id, throughput);
bmw256_midstate_init(thr_id, throughput);
@@ -72,22 +72,39 @@ extern "C" int scanhash_bmw(int thr_id, struct work* work, uint32_t max_nonce, u
cuda_check_cpu_setTarget(ptarget);
do {
bmw256_cpu_hash_80(thr_id, (int) throughput, pdata[19], d_hash[thr_id], (int) swapnonce);
uint32_t foundNonce = cuda_check_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
uint32_t _ALIGN(64) vhash64[8];
endiandata[19] = swab32_if(foundNonce, swapnonce);
bmw_hash(vhash64, endiandata);
bmw256_cpu_hash_80(thr_id, (int) throughput, pdata[19], d_hash[thr_id], 1);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
*hashes_done = foundNonce - first_nonce + 1;
pdata[19] = swab32_if(foundNonce,!swapnonce);
work_set_target_ratio(work, vhash64);
return 1;
*hashes_done = pdata[19] - first_nonce + throughput;
work->nonces[0] = cuda_check_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (work->nonces[0] != UINT32_MAX)
{
const uint32_t Htarg = ptarget[7];
uint32_t _ALIGN(64) vhash[8];
be32enc(&endiandata[19], work->nonces[0]);
bmw_hash(vhash, endiandata);
if (vhash[7] <= Htarg && fulltest(vhash, ptarget)) {
work->valid_nonces = 1;
work_set_target_ratio(work, vhash);
work->nonces[1] = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
if (work->nonces[1] != 0) {
be32enc(&endiandata[19], work->nonces[1]);
bmw_hash(vhash, endiandata);
bn_set_target_ratio(work, vhash, 1);
work->valid_nonces++;
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
} else {
pdata[19] = work->nonces[0] + 1; // cursor
}
return work->valid_nonces;
}
else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNonce);
else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
pdata[19] = work->nonces[0] + 1;
continue;
}
}
+341 -8
View File
@@ -2,6 +2,8 @@
* Blake-256 Cuda Kernel (Tested on SM 5.0)
*
* Tanguy Pruvot - Nov. 2014
*
* + merged blake+keccak kernel for lyra2v2
*/
extern "C" {
#include "sph/sph_blake.h"
@@ -11,11 +13,25 @@ extern "C" {
#include <memory.h>
#ifdef __INTELLISENSE__
/* just for vstudio code colors */
__device__ uint32_t __byte_perm(uint32_t a, uint32_t b, uint32_t c);
#endif
#define UINT2(x,y) make_uint2(x,y)
__device__ __inline__ uint2 ROR8(const uint2 a) {
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x0765);
result.y = __byte_perm(a.x, a.y, 0x0765);
return result;
}
static __device__ uint64_t cuda_swab32ll(uint64_t x) {
return MAKE_ULONGLONG(cuda_swab32(_LODWORD(x)), cuda_swab32(_HIDWORD(x)));
}
__constant__ static uint32_t c_data[3+1];
__constant__ static uint32_t c_data[3 + 1];
__constant__ static uint32_t sigma[16][16];
static uint32_t c_sigma[16][16] = {
@@ -58,9 +74,25 @@ static const uint32_t c_u256[16] = {
0x3F84D5B5, 0xB5470917
};
__constant__ uint2 keccak_round_constants35[24] = {
{ 0x00000001ul, 0x00000000 }, { 0x00008082ul, 0x00000000 },
{ 0x0000808aul, 0x80000000 }, { 0x80008000ul, 0x80000000 },
{ 0x0000808bul, 0x00000000 }, { 0x80000001ul, 0x00000000 },
{ 0x80008081ul, 0x80000000 }, { 0x00008009ul, 0x80000000 },
{ 0x0000008aul, 0x00000000 }, { 0x00000088ul, 0x00000000 },
{ 0x80008009ul, 0x00000000 }, { 0x8000000aul, 0x00000000 },
{ 0x8000808bul, 0x00000000 }, { 0x0000008bul, 0x80000000 },
{ 0x00008089ul, 0x80000000 }, { 0x00008003ul, 0x80000000 },
{ 0x00008002ul, 0x80000000 }, { 0x00000080ul, 0x80000000 },
{ 0x0000800aul, 0x00000000 }, { 0x8000000aul, 0x80000000 },
{ 0x80008081ul, 0x80000000 }, { 0x00008080ul, 0x80000000 },
{ 0x80000001ul, 0x00000000 }, { 0x80008008ul, 0x80000000 }
};
#define GS2(a,b,c,d,x) { \
const uint32_t idx1 = sigma[r][x]; \
const uint32_t idx2 = sigma[r][x+1]; \
const uint32_t idx2 = sigma[r][(x)+1]; \
v[a] += (m[idx1] ^ u256[idx2]) + v[b]; \
v[d] = SPH_ROTL32(v[d] ^ v[a], 16); \
v[c] += v[d]; \
@@ -76,7 +108,7 @@ static const uint32_t c_u256[16] = {
//#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#define hostGS(a,b,c,d,x) { \
const uint32_t idx1 = c_sigma[r][x]; \
const uint32_t idx2 = c_sigma[r][x+1]; \
const uint32_t idx2 = c_sigma[r][(x)+1]; \
v[a] += (m[idx1] ^ c_u256[idx2]) + v[b]; \
v[d] = ROTR32(v[d] ^ v[a], 16); \
v[c] += v[d]; \
@@ -88,6 +120,17 @@ static const uint32_t c_u256[16] = {
v[b] = ROTR32(v[b] ^ v[c], 7); \
}
#define GSPREC(a,b,c,d,x,y) { \
v[a] += (m[x] ^ u256[y]) + v[b]; \
v[d] = __byte_perm(v[d] ^ v[a],0, 0x1032); \
v[c] += v[d]; \
v[b] = SPH_ROTR32(v[b] ^ v[c], 12); \
v[a] += (m[y] ^ u256[x]) + v[b]; \
v[d] = __byte_perm(v[d] ^ v[a],0, 0x0321); \
v[c] += v[d]; \
v[b] = SPH_ROTR32(v[b] ^ v[c], 7); \
}
/* Second part (64-80) msg never change, store it */
__device__ __constant__ static const uint32_t c_Padding[16] = {
0, 0, 0, 0,
@@ -158,8 +201,8 @@ static void blake256_compress2nd(uint32_t *h, const uint32_t *block, const uint3
for (int i = 0; i < 8; i++)
v[i] = h[i];
v[8] = u256[0];
v[9] = u256[1];
v[8] = u256[0];
v[9] = u256[1];
v[10] = u256[2];
v[11] = u256[3];
@@ -189,7 +232,273 @@ static void blake256_compress2nd(uint32_t *h, const uint32_t *block, const uint3
}
}
__global__ __launch_bounds__(256,3)
static void __forceinline__ __device__ keccak_block(uint2 *s)
{
uint2 bc[5], tmpxor[5], u, v;
// uint2 s[25];
#pragma unroll 1
for (int i = 0; i < 24; i++)
{
#pragma unroll
for (uint32_t x = 0; x < 5; x++)
tmpxor[x] = s[x] ^ s[x + 5] ^ s[x + 10] ^ s[x + 15] ^ s[x + 20];
bc[0] = tmpxor[0] ^ ROL2(tmpxor[2], 1);
bc[1] = tmpxor[1] ^ ROL2(tmpxor[3], 1);
bc[2] = tmpxor[2] ^ ROL2(tmpxor[4], 1);
bc[3] = tmpxor[3] ^ ROL2(tmpxor[0], 1);
bc[4] = tmpxor[4] ^ ROL2(tmpxor[1], 1);
u = s[1] ^ bc[0];
s[0] ^= bc[4];
s[1] = ROL2(s[6] ^ bc[0], 44);
s[6] = ROL2(s[9] ^ bc[3], 20);
s[9] = ROL2(s[22] ^ bc[1], 61);
s[22] = ROL2(s[14] ^ bc[3], 39);
s[14] = ROL2(s[20] ^ bc[4], 18);
s[20] = ROL2(s[2] ^ bc[1], 62);
s[2] = ROL2(s[12] ^ bc[1], 43);
s[12] = ROL2(s[13] ^ bc[2], 25);
s[13] = ROL8(s[19] ^ bc[3]);
s[19] = ROR8(s[23] ^ bc[2]);
s[23] = ROL2(s[15] ^ bc[4], 41);
s[15] = ROL2(s[4] ^ bc[3], 27);
s[4] = ROL2(s[24] ^ bc[3], 14);
s[24] = ROL2(s[21] ^ bc[0], 2);
s[21] = ROL2(s[8] ^ bc[2], 55);
s[8] = ROL2(s[16] ^ bc[0], 45);
s[16] = ROL2(s[5] ^ bc[4], 36);
s[5] = ROL2(s[3] ^ bc[2], 28);
s[3] = ROL2(s[18] ^ bc[2], 21);
s[18] = ROL2(s[17] ^ bc[1], 15);
s[17] = ROL2(s[11] ^ bc[0], 10);
s[11] = ROL2(s[7] ^ bc[1], 6);
s[7] = ROL2(s[10] ^ bc[4], 3);
s[10] = ROL2(u, 1);
u = s[0]; v = s[1]; s[0] ^= (~v) & s[2]; s[1] ^= (~s[2]) & s[3]; s[2] ^= (~s[3]) & s[4]; s[3] ^= (~s[4]) & u; s[4] ^= (~u) & v;
u = s[5]; v = s[6]; s[5] ^= (~v) & s[7]; s[6] ^= (~s[7]) & s[8]; s[7] ^= (~s[8]) & s[9]; s[8] ^= (~s[9]) & u; s[9] ^= (~u) & v;
u = s[10]; v = s[11]; s[10] ^= (~v) & s[12]; s[11] ^= (~s[12]) & s[13]; s[12] ^= (~s[13]) & s[14]; s[13] ^= (~s[14]) & u; s[14] ^= (~u) & v;
u = s[15]; v = s[16]; s[15] ^= (~v) & s[17]; s[16] ^= (~s[17]) & s[18]; s[17] ^= (~s[18]) & s[19]; s[18] ^= (~s[19]) & u; s[19] ^= (~u) & v;
u = s[20]; v = s[21]; s[20] ^= (~v) & s[22]; s[21] ^= (~s[22]) & s[23]; s[22] ^= (~s[23]) & s[24]; s[23] ^= (~s[24]) & u; s[24] ^= (~u) & v;
s[0] ^= keccak_round_constants35[i];
}
}
//__launch_bounds__(256)
__global__
void blakeKeccak256_gpu_hash_80(const uint32_t threads, const uint32_t startNonce, uint32_t * Hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
const uint32_t nonce = startNonce + thread;
const uint32_t T0 = 640;
uint32_t h[8];
#pragma unroll 8
for (int i = 0; i<8; i++) { h[i] = cpu_h[i]; }
uint32_t v[16];
const uint32_t c_Padding[12] = {
0x80000000, 0, 0, 0,
0, 0, 0, 0,
0, 1, 0, 640
};
const uint32_t u256[16] = {
0x243F6A88, 0x85A308D3,
0x13198A2E, 0x03707344,
0xA4093822, 0x299F31D0,
0x082EFA98, 0xEC4E6C89,
0x452821E6, 0x38D01377,
0xBE5466CF, 0x34E90C6C,
0xC0AC29B7, 0xC97C50DD,
0x3F84D5B5, 0xB5470917
};
uint32_t m[16] = {
c_data[0], c_data[1], c_data[2], nonce,
c_Padding[0], c_Padding[1], c_Padding[2], c_Padding[3],
c_Padding[4], c_Padding[5], c_Padding[6], c_Padding[7],
c_Padding[8], c_Padding[9], c_Padding[10], c_Padding[11]
};
#pragma unroll 8
for (int i = 0; i < 8; i++)
v[i] = h[i];
v[8] = u256[0];
v[9] = u256[1];
v[10] = u256[2];
v[11] = u256[3];
v[12] = u256[4] ^ T0;
v[13] = u256[5] ^ T0;
v[14] = u256[6];
v[15] = u256[7];
// { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
GSPREC(0, 4, 0x8, 0xC, 0, 1);
GSPREC(1, 5, 0x9, 0xD, 2, 3);
GSPREC(2, 6, 0xA, 0xE, 4, 5);
GSPREC(3, 7, 0xB, 0xF, 6, 7);
GSPREC(0, 5, 0xA, 0xF, 8, 9);
GSPREC(1, 6, 0xB, 0xC, 10, 11);
GSPREC(2, 7, 0x8, 0xD, 12, 13);
GSPREC(3, 4, 0x9, 0xE, 14, 15);
// { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
GSPREC(0, 4, 0x8, 0xC, 14, 10);
GSPREC(1, 5, 0x9, 0xD, 4, 8);
GSPREC(2, 6, 0xA, 0xE, 9, 15);
GSPREC(3, 7, 0xB, 0xF, 13, 6);
GSPREC(0, 5, 0xA, 0xF, 1, 12);
GSPREC(1, 6, 0xB, 0xC, 0, 2);
GSPREC(2, 7, 0x8, 0xD, 11, 7);
GSPREC(3, 4, 0x9, 0xE, 5, 3);
// { 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
GSPREC(0, 4, 0x8, 0xC, 11, 8);
GSPREC(1, 5, 0x9, 0xD, 12, 0);
GSPREC(2, 6, 0xA, 0xE, 5, 2);
GSPREC(3, 7, 0xB, 0xF, 15, 13);
GSPREC(0, 5, 0xA, 0xF, 10, 14);
GSPREC(1, 6, 0xB, 0xC, 3, 6);
GSPREC(2, 7, 0x8, 0xD, 7, 1);
GSPREC(3, 4, 0x9, 0xE, 9, 4);
// { 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
GSPREC(0, 4, 0x8, 0xC, 7, 9);
GSPREC(1, 5, 0x9, 0xD, 3, 1);
GSPREC(2, 6, 0xA, 0xE, 13, 12);
GSPREC(3, 7, 0xB, 0xF, 11, 14);
GSPREC(0, 5, 0xA, 0xF, 2, 6);
GSPREC(1, 6, 0xB, 0xC, 5, 10);
GSPREC(2, 7, 0x8, 0xD, 4, 0);
GSPREC(3, 4, 0x9, 0xE, 15, 8);
// { 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 },
GSPREC(0, 4, 0x8, 0xC, 9, 0);
GSPREC(1, 5, 0x9, 0xD, 5, 7);
GSPREC(2, 6, 0xA, 0xE, 2, 4);
GSPREC(3, 7, 0xB, 0xF, 10, 15);
GSPREC(0, 5, 0xA, 0xF, 14, 1);
GSPREC(1, 6, 0xB, 0xC, 11, 12);
GSPREC(2, 7, 0x8, 0xD, 6, 8);
GSPREC(3, 4, 0x9, 0xE, 3, 13);
// { 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 },
GSPREC(0, 4, 0x8, 0xC, 2, 12);
GSPREC(1, 5, 0x9, 0xD, 6, 10);
GSPREC(2, 6, 0xA, 0xE, 0, 11);
GSPREC(3, 7, 0xB, 0xF, 8, 3);
GSPREC(0, 5, 0xA, 0xF, 4, 13);
GSPREC(1, 6, 0xB, 0xC, 7, 5);
GSPREC(2, 7, 0x8, 0xD, 15, 14);
GSPREC(3, 4, 0x9, 0xE, 1, 9);
// { 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 },
GSPREC(0, 4, 0x8, 0xC, 12, 5);
GSPREC(1, 5, 0x9, 0xD, 1, 15);
GSPREC(2, 6, 0xA, 0xE, 14, 13);
GSPREC(3, 7, 0xB, 0xF, 4, 10);
GSPREC(0, 5, 0xA, 0xF, 0, 7);
GSPREC(1, 6, 0xB, 0xC, 6, 3);
GSPREC(2, 7, 0x8, 0xD, 9, 2);
GSPREC(3, 4, 0x9, 0xE, 8, 11);
// { 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 },
GSPREC(0, 4, 0x8, 0xC, 13, 11);
GSPREC(1, 5, 0x9, 0xD, 7, 14);
GSPREC(2, 6, 0xA, 0xE, 12, 1);
GSPREC(3, 7, 0xB, 0xF, 3, 9);
GSPREC(0, 5, 0xA, 0xF, 5, 0);
GSPREC(1, 6, 0xB, 0xC, 15, 4);
GSPREC(2, 7, 0x8, 0xD, 8, 6);
GSPREC(3, 4, 0x9, 0xE, 2, 10);
// { 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 },
GSPREC(0, 4, 0x8, 0xC, 6, 15);
GSPREC(1, 5, 0x9, 0xD, 14, 9);
GSPREC(2, 6, 0xA, 0xE, 11, 3);
GSPREC(3, 7, 0xB, 0xF, 0, 8);
GSPREC(0, 5, 0xA, 0xF, 12, 2);
GSPREC(1, 6, 0xB, 0xC, 13, 7);
GSPREC(2, 7, 0x8, 0xD, 1, 4);
GSPREC(3, 4, 0x9, 0xE, 10, 5);
// { 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0 },
GSPREC(0, 4, 0x8, 0xC, 10, 2);
GSPREC(1, 5, 0x9, 0xD, 8, 4);
GSPREC(2, 6, 0xA, 0xE, 7, 6);
GSPREC(3, 7, 0xB, 0xF, 1, 5);
GSPREC(0, 5, 0xA, 0xF, 15, 11);
GSPREC(1, 6, 0xB, 0xC, 9, 14);
GSPREC(2, 7, 0x8, 0xD, 3, 12);
GSPREC(3, 4, 0x9, 0xE, 13, 0);
// { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 },
GSPREC(0, 4, 0x8, 0xC, 0, 1);
GSPREC(1, 5, 0x9, 0xD, 2, 3);
GSPREC(2, 6, 0xA, 0xE, 4, 5);
GSPREC(3, 7, 0xB, 0xF, 6, 7);
GSPREC(0, 5, 0xA, 0xF, 8, 9);
GSPREC(1, 6, 0xB, 0xC, 10, 11);
GSPREC(2, 7, 0x8, 0xD, 12, 13);
GSPREC(3, 4, 0x9, 0xE, 14, 15);
// { 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 },
GSPREC(0, 4, 0x8, 0xC, 14, 10);
GSPREC(1, 5, 0x9, 0xD, 4, 8);
GSPREC(2, 6, 0xA, 0xE, 9, 15);
GSPREC(3, 7, 0xB, 0xF, 13, 6);
GSPREC(0, 5, 0xA, 0xF, 1, 12);
GSPREC(1, 6, 0xB, 0xC, 0, 2);
GSPREC(2, 7, 0x8, 0xD, 11, 7);
GSPREC(3, 4, 0x9, 0xE, 5, 3);
// { 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 },
GSPREC(0, 4, 0x8, 0xC, 11, 8);
GSPREC(1, 5, 0x9, 0xD, 12, 0);
GSPREC(2, 6, 0xA, 0xE, 5, 2);
GSPREC(3, 7, 0xB, 0xF, 15, 13);
GSPREC(0, 5, 0xA, 0xF, 10, 14);
GSPREC(1, 6, 0xB, 0xC, 3, 6);
GSPREC(2, 7, 0x8, 0xD, 7, 1);
GSPREC(3, 4, 0x9, 0xE, 9, 4);
// { 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 },
GSPREC(0, 4, 0x8, 0xC, 7, 9);
GSPREC(1, 5, 0x9, 0xD, 3, 1);
GSPREC(2, 6, 0xA, 0xE, 13, 12);
GSPREC(3, 7, 0xB, 0xF, 11, 14);
GSPREC(0, 5, 0xA, 0xF, 2, 6);
GSPREC(1, 6, 0xB, 0xC, 5, 10);
GSPREC(2, 7, 0x8, 0xD, 4, 0);
GSPREC(3, 4, 0x9, 0xE, 15, 8);
h[0] = cuda_swab32(h[0] ^ v[0] ^ v[8]);
h[1] = cuda_swab32(h[1] ^ v[1] ^ v[9]);
h[2] = cuda_swab32(h[2] ^ v[2] ^ v[10]);
h[3] = cuda_swab32(h[3] ^ v[3] ^ v[11]);
h[4] = cuda_swab32(h[4] ^ v[4] ^ v[12]);
h[5] = cuda_swab32(h[5] ^ v[5] ^ v[13]);
h[6] = cuda_swab32(h[6] ^ v[6] ^ v[14]);
h[7] = cuda_swab32(h[7] ^ v[7] ^ v[15]);
uint2 keccak_gpu_state[25] = { 0 };
keccak_gpu_state[0].x = h[0];
keccak_gpu_state[0].y = h[1];
keccak_gpu_state[1].x = h[2];
keccak_gpu_state[1].y = h[3];
keccak_gpu_state[2].x = h[4];
keccak_gpu_state[2].y = h[5];
keccak_gpu_state[3].x = h[6];
keccak_gpu_state[3].y = h[7];
keccak_gpu_state[4] = UINT2(1, 0);
keccak_gpu_state[16] = UINT2(0, 0x80000000);
keccak_block(keccak_gpu_state);
uint64_t *outputHash = (uint64_t *)Hash;
#pragma unroll 4
for (int i = 0; i<4; i++)
outputHash[i*threads + thread] = devectorize(keccak_gpu_state[i]);
}
}
__global__ __launch_bounds__(256, 3)
void blake256_gpu_hash_80(const uint32_t threads, const uint32_t startNonce, uint64_t * Hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
@@ -209,7 +518,7 @@ void blake256_gpu_hash_80(const uint32_t threads, const uint32_t startNonce, uin
#pragma unroll
for (int i = 0; i<4; i++) {
Hash[i*threads + thread] = cuda_swab32ll(MAKE_ULONGLONG(h[2 * i], h[2*i+1]));
Hash[i*threads + thread] = cuda_swab32ll(MAKE_ULONGLONG(h[2 * i], h[2 * i + 1]));
}
}
}
@@ -222,7 +531,7 @@ void blake256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
blake256_gpu_hash_80 <<<grid, block>>> (threads, startNonce, Hash);
blake256_gpu_hash_80 << <grid, block >> > (threads, startNonce, Hash);
MyStreamSynchronize(NULL, order, thr_id);
}
@@ -245,3 +554,27 @@ void blake256_cpu_init(int thr_id, uint32_t threads)
cudaMemcpyToSymbol(u256, c_u256, sizeof(c_u256), 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(sigma, c_sigma, sizeof(c_sigma), 0, cudaMemcpyHostToDevice);
}
/** for lyra2v2 **/
__host__
void blakeKeccak256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32_t startNonce, uint64_t *Hash, int order)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
blakeKeccak256_gpu_hash_80 <<<grid, block>>> (threads, startNonce, (uint32_t *)Hash);
}
__host__
void blakeKeccak256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32_t startNonce, uint64_t *Hash, int order, cudaStream_t stream)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
blakeKeccak256_gpu_hash_80 <<<grid, block, 0, stream>>> (threads, startNonce, (uint32_t *)Hash);
}
+268 -203
View File
@@ -3,181 +3,249 @@
#define CUBEHASH_ROUNDS 16 /* this is r for CubeHashr/b */
#define CUBEHASH_BLOCKBYTES 32 /* this is b for CubeHashr/b */
#ifdef __INTELLISENSE__
/* just for vstudio code colors */
#define __CUDA_ARCH__ 520
#endif
#if __CUDA_ARCH__ < 350
#define LROT(x,bits) ((x << bits) | (x >> (32 - bits)))
#else
#define LROT(x, bits) __funnelshift_l(x, x, bits)
#endif
#if __CUDA_ARCH__ < 500
#define TPB 576
#else
#define TPB 1024
#endif
#define TPB35 576
#define TPB50 1024
#define ROTATEUPWARDS7(a) LROT(a,7)
#define ROTATEUPWARDS11(a) LROT(a,11)
//#define SWAP(a,b) { uint32_t u = a; a = b; b = u; }
#define SWAP(a,b) { a ^= b; b ^= a; a ^= b; }
__device__ __forceinline__ void rrounds(uint32_t x[2][2][2][2][2])
{
int r;
int j;
int k;
int l;
int m;
#pragma unroll 2
for (r = 0; r < CUBEHASH_ROUNDS; ++r) {
/* "add x_0jklm into x_1jklmn modulo 2^32" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[1][j][k][l][m] += x[0][j][k][l][m];
uint32_t x0[2][2][2][2];
uint32_t x1[2][2][2][2];
for (r = 0; r < CUBEHASH_ROUNDS; r += 2) {
/* "rotate x_0jklm upwards by 7 bits" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] = ROTATEUPWARDS7(x[0][j][k][l][m]);
x0[0][0][0][0] = ROTATEUPWARDS7(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS7(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS7(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS7(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS7(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS7(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS7(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS7(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS7(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS7(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS7(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS7(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS7(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS7(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS7(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS7(x[0][1][1][1][1]);
/* "swap x_00klm with x_01klm" */
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
SWAP(x[0][0][k][l][m], x[0][1][k][l][m])
/* "add x_0jklm into x_1jklm modulo 2^32" */
x1[0][0][0][0] = x[1][0][0][0][0] + x[0][0][0][0][0];
x1[0][0][0][1] = x[1][0][0][0][1] + x[0][0][0][0][1];
x1[0][0][1][0] = x[1][0][0][1][0] + x[0][0][0][1][0];
x1[0][0][1][1] = x[1][0][0][1][1] + x[0][0][0][1][1];
x1[0][1][0][0] = x[1][0][1][0][0] + x[0][0][1][0][0];
x1[0][1][0][1] = x[1][0][1][0][1] + x[0][0][1][0][1];
x1[0][1][1][0] = x[1][0][1][1][0] + x[0][0][1][1][0];
x1[0][1][1][1] = x[1][0][1][1][1] + x[0][0][1][1][1];
x1[1][0][0][0] = x[1][1][0][0][0] + x[0][1][0][0][0];
x1[1][0][0][1] = x[1][1][0][0][1] + x[0][1][0][0][1];
x1[1][0][1][0] = x[1][1][0][1][0] + x[0][1][0][1][0];
x1[1][0][1][1] = x[1][1][0][1][1] + x[0][1][0][1][1];
x1[1][1][0][0] = x[1][1][1][0][0] + x[0][1][1][0][0];
x1[1][1][0][1] = x[1][1][1][0][1] + x[0][1][1][0][1];
x1[1][1][1][0] = x[1][1][1][1][0] + x[0][1][1][1][0];
x1[1][1][1][1] = x[1][1][1][1][1] + x[0][1][1][1][1];
/* "xor x_1jklm into x_0jklm" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] ^= x[1][j][k][l][m];
/* "swap x_1jk0m with x_1jk1m" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (m = 0; m < 2; ++m)
SWAP(x[1][j][k][0][m], x[1][j][k][1][m])
/* "add x_0jklm into x_1jklm modulo 2^32" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[1][j][k][l][m] += x[0][j][k][l][m];
/* "xor x_1~jklm into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x1[1][0][0][0];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x1[1][0][0][1];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x1[1][0][1][0];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x1[1][0][1][1];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x1[1][1][0][0];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x1[1][1][0][1];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x1[1][1][1][0];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x1[1][1][1][1];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x1[0][0][0][0];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x1[0][0][0][1];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x1[0][0][1][0];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x1[0][0][1][1];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x1[0][1][0][0];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x1[0][1][0][1];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x1[0][1][1][0];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x1[0][1][1][1];
/* "rotate x_0jklm upwards by 11 bits" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] = ROTATEUPWARDS11(x[0][j][k][l][m]);
x0[0][0][0][0] = ROTATEUPWARDS11(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS11(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS11(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS11(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS11(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS11(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS11(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS11(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS11(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS11(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS11(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS11(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS11(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS11(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS11(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS11(x[0][1][1][1][1]);
/* "swap x_0j0lm with x_0j1lm" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
SWAP(x[0][j][0][l][m], x[0][j][1][l][m])
/* "add x_0jklm into x_1~jk~lm modulo 2^32" */
x[1][1][0][1][0] = x1[1][0][1][0] + x[0][0][0][0][0];
x[1][1][0][1][1] = x1[1][0][1][1] + x[0][0][0][0][1];
x[1][1][0][0][0] = x1[1][0][0][0] + x[0][0][0][1][0];
x[1][1][0][0][1] = x1[1][0][0][1] + x[0][0][0][1][1];
x[1][1][1][1][0] = x1[1][1][1][0] + x[0][0][1][0][0];
x[1][1][1][1][1] = x1[1][1][1][1] + x[0][0][1][0][1];
x[1][1][1][0][0] = x1[1][1][0][0] + x[0][0][1][1][0];
x[1][1][1][0][1] = x1[1][1][0][1] + x[0][0][1][1][1];
x[1][0][0][1][0] = x1[0][0][1][0] + x[0][1][0][0][0];
x[1][0][0][1][1] = x1[0][0][1][1] + x[0][1][0][0][1];
x[1][0][0][0][0] = x1[0][0][0][0] + x[0][1][0][1][0];
x[1][0][0][0][1] = x1[0][0][0][1] + x[0][1][0][1][1];
x[1][0][1][1][0] = x1[0][1][1][0] + x[0][1][1][0][0];
x[1][0][1][1][1] = x1[0][1][1][1] + x[0][1][1][0][1];
x[1][0][1][0][0] = x1[0][1][0][0] + x[0][1][1][1][0];
x[1][0][1][0][1] = x1[0][1][0][1] + x[0][1][1][1][1];
/* "xor x_1jklm into x_0jklm" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
#pragma unroll 2
for (m = 0; m < 2; ++m)
x[0][j][k][l][m] ^= x[1][j][k][l][m];
/* "xor x_1~j~k~lm into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x[1][1][1][1][0];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x[1][1][1][1][1];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x[1][1][1][0][0];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x[1][1][1][0][1];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x[1][1][0][1][0];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x[1][1][0][1][1];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x[1][1][0][0][0];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x[1][1][0][0][1];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x[1][0][1][1][0];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x[1][0][1][1][1];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x[1][0][1][0][0];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x[1][0][1][0][1];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x[1][0][0][1][0];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x[1][0][0][1][1];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x[1][0][0][0][0];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x[1][0][0][0][1];
/* "swap x_1jkl0 with x_1jkl1" */
#pragma unroll 2
for (j = 0; j < 2; ++j)
#pragma unroll 2
for (k = 0; k < 2; ++k)
#pragma unroll 2
for (l = 0; l < 2; ++l)
SWAP(x[1][j][k][l][0], x[1][j][k][l][1])
/* "rotate x_0jklm upwards by 7 bits" */
x0[0][0][0][0] = ROTATEUPWARDS7(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS7(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS7(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS7(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS7(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS7(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS7(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS7(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS7(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS7(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS7(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS7(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS7(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS7(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS7(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS7(x[0][1][1][1][1]);
/* "add x_0jklm into x_1~j~k~l~m modulo 2^32" */
x1[1][1][1][1] = x[1][1][1][1][1] + x[0][0][0][0][0];
x1[1][1][1][0] = x[1][1][1][1][0] + x[0][0][0][0][1];
x1[1][1][0][1] = x[1][1][1][0][1] + x[0][0][0][1][0];
x1[1][1][0][0] = x[1][1][1][0][0] + x[0][0][0][1][1];
x1[1][0][1][1] = x[1][1][0][1][1] + x[0][0][1][0][0];
x1[1][0][1][0] = x[1][1][0][1][0] + x[0][0][1][0][1];
x1[1][0][0][1] = x[1][1][0][0][1] + x[0][0][1][1][0];
x1[1][0][0][0] = x[1][1][0][0][0] + x[0][0][1][1][1];
x1[0][1][1][1] = x[1][0][1][1][1] + x[0][1][0][0][0];
x1[0][1][1][0] = x[1][0][1][1][0] + x[0][1][0][0][1];
x1[0][1][0][1] = x[1][0][1][0][1] + x[0][1][0][1][0];
x1[0][1][0][0] = x[1][0][1][0][0] + x[0][1][0][1][1];
x1[0][0][1][1] = x[1][0][0][1][1] + x[0][1][1][0][0];
x1[0][0][1][0] = x[1][0][0][1][0] + x[0][1][1][0][1];
x1[0][0][0][1] = x[1][0][0][0][1] + x[0][1][1][1][0];
x1[0][0][0][0] = x[1][0][0][0][0] + x[0][1][1][1][1];
/* "xor x_1j~k~l~m into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x1[0][1][1][1];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x1[0][1][1][0];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x1[0][1][0][1];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x1[0][1][0][0];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x1[0][0][1][1];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x1[0][0][1][0];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x1[0][0][0][1];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x1[0][0][0][0];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x1[1][1][1][1];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x1[1][1][1][0];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x1[1][1][0][1];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x1[1][1][0][0];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x1[1][0][1][1];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x1[1][0][1][0];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x1[1][0][0][1];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x1[1][0][0][0];
/* "rotate x_0jklm upwards by 11 bits" */
x0[0][0][0][0] = ROTATEUPWARDS11(x[0][0][0][0][0]);
x0[0][0][0][1] = ROTATEUPWARDS11(x[0][0][0][0][1]);
x0[0][0][1][0] = ROTATEUPWARDS11(x[0][0][0][1][0]);
x0[0][0][1][1] = ROTATEUPWARDS11(x[0][0][0][1][1]);
x0[0][1][0][0] = ROTATEUPWARDS11(x[0][0][1][0][0]);
x0[0][1][0][1] = ROTATEUPWARDS11(x[0][0][1][0][1]);
x0[0][1][1][0] = ROTATEUPWARDS11(x[0][0][1][1][0]);
x0[0][1][1][1] = ROTATEUPWARDS11(x[0][0][1][1][1]);
x0[1][0][0][0] = ROTATEUPWARDS11(x[0][1][0][0][0]);
x0[1][0][0][1] = ROTATEUPWARDS11(x[0][1][0][0][1]);
x0[1][0][1][0] = ROTATEUPWARDS11(x[0][1][0][1][0]);
x0[1][0][1][1] = ROTATEUPWARDS11(x[0][1][0][1][1]);
x0[1][1][0][0] = ROTATEUPWARDS11(x[0][1][1][0][0]);
x0[1][1][0][1] = ROTATEUPWARDS11(x[0][1][1][0][1]);
x0[1][1][1][0] = ROTATEUPWARDS11(x[0][1][1][1][0]);
x0[1][1][1][1] = ROTATEUPWARDS11(x[0][1][1][1][1]);
/* "add x_0jklm into x_1j~kl~m modulo 2^32" */
x[1][0][1][0][1] = x1[0][1][0][1] + x[0][0][0][0][0];
x[1][0][1][0][0] = x1[0][1][0][0] + x[0][0][0][0][1];
x[1][0][1][1][1] = x1[0][1][1][1] + x[0][0][0][1][0];
x[1][0][1][1][0] = x1[0][1][1][0] + x[0][0][0][1][1];
x[1][0][0][0][1] = x1[0][0][0][1] + x[0][0][1][0][0];
x[1][0][0][0][0] = x1[0][0][0][0] + x[0][0][1][0][1];
x[1][0][0][1][1] = x1[0][0][1][1] + x[0][0][1][1][0];
x[1][0][0][1][0] = x1[0][0][1][0] + x[0][0][1][1][1];
x[1][1][1][0][1] = x1[1][1][0][1] + x[0][1][0][0][0];
x[1][1][1][0][0] = x1[1][1][0][0] + x[0][1][0][0][1];
x[1][1][1][1][1] = x1[1][1][1][1] + x[0][1][0][1][0];
x[1][1][1][1][0] = x1[1][1][1][0] + x[0][1][0][1][1];
x[1][1][0][0][1] = x1[1][0][0][1] + x[0][1][1][0][0];
x[1][1][0][0][0] = x1[1][0][0][0] + x[0][1][1][0][1];
x[1][1][0][1][1] = x1[1][0][1][1] + x[0][1][1][1][0];
x[1][1][0][1][0] = x1[1][0][1][0] + x[0][1][1][1][1];
/* "xor x_1jkl~m into x_0jklm" */
x[0][0][0][0][0] = x0[0][0][0][0] ^ x[1][0][0][0][1];
x[0][0][0][0][1] = x0[0][0][0][1] ^ x[1][0][0][0][0];
x[0][0][0][1][0] = x0[0][0][1][0] ^ x[1][0][0][1][1];
x[0][0][0][1][1] = x0[0][0][1][1] ^ x[1][0][0][1][0];
x[0][0][1][0][0] = x0[0][1][0][0] ^ x[1][0][1][0][1];
x[0][0][1][0][1] = x0[0][1][0][1] ^ x[1][0][1][0][0];
x[0][0][1][1][0] = x0[0][1][1][0] ^ x[1][0][1][1][1];
x[0][0][1][1][1] = x0[0][1][1][1] ^ x[1][0][1][1][0];
x[0][1][0][0][0] = x0[1][0][0][0] ^ x[1][1][0][0][1];
x[0][1][0][0][1] = x0[1][0][0][1] ^ x[1][1][0][0][0];
x[0][1][0][1][0] = x0[1][0][1][0] ^ x[1][1][0][1][1];
x[0][1][0][1][1] = x0[1][0][1][1] ^ x[1][1][0][1][0];
x[0][1][1][0][0] = x0[1][1][0][0] ^ x[1][1][1][0][1];
x[0][1][1][0][1] = x0[1][1][0][1] ^ x[1][1][1][0][0];
x[0][1][1][1][0] = x0[1][1][1][0] ^ x[1][1][1][1][1];
x[0][1][1][1][1] = x0[1][1][1][1] ^ x[1][1][1][1][0];
}
}
__device__ __forceinline__ void block_tox(const uint32_t *in, uint32_t x[2][2][2][2][2])
{
x[0][0][0][0][0] ^= in[0];
x[0][0][0][0][1] ^= in[1];
x[0][0][0][1][0] ^= in[2];
x[0][0][0][1][1] ^= in[3];
x[0][0][1][0][0] ^= in[4];
x[0][0][1][0][1] ^= in[5];
x[0][0][1][1][0] ^= in[6];
x[0][0][1][1][1] ^= in[7];
}
__device__ __forceinline__ void hash_fromx(uint32_t *out, uint32_t x[2][2][2][2][2])
{
out[0] = x[0][0][0][0][0];
out[1] = x[0][0][0][0][1];
out[2] = x[0][0][0][1][0];
out[3] = x[0][0][0][1][1];
out[4] = x[0][0][1][0][0];
out[5] = x[0][0][1][0][1];
out[6] = x[0][0][1][1][0];
out[7] = x[0][0][1][1][1];
}
__device__ __forceinline__
void Update32(uint32_t x[2][2][2][2][2], const uint32_t *data)
{
/* "xor the block into the first b bytes of the state" */
/* "and then transform the state invertibly through r identical rounds" */
block_tox(data, x);
rrounds(x);
}
__device__ __forceinline__
void Update32_const(uint32_t x[2][2][2][2][2])
{
x[0][0][0][0][0] ^= 0x80;
rrounds(x);
}
__device__ __forceinline__
void Final(uint32_t x[2][2][2][2][2], uint32_t *hashval)
{
@@ -185,27 +253,44 @@ void Final(uint32_t x[2][2][2][2][2], uint32_t *hashval)
x[1][1][1][1][1] ^= 1U;
/* "the state is then transformed invertibly through 10r identical rounds" */
#pragma unroll 2
for (int i = 0; i < 10; ++i) rrounds(x);
/* "output the first h/8 bytes of the state" */
hash_fromx(hashval, x);
hashval[0] = x[0][0][0][0][0];
hashval[1] = x[0][0][0][0][1];
hashval[2] = x[0][0][0][1][0];
hashval[3] = x[0][0][0][1][1];
hashval[4] = x[0][0][1][0][0];
hashval[5] = x[0][0][1][0][1];
hashval[6] = x[0][0][1][1][0];
hashval[7] = x[0][0][1][1][1];
}
#if __CUDA_ARCH__ >= 500
__global__ __launch_bounds__(TPB, 1)
__global__ __launch_bounds__(TPB50, 1)
#else
__global__ __launch_bounds__(TPB35, 1)
#endif
void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
#if __CUDA_ARCH__ >= 500
uint2 Hash[4];
Hash[0] = __ldg(&g_hash[thread]);
Hash[1] = __ldg(&g_hash[thread + 1 * threads]);
Hash[2] = __ldg(&g_hash[thread + 2 * threads]);
Hash[3] = __ldg(&g_hash[thread + 3 * threads]);
#else
uint32_t Hash[8];
LOHI(Hash[0], Hash[1], __ldg(&((uint64_t*)g_hash)[thread]));
LOHI(Hash[2], Hash[3], __ldg(&((uint64_t*)g_hash)[thread + 1 * threads]));
LOHI(Hash[4], Hash[5], __ldg(&((uint64_t*)g_hash)[thread + 2 * threads]));
LOHI(Hash[6], Hash[7], __ldg(&((uint64_t*)g_hash)[thread + 3 * threads]));
#endif
uint32_t x[2][2][2][2][2] =
{
@@ -219,6 +304,7 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_ha
0x15815AEB, 0x4AB6AAD6, 0x9CDAF8AF, 0xD6032C0A
};
#if __CUDA_ARCH__ >= 500
x[0][0][0][0][0] ^= Hash[0].x;
x[0][0][0][0][1] ^= Hash[0].y;
x[0][0][0][1][0] ^= Hash[1].x;
@@ -227,48 +313,7 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *g_ha
x[0][0][1][0][1] ^= Hash[2].y;
x[0][0][1][1][0] ^= Hash[3].x;
x[0][0][1][1][1] ^= Hash[3].y;
rrounds(x);
x[0][0][0][0][0] ^= 0x80U;
rrounds(x);
Final(x, (uint32_t*) Hash);
g_hash[thread] = Hash[0];
g_hash[1 * threads + thread] = Hash[1];
g_hash[2 * threads + thread] = Hash[2];
g_hash[3 * threads + thread] = Hash[3];
}
}
#else
__global__ __launch_bounds__(TPB, 1)
void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *d_hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t Hash[8];
uint64_t* g_hash = (uint64_t*) d_hash;
LOHI(Hash[0], Hash[1], __ldg(&g_hash[thread]));
LOHI(Hash[2], Hash[3], __ldg(&g_hash[thread + 1 * threads]));
LOHI(Hash[4], Hash[5], __ldg(&g_hash[thread + 2 * threads]));
LOHI(Hash[6], Hash[7], __ldg(&g_hash[thread + 3 * threads]));
uint32_t x[2][2][2][2][2] =
{
0xEA2BD4B4, 0xCCD6F29F, 0x63117E71, 0x35481EAE,
0x22512D5B, 0xE5D94E63, 0x7E624131, 0xF4CC12BE,
0xC2D0B696, 0x42AF2070, 0xD0720C35, 0x3361DA8C,
0x28CCECA4, 0x8EF8AD83, 0x4680AC00, 0x40E5FBAB,
0xD89041C3, 0x6107FBD5, 0x6C859D41, 0xF0B26679,
0x09392549, 0x5FA25603, 0x65C892FD, 0x93CB6285,
0x2AF2B5AE, 0x9E4B4E60, 0x774ABFDD, 0x85254725,
0x15815AEB, 0x4AB6AAD6, 0x9CDAF8AF, 0xD6032C0A
};
x[0][0][0][0][0] ^= Hash[0];
x[0][0][0][0][1] ^= Hash[1];
x[0][0][0][1][0] ^= Hash[2];
@@ -277,29 +322,49 @@ void cubehash256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *d_ha
x[0][0][1][0][1] ^= Hash[5];
x[0][0][1][1][0] ^= Hash[6];
x[0][0][1][1][1] ^= Hash[7];
#endif
rrounds(x);
x[0][0][0][0][0] ^= 0x80U;
rrounds(x);
#if __CUDA_ARCH__ >= 500
Final(x, (uint32_t*)Hash);
g_hash[thread] = Hash[0];
g_hash[1 * threads + thread] = Hash[1];
g_hash[2 * threads + thread] = Hash[2];
g_hash[3 * threads + thread] = Hash[3];
#else
Final(x, Hash);
g_hash[thread] = ((uint64_t*)Hash)[0];
g_hash[1 * threads + thread] = ((uint64_t*)Hash)[1];
g_hash[2 * threads + thread] = ((uint64_t*)Hash)[2];
g_hash[3 * threads + thread] = ((uint64_t*)Hash)[3];
((uint64_t*)g_hash)[thread] = ((uint64_t*)Hash)[0];
((uint64_t*)g_hash)[1 * threads + thread] = ((uint64_t*)Hash)[1];
((uint64_t*)g_hash)[2 * threads + thread] = ((uint64_t*)Hash)[2];
((uint64_t*)g_hash)[3 * threads + thread] = ((uint64_t*)Hash)[3];
#endif
}
}
#endif
__host__
void cubehash256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_hash, int order)
{
uint32_t tpb = TPB;
uint32_t tpb = TPB35;
if (cuda_arch[thr_id] >= 500) tpb = TPB50;
dim3 grid((threads + tpb-1)/tpb);
dim3 grid((threads + tpb - 1) / tpb);
dim3 block(tpb);
cubehash256_gpu_hash_32 <<<grid, block>>> (threads, startNounce, (uint2*) d_hash);
cubehash256_gpu_hash_32 <<<grid, block >>> (threads, startNounce, (uint2*)d_hash);
}
__host__
void cubehash256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_hash, int order, cudaStream_t stream)
{
uint32_t tpb = TPB35;
if (cuda_arch[thr_id] >= 500) tpb = TPB50;
dim3 grid((threads + tpb - 1) / tpb);
dim3 block(tpb);
cubehash256_gpu_hash_32 <<<grid, block, 0, stream >>> (threads, startNounce, (uint2*)d_hash);
}
+345 -271
View File
@@ -1,309 +1,383 @@
#include "miner.h"
/**
* KECCAK-256 CUDA optimised implementation, based on ccminer-alexis code
*/
#include <miner.h>
extern "C" {
#include <stdint.h>
#include <memory.h>
}
#include "cuda_helper.h"
#include <cuda_helper.h>
#include <cuda_vectors.h>
static const uint64_t host_keccak_round_constants[24] = {
0x0000000000000001ull, 0x0000000000008082ull,
0x800000000000808aull, 0x8000000080008000ull,
0x000000000000808bull, 0x0000000080000001ull,
0x8000000080008081ull, 0x8000000000008009ull,
0x000000000000008aull, 0x0000000000000088ull,
0x0000000080008009ull, 0x000000008000000aull,
0x000000008000808bull, 0x800000000000008bull,
0x8000000000008089ull, 0x8000000000008003ull,
0x8000000000008002ull, 0x8000000000000080ull,
0x000000000000800aull, 0x800000008000000aull,
0x8000000080008081ull, 0x8000000000008080ull,
0x0000000080000001ull, 0x8000000080008008ull
#define TPB52 1024
#define TPB50 384
#define NPT 2
#define NBN 2
static uint32_t *d_nonces[MAX_GPUS];
static uint32_t *h_nonces[MAX_GPUS];
__constant__ uint2 c_message48[6];
__constant__ uint2 c_mid[17];
__constant__ uint2 keccak_round_constants[24] = {
{ 0x00000001, 0x00000000 }, { 0x00008082, 0x00000000 }, { 0x0000808a, 0x80000000 }, { 0x80008000, 0x80000000 },
{ 0x0000808b, 0x00000000 }, { 0x80000001, 0x00000000 }, { 0x80008081, 0x80000000 }, { 0x00008009, 0x80000000 },
{ 0x0000008a, 0x00000000 }, { 0x00000088, 0x00000000 }, { 0x80008009, 0x00000000 }, { 0x8000000a, 0x00000000 },
{ 0x8000808b, 0x00000000 }, { 0x0000008b, 0x80000000 }, { 0x00008089, 0x80000000 }, { 0x00008003, 0x80000000 },
{ 0x00008002, 0x80000000 }, { 0x00000080, 0x80000000 }, { 0x0000800a, 0x00000000 }, { 0x8000000a, 0x80000000 },
{ 0x80008081, 0x80000000 }, { 0x00008080, 0x80000000 }, { 0x80000001, 0x00000000 }, { 0x80008008, 0x80000000 }
};
uint32_t *d_nounce[MAX_GPUS];
uint32_t *d_KNonce[MAX_GPUS];
__constant__ uint32_t pTarget[8];
__constant__ uint64_t keccak_round_constants[24];
__constant__ uint64_t c_PaddedMessage80[10]; // padded message (80 bytes + padding?)
#if __CUDA_ARCH__ >= 350
__device__ __forceinline__
static void keccak_blockv35(uint2 *s, const uint64_t *keccak_round_constants)
{
size_t i;
uint2 t[5], u[5], v, w;
#pragma unroll
for (i = 0; i < 24; i++) {
/* theta: c = a[0,i] ^ a[1,i] ^ .. a[4,i] */
t[0] = s[0] ^ s[5] ^ s[10] ^ s[15] ^ s[20];
t[1] = s[1] ^ s[6] ^ s[11] ^ s[16] ^ s[21];
t[2] = s[2] ^ s[7] ^ s[12] ^ s[17] ^ s[22];
t[3] = s[3] ^ s[8] ^ s[13] ^ s[18] ^ s[23];
t[4] = s[4] ^ s[9] ^ s[14] ^ s[19] ^ s[24];
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
u[0] = t[4] ^ ROL2(t[1], 1);
u[1] = t[0] ^ ROL2(t[2], 1);
u[2] = t[1] ^ ROL2(t[3], 1);
u[3] = t[2] ^ ROL2(t[4], 1);
u[4] = t[3] ^ ROL2(t[0], 1);
/* theta: a[0,i], a[1,i], .. a[4,i] ^= d[i] */
s[0] ^= u[0]; s[5] ^= u[0]; s[10] ^= u[0]; s[15] ^= u[0]; s[20] ^= u[0];
s[1] ^= u[1]; s[6] ^= u[1]; s[11] ^= u[1]; s[16] ^= u[1]; s[21] ^= u[1];
s[2] ^= u[2]; s[7] ^= u[2]; s[12] ^= u[2]; s[17] ^= u[2]; s[22] ^= u[2];
s[3] ^= u[3]; s[8] ^= u[3]; s[13] ^= u[3]; s[18] ^= u[3]; s[23] ^= u[3];
s[4] ^= u[4]; s[9] ^= u[4]; s[14] ^= u[4]; s[19] ^= u[4]; s[24] ^= u[4];
/* rho pi: b[..] = rotl(a[..], ..) */
v = s[1];
s[1] = ROL2(s[6], 44);
s[6] = ROL2(s[9], 20);
s[9] = ROL2(s[22], 61);
s[22] = ROL2(s[14], 39);
s[14] = ROL2(s[20], 18);
s[20] = ROL2(s[2], 62);
s[2] = ROL2(s[12], 43);
s[12] = ROL2(s[13], 25);
s[13] = ROL2(s[19], 8);
s[19] = ROL2(s[23], 56);
s[23] = ROL2(s[15], 41);
s[15] = ROL2(s[4], 27);
s[4] = ROL2(s[24], 14);
s[24] = ROL2(s[21], 2);
s[21] = ROL2(s[8], 55);
s[8] = ROL2(s[16], 45);
s[16] = ROL2(s[5], 36);
s[5] = ROL2(s[3], 28);
s[3] = ROL2(s[18], 21);
s[18] = ROL2(s[17], 15);
s[17] = ROL2(s[11], 10);
s[11] = ROL2(s[7], 6);
s[7] = ROL2(s[10], 3);
s[10] = ROL2(v, 1);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
v = s[0]; w = s[1]; s[0] ^= (~w) & s[2]; s[1] ^= (~s[2]) & s[3]; s[2] ^= (~s[3]) & s[4]; s[3] ^= (~s[4]) & v; s[4] ^= (~v) & w;
v = s[5]; w = s[6]; s[5] ^= (~w) & s[7]; s[6] ^= (~s[7]) & s[8]; s[7] ^= (~s[8]) & s[9]; s[8] ^= (~s[9]) & v; s[9] ^= (~v) & w;
v = s[10]; w = s[11]; s[10] ^= (~w) & s[12]; s[11] ^= (~s[12]) & s[13]; s[12] ^= (~s[13]) & s[14]; s[13] ^= (~s[14]) & v; s[14] ^= (~v) & w;
v = s[15]; w = s[16]; s[15] ^= (~w) & s[17]; s[16] ^= (~s[17]) & s[18]; s[17] ^= (~s[18]) & s[19]; s[18] ^= (~s[19]) & v; s[19] ^= (~v) & w;
v = s[20]; w = s[21]; s[20] ^= (~w) & s[22]; s[21] ^= (~s[22]) & s[23]; s[22] ^= (~s[23]) & s[24]; s[23] ^= (~s[24]) & v; s[24] ^= (~v) & w;
/* iota: a[0,0] ^= round constant */
s[0] ^= vectorize(keccak_round_constants[i]);
}
}
uint2 xor3x(const uint2 a,const uint2 b,const uint2 c) {
uint2 result;
#if __CUDA_ARCH__ >= 500 && CUDA_VERSION >= 7050
asm ("lop3.b32 %0, %1, %2, %3, 0x96;" : "=r"(result.x) : "r"(a.x), "r"(b.x),"r"(c.x)); //0x96 = 0xF0 ^ 0xCC ^ 0xAA
asm ("lop3.b32 %0, %1, %2, %3, 0x96;" : "=r"(result.y) : "r"(a.y), "r"(b.y),"r"(c.y)); //0x96 = 0xF0 ^ 0xCC ^ 0xAA
#else
__device__ __forceinline__
static void keccak_blockv30(uint64_t *s, const uint64_t *keccak_round_constants)
{
size_t i;
uint64_t t[5], u[5], v, w;
/* absorb input */
for (i = 0; i < 24; i++) {
/* theta: c = a[0,i] ^ a[1,i] ^ .. a[4,i] */
t[0] = s[0] ^ s[5] ^ s[10] ^ s[15] ^ s[20];
t[1] = s[1] ^ s[6] ^ s[11] ^ s[16] ^ s[21];
t[2] = s[2] ^ s[7] ^ s[12] ^ s[17] ^ s[22];
t[3] = s[3] ^ s[8] ^ s[13] ^ s[18] ^ s[23];
t[4] = s[4] ^ s[9] ^ s[14] ^ s[19] ^ s[24];
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
u[0] = t[4] ^ ROTL64(t[1], 1);
u[1] = t[0] ^ ROTL64(t[2], 1);
u[2] = t[1] ^ ROTL64(t[3], 1);
u[3] = t[2] ^ ROTL64(t[4], 1);
u[4] = t[3] ^ ROTL64(t[0], 1);
/* theta: a[0,i], a[1,i], .. a[4,i] ^= d[i] */
s[0] ^= u[0]; s[5] ^= u[0]; s[10] ^= u[0]; s[15] ^= u[0]; s[20] ^= u[0];
s[1] ^= u[1]; s[6] ^= u[1]; s[11] ^= u[1]; s[16] ^= u[1]; s[21] ^= u[1];
s[2] ^= u[2]; s[7] ^= u[2]; s[12] ^= u[2]; s[17] ^= u[2]; s[22] ^= u[2];
s[3] ^= u[3]; s[8] ^= u[3]; s[13] ^= u[3]; s[18] ^= u[3]; s[23] ^= u[3];
s[4] ^= u[4]; s[9] ^= u[4]; s[14] ^= u[4]; s[19] ^= u[4]; s[24] ^= u[4];
/* rho pi: b[..] = rotl(a[..], ..) */
v = s[ 1];
s[ 1] = ROTL64(s[ 6], 44);
s[ 6] = ROTL64(s[ 9], 20);
s[ 9] = ROTL64(s[22], 61);
s[22] = ROTL64(s[14], 39);
s[14] = ROTL64(s[20], 18);
s[20] = ROTL64(s[ 2], 62);
s[ 2] = ROTL64(s[12], 43);
s[12] = ROTL64(s[13], 25);
s[13] = ROTL64(s[19], 8);
s[19] = ROTL64(s[23], 56);
s[23] = ROTL64(s[15], 41);
s[15] = ROTL64(s[ 4], 27);
s[ 4] = ROTL64(s[24], 14);
s[24] = ROTL64(s[21], 2);
s[21] = ROTL64(s[ 8], 55);
s[ 8] = ROTL64(s[16], 45);
s[16] = ROTL64(s[ 5], 36);
s[ 5] = ROTL64(s[ 3], 28);
s[ 3] = ROTL64(s[18], 21);
s[18] = ROTL64(s[17], 15);
s[17] = ROTL64(s[11], 10);
s[11] = ROTL64(s[ 7], 6);
s[ 7] = ROTL64(s[10], 3);
s[10] = ROTL64( v, 1);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
v = s[ 0]; w = s[ 1]; s[ 0] ^= (~w) & s[ 2]; s[ 1] ^= (~s[ 2]) & s[ 3]; s[ 2] ^= (~s[ 3]) & s[ 4]; s[ 3] ^= (~s[ 4]) & v; s[ 4] ^= (~v) & w;
v = s[ 5]; w = s[ 6]; s[ 5] ^= (~w) & s[ 7]; s[ 6] ^= (~s[ 7]) & s[ 8]; s[ 7] ^= (~s[ 8]) & s[ 9]; s[ 8] ^= (~s[ 9]) & v; s[ 9] ^= (~v) & w;
v = s[10]; w = s[11]; s[10] ^= (~w) & s[12]; s[11] ^= (~s[12]) & s[13]; s[12] ^= (~s[13]) & s[14]; s[13] ^= (~s[14]) & v; s[14] ^= (~v) & w;
v = s[15]; w = s[16]; s[15] ^= (~w) & s[17]; s[16] ^= (~s[17]) & s[18]; s[17] ^= (~s[18]) & s[19]; s[18] ^= (~s[19]) & v; s[19] ^= (~v) & w;
v = s[20]; w = s[21]; s[20] ^= (~w) & s[22]; s[21] ^= (~s[22]) & s[23]; s[22] ^= (~s[23]) & s[24]; s[23] ^= (~s[24]) & v; s[24] ^= (~v) & w;
/* iota: a[0,0] ^= round constant */
s[0] ^= keccak_round_constants[i];
}
}
result = a^b^c;
#endif
__global__ __launch_bounds__(128,5)
void keccak256_gpu_hash_80(uint32_t threads, uint32_t startNounce, void *outputHash, uint32_t *resNounce)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t nounce = startNounce + thread;
#if __CUDA_ARCH__ >= 350
uint2 keccak_gpu_state[25];
#pragma unroll 25
for (int i=0; i<25; i++) {
if (i<9) keccak_gpu_state[i] = vectorize(c_PaddedMessage80[i]);
else keccak_gpu_state[i] = make_uint2(0, 0);
}
keccak_gpu_state[9]= vectorize(c_PaddedMessage80[9]);
keccak_gpu_state[9].y = cuda_swab32(nounce);
keccak_gpu_state[10] = make_uint2(1, 0);
keccak_gpu_state[16] = make_uint2(0, 0x80000000);
keccak_blockv35(keccak_gpu_state,keccak_round_constants);
if (devectorize(keccak_gpu_state[3]) <= ((uint64_t*)pTarget)[3]) {resNounce[0] = nounce;}
#else
uint64_t keccak_gpu_state[25];
#pragma unroll 25
for (int i=0; i<25; i++) {
if (i<9) keccak_gpu_state[i] = c_PaddedMessage80[i];
else keccak_gpu_state[i] = 0;
}
keccak_gpu_state[9] = REPLACE_HIDWORD(c_PaddedMessage80[9], cuda_swab32(nounce));
keccak_gpu_state[10] = 0x0000000000000001;
keccak_gpu_state[16] = 0x8000000000000000;
keccak_blockv30(keccak_gpu_state, keccak_round_constants);
if (keccak_gpu_state[3] <= ((uint64_t*)pTarget)[3]) { resNounce[0] = nounce; }
#endif
}
}
__host__
uint32_t keccak256_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_outputHash, int order)
{
uint32_t result = UINT32_MAX;
cudaMemset(d_KNonce[thr_id], 0xff, sizeof(uint32_t));
const uint32_t threadsperblock = 128;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
size_t shared_size = 0;
keccak256_gpu_hash_80<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash, d_KNonce[thr_id]);
MyStreamSynchronize(NULL, order, thr_id);
cudaMemcpy(d_nounce[thr_id], d_KNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
cudaThreadSynchronize();
result = *d_nounce[thr_id];
return result;
}
__global__ __launch_bounds__(256,3)
void keccak256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint64_t *outputHash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
#if __CUDA_ARCH__ >= 350 /* tpr: to double check if faster on SM5+ */
uint2 keccak_gpu_state[25];
#pragma unroll 25
for (int i = 0; i<25; i++) {
if (i<4) keccak_gpu_state[i] = vectorize(outputHash[i*threads+thread]);
else keccak_gpu_state[i] = make_uint2(0, 0);
}
keccak_gpu_state[4] = make_uint2(1, 0);
keccak_gpu_state[16] = make_uint2(0, 0x80000000);
keccak_blockv35(keccak_gpu_state, keccak_round_constants);
#pragma unroll 4
for (int i=0; i<4; i++)
outputHash[i*threads+thread] = devectorize(keccak_gpu_state[i]);
__device__ __forceinline__
uint2 chi(const uint2 a,const uint2 b,const uint2 c) { // keccak chi
uint2 result;
#if __CUDA_ARCH__ >= 500 && CUDA_VERSION >= 7050
asm ("lop3.b32 %0, %1, %2, %3, 0xD2;" : "=r"(result.x) : "r"(a.x), "r"(b.x),"r"(c.x)); //0xD2 = 0xF0 ^ ((~0xCC) & 0xAA)
asm ("lop3.b32 %0, %1, %2, %3, 0xD2;" : "=r"(result.y) : "r"(a.y), "r"(b.y),"r"(c.y)); //0xD2 = 0xF0 ^ ((~0xCC) & 0xAA)
#else
uint64_t keccak_gpu_state[25];
#pragma unroll 25
for (int i = 0; i<25; i++) {
if (i<4)
keccak_gpu_state[i] = outputHash[i*threads+thread];
else
keccak_gpu_state[i] = 0;
}
keccak_gpu_state[4] = 0x0000000000000001;
keccak_gpu_state[16] = 0x8000000000000000;
keccak_blockv30(keccak_gpu_state, keccak_round_constants);
#pragma unroll 4
for (int i = 0; i<4; i++)
outputHash[i*threads + thread] = keccak_gpu_state[i];
result = a ^ (~b) & c;
#endif
return result;
}
__device__ __forceinline__
uint64_t xor5(uint64_t a, uint64_t b, uint64_t c, uint64_t d, uint64_t e)
{
uint64_t result;
asm("xor.b64 %0, %1, %2;" : "=l"(result) : "l"(d) ,"l"(e));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(c));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(b));
asm("xor.b64 %0, %0, %1;" : "+l"(result) : "l"(a));
return result;
}
#if __CUDA_ARCH__ <= 500
__global__ __launch_bounds__(TPB50, 2)
#else
__global__ __launch_bounds__(TPB52, 1)
#endif
void keccak256_gpu_hash_80(uint32_t threads, uint32_t startNonce, uint32_t *resNounce, const uint2 highTarget)
{
uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
uint2 s[25], t[5], v, w, u[5];
#if __CUDA_ARCH__ > 500
uint64_t step = gridDim.x * blockDim.x;
uint64_t maxNonce = startNonce + threads;
for(uint64_t nounce = startNonce + thread; nounce<maxNonce;nounce+=step) {
#else
uint32_t nounce = startNonce+thread;
if(thread<threads) {
#endif
s[ 9] = make_uint2(c_message48[0].x,cuda_swab32(nounce));
s[10] = keccak_round_constants[0];
t[ 4] = c_message48[1]^s[ 9];
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
u[ 0] = t[4] ^ c_mid[ 0];
u[ 1] = c_mid[ 1] ^ ROL2(t[4],1);
u[ 2] = c_mid[ 2];
/* thetarho pi: b[..] = rotl(a[..] ^ d[...], ..)*/
s[ 7] = ROL2(s[10]^u[0], 3);
s[10] = c_mid[ 3];
w = c_mid[ 4];
s[20] = c_mid[ 5];
s[ 6] = ROL2(s[ 9]^u[2],20);
s[ 9] = c_mid[ 6];
s[22] = c_mid[ 7];
s[14] = ROL2(u[0],18);
s[ 2] = c_mid[ 8];
s[12] = ROL2(u[1],25);
s[13] = c_mid[ 9];
s[19] = ROR8(u[1]);
s[23] = ROR2(u[0],23);
s[15] = c_mid[10];
s[ 4] = c_mid[11];
s[24] = c_mid[12];
s[21] = ROR2(c_message48[2]^u[1], 9);
s[ 8] = c_mid[13];
s[16] = ROR2(c_message48[3]^u[0],28);
s[ 5] = ROL2(c_message48[4]^u[1],28);
s[ 3] = ROL2(u[1],21);
s[18] = c_mid[14];
s[17] = c_mid[15];
s[11] = c_mid[16];
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
v = c_message48[5]^u[0];
s[ 0] = chi(v,w,s[ 2]);
s[ 1] = chi(w,s[ 2],s[ 3]);
s[ 2] = chi(s[ 2],s[ 3],s[ 4]);
s[ 3] = chi(s[ 3],s[ 4],v);
s[ 4] = chi(s[ 4],v,w);
v = s[ 5]; w = s[ 6]; s[ 5] = chi(v,w,s[ 7]); s[ 6] = chi(w,s[ 7],s[ 8]); s[ 7] = chi(s[ 7],s[ 8],s[ 9]); s[ 8] = chi(s[ 8],s[ 9],v);s[ 9] = chi(s[ 9],v,w);
v = s[10]; w = s[11]; s[10] = chi(v,w,s[12]); s[11] = chi(w,s[12],s[13]); s[12] = chi(s[12],s[13],s[14]); s[13] = chi(s[13],s[14],v);s[14] = chi(s[14],v,w);
v = s[15]; w = s[16]; s[15] = chi(v,w,s[17]); s[16] = chi(w,s[17],s[18]); s[17] = chi(s[17],s[18],s[19]); s[18] = chi(s[18],s[19],v);s[19] = chi(s[19],v,w);
v = s[20]; w = s[21]; s[20] = chi(v,w,s[22]); s[21] = chi(w,s[22],s[23]); s[22] = chi(s[22],s[23],s[24]); s[23] = chi(s[23],s[24],v);s[24] = chi(s[24],v,w);
/* iota: a[0,0] ^= round constant */
s[ 0] ^=keccak_round_constants[ 0];
#if __CUDA_ARCH__ > 500
#pragma unroll 22
#else
#pragma unroll 4
#endif
for (int i = 1; i < 23; i++) {
#pragma unroll
for(int j=0;j<5;j++) {
t[ j] = vectorize(xor5(devectorize(s[ j]),devectorize(s[j+5]),devectorize(s[j+10]),devectorize(s[j+15]),devectorize(s[j+20])));
}
/*theta*/
#pragma unroll
for(int j=0;j<5;j++) {
u[j] = ROL2(t[j], 1);
}
s[ 4] = xor3x(s[ 4], t[3], u[0]);s[ 9] = xor3x(s[ 9], t[3], u[0]);s[14] = xor3x(s[14], t[3], u[0]);s[19] = xor3x(s[19], t[3], u[0]);s[24] = xor3x(s[24], t[3], u[0]);
s[ 0] = xor3x(s[ 0], t[4], u[1]);s[ 5] = xor3x(s[ 5], t[4], u[1]);s[10] = xor3x(s[10], t[4], u[1]);s[15] = xor3x(s[15], t[4], u[1]);s[20] = xor3x(s[20], t[4], u[1]);
s[ 1] = xor3x(s[ 1], t[0], u[2]);s[ 6] = xor3x(s[ 6], t[0], u[2]);s[11] = xor3x(s[11], t[0], u[2]);s[16] = xor3x(s[16], t[0], u[2]);s[21] = xor3x(s[21], t[0], u[2]);
s[ 2] = xor3x(s[ 2], t[1], u[3]);s[ 7] = xor3x(s[ 7], t[1], u[3]);s[12] = xor3x(s[12], t[1], u[3]);s[17] = xor3x(s[17], t[1], u[3]);s[22] = xor3x(s[22], t[1], u[3]);
s[ 3] = xor3x(s[ 3], t[2], u[4]);s[ 8] = xor3x(s[ 8], t[2], u[4]);s[13] = xor3x(s[13], t[2], u[4]);s[18] = xor3x(s[18], t[2], u[4]);s[23] = xor3x(s[23], t[2], u[4]);
/*rho pi: b[..] = rotl(a[..] ^ d[...], ..)*/
v = s[ 1];
s[ 1] = ROL2(s[ 6],44); s[ 6] = ROL2(s[ 9],20); s[ 9] = ROL2(s[22],61); s[22] = ROL2(s[14],39);
s[14] = ROL2(s[20],18); s[20] = ROL2(s[ 2],62); s[ 2] = ROL2(s[12],43); s[12] = ROL2(s[13],25);
s[13] = ROL8(s[19]); s[19] = ROR8(s[23]); s[23] = ROL2(s[15],41); s[15] = ROL2(s[ 4],27);
s[ 4] = ROL2(s[24],14); s[24] = ROL2(s[21], 2); s[21] = ROL2(s[ 8],55); s[ 8] = ROL2(s[16],45);
s[16] = ROL2(s[ 5],36); s[ 5] = ROL2(s[ 3],28); s[ 3] = ROL2(s[18],21); s[18] = ROL2(s[17],15);
s[17] = ROL2(s[11],10); s[11] = ROL2(s[ 7], 6); s[ 7] = ROL2(s[10], 3); s[10] = ROL2(v, 1);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
#pragma unroll
for(int j=0;j<25;j+=5) {
v=s[j];w=s[j + 1];s[j] = chi(s[j],s[j+1],s[j+2]);s[j+1] = chi(s[j+1],s[j+2],s[j+3]);s[j+2]=chi(s[j+2],s[j+3],s[j+4]);s[j+3]=chi(s[j+3],s[j+4],v);s[j+4]=chi(s[j+4],v,w);
}
/* iota: a[0,0] ^= round constant */
s[ 0] ^=keccak_round_constants[ i];
}
/* theta: c = a[0,i] ^ a[1,i] ^ .. a[4,i] */
#pragma unroll 5
for(int j=0;j<5;j++) {
t[ j] = xor3x(xor3x(s[j+0],s[j+5],s[j+10]), s[j+15], s[j+20]);
}
s[24] = xor3x(s[24],t[3],ROL2(t[0],1));
s[18] = xor3x(s[18],t[2],ROL2(t[4],1));
s[ 0] = xor3x(s[ 0],t[4],ROL2(t[1],1));
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
s[24] = ROL2(s[24],14);
s[18] = ROL2(s[18],21);
if (devectorize(chi(s[18],s[24],s[ 0])) <= devectorize(highTarget)) {
// if(chi(s[18].x,s[24].x,s[0].x)<=highTarget.x) {
// if(chi(s[18].y,s[24].y,s[0].y)<=highTarget.y) {
const uint32_t tmp = atomicExch(&resNounce[0], nounce);
if (tmp != UINT32_MAX)
resNounce[1] = tmp;
// return;
// }
}
}
}
__host__
void keccak256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
void keccak256_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t* resNonces, const uint2 highTarget)
{
const uint32_t threadsperblock = 256;
uint32_t tpb;
dim3 grid;
if (device_sm[device_map[thr_id]] <= 500) {
tpb = TPB50;
grid.x = (threads + tpb-1)/tpb;
} else {
tpb = TPB52;
grid.x = (threads + (NPT*tpb)-1)/(NPT*tpb);
}
const dim3 block(tpb);
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
keccak256_gpu_hash_80<<<grid, block>>>(threads, startNonce, d_nonces[thr_id], highTarget);
// cudaThreadSynchronize();
cudaMemcpy(h_nonces[thr_id], d_nonces[thr_id], NBN*sizeof(uint32_t), cudaMemcpyDeviceToHost);
memcpy(resNonces, h_nonces[thr_id], NBN*sizeof(uint32_t));
}
keccak256_gpu_hash_32 <<<grid, block>>> (threads, startNounce, d_outputHash);
MyStreamSynchronize(NULL, order, thr_id);
#if 0
#if __CUDA_ARCH__ <= 500
__global__ __launch_bounds__(TPB50, 2)
#else
__global__ __launch_bounds__(TPB52, 1)
#endif
void keccak256_gpu_hash_32(uint32_t threads, uint2* outputHash)
{
uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
uint2 s[25], t[5], v, w, u[5];
if(thread < threads) {
#pragma unroll 25
for (int i = 0; i<25; i++) {
if (i<4) s[i] = __ldg(&outputHash[i*threads+thread]);
else s[i] = make_uint2(0, 0);
}
s[4] = keccak_round_constants[ 0];
s[16] = make_uint2(0, 0x80000000);
#if __CUDA_ARCH__ > 500
#pragma unroll
#else
#pragma unroll 4
#endif
for (uint32_t i = 0; i < 23; i++) {
/*theta*/
#pragma unroll 5
for(int j=0; j<5; j++) {
t[ j] = vectorize(xor5(devectorize(s[ j]),devectorize(s[j+5]),devectorize(s[j+10]),devectorize(s[j+15]),devectorize(s[j+20])));
}
/*theta*/
#pragma unroll 5
for(int j=0; j<5; j++) {
u[j] = ROL2(t[j], 1);
}
s[ 4] = xor3x(s[ 4], t[3], u[0]);s[ 9] = xor3x(s[ 9], t[3], u[0]);s[14] = xor3x(s[14], t[3], u[0]);s[19] = xor3x(s[19], t[3], u[0]);s[24] = xor3x(s[24], t[3], u[0]);
s[ 0] = xor3x(s[ 0], t[4], u[1]);s[ 5] = xor3x(s[ 5], t[4], u[1]);s[10] = xor3x(s[10], t[4], u[1]);s[15] = xor3x(s[15], t[4], u[1]);s[20] = xor3x(s[20], t[4], u[1]);
s[ 1] = xor3x(s[ 1], t[0], u[2]);s[ 6] = xor3x(s[ 6], t[0], u[2]);s[11] = xor3x(s[11], t[0], u[2]);s[16] = xor3x(s[16], t[0], u[2]);s[21] = xor3x(s[21], t[0], u[2]);
s[ 2] = xor3x(s[ 2], t[1], u[3]);s[ 7] = xor3x(s[ 7], t[1], u[3]);s[12] = xor3x(s[12], t[1], u[3]);s[17] = xor3x(s[17], t[1], u[3]);s[22] = xor3x(s[22], t[1], u[3]);
s[ 3] = xor3x(s[ 3], t[2], u[4]);s[ 8] = xor3x(s[ 8], t[2], u[4]);s[13] = xor3x(s[13], t[2], u[4]);s[18] = xor3x(s[18], t[2], u[4]);s[23] = xor3x(s[23], t[2], u[4]);
/*rho pi: b[..] = rotl(a[..] ^ d[...], ..)*/
v = s[ 1];
s[ 1] = ROL2(s[ 6],44); s[ 6] = ROL2(s[ 9],20); s[ 9] = ROL2(s[22],61); s[22] = ROL2(s[14],39);
s[14] = ROL2(s[20],18); s[20] = ROL2(s[ 2],62); s[ 2] = ROL2(s[12],43); s[12] = ROL2(s[13],25);
s[13] = ROL8(s[19]); s[19] = ROR8(s[23]); s[23] = ROL2(s[15],41); s[15] = ROL2(s[ 4],27);
s[ 4] = ROL2(s[24],14); s[24] = ROL2(s[21], 2); s[21] = ROL2(s[ 8],55); s[ 8] = ROL2(s[16],45);
s[16] = ROL2(s[ 5],36); s[ 5] = ROL2(s[ 3],28); s[ 3] = ROL2(s[18],21); s[18] = ROL2(s[17],15);
s[17] = ROL2(s[11],10); s[11] = ROL2(s[ 7], 6); s[ 7] = ROL2(s[10], 3); s[10] = ROL2(v, 1);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
#pragma unroll 5
for(int j=0; j<25; j+=5) {
v=s[j];w=s[j + 1]; s[j] = chi(v,w,s[j+2]); s[j+1] = chi(w,s[j+2],s[j+3]); s[j+2]=chi(s[j+2],s[j+3],s[j+4]); s[j+3]=chi(s[j+3],s[j+4],v); s[j+4]=chi(s[j+4],v,w);
}
/* iota: a[0,0] ^= round constant */
s[ 0] ^=keccak_round_constants[ i];
}
/* theta: c = a[0,i] ^ a[1,i] ^ .. a[4,i] */
#pragma unroll 5
for(int j=0;j<5;j++) {
t[ j] = xor3x(xor3x(s[j+0],s[j+5],s[j+10]), s[j+15], s[j+20]);
}
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
#pragma unroll 5
for(int j=0;j<5;j++) {
u[j] = ROL2(t[j],1);
}
/* thetarho pi: b[..] = rotl(a[..] ^ d[...], ..) //There's no need to perform theta and -store- the result since it's unique for each a[..]*/
s[ 4] = xor3x(s[24], t[3], u[0]);
s[ 0] = xor3x(s[ 0], t[4], u[1]);
s[ 1] = xor3x(s[ 6], t[0], u[2]);
s[ 2] = xor3x(s[12], t[1], u[3]);
s[ 3] = xor3x(s[18], t[2], u[4]);
s[ 1] = ROR2(s[ 1],20);
s[ 2] = ROR2(s[ 2],21);
s[ 3] = ROL2(s[ 3],21);
s[ 4] = ROL2(s[ 4],14);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
outputHash[0*threads+thread] = chi(s[ 0],s[ 1],s[ 2]) ^ keccak_round_constants[23];
outputHash[1*threads+thread] = chi(s[ 1],s[ 2],s[ 3]);
outputHash[2*threads+thread] = chi(s[ 2],s[ 3],s[ 4]);
outputHash[3*threads+thread] = chi(s[ 3],s[ 4],s[ 0]);
}
}
__host__
void keccak256_setBlock_80(void *pdata,const void *pTargetIn)
void keccak256_cpu_hash_32(const int thr_id,const uint32_t threads, uint2* d_hash)
{
unsigned char PaddedMessage[80];
memcpy(PaddedMessage, pdata, 80);
CUDA_SAFE_CALL(cudaMemcpyToSymbol(pTarget, pTargetIn, 8*sizeof(uint32_t), 0, cudaMemcpyHostToDevice));
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_PaddedMessage80, PaddedMessage, 10*sizeof(uint64_t), 0, cudaMemcpyHostToDevice));
uint32_t tpb = TPB52;
if (device_sm[device_map[thr_id]] == 500) tpb = TPB50;
const dim3 grid((threads + tpb-1)/tpb);
const dim3 block(tpb);
keccak256_gpu_hash_32 <<<grid, block>>> (threads, d_hash);
}
#endif
__host__
void keccak256_setBlock_80(uint64_t *endiandata)
{
uint64_t midstate[17], s[25];
uint64_t t[5], u[5];
s[10] = 1; //(uint64_t)make_uint2(1, 0);
s[16] = ((uint64_t)1)<<63; //(uint64_t)make_uint2(0, 0x80000000);
t[0] = endiandata[0] ^ endiandata[5] ^ s[10];
t[1] = endiandata[1] ^ endiandata[6] ^ s[16];
t[2] = endiandata[2] ^ endiandata[7];
t[3] = endiandata[3] ^ endiandata[8];
midstate[ 0] = ROTL64(t[1], 1); //u[0] -partial
u[1] = t[ 0] ^ ROTL64(t[2], 1); //u[1]
u[2] = t[ 1] ^ ROTL64(t[3], 1); //u[2]
midstate[ 1] = t[ 2]; //u[3] -partial
midstate[ 2] = t[ 3] ^ ROTL64(t[0], 1); //u[4]
midstate[ 3] = ROTL64(endiandata[1]^u[1], 1); //v
midstate[ 4] = ROTL64(endiandata[6]^u[1], 44);
midstate[ 5] = ROTL64(endiandata[2]^u[2], 62);
midstate[ 6] = ROTL64(u[2], 61);
midstate[ 7] = ROTL64(midstate[2], 39);
midstate[ 8] = ROTL64(u[2], 43);
midstate[ 9] = ROTL64(midstate[2], 8);
midstate[10] = ROTL64(endiandata[4]^midstate[ 2],27);
midstate[11] = ROTL64(midstate[2], 14);
midstate[12] = ROTL64(u[1], 2);
midstate[13] = ROTL64(s[16] ^ u[1], 45);
midstate[14] = ROTL64(u[2],15);
midstate[15] = ROTL64(u[1],10);
midstate[16] = ROTL64(endiandata[7]^u[2], 6);
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_mid, midstate,17*sizeof(uint64_t), 0, cudaMemcpyHostToDevice));
// pass only what's needed
uint64_t message48[6];
message48[0] = endiandata[9];
message48[1] = endiandata[4];
message48[2] = endiandata[8];
message48[3] = endiandata[5];
message48[4] = endiandata[3];
message48[5] = endiandata[0];
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_message48, message48, 6*sizeof(uint64_t), 0, cudaMemcpyHostToDevice));
}
__host__
void keccak256_cpu_init(int thr_id, uint32_t threads)
void keccak256_cpu_init(int thr_id)
{
CUDA_SAFE_CALL(cudaMemcpyToSymbol(keccak_round_constants, host_keccak_round_constants,
sizeof(host_keccak_round_constants), 0, cudaMemcpyHostToDevice));
CUDA_SAFE_CALL(cudaMalloc(&d_KNonce[thr_id], sizeof(uint32_t)));
CUDA_SAFE_CALL(cudaMallocHost(&d_nounce[thr_id], 1*sizeof(uint32_t)));
CUDA_SAFE_CALL(cudaMalloc(&d_nonces[thr_id], NBN*sizeof(uint32_t)));
//CUDA_SAFE_CALL(cudaMallocHost(&h_nonces[thr_id], NBN*sizeof(uint32_t)));
h_nonces[thr_id] = (uint32_t*) malloc(NBN * sizeof(uint32_t));
if(h_nonces[thr_id] == NULL) {
gpulog(LOG_ERR,thr_id,"Host memory allocation failed");
exit(EXIT_FAILURE);
}
}
__host__
void keccak256_setOutput(int thr_id)
{
CUDA_SAFE_CALL(cudaMemset(d_nonces[thr_id], 0xff, NBN*sizeof(uint32_t)));
}
__host__
void keccak256_cpu_free(int thr_id)
{
cudaFree(d_KNonce[thr_id]);
cudaFreeHost(d_nounce[thr_id]);
cudaFree(d_nonces[thr_id]);
//cudaFreeHost(h_nonces[thr_id]);
free(h_nonces[thr_id]);
}
+303
View File
@@ -0,0 +1,303 @@
#include "miner.h"
extern "C" {
#include <stdint.h>
#include <memory.h>
}
#include "cuda_helper.h"
static const uint64_t host_keccak_round_constants[24] = {
0x0000000000000001ull, 0x0000000000008082ull,
0x800000000000808aull, 0x8000000080008000ull,
0x000000000000808bull, 0x0000000080000001ull,
0x8000000080008081ull, 0x8000000000008009ull,
0x000000000000008aull, 0x0000000000000088ull,
0x0000000080008009ull, 0x000000008000000aull,
0x000000008000808bull, 0x800000000000008bull,
0x8000000000008089ull, 0x8000000000008003ull,
0x8000000000008002ull, 0x8000000000000080ull,
0x000000000000800aull, 0x800000008000000aull,
0x8000000080008081ull, 0x8000000000008080ull,
0x0000000080000001ull, 0x8000000080008008ull
};
static uint32_t *d_KNonce[MAX_GPUS];
__constant__ uint32_t pTarget[8];
__constant__ uint64_t keccak_round_constants[24];
__constant__ uint64_t c_PaddedMessage80[10]; // padded message (80 bytes + padding?)
#if __CUDA_ARCH__ >= 350
__device__ __forceinline__
static void keccak_blockv35(uint2 *s, const uint64_t *keccak_round_constants)
{
size_t i;
uint2 t[5], u[5], v, w;
#pragma unroll
for (i = 0; i < 24; i++) {
/* theta: c = a[0,i] ^ a[1,i] ^ .. a[4,i] */
t[0] = s[0] ^ s[5] ^ s[10] ^ s[15] ^ s[20];
t[1] = s[1] ^ s[6] ^ s[11] ^ s[16] ^ s[21];
t[2] = s[2] ^ s[7] ^ s[12] ^ s[17] ^ s[22];
t[3] = s[3] ^ s[8] ^ s[13] ^ s[18] ^ s[23];
t[4] = s[4] ^ s[9] ^ s[14] ^ s[19] ^ s[24];
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
u[0] = t[4] ^ ROL2(t[1], 1);
u[1] = t[0] ^ ROL2(t[2], 1);
u[2] = t[1] ^ ROL2(t[3], 1);
u[3] = t[2] ^ ROL2(t[4], 1);
u[4] = t[3] ^ ROL2(t[0], 1);
/* theta: a[0,i], a[1,i], .. a[4,i] ^= d[i] */
s[0] ^= u[0]; s[5] ^= u[0]; s[10] ^= u[0]; s[15] ^= u[0]; s[20] ^= u[0];
s[1] ^= u[1]; s[6] ^= u[1]; s[11] ^= u[1]; s[16] ^= u[1]; s[21] ^= u[1];
s[2] ^= u[2]; s[7] ^= u[2]; s[12] ^= u[2]; s[17] ^= u[2]; s[22] ^= u[2];
s[3] ^= u[3]; s[8] ^= u[3]; s[13] ^= u[3]; s[18] ^= u[3]; s[23] ^= u[3];
s[4] ^= u[4]; s[9] ^= u[4]; s[14] ^= u[4]; s[19] ^= u[4]; s[24] ^= u[4];
/* rho pi: b[..] = rotl(a[..], ..) */
v = s[1];
s[1] = ROL2(s[6], 44);
s[6] = ROL2(s[9], 20);
s[9] = ROL2(s[22], 61);
s[22] = ROL2(s[14], 39);
s[14] = ROL2(s[20], 18);
s[20] = ROL2(s[2], 62);
s[2] = ROL2(s[12], 43);
s[12] = ROL2(s[13], 25);
s[13] = ROL2(s[19], 8);
s[19] = ROL2(s[23], 56);
s[23] = ROL2(s[15], 41);
s[15] = ROL2(s[4], 27);
s[4] = ROL2(s[24], 14);
s[24] = ROL2(s[21], 2);
s[21] = ROL2(s[8], 55);
s[8] = ROL2(s[16], 45);
s[16] = ROL2(s[5], 36);
s[5] = ROL2(s[3], 28);
s[3] = ROL2(s[18], 21);
s[18] = ROL2(s[17], 15);
s[17] = ROL2(s[11], 10);
s[11] = ROL2(s[7], 6);
s[7] = ROL2(s[10], 3);
s[10] = ROL2(v, 1);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
v = s[0]; w = s[1]; s[0] ^= (~w) & s[2]; s[1] ^= (~s[2]) & s[3]; s[2] ^= (~s[3]) & s[4]; s[3] ^= (~s[4]) & v; s[4] ^= (~v) & w;
v = s[5]; w = s[6]; s[5] ^= (~w) & s[7]; s[6] ^= (~s[7]) & s[8]; s[7] ^= (~s[8]) & s[9]; s[8] ^= (~s[9]) & v; s[9] ^= (~v) & w;
v = s[10]; w = s[11]; s[10] ^= (~w) & s[12]; s[11] ^= (~s[12]) & s[13]; s[12] ^= (~s[13]) & s[14]; s[13] ^= (~s[14]) & v; s[14] ^= (~v) & w;
v = s[15]; w = s[16]; s[15] ^= (~w) & s[17]; s[16] ^= (~s[17]) & s[18]; s[17] ^= (~s[18]) & s[19]; s[18] ^= (~s[19]) & v; s[19] ^= (~v) & w;
v = s[20]; w = s[21]; s[20] ^= (~w) & s[22]; s[21] ^= (~s[22]) & s[23]; s[22] ^= (~s[23]) & s[24]; s[23] ^= (~s[24]) & v; s[24] ^= (~v) & w;
/* iota: a[0,0] ^= round constant */
s[0] ^= vectorize(keccak_round_constants[i]);
}
}
#else
__device__ __forceinline__
static void keccak_blockv30(uint64_t *s, const uint64_t *keccak_round_constants)
{
size_t i;
uint64_t t[5], u[5], v, w;
/* absorb input */
for (i = 0; i < 24; i++) {
/* theta: c = a[0,i] ^ a[1,i] ^ .. a[4,i] */
t[0] = s[0] ^ s[5] ^ s[10] ^ s[15] ^ s[20];
t[1] = s[1] ^ s[6] ^ s[11] ^ s[16] ^ s[21];
t[2] = s[2] ^ s[7] ^ s[12] ^ s[17] ^ s[22];
t[3] = s[3] ^ s[8] ^ s[13] ^ s[18] ^ s[23];
t[4] = s[4] ^ s[9] ^ s[14] ^ s[19] ^ s[24];
/* theta: d[i] = c[i+4] ^ rotl(c[i+1],1) */
u[0] = t[4] ^ ROTL64(t[1], 1);
u[1] = t[0] ^ ROTL64(t[2], 1);
u[2] = t[1] ^ ROTL64(t[3], 1);
u[3] = t[2] ^ ROTL64(t[4], 1);
u[4] = t[3] ^ ROTL64(t[0], 1);
/* theta: a[0,i], a[1,i], .. a[4,i] ^= d[i] */
s[0] ^= u[0]; s[5] ^= u[0]; s[10] ^= u[0]; s[15] ^= u[0]; s[20] ^= u[0];
s[1] ^= u[1]; s[6] ^= u[1]; s[11] ^= u[1]; s[16] ^= u[1]; s[21] ^= u[1];
s[2] ^= u[2]; s[7] ^= u[2]; s[12] ^= u[2]; s[17] ^= u[2]; s[22] ^= u[2];
s[3] ^= u[3]; s[8] ^= u[3]; s[13] ^= u[3]; s[18] ^= u[3]; s[23] ^= u[3];
s[4] ^= u[4]; s[9] ^= u[4]; s[14] ^= u[4]; s[19] ^= u[4]; s[24] ^= u[4];
/* rho pi: b[..] = rotl(a[..], ..) */
v = s[ 1];
s[ 1] = ROTL64(s[ 6], 44);
s[ 6] = ROTL64(s[ 9], 20);
s[ 9] = ROTL64(s[22], 61);
s[22] = ROTL64(s[14], 39);
s[14] = ROTL64(s[20], 18);
s[20] = ROTL64(s[ 2], 62);
s[ 2] = ROTL64(s[12], 43);
s[12] = ROTL64(s[13], 25);
s[13] = ROTL64(s[19], 8);
s[19] = ROTL64(s[23], 56);
s[23] = ROTL64(s[15], 41);
s[15] = ROTL64(s[ 4], 27);
s[ 4] = ROTL64(s[24], 14);
s[24] = ROTL64(s[21], 2);
s[21] = ROTL64(s[ 8], 55);
s[ 8] = ROTL64(s[16], 45);
s[16] = ROTL64(s[ 5], 36);
s[ 5] = ROTL64(s[ 3], 28);
s[ 3] = ROTL64(s[18], 21);
s[18] = ROTL64(s[17], 15);
s[17] = ROTL64(s[11], 10);
s[11] = ROTL64(s[ 7], 6);
s[ 7] = ROTL64(s[10], 3);
s[10] = ROTL64( v, 1);
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
v = s[ 0]; w = s[ 1]; s[ 0] ^= (~w) & s[ 2]; s[ 1] ^= (~s[ 2]) & s[ 3]; s[ 2] ^= (~s[ 3]) & s[ 4]; s[ 3] ^= (~s[ 4]) & v; s[ 4] ^= (~v) & w;
v = s[ 5]; w = s[ 6]; s[ 5] ^= (~w) & s[ 7]; s[ 6] ^= (~s[ 7]) & s[ 8]; s[ 7] ^= (~s[ 8]) & s[ 9]; s[ 8] ^= (~s[ 9]) & v; s[ 9] ^= (~v) & w;
v = s[10]; w = s[11]; s[10] ^= (~w) & s[12]; s[11] ^= (~s[12]) & s[13]; s[12] ^= (~s[13]) & s[14]; s[13] ^= (~s[14]) & v; s[14] ^= (~v) & w;
v = s[15]; w = s[16]; s[15] ^= (~w) & s[17]; s[16] ^= (~s[17]) & s[18]; s[17] ^= (~s[18]) & s[19]; s[18] ^= (~s[19]) & v; s[19] ^= (~v) & w;
v = s[20]; w = s[21]; s[20] ^= (~w) & s[22]; s[21] ^= (~s[22]) & s[23]; s[22] ^= (~s[23]) & s[24]; s[23] ^= (~s[24]) & v; s[24] ^= (~v) & w;
/* iota: a[0,0] ^= round constant */
s[0] ^= keccak_round_constants[i];
}
}
#endif
__global__ __launch_bounds__(128,5)
void keccak256_sm3_gpu_hash_80(uint32_t threads, uint32_t startNounce, uint32_t *resNounce)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t nounce = startNounce + thread;
#if __CUDA_ARCH__ >= 350
uint2 keccak_gpu_state[25];
#pragma unroll 25
for (int i=0; i<25; i++) {
if (i<9) keccak_gpu_state[i] = vectorize(c_PaddedMessage80[i]);
else keccak_gpu_state[i] = make_uint2(0, 0);
}
keccak_gpu_state[9]= vectorize(c_PaddedMessage80[9]);
keccak_gpu_state[9].y = cuda_swab32(nounce);
keccak_gpu_state[10] = make_uint2(1, 0);
keccak_gpu_state[16] = make_uint2(0, 0x80000000);
keccak_blockv35(keccak_gpu_state,keccak_round_constants);
if (devectorize(keccak_gpu_state[3]) <= ((uint64_t*)pTarget)[3]) {resNounce[0] = nounce;}
#else
uint64_t keccak_gpu_state[25];
#pragma unroll 25
for (int i=0; i<25; i++) {
if (i<9) keccak_gpu_state[i] = c_PaddedMessage80[i];
else keccak_gpu_state[i] = 0;
}
keccak_gpu_state[9] = REPLACE_HIDWORD(c_PaddedMessage80[9], cuda_swab32(nounce));
keccak_gpu_state[10] = 0x0000000000000001;
keccak_gpu_state[16] = 0x8000000000000000;
keccak_blockv30(keccak_gpu_state, keccak_round_constants);
if (keccak_gpu_state[3] <= ((uint64_t*)pTarget)[3]) { resNounce[0] = nounce; }
#endif
}
}
__host__
void keccak256_sm3_hash_80(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *resNonces, int order)
{
cudaMemset(d_KNonce[thr_id], 0xff, 2*sizeof(uint32_t));
const uint32_t threadsperblock = 128;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
size_t shared_size = 0;
keccak256_sm3_gpu_hash_80<<<grid, block, shared_size>>>(threads, startNounce, d_KNonce[thr_id]);
cudaMemcpy(resNonces, d_KNonce[thr_id], 2*sizeof(uint32_t), cudaMemcpyDeviceToHost);
cudaThreadSynchronize();
}
#if 0
__global__ __launch_bounds__(256,3)
void keccak256_sm3_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint64_t *outputHash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
#if __CUDA_ARCH__ >= 350 /* tpr: to double check if faster on SM5+ */
uint2 keccak_gpu_state[25];
#pragma unroll 25
for (int i = 0; i<25; i++) {
if (i<4) keccak_gpu_state[i] = vectorize(outputHash[i*threads+thread]);
else keccak_gpu_state[i] = make_uint2(0, 0);
}
keccak_gpu_state[4] = make_uint2(1, 0);
keccak_gpu_state[16] = make_uint2(0, 0x80000000);
keccak_blockv35(keccak_gpu_state, keccak_round_constants);
#pragma unroll 4
for (int i=0; i<4; i++)
outputHash[i*threads+thread] = devectorize(keccak_gpu_state[i]);
#else
uint64_t keccak_gpu_state[25];
#pragma unroll 25
for (int i = 0; i<25; i++) {
if (i<4)
keccak_gpu_state[i] = outputHash[i*threads+thread];
else
keccak_gpu_state[i] = 0;
}
keccak_gpu_state[4] = 0x0000000000000001;
keccak_gpu_state[16] = 0x8000000000000000;
keccak_blockv30(keccak_gpu_state, keccak_round_constants);
#pragma unroll 4
for (int i = 0; i<4; i++)
outputHash[i*threads + thread] = keccak_gpu_state[i];
#endif
}
}
__host__
void keccak256_sm3_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
keccak256_sm3_gpu_hash_32 <<<grid, block>>> (threads, startNounce, d_outputHash);
MyStreamSynchronize(NULL, order, thr_id);
}
#endif
__host__
void keccak256_sm3_setBlock_80(void *pdata,const void *pTargetIn)
{
unsigned char PaddedMessage[80];
memcpy(PaddedMessage, pdata, 80);
CUDA_SAFE_CALL(cudaMemcpyToSymbol(pTarget, pTargetIn, 8*sizeof(uint32_t), 0, cudaMemcpyHostToDevice));
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_PaddedMessage80, PaddedMessage, 10*sizeof(uint64_t), 0, cudaMemcpyHostToDevice));
}
__host__
void keccak256_sm3_init(int thr_id, uint32_t threads)
{
CUDA_SAFE_CALL(cudaMemcpyToSymbol(keccak_round_constants, host_keccak_round_constants,
sizeof(host_keccak_round_constants), 0, cudaMemcpyHostToDevice));
CUDA_SAFE_CALL(cudaMalloc(&d_KNonce[thr_id], 2*sizeof(uint32_t)));
}
__host__
void keccak256_sm3_free(int thr_id)
{
cudaFree(d_KNonce[thr_id]);
}
+335 -329
View File
@@ -1,14 +1,11 @@
/**
* Blake-256 Decred 180-Bytes input Cuda Kernel (Tested on SM 5/5.2)
* Blake-256 Decred 180-Bytes input Cuda Kernel
*
* Tanguy Pruvot - Feb 2016
*
* Revised for optimisation by pallas @ bitcointalk - Apr 2016
* Tanguy Pruvot, Alexis Provos - Feb/Sep 2016
*/
#include <stdint.h>
#include <memory.h>
#include <miner.h>
extern "C" {
@@ -16,9 +13,10 @@ extern "C" {
}
/* threads per block */
#define TPB 512
/* nonces per round */
#define NPR 128
#define TPB 640
/* max count of found nonces in one call (like sgminer) */
#define MAX_RESULTS 4
/* hash by cpu with blake 256 */
extern "C" void decred_hash(void *output, const void *input)
@@ -36,255 +34,142 @@ extern "C" void decred_hash(void *output, const void *input)
#ifdef __INTELLISENSE__
#define __byte_perm(x, y, b) x
#define atomicInc(p, max) (*p)++
#endif
__constant__ uint32_t _ALIGN(4) d_data[24];
__constant__ uint32_t _ALIGN(16) c_h[2];
__constant__ uint32_t _ALIGN(16) c_data[32];
__constant__ uint32_t _ALIGN(16) c_xors[215];
/* 16 adapters max */
/* Buffers of candidate nonce(s) */
static uint32_t *d_resNonce[MAX_GPUS];
static uint32_t *h_resNonce[MAX_GPUS];
/* max count of found nonces in one call */
#define NBN 2
#if NBN > 1
static uint32_t extra_results[NBN] = { UINT32_MAX };
#endif
#define ROR8(a) __byte_perm(a, 0, 0x0321)
#define ROL16(a) __byte_perm(a, 0, 0x1032)
/* ############################################################################################################################### */
#define GSPREC(a,b,c,d,x,y) { \
v[a] += (m[x] ^ c_u256[y]) + v[b]; \
v[d] = __byte_perm(v[d] ^ v[a], 0, 0x1032); \
v[c] += v[d]; \
v[b] = SPH_ROTR32(v[b] ^ v[c], 12); \
v[a] += (m[y] ^ c_u256[x]) + v[b]; \
v[d] = __byte_perm(v[d] ^ v[a], 0, 0x0321); \
v[c] += v[d]; \
v[b] = SPH_ROTR32(v[b] ^ v[c], 7); \
/* macro bodies */
#define pxorGS(a,b,c,d) { \
v[a]+= c_xors[i++] + v[b]; \
v[d] = ROL16(v[d] ^ v[a]); \
v[c]+= v[d]; \
v[b] = ROTR32(v[b] ^ v[c], 12); \
v[a]+= c_xors[i++] + v[b]; \
v[d] = ROR8(v[d] ^ v[a]); \
v[c]+= v[d]; \
v[b] = ROTR32(v[b] ^ v[c], 7); \
}
#define GSPREC4(a0,b0,c0,d0,x0,y0,a1,b1,c1,d1,x1,y1,a2,b2,c2,d2,x2,y2,a3,b3,c3,d3,x3,y3) { \
v[a0] += (m[x0] ^ c_u256[y0]) + v[b0]; \
v[a1] += (m[x1] ^ c_u256[y1]) + v[b1]; \
v[a2] += (m[x2] ^ c_u256[y2]) + v[b2]; \
v[a3] += (m[x3] ^ c_u256[y3]) + v[b3]; \
v[d0] = __byte_perm(v[d0] ^ v[a0], 0, 0x1032); \
v[d1] = __byte_perm(v[d1] ^ v[a1], 0, 0x1032); \
v[d2] = __byte_perm(v[d2] ^ v[a2], 0, 0x1032); \
v[d3] = __byte_perm(v[d3] ^ v[a3], 0, 0x1032); \
v[c0] += v[d0]; \
v[c1] += v[d1]; \
v[c2] += v[d2]; \
v[c3] += v[d3]; \
v[b0] = SPH_ROTR32(v[b0] ^ v[c0], 12); \
v[b1] = SPH_ROTR32(v[b1] ^ v[c1], 12); \
v[b2] = SPH_ROTR32(v[b2] ^ v[c2], 12); \
v[b3] = SPH_ROTR32(v[b3] ^ v[c3], 12); \
v[a0] += (m[y0] ^ c_u256[x0]) + v[b0]; \
v[a1] += (m[y1] ^ c_u256[x1]) + v[b1]; \
v[a2] += (m[y2] ^ c_u256[x2]) + v[b2]; \
v[a3] += (m[y3] ^ c_u256[x3]) + v[b3]; \
v[d0] = __byte_perm(v[d0] ^ v[a0], 0, 0x0321); \
v[d1] = __byte_perm(v[d1] ^ v[a1], 0, 0x0321); \
v[d2] = __byte_perm(v[d2] ^ v[a2], 0, 0x0321); \
v[d3] = __byte_perm(v[d3] ^ v[a3], 0, 0x0321); \
v[c0] += v[d0]; \
v[c1] += v[d1]; \
v[c2] += v[d2]; \
v[c3] += v[d3]; \
v[b0] = SPH_ROTR32(v[b0] ^ v[c0], 7); \
v[b1] = SPH_ROTR32(v[b1] ^ v[c1], 7); \
v[b2] = SPH_ROTR32(v[b2] ^ v[c2], 7); \
v[b3] = SPH_ROTR32(v[b3] ^ v[c3], 7); \
#define pxorGS2(a,b,c,d, a1,b1,c1,d1) {\
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROL16(v[ d] ^ v[ a]); v[d1] = ROL16(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 12); v[b1] = ROTR32(v[b1] ^ v[c1], 12); \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROR8(v[ d] ^ v[ a]); v[d1] = ROR8(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 7); v[b1] = ROTR32(v[b1] ^ v[c1], 7); \
}
#define pxory1GS2(a,b,c,d, a1,b1,c1,d1) { \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROL16(v[ d] ^ v[ a]); v[d1] = ROL16(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 12); v[b1] = ROTR32(v[b1] ^ v[c1], 12); \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= (c_xors[i++]^nonce) + v[b1]; \
v[ d] = ROR8(v[ d] ^ v[ a]); v[d1] = ROR8(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 7); v[b1] = ROTR32(v[b1] ^ v[c1], 7); \
}
static const __constant__ uint32_t c_u256[16] = {
0x243F6A88, 0x85A308D3, 0x13198A2E, 0x03707344,
0xA4093822, 0x299F31D0, 0x082EFA98, 0xEC4E6C89,
0x452821E6, 0x38D01377, 0xBE5466CF, 0x34E90C6C,
0xC0AC29B7, 0xC97C50DD, 0x3F84D5B5, 0xB5470917
};
#define pxory0GS2(a,b,c,d, a1,b1,c1,d1) { \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROL16(v[ d] ^ v[ a]); v[d1] = ROL16(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 12); v[b1] = ROTR32(v[b1] ^ v[c1], 12); \
v[ a]+= (c_xors[i++]^nonce) + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROR8(v[ d] ^ v[ a]); v[d1] = ROR8(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 7); v[b1] = ROTR32(v[b1] ^ v[c1], 7); \
}
#define pxorx1GS2(a,b,c,d, a1,b1,c1,d1) { \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= (c_xors[i++]^nonce) + v[b1]; \
v[ d] = ROL16(v[ d] ^ v[ a]); v[d1] = ROL16(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 12); v[b1] = ROTR32(v[b1] ^ v[c1], 12); \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROR8(v[ d] ^ v[ a]); v[d1] = ROR8(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 7); v[b1] = ROTR32(v[b1] ^ v[c1], 7); \
}
__device__ __forceinline__
uint32_t blake256_compress_14(uint32_t *m, uint32_t *v_init, uint32_t d_data6, uint32_t d_data7)
#define pxorx0GS2(a,b,c,d, a1,b1,c1,d1) { \
v[ a]+= (c_xors[i++]^nonce) + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROL16(v[ d] ^ v[ a]); v[d1] = ROL16(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 12); v[b1] = ROTR32(v[b1] ^ v[c1], 12); \
v[ a]+= c_xors[i++] + v[ b]; v[a1]+= c_xors[i++] + v[b1]; \
v[ d] = ROR8(v[ d] ^ v[ a]); v[d1] = ROR8(v[d1] ^ v[a1]); \
v[ c]+= v[ d]; v[c1]+= v[d1]; \
v[ b] = ROTR32(v[ b] ^ v[ c], 7); v[b1] = ROTR32(v[b1] ^ v[c1], 7); \
}
__global__ __launch_bounds__(TPB,1)
void decred_gpu_hash_nonce(const uint32_t threads, const uint32_t startNonce, uint32_t *resNonce, const uint32_t highTarget)
{
uint32_t v[16];
const uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
#pragma unroll
for (uint32_t i = 0; i < 16; i++) v[i] = v_init[i];
// these two are not modified:
v[ 9] = 0x85A308D3;
v[13] = 0x299F31D0 ^ (180U*8U);
// round 1 with nonce
GSPREC(1, 5, 0x9, 0xD, 2, 3);
GSPREC(0, 5, 0xA, 0xF, 8, 9);
GSPREC(1, 6, 0xB, 0xC, 10, 11);
GSPREC(2, 7, 0x8, 0xD, 12, 13);
GSPREC(3, 4, 0x9, 0xE, 14, 15);
// round 2
GSPREC4(0, 4, 0x8, 0xC, 14, 10, 1, 5, 0x9, 0xD, 4, 8, 2, 6, 0xA, 0xE, 9, 15, 3, 7, 0xB, 0xF, 13, 6);
GSPREC4(0, 5, 0xA, 0xF, 1, 12, 1, 6, 0xB, 0xC, 0, 2, 2, 7, 0x8, 0xD, 11, 7, 3, 4, 0x9, 0xE, 5, 3);
// round 3
GSPREC4(0, 4, 0x8, 0xC, 11, 8, 1, 5, 0x9, 0xD, 12, 0, 2, 6, 0xA, 0xE, 5, 2, 3, 7, 0xB, 0xF, 15, 13);
GSPREC4(0, 5, 0xA, 0xF, 10, 14, 1, 6, 0xB, 0xC, 3, 6, 2, 7, 0x8, 0xD, 7, 1, 3, 4, 0x9, 0xE, 9, 4);
// round 4
GSPREC4(0, 4, 0x8, 0xC, 7, 9, 1, 5, 0x9, 0xD, 3, 1, 2, 6, 0xA, 0xE, 13, 12, 3, 7, 0xB, 0xF, 11, 14);
GSPREC4(0, 5, 0xA, 0xF, 2, 6, 1, 6, 0xB, 0xC, 5, 10, 2, 7, 0x8, 0xD, 4, 0, 3, 4, 0x9, 0xE, 15, 8);
// round 5
GSPREC4(0, 4, 0x8, 0xC, 9, 0, 1, 5, 0x9, 0xD, 5, 7, 2, 6, 0xA, 0xE, 2, 4, 3, 7, 0xB, 0xF, 10, 15);
GSPREC4(0, 5, 0xA, 0xF, 14, 1, 1, 6, 0xB, 0xC, 11, 12, 2, 7, 0x8, 0xD, 6, 8, 3, 4, 0x9, 0xE, 3, 13);
// round 6
GSPREC4(0, 4, 0x8, 0xC, 2, 12, 1, 5, 0x9, 0xD, 6, 10, 2, 6, 0xA, 0xE, 0, 11, 3, 7, 0xB, 0xF, 8, 3);
GSPREC4(0, 5, 0xA, 0xF, 4, 13, 1, 6, 0xB, 0xC, 7, 5, 2, 7, 0x8, 0xD, 15,14, 3, 4, 0x9, 0xE, 1, 9);
// round 7
GSPREC4(0, 4, 0x8, 0xC, 12, 5, 1, 5, 0x9, 0xD, 1, 15, 2, 6, 0xA, 0xE, 14,13, 3, 7, 0xB, 0xF, 4, 10);
GSPREC4(0, 5, 0xA, 0xF, 0, 7, 1, 6, 0xB, 0xC, 6, 3, 2, 7, 0x8, 0xD, 9, 2, 3, 4, 0x9, 0xE, 8, 11);
#ifdef FULL_4WAY
// round 8
GSPREC4(0, 4, 0x8, 0xC, 13,11, 1, 5, 0x9, 0xD, 7, 14, 2, 6, 0xA, 0xE, 12, 1, 3, 7, 0xB, 0xF, 3, 9);
GSPREC4(0, 5, 0xA, 0xF, 5, 0, 1, 6, 0xB, 0xC, 15, 4, 2, 7, 0x8, 0xD, 8, 6, 3, 4, 0x9, 0xE, 2, 10);
// round 9
GSPREC4(0, 4, 0x8, 0xC, 6, 15, 1, 5, 0x9, 0xD, 14, 9, 2, 6, 0xA, 0xE, 11, 3, 3, 7, 0xB, 0xF, 0, 8);
GSPREC4(0, 5, 0xA, 0xF, 12, 2, 1, 6, 0xB, 0xC, 13, 7, 2, 7, 0x8, 0xD, 1, 4, 3, 4, 0x9, 0xE, 10, 5);
// round 10
GSPREC4(0, 4, 0x8, 0xC, 10, 2, 1, 5, 0x9, 0xD, 8, 4, 2, 6, 0xA, 0xE, 7, 6, 3, 7, 0xB, 0xF, 1, 5);
GSPREC4(0, 5, 0xA, 0xF, 15,11, 1, 6, 0xB, 0xC, 9, 14, 2, 7, 0x8, 0xD, 3, 12, 3, 4, 0x9, 0xE, 13, 0);
// round 11
GSPREC4(0, 4, 0x8, 0xC, 0, 1, 1, 5, 0x9, 0xD, 2, 3, 2, 6, 0xA, 0xE, 4, 5, 3, 7, 0xB, 0xF, 6, 7);
GSPREC4(0, 5, 0xA, 0xF, 8, 9, 1, 6, 0xB, 0xC, 10,11, 2, 7, 0x8, 0xD, 12,13, 3, 4, 0x9, 0xE, 14,15);
// round 12
GSPREC4(0, 4, 0x8, 0xC, 14,10, 1, 5, 0x9, 0xD, 4, 8, 2, 6, 0xA, 0xE, 9, 15, 3, 7, 0xB, 0xF, 13, 6);
GSPREC4(0, 5, 0xA, 0xF, 1, 12, 1, 6, 0xB, 0xC, 0, 2, 2, 7, 0x8, 0xD, 11, 7, 3, 4, 0x9, 0xE, 5, 3);
// round 13
GSPREC4(0, 4, 0x8, 0xC, 11, 8, 1, 5, 0x9, 0xD, 12, 0, 2, 6, 0xA, 0xE, 5, 2, 3, 7, 0xB, 0xF, 15,13);
GSPREC4(0, 5, 0xA, 0xF, 10,14, 1, 6, 0xB, 0xC, 3, 6, 2, 7, 0x8, 0xD, 7, 1, 3, 4, 0x9, 0xE, 9, 4);
#else
// round 8
GSPREC(0, 4, 0x8, 0xC, 13,11);
GSPREC(1, 5, 0x9, 0xD, 7, 14);
GSPREC(2, 6, 0xA, 0xE, 12, 1);
GSPREC(3, 7, 0xB, 0xF, 3, 9);
GSPREC(0, 5, 0xA, 0xF, 5, 0);
GSPREC(1, 6, 0xB, 0xC, 15, 4);
GSPREC(2, 7, 0x8, 0xD, 8, 6);
GSPREC(3, 4, 0x9, 0xE, 2, 10);
// round 9
GSPREC(0, 4, 0x8, 0xC, 6, 15);
GSPREC(1, 5, 0x9, 0xD, 14, 9);
GSPREC(2, 6, 0xA, 0xE, 11, 3);
GSPREC(3, 7, 0xB, 0xF, 0, 8);
GSPREC(0, 5, 0xA, 0xF, 12, 2);
GSPREC(1, 6, 0xB, 0xC, 13, 7);
GSPREC(2, 7, 0x8, 0xD, 1, 4);
GSPREC(3, 4, 0x9, 0xE, 10, 5);
// round 10
GSPREC(0, 4, 0x8, 0xC, 10, 2);
GSPREC(1, 5, 0x9, 0xD, 8, 4);
GSPREC(2, 6, 0xA, 0xE, 7, 6);
GSPREC(3, 7, 0xB, 0xF, 1, 5);
GSPREC(0, 5, 0xA, 0xF, 15,11);
GSPREC(1, 6, 0xB, 0xC, 9, 14);
GSPREC(2, 7, 0x8, 0xD, 3, 12);
GSPREC(3, 4, 0x9, 0xE, 13, 0);
// round 11
GSPREC(0, 4, 0x8, 0xC, 0, 1);
GSPREC(1, 5, 0x9, 0xD, 2, 3);
GSPREC(2, 6, 0xA, 0xE, 4, 5);
GSPREC(3, 7, 0xB, 0xF, 6, 7);
GSPREC(0, 5, 0xA, 0xF, 8, 9);
GSPREC(1, 6, 0xB, 0xC, 10,11);
GSPREC(2, 7, 0x8, 0xD, 12,13);
GSPREC(3, 4, 0x9, 0xE, 14,15);
// round 12
GSPREC(0, 4, 0x8, 0xC, 14,10);
GSPREC(1, 5, 0x9, 0xD, 4, 8);
GSPREC(2, 6, 0xA, 0xE, 9, 15);
GSPREC(3, 7, 0xB, 0xF, 13, 6);
GSPREC(0, 5, 0xA, 0xF, 1, 12);
GSPREC(1, 6, 0xB, 0xC, 0, 2);
GSPREC(2, 7, 0x8, 0xD, 11, 7);
GSPREC(3, 4, 0x9, 0xE, 5, 3);
// round 13
GSPREC(0, 4, 0x8, 0xC, 11, 8);
GSPREC(1, 5, 0x9, 0xD, 12, 0);
GSPREC(2, 6, 0xA, 0xE, 5, 2);
GSPREC(3, 7, 0xB, 0xF, 15,13);
GSPREC(0, 5, 0xA, 0xF, 10,14);
GSPREC(1, 6, 0xB, 0xC, 3, 6);
GSPREC(2, 7, 0x8, 0xD, 7, 1);
GSPREC(3, 4, 0x9, 0xE, 9, 4);
#endif
// round 14
GSPREC(0, 4, 0x8, 0xC, 7, 9);
GSPREC(1, 5, 0x9, 0xD, 3, 1);
GSPREC(2, 6, 0xA, 0xE, 13,12);
GSPREC(3, 7, 0xB, 0xF, 11,14);
GSPREC(0, 5, 0xA, 0xF, 2, 6);
GSPREC(2, 7, 0x8, 0xD, 4, 0);
if ((d_data7 ^ v[7] ^ v[15]) == 0) {
GSPREC(1, 6, 0xB, 0xC, 5, 10);
GSPREC(3, 4, 0x9, 0xE, 15, 8);
return (d_data6 ^ v[6] ^ v[14]);
}
return UINT32_MAX;
}
/* ############################################################################################################################### */
// ------ Close: Last 52/64 bytes ------
__global__
void blake256_gpu_hash_nonce(const uint32_t threads, const uint32_t startNonce, uint32_t *resNonce, const uint64_t highTarget)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
// if (thread < threads)
if (thread < threads)
{
const uint32_t nonce = startNonce + thread * NPR;
uint32_t m[16], v[16], temp;
const uint32_t d_data6 = d_data[6], d_data7 = d_data[7];
uint32_t v[16];
#pragma unroll
for(int i = 0; i < 8; i++) v[i] = d_data[i];
for(int i=0; i<16; i+=4) {
*(uint4*)&v[i] = *(uint4*)&c_data[i];
}
#pragma unroll
for (uint32_t i = 0; i < 16; i++) m[i] = d_data[i+8U];
const uint32_t nonce = startNonce + thread;
v[ 1]+= (nonce ^ 0x13198A2E);
v[13] = ROR8(v[13] ^ v[1]);
v[ 9]+= v[13];
v[ 5] = ROTR32(v[5] ^ v[9], 7);
v[ 8] = 0x243F6A88;
v[ 9] = 0x85A308D3;
v[10] = 0x13198A2E;
v[11] = 0x03707344;
int i = 0;
v[ 1]+= c_xors[i++];// + v[ 6];
v[ 0]+= v[5];
v[12] = ROL16(v[12] ^ v[ 1]); v[15] = ROL16(v[15] ^ v[ 0]);
v[11]+= v[12]; v[10]+= v[15];
v[ 6] = ROTR32(v[ 6] ^ v[11], 12); v[ 5] = ROTR32(v[5] ^ v[10], 12);
v[ 1]+= c_xors[i++] + v[ 6]; v[ 0]+= c_xors[i++] + v[ 5];
v[12] = ROR8(v[12] ^ v[ 1]); v[15] = ROR8(v[15] ^ v[ 0]);
v[11]+= v[12]; v[10]+= v[15];
v[ 6] = ROTR32(v[ 6] ^ v[11], 7); v[ 5] = ROTR32(v[ 5] ^ v[10], 7);
v[12] = 0xA4093822 ^ (180U*8U);
v[13] = 0x299F31D0 ^ (180U*8U);
v[14] = 0x082EFA98;
v[15] = 0xEC4E6C89;
pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxory1GS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorx1GS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorx1GS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorx1GS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxory1GS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxory1GS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorx1GS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxory0GS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorx0GS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxory1GS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxory1GS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorGS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorx1GS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS2( 2, 7, 8, 13, 3, 4, 9, 14);
pxorx1GS2( 0, 4, 8, 12, 1, 5, 9, 13); pxorGS2( 2, 6, 10, 14, 3, 7, 11, 15); pxorGS2( 0, 5, 10, 15, 1, 6, 11, 12); pxorGS( 2, 7, 8, 13);
// round 1 without nonce
GSPREC(0, 4, 0x8, 0xC, 0, 1);
GSPREC(2, 6, 0xA, 0xE, 4, 5);
GSPREC(3, 7, 0xB, 0xF, 6, 7);
for (m[3] = nonce; m[3] < nonce + NPR; m[3]++) {
temp = blake256_compress_14(m, v, d_data6, d_data7);
if (temp != UINT32_MAX && cuda_swab32(temp) <= highTarget) {
#if NBN == 2
if (resNonce[0] != UINT32_MAX) resNonce[1] = m[3];
else resNonce[0] = m[3];
#else
resNonce[0] = m[3];
#endif
// from alexis78:
// return statement allows CUDA7.5 to :
// 1. Store the values fetched from constant memory in registers.
// 2. Perform more precomputations on the outside of the for loop.
// 3. Stop the continuous fetches from the constant memory while iterating
if ((c_h[1]^v[15]) == v[7]) {
v[ 3] += c_xors[i++] + v[4];
v[14] = ROL16(v[14] ^ v[3]);
v[ 9] += v[14];
v[ 4] = ROTR32(v[4] ^ v[9], 12);
v[ 3] += c_xors[i++] + v[4];
v[14] = ROR8(v[14] ^ v[3]);
if(cuda_swab32((c_h[0]^v[6]^v[14])) <= highTarget) {
uint32_t pos = atomicInc(&resNonce[0], UINT32_MAX)+1;
resNonce[pos] = nonce;
return;
}
}
@@ -292,60 +177,163 @@ void blake256_gpu_hash_nonce(const uint32_t threads, const uint32_t startNonce,
}
__host__
static uint32_t decred_cpu_hash_nonce(const int thr_id, const uint32_t threads, const uint32_t startNonce, const uint64_t highTarget)
void decred_cpu_setBlock_52(const uint32_t *input)
{
uint32_t result = UINT32_MAX;
const uint32_t real_threads = threads / NPR;
/*
Precompute everything possible and pass it on constant memory
*/
const uint32_t z[16] = {
0x243F6A88U, 0x85A308D3U, 0x13198A2EU, 0x03707344U,
0xA4093822U, 0x299F31D0U, 0x082EFA98U, 0xEC4E6C89U,
0x452821E6U, 0x38D01377U, 0xBE5466CFU, 0x34E90C6CU,
0xC0AC29B7U, 0xC97C50DDU, 0x3F84D5B5U, 0xB5470917U
};
dim3 grid((real_threads + TPB-1)/TPB);
dim3 block(TPB);
int i=0;
uint32_t _ALIGN(64) preXOR[215];
uint32_t _ALIGN(64) data[16];
uint32_t _ALIGN(64) m[16];
uint32_t _ALIGN(64) h[ 2];
/* Check error on Ctrl+C or kill to prevent segfaults on exit */
if (cudaMemset(d_resNonce[thr_id], 0xff, NBN*sizeof(uint32_t)) != cudaSuccess)
return result;
blake256_gpu_hash_nonce <<<grid, block>>> (real_threads, startNonce, d_resNonce[thr_id], highTarget);
cudaThreadSynchronize();
if (cudaSuccess == cudaMemcpy(h_resNonce[thr_id], d_resNonce[thr_id], NBN*sizeof(uint32_t), cudaMemcpyDeviceToHost)) {
result = h_resNonce[thr_id][0];
#if NBN > 1
for (int n=0; n < (NBN-1); n++)
extra_results[n] = h_resNonce[thr_id][n+1];
#endif
}
return result;
}
__host__
static void decred_midstate_128(uint32_t *output, const uint32_t *input)
{
sph_blake256_context ctx;
sph_blake256_set_rounds(14);
sph_blake256_init(&ctx);
sph_blake256(&ctx, input, 128);
memcpy(output, (void*)ctx.H, 32);
}
data[ 0] = ctx.H[0];
data[ 1] = ctx.H[1];
data[ 2] = ctx.H[2];
data[ 3] = ctx.H[3];
data[ 4] = ctx.H[4];
data[ 5] = ctx.H[5];
data[ 8] = ctx.H[6];
data[12] = swab32(input[35]);
data[13] = ctx.H[7];
__host__
void decred_cpu_setBlock_52(uint32_t *penddata, const uint32_t *midstate, const uint32_t *ptarget)
{
uint32_t _ALIGN(64) data[24];
memcpy(data, midstate, 32);
// pre swab32
for (int i=0; i<13; i++)
data[8+i] = swab32(penddata[i]);
data[21] = 0x80000001;
data[22] = 0;
data[23] = 0x000005a0;
CUDA_SAFE_CALL(cudaMemcpyToSymbol(d_data, data, 32 + 64, 0, cudaMemcpyHostToDevice));
}
m[ 0] = swab32(input[32]); m[ 1] = swab32(input[33]);
m[ 2] = swab32(input[34]); m[ 3] = 0;
m[ 4] = swab32(input[36]); m[ 5] = swab32(input[37]);
m[ 6] = swab32(input[38]); m[ 7] = swab32(input[39]);
m[ 8] = swab32(input[40]); m[ 9] = swab32(input[41]);
m[10] = swab32(input[42]); m[11] = swab32(input[43]);
m[12] = swab32(input[44]); m[13] = 0x80000001;
m[14] = 0;
m[15] = 0x000005a0;
h[ 0] = data[ 8];
h[ 1] = data[13];
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_h,h, 8, 0, cudaMemcpyHostToDevice));
data[ 0]+= (m[ 0] ^ z[1]) + data[ 4];
data[12] = SPH_ROTR32(z[4] ^ SPH_C32(0x5A0) ^ data[ 0], 16);
data[ 8] = z[0]+data[12];
data[ 4] = SPH_ROTR32(data[ 4] ^ data[ 8], 12);
data[ 0]+= (m[ 1] ^ z[0]) + data[ 4];
data[12] = SPH_ROTR32(data[12] ^ data[ 0],8);
data[ 8]+= data[12];
data[ 4] = SPH_ROTR32(data[ 4] ^ data[ 8], 7);
data[ 1]+= (m[ 2] ^ z[3]) + data[ 5];
data[13] = SPH_ROTR32((z[5] ^ SPH_C32(0x5A0)) ^ data[ 1], 16);
data[ 9] = z[1]+data[13];
data[ 5] = SPH_ROTR32(data[ 5] ^ data[ 9], 12);
data[ 1]+= data[ 5]; //+nonce ^ ...
data[ 2]+= (m[ 4] ^ z[5]) + h[ 0];
data[14] = SPH_ROTR32(z[6] ^ data[ 2],16);
data[10] = z[2] + data[14];
data[ 6] = SPH_ROTR32(h[ 0] ^ data[10], 12);
data[ 2]+= (m[ 5] ^ z[4]) + data[ 6];
data[14] = SPH_ROTR32(data[14] ^ data[ 2], 8);
data[10]+= data[14];
data[ 6] = SPH_ROTR32(data[ 6] ^ data[10], 7);
data[ 3]+= (m[ 6] ^ z[7]) + h[ 1];
data[15] = SPH_ROTR32(z[7] ^ data[ 3],16);
data[11] = z[3] + data[15];
data[ 7] = SPH_ROTR32(h[ 1] ^ data[11], 12);
data[ 3]+= (m[ 7] ^ z[6]) + data[ 7];
data[15] = SPH_ROTR32(data[15] ^ data[ 3],8);
data[11]+= data[15];
data[ 7] = SPH_ROTR32(data[11] ^ data[ 7], 7);
data[ 0]+= m[ 8] ^ z[9];
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_data, data, 64, 0, cudaMemcpyHostToDevice));
#define precalcXORGS(x,y) { \
preXOR[i++]= (m[x] ^ z[y]); \
preXOR[i++]= (m[y] ^ z[x]); \
}
#define precalcXORGS2(x,y,x1,y1){\
preXOR[i++] = (m[ x] ^ z[ y]);\
preXOR[i++] = (m[x1] ^ z[y1]);\
preXOR[i++] = (m[ y] ^ z[ x]);\
preXOR[i++] = (m[y1] ^ z[x1]);\
}
precalcXORGS(10,11);
preXOR[ 0]+=data[ 6];
preXOR[i++] = (m[9] ^ z[8]);
precalcXORGS2(12,13,14,15);
precalcXORGS2(14,10, 4, 8);
precalcXORGS2( 9,15,13, 6);
precalcXORGS2( 1,12, 0, 2);
precalcXORGS2(11, 7, 5, 3);
precalcXORGS2(11, 8,12, 0);
precalcXORGS2( 5, 2,15,13);
precalcXORGS2(10,14, 3, 6);
precalcXORGS2( 7, 1, 9, 4);
precalcXORGS2( 7, 9, 3, 1);
precalcXORGS2(13,12,11,14);
precalcXORGS2( 2, 6, 5,10);
precalcXORGS2( 4, 0,15, 8);
precalcXORGS2( 9, 0, 5, 7);
precalcXORGS2( 2, 4,10,15);
precalcXORGS2(14, 1,11,12);
precalcXORGS2( 6, 8, 3,13);
precalcXORGS2( 2,12, 6,10);
precalcXORGS2( 0,11, 8, 3);
precalcXORGS2( 4,13, 7, 5);
precalcXORGS2(15,14, 1, 9);
precalcXORGS2(12, 5, 1,15);
precalcXORGS2(14,13, 4,10);
precalcXORGS2( 0, 7, 6, 3);
precalcXORGS2( 9, 2, 8,11);
precalcXORGS2(13,11, 7,14);
precalcXORGS2(12, 1, 3, 9);
precalcXORGS2( 5, 0,15, 4);
precalcXORGS2( 8, 6, 2,10);
precalcXORGS2( 6,15,14, 9);
precalcXORGS2(11, 3, 0, 8);
precalcXORGS2(12, 2,13, 7);
precalcXORGS2( 1, 4,10, 5);
precalcXORGS2(10, 2, 8, 4);
precalcXORGS2( 7, 6, 1, 5);
precalcXORGS2(15,11, 9,14);
precalcXORGS2( 3,12,13, 0);
precalcXORGS2( 0, 1, 2, 3);
precalcXORGS2( 4, 5, 6, 7);
precalcXORGS2( 8, 9,10,11);
precalcXORGS2(12,13,14,15);
precalcXORGS2(14,10, 4, 8);
precalcXORGS2( 9,15,13, 6);
precalcXORGS2( 1,12, 0, 2);
precalcXORGS2(11, 7, 5, 3);
precalcXORGS2(11, 8,12, 0);
precalcXORGS2( 5, 2,15,13);
precalcXORGS2(10,14, 3, 6);
precalcXORGS2( 7, 1, 9, 4);
precalcXORGS2( 7, 9, 3, 1);
precalcXORGS2(13,12,11,14);
precalcXORGS2( 2, 6, 5,10);
precalcXORGS( 4, 0);
precalcXORGS(15, 8);
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_xors, preXOR, 215*sizeof(uint32_t), 0, cudaMemcpyHostToDevice));
}
/* ############################################################################################################################### */
@@ -357,31 +345,25 @@ static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_decred(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[48];
uint32_t _ALIGN(64) midstate[8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t *pnonce = &pdata[DCR_NONCE_OFT32];
const uint32_t first_nonce = *pnonce;
uint64_t targetHigh = ((uint64_t*)ptarget)[3];
const uint32_t targetHigh = opt_benchmark ? 0x1ULL : ptarget[6];
int dev_id = device_map[thr_id];
const int dev_id = device_map[thr_id];
int intensity = (device_sm[dev_id] > 500 && !is_windows()) ? 29 : 25;
if (device_sm[dev_id] < 350) intensity = 22;
uint32_t throughput = cuda_default_throughput(thr_id, 1U << intensity);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
int rc = 0;
const dim3 grid((throughput + TPB-1)/(TPB));
const dim3 block(TPB);
if (opt_benchmark) {
targetHigh = 0x1ULL << 32;
ptarget[6] = swab32(0xff);
}
if (!init[thr_id])
{
if (!init[thr_id]) {
cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
@@ -390,68 +372,93 @@ extern "C" int scanhash_decred(int thr_id, struct work* work, uint32_t max_nonce
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], MAX_RESULTS*sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], MAX_RESULTS*sizeof(uint32_t)), -1);
init[thr_id] = true;
}
memcpy(endiandata, pdata, 180);
decred_midstate_128(midstate, endiandata);
decred_cpu_setBlock_52(&pdata[32], midstate, ptarget);
decred_cpu_setBlock_52(endiandata);
cudaMemset(d_resNonce[thr_id], 0x00, sizeof(uint32_t));
do {
uint32_t* resNonces = h_resNonce[thr_id];
if (resNonces[0]) cudaMemset(d_resNonce[thr_id], 0x00, sizeof(uint32_t));
// GPU HASH
uint32_t foundNonce = decred_cpu_hash_nonce(thr_id, throughput, (*pnonce), targetHigh);
decred_gpu_hash_nonce <<<grid, block>>> (throughput, (*pnonce), d_resNonce[thr_id], targetHigh);
if (foundNonce != UINT32_MAX)
*hashes_done = (*pnonce) - first_nonce + throughput;
// first cell contains the valid nonces count
cudaMemcpy(resNonces, d_resNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
if (resNonces[0])
{
uint32_t vhashcpu[8];
uint32_t Htarg = ptarget[6];
uint32_t _ALIGN(64) vhash[8];
be32enc(&endiandata[DCR_NONCE_OFT32], foundNonce);
decred_hash(vhashcpu, endiandata);
cudaMemcpy(resNonces, d_resNonce[thr_id], (resNonces[0]+1)*sizeof(uint32_t), cudaMemcpyDeviceToHost);
if (vhashcpu[6] <= Htarg && fulltest(vhashcpu, ptarget))
be32enc(&endiandata[DCR_NONCE_OFT32], resNonces[1]);
decred_hash(vhash, endiandata);
if (vhash[6] <= ptarget[6] && fulltest(vhash, ptarget))
{
rc = 1;
work_set_target_ratio(work, vhashcpu);
*hashes_done = (*pnonce) - first_nonce + throughput;
work->nonces[0] = swab32(foundNonce);
#if NBN > 1
if (extra_results[0] != UINT32_MAX) {
be32enc(&endiandata[DCR_NONCE_OFT32], extra_results[0]);
decred_hash(vhashcpu, endiandata);
if (vhashcpu[6] <= Htarg && fulltest(vhashcpu, ptarget)) {
work->nonces[1] = swab32(extra_results[0]);
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio) {
work_set_target_ratio(work, vhashcpu);
xchg(work->nonces[1], work->nonces[0]);
}
rc = 2;
}
extra_results[0] = UINT32_MAX;
}
#endif
work->valid_nonces = 1;
work_set_target_ratio(work, vhash);
work->nonces[0] = swab32(resNonces[1]);
*pnonce = work->nonces[0];
return rc;
}
else if (opt_debug) {
applog_hash(ptarget);
applog_compare_hash(vhashcpu, ptarget);
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNonce);
// search for another nonce
for(uint32_t n=2; n <= resNonces[0]; n++)
{
be32enc(&endiandata[DCR_NONCE_OFT32], resNonces[n]);
decred_hash(vhash, endiandata);
if (vhash[6] <= ptarget[6] && fulltest(vhash, ptarget)) {
work->nonces[1] = swab32(resNonces[n]);
if (bn_hash_target_ratio(vhash, ptarget) > work->shareratio[0]) {
// we really want the best first ? depends...
work->shareratio[1] = work->shareratio[0];
work->sharediff[1] = work->sharediff[0];
xchg(work->nonces[1], work->nonces[0]);
work_set_target_ratio(work, vhash);
work->valid_nonces++;
} else if (work->valid_nonces == 1) {
bn_set_target_ratio(work, vhash, 1);
work->valid_nonces++;
}
work->valid_nonces = 2; // MAX_NONCES submit limited to 2
gpulog(LOG_DEBUG, thr_id, "multiple nonces 1:%08x (%g) %u:%08x (%g)",
work->nonces[0], work->sharediff[0], n, work->nonces[1], work->sharediff[1]);
} else if (vhash[6] > ptarget[6]) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result %u for %08x does not validate on CPU!", n, resNonces[n]);
}
}
return work->valid_nonces;
} else if (vhash[6] > ptarget[6]) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", resNonces[1]);
}
}
*pnonce += throughput;
} while (!work_restart[thr_id].restart && max_nonce > (uint64_t)throughput + (*pnonce));
*hashes_done = (*pnonce) - first_nonce;
return rc;
MyStreamSynchronize(NULL, 0, device_map[thr_id]);
return 0;
}
// cleanup
extern "C" void free_decred(int thr_id)
{
@@ -459,7 +466,6 @@ extern "C" void free_decred(int thr_id)
return;
cudaDeviceSynchronize();
cudaFreeHost(h_resNonce[thr_id]);
cudaFree(d_resNonce[thr_id]);
+88 -27
View File
@@ -14,12 +14,18 @@ extern "C"
#include "cuda_helper.h"
static uint32_t *d_hash[MAX_GPUS];
extern void keccak256_cpu_init(int thr_id, uint32_t threads);
// SM5+ cuda
extern void keccak256_cpu_init(int thr_id);
extern void keccak256_cpu_free(int thr_id);
extern void keccak256_setBlock_80(void *pdata,const void *ptarget);
extern uint32_t keccak256_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_hash, int order);
extern void keccak256_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t* resNonces, const uint2 highTarget);
extern void keccak256_setBlock_80(uint64_t *endiandata);
extern void keccak256_setOutput(int thr_id);
// compat
extern void keccak256_sm3_init(int thr_id, uint32_t threads);
extern void keccak256_sm3_free(int thr_id);
extern void keccak256_sm3_setBlock_80(void *pdata, const void *ptarget);
extern void keccak256_sm3_hash_80(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t* resNonces, int order);
// CPU Hash
extern "C" void keccak256_hash(void *state, const void *input)
@@ -35,6 +41,7 @@ extern "C" void keccak256_hash(void *state, const void *input)
}
static bool init[MAX_GPUS] = { 0 };
static bool use_compat_kernels[MAX_GPUS] = { 0 };
extern "C" int scanhash_keccak256(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
@@ -42,45 +49,97 @@ extern "C" int scanhash_keccak256(int thr_id, struct work* work, uint32_t max_no
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t throughput = cuda_default_throughput(thr_id, 1U << 21); // 256*256*8*4
const int dev_id = device_map[thr_id];
uint32_t throughput;
uint32_t intensity = 23;
if(!use_compat_kernels[thr_id]) {
if (strstr(device_name[dev_id], "GTX 1070")) intensity = 25;
if (strstr(device_name[dev_id], "GTX 1080")) intensity = 26;
}
throughput = cuda_default_throughput(thr_id, 1U << intensity);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark)
ptarget[7] = 0x000f;
if (!init[thr_id]) {
cudaSetDevice(device_map[thr_id]);
if (!init[thr_id])
{
cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_LOG_ERROR();
}
cuda_get_arch(thr_id);
use_compat_kernels[thr_id] = (cuda_arch[dev_id] < 500);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], throughput * 64));
keccak256_cpu_init(thr_id, throughput);
if(!use_compat_kernels[thr_id]) {
keccak256_cpu_init(thr_id);
} else {
// really useful ?
keccak256_sm3_init(thr_id, throughput);
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
init[thr_id] = true;
}
for (int k=0; k < 20; k++) {
for (int k=0; k < 19; k++) {
be32enc(&endiandata[k], pdata[k]);
}
keccak256_setBlock_80((void*)endiandata, ptarget);
const uint2 highTarget = make_uint2(ptarget[6], ptarget[7]);
if(use_compat_kernels[thr_id])
keccak256_sm3_setBlock_80((void*)endiandata, ptarget);
else {
keccak256_setBlock_80((uint64_t*)endiandata);
keccak256_setOutput(thr_id);
}
do {
int order = 0;
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = keccak256_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
if (foundNonce != UINT32_MAX && bench_algo < 0)
{
uint32_t _ALIGN(64) vhash64[8];
be32enc(&endiandata[19], foundNonce);
keccak256_hash(vhash64, endiandata);
if(use_compat_kernels[thr_id])
keccak256_sm3_hash_80(thr_id, throughput, pdata[19], work->nonces, order++);
else {
keccak256_cpu_hash_80(thr_id, throughput, pdata[19], work->nonces, highTarget);
}
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
work_set_target_ratio(work, vhash64);
pdata[19] = foundNonce;
return 1;
if (work->nonces[0] != UINT32_MAX && bench_algo < 0)
{
const uint32_t Htarg = ptarget[7];
uint32_t _ALIGN(64) vhash[8];
be32enc(&endiandata[19], work->nonces[0]);
keccak256_hash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work->valid_nonces = 1;
work_set_target_ratio(work, vhash);
if (!use_compat_kernels[thr_id] && work->nonces[1] != UINT32_MAX) {
be32enc(&endiandata[19], work->nonces[1]);
keccak256_hash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work->valid_nonces++;
bn_set_target_ratio(work, vhash, 1);
}
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
} else {
pdata[19] = work->nonces[0] + 1;
}
return work->valid_nonces;
}
else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNonce);
else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
pdata[19] = work->nonces[0] + 1;
//keccak256_setOutput(thr_id);
continue;
}
}
@@ -105,9 +164,11 @@ extern "C" void free_keccak256(int thr_id)
cudaThreadSynchronize();
cudaFree(d_hash[thr_id]);
keccak256_cpu_free(thr_id);
if(!use_compat_kernels[thr_id])
keccak256_cpu_free(thr_id);
else {
keccak256_sm3_free(thr_id);
}
cudaDeviceSynchronize();
init[thr_id] = false;
+33 -16
View File
@@ -378,6 +378,11 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc
const uint32_t targetHigh = ptarget[6];
int dev_id = device_map[thr_id];
int intensity = (device_sm[dev_id] > 500 && !is_windows()) ? 30 : 24;
if (device_sm[dev_id] < 350) intensity = 22;
uint32_t throughput = cuda_default_throughput(thr_id, 1U << intensity);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
if (!init[thr_id]) {
cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
@@ -387,6 +392,10 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
cudaStreamCreate(&streams[thr_id]);
@@ -402,11 +411,6 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc
vanilla_cpu_setBlock_16(thr_id,endiandata,&pdata[16]);
int intensity = (device_sm[dev_id] > 500 && !is_windows()) ? 30 : 24;
if (device_sm[dev_id] < 350) intensity = 22;
uint32_t throughput = cuda_default_throughput(thr_id, 1U << intensity);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
const dim3 grid((throughput + (NPT*TPB)-1)/(NPT*TPB));
const dim3 block(TPB);
int rc = 0;
@@ -414,6 +418,7 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc
do {
vanilla_gpu_hash_16_8<<<grid,block, 0, streams[thr_id]>>>(throughput, pdata[19], d_resNonce[thr_id], targetHigh);
cudaMemcpyAsync(h_resNonce[thr_id], d_resNonce[thr_id], NBN*sizeof(uint32_t), cudaMemcpyDeviceToHost,streams[thr_id]);
*hashes_done = pdata[19] - first_nonce + throughput;
cudaStreamSynchronize(streams[thr_id]);
if (h_resNonce[thr_id][0] != UINT32_MAX){
@@ -427,31 +432,43 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc
vanillahash(vhashcpu, endiandata, blakerounds);
if (vhashcpu[6] <= Htarg && fulltest(vhashcpu, ptarget)) {
rc = 1;
work->valid_nonces = 1;
work->nonces[0] = h_resNonce[thr_id][0];
work_set_target_ratio(work, vhashcpu);
*hashes_done = pdata[19] - first_nonce + throughput;
pdata[19] = h_resNonce[thr_id][0];
#if NBN > 1
if (h_resNonce[thr_id][1] != UINT32_MAX) {
work->nonces[1] = h_resNonce[thr_id][1];
be32enc(&endiandata[19], h_resNonce[thr_id][1]);
vanillahash(vhashcpu, endiandata, blakerounds);
pdata[21] = h_resNonce[thr_id][1];
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio) {
if (bn_hash_target_ratio(vhashcpu, ptarget) > work->shareratio[0]) {
work_set_target_ratio(work, vhashcpu);
xchg(pdata[19], pdata[21]);
xchg(work->nonces[0], work->nonces[1]);
}
rc = 2;
work->valid_nonces = 2;
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
} else {
pdata[19] = work->nonces[0] + 1; // cursor
}
#endif
return rc;
return work->valid_nonces;
}
else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", h_resNonce[thr_id][0]);
else if (vhashcpu[6] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", h_resNonce[thr_id][0]);
pdata[19] = work->nonces[0] + 1;
continue;
}
}
if ((uint64_t) throughput + pdata[19] >= max_nonce) {
pdata[19] = max_nonce;
break;
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart && ((uint64_t)max_nonce > ((uint64_t)(pdata[19]) + (uint64_t)throughput)));
} while (!work_restart[thr_id].restart);
*hashes_done = pdata[19] - first_nonce;
MyStreamSynchronize(NULL, 0, dev_id);
+97 -42
View File
@@ -1,42 +1,97 @@
You can use ./build.sh to configure and build with default options.
It is advised to run ./autogen.sh before ./configure (autoconf and automake
need to be installed on your system for autogen.sh to work)
./configure has an option named --with-cuda that allows you to specify
where your CUDA 6.5 toolkit is installed (usually /usr/local/cuda,
but some distros may have a different default location)
** How to compile on Fedora 22 (2015) **
# Step 1: gcc and dependencies
dnf install gcc gcc-c++ automake
dnf install jansson-devel openssl-devel libcurl-devel zlib-devel pciutils-devel
# Step 2: nvidia drivers (Download common linux drivers from nvidia site)
dnf install kernel-devel
modprobe -r nouveau
./NVIDIA-Linux-x86_64-346.72.run
# Step 3: CUDA SDK (Download from nvidia archive)
dnf install cuda-repo-fedora20-6-5-prod-6.5-19.x86_64.rpm
dnf install cuda-minimal-build-6-5.x86_64
# add the nvcc binary path to the system
ln -s /usr/local/cuda-6.5 /usr/local/cuda
echo 'export PATH=$PATH:/usr/local/cuda/bin' > /etc/profile.d/cuda.sh
# add the cudart library path to the system
echo /usr/local/cuda/lib64 > /etc/ld.so.conf.d/cuda.conf
ldconfig
# Step 4: Fix the toolkit warning with gcc 5.1
edit the file /usr/local/cuda/include/host_config.h
and comment/delete the line 82 : #error -- unsupported GNU version! gcc 4.9
./build.sh
./ccminer -n
You can use ./build.sh to configure and build with default options.
It is advised to run ./autogen.sh before ./configure (autoconf and automake
need to be installed on your system for autogen.sh to work)
./configure has an option named --with-cuda that allows you to specify
where your CUDA 6.5 toolkit is installed (usually /usr/local/cuda,
but some distros may have a different default location)
** How to compile on Ubuntu (16.04 LTS)
First, install Cuda toolkit and nVidia Driver, and type `nvidia-smi` to check if your card is detected.
Install dependencies
```sudo apt-get install libcurl4-openssl-dev libssl-dev libjansson-dev automake autotools-dev build-essential```
Ubuntu is now shipped with gcc 6 or 7 so please install gcc/g++ 5 and make it the default (required by the cuda toolkit)
```
sudo apt-get install gcc-5 g++-5
sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-5 1
```
Then use the helper ./build.sh in ccminer source folder, edit configure.sh and the Makefile.am if required.
```
./build.sh
./ccminer --version
```
** How to compile on Fedora 25 **
Note: You may find an alternative method via rpms :
see https://negativo17.org/nvidia-driver/ and https://negativo17.org/repos/multimedia/
# Step 1: gcc and dependencies
dnf install gcc gcc-c++ autoconf automake
dnf install jansson-devel openssl-devel libcurl-devel zlib-devel
# Step 2: nvidia drivers (Download common linux drivers from nvidia site)
dnf install kernel-devel
dnf install https://download1.rpmfusion.org/free/fedora/rpmfusion-free-release-$(rpm -E %fedora).noarch.rpm https://download1.rpmfusion.org/nonfree/fedora/rpmfusion-nonfree-release-$(rpm -E %fedora).noarch.rpm
dnf check-update
dnf install xorg-x11-drv-nvidia-cuda kmod-nvidia
ln -s libnvidia-ml.so.1 /usr/lib64/libnvidia-ml.so
# Step 3: CUDA SDK (Download from nvidia the generic ".run" archive)
# --override is required to ignore "too recent" gcc 6.3
# --silent is required to install only the toolkit (no kmod)
./cuda_8.0.61_375.26_linux.run --toolkit --silent --override
nvcc --version
# add the nvcc binary path to the system
ln -s /usr/local/cuda-8.0 /usr/local/cuda # (if not already made)
echo 'export PATH=$PATH:/usr/local/cuda/bin' > /etc/profile.d/cuda.sh
# add the cudart library path to the system
echo /usr/local/cuda/lib64 > /etc/ld.so.conf.d/cuda.conf
ldconfig
# Step 4: Fix the toolkit incompatibility with gcc 6
# You need to build yourself an older GCC/G++ version, i recommend the 5.4
# see https://gcc.gnu.org/mirrors.html
# Note: this manual method will override the default gcc, it could be better to use a custom toolchain prefix
wget ftp://ftp.lip6.fr/pub/gcc/releases/gcc-5.4.0/gcc-5.4.0.tar.bz2
dnf install libmpc-devel mpfr-devel gmp-devel
./configure --prefix=/usr/local --enable-languages=c,c++,lto --disable-multilib
make -j 8 && make install
(while this step, you have the time to cook something :p)
# or, for previous fedora versions, edit the file /usr/local/cuda/include/host_config.h
# and comment/delete the line 121 : #error -- unsupported GNU version! gcc versions later than 5 are not supported!
./build.sh
./ccminer -n
** How to compile on macOS **
# Step 1: download and install CUDA Toolkit 8 or more recent
# https://developer.nvidia.com/cuda-toolkit-archive
# Step 2: install Homebrew
ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)"
# Step 3: dependencies
brew install pkg-config autoconf automake curl openssl llvm
./build.sh
./ccminer -n
+31 -24
View File
@@ -30,8 +30,8 @@ extern void jackpot_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32
extern uint32_t cuda_check_hash_branch(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_inputHash, int order);
// Original jackpothash Funktion aus einem miner Quelltext
extern "C" unsigned int jackpothash(void *state, const void *input)
// CPU HASH JHA v8
extern "C" void jackpothash(void *state, const void *input)
{
uint32_t hash[16];
unsigned int rnd;
@@ -71,8 +71,6 @@ extern "C" unsigned int jackpothash(void *state, const void *input)
}
}
memcpy(state, hash, 32);
return rnd;
}
static bool init[MAX_GPUS] = { 0 };
@@ -105,6 +103,7 @@ extern "C" int scanhash_jackpot(int thr_id, struct work *work, uint32_t max_nonc
gpulog(LOG_ERR, thr_id, "Sorry, This algo is not supported by this GPU arch (SM 3.0 required)");
proper_exit(EXIT_CODE_CUDA_ERROR);
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t) 64 * throughput));
@@ -213,34 +212,42 @@ extern "C" int scanhash_jackpot(int thr_id, struct work *work, uint32_t max_nonc
CUDA_LOG_ERROR();
uint32_t foundNonce = cuda_check_hash_branch(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
work->nonces[0] = cuda_check_hash_branch(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
if (foundNonce != UINT32_MAX)
if (work->nonces[0] != UINT32_MAX)
{
uint32_t vhash64[8];
be32enc(&endiandata[19], foundNonce);
const uint32_t Htarg = ptarget[7];
uint32_t _ALIGN(64) vhash[8];
be32enc(&endiandata[19], work->nonces[0]);
// jackpothash function gibt die Zahl der Runden zurück
jackpothash(vhash64, endiandata);
jackpothash(vhash, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
int res = 1;
work_set_target_ratio(work, vhash64);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work->valid_nonces = 1;
work_set_target_ratio(work, vhash);
#if 0
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
nist5hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
work_set_target_ratio(work, vhash64);
pdata[21] = secNonce;
res++;
work->nonces[1] = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
if (work->nonces[1] != 0) {
be32enc(&endiandata[19], work->nonces[1]);
jackpothash(vhash, endiandata);
bn_set_target_ratio(work, vhash, 1);
work->valid_nonces++;
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
} else {
pdata[19] = work->nonces[0] + 1; // cursor
}
#else
pdata[19] = work->nonces[0] + 1; // cursor
#endif
pdata[19] = foundNonce;
return res;
} else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNonce);
return work->valid_nonces;
}
else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
pdata[19] = work->nonces[0] + 1;
continue;
}
}
+265
View File
@@ -0,0 +1,265 @@
/**
* JHA v8 algorithm - compatible implementation
* @author tpruvot@github 05-2017
*/
extern "C" {
#include "sph/sph_keccak.h"
#include "sph/sph_blake.h"
#include "sph/sph_groestl.h"
#include "sph/sph_jh.h"
#include "sph/sph_skein.h"
}
#include "miner.h"
#include "cuda_helper.h"
#include "quark/cuda_quark.h"
static uint32_t *d_hash[MAX_GPUS] = { 0 };
static uint32_t *d_hash_br2[MAX_GPUS];
static uint32_t *d_tempBranch[MAX_GPUS];
extern void jackpot_keccak512_cpu_init(int thr_id, uint32_t threads);
extern void jackpot_keccak512_cpu_setBlock(void *pdata, size_t inlen);
extern void jackpot_keccak512_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_hash, int order);
// CPU HASH
extern "C" void jha_hash(void *output, const void *input)
{
uint32_t hash[16];
sph_blake512_context ctx_blake;
sph_groestl512_context ctx_groestl;
sph_jh512_context ctx_jh;
sph_keccak512_context ctx_keccak;
sph_skein512_context ctx_skein;
sph_keccak512_init(&ctx_keccak);
sph_keccak512 (&ctx_keccak, input, 80);
sph_keccak512_close(&ctx_keccak, hash);
for (int rnd = 0; rnd < 3; rnd++)
{
if (hash[0] & 0x01) {
sph_groestl512_init(&ctx_groestl);
sph_groestl512 (&ctx_groestl, (&hash), 64);
sph_groestl512_close(&ctx_groestl, (&hash));
}
else {
sph_skein512_init(&ctx_skein);
sph_skein512 (&ctx_skein, (&hash), 64);
sph_skein512_close(&ctx_skein, (&hash));
}
if (hash[0] & 0x01) {
sph_blake512_init(&ctx_blake);
sph_blake512 (&ctx_blake, (&hash), 64);
sph_blake512_close(&ctx_blake, (&hash));
}
else {
sph_jh512_init(&ctx_jh);
sph_jh512 (&ctx_jh, (&hash), 64);
sph_jh512_close(&ctx_jh, (&hash));
}
}
memcpy(output, hash, 32);
}
__global__ __launch_bounds__(128, 8)
void jha_filter_gpu(const uint32_t threads, const uint32_t* d_hash, uint32_t* d_branch2, uint32_t* d_NonceBranch)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
const uint32_t offset = thread * 16U; // 64U / sizeof(uint32_t);
uint4 *psrc = (uint4*) (&d_hash[offset]);
d_NonceBranch[thread] = ((uint8_t*)psrc)[0] & 0x01;
if (d_NonceBranch[thread]) return;
// uint4 = 4x uint32_t = 16 bytes
uint4 *pdst = (uint4*) (&d_branch2[offset]);
pdst[0] = psrc[0];
pdst[1] = psrc[1];
pdst[2] = psrc[2];
pdst[3] = psrc[3];
}
}
__global__ __launch_bounds__(128, 8)
void jha_merge_gpu(const uint32_t threads, uint32_t* d_hash, uint32_t* d_branch2, uint32_t* const d_NonceBranch)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads && !d_NonceBranch[thread])
{
const uint32_t offset = thread * 16U;
uint4 *pdst = (uint4*) (&d_hash[offset]);
uint4 *psrc = (uint4*) (&d_branch2[offset]);
pdst[0] = psrc[0];
pdst[1] = psrc[1];
pdst[2] = psrc[2];
pdst[3] = psrc[3];
}
}
__host__
uint32_t jha_filter_cpu(const int thr_id, const uint32_t threads, const uint32_t *inpHashes, uint32_t* d_branch2)
{
const uint32_t threadsperblock = 128;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
// extract algo permution hashes to a second branch buffer
jha_filter_gpu <<<grid, block>>> (threads, inpHashes, d_branch2, d_tempBranch[thr_id]);
return threads;
}
__host__
void jha_merge_cpu(const int thr_id, const uint32_t threads, uint32_t *outpHashes, uint32_t* d_branch2)
{
const uint32_t threadsperblock = 128;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
// put back second branch hashes to the common buffer d_hash
jha_merge_gpu <<<grid, block>>> (threads, outpHashes, d_branch2, d_tempBranch[thr_id]);
}
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_jha(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[22];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int dev_id = device_map[thr_id];
uint32_t throughput = cuda_default_throughput(thr_id, 1U << 20);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark)
ptarget[7] = 0x000f;
if (!init[thr_id])
{
cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_LOG_ERROR();
}
cuda_get_arch(thr_id);
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t) 64 * throughput));
CUDA_SAFE_CALL(cudaMalloc(&d_hash_br2[thr_id], (size_t) 64 * throughput));
CUDA_SAFE_CALL(cudaMalloc(&d_tempBranch[thr_id], sizeof(uint32_t) * throughput));
jackpot_keccak512_cpu_init(thr_id, throughput);
quark_blake512_cpu_init(thr_id, throughput);
quark_groestl512_cpu_init(thr_id, throughput);
quark_jh512_cpu_init(thr_id, throughput);
quark_skein512_cpu_init(thr_id, throughput);
cuda_check_cpu_init(thr_id, throughput);
init[thr_id] = true;
}
for (int k=0; k < 22; k++)
be32enc(&endiandata[k], pdata[k]);
jackpot_keccak512_cpu_setBlock((void*)endiandata, 80);
cuda_check_cpu_setTarget(ptarget);
do {
int order = 0;
jackpot_keccak512_cpu_hash(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
for (int rnd = 0; rnd < 3; rnd++)
{
jha_filter_cpu(thr_id, throughput, d_hash[thr_id], d_hash_br2[thr_id]);
quark_groestl512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_skein512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash_br2[thr_id], order++);
jha_merge_cpu(thr_id, throughput, d_hash[thr_id], d_hash_br2[thr_id]);
jha_filter_cpu(thr_id, throughput, d_hash[thr_id], d_hash_br2[thr_id]);
quark_blake512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_jh512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash_br2[thr_id], order++);
jha_merge_cpu(thr_id, throughput, d_hash[thr_id], d_hash_br2[thr_id]);
}
*hashes_done = pdata[19] - first_nonce + throughput;
CUDA_LOG_ERROR();
work->nonces[0] = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (work->nonces[0] != UINT32_MAX)
{
const uint32_t Htarg = ptarget[7];
uint32_t _ALIGN(64) vhash[8];
be32enc(&endiandata[19], work->nonces[0]);
jha_hash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work->valid_nonces = 1;
work_set_target_ratio(work, vhash);
work->nonces[1] = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
if (work->nonces[1] != 0) {
be32enc(&endiandata[19], work->nonces[1]);
jha_hash(vhash, endiandata);
bn_set_target_ratio(work, vhash, 1);
work->valid_nonces++;
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1;
} else {
pdata[19] = work->nonces[0] + 1; // cursor
}
return work->valid_nonces;
}
else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
pdata[19] = work->nonces[0] + 1;
continue;
}
}
if ((uint64_t) throughput + pdata[19] >= max_nonce) {
pdata[19] = max_nonce;
break;
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart);
*hashes_done = pdata[19] - first_nonce;
CUDA_LOG_ERROR();
return 0;
}
// cleanup
extern "C" void free_jha(int thr_id)
{
if (!init[thr_id])
return;
cudaThreadSynchronize();
cudaFree(d_hash[thr_id]);
cudaFree(d_hash_br2[thr_id]);
cudaFree(d_tempBranch[thr_id]);
quark_blake512_cpu_free(thr_id);
quark_groestl512_cpu_free(thr_id);
cuda_check_cpu_free(thr_id);
CUDA_LOG_ERROR();
cudaDeviceSynchronize();
init[thr_id] = false;
}
+63 -15
View File
@@ -9,7 +9,7 @@ endif
EXTRA_DIST = autogen.sh README.md README.txt LICENSE.txt \
ccminer.sln ccminer.vcxproj ccminer.vcxproj.filters \
compat/gettimeofday.c compat/getopt/getopt_long.c cpuminer-config.h.in
compat/gettimeofday.c compat/getopt/getopt_long.c
SUBDIRS = compat
@@ -21,6 +21,9 @@ ccminer_SOURCES = elist.h miner.h compat.h \
crc32.c hefty1.c \
ccminer.cpp pools.cpp util.cpp bench.cpp bignum.cpp \
api.cpp hashlog.cpp nvml.cpp stats.cpp sysinfos.cpp cuda.cpp \
nvsettings.cpp \
equi/equi-stratum.cpp equi/equi.cpp equi/blake2/blake2bx.cpp \
equi/equihash.cpp equi/cuda_equi.cu \
heavy/heavy.cu \
heavy/cuda_blake512.cu heavy/cuda_blake512.h \
heavy/cuda_combine.cu heavy/cuda_combine.h \
@@ -28,37 +31,60 @@ ccminer_SOURCES = elist.h miner.h compat.h \
heavy/cuda_hefty1.cu heavy/cuda_hefty1.h \
heavy/cuda_keccak512.cu heavy/cuda_keccak512.h \
heavy/cuda_sha256.cu heavy/cuda_sha256.h \
heavy/bastion.cu heavy/cuda_bastion.cu \
fuguecoin.cpp Algo256/cuda_fugue256.cu sph/fugue.c uint256.h \
groestlcoin.cpp cuda_groestlcoin.cu cuda_groestlcoin.h \
myriadgroestl.cpp cuda_myriadgroestl.cu \
lyra2/Lyra2.c lyra2/Sponge.c \
lyra2/lyra2RE.cu lyra2/cuda_lyra2.cu \
lyra2/lyra2REv2.cu lyra2/cuda_lyra2v2.cu \
lyra2/lyra2REv3.cu lyra2/cuda_lyra2v3.cu \
lyra2/Lyra2Z.c lyra2/lyra2Z.cu lyra2/cuda_lyra2Z.cu \
lyra2/allium.cu \
Algo256/cuda_bmw256.cu Algo256/cuda_cubehash256.cu \
Algo256/cuda_blake256.cu Algo256/cuda_groestl256.cu Algo256/cuda_keccak256.cu Algo256/cuda_skein256.cu \
Algo256/cuda_blake256.cu Algo256/cuda_groestl256.cu \
Algo256/cuda_keccak256_sm3.cu Algo256/cuda_keccak256.cu Algo256/cuda_skein256.cu \
Algo256/blake256.cu Algo256/decred.cu Algo256/vanilla.cu Algo256/keccak256.cu \
Algo256/blake2s.cu sph/blake2s.c \
Algo256/bmw.cu Algo256/cuda_bmw.cu \
JHA/jackpotcoin.cu JHA/cuda_jha_keccak512.cu \
blake2b.cu \
crypto/xmr-rpc.cpp crypto/wildkeccak-cpu.cpp crypto/wildkeccak.cu \
crypto/cryptolight.cu crypto/cryptolight-core.cu crypto/cryptolight-cpu.cpp \
crypto/cryptonight.cu crypto/cryptonight-core.cu crypto/cryptonight-extra.cu \
crypto/cryptonight-cpu.cpp crypto/oaes_lib.cpp crypto/aesb.cpp crypto/cpu/c_keccak.c \
JHA/jha.cu JHA/jackpotcoin.cu JHA/cuda_jha_keccak512.cu \
JHA/cuda_jha_compactionTest.cu cuda_checkhash.cu \
quark/cuda_jh512.cu quark/cuda_quark_blake512.cu quark/cuda_quark_groestl512.cu quark/cuda_skein512.cu \
quark/cuda_bmw512.cu quark/cuda_quark_keccak512.cu \
quark/nist5.cu \
quark/quarkcoin.cu quark/cuda_quark_compactionTest.cu \
neoscrypt/neoscrypt.cpp neoscrypt/neoscrypt-cpu.c neoscrypt/cuda_neoscrypt.cu \
cuda_nist5.cu pentablake.cu skein.cu cuda_skeincoin.cu skein2.cpp zr5.cu \
pentablake.cu skein.cu cuda_skeincoin.cu skein2.cpp zr5.cu \
skunk/skunk.cu skunk/cuda_skunk.cu skunk/cuda_skunk_streebog.cu \
sha256/sha256d.cu sha256/cuda_sha256d.cu sha256/sha256t.cu sha256/cuda_sha256t.cu sha256/sha256q.cu sha256/cuda_sha256q.cu \
sia/sia.cu sia/sia-rpc.cpp sph/blake2b.c \
sph/bmw.c sph/blake.c sph/groestl.c sph/jh.c sph/keccak.c sph/skein.c \
sph/cubehash.c sph/echo.c sph/luffa.c sph/sha2.c sph/shavite.c sph/simd.c \
sph/hamsi.c sph/hamsi_helper.c sph/streebog.c \
sph/shabal.c sph/whirlpool.c sph/sha2big.c sph/haval.c \
sph/ripemd.c sph/sph_sha2.c \
polytimos.cu \
lbry/lbry.cu lbry/cuda_sha256_lbry.cu lbry/cuda_sha512_lbry.cu lbry/cuda_lbry_merged.cu \
qubit/qubit.cu qubit/qubit_luffa512.cu qubit/deep.cu qubit/luffa.cu \
x11/x11.cu x11/fresh.cu x11/cuda_x11_luffa512.cu x11/cuda_x11_cubehash512.cu \
x11/cuda_x11_shavite512.cu x11/cuda_x11_simd512.cu x11/cuda_x11_echo.cu \
x11/cuda_x11_luffa512_Cubehash.cu \
tribus/tribus.cu tribus/cuda_echo512_final.cu \
x11/x11.cu x12/x12.cu x11/fresh.cu x11/cuda_x11_luffa512.cu x11/cuda_x11_cubehash512.cu \
x11/cuda_x11_shavite512.cu x11/cuda_x11_simd512.cu x11/cuda_x11_echo.cu x11/exosis.cu \
x11/cuda_x11_luffa512_Cubehash.cu x11/x11evo.cu x11/timetravel.cu x11/bitcore.cu \
x13/x13.cu x13/cuda_x13_hamsi512.cu x13/cuda_x13_fugue512.cu \
x13/hsr.cu x13/cuda_hsr_sm3.cu x13/sm3.c \
x15/x14.cu x15/x15.cu x15/cuda_x14_shabal512.cu x15/cuda_x15_whirlpool.cu \
x15/whirlpool.cu \
x17/x17.cu x17/cuda_x17_haval256.cu x17/cuda_x17_sha512.cu \
x11/c11.cu x11/s3.cu x11/sib.cu x11/cuda_streebog.cu
x15/whirlpool.cu x15/cuda_x15_whirlpool_sm3.cu \
x16/x16r.cu x16/x16s.cu x16/cuda_x16_echo512.cu x16/cuda_x16_fugue512.cu \
x16/cuda_x16_shabal512.cu x16/cuda_x16_simd512_80.cu \
x16/cuda_x16_echo512_64.cu \
x17/x17.cu x17/hmq17.cu x17/sonoa.cu x17/cuda_x17_haval256.cu x17/cuda_x17_sha512.cu \
phi/phi.cu phi/phi2.cu phi/cuda_phi2.cu phi/cuda_phi2_cubehash512.cu x11/cuda_streebog_maxwell.cu \
x11/c11.cu x11/s3.cu x11/sib.cu x11/veltor.cu x11/cuda_streebog.cu
# scrypt
ccminer_SOURCES += scrypt.cpp scrypt-jane.cpp \
@@ -80,12 +106,25 @@ ccminer_LDFLAGS = $(PTHREAD_FLAGS) @CUDA_LDFLAGS@
ccminer_LDADD = @LIBCURL@ @JANSSON_LIBS@ @PTHREAD_LIBS@ @WS2_LIBS@ @CUDA_LIBS@ @OPENMP_CFLAGS@ @LIBS@ $(nvml_libs)
ccminer_CPPFLAGS = @LIBCURL_CPPFLAGS@ @OPENMP_CFLAGS@ $(CPPFLAGS) $(PTHREAD_FLAGS) -fno-strict-aliasing $(JANSSON_INCLUDES) $(DEF_INCLUDES) $(nvml_defs)
nvcc_ARCH = -gencode=arch=compute_50,code=\"sm_50,compute_50\"
if HAVE_OSX
ccminer_CPPFLAGS += -I/usr/local/llvm/lib/clang/4.0.0/include
ccminer_LDFLAGS += -L/usr/local/llvm/lib
ccminer_LDADD += -lomp
endif
#ccminer_CPPFLAGS += -DUSE_LIBSODIUM
#ccminer_LDFLAGS += -Lequi/lib
#ccminer_LDADD += -lsodium
ccminer_LDADD += -lcuda
nvcc_ARCH :=
#nvcc_ARCH += -gencode=arch=compute_75,code=\"sm_75,compute_75\" # CUDA 10 req.
#nvcc_ARCH += -gencode=arch=compute_70,code=\"sm_70,compute_70\" # CUDA 9.1
#nvcc_ARCH += -gencode=arch=compute_61,code=\"sm_61,compute_61\" # CUDA 8
nvcc_ARCH += -gencode=arch=compute_52,code=\"sm_52,compute_52\"
#nvcc_ARCH += -gencode=arch=compute_50,code=\"sm_50,compute_50\"
#nvcc_ARCH += -gencode=arch=compute_35,code=\"sm_35,compute_35\"
#nvcc_ARCH += -gencode=arch=compute_30,code=\"sm_30,compute_30\"
#nvcc_ARCH += -gencode=arch=compute_20,code=\"sm_21,compute_20\"
nvcc_FLAGS = $(nvcc_ARCH) @CUDA_INCLUDES@ -I. @CUDA_CFLAGS@
nvcc_FLAGS += $(JANSSON_INCLUDES) --ptxas-options="-v"
@@ -100,6 +139,12 @@ Algo256/blake256.o: Algo256/blake256.cu
Algo256/cuda_bmw.o: Algo256/cuda_bmw.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=76 -o $@ -c $<
crypto/cryptonight-core.o: crypto/cryptonight-core.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=64 -o $@ -c $<
crypto/cryptonight-extra.o: crypto/cryptonight-extra.cu
$(NVCC) $(nvcc_FLAGS) -o $@ -c $<
heavy/cuda_hefty1.o: heavy/cuda_hefty1.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
@@ -125,6 +170,9 @@ x17/cuda_x17_sha512.o: x17/cuda_x17_sha512.cu
quark/cuda_quark_blake512.o: quark/cuda_quark_blake512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
quark/cuda_jh512.o: quark/cuda_jh512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
quark/cuda_quark_keccak512.o: quark/cuda_quark_keccak512.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=88 -o $@ -c $<
@@ -136,15 +184,15 @@ JHA/cuda_jha_compactionTest.o: JHA/cuda_jha_compactionTest.cu
# This object does not use cuda device code but call the different kernels (autotune)
scrypt/salsa_kernel.o: scrypt/salsa_kernel.cu
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_20,code=\"sm_21,compute_20\" -o $@ -c $<
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_30,code=\"sm_30,compute_30\" -o $@ -c $<
# These kernels are for older devices (SM)
scrypt/test_kernel.o: scrypt/test_kernel.cu
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_20,code=\"sm_20,compute_20\" -o $@ -c $<
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_30,code=\"sm_30,compute_30\" -o $@ -c $<
scrypt/fermi_kernel.o: scrypt/fermi_kernel.cu
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_20,code=\"sm_21,compute_20\" -o $@ -c $<
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_30,code=\"sm_30,compute_30\" -o $@ -c $<
scrypt/kepler_kernel.o: scrypt/kepler_kernel.cu
$(NVCC) $(JANSSON_INCLUDES) -I. @CUDA_INCLUDES@ @CUDA_CFLAGS@ -gencode=arch=compute_30,code=\"sm_30,compute_30\" -o $@ -c $<
+12 -14
View File
@@ -1,20 +1,19 @@
ccminer
=======
# ccminer
Based on Christian Buchner's &amp; Christian H.'s CUDA project, no more active on github recently.
Based on Christian Buchner's &amp; Christian H.'s CUDA project, no more active on github since 2014.
Fork by tpruvot@github with X14,X15,X17,Blake256,BlakeCoin,Lyra2RE,Skein,ZR5 and others, check the [README.txt](README.txt)
Check the [README.txt](README.txt) for the additions
BTC donation address: 1AJdfCpLWPNoAMDfHF1wD5y8VgKSSTHxPo
[![tip for next commit](https://tip4commit.com/projects/927.svg)](https://tip4commit.com/github/tpruvot/ccminer)
BTC donation address: 1AJdfCpLWPNoAMDfHF1wD5y8VgKSSTHxPo (tpruvot)
A part of the recent algos were originally written by [djm34](https://github.com/djm34).
A part of the recent algos were originally written by [djm34](https://github.com/djm34) and [alexis78](https://github.com/alexis78)
This variant was tested and built on Linux (ubuntu server 14.04) and VStudio 2013 on Windows 7.
This variant was tested and built on Linux (ubuntu server 14.04, 16.04, Fedora 22 to 25)
It is also built for Windows 7 to 10 with VStudio 2013, to stay compatible with Windows 7 and Vista.
Note that the x86 releases are generally faster than x64 ones on Windows.
Note that the x86 releases are generally faster than x64 ones on Windows, but that tend to change with the recent drivers.
The recommended CUDA Toolkit version is [6.5.19](http://developer.download.nvidia.com/compute/cuda/6_5/rel/installers/cuda_6.5.19_windows_general_64.exe), but some light algos could be faster with the version 7.5 (like blake and skein).
The recommended CUDA Toolkit version was the [6.5.19](http://developer.download.nvidia.com/compute/cuda/6_5/rel/installers/cuda_6.5.19_windows_general_64.exe), but some light algos could be faster with the version 7.5 and 8.0 (like lbry, decred and skein).
About source code dependencies
------------------------------
@@ -22,9 +21,7 @@ About source code dependencies
This project requires some libraries to be built :
- OpenSSL (prebuilt for win)
- Curl (prebuilt for win)
- pthreads (prebuilt for win)
The tree now contains recent prebuilt openssl and curl .lib for both x86 and x64 platforms (windows).
@@ -32,7 +29,8 @@ The tree now contains recent prebuilt openssl and curl .lib for both x86 and x64
To rebuild them, you need to clone this repository and its submodules :
git clone https://github.com/peters/curl-for-windows.git compat/curl-for-windows
On Linux, you can use the helper ./build.sh (edit it if required)
There is also an old [Tutorial for windows](http://cudamining.co.uk/url/tutorials/id/3) on [CudaMining](http://cudamining.co.uk) website.
Compile on Linux
----------------
Please see [INSTALL](https://github.com/tpruvot/ccminer/blob/linux/INSTALL) file or [project Wiki](https://github.com/tpruvot/ccminer/wiki/Compatibility)
+162 -22
View File
@@ -1,5 +1,5 @@
ccMiner release 1.7.6 (May 2016) "DCR vote & pool device stats"
ccminer 2.3.1 "lyra2v3, exosis and sha256q"
---------------------------------------------------------------
***************************************************************
@@ -8,8 +8,7 @@ If you find this tool useful and like to support its continuous
tpruvot@github:
BTC : 1AJdfCpLWPNoAMDfHF1wD5y8VgKSSTHxPo
DRK : XeVrkPrWB7pDbdFLfKhF1Z3xpqhsx6wkH3
ZRC : ZEcubH2xp2mpuwxMjy7wZThr5AzLGu3mqT
DCR : DsUCcACGcyP8McNMRXQwbtpDxaVUYLDQDeU
DJM34:
BTC donation address: 1NENYmxwZGHsKFmyjTc5WferTn5VTFb7Ze
@@ -28,28 +27,35 @@ Decred (Blake256 14-rounds - 180 bytes)
HeavyCoin & MjollnirCoin
FugueCoin
GroestlCoin & Myriad-Groestl
JackpotCoin
Lbry Credits
JackpotCoin (JHA)
QuarkCoin family & AnimeCoin
TalkCoin
DarkCoin and other X11 coins
Chaincoin and Flaxscript (C11)
Saffroncoin blake (256 14-rounds)
BlakeCoin (256 8-rounds)
Midnight (BMW 256)
Qubit (Digibyte, ...)
Luffa (Joincoin)
Keccak (Maxcoin)
Pentablake (Blake 512 x5)
1Coin Triple S
Neoscrypt (FeatherCoin)
x11evo (Revolver)
phi2 (LUXCoin)
Scrypt and Scrypt:N
Scrypt-Jane (Chacha)
Sibcoin (sib)
sib (Sibcoin)
Skein (Skein + SHA)
Signatum (Skein cubehash fugue Streebog)
SonoA (Sono)
Tribus (JH, keccak, simd)
Woodcoin (Double Skein)
Vanilla (Blake256 8-rounds - double sha256)
Vertcoin Lyra2RE
Ziftrcoin (ZR5)
Vertcoin Lyra2REv3
Boolberry (Wild Keccak)
Monero (Cryptonight v7 with -a monero)
Aeon (Cryptonight-lite)
where some of these coins have a VERY NOTABLE nVidia advantage
over competing AMD (OpenCL Only) implementations.
@@ -68,43 +74,72 @@ This code is based on the pooler cpuminer and inherits
its command line interface and options.
-a, --algo=ALGO specify the algorithm to use
allium use to mine Garlic
bastion use to mine Joincoin
bitcore use to mine Bitcore's Timetravel10
blake use to mine Saffroncoin (Blake256)
blakecoin use to mine Old Blake 256
blake2s use to mine Nevacoin (Blake2-S 256)
bmw use to mine Midnight
cryptolight use to mine AEON cryptonight variant 1 (MEM/2)
cryptonight use to mine original cryptonight
c11/flax use to mine Chaincoin and Flax
decred use to mine Decred 180 bytes Blake256-14
deep use to mine Deepcoin
dmd-gr use to mine Diamond-Groestl
equihash use to mine ZEC, HUSH and KMD
exosis use to mine EXO
fresh use to mine Freshcoin
fugue256 use to mine Fuguecoin
groestl use to mine Groestlcoin
heavy use to mine Heavycoin
jackpot use to mine Jackpotcoin
hsr use to mine Hshare
jackpot use to mine Sweepcoin
keccak use to mine Maxcoin
keccakc use to mine CreativeCoin
lbry use to mine LBRY Credits
luffa use to mine Joincoin
lyra2 use to mine Vertcoin
mjollnir use to mine Mjollnircoin
lyra2 use to mine CryptoCoin
lyra2v2 use to mine Monacoin
lyra2v3 use to mine Vertcoin
lyra2z use to mine Zerocoin (XZC)
monero use to mine Monero (XMR)
myr-gr use to mine Myriad-Groest
neoscrypt use to mine FeatherCoin
neoscrypt use to mine FeatherCoin, Trezarcoin, Orbitcoin, etc
nist5 use to mine TalkCoin
penta use to mine Joincoin / Pentablake
phi1612 use to mine Seraph
phi2 use to mine LUXCoin
polytimos use to mine Polytimos
quark use to mine Quarkcoin
qubit use to mine Qubit
scrypt use to mine Scrypt coins
scrypt use to mine Scrypt coins (Litecoin, Dogecoin, etc)
scrypt:N use to mine Scrypt-N (:10 for 2048 iterations)
scrypt-jane use to mine Chacha coins like Cache and Ultracoin
s3 use to mine 1coin
s3 use to mine 1coin (ONE)
sha256t use to mine OneCoin (OC)
sha256q use to mine Pyrite
sia use to mine SIA
sib use to mine Sibcoin
skein use to mine Skeincoin
skein2 use to mine Woodcoin
skunk use to mine Signatum
sonoa use to mine Sono
stellite use to mine Stellite (a cryptonight variant)
timetravel use to mine MachineCoin
tribus use to mine Denarius
x11evo use to mine Revolver
x11 use to mine DarkCoin
x14 use to mine X14Coin
x12 use to mine GalaxyCash
x13 use to mine X13
x14 use to mine X14
x15 use to mine Halcyon
x17 use to mine X17
x16r use to mine Raven
x16s use to mine Pigeon and Eden
x17 use to mine X17
vanilla use to mine Vanilla (Blake256)
veltor use to mine VeltorCoin
whirlpool use to mine Joincoin
wildkeccak use to mine Boolberry (Stratum only)
zr5 use to mine ZiftrCoin
-d, --devices gives a comma separated list of CUDA device IDs
@@ -117,8 +152,6 @@ its command line interface and options.
--cuda-schedule Set device threads scheduling mode (default: auto)
-f, --diff-factor Divide difficulty by this factor (default 1.0)
-m, --diff-multiplier Multiply difficulty by this value (default 1.0)
--vote=VOTE block reward vote (for HeavyCoin)
--trust-pool trust the max block reward vote (maxvote) sent by the pool
-o, --url=URL URL of mining server
-O, --userpass=U:P username:password pair for mining server
-u, --user=USERNAME username for mining server
@@ -134,6 +167,7 @@ its command line interface and options.
-T, --timeout=N network timeout, in seconds (default: 300)
-s, --scantime=N upper bound on time spent scanning current work when
long polling is unavailable, in seconds (default: 5)
--submit-stale ignore stale job checks, may create more rejected shares
-n, --ndevs list cuda devices
-N, --statsavg number of samples used to display hashrate (default: 30)
--no-gbt disable getblocktemplate support (height check in solo)
@@ -143,15 +177,19 @@ its command line interface and options.
--no-color disable colored output
-D, --debug enable debug output
-P, --protocol-dump verbose dump of protocol-level activities
-b, --api-bind IP/Port for the miner API (default: 127.0.0.1:4068)
--api-remote Allow remote control
-b, --api-bind=port IP:port for the miner API (default: 127.0.0.1:4068), 0 disabled
--api-remote Allow remote control, like pool switching, imply --api-allow=0/0
--api-allow=... IP/mask of the allowed api client(s), 0/0 for all
--max-temp=N Only mine if gpu temp is less than specified value
--max-rate=N[KMG] Only mine if net hashrate is less than specified value
--max-diff=N Only mine if net difficulty is less than specified value
--max-log-rate Interval to reduce per gpu hashrate logs (default: 3)
--pstate=0 will force the Geforce 9xx to run in P0 P-State
--plimit=150W set the gpu power limit, allow multiple values for N cards
on windows this parameter use percentages (like OC tools)
--tlimit=85 Set the gpu thermal limit (windows only)
--keep-clocks prevent reset clocks and/or power limit on exit
--show-diff display submitted block and net difficulty
--hide-diff Hide submitted shares diff and net difficulty
-B, --background run the miner in the background
--benchmark run in offline benchmark mode
--cputest debug hashes from cpu algorithms
@@ -177,6 +215,18 @@ Scrypt specific options:
cache for mining. Kepler devices may profit.
--no-autotune disable auto-tuning of kernel launch parameters
CryptoNight specific options:
-l, --launch-config gives the launch configuration for each kernel
in a comma separated list, one per device.
--bfactor=[0-12] Run Cryptonight core kernel in smaller pieces,
From 0 (ui freeze) to 12 (smooth), win default is 11
This is a per-device setting like the launch config.
Wildkeccak specific:
-l, --launch-config gives the launch configuration for each kernel
in a comma separated list, one per device.
-k, --scratchpad url Url used to download the scratchpad cache.
>>> Examples <<<
@@ -200,6 +250,8 @@ Example for Fuguecoin pool mining on dwarfpool.com with all your GPUs
Example for Groestlcoin solo mining
ccminer -q -s 1 -a groestl -o http://127.0.0.1:1441/ -u USERNAME -p PASSWORD
Example for Boolberry
ccminer -a wildkeccak -o stratum+tcp://bbr.suprnova.cc:7777 -u tpruvot.donate -p x -k http://bbr.suprnova.cc/scratchpad.bin -l 64x360
Example for Scrypt-N (2048) on Nicehash
ccminer -a scrypt:10 -o stratum+tcp://stratum.nicehash.com:3335 -u 3EujYFcoBzWvpUEvbe3obEG95mBuU88QBD -p x
@@ -238,6 +290,94 @@ so we can more efficiently implement new algorithms using the latest hardware
features.
>>> RELEASE HISTORY <<<
Jan. 30th 2019 v2.3.1
Handle Lyra2v3 algo
Handle sha256q algo
Handle exosis algo
Handle blake2b standard algo
June 23th 2018 v2.3
Handle phi2 header variation for smart contracts
Handle monero, stellite, graft and cryptolight variants
Handle SonoA algo
June 10th 2018 v2.2.6
New phi2 algo for LUX
New allium algo for Garlic
Apr. 02nd 2018 v2.2.5
New x16r algo for Raven
New x16s algo for Pigeon and Eden
New x12 algo for Galaxycash
Equihash (SIMT) sync issues for the Volta generation
Jan. 04th 2018 v2.2.4
Improve lyra2v2
Higher keccak default intensity
Drop SM 2.x support by default, for CUDA 9 and more recent
Dec. 04th 2017 v2.2.3
Polytimos Algo
Handle keccakc variant (with refreshed sha256d merkle)
Optimised keccak for SM5+, based on alexis improvements
Oct. 09th 2017 v2.2.2
Import and clean the hsr algo (x13 + custom hash)
Import and optimise phi algo from LuxCoin repository
Improve sib algo too for maxwell and pascal cards
Small fix to handle more than 9 cards on linux (-d 10+)
Attempt to free equihash memory "properly"
--submit-stale parameter for supernova pool (which change diff too fast)
Sep. 01st 2017 v2.2.1
Improve tribus algo on recent cards (up to +10%)
Aug. 13th 2017 v2.2
New skunk algo, using the heavy streebog algorithm
Enhance tribus algo (+10%)
equihash protocol enhancement on yiimp.ccminer.org and zpool.ca
June 16th 2017 v2.1-tribus
Interface equihash algo with djeZo solver (from nheqminer 0.5c)
New api parameters (and multicast announces for local networks)
New tribus algo
May. 14th 2017 v2.0
Handle cryptonight, wildkeccak and cryptonight-lite
Add a serie of new algos: timetravel, bastion, hmq1725, sha256t
Import lyra2z from djm34 work...
Rework the common skein512 (used in most algos except skein ;)
Upgrade whirlpool algo with alexis version (2x faster)
Store the share diff of second nonce(s) in most algos
Hardware monitoring thread to get more accurate power readings
Small changes for the quiet mode & max-log-rate to reduce logs
Add bitcore and a compatible jha algo
Dec. 21th 2016 v1.8.4
Improve streebog based algos, veltor and sib (from alexis work)
Blake2s greetly improved (3x), thanks to alexis too...
Sep. 28th 2016 v1.8.3
show intensity on startup for each cards
show-diff is now used by default, use --hide-diff if not wanted
Sep. 22th 2016 v1.8.2
lbry improvements by Alexis Provos
Prevent Windows hibernate while mining
veltor algo (basic implementation)
Aug. 10th 2016 v1.8.1
SIA Blake2-B Algo (getwork over stratum for Suprnova)
SIA Nanopool RPC (getwork over http)
Update also the older lyra2 with Nanashi version
July 20th 2016 v1.8.0
Pascal support with cuda 8
lbry new multi sha / ripemd algo (LBC)
x11evo algo (XRE)
Lyra2v2, Neoscrypt and Decred improvements
Enhance windows NVAPI clock and power limits
Led support for mining/shares activity on windows
May 18th 2016 v1.7.6
Decred vote support
+117
View File
@@ -7,44 +7,79 @@
enum sha_algos {
ALGO_BLAKECOIN = 0,
ALGO_BLAKE,
ALGO_BLAKE2B,
ALGO_BLAKE2S,
ALGO_ALLIUM,
ALGO_BMW,
ALGO_BASTION,
ALGO_C11,
ALGO_CRYPTOLIGHT,
ALGO_CRYPTONIGHT,
ALGO_DEEP,
ALGO_DECRED,
ALGO_DMD_GR,
ALGO_EQUIHASH,
ALGO_EXOSIS,
ALGO_FRESH,
ALGO_FUGUE256, /* Fugue256 */
ALGO_GROESTL,
ALGO_HEAVY, /* Heavycoin hash */
ALGO_HMQ1725,
ALGO_HSR,
ALGO_KECCAK,
ALGO_KECCAKC, /* refreshed Keccak with pool factor 256 */
ALGO_JACKPOT,
ALGO_JHA,
ALGO_LBRY,
ALGO_LUFFA,
ALGO_LYRA2,
ALGO_LYRA2v2,
ALGO_LYRA2v3,
ALGO_LYRA2Z,
ALGO_MJOLLNIR, /* Hefty hash */
ALGO_MYR_GR,
ALGO_NEOSCRYPT,
ALGO_NIST5,
ALGO_PENTABLAKE,
ALGO_PHI,
ALGO_PHI2,
ALGO_POLYTIMOS,
ALGO_QUARK,
ALGO_QUBIT,
ALGO_SCRYPT,
ALGO_SCRYPT_JANE,
ALGO_SHA256D,
ALGO_SHA256T,
ALGO_SHA256Q,
ALGO_SIA,
ALGO_SIB,
ALGO_SKEIN,
ALGO_SKEIN2,
ALGO_SKUNK,
ALGO_SONOA,
ALGO_S3,
ALGO_TIMETRAVEL,
ALGO_TRIBUS,
ALGO_BITCORE,
ALGO_X11EVO,
ALGO_X11,
ALGO_X12,
ALGO_X13,
ALGO_X14,
ALGO_X15,
ALGO_X16R,
ALGO_X16S,
ALGO_X17,
ALGO_VANILLA,
ALGO_VELTOR,
ALGO_WHIRLCOIN,
ALGO_WHIRLPOOL,
ALGO_WHIRLPOOLX,
ALGO_WILDKECCAK,
ALGO_ZR5,
ALGO_MONERO,
ALGO_GRAFT,
ALGO_STELLITE,
ALGO_AUTO,
ALGO_COUNT
};
@@ -54,44 +89,79 @@ extern volatile enum sha_algos opt_algo;
static const char *algo_names[] = {
"blakecoin",
"blake",
"blake2b",
"blake2s",
"allium",
"bmw",
"bastion",
"c11",
"cryptolight",
"cryptonight",
"deep",
"decred",
"dmd-gr",
"equihash",
"exosis",
"fresh",
"fugue256",
"groestl",
"heavy",
"hmq1725",
"hsr",
"keccak",
"keccakc",
"jackpot",
"jha",
"lbry",
"luffa",
"lyra2",
"lyra2v2",
"lyra2v3",
"lyra2z",
"mjollnir",
"myr-gr",
"neoscrypt",
"nist5",
"penta",
"phi",
"phi2",
"polytimos",
"quark",
"qubit",
"scrypt",
"scrypt-jane",
"sha256d",
"sha256t",
"sha256q",
"sia",
"sib",
"skein",
"skein2",
"skunk",
"sonoa",
"s3",
"timetravel",
"tribus",
"bitcore",
"x11evo",
"x11",
"x12",
"x13",
"x14",
"x15",
"x16r",
"x16s",
"x17",
"vanilla",
"veltor",
"whirlcoin",
"whirlpool",
"whirlpoolx",
"wildkeccak",
"zr5",
"monero",
"graft",
"stellite",
"auto", /* reserved for multi algo */
""
};
@@ -111,16 +181,38 @@ static inline int algo_to_int(char* arg)
// some aliases...
if (!strcasecmp("all", arg))
i = ALGO_AUTO;
else if (!strcasecmp("cryptonight-light", arg))
i = ALGO_CRYPTOLIGHT;
else if (!strcasecmp("cryptonight-lite", arg))
i = ALGO_CRYPTOLIGHT;
else if (!strcasecmp("flax", arg))
i = ALGO_C11;
else if (!strcasecmp("diamond", arg))
i = ALGO_DMD_GR;
else if (!strcasecmp("equi", arg))
i = ALGO_EQUIHASH;
else if (!strcasecmp("doom", arg))
i = ALGO_LUFFA;
else if (!strcasecmp("hmq17", arg))
i = ALGO_HMQ1725;
else if (!strcasecmp("hshare", arg))
i = ALGO_HSR;
else if (!strcasecmp("lyra2re", arg))
i = ALGO_LYRA2;
else if (!strcasecmp("lyra2rev2", arg))
i = ALGO_LYRA2v2;
else if (!strcasecmp("lyra2rev3", arg))
i = ALGO_LYRA2v3;
else if (!strcasecmp("phi1612", arg))
i = ALGO_PHI;
else if (!strcasecmp("bitcoin", arg))
i = ALGO_SHA256D;
else if (!strcasecmp("sha256", arg))
i = ALGO_SHA256D;
else if (!strcasecmp("thorsriddle", arg))
i = ALGO_VELTOR;
else if (!strcasecmp("timetravel10", arg))
i = ALGO_BITCORE;
else if (!strcasecmp("whirl", arg))
i = ALGO_WHIRLPOOL;
else if (!strcasecmp("ziftr", arg))
@@ -132,4 +224,29 @@ static inline int algo_to_int(char* arg)
return i;
}
static inline int get_cryptonight_algo(int fork)
{
int algo = ALGO_COUNT;
switch (fork) {
case 8:
algo = ALGO_GRAFT;
break;
case 7:
algo = ALGO_MONERO;
break;
case 3:
algo = ALGO_STELLITE;
break;
default:
algo = ALGO_CRYPTONIGHT;
break;
}
return algo;
}
#endif
+408 -76
View File
@@ -8,7 +8,7 @@
* Software Foundation; either version 2 of the License, or (at your option)
* any later version. See COPYING for more details.
*/
#define APIVERSION "1.8"
#define APIVERSION "1.9"
#ifdef WIN32
# define _WINSOCK_DEPRECATED_NO_WARNINGS
@@ -86,13 +86,20 @@ static struct IP4ACCESS *ipaccess = NULL;
#define ALLIP4 "0.0.0.0"
static const char *localaddr = "127.0.0.1";
static const char *UNAVAILABLE = " - API will not be available";
static const char *MUNAVAILABLE = " - API multicast listener will not be available";
static char *buffer = NULL;
static time_t startup = 0;
static int bye = 0;
extern char *opt_api_bind;
extern int opt_api_port;
extern char *opt_api_allow;
extern int opt_api_listen; /* port */
extern int opt_api_remote;
extern char *opt_api_groups;
extern bool opt_api_mcast;
extern char *opt_api_mcast_addr;
extern char *opt_api_mcast_code;
extern char *opt_api_mcast_des;
extern int opt_api_mcast_port;
// current stratum...
extern struct stratum_ctx stratum;
@@ -112,10 +119,14 @@ static void gpustatus(int thr_id)
if (thr_id >= 0 && thr_id < opt_n_threads) {
struct cgpu_info *cgpu = &thr_info[thr_id].gpu;
double khashes_per_watt = 0;
int gpuid = cgpu->gpu_id;
char buf[512]; *buf = '\0';
char* card;
cuda_gpu_info(cgpu);
cgpu->gpu_plimit = device_plimit[cgpu->gpu_id];
#ifdef USE_WRAPNVML
cgpu->has_monitoring = true;
cgpu->gpu_bus = gpu_busid(cgpu);
@@ -123,23 +134,30 @@ static void gpustatus(int thr_id)
cgpu->gpu_fan = (uint16_t) gpu_fanpercent(cgpu);
cgpu->gpu_fan_rpm = (uint16_t) gpu_fanrpm(cgpu);
cgpu->gpu_power = gpu_power(cgpu); // mWatts
cgpu->gpu_plimit = gpu_plimit(cgpu); // mW or %
#endif
cuda_gpu_clocks(cgpu);
// todo: per gpu
cgpu->accepted = p->accepted_count;
cgpu->rejected = p->rejected_count;
cgpu->khashes = stats_get_speed(thr_id, 0.0) / 1000.0;
if (cgpu->monitor.gpu_power) {
cgpu->gpu_power = cgpu->monitor.gpu_power;
khashes_per_watt = (double)cgpu->khashes / cgpu->monitor.gpu_power;
khashes_per_watt *= 1000; // power in mW
//gpulog(LOG_BLUE, thr_id, "KHW: %g", khashes_per_watt);
}
card = device_name[gpuid];
snprintf(buf, sizeof(buf), "GPU=%d;BUS=%hd;CARD=%s;TEMP=%.1f;"
"POWER=%u;FAN=%hu;RPM=%hu;FREQ=%d;KHS=%.2f;HWF=%d;I=%.1f;THR=%u|",
"POWER=%u;FAN=%hu;RPM=%hu;"
"FREQ=%u;MEMFREQ=%u;GPUF=%u;MEMF=%u;"
"KHS=%.2f;KHW=%.5f;PLIM=%u;"
"ACC=%u;REJ=%u;HWF=%u;I=%.1f;THR=%u|",
gpuid, cgpu->gpu_bus, card, cgpu->gpu_temp,
cgpu->gpu_power, cgpu->gpu_fan, cgpu->gpu_fan_rpm,
cgpu->gpu_clock, cgpu->khashes,
cgpu->hw_errors, cgpu->intensity, cgpu->throughput);
cgpu->gpu_clock/1000, cgpu->gpu_memclock/1000, // base freqs in MHz
cgpu->monitor.gpu_clock, cgpu->monitor.gpu_memclock, // current
cgpu->khashes, khashes_per_watt, cgpu->gpu_plimit,
cgpu->accepted, (unsigned) cgpu->rejected, (unsigned) cgpu->hw_errors,
cgpu->intensity, cgpu->throughput);
// append to buffer for multi gpus
strcat(buffer, buf);
@@ -239,7 +257,7 @@ static char *getpoolnfo(char *params)
static void gpuhwinfos(int gpu_id)
{
char buf[256];
char buf[512];
char pstate[8];
char* card;
struct cgpu_info *cgpu = NULL;
@@ -254,6 +272,9 @@ static void gpuhwinfos(int gpu_id)
if (cgpu == NULL)
return;
cuda_gpu_info(cgpu);
cgpu->gpu_plimit = device_plimit[cgpu->gpu_id];
#ifdef USE_WRAPNVML
cgpu->has_monitoring = true;
cgpu->gpu_bus = gpu_busid(cgpu);
@@ -262,10 +283,12 @@ static void gpuhwinfos(int gpu_id)
cgpu->gpu_fan_rpm = (uint16_t) gpu_fanrpm(cgpu);
cgpu->gpu_pstate = (int16_t) gpu_pstate(cgpu);
cgpu->gpu_power = gpu_power(cgpu);
cgpu->gpu_plimit = gpu_plimit(cgpu);
gpu_info(cgpu);
#ifdef WIN32
if (opt_debug) nvapi_pstateinfo(cgpu->gpu_id);
#endif
#endif
cuda_gpu_clocks(cgpu);
memset(pstate, 0, sizeof(pstate));
if (cgpu->gpu_pstate != -1)
@@ -273,13 +296,15 @@ static void gpuhwinfos(int gpu_id)
card = device_name[gpu_id];
snprintf(buf, sizeof(buf), "GPU=%d;BUS=%hd;CARD=%s;SM=%u;MEM=%lu;"
"TEMP=%.1f;FAN=%hu;RPM=%hu;FREQ=%d;MEMFREQ=%d;PST=%s;POWER=%u;"
snprintf(buf, sizeof(buf), "GPU=%d;BUS=%hd;CARD=%s;SM=%hu;MEM=%u;"
"TEMP=%.1f;FAN=%hu;RPM=%hu;FREQ=%u;MEMFREQ=%u;GPUF=%u;MEMF=%u;"
"PST=%s;POWER=%u;PLIM=%u;"
"VID=%hx;PID=%hx;NVML=%d;NVAPI=%d;SN=%s;BIOS=%s|",
gpu_id, cgpu->gpu_bus, card, cgpu->gpu_arch, cgpu->gpu_mem,
gpu_id, cgpu->gpu_bus, card, cgpu->gpu_arch, (uint32_t) cgpu->gpu_mem,
cgpu->gpu_temp, cgpu->gpu_fan, cgpu->gpu_fan_rpm,
cgpu->gpu_clock, cgpu->gpu_memclock,
pstate, cgpu->gpu_power,
cgpu->gpu_clock/1000U, cgpu->gpu_memclock/1000U, // base clocks
cgpu->monitor.gpu_clock, cgpu->monitor.gpu_memclock, // current
pstate, cgpu->gpu_power, cgpu->gpu_plimit,
cgpu->gpu_vid, cgpu->gpu_pid, cgpu->nvml_id, cgpu->nvapi_id,
cgpu->gpu_sn, cgpu->gpu_desc);
@@ -296,8 +321,12 @@ static const char* os_name()
return "windows";
#else
FILE *fd = fopen("/proc/version", "r");
if (!fd || !fscanf(fd, "Linux version %48s", &os_version[6]))
if (!fd)
return "linux";
if (!fscanf(fd, "Linux version %48s", &os_version[6])) {
fclose(fd);
return "linux";
}
fclose(fd);
os_version[48] = '\0';
return (const char*) os_version;
@@ -316,7 +345,7 @@ static void syshwinfos()
memset(buf, 0, sizeof(buf));
snprintf(buf, sizeof(buf), "OS=%s;NVDRIVER=%s;CPUS=%d;CPUTEMP=%d;CPUFREQ=%d|",
os_name(), driver_version, num_cpus, cputc, cpuclk);
os_name(), driver_version, num_cpus, cputc, cpuclk/1000);
strcat(buffer, buf);
}
@@ -347,7 +376,7 @@ static char *gethistory(char *params)
*buffer = '\0';
for (int i = 0; i < records; i++) {
time_t ts = data[i].tm_stat;
p += sprintf(p, "GPU=%d;H=%u;KHS=%.2f;DIFF=%.6f;"
p += sprintf(p, "GPU=%d;H=%u;KHS=%.2f;DIFF=%g;"
"COUNT=%u;FOUND=%u;ID=%u;TS=%u|",
data[i].gpu_id, data[i].height, data[i].hashrate, data[i].difficulty,
data[i].hashcount, data[i].hashfound, data[i].uid, (uint32_t)ts);
@@ -356,7 +385,7 @@ static char *gethistory(char *params)
}
/**
* Returns the job scans ranges (debug purpose)
* Returns the job scans ranges (debug purpose, only with -D)
*/
static char *getscanlog(char *params)
{
@@ -366,9 +395,11 @@ static char *getscanlog(char *params)
*buffer = '\0';
for (int i = 0; i < records; i++) {
time_t ts = data[i].tm_upd;
p += sprintf(p, "H=%u;P=%u;JOB=%u;N=%u;FROM=0x%x;SCANTO=0x%x;"
p += sprintf(p, "H=%u;P=%u;JOB=%u;ID=%d;DIFF=%g;"
"N=0x%x;FROM=0x%x;SCANTO=0x%x;"
"COUNT=0x%x;FOUND=%u;TS=%u|",
data[i].height, data[i].npool, data[i].njobid, data[i].nonce, data[i].scanned_from, data[i].scanned_to,
data[i].height, data[i].npool, data[i].njobid, (int)data[i].job_nonce_id, data[i].sharediff,
data[i].nonce, data[i].scanned_from, data[i].scanned_to,
(data[i].scanned_to - data[i].scanned_from), data[i].tm_sent ? 1 : 0, (uint32_t)ts);
}
return buffer;
@@ -395,15 +426,6 @@ static char *getmeminfo(char *params)
/*****************************************************************************/
/**
* Remote control allowed ?
* TODO: ip filters
*/
static bool check_remote_access(void)
{
return (opt_api_remote > 0);
}
/**
* Set pool by index (pools array in json config)
* switchpool|1|
@@ -412,8 +434,6 @@ static char *remote_switchpool(char *params)
{
bool ret = false;
*buffer = '\0';
if (!check_remote_access())
return buffer;
if (!params || strlen(params) == 0) {
// rotate pool test
ret = pool_switch_next(-1);
@@ -436,8 +456,6 @@ static char *remote_seturl(char *params)
{
bool ret;
*buffer = '\0';
if (!check_remote_access())
return buffer;
if (!params || strlen(params) == 0) {
// rotate pool test
ret = pool_switch_next(-1);
@@ -454,8 +472,6 @@ static char *remote_seturl(char *params)
static char *remote_quit(char *params)
{
*buffer = '\0';
if (!check_remote_access())
return buffer;
bye = 1;
sprintf(buffer, "%s", "bye|");
return buffer;
@@ -467,22 +483,23 @@ static char *gethelp(char *params);
struct CMDS {
const char *name;
char *(*func)(char *);
bool iswritemode;
} cmds[] = {
{ "summary", getsummary },
{ "threads", getthreads },
{ "pool", getpoolnfo },
{ "histo", gethistory },
{ "hwinfo", gethwinfos },
{ "meminfo", getmeminfo },
{ "scanlog", getscanlog },
{ "summary", getsummary, false },
{ "threads", getthreads, false },
{ "pool", getpoolnfo, false },
{ "histo", gethistory, false },
{ "hwinfo", gethwinfos, false },
{ "meminfo", getmeminfo, false },
{ "scanlog", getscanlog, false },
/* remote functions */
{ "seturl", remote_seturl }, /* prefer switchpool, deprecated */
{ "switchpool", remote_switchpool },
{ "quit", remote_quit },
{ "seturl", remote_seturl, true }, /* prefer switchpool, deprecated */
{ "switchpool", remote_switchpool, true },
{ "quit", remote_quit, true },
/* keep it the last */
{ "help", gethelp },
{ "help", gethelp, false },
};
#define CMDMAX ARRAY_SIZE(cmds)
@@ -490,8 +507,10 @@ static char *gethelp(char *params)
{
*buffer = '\0';
char * p = buffer;
for (int i = 0; i < CMDMAX-1; i++)
p += sprintf(p, "%s\n", cmds[i].name);
for (int i = 0; i < CMDMAX-1; i++) {
bool displayed = !cmds[i].iswritemode || opt_api_allow;
if (displayed) p += sprintf(p, "%s\n", cmds[i].name);
}
sprintf(p, "|");
return buffer;
}
@@ -654,6 +673,146 @@ static int websocket_handshake(SOCKETTYPE c, char *result, char *clientkey)
return 0;
}
/*
* Interpret --api-groups G:cmd1:cmd2:cmd3,P:cmd4,*,...
*/
static void setup_groups()
{
const char *api_groups = opt_api_groups ? opt_api_groups : "";
char *buf, *cmd, *ptr, *next, *colon;
char commands[512] = { 0 };
char cmdbuf[128] = { 0 };
char group;
bool addstar;
int i;
buf = (char *)calloc(1, strlen(api_groups) + 1);
if (unlikely(!buf))
proper_exit(1); //, "Failed to malloc ipgroups buf");
strcpy(buf, api_groups);
next = buf;
// for each group defined
while (next && *next) {
ptr = next;
next = strchr(ptr, ',');
if (next)
*(next++) = '\0';
// Validate the group
if (*(ptr+1) != ':') {
colon = strchr(ptr, ':');
if (colon)
*colon = '\0';
proper_exit(1); //, "API invalid group name '%s'", ptr);
}
group = GROUP(*ptr);
if (!VALIDGROUP(group))
proper_exit(1); //, "API invalid group name '%c'", *ptr);
if (group == PRIVGROUP)
proper_exit(1); //, "API group name can't be '%c'", PRIVGROUP);
if (group == NOPRIVGROUP)
proper_exit(1); //, "API group name can't be '%c'", NOPRIVGROUP);
if (apigroups[GROUPOFFSET(group)].commands != NULL)
proper_exit(1); //, "API duplicate group name '%c'", *ptr);
ptr += 2;
// Validate the command list (and handle '*')
cmd = &(commands[0]);
*(cmd++) = '|';
*cmd = '\0';
addstar = false;
while (ptr && *ptr) {
colon = strchr(ptr, ':');
if (colon)
*(colon++) = '\0';
if (strcmp(ptr, "*") == 0)
addstar = true;
else {
bool did = false;
for (i = 0; i < CMDMAX-1; i++) {
if (strcasecmp(ptr, cmds[i].name) == 0) {
did = true;
break;
}
}
if (did) {
// skip duplicates
sprintf(cmdbuf, "|%s|", cmds[i].name);
if (strstr(commands, cmdbuf) == NULL) {
strcpy(cmd, cmds[i].name);
cmd += strlen(cmds[i].name);
*(cmd++) = '|';
*cmd = '\0';
}
} else {
proper_exit(1); //, "API unknown command '%s' in group '%c'", ptr, group);
}
}
ptr = colon;
}
// * = allow all non-iswritemode commands
if (addstar) {
for (i = 0; i < CMDMAX-1; i++) {
if (cmds[i].iswritemode == false) {
// skip duplicates
sprintf(cmdbuf, "|%s|", cmds[i].name);
if (strstr(commands, cmdbuf) == NULL) {
strcpy(cmd, cmds[i].name);
cmd += strlen(cmds[i].name);
*(cmd++) = '|';
*cmd = '\0';
}
}
}
}
ptr = apigroups[GROUPOFFSET(group)].commands = (char *)calloc(1, strlen(commands) + 1);
if (unlikely(!ptr))
proper_exit(1); //, "Failed to malloc group commands buf");
strcpy(ptr, commands);
}
// Now define R (NOPRIVGROUP) as all non-iswritemode commands
cmd = &(commands[0]);
*(cmd++) = '|';
*cmd = '\0';
for (i = 0; i < CMDMAX-1; i++) {
if (cmds[i].iswritemode == false) {
strcpy(cmd, cmds[i].name);
cmd += strlen(cmds[i].name);
*(cmd++) = '|';
*cmd = '\0';
}
}
ptr = apigroups[GROUPOFFSET(NOPRIVGROUP)].commands = (char *)calloc(1, strlen(commands) + 1);
if (unlikely(!ptr))
proper_exit(1); //, "Failed to malloc noprivgroup commands buf");
strcpy(ptr, commands);
// W (PRIVGROUP) is handled as a special case since it simply means all commands
free(buf);
return;
}
/*
* Interpret [W:]IP[/Prefix][,[R|W:]IP2[/Prefix2][,...]] --api-allow option
* special case of 0/0 allows /0 (means all IP addresses)
*/
#define ALLIPS "0/0"
/*
* N.B. IP4 addresses are by Definition 32bit big endian on all platforms
*/
@@ -703,7 +862,7 @@ static void setup_ipaccess()
ipaccess[ips].group = group;
if (strcmp(ptr, ALLIP4) == 0)
if (strcmp(ptr, ALLIPS) == 0 || strcmp(ptr, ALLIP4) == 0)
ipaccess[ips].ip = ipaccess[ips].mask = 0;
else
{
@@ -766,16 +925,184 @@ static bool check_connect(struct sockaddr_in *cli, char **connectaddr, char *gro
}
}
}
else if (strcmp(opt_api_bind, ALLIP4) == 0)
addrok = true;
else
addrok = (strcmp(*connectaddr, localaddr) == 0);
return addrok;
}
static void mcast()
{
struct sockaddr_in listen;
struct ip_mreq grp;
struct sockaddr_in came_from;
time_t bindstart;
char *binderror;
SOCKETTYPE mcast_sock;
SOCKETTYPE reply_sock;
socklen_t came_from_siz;
char *connectaddr;
ssize_t rep;
int bound;
int count;
int reply_port;
bool addrok;
char group;
char expect[] = "ccminer-"; // first 8 bytes constant
char *expect_code;
size_t expect_code_len;
char buf[1024];
char replybuf[1024];
memset(&grp, 0, sizeof(grp));
grp.imr_multiaddr.s_addr = inet_addr(opt_api_mcast_addr);
if (grp.imr_multiaddr.s_addr == INADDR_NONE)
proper_exit(1); //, "Invalid Multicast Address");
grp.imr_interface.s_addr = INADDR_ANY;
mcast_sock = socket(AF_INET, SOCK_DGRAM, 0);
int optval = 1;
if (SOCKETFAIL(setsockopt(mcast_sock, SOL_SOCKET, SO_REUSEADDR, (const char *)(&optval), sizeof(optval)))) {
applog(LOG_ERR, "API mcast setsockopt SO_REUSEADDR failed (%s)%s", strerror(errno), MUNAVAILABLE);
goto die;
}
memset(&listen, 0, sizeof(listen));
listen.sin_family = AF_INET;
listen.sin_addr.s_addr = INADDR_ANY;
listen.sin_port = htons(opt_api_mcast_port);
// try for more than 1 minute ... in case the old one hasn't completely gone yet
bound = 0;
bindstart = time(NULL);
while (bound == 0) {
if (SOCKETFAIL(bind(mcast_sock, (struct sockaddr *)(&listen), sizeof(listen)))) {
binderror = strerror(errno);;
if ((time(NULL) - bindstart) > 61)
break;
else
sleep(30);
}
else
bound = 1;
}
if (bound == 0) {
applog(LOG_ERR, "API mcast bind to port %d failed (%s)%s", opt_api_port, binderror, MUNAVAILABLE);
goto die;
}
if (SOCKETFAIL(setsockopt(mcast_sock, IPPROTO_IP, IP_ADD_MEMBERSHIP, (const char *)(&grp), sizeof(grp)))) {
applog(LOG_ERR, "API mcast join failed (%s)%s", strerror(errno), MUNAVAILABLE);
goto die;
}
expect_code_len = sizeof(expect) + strlen(opt_api_mcast_code);
expect_code = (char *)calloc(1, expect_code_len + 1);
if (!expect_code)
proper_exit(1); //, "Failed to malloc mcast expect_code");
snprintf(expect_code, expect_code_len + 1, "%s%s-", expect, opt_api_mcast_code);
count = 0;
while (42) {
sleep(1);
count++;
came_from_siz = sizeof(came_from);
if (SOCKETFAIL(rep = recvfrom(mcast_sock, buf, sizeof(buf) - 1,
0, (struct sockaddr *)(&came_from), &came_from_siz))) {
applog(LOG_DEBUG, "API mcast failed count=%d (%s) (%d)",
count, strerror(errno), (int)mcast_sock);
continue;
}
addrok = check_connect(&came_from, &connectaddr, &group);
applog(LOG_DEBUG, "API mcast from %s - %s",
connectaddr, addrok ? "Accepted" : "Ignored");
if (!addrok) {
continue;
}
buf[rep] = '\0';
if (rep > 0 && buf[rep - 1] == '\n')
buf[--rep] = '\0';
applog(LOG_DEBUG, "API mcast request rep=%d (%s) from %s:%d",
(int)rep, buf,
inet_ntoa(came_from.sin_addr),
ntohs(came_from.sin_port));
if ((size_t)rep > expect_code_len && memcmp(buf, expect_code, expect_code_len) == 0) {
reply_port = atoi(&buf[expect_code_len]);
if (reply_port < 1 || reply_port > 65535) {
applog(LOG_DEBUG, "API mcast request ignored - invalid port (%s)",
&buf[expect_code_len]);
}
else {
applog(LOG_DEBUG, "API mcast request OK port %s=%d",
&buf[expect_code_len], reply_port);
came_from.sin_port = htons(reply_port);
reply_sock = socket(AF_INET, SOCK_DGRAM, 0);
snprintf(replybuf, sizeof(replybuf),
"ccm-%s-%d-%s", opt_api_mcast_code, opt_api_port, opt_api_mcast_des);
rep = sendto(reply_sock, replybuf, (int) strlen(replybuf) + 1,
0, (struct sockaddr *)(&came_from), (int) sizeof(came_from));
if (SOCKETFAIL(rep)) {
applog(LOG_DEBUG, "API mcast send reply failed (%s) (%d)",
strerror(errno), (int)reply_sock);
} else {
applog(LOG_DEBUG, "API mcast send reply (%s) succeeded (%d) (%d)",
replybuf, (int)rep, (int)reply_sock);
}
CLOSESOCKET(reply_sock);
}
}
else
applog(LOG_DEBUG, "API mcast request was no good");
}
die:
CLOSESOCKET(mcast_sock);
}
static void *mcast_thread(void *userdata)
{
struct thr_info *mythr = (struct thr_info *)userdata;
pthread_detach(pthread_self());
pthread_setcanceltype(PTHREAD_CANCEL_ASYNCHRONOUS, NULL);
mcast();
//PTH(mythr) = 0L;
return NULL;
}
void mcast_init()
{
struct thr_info *thr;
thr = (struct thr_info *)calloc(1, sizeof(*thr));
if (!thr)
proper_exit(1); //, "Failed to calloc mcast thr");
if (unlikely(pthread_create(&thr->pth, NULL, mcast_thread, thr)))
proper_exit(1); //, "API mcast thread create failed");
}
static void api()
{
const char *addr = opt_api_allow;
unsigned short port = (unsigned short) opt_api_listen; // 4068
const char *addr = opt_api_bind;
unsigned short port = (unsigned short) opt_api_port; // 4068
char buf[MYBUFSIZ];
int n, bound;
char *connectaddr;
@@ -793,11 +1120,13 @@ static void api()
SOCKETTYPE c;
SOCKETTYPE *apisock;
if (!opt_api_listen && opt_debug) {
if (!opt_api_port && opt_debug) {
applog(LOG_DEBUG, "API disabled");
return;
}
setup_groups();
if (opt_api_allow) {
setup_ipaccess();
if (ips == 0) {
@@ -818,9 +1147,10 @@ static void api()
memset(&serv, 0, sizeof(serv));
serv.sin_family = AF_INET;
serv.sin_addr.s_addr = inet_addr(addr);
serv.sin_addr.s_addr = inet_addr(addr); // TODO: allow bind to ip/interface
if (serv.sin_addr.s_addr == (in_addr_t)INVINETADDR) {
applog(LOG_ERR, "API initialisation 2 failed (%s)%s", strerror(errno), UNAVAILABLE);
// free(apisock); FIXME!!
return;
}
@@ -848,7 +1178,7 @@ static void api()
binderror = strerror(errno);
if ((time(NULL) - bindstart) > 61)
break;
else if (opt_api_listen == 4068) {
else if (opt_api_port == 4068) {
/* when port is default one, use first available */
if (opt_debug)
applog(LOG_DEBUG, "API bind to port %d failed, trying port %u",
@@ -865,10 +1195,10 @@ static void api()
}
else {
bound = 1;
if (opt_api_listen != port) {
if (opt_api_port != port) {
applog(LOG_WARNING, "API bind to port %d failed - using port %u",
opt_api_listen, (uint32_t) port);
opt_api_listen = port;
opt_api_port, (uint32_t)port);
opt_api_port = port;
}
}
}
@@ -886,6 +1216,16 @@ static void api()
return;
}
if (opt_api_allow && strcmp(opt_api_bind, "127.0.0.1") == 0)
applog(LOG_WARNING, "API open locally in full access mode on port %d", opt_api_port);
else if (opt_api_allow)
applog(LOG_WARNING, "API open in full access mode to %s on port %d", opt_api_allow, opt_api_port);
else if (strcmp(opt_api_bind, "127.0.0.1") != 0)
applog(LOG_INFO, "API open to the network in read-only mode on port %d", opt_api_port);
if (opt_api_mcast)
mcast_init();
buffer = (char *) calloc(1, MYBUFSIZ + 1);
counter = 0;
@@ -912,7 +1252,7 @@ static void api()
char *wskey = NULL;
n = recv(c, &buf[0], SOCK_REC_BUFSZ, 0);
fail = SOCKETFAIL(n);
fail = SOCKETFAIL(n) || n < 0;
if (fail)
buf[0] = '\0';
else if (n > 0 && buf[n-1] == '\n') {
@@ -921,7 +1261,7 @@ static void api()
if (n > 0 && buf[n-1] == '\r')
buf[n-1] = '\0';
}
buf[n] = '\0';
else buf[n] = '\0';
//if (opt_debug && opt_protocol && n > 0)
// applog(LOG_DEBUG, "API: recv command: (%d) '%s'+char(%x)", n, buf, buf[n-1]);
@@ -972,8 +1312,8 @@ static void api()
break;
}
}
CLOSESOCKET(c);
}
CLOSESOCKET(c);
}
}
@@ -1004,16 +1344,8 @@ void api_set_throughput(int thr_id, uint32_t throughput)
{
if (thr_id < MAX_GPUS && thr_info) {
struct cgpu_info *cgpu = &thr_info[thr_id].gpu;
uint32_t ws = throughput;
uint8_t i = 0;
cgpu->throughput = throughput;
while (ws > 1 && i++ < 32)
ws = ws >> 1;
cgpu->intensity_int = i;
cgpu->intensity = (float) i;
if (i && (1U << i) < throughput) {
cgpu->intensity += ((float) (throughput-(1U << i)) / (1U << i));
}
cgpu->intensity = throughput2intensity(throughput);
if (cgpu->throughput != throughput) cgpu->throughput = throughput;
}
// to display in bench results
if (opt_benchmark)
+16 -3
View File
@@ -1,5 +1,5 @@
<?php
/* ccminer API sample UI (API 1.7) */
/* ccminer API sample UI (API 1.9) */
$host = 'http://localhost/api/'; // 'http://'.$_SERVER['SERVER_NAME'].'/api/';
$configs = array(
@@ -65,8 +65,14 @@ function translateField($key)
$intl['TEMP'] = 'T°c';
$intl['FAN'] = 'Fan %';
$intl['FREQ'] = 'Freq.';
$intl['CPUFREQ'] = 'CPU Freq.';
$intl['FREQ'] = 'Base Freq.';
$intl['MEMFREQ'] = 'Mem. Freq.';
$intl['GPUF'] = 'Curr Freq.';
$intl['MEMF'] = 'Mem. Freq.';
$intl['KHW'] = 'Efficiency';
$intl['POWER'] = 'Power';
$intl['PLIM'] = 'P.Limit';
$intl['PST'] = 'P-State';
// pool infos
@@ -98,8 +104,12 @@ function translateValue($key,$val,$data=array())
case 'NAME':
$val = $data['NAME'].'&nbsp;'.$data['VER'];
break;
case 'CPUFREQ':
case 'FREQ':
$val = sprintf("%d MHz", round(floatval($val)/1000.0));
case 'MEMFREQ':
case 'GPUF':
case 'MEMF':
$val = sprintf("%d MHz", $val);
break;
case 'POWER':
$val = sprintf("%d W", round(floatval($val)/1000.0));
@@ -111,6 +121,9 @@ function translateValue($key,$val,$data=array())
case 'NETKHS':
$val = '<span class="bold">'.$val.'</span> kH/s';
break;
case 'KHW':
$val = $val.' kH/W';
break;
case 'NAME':
case 'POOL';
case 'USER':
Executable
+34
View File
@@ -0,0 +1,34 @@
#!/usr/bin/perl
# sample script to query ccminer API
my $command = "summary|";
use Socket;
use IO::Socket::INET;
my $sock = new IO::Socket::INET (
PeerAddr => '127.0.0.1',
PeerPort => 4068,
Proto => 'tcp',
ReuseAddr => 1,
Timeout => 10,
);
if ($sock) {
print $sock $command;
my $res = "";
while(<$sock>) {
$res .= $_;
}
close($sock);
print("$res\n");
} else {
print("ccminer socket failed\n");
}
+1 -1
View File
@@ -1,7 +1,7 @@
<!DOCTYPE HTML>
<html>
<head>
<script src="//ajax.googleapis.com/ajax/libs/jquery/1.11.1/jquery.min.js"></script>
<script src="http://ajax.googleapis.com/ajax/libs/jquery/1.11.1/jquery.min.js"></script>
<script src="http://code.highcharts.com/highcharts.js"></script>
<script src="http://code.highcharts.com/modules/exporting.js"></script>
</head>
+65 -2
View File
@@ -8,6 +8,11 @@
#include "miner.h"
#include "algos.h"
#include <cuda_runtime.h>
#ifdef __APPLE__
#include "compat/pthreads/pthread_barrier.hpp"
#endif
int bench_algo = -1;
@@ -44,42 +49,76 @@ void bench_free()
void algo_free_all(int thr_id)
{
// only initialized algos will be freed
free_allium(thr_id);
free_bastion(thr_id);
free_bitcore(thr_id);
free_blake256(thr_id);
free_blake2b(thr_id);
free_blake2s(thr_id);
free_bmw(thr_id);
free_c11(thr_id);
free_cryptolight(thr_id);
free_cryptonight(thr_id);
free_decred(thr_id);
free_deep(thr_id);
free_equihash(thr_id);
free_exosis(thr_id);
free_keccak256(thr_id);
free_fresh(thr_id);
free_fugue256(thr_id);
free_groestlcoin(thr_id);
#ifdef WITH_HEAVY_ALGO
free_heavy(thr_id);
#endif
free_hmq17(thr_id);
free_hsr(thr_id);
free_jackpot(thr_id);
free_jha(thr_id);
free_lbry(thr_id);
free_luffa(thr_id);
free_lyra2(thr_id);
free_lyra2v2(thr_id);
free_lyra2v3(thr_id);
free_lyra2Z(thr_id);
free_myriad(thr_id);
free_neoscrypt(thr_id);
free_nist5(thr_id);
free_pentablake(thr_id);
free_phi(thr_id);
free_phi2(thr_id);
free_polytimos(thr_id);
free_quark(thr_id);
free_qubit(thr_id);
free_skeincoin(thr_id);
free_skein2(thr_id);
free_skunk(thr_id);
free_sha256d(thr_id);
free_sha256t(thr_id);
free_sha256q(thr_id);
free_sia(thr_id);
free_sib(thr_id);
free_sonoa(thr_id);
free_s3(thr_id);
free_vanilla(thr_id);
free_veltor(thr_id);
free_whirl(thr_id);
//free_whirlx(thr_id);
free_wildkeccak(thr_id);
free_x11evo(thr_id);
free_x11(thr_id);
free_x12(thr_id);
free_x13(thr_id);
free_x14(thr_id);
free_x15(thr_id);
free_x16r(thr_id);
free_x16s(thr_id);
free_x17(thr_id);
free_zr5(thr_id);
//free_sha256d(thr_id);
free_scrypt(thr_id);
free_scrypt_jane(thr_id);
free_timetravel(thr_id);
free_tribus(thr_id);
free_bitcore(thr_id);
}
// benchmark all algos (called once per mining thread)
@@ -98,14 +137,22 @@ bool bench_algo_switch_next(int thr_id)
// skip some duplicated algos
if (algo == ALGO_C11) algo++; // same as x11
if (algo == ALGO_DMD_GR) algo++; // same as groestl
if (algo == ALGO_HEAVY) algo++; // dead
if (algo == ALGO_MJOLLNIR) algo++; // same as heavy
if (algo == ALGO_KECCAKC) algo++; // same as keccak
if (algo == ALGO_WHIRLCOIN) algo++; // same as whirlpool
if (algo == ALGO_WHIRLPOOLX) algo++; // disabled
// todo: algo switch from RPC 2.0
if (algo == ALGO_CRYPTOLIGHT) algo++;
if (algo == ALGO_CRYPTONIGHT) algo++;
if (algo == ALGO_WILDKECCAK) algo++;
if (algo == ALGO_QUARK) algo++; // to fix
if (algo == ALGO_LBRY && CUDART_VERSION < 7000) algo++;
if (device_sm[dev_id] && device_sm[dev_id] < 300) {
// incompatible SM 2.1 kernels...
if (algo == ALGO_GROESTL) algo++;
if (algo == ALGO_MYR_GR) algo++;
if (algo == ALGO_JACKPOT) algo++; // compact shuffle
if (algo == ALGO_NEOSCRYPT) algo++;
if (algo == ALGO_WHIRLPOOLX) algo++;
}
@@ -113,6 +160,22 @@ bool bench_algo_switch_next(int thr_id)
if (algo == ALGO_SCRYPT) algo++;
if (algo == ALGO_SCRYPT_JANE) algo++;
// Set cryptonight variant
switch (algo) {
case ALGO_MONERO:
cryptonight_fork = 7;
break;
case ALGO_GRAFT:
cryptonight_fork = 8;
break;
case ALGO_STELLITE:
cryptonight_fork = 3;
break;
case ALGO_CRYPTONIGHT:
cryptonight_fork = 1;
break;
}
// free current algo memory and track mem usage
mused = cuda_available_memory(thr_id);
algo_free_all(thr_id);
+20 -6
View File
@@ -5,7 +5,13 @@
#include <stdio.h>
#include "uint256.h"
#include <openssl/opensslv.h>
#if OPENSSL_VERSION_NUMBER < 0x10100000L
#include "compat/bignum_ssl10.hpp"
#else
#include "bignum.hpp"
#endif
#include "miner.h" // hex2bin
@@ -70,17 +76,25 @@ extern "C" double bn_hash_target_ratio(uint32_t* hash, uint32_t* target)
}
// store ratio in work struct
extern "C" void bn_store_hash_target_ratio(uint32_t* hash, uint32_t* target, struct work* work)
extern "C" void bn_store_hash_target_ratio(uint32_t* hash, uint32_t* target, struct work* work, int nonce)
{
// only if the option is enabled (to reduce cpu usage)
if (opt_showdiff) {
work->shareratio = bn_hash_target_ratio(hash, target);
work->sharediff = work->targetdiff * work->shareratio;
}
if (!opt_showdiff) return;
if (nonce < 0 || nonce >= MAX_NONCES) return;
work->shareratio[nonce] = bn_hash_target_ratio(hash, target);
work->sharediff[nonce] = work->targetdiff * work->shareratio[nonce];
}
// new method to save all nonce(s) share diff/ration
extern "C" void bn_set_target_ratio(struct work* work, uint32_t* hash, int nonce)
{
bn_store_hash_target_ratio(hash, work->target, work, nonce);
}
// compat (only store single nonce share diff per work)
extern "C" void work_set_target_ratio(struct work* work, uint32_t* hash)
{
bn_store_hash_target_ratio(hash, work->target, work);
bn_store_hash_target_ratio(hash, work->target, work, work->submit_nonce_id);
}
+111 -129
View File
@@ -9,20 +9,13 @@
#define PROTOCOL_VERSION 70001
#endif
#include <map>
#include <limits>
#include <algorithm>
#include <stdexcept>
#include <vector>
#include "uint256.h"
#include <openssl/bn.h>
#include "serialize.hpp"
using namespace std;
/** Errors thrown by the bignum class */
class bignum_error : public std::runtime_error
{
@@ -60,75 +53,90 @@ public:
/** C++ wrapper for BIGNUM (OpenSSL bignum) */
class CBigNum : public BIGNUM
class CBigNum
{
private:
BIGNUM *self = NULL;
void init()
{
if (self)
BN_clear_free(self);
self = BN_new();
if (!self)
throw bignum_error("CBigNum::init(): BN_new() returned NULL");
}
public:
BIGNUM *get() { return self; }
const BIGNUM *getc() const { return self; }
CBigNum()
{
BN_init(this);
init();
}
CBigNum(const CBigNum& b)
{
BN_init(this);
if (!BN_copy(this, &b))
init();
if (!BN_copy(self, b.getc()))
{
BN_clear_free(this);
BN_clear_free(self);
throw bignum_error("CBigNum::CBigNum(const CBigNum&) : BN_copy failed");
}
}
CBigNum& operator=(const CBigNum& b)
{
if (!BN_copy(this, &b))
if (!BN_copy(self, b.getc()))
throw bignum_error("CBigNum::operator= : BN_copy failed");
return (*this);
}
~CBigNum()
{
BN_clear_free(this);
BN_clear_free(self);
}
//CBigNum(char n) is not portable. Use 'signed char' or 'unsigned char'.
CBigNum(signed char n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(short n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(int n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(long n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(int64 n) { BN_init(this); setint64(n); }
CBigNum(unsigned char n) { BN_init(this); setulong(n); }
CBigNum(unsigned short n) { BN_init(this); setulong(n); }
CBigNum(unsigned int n) { BN_init(this); setulong(n); }
CBigNum(unsigned long n) { BN_init(this); setulong(n); }
CBigNum(uint64 n) { BN_init(this); setuint64(n); }
explicit CBigNum(uint256 n) { BN_init(this); setuint256(n); }
CBigNum(signed char n) { init(); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(short n) { init(); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(int n) { init(); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(long n) { init(); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(int64 n) { init(); setint64(n); }
CBigNum(unsigned char n) { init(); setulong(n); }
CBigNum(unsigned short n) { init(); setulong(n); }
CBigNum(unsigned int n) { init(); setulong(n); }
CBigNum(unsigned long n) { init(); setulong(n); }
CBigNum(uint64 n) { init(); setuint64(n); }
explicit CBigNum(uint256 n) { init(); setuint256(n); }
explicit CBigNum(const std::vector<unsigned char>& vch)
{
BN_init(this);
init();
setvch(vch);
}
void setulong(unsigned long n)
{
if (!BN_set_word(this, n))
if (!BN_set_word(self, n))
throw bignum_error("CBigNum conversion from unsigned long : BN_set_word failed");
}
unsigned long getulong() const
{
return (unsigned long) BN_get_word(this);
return BN_get_word(self);
}
unsigned int getuint() const
{
return (unsigned int) BN_get_word(this);
return BN_get_word(self);
}
int getint() const
{
unsigned long n = (unsigned long) BN_get_word(this);
if (!BN_is_negative(this))
unsigned long n = BN_get_word(self);
if (!BN_is_negative(self))
return (n > (unsigned long)std::numeric_limits<int>::max() ? std::numeric_limits<int>::max() : n);
else
return (n > (unsigned long)std::numeric_limits<int>::max() ? std::numeric_limits<int>::min() : -(int)n);
@@ -141,11 +149,9 @@ public:
bool fNegative;
uint64 n;
if (sn < 0LL)
if (sn < (int64)0)
{
// Since the minimum signed integer cannot be represented as positive so long as its type is signed,
// and it's not well-defined what happens if you make it unsigned before negating it,
// we instead increment the negative integer by 1, convert it, then increment the (now positive) unsigned integer by 1 to compensate
// Since the minimum signed integer cannot be represented as positive so long as its type is signed, and it's not well-defined what happens if you make it unsigned before negating it, we instead increment the negative integer by 1, convert it, then increment the (now positive) unsigned integer by 1 to compensate
n = -(sn + 1);
++n;
fNegative = true;
@@ -171,12 +177,27 @@ public:
}
*p++ = c;
}
unsigned int nSize = (unsigned int) (p - (pch + 4));
unsigned int nSize = p - (pch + 4);
pch[0] = (nSize >> 24) & 0xff;
pch[1] = (nSize >> 16) & 0xff;
pch[2] = (nSize >> 8) & 0xff;
pch[3] = (nSize) & 0xff;
BN_mpi2bn(pch, (int) (p - pch), this);
BN_mpi2bn(pch, p - pch, self);
}
uint64 getuint64()
{
unsigned int nSize = BN_bn2mpi(self, NULL);
if (nSize < 4)
return 0;
std::vector<unsigned char> vch(nSize);
BN_bn2mpi(self, &vch[0]);
if (vch.size() > 4)
vch[4] &= 0x7f;
uint64 n = 0;
for (unsigned int i = 0, j = vch.size()-1; i < sizeof(n) && j >= 4; i++, j--)
((unsigned char*)&n)[i] = vch[j];
return n;
}
void setuint64(uint64 n)
@@ -198,12 +219,12 @@ public:
}
*p++ = c;
}
unsigned int nSize = (unsigned int) (p - (pch + 4));
unsigned int nSize = p - (pch + 4);
pch[0] = (nSize >> 24) & 0xff;
pch[1] = (nSize >> 16) & 0xff;
pch[2] = (nSize >> 8) & 0xff;
pch[3] = (nSize) & 0xff;
BN_mpi2bn(pch, (int) (p - pch), this);
BN_mpi2bn(pch, p - pch, self);
}
void setuint256(uint256 n)
@@ -226,33 +247,34 @@ public:
}
*p++ = c;
}
unsigned int nSize = (unsigned int) (p - (pch + 4));
unsigned int nSize = p - (pch + 4);
pch[0] = (nSize >> 24) & 0xff;
pch[1] = (nSize >> 16) & 0xff;
pch[2] = (nSize >> 8) & 0xff;
pch[3] = (nSize >> 0) & 0xff;
BN_mpi2bn(pch, (int) (p - pch), this);
BN_mpi2bn(pch, p - pch, self);
}
uint256 getuint256() const
uint256 getuint256()
{
unsigned int nSize = BN_bn2mpi(this, NULL);
unsigned int nSize = BN_bn2mpi(self, NULL);
if (nSize < 4)
return 0;
std::vector<unsigned char> vch(nSize);
BN_bn2mpi(this, &vch[0]);
BN_bn2mpi(self, &vch[0]);
if (vch.size() > 4)
vch[4] &= 0x7f;
uint256 n = 0;
for (unsigned int i = 0, j = (unsigned int) vch.size()-1; i < sizeof(n) && j >= 4; i++, j--)
for (unsigned int i = 0, j = vch.size()-1; i < sizeof(n) && j >= 4; i++, j--)
((unsigned char*)&n)[i] = vch[j];
return n;
}
void setvch(const std::vector<unsigned char>& vch)
{
std::vector<unsigned char> vch2(vch.size() + 4);
unsigned int nSize = (unsigned int) vch.size();
unsigned int nSize = vch.size();
// BIGNUM's byte stream format expects 4 bytes of
// big endian size data info at the front
vch2[0] = (nSize >> 24) & 0xff;
@@ -261,83 +283,43 @@ public:
vch2[3] = (nSize >> 0) & 0xff;
// swap data to big endian
reverse_copy(vch.begin(), vch.end(), vch2.begin() + 4);
BN_mpi2bn(&vch2[0], (int) vch2.size(), this);
BN_mpi2bn(&vch2[0], vch2.size(), self);
}
std::vector<unsigned char> getvch() const
{
unsigned int nSize = BN_bn2mpi(this, NULL);
unsigned int nSize = BN_bn2mpi(self, NULL);
if (nSize <= 4)
return std::vector<unsigned char>();
std::vector<unsigned char> vch(nSize);
BN_bn2mpi(this, &vch[0]);
BN_bn2mpi(self, &vch[0]);
vch.erase(vch.begin(), vch.begin() + 4);
reverse(vch.begin(), vch.end());
return vch;
}
// The "compact" format is a representation of a whole
// number N using an unsigned 32bit number similar to a
// floating point format.
// The most significant 8 bits are the unsigned exponent of base 256.
// This exponent can be thought of as "number of bytes of N".
// The lower 23 bits are the mantissa.
// Bit number 24 (0x800000) represents the sign of N.
// N = (-1^sign) * mantissa * 256^(exponent-3)
//
// Satoshi's original implementation used BN_bn2mpi() and BN_mpi2bn().
// MPI uses the most significant bit of the first byte as sign.
// Thus 0x1234560000 is compact (0x05123456)
// and 0xc0de000000 is compact (0x0600c0de)
// (0x05c0de00) would be -0x40de000000
//
// Bitcoin only uses this "compact" format for encoding difficulty
// targets, which are unsigned 256bit quantities. Thus, all the
// complexities of the sign bit and using base 256 are probably an
// implementation accident.
//
// This implementation directly uses shifts instead of going
// through an intermediate MPI representation.
CBigNum& SetCompact(unsigned int nCompact)
{
unsigned int nSize = nCompact >> 24;
bool fNegative =(nCompact & 0x00800000) != 0;
unsigned int nWord = nCompact & 0x007fffff;
if (nSize <= 3)
{
nWord >>= 8*(3-nSize);
BN_set_word(this, nWord);
}
else
{
BN_set_word(this, nWord);
BN_lshift(this, this, 8*(nSize-3));
}
BN_set_negative(this, fNegative);
std::vector<unsigned char> vch(4 + nSize);
vch[3] = nSize;
if (nSize >= 1) vch[4] = (nCompact >> 16) & 0xff;
if (nSize >= 2) vch[5] = (nCompact >> 8) & 0xff;
if (nSize >= 3) vch[6] = (nCompact >> 0) & 0xff;
BN_mpi2bn(&vch[0], vch.size(), self);
return *this;
}
unsigned int GetCompact() const
{
unsigned int nSize = BN_num_bytes(this);
unsigned int nCompact = 0;
if (nSize <= 3)
nCompact = (unsigned int) BN_get_word(this) << 8*(3-nSize);
else
{
CBigNum bn;
BN_rshift(&bn, this, 8*(nSize-3));
nCompact = (unsigned int) BN_get_word(&bn);
}
// The 0x00800000 bit denotes the sign.
// Thus, if it is already set, divide the mantissa by 256 and increase the exponent.
if (nCompact & 0x00800000)
{
nCompact >>= 8;
nSize++;
}
nCompact |= nSize << 24;
nCompact |= (BN_is_negative(this) ? 0x00800000 : 0);
unsigned int nSize = BN_bn2mpi(self, NULL);
std::vector<unsigned char> vch(nSize);
nSize -= 4;
BN_bn2mpi(self, &vch[0]);
unsigned int nCompact = nSize << 24;
if (nSize >= 1) nCompact |= (vch[4] << 16);
if (nSize >= 2) nCompact |= (vch[5] << 8);
if (nSize >= 3) nCompact |= (vch[6] << 0);
return nCompact;
}
@@ -378,20 +360,20 @@ public:
CBigNum bn0 = 0;
std::string str;
CBigNum bn = *this;
BN_set_negative(&bn, false);
BN_set_negative(bn.get(), false);
CBigNum dv;
CBigNum rem;
if (BN_cmp(&bn, &bn0) == 0)
if (BN_cmp(bn.getc(), bn0.getc()) == 0)
return "0";
while (BN_cmp(&bn, &bn0) > 0)
while (BN_cmp(bn.getc(), bn0.getc()) > 0)
{
if (!BN_div(&dv, &rem, &bn, &bnBase, pctx))
if (!BN_div(dv.get(), rem.get(), bn.getc(), bnBase.getc(), pctx))
throw bignum_error("CBigNum::ToString() : BN_div failed");
bn = dv;
unsigned int c = rem.getulong();
str += "0123456789abcdef"[c];
}
if (BN_is_negative(this))
if (BN_is_negative(self))
str += "-";
reverse(str.begin(), str.end());
return str;
@@ -424,12 +406,12 @@ public:
bool operator!() const
{
return BN_is_zero(this);
return BN_is_zero(self);
}
CBigNum& operator+=(const CBigNum& b)
{
if (!BN_add(this, this, &b))
if (!BN_add(self, self, b.getc()))
throw bignum_error("CBigNum::operator+= : BN_add failed");
return *this;
}
@@ -443,7 +425,7 @@ public:
CBigNum& operator*=(const CBigNum& b)
{
CAutoBN_CTX pctx;
if (!BN_mul(this, this, &b, pctx))
if (!BN_mul(self, self, b.getc(), pctx))
throw bignum_error("CBigNum::operator*= : BN_mul failed");
return *this;
}
@@ -462,7 +444,7 @@ public:
CBigNum& operator<<=(unsigned int shift)
{
if (!BN_lshift(this, this, shift))
if (!BN_lshift(self, self, shift))
throw bignum_error("CBigNum:operator<<= : BN_lshift failed");
return *this;
}
@@ -473,13 +455,13 @@ public:
// if built on ubuntu 9.04 or 9.10, probably depends on version of OpenSSL
CBigNum a = 1;
a <<= shift;
if (BN_cmp(&a, this) > 0)
if (BN_cmp(a.getc(), self) > 0)
{
*this = 0;
return *this;
}
if (!BN_rshift(this, this, shift))
if (!BN_rshift(self, self, shift))
throw bignum_error("CBigNum:operator>>= : BN_rshift failed");
return *this;
}
@@ -488,7 +470,7 @@ public:
CBigNum& operator++()
{
// prefix operator
if (!BN_add(this, this, BN_value_one()))
if (!BN_add(self, self, BN_value_one()))
throw bignum_error("CBigNum::operator++ : BN_add failed");
return *this;
}
@@ -505,7 +487,7 @@ public:
{
// prefix operator
CBigNum r;
if (!BN_sub(&r, this, BN_value_one()))
if (!BN_sub(r.get(), self, BN_value_one()))
throw bignum_error("CBigNum::operator-- : BN_sub failed");
*this = r;
return *this;
@@ -530,7 +512,7 @@ public:
inline const CBigNum operator+(const CBigNum& a, const CBigNum& b)
{
CBigNum r;
if (!BN_add(&r, &a, &b))
if (!BN_add(r.get(), a.getc(), b.getc()))
throw bignum_error("CBigNum::operator+ : BN_add failed");
return r;
}
@@ -538,7 +520,7 @@ inline const CBigNum operator+(const CBigNum& a, const CBigNum& b)
inline const CBigNum operator-(const CBigNum& a, const CBigNum& b)
{
CBigNum r;
if (!BN_sub(&r, &a, &b))
if (!BN_sub(r.get(), a.getc(), b.getc()))
throw bignum_error("CBigNum::operator- : BN_sub failed");
return r;
}
@@ -546,7 +528,7 @@ inline const CBigNum operator-(const CBigNum& a, const CBigNum& b)
inline const CBigNum operator-(const CBigNum& a)
{
CBigNum r(a);
BN_set_negative(&r, !BN_is_negative(&r));
BN_set_negative(r.get(), !BN_is_negative(r.getc()));
return r;
}
@@ -554,7 +536,7 @@ inline const CBigNum operator*(const CBigNum& a, const CBigNum& b)
{
CAutoBN_CTX pctx;
CBigNum r;
if (!BN_mul(&r, &a, &b, pctx))
if (!BN_mul(r.get(), a.getc(), b.getc(), pctx))
throw bignum_error("CBigNum::operator* : BN_mul failed");
return r;
}
@@ -563,7 +545,7 @@ inline const CBigNum operator/(const CBigNum& a, const CBigNum& b)
{
CAutoBN_CTX pctx;
CBigNum r;
if (!BN_div(&r, NULL, &a, &b, pctx))
if (!BN_div(r.get(), NULL, a.getc(), b.getc(), pctx))
throw bignum_error("CBigNum::operator/ : BN_div failed");
return r;
}
@@ -572,7 +554,7 @@ inline const CBigNum operator%(const CBigNum& a, const CBigNum& b)
{
CAutoBN_CTX pctx;
CBigNum r;
if (!BN_mod(&r, &a, &b, pctx))
if (!BN_mod(r.get(), a.getc(), b.getc(), pctx))
throw bignum_error("CBigNum::operator% : BN_div failed");
return r;
}
@@ -580,7 +562,7 @@ inline const CBigNum operator%(const CBigNum& a, const CBigNum& b)
inline const CBigNum operator<<(const CBigNum& a, unsigned int shift)
{
CBigNum r;
if (!BN_lshift(&r, &a, shift))
if (!BN_lshift(r.get(), a.getc(), shift))
throw bignum_error("CBigNum:operator<< : BN_lshift failed");
return r;
}
@@ -592,11 +574,11 @@ inline const CBigNum operator>>(const CBigNum& a, unsigned int shift)
return r;
}
inline bool operator==(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) == 0); }
inline bool operator!=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) != 0); }
inline bool operator<=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) <= 0); }
inline bool operator>=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) >= 0); }
inline bool operator<(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) < 0); }
inline bool operator>(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) > 0); }
inline bool operator==(const CBigNum& a, const CBigNum& b) { return (BN_cmp(a.getc(), b.getc()) == 0); }
inline bool operator!=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(a.getc(), b.getc()) != 0); }
inline bool operator<=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(a.getc(), b.getc()) <= 0); }
inline bool operator>=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(a.getc(), b.getc()) >= 0); }
inline bool operator<(const CBigNum& a, const CBigNum& b) { return (BN_cmp(a.getc(), b.getc()) < 0); }
inline bool operator>(const CBigNum& a, const CBigNum& b) { return (BN_cmp(a.getc(), b.getc()) > 0); }
#endif
+273
View File
@@ -0,0 +1,273 @@
/**
* Blake2-B CUDA Implementation
*
* tpruvot@github July 2016
*
*/
#include <miner.h>
#include <string.h>
#include <stdint.h>
#include <sph/blake2b.h>
#include <cuda_helper.h>
#include <cuda_vector_uint2x4.h>
#define TPB 512
#define NBN 2
static uint32_t *d_resNonces[MAX_GPUS];
__device__ uint64_t d_data[10];
static __constant__ const int8_t blake2b_sigma[12][16] = {
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 } ,
{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 } ,
{ 11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4 } ,
{ 7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8 } ,
{ 9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13 } ,
{ 2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9 } ,
{ 12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11 } ,
{ 13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10 } ,
{ 6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5 } ,
{ 10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0 } ,
{ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 } ,
{ 14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3 }
};
// host mem align
#define A 64
extern "C" void blake2b_hash(void *output, const void *input)
{
uint8_t _ALIGN(A) hash[32];
blake2b_ctx ctx;
blake2b_init(&ctx, 32, NULL, 0);
blake2b_update(&ctx, input, 80);
blake2b_final(&ctx, hash);
memcpy(output, hash, 32);
}
// ----------------------------------------------------------------
__device__ __forceinline__
static void G(const int r, const int i, uint64_t &a, uint64_t &b, uint64_t &c, uint64_t &d, uint64_t const m[16])
{
a = a + b + m[ blake2b_sigma[r][2*i] ];
((uint2*)&d)[0] = SWAPUINT2( ((uint2*)&d)[0] ^ ((uint2*)&a)[0] );
c = c + d;
((uint2*)&b)[0] = ROR24( ((uint2*)&b)[0] ^ ((uint2*)&c)[0] );
a = a + b + m[ blake2b_sigma[r][2*i+1] ];
((uint2*)&d)[0] = ROR16( ((uint2*)&d)[0] ^ ((uint2*)&a)[0] );
c = c + d;
((uint2*)&b)[0] = ROR2( ((uint2*)&b)[0] ^ ((uint2*)&c)[0], 63U);
}
#define ROUND(r) \
G(r, 0, v[0], v[4], v[ 8], v[12], m); \
G(r, 1, v[1], v[5], v[ 9], v[13], m); \
G(r, 2, v[2], v[6], v[10], v[14], m); \
G(r, 3, v[3], v[7], v[11], v[15], m); \
G(r, 4, v[0], v[5], v[10], v[15], m); \
G(r, 5, v[1], v[6], v[11], v[12], m); \
G(r, 6, v[2], v[7], v[ 8], v[13], m); \
G(r, 7, v[3], v[4], v[ 9], v[14], m);
__global__
//__launch_bounds__(128, 8) /* to force 64 regs */
void blake2b_gpu_hash(const uint32_t threads, const uint32_t startNonce, uint32_t *resNonce, const uint2 target2)
{
const uint32_t nonce = (blockDim.x * blockIdx.x + threadIdx.x) + startNonce;
uint64_t m[16];
m[0] = d_data[0];
m[1] = d_data[1];
m[2] = d_data[2];
m[3] = d_data[3];
m[4] = d_data[4];
m[5] = d_data[5];
m[6] = d_data[6];
m[7] = d_data[7];
m[8] = d_data[8];
((uint32_t*)m)[18] = AS_U32(&d_data[9]);
((uint32_t*)m)[19] = nonce;
m[10] = m[11] = 0;
m[12] = m[13] = 0;
m[14] = m[15] = 0;
uint64_t v[16] = {
0x6a09e667f2bdc928, 0xbb67ae8584caa73b, 0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
0x510e527fade682d1, 0x9b05688c2b3e6c1f, 0x1f83d9abfb41bd6b, 0x5be0cd19137e2179,
0x6a09e667f3bcc908, 0xbb67ae8584caa73b, 0x3c6ef372fe94f82b, 0xa54ff53a5f1d36f1,
0x510e527fade68281, 0x9b05688c2b3e6c1f, 0xe07c265404be4294, 0x5be0cd19137e2179
};
ROUND( 0);
ROUND( 1);
ROUND( 2);
ROUND( 3);
ROUND( 4);
ROUND( 5);
ROUND( 6);
ROUND( 7);
ROUND( 8);
ROUND( 9);
ROUND(10);
ROUND(11);
uint2 last = vectorize(v[3] ^ v[11] ^ 0xa54ff53a5f1d36f1);
if (last.y <= target2.y && last.x <= target2.x) {
resNonce[1] = resNonce[0];
resNonce[0] = nonce;
}
}
__host__
uint32_t blake2b_hash_cuda(const int thr_id, const uint32_t threads, const uint32_t startNonce, const uint2 target2, uint32_t &secNonce)
{
uint32_t resNonces[NBN] = { UINT32_MAX, UINT32_MAX };
uint32_t result = UINT32_MAX;
dim3 grid((threads + TPB-1)/TPB);
dim3 block(TPB);
/* Check error on Ctrl+C or kill to prevent segfaults on exit */
if (cudaMemset(d_resNonces[thr_id], 0xff, NBN*sizeof(uint32_t)) != cudaSuccess)
return result;
blake2b_gpu_hash <<<grid, block, 8>>> (threads, startNonce, d_resNonces[thr_id], target2);
cudaThreadSynchronize();
if (cudaSuccess == cudaMemcpy(resNonces, d_resNonces[thr_id], NBN*sizeof(uint32_t), cudaMemcpyDeviceToHost)) {
result = resNonces[0];
secNonce = resNonces[1];
if (secNonce == result) secNonce = UINT32_MAX;
}
return result;
}
__host__
void blake2b_setBlock(uint32_t *data)
{
CUDA_SAFE_CALL(cudaMemcpyToSymbol(d_data, data, 80, 0, cudaMemcpyHostToDevice));
}
static bool init[MAX_GPUS] = { 0 };
int scanhash_blake2b(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(A) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int dev_id = device_map[thr_id];
int intensity = (device_sm[dev_id] >= 500 && !is_windows()) ? 28 : 25;
if (device_sm[dev_id] >= 520 && is_windows()) intensity = 26;
if (device_sm[dev_id] < 350) intensity = 22;
uint32_t throughput = cuda_default_throughput(thr_id, 1U << intensity);
if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
if (!init[thr_id])
{
cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage (linux)
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonces[thr_id], NBN * sizeof(uint32_t)), -1);
init[thr_id] = true;
}
for (int i=0; i < 20; i++)
be32enc(&endiandata[i], pdata[i]);
const uint2 target = make_uint2(ptarget[6], ptarget[7]);
blake2b_setBlock(endiandata);
do {
work->nonces[0] = blake2b_hash_cuda(thr_id, throughput, pdata[19], target, work->nonces[1]);
*hashes_done = pdata[19] - first_nonce + throughput;
if (work->nonces[0] != UINT32_MAX)
{
const uint32_t Htarg = ptarget[7];
uint32_t _ALIGN(A) vhash[8];
work->valid_nonces = 0;
endiandata[19] = work->nonces[0];
blake2b_hash(vhash, endiandata);
if (vhash[7] <= Htarg && fulltest(vhash, ptarget)) {
work_set_target_ratio(work, vhash);
work->valid_nonces++;
pdata[19] = work->nonces[0] + 1;
} else {
gpu_increment_reject(thr_id);
}
if (work->nonces[1] != UINT32_MAX) {
endiandata[19] = work->nonces[1];
blake2b_hash(vhash, endiandata);
if (vhash[7] <= Htarg && fulltest(vhash, ptarget)) {
if (bn_hash_target_ratio(vhash, ptarget) > work->shareratio[0]) {
work->sharediff[1] = work->sharediff[0];
work->shareratio[1] = work->shareratio[0];
xchg(work->nonces[1], work->nonces[0]);
work_set_target_ratio(work, vhash);
} else {
bn_set_target_ratio(work, vhash, 1);
}
work->valid_nonces++;
pdata[19] = max(work->nonces[0], work->nonces[1]) + 1; // next scan start
} else {
gpu_increment_reject(thr_id);
}
}
if (work->valid_nonces) {
work->nonces[0] = cuda_swab32(work->nonces[0]);
work->nonces[1] = cuda_swab32(work->nonces[1]);
return work->valid_nonces;
}
}
if ((uint64_t) throughput + pdata[19] >= max_nonce) {
pdata[19] = max_nonce;
break;
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart);
*hashes_done = pdata[19] - first_nonce;
return 0;
}
// cleanup
extern "C" void free_blake2b(int thr_id)
{
if (!init[thr_id])
return;
//cudaThreadSynchronize();
cudaFree(d_resNonces[thr_id]);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+6 -5
View File
@@ -2,18 +2,19 @@
"_comment1" : "Possible keys are the long options (ccminer --help)",
"_comment2" : "todo: support /* comments */",
"algo" : "blake",
"intensity": 19,
"algo" : "lbry",
"intensity": 22,
"api-bind": "127.0.0.1:4068",
"statsavg": 20,
"max-log-rate": 60,
"quiet" : false,
"debug" : false,
"protocol" : false,
"cpu-priority" : 3,
"url" : "stratum+tcp://stratum.eu.miners-pool.eu:8421",
"user" : "ShGCAFrspnLDo414EzUyJm4k5Qr2Kkrmvh",
"pass" : "x"
"url" : "stratum+tcp://yiimp.ccminer.org:3334",
"user" : "bK2wcSFJv2nnNjFvTN5Q7VX8X8unJktJsa",
"pass" : "donate"
}
+925 -135
View File
File diff suppressed because it is too large. Load diff
+157 -58
View File
@@ -23,29 +23,23 @@
<RootNamespace>ccminer</RootNamespace>
</PropertyGroup>
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.Default.props" />
<PropertyGroup Condition="'$(Configuration)'=='Release'" Label="Configuration">
<PropertyGroup Label="Configuration">
<ConfigurationType>Application</ConfigurationType>
<UseDebugLibraries>false</UseDebugLibraries>
<CharacterSet>MultiByte</CharacterSet>
<PlatformToolset>v120</PlatformToolset>
<CLRSupport>false</CLRSupport>
<LinkIncremental>false</LinkIncremental>
</PropertyGroup>
<PropertyGroup Condition="'$(Configuration)'=='Release'" Label="Configuration">
<UseDebugLibraries>false</UseDebugLibraries>
<WholeProgramOptimization>true</WholeProgramOptimization>
</PropertyGroup>
<PropertyGroup Condition="'$(Configuration)'=='Debug'" Label="Configuration">
<ConfigurationType>Application</ConfigurationType>
<UseDebugLibraries>true</UseDebugLibraries>
<CharacterSet>MultiByte</CharacterSet>
<PlatformToolset>v120</PlatformToolset>
<CLRSupport>false</CLRSupport>
<LinkIncremental>false</LinkIncremental>
</PropertyGroup>
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.props" />
<ImportGroup Label="ExtensionSettings" Condition="'$(Platform)'=='Win32'">
<Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 6.5.props" />
</ImportGroup>
<ImportGroup Label="ExtensionSettings" Condition="'$(Platform)'=='x64'">
<Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 7.5.props" />
<ImportGroup Label="ExtensionSettings">
<Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 9.0.props" />
</ImportGroup>
<ImportGroup Label="PropertySheets" Condition="'$(Configuration)|$(Platform)'=='Debug|Win32'">
<Import Project="$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props" Condition="exists('$(UserRootDir)\Microsoft.Cpp.$(Platform).user.props')" Label="LocalAppDataPlatform" />
@@ -118,7 +112,7 @@
<MaxRegCount>80</MaxRegCount>
<PtxAsOptionV>true</PtxAsOptionV>
<Keep>true</Keep>
<CodeGeneration>compute_50,sm_50</CodeGeneration>
<CodeGeneration>compute_52,sm_52</CodeGeneration>
<Include>$(NVTOOLSEXT_PATH)\include;..\..\..\Common\C99</Include>
<TargetMachinePlatform>64</TargetMachinePlatform>
</CudaCompile>
@@ -161,8 +155,8 @@
<MaxRegCount>80</MaxRegCount>
<PtxAsOptionV>true</PtxAsOptionV>
<Keep>true</Keep>
<CodeGeneration>compute_50,sm_50;compute_52,sm_52;compute_30,sm_30;compute_20,sm_21</CodeGeneration>
<AdditionalOptions>--ptxas-options="-O2" %(AdditionalOptions)</AdditionalOptions>
<CodeGeneration>compute_50,sm_50;compute_52,sm_52;compute_30,sm_30</CodeGeneration>
<AdditionalOptions>--ptxas-options="-O2" --Wno-deprecated-gpu-targets %(AdditionalOptions)</AdditionalOptions>
<Optimization>O2</Optimization>
</CudaCompile>
<CudaLink>
@@ -204,10 +198,11 @@
<MaxRegCount>80</MaxRegCount>
<PtxAsOptionV>true</PtxAsOptionV>
<Keep>true</Keep>
<CodeGeneration>compute_52,sm_52;compute_50,sm_50;compute_35,sm_35;compute_30,sm_30;compute_20,sm_21</CodeGeneration>
<Include>$(NVTOOLSEXT_PATH)\include;..\..\..\Common\C99</Include>
<CodeGeneration>compute_52,sm_52;compute_50,sm_50;compute_35,sm_35;compute_30,sm_30</CodeGeneration>
<Include>$(NVTOOLSEXT_PATH)\include</Include>
<Optimization>O3</Optimization>
<TargetMachinePlatform>64</TargetMachinePlatform>
<AdditionalOptions>--Wno-deprecated-gpu-targets %(AdditionalOptions)</AdditionalOptions>
</CudaCompile>
<CudaLink>
<Optimization>O3</Optimization>
@@ -227,11 +222,28 @@
<ClCompile Include="compat\jansson\strconv.c" />
<ClCompile Include="compat\jansson\utf.c" />
<ClCompile Include="compat\jansson\value.c" />
<ClInclude Include="compat\pthreads\pthread.h" />
<ClCompile Include="compat\winansi.c" />
<ClCompile Include="ccminer.cpp">
<TreatWChar_tAsBuiltInType>false</TreatWChar_tAsBuiltInType>
<Optimization Condition="'$(Configuration)'=='Release'">Full</Optimization>
</ClCompile>
<ClCompile Include="crypto\mman.c" />
<ClCompile Include="crypto\wildkeccak-cpu.cpp" />
<ClCompile Include="crypto\xmr-rpc.cpp" />
<ClCompile Include="crypto\aesb.cpp" />
<ClCompile Include="crypto\oaes_lib.cpp" />
<ClCompile Include="crypto\cryptolight-cpu.cpp" />
<ClCompile Include="crypto\cryptonight-cpu.cpp" />
<ClCompile Include="crypto\cpu\c_keccak.c" />
<ClCompile Include="equi\blake2\blake2bx.cpp">
<EnableEnhancedInstructionSet Condition="'$(Platform)'=='Win32'">StreamingSIMDExtensions</EnableEnhancedInstructionSet>
</ClCompile>
<ClCompile Include="equi\equi-stratum.cpp" />
<ClCompile Include="equi\equi.cpp" />
<ClCompile Include="equi\equihash.cpp" />
<ClCompile Include="nvapi.cpp" />
<ClCompile Include="nvsettings.cpp" />
<ClCompile Include="pools.cpp" />
<ClCompile Include="util.cpp" />
<ClCompile Include="bench.cpp" />
@@ -249,22 +261,55 @@
<ClCompile Include="myriadgroestl.cpp" />
<ClCompile Include="lyra2\Lyra2.c" />
<ClCompile Include="lyra2\Sponge.c" />
<ClInclude Include="lyra2\cuda_lyra2_sm2.cuh" />
<ClCompile Include="lyra2\Lyra2Z.c" />
<ClInclude Include="equi\eqcuda.hpp" />
<ClInclude Include="equi\equihash.h" />
<ClInclude Include="neoscrypt\neoscrypt.h" />
<ClCompile Include="neoscrypt\neoscrypt.cpp" />
<ClCompile Include="neoscrypt\neoscrypt-cpu.c" />
<ClInclude Include="neoscrypt\cuda_vectors.h" />
<ClInclude Include="x11\cuda_x11_simd512_sm2.cuh" />
<ClInclude Include="x16\cuda_x16.h" />
<CudaCompile Include="Algo256\bmw.cu" />
<CudaCompile Include="Algo256\cuda_bmw.cu">
<MaxRegCount>76</MaxRegCount>
</CudaCompile>
<CudaCompile Include="neoscrypt\cuda_neoscrypt.cu" />
<CudaCompile Include="crypto\cryptolight.cu" />
<CudaCompile Include="crypto\cryptolight-core.cu">
<MaxRegCount>64</MaxRegCount>
</CudaCompile>
<CudaCompile Include="crypto\cryptonight.cu" />
<CudaCompile Include="crypto\cryptonight-core.cu">
<MaxRegCount>64</MaxRegCount>
</CudaCompile>
<CudaCompile Include="crypto\cryptonight-extra.cu">
<MaxRegCount>255</MaxRegCount>
</CudaCompile>
<ClInclude Include="crypto\cn_aes.cuh" />
<ClInclude Include="crypto\cn_blake.cuh" />
<ClInclude Include="crypto\cn_groestl.cuh" />
<ClInclude Include="crypto\cn_jh.cuh" />
<ClInclude Include="crypto\cn_keccak.cuh" />
<ClInclude Include="crypto\cn_skein.cuh" />
<CudaCompile Include="crypto\wildkeccak.cu">
<MaxRegCount>128</MaxRegCount>
</CudaCompile>
<CudaCompile Include="equi\cuda_equi.cu">
<CodeGeneration>compute_52,sm_52;compute_50,sm_50;compute_30,sm_30</CodeGeneration>
<AdditionalOptions> -Xptxas -dlcm=ca -Xptxas -dscm=cs %(AdditionalOptions)</AdditionalOptions>
<MaxRegCount>0</MaxRegCount>
<CodeGeneration Condition="'$(Configuration)|$(Platform)'=='Release|x64'">compute_61,sm_61;compute_52,sm_52;compute_50,sm_50;compute_30,sm_30</CodeGeneration>
</CudaCompile>
<CudaCompile Include="neoscrypt\cuda_neoscrypt.cu">
<MaxRegCount>160</MaxRegCount>
</CudaCompile>
<ClCompile Include="scrypt-jane.cpp" />
<ClCompile Include="scrypt.cpp" />
<ClCompile Include="sia\sia-rpc.cpp" />
<ClCompile Include="skein2.cpp" />
<ClCompile Include="sph\aes_helper.c" />
<ClCompile Include="sph\blake.c" />
<ClCompile Include="sph\blake2b.c" />
<ClCompile Include="sph\blake2s.c" />
<ClCompile Include="sph\bmw.c" />
<ClCompile Include="sph\cubehash.c" />
@@ -272,15 +317,11 @@
<ClCompile Include="sph\fugue.c" />
<ClCompile Include="sph\groestl.c" />
<ClCompile Include="sph\haval.c" />
<ClCompile Include="sph\haval_helper.c">
<ExcludedFromBuild>true</ExcludedFromBuild>
</ClCompile>
<ClCompile Include="sph\jh.c" />
<ClCompile Include="sph\keccak.c" />
<ClCompile Include="sph\luffa.c" />
<ClCompile Include="sph\md_helper.c">
<ExcludedFromBuild>true</ExcludedFromBuild>
</ClCompile>
<ClCompile Include="sph\ripemd.c" />
<ClCompile Include="sph\sph_sha2.c" />
<ClCompile Include="sph\sha2.c" />
<ClCompile Include="sph\sha2big.c" />
<ClCompile Include="sph\shabal.c" />
@@ -291,6 +332,7 @@
<ClCompile Include="sph\hamsi.c" />
<ClCompile Include="sph\hamsi_helper.c" />
<ClCompile Include="sph\whirlpool.c" />
<ClCompile Include="x13\sm3.c" />
</ItemGroup>
<ItemGroup>
<ClInclude Include="compat.h" />
@@ -310,7 +352,12 @@
<ClInclude Include="compat\sys\time.h" />
<ClInclude Include="compat\unistd.h" />
<ClInclude Include="compat\winansi.h" />
<ClInclude Include="cpuminer-config.h" />
<ClInclude Include="compat\ccminer-config.h" />
<ClInclude Include="crypto\cryptolight.h" />
<ClInclude Include="crypto\cryptonight.h" />
<ClInclude Include="crypto\mman.h" />
<ClInclude Include="crypto\wildkeccak.h" />
<ClInclude Include="crypto\xmr-rpc.h" />
<ClInclude Include="cuda_groestlcoin.h" />
<ClInclude Include="cuda_helper.h" />
<ClInclude Include="cuda_vector_uint2x4.h" />
@@ -321,10 +368,13 @@
<ClInclude Include="miner.h" />
<ClInclude Include="nvml.h" />
<ClInclude Include="quark\cuda_bmw512_sm3.cuh" />
<ClInclude Include="quark\cuda_quark_groestl512_sm2.cuh" />
<ClInclude Include="quark\cuda_quark_blake512_sp.cuh" />
<ClInclude Include="quark\cuda_skein512_sp.cuh" />
<ClInclude Include="res\resource.h" />
<ClInclude Include="sia\sia-rpc.h" />
<ClInclude Include="scrypt\salsa_kernel.h" />
<ClInclude Include="sph\blake2b.h" />
<ClInclude Include="sph\blake2s.h" />
<ClInclude Include="sph\sph_blake.h" />
<ClInclude Include="sph\sph_bmw.h" />
<ClInclude Include="sph\sph_cubehash.h" />
@@ -347,7 +397,7 @@
<ClInclude Include="uint256.h" />
<ClInclude Include="lyra2\Lyra2.h" />
<ClInclude Include="lyra2\Sponge.h" />
<ClInclude Include="lyra2\cuda_lyra2v2_sm3.cuh" />
<ClInclude Include="lyra2\Lyra2Z.h" />
<ClInclude Include="quark\groestl_transf_quad.h" />
<ClInclude Include="quark\groestl_functions_quad.h" />
<ClInclude Include="quark\cuda_quark.h" />
@@ -357,16 +407,14 @@
<CudaCompile Include="cuda.cpp" />
<CudaCompile Include="cuda_groestlcoin.cu" />
<CudaCompile Include="cuda_myriadgroestl.cu" />
<CudaCompile Include="cuda_nist5.cu">
</CudaCompile>
<CudaCompile Include="scrypt\blake.cu" />
<CudaCompile Include="scrypt\keccak.cu" />
<CudaCompile Include="scrypt\sha256.cu" />
<CudaCompile Include="scrypt\salsa_kernel.cu">
<CodeGeneration>compute_20,sm_21</CodeGeneration>
<CodeGeneration>compute_30,sm_30</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\fermi_kernel.cu">
<CodeGeneration>compute_20,sm_21</CodeGeneration>
<CodeGeneration>compute_30,sm_30</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\kepler_kernel.cu">
<CodeGeneration>compute_30,sm_30</CodeGeneration>
@@ -378,11 +426,17 @@
<CodeGeneration>compute_35,sm_35;compute_50,sm_50;compute_52,sm_52</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\test_kernel.cu">
<CodeGeneration>compute_20,sm_21</CodeGeneration>
<CodeGeneration>compute_30,sm_30</CodeGeneration>
</CudaCompile>
<CudaCompile Include="scrypt\titan_kernel.cu">
<CodeGeneration>compute_35,sm_35;compute_50,sm_50</CodeGeneration>
</CudaCompile>
<CudaCompile Include="sha256\cuda_sha256d.cu" />
<CudaCompile Include="sha256\sha256d.cu" />
<CudaCompile Include="sha256\cuda_sha256t.cu" />
<CudaCompile Include="sha256\sha256t.cu" />
<CudaCompile Include="sha256\cuda_sha256q.cu" />
<CudaCompile Include="sha256\sha256q.cu" />
<CudaCompile Include="zr5.cu" />
<CudaCompile Include="heavy\cuda_blake512.cu">
</CudaCompile>
@@ -399,14 +453,16 @@
</CudaCompile>
<CudaCompile Include="heavy\heavy.cu">
</CudaCompile>
<CudaCompile Include="heavy\bastion.cu" />
<CudaCompile Include="heavy\cuda_bastion.cu" />
<CudaCompile Include="JHA\jha.cu" />
<CudaCompile Include="JHA\jackpotcoin.cu" />
<CudaCompile Include="JHA\cuda_jha_compactionTest.cu">
<AdditionalOptions Condition="'$(Configuration)'=='Release'">-Xptxas "-abi=yes" %(AdditionalOptions)</AdditionalOptions>
<AdditionalOptions Condition="'$(Configuration)'=='Debug'">-Xptxas "-abi=yes" %(AdditionalOptions)</AdditionalOptions>
<AdditionalOptions>-Xptxas "-abi=yes" %(AdditionalOptions)</AdditionalOptions>
</CudaCompile>
<CudaCompile Include="JHA\cuda_jha_keccak512.cu">
</CudaCompile>
<CudaCompile Include="JHA\jackpotcoin.cu">
</CudaCompile>
<CudaCompile Include="blake2b.cu" />
<CudaCompile Include="Algo256\blake256.cu">
<MaxRegCount>64</MaxRegCount>
<AdditionalOptions Condition="'$(Configuration)'=='Release'">--ptxas-options="-dlcm=cg" %(AdditionalOptions)</AdditionalOptions>
@@ -423,10 +479,15 @@
<CudaCompile Include="Algo256\cuda_cubehash256.cu" />
<CudaCompile Include="Algo256\cuda_fugue256.cu" />
<CudaCompile Include="Algo256\cuda_groestl256.cu" />
<CudaCompile Include="Algo256\cuda_keccak256.cu">
<CudaCompile Include="Algo256\cuda_keccak256.cu" />
<CudaCompile Include="Algo256\cuda_keccak256_sm3.cu">
<MaxRegCount>92</MaxRegCount>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_skein256.cu" />
<CudaCompile Include="lbry\cuda_sha256_lbry.cu" />
<CudaCompile Include="lbry\cuda_sha512_lbry.cu" />
<CudaCompile Include="lbry\cuda_lbry_merged.cu" />
<CudaCompile Include="lbry\lbry.cu" />
<CudaCompile Include="pentablake.cu">
<MaxRegCount>80</MaxRegCount>
<AdditionalOptions Condition="'$(Configuration)'=='Release'">--ptxas-options="-dlcm=cg" %(AdditionalOptions)</AdditionalOptions>
@@ -455,26 +516,49 @@
</CudaCompile>
<CudaCompile Include="quark\cuda_skein512.cu">
</CudaCompile>
<CudaCompile Include="quark\quarkcoin.cu">
</CudaCompile>
<CudaCompile Include="qubit\deep.cu">
</CudaCompile>
<CudaCompile Include="qubit\luffa.cu">
</CudaCompile>
<CudaCompile Include="qubit\qubit.cu">
</CudaCompile>
<CudaCompile Include="qubit\qubit_luffa512.cu">
</CudaCompile>
<CudaCompile Include="quark\quarkcoin.cu" />
<CudaCompile Include="quark\nist5.cu" />
<CudaCompile Include="qubit\deep.cu" />
<CudaCompile Include="qubit\luffa.cu" />
<CudaCompile Include="qubit\qubit.cu" />
<CudaCompile Include="qubit\qubit_luffa512.cu" />
<CudaCompile Include="lyra2\allium.cu" />
<CudaCompile Include="lyra2\lyra2RE.cu" />
<CudaCompile Include="lyra2\cuda_lyra2.cu" />
<CudaCompile Include="lyra2\lyra2REv2.cu" />
<CudaCompile Include="lyra2\cuda_lyra2v2.cu" />
<ClInclude Include="lyra2\cuda_lyra2_sm2.cuh" />
<ClInclude Include="lyra2\cuda_lyra2_sm5.cuh" />
<ClInclude Include="lyra2\cuda_lyra2v2_sm3.cuh" />
<CudaCompile Include="lyra2\lyra2REv3.cu" />
<CudaCompile Include="lyra2\cuda_lyra2v3.cu" />
<ClInclude Include="lyra2\cuda_lyra2v3_sm3.cuh" />
<CudaCompile Include="lyra2\lyra2Z.cu" />
<CudaCompile Include="lyra2\cuda_lyra2Z.cu" />
<ClInclude Include="lyra2\cuda_lyra2Z_sm5.cuh" />
<CudaCompile Include="polytimos.cu" />
<CudaCompile Include="sia\sia.cu" />
<CudaCompile Include="skein.cu">
<MaxRegCount>64</MaxRegCount>
</CudaCompile>
<CudaCompile Include="cuda_skeincoin.cu">
<MaxRegCount>48</MaxRegCount>
</CudaCompile>
<CudaCompile Include="phi\phi.cu" />
<CudaCompile Include="phi\phi2.cu" />
<CudaCompile Include="phi\cuda_phi2.cu" />
<CudaCompile Include="phi\cuda_phi2_cubehash512.cu" />
<CudaCompile Include="skunk\skunk.cu" />
<CudaCompile Include="skunk\cuda_skunk.cu">
<CodeGeneration>compute_50,sm_50;compute_52,sm_52</CodeGeneration>
<MaxRegCount>64</MaxRegCount>
</CudaCompile>
<CudaCompile Include="skunk\cuda_skunk_streebog.cu" />
<ClInclude Include="tribus\cuda_echo512_aes.cuh" />
<CudaCompile Include="tribus\cuda_echo512_final.cu">
<CodeGeneration>compute_50,sm_50;compute_52,sm_52</CodeGeneration>
</CudaCompile>
<CudaCompile Include="tribus\tribus.cu" />
<ClInclude Include="x11\cuda_x11_aes.cuh" />
<CudaCompile Include="x11\cuda_x11_cubehash512.cu" />
<CudaCompile Include="x11\cuda_x11_echo.cu">
@@ -491,26 +575,44 @@
<MaxRegCount>64</MaxRegCount>
</CudaCompile>
<CudaCompile Include="x11\cuda_streebog.cu" />
<CudaCompile Include="x11\cuda_streebog_maxwell.cu" />
<CudaCompile Include="x11\c11.cu" />
<CudaCompile Include="x11\fresh.cu" />
<CudaCompile Include="x11\sib.cu" />
<CudaCompile Include="x11\s3.cu" />
<CudaCompile Include="x11\timetravel.cu" />
<CudaCompile Include="x11\bitcore.cu" />
<CudaCompile Include="x11\exosis.cu" />
<CudaCompile Include="x11\veltor.cu" />
<CudaCompile Include="x11\x11.cu" />
<CudaCompile Include="x11\x11evo.cu" />
<CudaCompile Include="x12\x12.cu" />
<CudaCompile Include="x13\cuda_x13_hamsi512.cu">
<MaxRegCount>72</MaxRegCount>
</CudaCompile>
<CudaCompile Include="x13\cuda_x13_fugue512.cu">
</CudaCompile>
<CudaCompile Include="x13\x13.cu">
</CudaCompile>
<CudaCompile Include="x15\x14.cu">
</CudaCompile>
<CudaCompile Include="x13\x13.cu" />
<CudaCompile Include="x13\cuda_hsr_sm3.cu" />
<CudaCompile Include="x13\hsr.cu" />
<CudaCompile Include="x15\x14.cu" />
<CudaCompile Include="x15\cuda_x14_shabal512.cu" />
<CudaCompile Include="x15\cuda_x15_whirlpool.cu" />
<CudaCompile Include="x15\x15.cu" />
<CudaCompile Include="x15\whirlpool.cu" />
<CudaCompile Include="x17\x17.cu">
<CudaCompile Include="x15\cuda_x15_whirlpool_sm3.cu" />
<CudaCompile Include="x16\x16r.cu" />
<CudaCompile Include="x16\x16s.cu" />
<CudaCompile Include="x16\cuda_x16_echo512.cu" />
<CudaCompile Include="x16\cuda_x16_fugue512.cu" />
<CudaCompile Include="x16\cuda_x16_shabal512.cu" />
<CudaCompile Include="x16\cuda_x16_simd512_80.cu" />
<CudaCompile Include="x16\cuda_x16_echo512_64.cu">
<CodeGeneration>compute_50,sm_50;compute_52,sm_52</CodeGeneration>
</CudaCompile>
<CudaCompile Include="x17\hmq17.cu" />
<CudaCompile Include="x17\sonoa.cu" />
<CudaCompile Include="x17\x17.cu" />
<CudaCompile Include="x17\cuda_x17_haval256.cu">
</CudaCompile>
<CudaCompile Include="x17\cuda_x17_sha512.cu">
@@ -527,11 +629,8 @@
<Text Include="README.txt" />
</ItemGroup>
<Import Project="$(VCTargetsPath)\Microsoft.Cpp.targets" />
<ImportGroup Label="ExtensionTargets" Condition="'$(Platform)'=='Win32'">
<Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 6.5.targets" />
</ImportGroup>
<ImportGroup Label="ExtensionTargets" Condition="'$(Platform)'=='x64'">
<Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 7.5.targets" />
<ImportGroup Label="ExtensionTargets">
<Import Project="$(VCTargetsPath)\BuildCustomizations\CUDA 9.0.targets" />
</ImportGroup>
<!-- Copy the required dlls -->
<Target Name="AfterBuild">
+350 -17
View File
@@ -46,6 +46,9 @@
<Filter Include="Source Files\CUDA\qubit">
<UniqueIdentifier>{f3ed23a2-8ce7-41a5-b051-6da56047dc35}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\skunk">
<UniqueIdentifier>{23ed23a2-8e7a-65a2-b051-8aa63047d352}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\x11">
<UniqueIdentifier>{dd751f2d-bfd6-42c1-8f9b-cbe94e539353}</UniqueIdentifier>
</Filter>
@@ -55,6 +58,9 @@
<Filter Include="Source Files\CUDA\x15">
<UniqueIdentifier>{a2403c22-6777-46ab-a55a-3fcc7386c974}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\x16">
<UniqueIdentifier>{55dfae6a-66ba-43e2-8ceb-98ee70cbdf16}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\x17">
<UniqueIdentifier>{85dfae6a-66ca-4332-8cec-98ee70cbdf2f}</UniqueIdentifier>
</Filter>
@@ -82,6 +88,39 @@
<Filter Include="Source Files\CUDA\lyra2">
<UniqueIdentifier>{1613763f-895c-4321-b58b-6f5849868956}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\lbry">
<UniqueIdentifier>{3079ea1f-f768-455a-acd6-f517fac535b4}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\sha256">
<UniqueIdentifier>{86a896c0-1688-4854-98e3-285d166069a3}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\sia">
<UniqueIdentifier>{86a896c0-1688-4854-98e0-285d166069a3}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\crypto">
<UniqueIdentifier>{fea0fce3-c0fe-42f7-aa37-0cbba10b008a}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\crypto\xmr">
<UniqueIdentifier>{af52b078-ed91-4c6e-b07a-e9243acc85d2}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\crypto\bbr">
<UniqueIdentifier>{af387eac-e9e6-4e91-a5e8-637b1e7a8d93}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\xmr">
<UniqueIdentifier>{0f9aec5e-5409-488f-992a-2c108590d1ac}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\equi">
<UniqueIdentifier>{031afae7-2a78-4e32-9738-4b589b6f7ff3}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\tribus">
<UniqueIdentifier>{1e548d79-c217-4203-989a-a592fe2b2de3}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\phi">
<UniqueIdentifier>{311e8d79-1612-4f0f-8591-23a592f2b2d3}</UniqueIdentifier>
</Filter>
<Filter Include="Source Files\CUDA\x12">
<UniqueIdentifier>{xde48d89-fx12-1323-129a-b592fe2b2de3}</UniqueIdentifier>
</Filter>
</ItemGroup>
<ItemGroup>
<ClCompile Include="compat\jansson\dump.c">
@@ -126,6 +165,12 @@
<ClCompile Include="sph\blake.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\blake2b.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\blake2s.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\bmw.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
@@ -150,9 +195,6 @@
<ClCompile Include="sph\luffa.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\md_helper.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\sha2.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
@@ -180,7 +222,10 @@
<ClCompile Include="sph\haval.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\haval_helper.c">
<ClCompile Include="sph\ripemd.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\sph_sha2.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="sph\sha2big.c">
@@ -189,6 +234,9 @@
<ClCompile Include="sph\streebog.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="x13\sm3.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="compat\winansi.c">
<Filter>Source Files</Filter>
</ClCompile>
@@ -219,12 +267,18 @@
<ClCompile Include="compat\jansson\error.c">
<Filter>Source Files\jansson</Filter>
</ClCompile>
<ClInclude Include="compat\pthreads\pthread.h">
<Filter>Header Files\compat</Filter>
</ClInclude>
<ClCompile Include="lyra2\Lyra2.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="lyra2\Sponge.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="lyra2\Lyra2Z.c">
<Filter>Source Files\sph</Filter>
</ClCompile>
<ClCompile Include="scrypt.cpp">
<Filter>Source Files\CUDA\scrypt</Filter>
</ClCompile>
@@ -258,8 +312,50 @@
<ClCompile Include="bignum.cpp">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="sph\blake2s.c">
<Filter>Source Files\sph</Filter>
<ClCompile Include="nvapi.cpp">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="nvsettings.cpp">
<Filter>Source Files</Filter>
</ClCompile>
<ClCompile Include="sia\sia-rpc.cpp">
<Filter>Source Files\sia</Filter>
</ClCompile>
<ClCompile Include="crypto\xmr-rpc.cpp">
<Filter>Source Files\crypto</Filter>
</ClCompile>
<ClCompile Include="crypto\aesb.cpp">
<Filter>Source Files\crypto\xmr</Filter>
</ClCompile>
<ClCompile Include="crypto\cpu\c_keccak.c">
<Filter>Source Files\crypto\xmr</Filter>
</ClCompile>
<ClCompile Include="crypto\oaes_lib.cpp">
<Filter>Source Files\crypto\xmr</Filter>
</ClCompile>
<ClCompile Include="crypto\cryptolight-cpu.cpp">
<Filter>Source Files\crypto\xmr</Filter>
</ClCompile>
<ClCompile Include="crypto\cryptonight-cpu.cpp">
<Filter>Source Files\crypto\xmr</Filter>
</ClCompile>
<ClCompile Include="crypto\mman.c">
<Filter>Source Files\crypto\bbr</Filter>
</ClCompile>
<ClCompile Include="crypto\wildkeccak-cpu.cpp">
<Filter>Source Files\crypto\bbr</Filter>
</ClCompile>
<ClCompile Include="equi\equi.cpp">
<Filter>Source Files\equi</Filter>
</ClCompile>
<ClCompile Include="equi\equihash.cpp">
<Filter>Source Files\equi</Filter>
</ClCompile>
<ClCompile Include="equi\equi-stratum.cpp">
<Filter>Source Files\equi</Filter>
</ClCompile>
<ClCompile Include="equi\blake2\blake2bx.cpp">
<Filter>Source Files\equi</Filter>
</ClCompile>
</ItemGroup>
<ItemGroup>
@@ -267,7 +363,7 @@
<Filter>Header Files</Filter>
</ClInclude>
<ClInclude Include="compat.h">
<Filter>Header Files</Filter>
<Filter>Header Files\compat</Filter>
</ClInclude>
<ClInclude Include="elist.h">
<Filter>Header Files</Filter>
@@ -290,8 +386,8 @@
<ClInclude Include="compat\stdbool.h">
<Filter>Header Files\compat</Filter>
</ClInclude>
<ClInclude Include="cpuminer-config.h">
<Filter>Header Files</Filter>
<ClInclude Include="compat\ccminer-config.h">
<Filter>Header Files\compat</Filter>
</ClInclude>
<ClInclude Include="uint256.h">
<Filter>Header Files</Filter>
@@ -305,6 +401,12 @@
<ClInclude Include="x11\cuda_x11.h">
<Filter>Header Files\CUDA</Filter>
</ClInclude>
<ClInclude Include="sph\blake2b.h">
<Filter>Header Files\sph</Filter>
</ClInclude>
<ClInclude Include="sph\blake2s.h">
<Filter>Header Files\sph</Filter>
</ClInclude>
<ClInclude Include="sph\sph_blake.h">
<Filter>Header Files\sph</Filter>
</ClInclude>
@@ -410,6 +512,9 @@
<ClInclude Include="lyra2\Lyra2.h">
<Filter>Header Files\lyra2</Filter>
</ClInclude>
<ClInclude Include="lyra2\Lyra2Z.h">
<Filter>Header Files\lyra2</Filter>
</ClInclude>
<ClInclude Include="lyra2\Sponge.h">
<Filter>Header Files\lyra2</Filter>
</ClInclude>
@@ -443,18 +548,69 @@
<ClInclude Include="lyra2\cuda_lyra2_sm2.cuh">
<Filter>Source Files\CUDA\lyra2</Filter>
</ClInclude>
<ClInclude Include="quark\cuda_quark_blake512_sp.cuh">
<Filter>Source Files\CUDA\quark</Filter>
<ClInclude Include="lyra2\cuda_lyra2_sm5.cuh">
<Filter>Source Files\CUDA\lyra2</Filter>
</ClInclude>
<ClInclude Include="quark\cuda_skein512_sp.cuh">
<ClInclude Include="lyra2\cuda_lyra2Z_sm5.cuh">
<Filter>Source Files\CUDA\lyra2</Filter>
</ClInclude>
<ClInclude Include="quark\cuda_quark_blake512_sp.cuh">
<Filter>Source Files\CUDA\quark</Filter>
</ClInclude>
<ClInclude Include="quark\cuda_bmw512_sm3.cuh">
<Filter>Source Files\CUDA\quark</Filter>
</ClInclude>
<ClInclude Include="quark\cuda_quark_groestl512_sm2.cuh">
<Filter>Source Files\CUDA\quark</Filter>
</ClInclude>
<ClInclude Include="x11\cuda_x11_simd512_sm2.cuh">
<Filter>Source Files\CUDA\x11</Filter>
</ClInclude>
<ClInclude Include="sia\sia-rpc.h">
<Filter>Source Files\sia</Filter>
</ClInclude>
<ClInclude Include="crypto\xmr-rpc.h">
<Filter>Source Files\crypto</Filter>
</ClInclude>
<ClInclude Include="crypto\mman.h">
<Filter>Source Files\crypto\bbr</Filter>
</ClInclude>
<ClInclude Include="crypto\wildkeccak.h">
<Filter>Source Files\crypto\bbr</Filter>
</ClInclude>
<ClInclude Include="crypto\cn_aes.cuh">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cn_blake.cuh">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cn_groestl.cuh">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cn_jh.cuh">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cn_keccak.cuh">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cn_skein.cuh">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cryptolight.h">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="crypto\cryptonight.h">
<Filter>Source Files\CUDA\xmr</Filter>
</ClInclude>
<ClInclude Include="equi\eqcuda.hpp">
<Filter>Source Files\equi</Filter>
</ClInclude>
<ClInclude Include="equi\equihash.h">
<Filter>Source Files\equi</Filter>
</ClInclude>
<ClInclude Include="x16\cuda_x16.h">
<Filter>Header Files\CUDA</Filter>
</ClInclude>
</ItemGroup>
<ItemGroup>
<CudaCompile Include="cuda.cpp">
@@ -466,6 +622,9 @@
<CudaCompile Include="JHA\cuda_jha_keccak512.cu">
<Filter>Source Files\CUDA\JHA</Filter>
</CudaCompile>
<CudaCompile Include="JHA\jha.cu">
<Filter>Source Files\CUDA\JHA</Filter>
</CudaCompile>
<CudaCompile Include="JHA\jackpotcoin.cu">
<Filter>Source Files\CUDA\JHA</Filter>
</CudaCompile>
@@ -508,6 +667,12 @@
<CudaCompile Include="heavy\heavy.cu">
<Filter>Source Files\CUDA\heavy</Filter>
</CudaCompile>
<CudaCompile Include="heavy\bastion.cu">
<Filter>Source Files\CUDA\heavy</Filter>
</CudaCompile>
<CudaCompile Include="heavy\cuda_bastion.cu">
<Filter>Source Files\CUDA\heavy</Filter>
</CudaCompile>
<CudaCompile Include="quark\cuda_bmw512.cu">
<Filter>Source Files\CUDA\quark</Filter>
</CudaCompile>
@@ -517,6 +682,9 @@
<CudaCompile Include="quark\cuda_quark_keccak512.cu">
<Filter>Source Files\CUDA\quark</Filter>
</CudaCompile>
<CudaCompile Include="quark\nist5.cu">
<Filter>Source Files\CUDA\quark</Filter>
</CudaCompile>
<CudaCompile Include="quark\quarkcoin.cu">
<Filter>Source Files\CUDA\quark</Filter>
</CudaCompile>
@@ -532,9 +700,6 @@
<CudaCompile Include="qubit\qubit_luffa512.cu">
<Filter>Source Files\CUDA\qubit</Filter>
</CudaCompile>
<CudaCompile Include="cuda_nist5.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<ClInclude Include="x11\cuda_x11_aes.cuh">
<Filter>Source Files\CUDA\x11</Filter>
</ClInclude>
@@ -568,6 +733,9 @@
<CudaCompile Include="x11\x11.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\x11evo.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x13\cuda_x13_fugue512.cu">
<Filter>Source Files\CUDA\x13</Filter>
</CudaCompile>
@@ -577,6 +745,12 @@
<CudaCompile Include="x13\x13.cu">
<Filter>Source Files\CUDA\x13</Filter>
</CudaCompile>
<CudaCompile Include="x13\cuda_hsr_sm3.cu">
<Filter>Source Files\CUDA\x13</Filter>
</CudaCompile>
<CudaCompile Include="x13\hsr.cu">
<Filter>Source Files\CUDA\x13</Filter>
</CudaCompile>
<CudaCompile Include="cuda_checkhash.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
@@ -601,21 +775,78 @@
<CudaCompile Include="x17\cuda_x17_sha512.cu">
<Filter>Source Files\CUDA\x17</Filter>
</CudaCompile>
<CudaCompile Include="x17\hmq17.cu">
<Filter>Source Files\CUDA\x17</Filter>
</CudaCompile>
<CudaCompile Include="x17\sonoa.cu">
<Filter>Source Files\CUDA\x17</Filter>
</CudaCompile>
<CudaCompile Include="x17\x17.cu">
<Filter>Source Files\CUDA\x17</Filter>
</CudaCompile>
<CudaCompile Include="pentablake.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<CudaCompile Include="polytimos.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<CudaCompile Include="phi\phi.cu">
<Filter>Source Files\CUDA\phi</Filter>
</CudaCompile>
<CudaCompile Include="phi\phi2.cu">
<Filter>Source Files\CUDA\phi</Filter>
</CudaCompile>
<CudaCompile Include="phi\cuda_phi2.cu">
<Filter>Source Files\CUDA\phi</Filter>
</CudaCompile>
<CudaCompile Include="phi\cuda_phi2_cubehash512.cu">
<Filter>Source Files\CUDA\phi</Filter>
</CudaCompile>
<CudaCompile Include="skunk\skunk.cu">
<Filter>Source Files\CUDA\skunk</Filter>
</CudaCompile>
<CudaCompile Include="skunk\cuda_skunk.cu">
<Filter>Source Files\CUDA\skunk</Filter>
</CudaCompile>
<CudaCompile Include="skunk\cuda_skunk_streebog.cu">
<Filter>Source Files\CUDA\skunk</Filter>
</CudaCompile>
<CudaCompile Include="tribus\tribus.cu">
<Filter>Source Files\CUDA\tribus</Filter>
</CudaCompile>
<CudaCompile Include="tribus\cuda_echo512_final.cu">
<Filter>Source Files\CUDA\tribus</Filter>
</CudaCompile>
<ClInclude Include="tribus\cuda_echo512_aes.cuh">
<Filter>Source Files\CUDA\tribus</Filter>
</ClInclude>
<CudaCompile Include="x11\sib.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\cuda_streebog.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\cuda_streebog_maxwell.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\s3.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x12\x12.cu">
<Filter>Source Files\CUDA\x12</Filter>
</CudaCompile>
<CudaCompile Include="x11\timetravel.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\bitcore.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\exosis.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="x11\veltor.cu">
<Filter>Source Files\CUDA\x11</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\blake256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
@@ -626,7 +857,7 @@
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\keccak256.cu">
<Filter>Source Files\CUDA</Filter>
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_blake256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
@@ -646,6 +877,9 @@
<CudaCompile Include="Algo256\cuda_keccak256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_keccak256_sm3.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_skein256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
@@ -697,6 +931,9 @@
<CudaCompile Include="Algo256\cuda_bmw.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\allium.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\cuda_lyra2.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
@@ -709,9 +946,105 @@
<CudaCompile Include="lyra2\lyra2REv2.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\cuda_lyra2v3.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
<ClInclude Include="lyra2\cuda_lyra2v3_sm3.cuh">
<Filter>Source Files\CUDA\lyra2</Filter>
</ClInclude>
<CudaCompile Include="lyra2\lyra2REv3.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\cuda_lyra2Z.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\lyra2Z.cu">
<Filter>Source Files\CUDA\lyra2</Filter>
</CudaCompile>
<CudaCompile Include="blake2b.cu">
<Filter>Source Files\CUDA\</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\blake2s.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="lbry\cuda_sha256_lbry.cu">
<Filter>Source Files\CUDA\lbry</Filter>
</CudaCompile>
<CudaCompile Include="lbry\cuda_sha512_lbry.cu">
<Filter>Source Files\CUDA\lbry</Filter>
</CudaCompile>
<CudaCompile Include="lbry\cuda_lbry_merged.cu">
<Filter>Source Files\CUDA\lbry</Filter>
</CudaCompile>
<CudaCompile Include="lbry\lbry.cu">
<Filter>Source Files\CUDA\lbry</Filter>
</CudaCompile>
<CudaCompile Include="sha256\cuda_sha256d.cu">
<Filter>Source Files\CUDA\sha256</Filter>
</CudaCompile>
<CudaCompile Include="sha256\sha256d.cu">
<Filter>Source Files\CUDA\sha256</Filter>
</CudaCompile>
<CudaCompile Include="sha256\cuda_sha256t.cu">
<Filter>Source Files\CUDA\sha256</Filter>
</CudaCompile>
<CudaCompile Include="sha256\sha256t.cu">
<Filter>Source Files\CUDA\sha256</Filter>
</CudaCompile>
<CudaCompile Include="sha256\cuda_sha256q.cu">
<Filter>Source Files\CUDA\sha256</Filter>
</CudaCompile>
<CudaCompile Include="sha256\sha256q.cu">
<Filter>Source Files\CUDA\sha256</Filter>
</CudaCompile>
<CudaCompile Include="sia\sia.cu">
<Filter>Source Files\sia</Filter>
</CudaCompile>
<CudaCompile Include="crypto\cryptolight.cu">
<Filter>Source Files\CUDA\xmr</Filter>
</CudaCompile>
<CudaCompile Include="crypto\cryptolight-core.cu">
<Filter>Source Files\CUDA\xmr</Filter>
</CudaCompile>
<CudaCompile Include="crypto\cryptonight.cu">
<Filter>Source Files\CUDA\xmr</Filter>
</CudaCompile>
<CudaCompile Include="crypto\cryptonight-core.cu">
<Filter>Source Files\CUDA\xmr</Filter>
</CudaCompile>
<CudaCompile Include="crypto\cryptonight-extra.cu">
<Filter>Source Files\CUDA\xmr</Filter>
</CudaCompile>
<CudaCompile Include="crypto\wildkeccak.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<CudaCompile Include="equi\cuda_equi.cu">
<Filter>Source Files\equi</Filter>
</CudaCompile>
<CudaCompile Include="x15\cuda_x15_whirlpool_sm3.cu">
<Filter>Source Files\CUDA\x15</Filter>
</CudaCompile>
<CudaCompile Include="x16\cuda_x16_echo512.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
<CudaCompile Include="x16\cuda_x16_echo512_64.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
<CudaCompile Include="x16\cuda_x16_fugue512.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
<CudaCompile Include="x16\cuda_x16_shabal512.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
<CudaCompile Include="x16\cuda_x16_simd512_80.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
<CudaCompile Include="x16\x16r.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
<CudaCompile Include="x16\x16s.cu">
<Filter>Source Files\CUDA\x16</Filter>
</CudaCompile>
</ItemGroup>
<ItemGroup>
<Image Include="res\ccminer.ico">
@@ -728,4 +1061,4 @@
<Filter>Ressources</Filter>
</Text>
</ItemGroup>
</Project>
</Project>
+11
View File
@@ -51,6 +51,17 @@ static __inline int setpriority(int which, int who, int prio)
#define strcasecmp(x,y) _stricmp(x,y)
typedef int ssize_t;
__inline int msver(void) {
switch (_MSC_VER) {
case 1500: return 2008;
case 1600: return 2010;
case 1700: return 2012;
case 1800: return 2013;
case 1900: return 2015;
default: return (_MSC_VER/100);
}
}
#include <stdlib.h>
static __inline char * dirname(char *file) {
char buffer[_MAX_PATH] = { 0 };
+602
View File
@@ -0,0 +1,602 @@
// Copyright (c) 2009-2010 Satoshi Nakamoto
// Copyright (c) 2009-2012 The Bitcoin developers
// Distributed under the MIT/X11 software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
#ifndef BITCOIN_BIGNUM_H
#define BITCOIN_BIGNUM_H
#ifndef PROTOCOL_VERSION
#define PROTOCOL_VERSION 70001
#endif
#include <map>
#include <limits>
#include <algorithm>
#include <stdexcept>
#include <vector>
#include "uint256.h"
#include <openssl/bn.h>
#include "serialize.hpp"
using namespace std;
/** Errors thrown by the bignum class */
class bignum_error : public std::runtime_error
{
public:
explicit bignum_error(const std::string& str) : std::runtime_error(str) {}
};
/** RAII encapsulated BN_CTX (OpenSSL bignum context) */
class CAutoBN_CTX
{
protected:
BN_CTX* pctx;
BN_CTX* operator=(BN_CTX* pnew) { return pctx = pnew; }
public:
CAutoBN_CTX()
{
pctx = BN_CTX_new();
if (pctx == NULL)
throw bignum_error("CAutoBN_CTX : BN_CTX_new() returned NULL");
}
~CAutoBN_CTX()
{
if (pctx != NULL)
BN_CTX_free(pctx);
}
operator BN_CTX*() { return pctx; }
BN_CTX& operator*() { return *pctx; }
BN_CTX** operator&() { return &pctx; }
bool operator!() { return (pctx == NULL); }
};
/** C++ wrapper for BIGNUM (OpenSSL bignum) */
class CBigNum : public BIGNUM
{
public:
CBigNum()
{
BN_init(this);
}
CBigNum(const CBigNum& b)
{
BN_init(this);
if (!BN_copy(this, &b))
{
BN_clear_free(this);
throw bignum_error("CBigNum::CBigNum(const CBigNum&) : BN_copy failed");
}
}
CBigNum& operator=(const CBigNum& b)
{
if (!BN_copy(this, &b))
throw bignum_error("CBigNum::operator= : BN_copy failed");
return (*this);
}
~CBigNum()
{
BN_clear_free(this);
}
//CBigNum(char n) is not portable. Use 'signed char' or 'unsigned char'.
CBigNum(signed char n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(short n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(int n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(long n) { BN_init(this); if (n >= 0) setulong(n); else setint64(n); }
CBigNum(int64 n) { BN_init(this); setint64(n); }
CBigNum(unsigned char n) { BN_init(this); setulong(n); }
CBigNum(unsigned short n) { BN_init(this); setulong(n); }
CBigNum(unsigned int n) { BN_init(this); setulong(n); }
CBigNum(unsigned long n) { BN_init(this); setulong(n); }
CBigNum(uint64 n) { BN_init(this); setuint64(n); }
explicit CBigNum(uint256 n) { BN_init(this); setuint256(n); }
explicit CBigNum(const std::vector<unsigned char>& vch)
{
BN_init(this);
setvch(vch);
}
void setulong(unsigned long n)
{
if (!BN_set_word(this, n))
throw bignum_error("CBigNum conversion from unsigned long : BN_set_word failed");
}
unsigned long getulong() const
{
return (unsigned long) BN_get_word(this);
}
unsigned int getuint() const
{
return (unsigned int) BN_get_word(this);
}
int getint() const
{
unsigned long n = (unsigned long) BN_get_word(this);
if (!BN_is_negative(this))
return (n > (unsigned long)std::numeric_limits<int>::max() ? std::numeric_limits<int>::max() : n);
else
return (n > (unsigned long)std::numeric_limits<int>::max() ? std::numeric_limits<int>::min() : -(int)n);
}
void setint64(int64 sn)
{
unsigned char pch[sizeof(sn) + 6];
unsigned char* p = pch + 4;
bool fNegative;
uint64 n;
if (sn < 0LL)
{
// Since the minimum signed integer cannot be represented as positive so long as its type is signed,
// and it's not well-defined what happens if you make it unsigned before negating it,
// we instead increment the negative integer by 1, convert it, then increment the (now positive) unsigned integer by 1 to compensate
n = -(sn + 1);
++n;
fNegative = true;
} else {
n = sn;
fNegative = false;
}
bool fLeadingZeroes = true;
for (int i = 0; i < 8; i++)
{
unsigned char c = (n >> 56) & 0xff;
n <<= 8;
if (fLeadingZeroes)
{
if (c == 0)
continue;
if (c & 0x80)
*p++ = (fNegative ? 0x80 : 0);
else if (fNegative)
c |= 0x80;
fLeadingZeroes = false;
}
*p++ = c;
}
unsigned int nSize = (unsigned int) (p - (pch + 4));
pch[0] = (nSize >> 24) & 0xff;
pch[1] = (nSize >> 16) & 0xff;
pch[2] = (nSize >> 8) & 0xff;
pch[3] = (nSize) & 0xff;
BN_mpi2bn(pch, (int) (p - pch), this);
}
void setuint64(uint64 n)
{
unsigned char pch[sizeof(n) + 6];
unsigned char* p = pch + 4;
bool fLeadingZeroes = true;
for (int i = 0; i < 8; i++)
{
unsigned char c = (n >> 56) & 0xff;
n <<= 8;
if (fLeadingZeroes)
{
if (c == 0)
continue;
if (c & 0x80)
*p++ = 0;
fLeadingZeroes = false;
}
*p++ = c;
}
unsigned int nSize = (unsigned int) (p - (pch + 4));
pch[0] = (nSize >> 24) & 0xff;
pch[1] = (nSize >> 16) & 0xff;
pch[2] = (nSize >> 8) & 0xff;
pch[3] = (nSize) & 0xff;
BN_mpi2bn(pch, (int) (p - pch), this);
}
void setuint256(uint256 n)
{
unsigned char pch[sizeof(n) + 6];
unsigned char* p = pch + 4;
bool fLeadingZeroes = true;
unsigned char* pbegin = (unsigned char*)&n;
unsigned char* psrc = pbegin + sizeof(n);
while (psrc != pbegin)
{
unsigned char c = *(--psrc);
if (fLeadingZeroes)
{
if (c == 0)
continue;
if (c & 0x80)
*p++ = 0;
fLeadingZeroes = false;
}
*p++ = c;
}
unsigned int nSize = (unsigned int) (p - (pch + 4));
pch[0] = (nSize >> 24) & 0xff;
pch[1] = (nSize >> 16) & 0xff;
pch[2] = (nSize >> 8) & 0xff;
pch[3] = (nSize >> 0) & 0xff;
BN_mpi2bn(pch, (int) (p - pch), this);
}
uint256 getuint256() const
{
unsigned int nSize = BN_bn2mpi(this, NULL);
if (nSize < 4)
return 0;
std::vector<unsigned char> vch(nSize);
BN_bn2mpi(this, &vch[0]);
if (vch.size() > 4)
vch[4] &= 0x7f;
uint256 n = 0;
for (unsigned int i = 0, j = (unsigned int) vch.size()-1; i < sizeof(n) && j >= 4; i++, j--)
((unsigned char*)&n)[i] = vch[j];
return n;
}
void setvch(const std::vector<unsigned char>& vch)
{
std::vector<unsigned char> vch2(vch.size() + 4);
unsigned int nSize = (unsigned int) vch.size();
// BIGNUM's byte stream format expects 4 bytes of
// big endian size data info at the front
vch2[0] = (nSize >> 24) & 0xff;
vch2[1] = (nSize >> 16) & 0xff;
vch2[2] = (nSize >> 8) & 0xff;
vch2[3] = (nSize >> 0) & 0xff;
// swap data to big endian
reverse_copy(vch.begin(), vch.end(), vch2.begin() + 4);
BN_mpi2bn(&vch2[0], (int) vch2.size(), this);
}
std::vector<unsigned char> getvch() const
{
unsigned int nSize = BN_bn2mpi(this, NULL);
if (nSize <= 4)
return std::vector<unsigned char>();
std::vector<unsigned char> vch(nSize);
BN_bn2mpi(this, &vch[0]);
vch.erase(vch.begin(), vch.begin() + 4);
reverse(vch.begin(), vch.end());
return vch;
}
// The "compact" format is a representation of a whole
// number N using an unsigned 32bit number similar to a
// floating point format.
// The most significant 8 bits are the unsigned exponent of base 256.
// This exponent can be thought of as "number of bytes of N".
// The lower 23 bits are the mantissa.
// Bit number 24 (0x800000) represents the sign of N.
// N = (-1^sign) * mantissa * 256^(exponent-3)
//
// Satoshi's original implementation used BN_bn2mpi() and BN_mpi2bn().
// MPI uses the most significant bit of the first byte as sign.
// Thus 0x1234560000 is compact (0x05123456)
// and 0xc0de000000 is compact (0x0600c0de)
// (0x05c0de00) would be -0x40de000000
//
// Bitcoin only uses this "compact" format for encoding difficulty
// targets, which are unsigned 256bit quantities. Thus, all the
// complexities of the sign bit and using base 256 are probably an
// implementation accident.
//
// This implementation directly uses shifts instead of going
// through an intermediate MPI representation.
CBigNum& SetCompact(unsigned int nCompact)
{
unsigned int nSize = nCompact >> 24;
bool fNegative =(nCompact & 0x00800000) != 0;
unsigned int nWord = nCompact & 0x007fffff;
if (nSize <= 3)
{
nWord >>= 8*(3-nSize);
BN_set_word(this, nWord);
}
else
{
BN_set_word(this, nWord);
BN_lshift(this, this, 8*(nSize-3));
}
BN_set_negative(this, fNegative);
return *this;
}
unsigned int GetCompact() const
{
unsigned int nSize = BN_num_bytes(this);
unsigned int nCompact = 0;
if (nSize <= 3)
nCompact = (unsigned int) BN_get_word(this) << 8*(3-nSize);
else
{
CBigNum bn;
BN_rshift(&bn, this, 8*(nSize-3));
nCompact = (unsigned int) BN_get_word(&bn);
}
// The 0x00800000 bit denotes the sign.
// Thus, if it is already set, divide the mantissa by 256 and increase the exponent.
if (nCompact & 0x00800000)
{
nCompact >>= 8;
nSize++;
}
nCompact |= nSize << 24;
nCompact |= (BN_is_negative(this) ? 0x00800000 : 0);
return nCompact;
}
void SetHex(const std::string& str)
{
// skip 0x
const char* psz = str.c_str();
while (isspace(*psz))
psz++;
bool fNegative = false;
if (*psz == '-')
{
fNegative = true;
psz++;
}
if (psz[0] == '0' && tolower(psz[1]) == 'x')
psz += 2;
while (isspace(*psz))
psz++;
// hex string to bignum
static const signed char phexdigit[256] = { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,1,2,3,4,5,6,7,8,9,0,0,0,0,0,0, 0,0xa,0xb,0xc,0xd,0xe,0xf,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0xa,0xb,0xc,0xd,0xe,0xf,0,0,0,0,0,0,0,0,0 };
*this = 0;
while (isxdigit(*psz))
{
*this <<= 4;
int n = phexdigit[(unsigned char)*psz++];
*this += n;
}
if (fNegative)
*this = 0 - *this;
}
std::string ToString(int nBase=10) const
{
CAutoBN_CTX pctx;
CBigNum bnBase = nBase;
CBigNum bn0 = 0;
std::string str;
CBigNum bn = *this;
BN_set_negative(&bn, false);
CBigNum dv;
CBigNum rem;
if (BN_cmp(&bn, &bn0) == 0)
return "0";
while (BN_cmp(&bn, &bn0) > 0)
{
if (!BN_div(&dv, &rem, &bn, &bnBase, pctx))
throw bignum_error("CBigNum::ToString() : BN_div failed");
bn = dv;
unsigned int c = rem.getulong();
str += "0123456789abcdef"[c];
}
if (BN_is_negative(this))
str += "-";
reverse(str.begin(), str.end());
return str;
}
std::string GetHex() const
{
return ToString(16);
}
unsigned int GetSerializeSize(int nType=0, int nVersion=PROTOCOL_VERSION) const
{
return ::GetSerializeSize(getvch(), nType, nVersion);
}
template<typename Stream>
void Serialize(Stream& s, int nType=0, int nVersion=PROTOCOL_VERSION) const
{
::Serialize(s, getvch(), nType, nVersion);
}
template<typename Stream>
void Unserialize(Stream& s, int nType=0, int nVersion=PROTOCOL_VERSION)
{
std::vector<unsigned char> vch;
::Unserialize(s, vch, nType, nVersion);
setvch(vch);
}
bool operator!() const
{
return BN_is_zero(this);
}
CBigNum& operator+=(const CBigNum& b)
{
if (!BN_add(this, this, &b))
throw bignum_error("CBigNum::operator+= : BN_add failed");
return *this;
}
CBigNum& operator-=(const CBigNum& b)
{
*this = *this - b;
return *this;
}
CBigNum& operator*=(const CBigNum& b)
{
CAutoBN_CTX pctx;
if (!BN_mul(this, this, &b, pctx))
throw bignum_error("CBigNum::operator*= : BN_mul failed");
return *this;
}
CBigNum& operator/=(const CBigNum& b)
{
*this = *this / b;
return *this;
}
CBigNum& operator%=(const CBigNum& b)
{
*this = *this % b;
return *this;
}
CBigNum& operator<<=(unsigned int shift)
{
if (!BN_lshift(this, this, shift))
throw bignum_error("CBigNum:operator<<= : BN_lshift failed");
return *this;
}
CBigNum& operator>>=(unsigned int shift)
{
// Note: BN_rshift segfaults on 64-bit if 2^shift is greater than the number
// if built on ubuntu 9.04 or 9.10, probably depends on version of OpenSSL
CBigNum a = 1;
a <<= shift;
if (BN_cmp(&a, this) > 0)
{
*this = 0;
return *this;
}
if (!BN_rshift(this, this, shift))
throw bignum_error("CBigNum:operator>>= : BN_rshift failed");
return *this;
}
CBigNum& operator++()
{
// prefix operator
if (!BN_add(this, this, BN_value_one()))
throw bignum_error("CBigNum::operator++ : BN_add failed");
return *this;
}
const CBigNum operator++(int)
{
// postfix operator
const CBigNum ret = *this;
++(*this);
return ret;
}
CBigNum& operator--()
{
// prefix operator
CBigNum r;
if (!BN_sub(&r, this, BN_value_one()))
throw bignum_error("CBigNum::operator-- : BN_sub failed");
*this = r;
return *this;
}
const CBigNum operator--(int)
{
// postfix operator
const CBigNum ret = *this;
--(*this);
return ret;
}
friend inline const CBigNum operator-(const CBigNum& a, const CBigNum& b);
friend inline const CBigNum operator/(const CBigNum& a, const CBigNum& b);
friend inline const CBigNum operator%(const CBigNum& a, const CBigNum& b);
};
inline const CBigNum operator+(const CBigNum& a, const CBigNum& b)
{
CBigNum r;
if (!BN_add(&r, &a, &b))
throw bignum_error("CBigNum::operator+ : BN_add failed");
return r;
}
inline const CBigNum operator-(const CBigNum& a, const CBigNum& b)
{
CBigNum r;
if (!BN_sub(&r, &a, &b))
throw bignum_error("CBigNum::operator- : BN_sub failed");
return r;
}
inline const CBigNum operator-(const CBigNum& a)
{
CBigNum r(a);
BN_set_negative(&r, !BN_is_negative(&r));
return r;
}
inline const CBigNum operator*(const CBigNum& a, const CBigNum& b)
{
CAutoBN_CTX pctx;
CBigNum r;
if (!BN_mul(&r, &a, &b, pctx))
throw bignum_error("CBigNum::operator* : BN_mul failed");
return r;
}
inline const CBigNum operator/(const CBigNum& a, const CBigNum& b)
{
CAutoBN_CTX pctx;
CBigNum r;
if (!BN_div(&r, NULL, &a, &b, pctx))
throw bignum_error("CBigNum::operator/ : BN_div failed");
return r;
}
inline const CBigNum operator%(const CBigNum& a, const CBigNum& b)
{
CAutoBN_CTX pctx;
CBigNum r;
if (!BN_mod(&r, &a, &b, pctx))
throw bignum_error("CBigNum::operator% : BN_div failed");
return r;
}
inline const CBigNum operator<<(const CBigNum& a, unsigned int shift)
{
CBigNum r;
if (!BN_lshift(&r, &a, shift))
throw bignum_error("CBigNum:operator<< : BN_lshift failed");
return r;
}
inline const CBigNum operator>>(const CBigNum& a, unsigned int shift)
{
CBigNum r = a;
r >>= shift;
return r;
}
inline bool operator==(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) == 0); }
inline bool operator!=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) != 0); }
inline bool operator<=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) <= 0); }
inline bool operator>=(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) >= 0); }
inline bool operator<(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) < 0); }
inline bool operator>(const CBigNum& a, const CBigNum& b) { return (BN_cmp(&a, &b) > 0); }
#endif
+188 -194
View File
@@ -1,194 +1,188 @@
/* cpuminer-config.h. Generated from cpuminer-config.h.in by configure. */
/* cpuminer-config.h.in. Generated from configure.ac by autoheader. */
/* Define to one of `_getb67', `GETB67', `getb67' for Cray-2 and Cray-YMP
systems. This function is required for `alloca.c' support on those systems.
*/
/* #undef CRAY_STACKSEG_END */
/* Define to 1 if using `alloca.c'. */
/* #undef C_ALLOCA */
/* Define to 1 if you have `alloca', as a function or macro. */
#define HAVE_ALLOCA 1
/* Define to 1 if you have <alloca.h> and it should be used (not on Ultrix).
*/
#define HAVE_ALLOCA_H 1
/* Define to 1 if you have the declaration of `be32dec', and to 0 if you
don't. */
#define HAVE_DECL_BE32DEC 0
/* Define to 1 if you have the declaration of `be32enc', and to 0 if you
don't. */
#define HAVE_DECL_BE32ENC 0
/* Define to 1 if you have the declaration of `le32dec', and to 0 if you
don't. */
#define HAVE_DECL_LE32DEC 0
/* Define to 1 if you have the declaration of `le32enc', and to 0 if you
don't. */
#define HAVE_DECL_LE32ENC 0
/* Define to 1 if you have the `getopt_long' function. */
#define HAVE_GETOPT_LONG 1
/* Define to 1 if you have the <inttypes.h> header file. */
#define HAVE_INTTYPES_H 1
/* Define to 1 if you have the `crypto' library (-lcrypto). */
#define HAVE_LIBCRYPTO 1
/* Define to 1 if you have a functional curl library. */
#define HAVE_LIBCURL 1
/* Define to 1 if you have the `ssl' library (-lssl). */
#define HAVE_LIBSSL 1
/* Define to 1 if you have the `z' library (-lz). */
#define HAVE_LIBZ 1
/* Define to 1 if you have the <memory.h> header file. */
#define HAVE_MEMORY_H 1
/* Define to 1 if you have the <stdint.h> header file. */
#define HAVE_STDINT_H 1
/* Define to 1 if you have the <stdlib.h> header file. */
#define HAVE_STDLIB_H 1
/* Define to 1 if you have the <strings.h> header file. */
#define HAVE_STRINGS_H 1
/* Define to 1 if you have the <string.h> header file. */
#define HAVE_STRING_H 1
/* Define to 1 if you have the <syslog.h> header file. */
/* #undef HAVE_SYSLOG_H */
/* Define to 1 if you have the <sys/endian.h> header file. */
/* #undef HAVE_SYS_ENDIAN_H */
/* Define to 1 if you have the <sys/param.h> header file. */
#define HAVE_SYS_PARAM_H 1
/* Define to 1 if you have the <sys/stat.h> header file. */
#define HAVE_SYS_STAT_H 1
/* Define to 1 if you have the <sys/sysctl.h> header file. */
#define HAVE_SYS_SYSCTL_H 1
/* Define to 1 if you have the <sys/types.h> header file. */
#define HAVE_SYS_TYPES_H 1
/* Define to 1 if you have the <unistd.h> header file. */
#define HAVE_UNISTD_H 1
/* Defined if libcurl supports AsynchDNS */
/* #undef LIBCURL_FEATURE_ASYNCHDNS */
/* Defined if libcurl supports IDN */
#define LIBCURL_FEATURE_IDN 1
/* Defined if libcurl supports IPv6 */
#define LIBCURL_FEATURE_IPV6 1
/* Defined if libcurl supports KRB4 */
/* #undef LIBCURL_FEATURE_KRB4 */
/* Defined if libcurl supports libz */
#define LIBCURL_FEATURE_LIBZ 1
/* Defined if libcurl supports NTLM */
#define LIBCURL_FEATURE_NTLM 1
/* Defined if libcurl supports SSL */
#define LIBCURL_FEATURE_SSL 1
/* Defined if libcurl supports SSPI */
/* #undef LIBCURL_FEATURE_SSPI */
/* Defined if libcurl supports DICT */
/* #undef LIBCURL_PROTOCOL_DICT */
/* Defined if libcurl supports FILE */
#define LIBCURL_PROTOCOL_FILE 1
/* Defined if libcurl supports FTP */
#define LIBCURL_PROTOCOL_FTP 1
/* Defined if libcurl supports FTPS */
#define LIBCURL_PROTOCOL_FTPS 1
/* Defined if libcurl supports HTTP */
#define LIBCURL_PROTOCOL_HTTP 1
/* Defined if libcurl supports HTTPS */
#define LIBCURL_PROTOCOL_HTTPS 1
/* Defined if libcurl supports IMAP */
/* #undef LIBCURL_PROTOCOL_IMAP */
/* Defined if libcurl supports LDAP */
/* #undef LIBCURL_PROTOCOL_LDAP */
/* Defined if libcurl supports POP3 */
/* #undef LIBCURL_PROTOCOL_POP3 */
/* Defined if libcurl supports RTSP */
/* #undef LIBCURL_PROTOCOL_RTSP */
/* Defined if libcurl supports SMTP */
/* #undef LIBCURL_PROTOCOL_SMTP */
/* Defined if libcurl supports TELNET */
/* #undef LIBCURL_PROTOCOL_TELNET */
/* Defined if libcurl supports TFTP */
/* #undef LIBCURL_PROTOCOL_TFTP */
/* Define to 1 if your C compiler doesn't accept -c and -o together. */
/* #undef NO_MINUS_C_MINUS_O */
/* Name of package */
#define PACKAGE "ccminer"
/* Define to the address where bug reports for this package should be sent. */
#define PACKAGE_BUGREPORT ""
/* Define to the full name of this package. */
#define PACKAGE_NAME "ccminer"
/* Define to the full name and version of this package. */
#define PACKAGE_STRING "ccminer 1.7.6"
/* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "ccminer"
/* Define to the home page for this package. */
#define PACKAGE_URL "http://github.com/tpruvot/ccminer"
/* Define to the version of this package. */
#define PACKAGE_VERSION "1.7.6"
/* If using the C implementation of alloca, define if you know the
direction of stack growth for your system; otherwise it will be
automatically deduced at runtime.
STACK_DIRECTION > 0 => grows toward higher addresses
STACK_DIRECTION < 0 => grows toward lower addresses
STACK_DIRECTION = 0 => direction of growth unknown */
/* #undef STACK_DIRECTION */
/* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1
/* Version number of package */
#define VERSION "1.7.6"
/* Define curl_free() as free() if our version of curl lacks curl_free. */
/* #undef curl_free */
/* Define to `unsigned int' if <sys/types.h> does not define. */
//#define size_t unsigned int
/* CONFIG ONLY FOR MS VC++ BUILD */
/* Define to one of `_getb67', `GETB67', `getb67' for Cray-2 and Cray-YMP
systems. This function is required for `alloca.c' support on those systems.
*/
/* #undef CRAY_STACKSEG_END */
/* Define to 1 if using `alloca.c'. */
/* #undef C_ALLOCA */
/* Define to 1 if you have `alloca', as a function or macro. */
#define HAVE_ALLOCA 1
/* Define to 1 if you have <alloca.h> and it should be used (not on Ultrix).
*/
#define HAVE_ALLOCA_H 1
/* Define to 1 if you have the declaration of `be32dec', and to 0 if you
don't. */
#define HAVE_DECL_BE32DEC 0
/* Define to 1 if you have the declaration of `be32enc', and to 0 if you
don't. */
#define HAVE_DECL_BE32ENC 0
/* Define to 1 if you have the declaration of `le32dec', and to 0 if you
don't. */
#define HAVE_DECL_LE32DEC 0
/* Define to 1 if you have the declaration of `le32enc', and to 0 if you
don't. */
#define HAVE_DECL_LE32ENC 0
/* Define to 1 if you have the `getopt_long' function. */
#define HAVE_GETOPT_LONG 1
/* Define to 1 if you have the <inttypes.h> header file. */
#define HAVE_INTTYPES_H 1
/* Define to 1 if you have the `crypto' library (-lcrypto). */
#define HAVE_LIBCRYPTO 1
/* Define to 1 if you have a functional curl library. */
#define HAVE_LIBCURL 1
/* Define to 1 if you have the `ssl' library (-lssl). */
#define HAVE_LIBSSL 1
/* Define to 1 if you have the `z' library (-lz). */
#define HAVE_LIBZ 1
/* Define to 1 if you have the <memory.h> header file. */
#define HAVE_MEMORY_H 1
/* Define to 1 if you have the <stdint.h> header file. */
#define HAVE_STDINT_H 1
/* Define to 1 if you have the <stdlib.h> header file. */
#define HAVE_STDLIB_H 1
/* Define to 1 if you have the <strings.h> header file. */
#define HAVE_STRINGS_H 1
/* Define to 1 if you have the <string.h> header file. */
#define HAVE_STRING_H 1
/* Define to 1 if you have the <syslog.h> header file. */
/* #undef HAVE_SYSLOG_H */
/* Define to 1 if you have the <sys/endian.h> header file. */
/* #undef HAVE_SYS_ENDIAN_H */
/* Define to 1 if you have the <sys/param.h> header file. */
#define HAVE_SYS_PARAM_H 1
/* Define to 1 if you have the <sys/stat.h> header file. */
#define HAVE_SYS_STAT_H 1
/* Define to 1 if you have the <sys/sysctl.h> header file. */
#define HAVE_SYS_SYSCTL_H 1
/* Define to 1 if you have the <sys/types.h> header file. */
#define HAVE_SYS_TYPES_H 1
/* Define to 1 if you have the <unistd.h> header file. */
#define HAVE_UNISTD_H 1
/* Defined if libcurl supports AsynchDNS */
/* #undef LIBCURL_FEATURE_ASYNCHDNS */
/* Defined if libcurl supports IDN */
#define LIBCURL_FEATURE_IDN 1
/* Defined if libcurl supports IPv6 */
#define LIBCURL_FEATURE_IPV6 1
/* Defined if libcurl supports KRB4 */
/* #undef LIBCURL_FEATURE_KRB4 */
/* Defined if libcurl supports libz */
#define LIBCURL_FEATURE_LIBZ 1
/* Defined if libcurl supports NTLM */
#define LIBCURL_FEATURE_NTLM 1
/* Defined if libcurl supports SSL */
#define LIBCURL_FEATURE_SSL 1
/* Defined if libcurl supports SSPI */
/* #undef LIBCURL_FEATURE_SSPI */
/* Defined if libcurl supports DICT */
/* #undef LIBCURL_PROTOCOL_DICT */
/* Defined if libcurl supports FILE */
#define LIBCURL_PROTOCOL_FILE 1
/* Defined if libcurl supports FTP */
#define LIBCURL_PROTOCOL_FTP 1
/* Defined if libcurl supports FTPS */
#define LIBCURL_PROTOCOL_FTPS 1
/* Defined if libcurl supports HTTP */
#define LIBCURL_PROTOCOL_HTTP 1
/* Defined if libcurl supports HTTPS */
#define LIBCURL_PROTOCOL_HTTPS 1
/* Defined if libcurl supports IMAP */
/* #undef LIBCURL_PROTOCOL_IMAP */
/* Defined if libcurl supports LDAP */
/* #undef LIBCURL_PROTOCOL_LDAP */
/* Defined if libcurl supports POP3 */
/* #undef LIBCURL_PROTOCOL_POP3 */
/* Defined if libcurl supports RTSP */
/* #undef LIBCURL_PROTOCOL_RTSP */
/* Defined if libcurl supports SMTP */
/* #undef LIBCURL_PROTOCOL_SMTP */
/* Defined if libcurl supports TELNET */
/* #undef LIBCURL_PROTOCOL_TELNET */
/* Defined if libcurl supports TFTP */
/* #undef LIBCURL_PROTOCOL_TFTP */
/* Define to 1 if your C compiler doesn't accept -c and -o together. */
/* #undef NO_MINUS_C_MINUS_O */
/* Name of package */
#define PACKAGE "ccminer"
/* Define to the address where bug reports for this package should be sent. */
#define PACKAGE_BUGREPORT ""
/* Define to the full name of this package. */
#define PACKAGE_NAME "ccminer"
/* Define to the home page for this package. */
#define PACKAGE_URL "http://github.com/tpruvot/ccminer"
/* Define to the version of this package. */
#define PACKAGE_VERSION "2.3.1"
/* If using the C implementation of alloca, define if you know the
direction of stack growth for your system; otherwise it will be
automatically deduced at runtime.
STACK_DIRECTION > 0 => grows toward higher addresses
STACK_DIRECTION < 0 => grows toward lower addresses
STACK_DIRECTION = 0 => direction of growth unknown */
/* #undef STACK_DIRECTION */
/* Define to 1 if you have the ANSI C header files. */
#define STDC_HEADERS 1
/* Define curl_free() as free() if our version of curl lacks curl_free. */
/* #undef curl_free */
/* Define to `unsigned int' if <sys/types.h> does not define. */
//#define size_t unsigned int
#if !defined(HAVE_STRUCT_TIMESPEC) && _MSC_VER >= 1900
#define HAVE_STRUCT_TIMESPEC
#endif
+3
View File
@@ -23,6 +23,9 @@
#ifdef _MSC_VER
#define inline __inline
#if !defined(HAVE_STRUCT_TIMESPEC) && _MSC_VER >= 1900
#define HAVE_STRUCT_TIMESPEC
#endif
#endif
#ifdef __cplusplus
Binary file not shown.
+1197 -193
View File
File diff suppressed because it is too large. Load diff
+371 -2
View File
@@ -1,2 +1,371 @@
/* todo: stripped version... or not */
#include "nvapi.h"
#pragma once
#define NVAPI_INTERNAL
#include "nvapi.h"
NvAPI_Status nvapi_dll_init();
typedef struct {
NvU32 version;
NvU8 valid;
NvU8 count;
struct
{
NvU32 pstate; // Assumption
NvU32 unknown1[2];
NvU32 min_power;
NvU32 unknown2[2];
NvU32 def_power;
NvU32 unknown3[2];
NvU32 max_power;
NvU32 unknown4; // 0
} entries[4];
} NVAPI_GPU_POWER_INFO;
#define NVAPI_GPU_POWER_INFO_VER MAKE_NVAPI_VERSION(NVAPI_GPU_POWER_INFO, 1)
typedef struct {
NvU32 version;
NvU32 flags;
struct {
NvU32 unknown1;
NvU32 unknown2;
NvU32 power; // percent * 1000
NvU32 unknown4;
} entries[4];
} NVAPI_GPU_POWER_STATUS;
#define NVAPI_GPU_POWER_STATUS_VER MAKE_NVAPI_VERSION(NVAPI_GPU_POWER_STATUS, 1)
typedef struct {
NvU32 version;
NvU32 count;
struct {
NvU32 unknown1;
NvU32 unknown2;
NvU32 power; // unsure ?? 85536 to 95055 on 1080, 104825+ on 970
NvU32 unknown4;
} entries[4];
} NVAPI_GPU_POWER_TOPO;
#define NVAPI_GPU_POWER_TOPO_VER MAKE_NVAPI_VERSION(NVAPI_GPU_POWER_TOPO, 1)
typedef struct {
NvU32 version;
NvU32 flags;
struct {
NvU32 controller;
NvU32 unknown;
NvS32 min_temp;
NvS32 def_temp;
NvS32 max_temp;
NvU32 defaultFlags;
} entries[4];
} NVAPI_GPU_THERMAL_INFO;
#define NVAPI_GPU_THERMAL_INFO_VER MAKE_NVAPI_VERSION(NVAPI_GPU_THERMAL_INFO, 2)
typedef struct {
NvU32 version;
NvU32 flags;
struct {
NvU32 controller;
NvU32 value;
NvU32 flags;
} entries[4];
} NVAPI_GPU_THERMAL_LIMIT;
#define NVAPI_GPU_THERMAL_LIMIT_VER MAKE_NVAPI_VERSION(NVAPI_GPU_THERMAL_LIMIT, 2)
// Maxwell gpu core voltage reading
typedef struct {
NvU32 version;
NvU32 flags;
NvU32 count; // unsure
NvU32 unknown;
NvU32 value_uV;
NvU32 buf1[30];
} NVAPI_VOLT_STATUS; // 140 bytes (1-008c)
#define NVAPI_VOLT_STATUS_VER MAKE_NVAPI_VERSION(NVAPI_VOLT_STATUS, 1)
// Pascal gpu core voltage reading
typedef struct {
NvU32 version;
NvU32 flags;
NvU32 nul[8];
NvU32 value_uV;
NvU32 buf1[8];
} NVAPI_VOLTAGE_STATUS; // 76 bytes (1-004c)
#define NVAPI_VOLTAGE_STATUS_VER MAKE_NVAPI_VERSION(NVAPI_VOLTAGE_STATUS, 1)
typedef struct {
NvU32 version;
NvU32 numClocks; // unsure
NvU32 nul[8];
struct {
NvU32 a;
NvU32 clockType;
NvU32 c;
NvU32 d;
NvU32 e;
NvU32 f;
NvU32 g;
NvU32 h;
NvU32 i;
NvU32 j;
NvS32 rangeMax;
NvS32 rangeMin;
NvS32 tempMax; // ? unsure
NvU32 n;
NvU32 o;
NvU32 p;
NvU32 q;
NvU32 r;
} entries[32]; // NVAPI_MAX_GPU_CLOCKS ?
} NVAPI_CLOCKS_RANGE; // 2344 bytes
#define NVAPI_CLOCKS_RANGE_VER MAKE_NVAPI_VERSION(NVAPI_CLOCKS_RANGE, 1)
// seems to return a clock table mask
typedef struct {
NvU32 version;
NvU32 mask[4]; // 80 bits mask
NvU32 buf0[8];
struct {
NvU32 a;
NvU32 b;
NvU32 c;
NvU32 d;
NvU32 memDelta; // 1 for mem
NvU32 gpuDelta; // 1 for gpu
} clocks[80 + 23];
NvU32 buf1[916];
} NVAPI_CLOCK_MASKS; // 6188 bytes
#define NVAPI_CLOCK_MASKS_VER MAKE_NVAPI_VERSION(NVAPI_CLOCK_MASKS, 1)
// contains the gpu/mem clocks deltas
typedef struct {
NvU32 version;
NvU32 mask[4]; // 80 bits mask (could be 8x 32bits)
NvU32 buf0[12];
struct {
NvU32 a;
NvU32 b;
NvU32 c;
NvU32 d;
NvU32 e;
NvS32 freqDelta; // 84000 = +84MHz
NvU32 g;
NvU32 h;
NvU32 i;
} gpuDeltas[80];
NvU32 memFilled[23]; // maybe only 4 max
NvS32 memDeltas[23];
NvU32 buf1[1529];
} NVAPI_CLOCK_TABLE; // 9248 bytes
#define NVAPI_CLOCK_TABLE_VER MAKE_NVAPI_VERSION(NVAPI_CLOCK_TABLE, 1)
typedef struct {
NvU32 version;
NvU32 mask[4]; // 80 bits mask
NvU32 buf0[12];
struct {
NvU32 a; // 0
NvU32 freq_kHz;
NvU32 volt_uV;
NvU32 d;
NvU32 e;
NvU32 f;
NvU32 g;
} gpuEntries[80];
struct {
NvU32 a; // 1 for idle values ?
NvU32 freq_kHz;
NvU32 volt_uV;
NvU32 d;
NvU32 e;
NvU32 f;
NvU32 g;
} memEntries[23];
NvU32 buf1[1064];
} NVAPI_VFP_CURVE; // 7208 bytes (1-1c28)
#define NVAPI_VFP_CURVE_VER MAKE_NVAPI_VERSION(NVAPI_VFP_CURVE, 1)
typedef struct {
NvU32 version;
NvS32 percent;
NvU32 pad[8];
} NVAPI_VOLTBOOST_PERCENT; // 40 bytes (1-0028)
#define NVAPI_VOLTBOOST_PERCENT_VER MAKE_NVAPI_VERSION(NVAPI_VOLTBOOST_PERCENT, 1)
typedef struct {
NvU32 version;
NvU32 flags;
NvU32 filled; // 1
struct {
NvU32 volt_uV;
NvU32 unknown;
} entries[128];
// some empty tables then...
NvU32 buf1[3888];
} NVAPI_VOLTAGES_TABLE; // 16588 bytes (1-40cc)
#define NVAPI_VOLTAGES_TABLE_VER MAKE_NVAPI_VERSION(NVAPI_VOLTAGES_TABLE, 1)
typedef struct {
NvU32 version;
NvU32 val1; // 7
NvU32 val2; // 0x3F (63.)
NvU32 pad[16];
} NVAPI_GPU_PERF_INFO; // 76 bytes (1-004c)
#define NVAPI_GPU_PERF_INFO_VER MAKE_NVAPI_VERSION(NVAPI_GPU_PERF_INFO, 1)
typedef struct {
NvU32 version;
NvU32 flags; // 0
NvU64 timeRef; // increment with time
NvU64 val1; // seen 1 4 5 while mining, 16 else
NvU64 val2; // seen 7 and 3
NvU64 values[3]; // increment with time
NvU32 pad[326]; // empty
} NVAPI_GPU_PERF_STATUS; // 1360 bytes (1-0550)
#define NVAPI_GPU_PERF_STATUS_VER MAKE_NVAPI_VERSION(NVAPI_GPU_PERF_STATUS, 1)
typedef struct {
NvU32 version;
NvU32 val1; // 4
NvU32 val2; // 2 or 0
NvU32 val3; // 2
NvU32 val4; // 3
NV_GPU_PERF_PSTATE_ID pStateId;
NvU32 val6; // 0 or 2
NvU32 val7; // 4
NvU32 val8; // 0
NvU32 memFreq1; // 405000.
NvU32 memFreq2; // 405000.
NvU32 memFreqMin;// 101250.
NvU32 memFreqMax;// 486000.
NvU32 zeros[3];
NvU32 gpuFreq1; // 696000. Unsure about those
NvU32 gpuFreq2; // 696000.
NvU32 gpuFreqMin;// 174000.
NvU32 gpuFreqMax;// 658000.
NvU32 pad[2697];
} NVAPI_GPU_PERF_CLOCKS; // 10868 bytes (2-2a74)
#define NVAPI_GPU_PERF_CLOCKS_VER MAKE_NVAPI_VERSION(NVAPI_GPU_PERF_CLOCKS, 2)
typedef struct {
NvU32 version;
NvU32 level;
NvU32 count;
NvU32 pad[339]; // (4-0558)
} NVAPI_COOLER_SETTINGS;
#define NVAPI_COOLER_SETTINGS_VER MAKE_NVAPI_VERSION(NVAPI_COOLER_SETTINGS, 4)
typedef struct {
NvU32 version;
NvU32 level; // 0 = auto ?
NvU32 count; // 1
NvU32 pad[38]; // (1-00a4)
} NVAPI_COOLER_LEVEL;
#define NVAPI_COOLER_LEVEL_VER MAKE_NVAPI_VERSION(NVAPI_COOLER_LEVEL, 1)
NvAPI_Status NvAPI_DLL_GetInterfaceVersionString(NvAPI_ShortString string);
NvAPI_Status NvAPI_DLL_PerfPoliciesGetInfo(NvPhysicalGpuHandle, NVAPI_GPU_PERF_INFO*); // 409D9841 1-004c
NvAPI_Status NvAPI_DLL_PerfPoliciesGetStatus(NvPhysicalGpuHandle, NVAPI_GPU_PERF_STATUS*); // 3D358A0C 1-0550
NvAPI_Status NvAPI_DLL_ClientPowerPoliciesGetInfo(NvPhysicalGpuHandle, NVAPI_GPU_POWER_INFO*);
NvAPI_Status NvAPI_DLL_ClientPowerPoliciesGetStatus(NvPhysicalGpuHandle, NVAPI_GPU_POWER_STATUS*);
NvAPI_Status NvAPI_DLL_ClientPowerPoliciesSetStatus(NvPhysicalGpuHandle, NVAPI_GPU_POWER_STATUS*);
NvAPI_Status NvAPI_DLL_ClientPowerTopologyGetStatus(NvPhysicalGpuHandle, NVAPI_GPU_POWER_TOPO*); // EDCF624E 1-0048
NvAPI_Status NvAPI_DLL_ClientThermalPoliciesGetInfo(NvPhysicalGpuHandle, NVAPI_GPU_THERMAL_INFO*);
NvAPI_Status NvAPI_DLL_ClientThermalPoliciesGetLimit(NvPhysicalGpuHandle, NVAPI_GPU_THERMAL_LIMIT*);
NvAPI_Status NvAPI_DLL_ClientThermalPoliciesSetLimit(NvPhysicalGpuHandle, NVAPI_GPU_THERMAL_LIMIT*);
// Pascal GTX only
NvAPI_Status NvAPI_DLL_GetClockBoostRanges(NvPhysicalGpuHandle, NVAPI_CLOCKS_RANGE*);
NvAPI_Status NvAPI_DLL_GetClockBoostMask(NvPhysicalGpuHandle, NVAPI_CLOCK_MASKS*); // 0x507B4B59
NvAPI_Status NvAPI_DLL_GetClockBoostTable(NvPhysicalGpuHandle, NVAPI_CLOCK_TABLE*); // 0x23F1B133
NvAPI_Status NvAPI_DLL_SetClockBoostTable(NvPhysicalGpuHandle, NVAPI_CLOCK_TABLE*); // 0x0733E009
NvAPI_Status NvAPI_DLL_GetVFPCurve(NvPhysicalGpuHandle, NVAPI_VFP_CURVE*); // 0x21537AD4
NvAPI_Status NvAPI_DLL_GetCurrentVoltage(NvPhysicalGpuHandle, NVAPI_VOLTAGE_STATUS*); // 0x465F9BCF 1-004c
NvAPI_Status NvAPI_DLL_GetCoreVoltageBoostPercent(NvPhysicalGpuHandle, NVAPI_VOLTBOOST_PERCENT*);
NvAPI_Status NvAPI_DLL_SetCoreVoltageBoostPercent(NvPhysicalGpuHandle, NVAPI_VOLTBOOST_PERCENT*);
// Maxwell only
NvAPI_Status NvAPI_DLL_GetVoltageDomainsStatus(NvPhysicalGpuHandle, NVAPI_VOLT_STATUS*); // 0xC16C7E2C
NvAPI_Status NvAPI_DLL_GetVoltages(NvPhysicalGpuHandle, NVAPI_VOLTAGES_TABLE*); // 0x7D656244 1-40CC
NvAPI_Status NvAPI_DLL_GetVoltageStep(NvPhysicalGpuHandle, NVAPI_VOLT_STATUS*); // 0x28766157 1-008C unsure of the name
NvAPI_Status NvAPI_DLL_GetCoolerSettings(NvPhysicalGpuHandle, uint32_t, NVAPI_COOLER_SETTINGS*); // 0xDA141340 4-0558
NvAPI_Status NvAPI_DLL_SetCoolerLevels(NvPhysicalGpuHandle, uint32_t, NVAPI_COOLER_LEVEL*); // 0x891FA0AE 1-00A4
NvAPI_Status NvAPI_DLL_RestoreCoolerSettings(NvPhysicalGpuHandle, NVAPI_COOLER_SETTINGS*, uint32_t);
NvAPI_Status NvAPI_DLL_GetSerialNumber(NvPhysicalGpuHandle, NvAPI_ShortString serial);
NvAPI_Status NvAPI_DLL_GetPerfClocks(NvPhysicalGpuHandle, uint32_t num, NVAPI_GPU_PERF_CLOCKS* pClocks); // 2-2A74
//NvAPI_Status NvAPI_DLL_SetPerfClocks(NvPhysicalGpuHandle, uint32_t num, NVAPI_GPU_PERF_CLOCKS* pClocks); // error
//invalid..
//NvAPI_Status NvAPI_DLL_GetPstateClientLimits(NvPhysicalGpuHandle, NV_GPU_PERF_PSTATE_ID, uint32_t* pLimits);
//NvAPI_Status NvAPI_DLL_SetPstateClientLimits(NvPhysicalGpuHandle, NV_GPU_PERF_PSTATE_ID, uint32_t* pLimits);
NvAPI_Status NvAPI_DLL_SetPstates20v1(NvPhysicalGpuHandle handle, NV_GPU_PERF_PSTATES20_INFO_V1 *pSet);
NvAPI_Status NvAPI_DLL_SetPstates20v2(NvPhysicalGpuHandle handle, NV_GPU_PERF_PSTATES20_INFO_V2 *pSet);
NvAPI_Status NvAPI_DLL_Unload();
#define NV_ASSERT(x) { NvAPI_Status ret = x; if(ret != NVAPI_OK) return ret; }
// to reduce stack size, allow to reuse a mem buffer
#define NV_INIT_STRUCT_ON(TYPE, var, mem) { \
var = (TYPE*) mem; \
memset(var, 0, sizeof(TYPE)); \
var->version = TYPE##_VER; \
}
// alloc a struct, need free(var)
#define NV_INIT_STRUCT_ALLOC(TYPE, var) { \
var = (TYPE*) calloc(1, TYPE##_VER & 0xFFFF); \
if (var) var->version = TYPE##_VER; \
}
//! Used in NvAPI_I2CReadEx()
typedef struct
{
NvU32 version;
NvU32 displayMask; // Display Mask of the concerned display.
NvU8 bIsDDCPort; // indicates either the DDC port (TRUE) or the communication port (FALSE) of the concerned display.
NvU8 i2cDevAddress; // address of the I2C slave. The address should be shifted left by one. 0x50 -> 0xA0.
NvU8* pbI2cRegAddress; // I2C target register address. May be NULL, which indicates no register address should be sent.
NvU32 regAddrSize; // size in bytes of target register address. If pbI2cRegAddress is NULL, this field must be 0.
NvU8* pbData; // buffer of data which is to be read or written (depending on the command).
NvU32 cbRead; // bytes to read ??? seems required on write too
NvU32 cbSize; // full size of the data buffer, pbData, to be read or written.
NV_I2C_SPEED i2cSpeedKhz; // target speed of the transaction in (kHz) (Chosen from the enum NV_I2C_SPEED).
NvU8 portId; // portid on which device is connected (remember to set bIsPortIdSet if this value is set)
NvU32 bIsPortIdSet; // set this flag on if and only if portid value is set
} NV_I2C_INFO_EX;
#define NV_I2C_INFO_EX_VER MAKE_NVAPI_VERSION(NV_I2C_INFO_EX,3)
/*
sample evga x64 call (struct of 0x40 bytes)
ReadEx
$ ==> 40 00 03 00 00 00 00 00 00 40 00 00 00 00 00 00
$+10 58 F9 2B 00 00 00 00 00 01 00 00 00 00 00 00 00
$+20 C0 F9 2B 00 00 00 00 00 02 00 00 00 FF FF 00 00
$+30 00 00 00 00 02 00 00 00 01 00 00 00 00 00 00 00
$ ==> 40 00 03 00 00 00 00 00 00 10 00 00 00 00 00 00
$+10 68 F9 2B 00 00 00 00 00 01 00 00 00 00 00 00 00
$+20 C0 F9 2B 00 00 00 00 00 01 00 00 00 FF FF 00 00
$+30 00 00 00 00 01 00 00 00 01 00 00 00 00 00 00 00
00000000002BF968 > 75 83 CF 3F 01 00 00 00
00000000002BF9C0 > 0
WriteEx
$ ==> 40 00 03 00 00 00 00 00 00 8C 00 00 00 00 00 00
$+10 30 F9 2B 00 00 00 00 00 01 00 00 00 00 00 00 00
$+20 38 F9 2B 00 00 00 00 00 02 00 00 00 FF FF 00 00
$+30 00 00 00 00 01 00 00 00 01 00 00 00 00 00 00 00
00000000002BF930 > D1 00 00 00 00 00 00 00
00000000002BF938 > 38 00 00 00 00 00 00 00
*/
NvAPI_Status NvAPI_DLL_I2CReadEx(NvPhysicalGpuHandle, NV_I2C_INFO_EX*, NvU32*);
NvAPI_Status NvAPI_DLL_I2CWriteEx(NvPhysicalGpuHandle, NV_I2C_INFO_EX*, NvU32*);
+78 -6
View File
@@ -1,3 +1,36 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
#pragma once
#include"nvapi_lite_salstart.h"
#pragma pack(push,8)
@@ -20,6 +53,7 @@ extern "C" {
#endif
#endif
/* 64-bit types for compilers that support them, plus some obsolete variants */
#if defined(__GNUC__) || defined(__arm) || defined(__IAR_SYSTEMS_ICC__) || defined(__ghs__) || defined(_WIN64)
typedef unsigned long long NvU64; /* 0 to 18446744073709551615 */
@@ -36,6 +70,7 @@ typedef signed long NvS32; /* -2147483648 to 2147483647 */
typedef signed int NvS32; /* -2147483648 to 2147483647 */
#endif
#ifndef __unix
// mac os 32-bit still needs this
#if ( (defined(macintosh) && defined(__LP64__) && (__NVAPI_RESERVED0__)) || \
(!defined(macintosh) && defined(__NVAPI_RESERVED0__)) )
@@ -43,12 +78,22 @@ typedef unsigned int NvU32; /* 0 to 4294967295 */
#else
typedef unsigned long NvU32; /* 0 to 4294967295 */
#endif
#else
typedef unsigned int NvU32; /* 0 to 4294967295 */
#endif
typedef unsigned long temp_NvU32; /* 0 to 4294967295 */
typedef signed short NvS16;
typedef unsigned short NvU16;
typedef unsigned char NvU8;
typedef signed char NvS8;
/* Boolean type */
typedef NvU8 NvBool;
#define NV_TRUE ((NvBool)(0 == 0))
#define NV_FALSE ((NvBool)(0 != 0))
typedef struct _NV_RECT
{
NvU32 left;
@@ -299,11 +344,20 @@ typedef enum _NvAPI_Status
NVAPI_TIMEOUT = -191, //!< Requested operation timed out
NVAPI_GPU_WORKSTATION_FEATURE_INCOMPLETE = -192, //!< The requested workstation feature set has incomplete driver internal allocation resources
NVAPI_STEREO_INIT_ACTIVATION_NOT_DONE = -193, //!< Call failed because InitActivation was not called.
NVAPI_SYNC_NOT_ACTIVE = -194, //!< The requested action cannot be performed without Sync being enabled.
NVAPI_SYNC_MASTER_NOT_FOUND = -195, //!< The requested action cannot be performed without Sync Master being enabled.
NVAPI_INVALID_SYNC_TOPOLOGY = -196, //!< Invalid displays passed in the NV_GSYNC_DISPLAY pointer.
NVAPI_SYNC_NOT_ACTIVE = -194, //!< The requested action cannot be performed without Sync being enabled.
NVAPI_SYNC_MASTER_NOT_FOUND = -195, //!< The requested action cannot be performed without Sync Master being enabled.
NVAPI_INVALID_SYNC_TOPOLOGY = -196, //!< Invalid displays passed in the NV_GSYNC_DISPLAY pointer.
NVAPI_ECID_SIGN_ALGO_UNSUPPORTED = -197, //!< The specified signing algorithm is not supported. Either an incorrect value was entered or the current installed driver/hardware does not support the input value.
NVAPI_ECID_KEY_VERIFICATION_FAILED = -198, //!< The encrypted public key verification has failed.
NVAPI_ECID_KEY_VERIFICATION_FAILED = -198, //!< The encrypted public key verification has failed.
NVAPI_FIRMWARE_OUT_OF_DATE = -199, //!< The device's firmware is out of date.
NVAPI_FIRMWARE_REVISION_NOT_SUPPORTED = -200, //!< The device's firmware is not supported.
NVAPI_LICENSE_CALLER_AUTHENTICATION_FAILED = -201, //!< The caller is not authorized to modify the License.
NVAPI_D3D_DEVICE_NOT_REGISTERED = -202, //!< The user tried to use a deferred context without registering the device first
NVAPI_RESOURCE_NOT_ACQUIRED = -203, //!< Head or SourceId was not reserved for the VR Display before doing the Modeset.
NVAPI_TIMING_NOT_SUPPORTED = -204, //!< Provided timing is not supported.
NVAPI_HDCP_ENCRYPTION_FAILED = -205, //!< HDCP Encryption Failed for the device. Would be applicable when the device is HDCP Capable.
NVAPI_PCLK_LIMITATION_FAILED = -206, //!< Provided mode is over sink device pclk limitation.
NVAPI_NO_CONNECTOR_FOUND = -207, //!< No connector on GPU found.
} NvAPI_Status;
///////////////////////////////////////////////////////////////////////////////
@@ -352,9 +406,23 @@ typedef struct
} NV_DISPLAY_DRIVER_MEMORY_INFO_V2;
//! \ingroup driverapi
//! Used in NvAPI_GPU_GetMemoryInfo().
typedef struct
{
NvU32 version; //!< Version info
NvU32 dedicatedVideoMemory; //!< Size(in kb) of the physical framebuffer.
NvU32 availableDedicatedVideoMemory; //!< Size(in kb) of the available physical framebuffer for allocating video memory surfaces.
NvU32 systemVideoMemory; //!< Size(in kb) of system memory the driver allocates at load time.
NvU32 sharedSystemMemory; //!< Size(in kb) of shared system memory that driver is allowed to commit for surfaces across all allocations.
NvU32 curAvailableDedicatedVideoMemory; //!< Size(in kb) of the current available physical framebuffer for allocating video memory surfaces.
NvU32 dedicatedVideoMemoryEvictionsSize; //!< Size(in kb) of the total size of memory released as a result of the evictions.
NvU32 dedicatedVideoMemoryEvictionCount; //!< Indicates the number of eviction events that caused an allocation to be removed from dedicated video memory to free GPU
//!< video memory to make room for other allocations.
} NV_DISPLAY_DRIVER_MEMORY_INFO_V3;
//! \ingroup driverapi
typedef NV_DISPLAY_DRIVER_MEMORY_INFO_V2 NV_DISPLAY_DRIVER_MEMORY_INFO;
typedef NV_DISPLAY_DRIVER_MEMORY_INFO_V3 NV_DISPLAY_DRIVER_MEMORY_INFO;
//! \ingroup driverapi
//! Macro for constructing the version field of NV_DISPLAY_DRIVER_MEMORY_INFO_V1
@@ -365,7 +433,11 @@ typedef NV_DISPLAY_DRIVER_MEMORY_INFO_V2 NV_DISPLAY_DRIVER_MEMORY_INFO;
#define NV_DISPLAY_DRIVER_MEMORY_INFO_VER_2 MAKE_NVAPI_VERSION(NV_DISPLAY_DRIVER_MEMORY_INFO_V2,2)
//! \ingroup driverapi
#define NV_DISPLAY_DRIVER_MEMORY_INFO_VER NV_DISPLAY_DRIVER_MEMORY_INFO_VER_2
//! Macro for constructing the version field of NV_DISPLAY_DRIVER_MEMORY_INFO_V3
#define NV_DISPLAY_DRIVER_MEMORY_INFO_VER_3 MAKE_NVAPI_VERSION(NV_DISPLAY_DRIVER_MEMORY_INFO_V3,3)
//! \ingroup driverapi
#define NV_DISPLAY_DRIVER_MEMORY_INFO_VER NV_DISPLAY_DRIVER_MEMORY_INFO_VER_3
+37
View File
@@ -1,3 +1,36 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
#pragma once
#include"nvapi_lite_salstart.h"
#include"nvapi_lite_common.h"
@@ -29,6 +62,10 @@ typedef enum
//! argument (D3D_FEATURE_LEVEL supported by the device) that the function fills in. This argument
//! can contain -1 (NVAPI_DEVICE_FEATURE_LEVEL_NULL), if the requested featureLevel is less than DirecX 10.0.
//!
//! NOTE: When NvAPI_D3D11_CreateDevice is called with 10+ feature level we have an issue on few set of
//! tesla hardware (G80/G84/G86/G92/G94/G96) which does not support all feature level 10+ functionality
//! e.g. calling driver with mismatch between RenderTarget and Depth Buffer. App developers should
//! take into consideration such limitation when using NVAPI on such tesla hardwares.
//! SUPPORTED OS: Windows Vista and higher
//!
//!
+33
View File
@@ -1,3 +1,36 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
#ifndef __NVAPI_EMPTY_SAL
#ifdef __nvapi_undef__ecount
#undef __ecount
+34
View File
@@ -1,3 +1,37 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
// ====================================================
// SAL related support
// ====================================================
+42 -1
View File
@@ -1,3 +1,37 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
#pragma once
#include"nvapi_lite_salstart.h"
#include"nvapi_lite_common.h"
@@ -93,6 +127,7 @@ typedef enum _NVAPI_D3D_SETRESOURCEHINT_CATEGORY
//
// NVAPI_D3D_SRH_SLI_APP_CONTROLLED_INTERFRAME_CONTENT_SYNC:
// NVAPI_D3D_SRH_SLI_ASK_FOR_BROADCAST_USING:
//! \ingroup dx
@@ -101,9 +136,15 @@ typedef enum _NVAPI_D3D_SETRESOURCEHINT_CATEGORY
//! Default value: 0 \n
//! Explanation: If the value is 1, the driver will not track any rendering operations that would mark this resource as dirty,
//! avoiding any form of synchronization across frames rendered in parallel in multiple GPUs in AFR mode.
//!
//! NVAPI_D3D_SRH_SLI_ASK_FOR_BROADCAST_USAGE: Valid values : 0 or 1 \n
//! Default value: 0 \n
//! Explanation: If the value is 1, the driver will try to perform operations which involved target resource in broadcast,
//! where its possible. Hint is static and must be set before resource starts using.
typedef enum _NVAPI_D3D_SETRESOURCEHINT_SLI
{
NVAPI_D3D_SRH_SLI_APP_CONTROLLED_INTERFRAME_CONTENT_SYNC = 1
NVAPI_D3D_SRH_SLI_APP_CONTROLLED_INTERFRAME_CONTENT_SYNC = 1,
NVAPI_D3D_SRH_SLI_ASK_FOR_BROADCAST_USAGE = 2
} NVAPI_D3D_SETRESOURCEHINT_SLI;
//! \ingroup dx
+34
View File
@@ -1,3 +1,37 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
#pragma once
#include"nvapi_lite_salstart.h"
#include"nvapi_lite_common.h"
+34
View File
@@ -1,3 +1,37 @@
/************************************************************************************************************************************\
|* *|
|* Copyright © 2012 NVIDIA Corporation. All rights reserved. *|
|* *|
|* NOTICE TO USER: *|
|* *|
|* This software is subject to NVIDIA ownership rights under U.S. and international Copyright laws. *|
|* *|
|* This software and the information contained herein are PROPRIETARY and CONFIDENTIAL to NVIDIA *|
|* and are being provided solely under the terms and conditions of an NVIDIA software license agreement. *|
|* Otherwise, you have no rights to use or access this software in any manner. *|
|* *|
|* If not covered by the applicable NVIDIA software license agreement: *|
|* NVIDIA MAKES NO REPRESENTATION ABOUT THE SUITABILITY OF THIS SOFTWARE FOR ANY PURPOSE. *|
|* IT IS PROVIDED "AS IS" WITHOUT EXPRESS OR IMPLIED WARRANTY OF ANY KIND. *|
|* NVIDIA DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, *|
|* INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY, NONINFRINGEMENT, AND FITNESS FOR A PARTICULAR PURPOSE. *|
|* IN NO EVENT SHALL NVIDIA BE LIABLE FOR ANY SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, *|
|* OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, *|
|* NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOURCE CODE. *|
|* *|
|* U.S. Government End Users. *|
|* This software is a "commercial item" as that term is defined at 48 C.F.R. 2.101 (OCT 1995), *|
|* consisting of "commercial computer software" and "commercial computer software documentation" *|
|* as such terms are used in 48 C.F.R. 12.212 (SEPT 1995) and is provided to the U.S. Government only as a commercial end item. *|
|* Consistent with 48 C.F.R.12.212 and 48 C.F.R. 227.7202-1 through 227.7202-4 (JUNE 1995), *|
|* all U.S. Government End Users acquire the software with only those rights set forth herein. *|
|* *|
|* Any use of this software in individual and commercial software must include, *|
|* in the user documentation and internal comments to the code, *|
|* the above Disclaimer (as applicable) and U.S. Government End Users Notice. *|
|* *|
\************************************************************************************************************************************/
#pragma once
#include"nvapi_lite_salstart.h"
#include"nvapi_lite_common.h"
Binary file not shown.
+4
View File
@@ -37,6 +37,10 @@
#if !defined( PTHREAD_H )
#define PTHREAD_H
#if !defined(HAVE_STRUCT_TIMESPEC) && _MSC_VER >= 1900
#define HAVE_STRUCT_TIMESPEC
#endif
/*
* See the README file for an explanation of the pthreads-win32 version
* numbering scheme and how the DLL is named etc.
+69
View File
@@ -0,0 +1,69 @@
/**
* Meant to improve clang 4 / macos compatibility (untested)
*/
#ifndef PTHREAD_BARRIER_H_
#define PTHREAD_BARRIER_H_
#include <pthread.h>
#include <errno.h>
typedef int pthread_barrierattr_t;
typedef struct
{
pthread_mutex_t mutex;
pthread_cond_t cond;
int count;
int tripCount;
} pthread_barrier_t;
int pthread_barrier_init(pthread_barrier_t *barrier, const pthread_barrierattr_t *attr, unsigned int count)
{
if(count == 0)
{
errno = EINVAL;
return -1;
}
if(pthread_mutex_init(&barrier->mutex, 0) < 0)
{
return -1;
}
if(pthread_cond_init(&barrier->cond, 0) < 0)
{
pthread_mutex_destroy(&barrier->mutex);
return -1;
}
barrier->tripCount = count;
barrier->count = 0;
return 0;
}
int pthread_barrier_destroy(pthread_barrier_t *barrier)
{
pthread_cond_destroy(&barrier->cond);
pthread_mutex_destroy(&barrier->mutex);
return 0;
}
int pthread_barrier_wait(pthread_barrier_t *barrier)
{
pthread_mutex_lock(&barrier->mutex);
++(barrier->count);
if(barrier->count >= barrier->tripCount)
{
barrier->count = 0;
pthread_cond_broadcast(&barrier->cond);
pthread_mutex_unlock(&barrier->mutex);
return 1;
}
else
{
pthread_cond_wait(&barrier->cond, &(barrier->mutex));
pthread_mutex_unlock(&barrier->mutex);
return 0;
}
}
#endif // PTHREAD_BARRIER_H_
+13 -3
View File
@@ -1,10 +1,10 @@
AC_INIT([ccminer], [1.7.6], [], [ccminer], [http://github.com/tpruvot/ccminer])
AC_INIT([ccminer], [2.3.1], [], [ccminer], [http://github.com/tpruvot/ccminer])
AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM
AC_CONFIG_SRCDIR([ccminer.cpp])
AM_INIT_AUTOMAKE([foreign subdir-objects])
AC_CONFIG_HEADERS([cpuminer-config.h])
AC_CONFIG_HEADERS([ccminer-config.h])
dnl Make sure anyone changing configure.ac/Makefile.am has a clue
AM_MAINTAINER_MODE
@@ -63,6 +63,10 @@ case $target in
PTHREAD_FLAGS=""
WS2_LIBS="-lws2_32"
;;
*86*-apple-darwin*)
have_osx=true
PTHREAD_FLAGS=""
;;
esac
AC_CHECK_LIB(jansson, json_loads, request_jansson=false, request_jansson=true)
@@ -73,11 +77,12 @@ AC_CHECK_LIB([pthread], [pthread_create], PTHREAD_LIBS="-lpthread",
))))
AC_CHECK_LIB([z],[gzopen], [], [])
AC_CHECK_LIB([ssl],[SSL_library_init], [], [AC_MSG_ERROR([OpenSSL library required])])
AC_CHECK_LIB([ssl],[SSL_free], [], [AC_MSG_ERROR([OpenSSL library required])])
AC_CHECK_LIB([crypto],[EVP_DigestFinal_ex], [], [AC_MSG_ERROR([OpenSSL library required])])
AM_CONDITIONAL([WANT_JANSSON], [test x$request_jansson = xtrue])
AM_CONDITIONAL([HAVE_WINDOWS], [test x$have_win32 = xtrue])
AM_CONDITIONAL([HAVE_OSX], [test x$have_osx = xtrue])
AM_CONDITIONAL([ARCH_x86], [test x$have_x86 = xtrue])
AM_CONDITIONAL([ARCH_x86_64], [test x$have_x86_64 = xtrue])
@@ -111,6 +116,11 @@ else
SUFFIX=""
fi
if test x$have_osx = xtrue
then
SUFFIX=""
fi
dnl Setup CUDA paths
AC_ARG_WITH([cuda],
[ --with-cuda=PATH prefix where cuda is installed [default=/usr/local/cuda]])
+1 -7
View File
@@ -1,10 +1,4 @@
# possible additional CUDA_CFLAGS
#-gencode=arch=compute_52,code=\"sm_52,compute_52\"
#-gencode=arch=compute_50,code=\"sm_50,compute_50\"
#-gencode=arch=compute_35,code=\"sm_35,compute_35\"
#-gencode=arch=compute_30,code=\"sm_30,compute_30\"
#--ptxas-options=\"-v -dlcm=cg\""
# To change the cuda arch, edit Makefile.am and run ./build.sh
extracflags="-march=native -D_REENTRANT -falign-functions=16 -falign-jumps=16 -falign-labels=16"
+170
View File
@@ -0,0 +1,170 @@
/*
---------------------------------------------------------------------------
Copyright (c) 1998-2013, Brian Gladman, Worcester, UK. All rights reserved.
The redistribution and use of this software (with or without changes)
is allowed without the payment of fees or royalties provided that:
source code distributions include the above copyright notice, this
list of conditions and the following disclaimer;
binary distributions include the above copyright notice, this list
of conditions and the following disclaimer in their documentation.
This software is provided 'as is' with no explicit or implied warranties
in respect of its operation, including, but not limited to, correctness
and fitness for purpose.
---------------------------------------------------------------------------
Issue Date: 20/12/2007
*/
#include <stdint.h>
#if defined(__cplusplus)
extern "C"
{
#endif
#define TABLE_ALIGN 32
#define WPOLY 0x011b
#define N_COLS 4
#define AES_BLOCK_SIZE 16
#define RC_LENGTH (5 * (AES_BLOCK_SIZE / 4 - 2))
#if defined(_MSC_VER)
#define ALIGN __declspec(align(TABLE_ALIGN))
#elif defined(__GNUC__)
#define ALIGN __attribute__ ((aligned(16)))
#else
#define ALIGN
#endif
#define rf1(r,c) (r)
#define word_in(x,c) (*((uint32_t*)(x)+(c)))
#define word_out(x,c,v) (*((uint32_t*)(x)+(c)) = (v))
#define s(x,c) x[c]
#define si(y,x,c) (s(y,c) = word_in(x, c))
#define so(y,x,c) word_out(y, c, s(x,c))
#define state_in(y,x) si(y,x,0); si(y,x,1); si(y,x,2); si(y,x,3)
#define state_out(y,x) so(y,x,0); so(y,x,1); so(y,x,2); so(y,x,3)
#define round(y,x,k) \
y[0] = (k)[0] ^ (t_fn[0][x[0] & 0xff] ^ t_fn[1][(x[1] >> 8) & 0xff] ^ t_fn[2][(x[2] >> 16) & 0xff] ^ t_fn[3][x[3] >> 24]); \
y[1] = (k)[1] ^ (t_fn[0][x[1] & 0xff] ^ t_fn[1][(x[2] >> 8) & 0xff] ^ t_fn[2][(x[3] >> 16) & 0xff] ^ t_fn[3][x[0] >> 24]); \
y[2] = (k)[2] ^ (t_fn[0][x[2] & 0xff] ^ t_fn[1][(x[3] >> 8) & 0xff] ^ t_fn[2][(x[0] >> 16) & 0xff] ^ t_fn[3][x[1] >> 24]); \
y[3] = (k)[3] ^ (t_fn[0][x[3] & 0xff] ^ t_fn[1][(x[0] >> 8) & 0xff] ^ t_fn[2][(x[1] >> 16) & 0xff] ^ t_fn[3][x[2] >> 24]);
#define to_byte(x) ((x) & 0xff)
#define bval(x,n) to_byte((x) >> (8 * (n)))
#define fwd_var(x,r,c)\
( r == 0 ? ( c == 0 ? s(x,0) : c == 1 ? s(x,1) : c == 2 ? s(x,2) : s(x,3))\
: r == 1 ? ( c == 0 ? s(x,1) : c == 1 ? s(x,2) : c == 2 ? s(x,3) : s(x,0))\
: r == 2 ? ( c == 0 ? s(x,2) : c == 1 ? s(x,3) : c == 2 ? s(x,0) : s(x,1))\
: ( c == 0 ? s(x,3) : c == 1 ? s(x,0) : c == 2 ? s(x,1) : s(x,2)))
#define fwd_rnd(y,x,k,c) (s(y,c) = (k)[c] ^ four_tables(x,t_use(f,n),fwd_var,rf1,c))
#define sb_data(w) {\
w(0x63), w(0x7c), w(0x77), w(0x7b), w(0xf2), w(0x6b), w(0x6f), w(0xc5),\
w(0x30), w(0x01), w(0x67), w(0x2b), w(0xfe), w(0xd7), w(0xab), w(0x76),\
w(0xca), w(0x82), w(0xc9), w(0x7d), w(0xfa), w(0x59), w(0x47), w(0xf0),\
w(0xad), w(0xd4), w(0xa2), w(0xaf), w(0x9c), w(0xa4), w(0x72), w(0xc0),\
w(0xb7), w(0xfd), w(0x93), w(0x26), w(0x36), w(0x3f), w(0xf7), w(0xcc),\
w(0x34), w(0xa5), w(0xe5), w(0xf1), w(0x71), w(0xd8), w(0x31), w(0x15),\
w(0x04), w(0xc7), w(0x23), w(0xc3), w(0x18), w(0x96), w(0x05), w(0x9a),\
w(0x07), w(0x12), w(0x80), w(0xe2), w(0xeb), w(0x27), w(0xb2), w(0x75),\
w(0x09), w(0x83), w(0x2c), w(0x1a), w(0x1b), w(0x6e), w(0x5a), w(0xa0),\
w(0x52), w(0x3b), w(0xd6), w(0xb3), w(0x29), w(0xe3), w(0x2f), w(0x84),\
w(0x53), w(0xd1), w(0x00), w(0xed), w(0x20), w(0xfc), w(0xb1), w(0x5b),\
w(0x6a), w(0xcb), w(0xbe), w(0x39), w(0x4a), w(0x4c), w(0x58), w(0xcf),\
w(0xd0), w(0xef), w(0xaa), w(0xfb), w(0x43), w(0x4d), w(0x33), w(0x85),\
w(0x45), w(0xf9), w(0x02), w(0x7f), w(0x50), w(0x3c), w(0x9f), w(0xa8),\
w(0x51), w(0xa3), w(0x40), w(0x8f), w(0x92), w(0x9d), w(0x38), w(0xf5),\
w(0xbc), w(0xb6), w(0xda), w(0x21), w(0x10), w(0xff), w(0xf3), w(0xd2),\
w(0xcd), w(0x0c), w(0x13), w(0xec), w(0x5f), w(0x97), w(0x44), w(0x17),\
w(0xc4), w(0xa7), w(0x7e), w(0x3d), w(0x64), w(0x5d), w(0x19), w(0x73),\
w(0x60), w(0x81), w(0x4f), w(0xdc), w(0x22), w(0x2a), w(0x90), w(0x88),\
w(0x46), w(0xee), w(0xb8), w(0x14), w(0xde), w(0x5e), w(0x0b), w(0xdb),\
w(0xe0), w(0x32), w(0x3a), w(0x0a), w(0x49), w(0x06), w(0x24), w(0x5c),\
w(0xc2), w(0xd3), w(0xac), w(0x62), w(0x91), w(0x95), w(0xe4), w(0x79),\
w(0xe7), w(0xc8), w(0x37), w(0x6d), w(0x8d), w(0xd5), w(0x4e), w(0xa9),\
w(0x6c), w(0x56), w(0xf4), w(0xea), w(0x65), w(0x7a), w(0xae), w(0x08),\
w(0xba), w(0x78), w(0x25), w(0x2e), w(0x1c), w(0xa6), w(0xb4), w(0xc6),\
w(0xe8), w(0xdd), w(0x74), w(0x1f), w(0x4b), w(0xbd), w(0x8b), w(0x8a),\
w(0x70), w(0x3e), w(0xb5), w(0x66), w(0x48), w(0x03), w(0xf6), w(0x0e),\
w(0x61), w(0x35), w(0x57), w(0xb9), w(0x86), w(0xc1), w(0x1d), w(0x9e),\
w(0xe1), w(0xf8), w(0x98), w(0x11), w(0x69), w(0xd9), w(0x8e), w(0x94),\
w(0x9b), w(0x1e), w(0x87), w(0xe9), w(0xce), w(0x55), w(0x28), w(0xdf),\
w(0x8c), w(0xa1), w(0x89), w(0x0d), w(0xbf), w(0xe6), w(0x42), w(0x68),\
w(0x41), w(0x99), w(0x2d), w(0x0f), w(0xb0), w(0x54), w(0xbb), w(0x16) }
#define rc_data(w) {\
w(0x01), w(0x02), w(0x04), w(0x08), w(0x10),w(0x20), w(0x40), w(0x80),\
w(0x1b), w(0x36) }
#define bytes2word(b0, b1, b2, b3) (((uint32_t)(b3) << 24) | \
((uint32_t)(b2) << 16) | ((uint32_t)(b1) << 8) | (b0))
#define h0(x) (x)
#define w0(p) bytes2word(p, 0, 0, 0)
#define w1(p) bytes2word(0, p, 0, 0)
#define w2(p) bytes2word(0, 0, p, 0)
#define w3(p) bytes2word(0, 0, 0, p)
#define u0(p) bytes2word(f2(p), p, p, f3(p))
#define u1(p) bytes2word(f3(p), f2(p), p, p)
#define u2(p) bytes2word(p, f3(p), f2(p), p)
#define u3(p) bytes2word(p, p, f3(p), f2(p))
#define v0(p) bytes2word(fe(p), f9(p), fd(p), fb(p))
#define v1(p) bytes2word(fb(p), fe(p), f9(p), fd(p))
#define v2(p) bytes2word(fd(p), fb(p), fe(p), f9(p))
#define v3(p) bytes2word(f9(p), fd(p), fb(p), fe(p))
#define f2(x) ((x<<1) ^ (((x>>7) & 1) * WPOLY))
#define f4(x) ((x<<2) ^ (((x>>6) & 1) * WPOLY) ^ (((x>>6) & 2) * WPOLY))
#define f8(x) ((x<<3) ^ (((x>>5) & 1) * WPOLY) ^ (((x>>5) & 2) * WPOLY) ^ (((x>>5) & 4) * WPOLY))
#define f3(x) (f2(x) ^ x)
#define f9(x) (f8(x) ^ x)
#define fb(x) (f8(x) ^ f2(x) ^ x)
#define fd(x) (f8(x) ^ f4(x) ^ x)
#define fe(x) (f8(x) ^ f4(x) ^ f2(x))
#define t_dec(m,n) t_##m##n
#define t_set(m,n) t_##m##n
#define t_use(m,n) t_##m##n
#define d_4(t,n,b,e,f,g,h) ALIGN const t n[4][256] = { b(e), b(f), b(g), b(h) }
#define four_tables(x,tab,vf,rf,c) \
(tab[0][bval(vf(x,0,c),rf(0,c))] \
^ tab[1][bval(vf(x,1,c),rf(1,c))] \
^ tab[2][bval(vf(x,2,c),rf(2,c))] \
^ tab[3][bval(vf(x,3,c),rf(3,c))])
d_4(uint32_t, t_dec(f,n), sb_data, u0, u1, u2, u3);
void aesb_single_round(const uint8_t *in, uint8_t *out, uint8_t *expandedKey)
{
round(((uint32_t*) out), ((uint32_t*) in), ((uint32_t*) expandedKey));
}
void aesb_pseudo_round_mut(uint8_t *val, uint8_t *expandedKey)
{
uint32_t b1[4];
round(b1, ((uint32_t*) val), ((const uint32_t *) expandedKey));
round(((uint32_t*) val), b1, ((const uint32_t *) expandedKey) + 1 * N_COLS);
round(b1, ((uint32_t*) val), ((const uint32_t *) expandedKey) + 2 * N_COLS);
round(((uint32_t*) val), b1, ((const uint32_t *) expandedKey) + 3 * N_COLS);
round(b1, ((uint32_t*) val), ((const uint32_t *) expandedKey) + 4 * N_COLS);
round(((uint32_t*) val), b1, ((const uint32_t *) expandedKey) + 5 * N_COLS);
round(b1, ((uint32_t*) val), ((const uint32_t *) expandedKey) + 6 * N_COLS);
round(((uint32_t*) val), b1, ((const uint32_t *) expandedKey) + 7 * N_COLS);
round(b1, ((uint32_t*) val), ((const uint32_t *) expandedKey) + 8 * N_COLS);
round(((uint32_t*) val), b1, ((const uint32_t *) expandedKey) + 9 * N_COLS);
}
#if defined(__cplusplus)
}
#endif
+990
View File
@@ -0,0 +1,990 @@
#define N_COLS 4
#define WPOLY 0x011b
/*
static const __device__ __align__(16) uint32_t d_t_fn[1024] = {
0xa56363c6U, 0x847c7cf8U, 0x997777eeU, 0x8d7b7bf6U, 0x0df2f2ffU, 0xbd6b6bd6U, 0xb16f6fdeU, 0x54c5c591U,
0x50303060U, 0x03010102U, 0xa96767ceU, 0x7d2b2b56U, 0x19fefee7U, 0x62d7d7b5U, 0xe6abab4dU, 0x9a7676ecU,
0x45caca8fU, 0x9d82821fU, 0x40c9c989U, 0x877d7dfaU, 0x15fafaefU, 0xeb5959b2U, 0xc947478eU, 0x0bf0f0fbU,
0xecadad41U, 0x67d4d4b3U, 0xfda2a25fU, 0xeaafaf45U, 0xbf9c9c23U, 0xf7a4a453U, 0x967272e4U, 0x5bc0c09bU,
0xc2b7b775U, 0x1cfdfde1U, 0xae93933dU, 0x6a26264cU, 0x5a36366cU, 0x413f3f7eU, 0x02f7f7f5U, 0x4fcccc83U,
0x5c343468U, 0xf4a5a551U, 0x34e5e5d1U, 0x08f1f1f9U, 0x937171e2U, 0x73d8d8abU, 0x53313162U, 0x3f15152aU,
0x0c040408U, 0x52c7c795U, 0x65232346U, 0x5ec3c39dU, 0x28181830U, 0xa1969637U, 0x0f05050aU, 0xb59a9a2fU,
0x0907070eU, 0x36121224U, 0x9b80801bU, 0x3de2e2dfU, 0x26ebebcdU, 0x6927274eU, 0xcdb2b27fU, 0x9f7575eaU,
0x1b090912U, 0x9e83831dU, 0x742c2c58U, 0x2e1a1a34U, 0x2d1b1b36U, 0xb26e6edcU, 0xee5a5ab4U, 0xfba0a05bU,
0xf65252a4U, 0x4d3b3b76U, 0x61d6d6b7U, 0xceb3b37dU, 0x7b292952U, 0x3ee3e3ddU, 0x712f2f5eU, 0x97848413U,
0xf55353a6U, 0x68d1d1b9U, 0x00000000U, 0x2cededc1U, 0x60202040U, 0x1ffcfce3U, 0xc8b1b179U, 0xed5b5bb6U,
0xbe6a6ad4U, 0x46cbcb8dU, 0xd9bebe67U, 0x4b393972U, 0xde4a4a94U, 0xd44c4c98U, 0xe85858b0U, 0x4acfcf85U,
0x6bd0d0bbU, 0x2aefefc5U, 0xe5aaaa4fU, 0x16fbfbedU, 0xc5434386U, 0xd74d4d9aU, 0x55333366U, 0x94858511U,
0xcf45458aU, 0x10f9f9e9U, 0x06020204U, 0x817f7ffeU, 0xf05050a0U, 0x443c3c78U, 0xba9f9f25U, 0xe3a8a84bU,
0xf35151a2U, 0xfea3a35dU, 0xc0404080U, 0x8a8f8f05U, 0xad92923fU, 0xbc9d9d21U, 0x48383870U, 0x04f5f5f1U,
0xdfbcbc63U, 0xc1b6b677U, 0x75dadaafU, 0x63212142U, 0x30101020U, 0x1affffe5U, 0x0ef3f3fdU, 0x6dd2d2bfU,
0x4ccdcd81U, 0x140c0c18U, 0x35131326U, 0x2fececc3U, 0xe15f5fbeU, 0xa2979735U, 0xcc444488U, 0x3917172eU,
0x57c4c493U, 0xf2a7a755U, 0x827e7efcU, 0x473d3d7aU, 0xac6464c8U, 0xe75d5dbaU, 0x2b191932U, 0x957373e6U,
0xa06060c0U, 0x98818119U, 0xd14f4f9eU, 0x7fdcdca3U, 0x66222244U, 0x7e2a2a54U, 0xab90903bU, 0x8388880bU,
0xca46468cU, 0x29eeeec7U, 0xd3b8b86bU, 0x3c141428U, 0x79dedea7U, 0xe25e5ebcU, 0x1d0b0b16U, 0x76dbdbadU,
0x3be0e0dbU, 0x56323264U, 0x4e3a3a74U, 0x1e0a0a14U, 0xdb494992U, 0x0a06060cU, 0x6c242448U, 0xe45c5cb8U,
0x5dc2c29fU, 0x6ed3d3bdU, 0xefacac43U, 0xa66262c4U, 0xa8919139U, 0xa4959531U, 0x37e4e4d3U, 0x8b7979f2U,
0x32e7e7d5U, 0x43c8c88bU, 0x5937376eU, 0xb76d6ddaU, 0x8c8d8d01U, 0x64d5d5b1U, 0xd24e4e9cU, 0xe0a9a949U,
0xb46c6cd8U, 0xfa5656acU, 0x07f4f4f3U, 0x25eaeacfU, 0xaf6565caU, 0x8e7a7af4U, 0xe9aeae47U, 0x18080810U,
0xd5baba6fU, 0x887878f0U, 0x6f25254aU, 0x722e2e5cU, 0x241c1c38U, 0xf1a6a657U, 0xc7b4b473U, 0x51c6c697U,
0x23e8e8cbU, 0x7cdddda1U, 0x9c7474e8U, 0x211f1f3eU, 0xdd4b4b96U, 0xdcbdbd61U, 0x868b8b0dU, 0x858a8a0fU,
0x907070e0U, 0x423e3e7cU, 0xc4b5b571U, 0xaa6666ccU, 0xd8484890U, 0x05030306U, 0x01f6f6f7U, 0x120e0e1cU,
0xa36161c2U, 0x5f35356aU, 0xf95757aeU, 0xd0b9b969U, 0x91868617U, 0x58c1c199U, 0x271d1d3aU, 0xb99e9e27U,
0x38e1e1d9U, 0x13f8f8ebU, 0xb398982bU, 0x33111122U, 0xbb6969d2U, 0x70d9d9a9U, 0x898e8e07U, 0xa7949433U,
0xb69b9b2dU, 0x221e1e3cU, 0x92878715U, 0x20e9e9c9U, 0x49cece87U, 0xff5555aaU, 0x78282850U, 0x7adfdfa5U,
0x8f8c8c03U, 0xf8a1a159U, 0x80898909U, 0x170d0d1aU, 0xdabfbf65U, 0x31e6e6d7U, 0xc6424284U, 0xb86868d0U,
0xc3414182U, 0xb0999929U, 0x772d2d5aU, 0x110f0f1eU, 0xcbb0b07bU, 0xfc5454a8U, 0xd6bbbb6dU, 0x3a16162cU,
0x6363c6a5U, 0x7c7cf884U, 0x7777ee99U, 0x7b7bf68dU, 0xf2f2ff0dU, 0x6b6bd6bdU, 0x6f6fdeb1U, 0xc5c59154U,
0x30306050U, 0x01010203U, 0x6767cea9U, 0x2b2b567dU, 0xfefee719U, 0xd7d7b562U, 0xabab4de6U, 0x7676ec9aU,
0xcaca8f45U, 0x82821f9dU, 0xc9c98940U, 0x7d7dfa87U, 0xfafaef15U, 0x5959b2ebU, 0x47478ec9U, 0xf0f0fb0bU,
0xadad41ecU, 0xd4d4b367U, 0xa2a25ffdU, 0xafaf45eaU, 0x9c9c23bfU, 0xa4a453f7U, 0x7272e496U, 0xc0c09b5bU,
0xb7b775c2U, 0xfdfde11cU, 0x93933daeU, 0x26264c6aU, 0x36366c5aU, 0x3f3f7e41U, 0xf7f7f502U, 0xcccc834fU,
0x3434685cU, 0xa5a551f4U, 0xe5e5d134U, 0xf1f1f908U, 0x7171e293U, 0xd8d8ab73U, 0x31316253U, 0x15152a3fU,
0x0404080cU, 0xc7c79552U, 0x23234665U, 0xc3c39d5eU, 0x18183028U, 0x969637a1U, 0x05050a0fU, 0x9a9a2fb5U,
0x07070e09U, 0x12122436U, 0x80801b9bU, 0xe2e2df3dU, 0xebebcd26U, 0x27274e69U, 0xb2b27fcdU, 0x7575ea9fU,
0x0909121bU, 0x83831d9eU, 0x2c2c5874U, 0x1a1a342eU, 0x1b1b362dU, 0x6e6edcb2U, 0x5a5ab4eeU, 0xa0a05bfbU,
0x5252a4f6U, 0x3b3b764dU, 0xd6d6b761U, 0xb3b37dceU, 0x2929527bU, 0xe3e3dd3eU, 0x2f2f5e71U, 0x84841397U,
0x5353a6f5U, 0xd1d1b968U, 0x00000000U, 0xededc12cU, 0x20204060U, 0xfcfce31fU, 0xb1b179c8U, 0x5b5bb6edU,
0x6a6ad4beU, 0xcbcb8d46U, 0xbebe67d9U, 0x3939724bU, 0x4a4a94deU, 0x4c4c98d4U, 0x5858b0e8U, 0xcfcf854aU,
0xd0d0bb6bU, 0xefefc52aU, 0xaaaa4fe5U, 0xfbfbed16U, 0x434386c5U, 0x4d4d9ad7U, 0x33336655U, 0x85851194U,
0x45458acfU, 0xf9f9e910U, 0x02020406U, 0x7f7ffe81U, 0x5050a0f0U, 0x3c3c7844U, 0x9f9f25baU, 0xa8a84be3U,
0x5151a2f3U, 0xa3a35dfeU, 0x404080c0U, 0x8f8f058aU, 0x92923fadU, 0x9d9d21bcU, 0x38387048U, 0xf5f5f104U,
0xbcbc63dfU, 0xb6b677c1U, 0xdadaaf75U, 0x21214263U, 0x10102030U, 0xffffe51aU, 0xf3f3fd0eU, 0xd2d2bf6dU,
0xcdcd814cU, 0x0c0c1814U, 0x13132635U, 0xececc32fU, 0x5f5fbee1U, 0x979735a2U, 0x444488ccU, 0x17172e39U,
0xc4c49357U, 0xa7a755f2U, 0x7e7efc82U, 0x3d3d7a47U, 0x6464c8acU, 0x5d5dbae7U, 0x1919322bU, 0x7373e695U,
0x6060c0a0U, 0x81811998U, 0x4f4f9ed1U, 0xdcdca37fU, 0x22224466U, 0x2a2a547eU, 0x90903babU, 0x88880b83U,
0x46468ccaU, 0xeeeec729U, 0xb8b86bd3U, 0x1414283cU, 0xdedea779U, 0x5e5ebce2U, 0x0b0b161dU, 0xdbdbad76U,
0xe0e0db3bU, 0x32326456U, 0x3a3a744eU, 0x0a0a141eU, 0x494992dbU, 0x06060c0aU, 0x2424486cU, 0x5c5cb8e4U,
0xc2c29f5dU, 0xd3d3bd6eU, 0xacac43efU, 0x6262c4a6U, 0x919139a8U, 0x959531a4U, 0xe4e4d337U, 0x7979f28bU,
0xe7e7d532U, 0xc8c88b43U, 0x37376e59U, 0x6d6ddab7U, 0x8d8d018cU, 0xd5d5b164U, 0x4e4e9cd2U, 0xa9a949e0U,
0x6c6cd8b4U, 0x5656acfaU, 0xf4f4f307U, 0xeaeacf25U, 0x6565caafU, 0x7a7af48eU, 0xaeae47e9U, 0x08081018U,
0xbaba6fd5U, 0x7878f088U, 0x25254a6fU, 0x2e2e5c72U, 0x1c1c3824U, 0xa6a657f1U, 0xb4b473c7U, 0xc6c69751U,
0xe8e8cb23U, 0xdddda17cU, 0x7474e89cU, 0x1f1f3e21U, 0x4b4b96ddU, 0xbdbd61dcU, 0x8b8b0d86U, 0x8a8a0f85U,
0x7070e090U, 0x3e3e7c42U, 0xb5b571c4U, 0x6666ccaaU, 0x484890d8U, 0x03030605U, 0xf6f6f701U, 0x0e0e1c12U,
0x6161c2a3U, 0x35356a5fU, 0x5757aef9U, 0xb9b969d0U, 0x86861791U, 0xc1c19958U, 0x1d1d3a27U, 0x9e9e27b9U,
0xe1e1d938U, 0xf8f8eb13U, 0x98982bb3U, 0x11112233U, 0x6969d2bbU, 0xd9d9a970U, 0x8e8e0789U, 0x949433a7U,
0x9b9b2db6U, 0x1e1e3c22U, 0x87871592U, 0xe9e9c920U, 0xcece8749U, 0x5555aaffU, 0x28285078U, 0xdfdfa57aU,
0x8c8c038fU, 0xa1a159f8U, 0x89890980U, 0x0d0d1a17U, 0xbfbf65daU, 0xe6e6d731U, 0x424284c6U, 0x6868d0b8U,
0x414182c3U, 0x999929b0U, 0x2d2d5a77U, 0x0f0f1e11U, 0xb0b07bcbU, 0x5454a8fcU, 0xbbbb6dd6U, 0x16162c3aU,
0x63c6a563U, 0x7cf8847cU, 0x77ee9977U, 0x7bf68d7bU, 0xf2ff0df2U, 0x6bd6bd6bU, 0x6fdeb16fU, 0xc59154c5U,
0x30605030U, 0x01020301U, 0x67cea967U, 0x2b567d2bU, 0xfee719feU, 0xd7b562d7U, 0xab4de6abU, 0x76ec9a76U,
0xca8f45caU, 0x821f9d82U, 0xc98940c9U, 0x7dfa877dU, 0xfaef15faU, 0x59b2eb59U, 0x478ec947U, 0xf0fb0bf0U,
0xad41ecadU, 0xd4b367d4U, 0xa25ffda2U, 0xaf45eaafU, 0x9c23bf9cU, 0xa453f7a4U, 0x72e49672U, 0xc09b5bc0U,
0xb775c2b7U, 0xfde11cfdU, 0x933dae93U, 0x264c6a26U, 0x366c5a36U, 0x3f7e413fU, 0xf7f502f7U, 0xcc834fccU,
0x34685c34U, 0xa551f4a5U, 0xe5d134e5U, 0xf1f908f1U, 0x71e29371U, 0xd8ab73d8U, 0x31625331U, 0x152a3f15U,
0x04080c04U, 0xc79552c7U, 0x23466523U, 0xc39d5ec3U, 0x18302818U, 0x9637a196U, 0x050a0f05U, 0x9a2fb59aU,
0x070e0907U, 0x12243612U, 0x801b9b80U, 0xe2df3de2U, 0xebcd26ebU, 0x274e6927U, 0xb27fcdb2U, 0x75ea9f75U,
0x09121b09U, 0x831d9e83U, 0x2c58742cU, 0x1a342e1aU, 0x1b362d1bU, 0x6edcb26eU, 0x5ab4ee5aU, 0xa05bfba0U,
0x52a4f652U, 0x3b764d3bU, 0xd6b761d6U, 0xb37dceb3U, 0x29527b29U, 0xe3dd3ee3U, 0x2f5e712fU, 0x84139784U,
0x53a6f553U, 0xd1b968d1U, 0x00000000U, 0xedc12cedU, 0x20406020U, 0xfce31ffcU, 0xb179c8b1U, 0x5bb6ed5bU,
0x6ad4be6aU, 0xcb8d46cbU, 0xbe67d9beU, 0x39724b39U, 0x4a94de4aU, 0x4c98d44cU, 0x58b0e858U, 0xcf854acfU,
0xd0bb6bd0U, 0xefc52aefU, 0xaa4fe5aaU, 0xfbed16fbU, 0x4386c543U, 0x4d9ad74dU, 0x33665533U, 0x85119485U,
0x458acf45U, 0xf9e910f9U, 0x02040602U, 0x7ffe817fU, 0x50a0f050U, 0x3c78443cU, 0x9f25ba9fU, 0xa84be3a8U,
0x51a2f351U, 0xa35dfea3U, 0x4080c040U, 0x8f058a8fU, 0x923fad92U, 0x9d21bc9dU, 0x38704838U, 0xf5f104f5U,
0xbc63dfbcU, 0xb677c1b6U, 0xdaaf75daU, 0x21426321U, 0x10203010U, 0xffe51affU, 0xf3fd0ef3U, 0xd2bf6dd2U,
0xcd814ccdU, 0x0c18140cU, 0x13263513U, 0xecc32fecU, 0x5fbee15fU, 0x9735a297U, 0x4488cc44U, 0x172e3917U,
0xc49357c4U, 0xa755f2a7U, 0x7efc827eU, 0x3d7a473dU, 0x64c8ac64U, 0x5dbae75dU, 0x19322b19U, 0x73e69573U,
0x60c0a060U, 0x81199881U, 0x4f9ed14fU, 0xdca37fdcU, 0x22446622U, 0x2a547e2aU, 0x903bab90U, 0x880b8388U,
0x468cca46U, 0xeec729eeU, 0xb86bd3b8U, 0x14283c14U, 0xdea779deU, 0x5ebce25eU, 0x0b161d0bU, 0xdbad76dbU,
0xe0db3be0U, 0x32645632U, 0x3a744e3aU, 0x0a141e0aU, 0x4992db49U, 0x060c0a06U, 0x24486c24U, 0x5cb8e45cU,
0xc29f5dc2U, 0xd3bd6ed3U, 0xac43efacU, 0x62c4a662U, 0x9139a891U, 0x9531a495U, 0xe4d337e4U, 0x79f28b79U,
0xe7d532e7U, 0xc88b43c8U, 0x376e5937U, 0x6ddab76dU, 0x8d018c8dU, 0xd5b164d5U, 0x4e9cd24eU, 0xa949e0a9U,
0x6cd8b46cU, 0x56acfa56U, 0xf4f307f4U, 0xeacf25eaU, 0x65caaf65U, 0x7af48e7aU, 0xae47e9aeU, 0x08101808U,
0xba6fd5baU, 0x78f08878U, 0x254a6f25U, 0x2e5c722eU, 0x1c38241cU, 0xa657f1a6U, 0xb473c7b4U, 0xc69751c6U,
0xe8cb23e8U, 0xdda17cddU, 0x74e89c74U, 0x1f3e211fU, 0x4b96dd4bU, 0xbd61dcbdU, 0x8b0d868bU, 0x8a0f858aU,
0x70e09070U, 0x3e7c423eU, 0xb571c4b5U, 0x66ccaa66U, 0x4890d848U, 0x03060503U, 0xf6f701f6U, 0x0e1c120eU,
0x61c2a361U, 0x356a5f35U, 0x57aef957U, 0xb969d0b9U, 0x86179186U, 0xc19958c1U, 0x1d3a271dU, 0x9e27b99eU,
0xe1d938e1U, 0xf8eb13f8U, 0x982bb398U, 0x11223311U, 0x69d2bb69U, 0xd9a970d9U, 0x8e07898eU, 0x9433a794U,
0x9b2db69bU, 0x1e3c221eU, 0x87159287U, 0xe9c920e9U, 0xce8749ceU, 0x55aaff55U, 0x28507828U, 0xdfa57adfU,
0x8c038f8cU, 0xa159f8a1U, 0x89098089U, 0x0d1a170dU, 0xbf65dabfU, 0xe6d731e6U, 0x4284c642U, 0x68d0b868U,
0x4182c341U, 0x9929b099U, 0x2d5a772dU, 0x0f1e110fU, 0xb07bcbb0U, 0x54a8fc54U, 0xbb6dd6bbU, 0x162c3a16U,
0xc6a56363U, 0xf8847c7cU, 0xee997777U, 0xf68d7b7bU, 0xff0df2f2U, 0xd6bd6b6bU, 0xdeb16f6fU, 0x9154c5c5U,
0x60503030U, 0x02030101U, 0xcea96767U, 0x567d2b2bU, 0xe719fefeU, 0xb562d7d7U, 0x4de6ababU, 0xec9a7676U,
0x8f45cacaU, 0x1f9d8282U, 0x8940c9c9U, 0xfa877d7dU, 0xef15fafaU, 0xb2eb5959U, 0x8ec94747U, 0xfb0bf0f0U,
0x41ecadadU, 0xb367d4d4U, 0x5ffda2a2U, 0x45eaafafU, 0x23bf9c9cU, 0x53f7a4a4U, 0xe4967272U, 0x9b5bc0c0U,
0x75c2b7b7U, 0xe11cfdfdU, 0x3dae9393U, 0x4c6a2626U, 0x6c5a3636U, 0x7e413f3fU, 0xf502f7f7U, 0x834fccccU,
0x685c3434U, 0x51f4a5a5U, 0xd134e5e5U, 0xf908f1f1U, 0xe2937171U, 0xab73d8d8U, 0x62533131U, 0x2a3f1515U,
0x080c0404U, 0x9552c7c7U, 0x46652323U, 0x9d5ec3c3U, 0x30281818U, 0x37a19696U, 0x0a0f0505U, 0x2fb59a9aU,
0x0e090707U, 0x24361212U, 0x1b9b8080U, 0xdf3de2e2U, 0xcd26ebebU, 0x4e692727U, 0x7fcdb2b2U, 0xea9f7575U,
0x121b0909U, 0x1d9e8383U, 0x58742c2cU, 0x342e1a1aU, 0x362d1b1bU, 0xdcb26e6eU, 0xb4ee5a5aU, 0x5bfba0a0U,
0xa4f65252U, 0x764d3b3bU, 0xb761d6d6U, 0x7dceb3b3U, 0x527b2929U, 0xdd3ee3e3U, 0x5e712f2fU, 0x13978484U,
0xa6f55353U, 0xb968d1d1U, 0x00000000U, 0xc12cededU, 0x40602020U, 0xe31ffcfcU, 0x79c8b1b1U, 0xb6ed5b5bU,
0xd4be6a6aU, 0x8d46cbcbU, 0x67d9bebeU, 0x724b3939U, 0x94de4a4aU, 0x98d44c4cU, 0xb0e85858U, 0x854acfcfU,
0xbb6bd0d0U, 0xc52aefefU, 0x4fe5aaaaU, 0xed16fbfbU, 0x86c54343U, 0x9ad74d4dU, 0x66553333U, 0x11948585U,
0x8acf4545U, 0xe910f9f9U, 0x04060202U, 0xfe817f7fU, 0xa0f05050U, 0x78443c3cU, 0x25ba9f9fU, 0x4be3a8a8U,
0xa2f35151U, 0x5dfea3a3U, 0x80c04040U, 0x058a8f8fU, 0x3fad9292U, 0x21bc9d9dU, 0x70483838U, 0xf104f5f5U,
0x63dfbcbcU, 0x77c1b6b6U, 0xaf75dadaU, 0x42632121U, 0x20301010U, 0xe51affffU, 0xfd0ef3f3U, 0xbf6dd2d2U,
0x814ccdcdU, 0x18140c0cU, 0x26351313U, 0xc32fececU, 0xbee15f5fU, 0x35a29797U, 0x88cc4444U, 0x2e391717U,
0x9357c4c4U, 0x55f2a7a7U, 0xfc827e7eU, 0x7a473d3dU, 0xc8ac6464U, 0xbae75d5dU, 0x322b1919U, 0xe6957373U,
0xc0a06060U, 0x19988181U, 0x9ed14f4fU, 0xa37fdcdcU, 0x44662222U, 0x547e2a2aU, 0x3bab9090U, 0x0b838888U,
0x8cca4646U, 0xc729eeeeU, 0x6bd3b8b8U, 0x283c1414U, 0xa779dedeU, 0xbce25e5eU, 0x161d0b0bU, 0xad76dbdbU,
0xdb3be0e0U, 0x64563232U, 0x744e3a3aU, 0x141e0a0aU, 0x92db4949U, 0x0c0a0606U, 0x486c2424U, 0xb8e45c5cU,
0x9f5dc2c2U, 0xbd6ed3d3U, 0x43efacacU, 0xc4a66262U, 0x39a89191U, 0x31a49595U, 0xd337e4e4U, 0xf28b7979U,
0xd532e7e7U, 0x8b43c8c8U, 0x6e593737U, 0xdab76d6dU, 0x018c8d8dU, 0xb164d5d5U, 0x9cd24e4eU, 0x49e0a9a9U,
0xd8b46c6cU, 0xacfa5656U, 0xf307f4f4U, 0xcf25eaeaU, 0xcaaf6565U, 0xf48e7a7aU, 0x47e9aeaeU, 0x10180808U,
0x6fd5babaU, 0xf0887878U, 0x4a6f2525U, 0x5c722e2eU, 0x38241c1cU, 0x57f1a6a6U, 0x73c7b4b4U, 0x9751c6c6U,
0xcb23e8e8U, 0xa17cddddU, 0xe89c7474U, 0x3e211f1fU, 0x96dd4b4bU, 0x61dcbdbdU, 0x0d868b8bU, 0x0f858a8aU,
0xe0907070U, 0x7c423e3eU, 0x71c4b5b5U, 0xccaa6666U, 0x90d84848U, 0x06050303U, 0xf701f6f6U, 0x1c120e0eU,
0xc2a36161U, 0x6a5f3535U, 0xaef95757U, 0x69d0b9b9U, 0x17918686U, 0x9958c1c1U, 0x3a271d1dU, 0x27b99e9eU,
0xd938e1e1U, 0xeb13f8f8U, 0x2bb39898U, 0x22331111U, 0xd2bb6969U, 0xa970d9d9U, 0x07898e8eU, 0x33a79494U,
0x2db69b9bU, 0x3c221e1eU, 0x15928787U, 0xc920e9e9U, 0x8749ceceU, 0xaaff5555U, 0x50782828U, 0xa57adfdfU,
0x038f8c8cU, 0x59f8a1a1U, 0x09808989U, 0x1a170d0dU, 0x65dabfbfU, 0xd731e6e6U, 0x84c64242U, 0xd0b86868U,
0x82c34141U, 0x29b09999U, 0x5a772d2dU, 0x1e110f0fU, 0x7bcbb0b0U, 0xa8fc5454U, 0x6dd6bbbbU, 0x2c3a1616U
};
*/
#define AS_U32(addr) *((uint32_t*)(addr))
#define AS_U64(addr) *((uint64_t*)(addr))
#define AS_UINT2(addr) *((uint2*)(addr))
#define AS_UINT4(addr) *((uint4*)(addr))
#define AS_UL2(addr) *((ulonglong2*)(addr))
#define t_fn0(x) (sharedMemory[x])
#define t_fn1(x) (sharedMemory[0x100U | (x)])
#define t_fn2(x) (sharedMemory[0x200U | (x)])
#define t_fn3(x) (sharedMemory[0x300U | (x)])
#define round(shared, out, x, k) \
out[0] = (k)[0] ^ (t_fn0(x[0] & 0xff) ^ t_fn1((x[1] >> 8) & 0xff) ^ t_fn2((x[2] >> 16) & 0xff) ^ t_fn3((x[3] >> 24) & 0xff)); \
out[1] = (k)[1] ^ (t_fn0(x[1] & 0xff) ^ t_fn1((x[2] >> 8) & 0xff) ^ t_fn2((x[3] >> 16) & 0xff) ^ t_fn3((x[0] >> 24) & 0xff)); \
out[2] = (k)[2] ^ (t_fn0(x[2] & 0xff) ^ t_fn1((x[3] >> 8) & 0xff) ^ t_fn2((x[0] >> 16) & 0xff) ^ t_fn3((x[1] >> 24) & 0xff)); \
out[3] = (k)[3] ^ (t_fn0(x[3] & 0xff) ^ t_fn1((x[0] >> 8) & 0xff) ^ t_fn2((x[1] >> 16) & 0xff) ^ t_fn3((x[2] >> 24) & 0xff));
#define round_u4(shared, out, in, k) \
((uint32_t*)out)[0] = (k)[0] ^ t_fn0(in[0].x) ^ t_fn1(in[1].y) ^ t_fn2(in[2].z) ^ t_fn3(in[3].w); \
((uint32_t*)out)[1] = (k)[1] ^ t_fn0(in[1].x) ^ t_fn1(in[2].y) ^ t_fn2(in[3].z) ^ t_fn3(in[0].w); \
((uint32_t*)out)[2] = (k)[2] ^ t_fn0(in[2].x) ^ t_fn1(in[3].y) ^ t_fn2(in[0].z) ^ t_fn3(in[1].w); \
((uint32_t*)out)[3] = (k)[3] ^ t_fn0(in[3].x) ^ t_fn1(in[0].y) ^ t_fn2(in[1].z) ^ t_fn3(in[2].w);
#ifdef __INTELLISENSE__
#define __byte_perm(a,b,c) a
#endif
#define OFF32_0(x) (x & 0xFFu)
#define OFF32_1(x) __byte_perm(x, 0x01, 0x5541)
#define OFF32_2(x) __byte_perm(x, 0x02, 0x5542)
#define OFF32_3(x) __byte_perm(x, 0x03, 0x5543)
#define SHARED_0(x) sharedMemory[OFF32_0(x)]
#define SHARED_1(x) sharedMemory[OFF32_1(x)]
#define SHARED_2(x) sharedMemory[OFF32_2(x)]
#define SHARED_3(x) sharedMemory[OFF32_3(x)]
__device__ __forceinline__
void cn_aes_single_round(uint32_t * const sharedMemory, uint32_t * const in, uint32_t * out, uint32_t* expandedKey)
{
asm("// aes_single_round");
out[0] = expandedKey[0] ^ SHARED_0(in[0]) ^ SHARED_1(in[1]) ^ SHARED_2(in[2]) ^ SHARED_3(in[3]);
out[1] = expandedKey[1] ^ SHARED_0(in[1]) ^ SHARED_1(in[2]) ^ SHARED_2(in[3]) ^ SHARED_3(in[0]);
out[2] = expandedKey[2] ^ SHARED_0(in[2]) ^ SHARED_1(in[3]) ^ SHARED_2(in[0]) ^ SHARED_3(in[1]);
out[3] = expandedKey[3] ^ SHARED_0(in[3]) ^ SHARED_1(in[0]) ^ SHARED_2(in[1]) ^ SHARED_3(in[2]);
}
//
#ifdef _WIN64
/* do a mul.wide.u32 to prevent a shl + cvt 32 to 64 on ld.shared [ptr] */
#define OFF8_0(x) (x & 0xFFu) * sizeof(uint32_t)
#define OFF8_1(x) __byte_perm(x, 0x01, 0x5541) * sizeof(uint32_t)
#define OFF8_2(x) __byte_perm(x, 0x02, 0x5542) * sizeof(uint32_t)
#define OFF8_3(x) __byte_perm(x, 0x03, 0x5543) * sizeof(uint32_t)
#else
#define OFF8_0(x) (x & 0xFFu) << 2
#define OFF8_1(x) __byte_perm(x, 0x01, 0x5541) << 2
#define OFF8_2(x) __byte_perm(x, 0x02, 0x5542) << 2
#define OFF8_3(x) __byte_perm(x, 0x03, 0x5543) << 2
#endif
#define SHAR8_0(x) AS_U32(&sharedMemory[OFF8_0(x)])
#define SHAR8_1(x) AS_U32(&sharedMemory[OFF8_1(x)])
#define SHAR8_2(x) AS_U32(&sharedMemory[OFF8_2(x)])
#define SHAR8_3(x) AS_U32(&sharedMemory[OFF8_3(x)])
__device__ __forceinline__
void cn_aes_single_round_b(uint8_t * const sharedMemory, void * const long_state, const uint4 key, uint4 *res)
{
asm("// aes_single_round_b");
uint4 in = AS_UINT4(long_state);
*res = key;
res->x ^= SHAR8_0(in.x) ^ SHAR8_1(in.y) ^ SHAR8_2(in.z) ^ SHAR8_3(in.w);
res->y ^= SHAR8_0(in.y) ^ SHAR8_1(in.z) ^ SHAR8_2(in.w) ^ SHAR8_3(in.x);
res->z ^= SHAR8_0(in.z) ^ SHAR8_1(in.w) ^ SHAR8_2(in.x) ^ SHAR8_3(in.y);
res->w ^= SHAR8_0(in.w) ^ SHAR8_1(in.x) ^ SHAR8_2(in.y) ^ SHAR8_3(in.z);
}
#define round_perm(shared, out, in, k) \
out[0] = (k)[0] ^ SHARED_0(in[0]) ^ SHARED_1(in[1]) ^ SHARED_2(in[2]) ^ SHARED_3(in[3]); \
out[1] = (k)[1] ^ SHARED_0(in[1]) ^ SHARED_1(in[2]) ^ SHARED_2(in[3]) ^ SHARED_3(in[0]); \
out[2] = (k)[2] ^ SHARED_0(in[2]) ^ SHARED_1(in[3]) ^ SHARED_2(in[0]) ^ SHARED_3(in[1]); \
out[3] = (k)[3] ^ SHARED_0(in[3]) ^ SHARED_1(in[0]) ^ SHARED_2(in[1]) ^ SHARED_3(in[2]);
__device__ __forceinline__
void cn_aes_pseudo_round_mut(const uint32_t * sharedMemory, uint32_t * val, uint32_t const * expandedKey)
{
asm("// aes_pseudo_round_mut");
uint32_t b[4];
round_perm(sharedMemory, b, val, expandedKey);
round_perm(sharedMemory, val, b, expandedKey + (1 * N_COLS));
round_perm(sharedMemory, b, val, expandedKey + (2 * N_COLS));
round_perm(sharedMemory, val, b, expandedKey + (3 * N_COLS));
round_perm(sharedMemory, b, val, expandedKey + (4 * N_COLS));
round_perm(sharedMemory, val, b, expandedKey + (5 * N_COLS));
round_perm(sharedMemory, b, val, expandedKey + (6 * N_COLS));
round_perm(sharedMemory, val, b, expandedKey + (7 * N_COLS));
round_perm(sharedMemory, b, val, expandedKey + (8 * N_COLS));
round_perm(sharedMemory, val, b, expandedKey + (9 * N_COLS));
}
static __forceinline__ __device__ uint4 operator ^ (const uint4 &a, const uint4 &b) {
return make_uint4(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w);
}
#define round_perm4(in, k) {\
uint4 tmp; \
tmp.x = SHARED_0(in.x) ^ SHARED_1(in.y) ^ SHARED_2(in.z) ^ SHARED_3(in.w); \
tmp.y = SHARED_0(in.y) ^ SHARED_1(in.z) ^ SHARED_2(in.w) ^ SHARED_3(in.x); \
tmp.z = SHARED_0(in.z) ^ SHARED_1(in.w) ^ SHARED_2(in.x) ^ SHARED_3(in.y); \
tmp.w = SHARED_0(in.w) ^ SHARED_1(in.x) ^ SHARED_2(in.y) ^ SHARED_3(in.z); \
val = tmp ^ key[k]; \
}
__device__ __forceinline__
void cn_aes_pseudo_round_mut_uint4(uint32_t * const sharedMemory, uint4 &val, uint4 const key[10])
{
asm("// aes_pseudo_round_mut_uint4");
round_perm4(val, 0);
round_perm4(val, 1);
round_perm4(val, 2);
round_perm4(val, 3);
round_perm4(val, 4);
round_perm4(val, 5);
round_perm4(val, 6);
round_perm4(val, 7);
round_perm4(val, 8);
round_perm4(val, 9);
}
/*
__device__ __forceinline__
void cn_aes_gpu_init2(uint32_t* sharedMemory)
{
#if 0
if(blockDim.x >= 64)
{
if(threadIdx.x < 64) {
#define thrX (threadIdx.x << 2U) // ensure offsets aligned (16) to vector
#pragma unroll 4
for (uint32_t i = 0; i < 1024U; i += 256U) // 32x32 = 1024, 4 * 256 also
AS_UINT4(&sharedMemory[i + thrX]) = AS_UINT4(&d_t_fn[i + thrX]);
}
} else
#endif
if(blockDim.x >= 32) {
if(threadIdx.x < 32) {
#if 0
#pragma unroll 32
for(uint32_t i = 0; i < 1024; i += 32)
sharedMemory[threadIdx.x + i] = d_t_fn[threadIdx.x + i];
#else
#define thrX (threadIdx.x << 2U) // ensure offsets aligned (16) to vector
#pragma unroll 8
for (uint32_t i = 0; i < 1024; i += 128U) // 32x32 = 1024, 8 * 128 also
AS_UINT4(&sharedMemory[i + thrX]) = AS_UINT4(&d_t_fn[i + thrX]);
#endif
}
} else {
if(threadIdx.x < 4) {
#if 0
for (uint32_t i = 0; i < 1024; i += 4)
sharedMemory[threadIdx.x + i] = d_t_fn[threadIdx.x + i];
#else
#define thrX (threadIdx.x << 2U) // ensure offsets aligned (16) to vector
#pragma unroll 64
for (uint32_t i = 0; i < 1024; i += 16U)
AS_UINT4(&sharedMemory[i + thrX]) = AS_UINT4(&d_t_fn[i + thrX]);
#endif
}
}
}
*/
__device__ __forceinline__
void cn_aes_gpu_init(uint32_t* sharedMemory)
{
// AES 0
switch (threadIdx.x) {
case 0:
AS_UL2(&sharedMemory[0x000]) = make_ulonglong2(0x847c7cf8a56363c6, 0x8d7b7bf6997777ee);
AS_UL2(&sharedMemory[0x004]) = make_ulonglong2(0xbd6b6bd60df2f2ff, 0x54c5c591b16f6fde);
AS_UL2(&sharedMemory[0x008]) = make_ulonglong2(0x0301010250303060, 0x7d2b2b56a96767ce);
AS_UL2(&sharedMemory[0x00C]) = make_ulonglong2(0x62d7d7b519fefee7, 0x9a7676ece6abab4d);
AS_UL2(&sharedMemory[0x010]) = make_ulonglong2(0x9d82821f45caca8f, 0x877d7dfa40c9c989);
AS_UL2(&sharedMemory[0x014]) = make_ulonglong2(0xeb5959b215fafaef, 0x0bf0f0fbc947478e);
AS_UL2(&sharedMemory[0x018]) = make_ulonglong2(0x67d4d4b3ecadad41, 0xeaafaf45fda2a25f);
AS_UL2(&sharedMemory[0x01C]) = make_ulonglong2(0xf7a4a453bf9c9c23, 0x5bc0c09b967272e4);
break;
case 1:
AS_UL2(&sharedMemory[0x020]) = make_ulonglong2(0x1cfdfde1c2b7b775, 0x6a26264cae93933d);
AS_UL2(&sharedMemory[0x024]) = make_ulonglong2(0x413f3f7e5a36366c, 0x4fcccc8302f7f7f5);
AS_UL2(&sharedMemory[0x028]) = make_ulonglong2(0xf4a5a5515c343468, 0x08f1f1f934e5e5d1);
AS_UL2(&sharedMemory[0x02C]) = make_ulonglong2(0x73d8d8ab937171e2, 0x3f15152a53313162);
AS_UL2(&sharedMemory[0x030]) = make_ulonglong2(0x52c7c7950c040408, 0x5ec3c39d65232346);
AS_UL2(&sharedMemory[0x034]) = make_ulonglong2(0xa196963728181830, 0xb59a9a2f0f05050a);
AS_UL2(&sharedMemory[0x038]) = make_ulonglong2(0x361212240907070e, 0x3de2e2df9b80801b);
AS_UL2(&sharedMemory[0x03C]) = make_ulonglong2(0x6927274e26ebebcd, 0x9f7575eacdb2b27f);
break;
case 2:
AS_UL2(&sharedMemory[0x040]) = make_ulonglong2(0x9e83831d1b090912, 0x2e1a1a34742c2c58);
AS_UL2(&sharedMemory[0x044]) = make_ulonglong2(0xb26e6edc2d1b1b36, 0xfba0a05bee5a5ab4);
AS_UL2(&sharedMemory[0x048]) = make_ulonglong2(0x4d3b3b76f65252a4, 0xceb3b37d61d6d6b7);
AS_UL2(&sharedMemory[0x04C]) = make_ulonglong2(0x3ee3e3dd7b292952, 0x97848413712f2f5e);
AS_UL2(&sharedMemory[0x050]) = make_ulonglong2(0x68d1d1b9f55353a6, 0x2cededc100000000);
AS_UL2(&sharedMemory[0x054]) = make_ulonglong2(0x1ffcfce360202040, 0xed5b5bb6c8b1b179);
AS_UL2(&sharedMemory[0x058]) = make_ulonglong2(0x46cbcb8dbe6a6ad4, 0x4b393972d9bebe67);
AS_UL2(&sharedMemory[0x05C]) = make_ulonglong2(0xd44c4c98de4a4a94, 0x4acfcf85e85858b0);
break;
case 3:
AS_UL2(&sharedMemory[0x060]) = make_ulonglong2(0x2aefefc56bd0d0bb, 0x16fbfbede5aaaa4f);
AS_UL2(&sharedMemory[0x064]) = make_ulonglong2(0xd74d4d9ac5434386, 0x9485851155333366);
AS_UL2(&sharedMemory[0x068]) = make_ulonglong2(0x10f9f9e9cf45458a, 0x817f7ffe06020204);
AS_UL2(&sharedMemory[0x06C]) = make_ulonglong2(0x443c3c78f05050a0, 0xe3a8a84bba9f9f25);
AS_UL2(&sharedMemory[0x070]) = make_ulonglong2(0xfea3a35df35151a2, 0x8a8f8f05c0404080);
AS_UL2(&sharedMemory[0x074]) = make_ulonglong2(0xbc9d9d21ad92923f, 0x04f5f5f148383870);
AS_UL2(&sharedMemory[0x078]) = make_ulonglong2(0xc1b6b677dfbcbc63, 0x6321214275dadaaf);
AS_UL2(&sharedMemory[0x07C]) = make_ulonglong2(0x1affffe530101020, 0x6dd2d2bf0ef3f3fd);
break;
case 4:
AS_UL2(&sharedMemory[0x080]) = make_ulonglong2(0x140c0c184ccdcd81, 0x2fececc335131326);
AS_UL2(&sharedMemory[0x084]) = make_ulonglong2(0xa2979735e15f5fbe, 0x3917172ecc444488);
AS_UL2(&sharedMemory[0x088]) = make_ulonglong2(0xf2a7a75557c4c493, 0x473d3d7a827e7efc);
AS_UL2(&sharedMemory[0x08C]) = make_ulonglong2(0xe75d5dbaac6464c8, 0x957373e62b191932);
AS_UL2(&sharedMemory[0x090]) = make_ulonglong2(0x98818119a06060c0, 0x7fdcdca3d14f4f9e);
AS_UL2(&sharedMemory[0x094]) = make_ulonglong2(0x7e2a2a5466222244, 0x8388880bab90903b);
AS_UL2(&sharedMemory[0x098]) = make_ulonglong2(0x29eeeec7ca46468c, 0x3c141428d3b8b86b);
AS_UL2(&sharedMemory[0x09C]) = make_ulonglong2(0xe25e5ebc79dedea7, 0x76dbdbad1d0b0b16);
break;
case 5:
AS_UL2(&sharedMemory[0x0A0]) = make_ulonglong2(0x563232643be0e0db, 0x1e0a0a144e3a3a74);
AS_UL2(&sharedMemory[0x0A4]) = make_ulonglong2(0x0a06060cdb494992, 0xe45c5cb86c242448);
AS_UL2(&sharedMemory[0x0A8]) = make_ulonglong2(0x6ed3d3bd5dc2c29f, 0xa66262c4efacac43);
AS_UL2(&sharedMemory[0x0AC]) = make_ulonglong2(0xa4959531a8919139, 0x8b7979f237e4e4d3);
AS_UL2(&sharedMemory[0x0B0]) = make_ulonglong2(0x43c8c88b32e7e7d5, 0xb76d6dda5937376e);
AS_UL2(&sharedMemory[0x0B4]) = make_ulonglong2(0x64d5d5b18c8d8d01, 0xe0a9a949d24e4e9c);
AS_UL2(&sharedMemory[0x0B8]) = make_ulonglong2(0xfa5656acb46c6cd8, 0x25eaeacf07f4f4f3);
AS_UL2(&sharedMemory[0x0BC]) = make_ulonglong2(0x8e7a7af4af6565ca, 0x18080810e9aeae47);
break;
case 6:
AS_UL2(&sharedMemory[0x0C0]) = make_ulonglong2(0x887878f0d5baba6f, 0x722e2e5c6f25254a);
AS_UL2(&sharedMemory[0x0C4]) = make_ulonglong2(0xf1a6a657241c1c38, 0x51c6c697c7b4b473);
AS_UL2(&sharedMemory[0x0C8]) = make_ulonglong2(0x7cdddda123e8e8cb, 0x211f1f3e9c7474e8);
AS_UL2(&sharedMemory[0x0CC]) = make_ulonglong2(0xdcbdbd61dd4b4b96, 0x858a8a0f868b8b0d);
AS_UL2(&sharedMemory[0x0D0]) = make_ulonglong2(0x423e3e7c907070e0, 0xaa6666ccc4b5b571);
AS_UL2(&sharedMemory[0x0D4]) = make_ulonglong2(0x05030306d8484890, 0x120e0e1c01f6f6f7);
AS_UL2(&sharedMemory[0x0D8]) = make_ulonglong2(0x5f35356aa36161c2, 0xd0b9b969f95757ae);
AS_UL2(&sharedMemory[0x0DC]) = make_ulonglong2(0x58c1c19991868617, 0xb99e9e27271d1d3a);
break;
case 7:
AS_UL2(&sharedMemory[0x0E0]) = make_ulonglong2(0x13f8f8eb38e1e1d9, 0x33111122b398982b);
AS_UL2(&sharedMemory[0x0E4]) = make_ulonglong2(0x70d9d9a9bb6969d2, 0xa7949433898e8e07);
AS_UL2(&sharedMemory[0x0E8]) = make_ulonglong2(0x221e1e3cb69b9b2d, 0x20e9e9c992878715);
AS_UL2(&sharedMemory[0x0EC]) = make_ulonglong2(0xff5555aa49cece87, 0x7adfdfa578282850);
AS_UL2(&sharedMemory[0x0F0]) = make_ulonglong2(0xf8a1a1598f8c8c03, 0x170d0d1a80898909);
AS_UL2(&sharedMemory[0x0F4]) = make_ulonglong2(0x31e6e6d7dabfbf65, 0xb86868d0c6424284);
AS_UL2(&sharedMemory[0x0F8]) = make_ulonglong2(0xb0999929c3414182, 0x110f0f1e772d2d5a);
AS_UL2(&sharedMemory[0x0FC]) = make_ulonglong2(0xfc5454a8cbb0b07b, 0x3a16162cd6bbbb6d);
break;
}
// AES 1
switch (threadIdx.x) {
case 0:
AS_UL2(&sharedMemory[0x100]) = make_ulonglong2(0x7c7cf8846363c6a5, 0x7b7bf68d7777ee99);
AS_UL2(&sharedMemory[0x104]) = make_ulonglong2(0x6b6bd6bdf2f2ff0d, 0xc5c591546f6fdeb1);
AS_UL2(&sharedMemory[0x108]) = make_ulonglong2(0x0101020330306050, 0x2b2b567d6767cea9);
AS_UL2(&sharedMemory[0x10C]) = make_ulonglong2(0xd7d7b562fefee719, 0x7676ec9aabab4de6);
AS_UL2(&sharedMemory[0x110]) = make_ulonglong2(0x82821f9dcaca8f45, 0x7d7dfa87c9c98940);
AS_UL2(&sharedMemory[0x114]) = make_ulonglong2(0x5959b2ebfafaef15, 0xf0f0fb0b47478ec9);
AS_UL2(&sharedMemory[0x118]) = make_ulonglong2(0xd4d4b367adad41ec, 0xafaf45eaa2a25ffd);
AS_UL2(&sharedMemory[0x11C]) = make_ulonglong2(0xa4a453f79c9c23bf, 0xc0c09b5b7272e496);
break;
case 1:
AS_UL2(&sharedMemory[0x120]) = make_ulonglong2(0xfdfde11cb7b775c2, 0x26264c6a93933dae);
AS_UL2(&sharedMemory[0x124]) = make_ulonglong2(0x3f3f7e4136366c5a, 0xcccc834ff7f7f502);
AS_UL2(&sharedMemory[0x128]) = make_ulonglong2(0xa5a551f43434685c, 0xf1f1f908e5e5d134);
AS_UL2(&sharedMemory[0x12C]) = make_ulonglong2(0xd8d8ab737171e293, 0x15152a3f31316253);
AS_UL2(&sharedMemory[0x130]) = make_ulonglong2(0xc7c795520404080c, 0xc3c39d5e23234665);
AS_UL2(&sharedMemory[0x134]) = make_ulonglong2(0x969637a118183028, 0x9a9a2fb505050a0f);
AS_UL2(&sharedMemory[0x138]) = make_ulonglong2(0x1212243607070e09, 0xe2e2df3d80801b9b);
AS_UL2(&sharedMemory[0x13C]) = make_ulonglong2(0x27274e69ebebcd26, 0x7575ea9fb2b27fcd);
break;
case 2:
AS_UL2(&sharedMemory[0x140]) = make_ulonglong2(0x83831d9e0909121b, 0x1a1a342e2c2c5874);
AS_UL2(&sharedMemory[0x144]) = make_ulonglong2(0x6e6edcb21b1b362d, 0xa0a05bfb5a5ab4ee);
AS_UL2(&sharedMemory[0x148]) = make_ulonglong2(0x3b3b764d5252a4f6, 0xb3b37dced6d6b761);
AS_UL2(&sharedMemory[0x14C]) = make_ulonglong2(0xe3e3dd3e2929527b, 0x848413972f2f5e71);
AS_UL2(&sharedMemory[0x150]) = make_ulonglong2(0xd1d1b9685353a6f5, 0xededc12c00000000);
AS_UL2(&sharedMemory[0x154]) = make_ulonglong2(0xfcfce31f20204060, 0x5b5bb6edb1b179c8);
AS_UL2(&sharedMemory[0x158]) = make_ulonglong2(0xcbcb8d466a6ad4be, 0x3939724bbebe67d9);
AS_UL2(&sharedMemory[0x15C]) = make_ulonglong2(0x4c4c98d44a4a94de, 0xcfcf854a5858b0e8);
break;
case 3:
AS_UL2(&sharedMemory[0x160]) = make_ulonglong2(0xefefc52ad0d0bb6b, 0xfbfbed16aaaa4fe5);
AS_UL2(&sharedMemory[0x164]) = make_ulonglong2(0x4d4d9ad7434386c5, 0x8585119433336655);
AS_UL2(&sharedMemory[0x168]) = make_ulonglong2(0xf9f9e91045458acf, 0x7f7ffe8102020406);
AS_UL2(&sharedMemory[0x16C]) = make_ulonglong2(0x3c3c78445050a0f0, 0xa8a84be39f9f25ba);
AS_UL2(&sharedMemory[0x170]) = make_ulonglong2(0xa3a35dfe5151a2f3, 0x8f8f058a404080c0);
AS_UL2(&sharedMemory[0x174]) = make_ulonglong2(0x9d9d21bc92923fad, 0xf5f5f10438387048);
AS_UL2(&sharedMemory[0x178]) = make_ulonglong2(0xb6b677c1bcbc63df, 0x21214263dadaaf75);
AS_UL2(&sharedMemory[0x17C]) = make_ulonglong2(0xffffe51a10102030, 0xd2d2bf6df3f3fd0e);
break;
case 4:
AS_UL2(&sharedMemory[0x180]) = make_ulonglong2(0x0c0c1814cdcd814c, 0xececc32f13132635);
AS_UL2(&sharedMemory[0x184]) = make_ulonglong2(0x979735a25f5fbee1, 0x17172e39444488cc);
AS_UL2(&sharedMemory[0x188]) = make_ulonglong2(0xa7a755f2c4c49357, 0x3d3d7a477e7efc82);
AS_UL2(&sharedMemory[0x18C]) = make_ulonglong2(0x5d5dbae76464c8ac, 0x7373e6951919322b);
AS_UL2(&sharedMemory[0x190]) = make_ulonglong2(0x818119986060c0a0, 0xdcdca37f4f4f9ed1);
AS_UL2(&sharedMemory[0x194]) = make_ulonglong2(0x2a2a547e22224466, 0x88880b8390903bab);
AS_UL2(&sharedMemory[0x198]) = make_ulonglong2(0xeeeec72946468cca, 0x1414283cb8b86bd3);
AS_UL2(&sharedMemory[0x19C]) = make_ulonglong2(0x5e5ebce2dedea779, 0xdbdbad760b0b161d);
break;
case 5:
AS_UL2(&sharedMemory[0x1A0]) = make_ulonglong2(0x32326456e0e0db3b, 0x0a0a141e3a3a744e);
AS_UL2(&sharedMemory[0x1A4]) = make_ulonglong2(0x06060c0a494992db, 0x5c5cb8e42424486c);
AS_UL2(&sharedMemory[0x1A8]) = make_ulonglong2(0xd3d3bd6ec2c29f5d, 0x6262c4a6acac43ef);
AS_UL2(&sharedMemory[0x1AC]) = make_ulonglong2(0x959531a4919139a8, 0x7979f28be4e4d337);
AS_UL2(&sharedMemory[0x1B0]) = make_ulonglong2(0xc8c88b43e7e7d532, 0x6d6ddab737376e59);
AS_UL2(&sharedMemory[0x1B4]) = make_ulonglong2(0xd5d5b1648d8d018c, 0xa9a949e04e4e9cd2);
AS_UL2(&sharedMemory[0x1B8]) = make_ulonglong2(0x5656acfa6c6cd8b4, 0xeaeacf25f4f4f307);
AS_UL2(&sharedMemory[0x1BC]) = make_ulonglong2(0x7a7af48e6565caaf, 0x08081018aeae47e9);
break;
case 6:
AS_UL2(&sharedMemory[0x1C0]) = make_ulonglong2(0x7878f088baba6fd5, 0x2e2e5c7225254a6f);
AS_UL2(&sharedMemory[0x1C4]) = make_ulonglong2(0xa6a657f11c1c3824, 0xc6c69751b4b473c7);
AS_UL2(&sharedMemory[0x1C8]) = make_ulonglong2(0xdddda17ce8e8cb23, 0x1f1f3e217474e89c);
AS_UL2(&sharedMemory[0x1CC]) = make_ulonglong2(0xbdbd61dc4b4b96dd, 0x8a8a0f858b8b0d86);
AS_UL2(&sharedMemory[0x1D0]) = make_ulonglong2(0x3e3e7c427070e090, 0x6666ccaab5b571c4);
AS_UL2(&sharedMemory[0x1D4]) = make_ulonglong2(0x03030605484890d8, 0x0e0e1c12f6f6f701);
AS_UL2(&sharedMemory[0x1D8]) = make_ulonglong2(0x35356a5f6161c2a3, 0xb9b969d05757aef9);
AS_UL2(&sharedMemory[0x1DC]) = make_ulonglong2(0xc1c1995886861791, 0x9e9e27b91d1d3a27);
break;
case 7:
AS_UL2(&sharedMemory[0x1E0]) = make_ulonglong2(0xf8f8eb13e1e1d938, 0x1111223398982bb3);
AS_UL2(&sharedMemory[0x1E4]) = make_ulonglong2(0xd9d9a9706969d2bb, 0x949433a78e8e0789);
AS_UL2(&sharedMemory[0x1E8]) = make_ulonglong2(0x1e1e3c229b9b2db6, 0xe9e9c92087871592);
AS_UL2(&sharedMemory[0x1EC]) = make_ulonglong2(0x5555aaffcece8749, 0xdfdfa57a28285078);
AS_UL2(&sharedMemory[0x1F0]) = make_ulonglong2(0xa1a159f88c8c038f, 0x0d0d1a1789890980);
AS_UL2(&sharedMemory[0x1F4]) = make_ulonglong2(0xe6e6d731bfbf65da, 0x6868d0b8424284c6);
AS_UL2(&sharedMemory[0x1F8]) = make_ulonglong2(0x999929b0414182c3, 0x0f0f1e112d2d5a77);
AS_UL2(&sharedMemory[0x1FC]) = make_ulonglong2(0x5454a8fcb0b07bcb, 0x16162c3abbbb6dd6);
break;
}
// AES 2
switch (threadIdx.x) {
case 0:
AS_UL2(&sharedMemory[0x200]) = make_ulonglong2(0x7cf8847c63c6a563, 0x7bf68d7b77ee9977);
AS_UL2(&sharedMemory[0x204]) = make_ulonglong2(0x6bd6bd6bf2ff0df2, 0xc59154c56fdeb16f);
AS_UL2(&sharedMemory[0x208]) = make_ulonglong2(0x0102030130605030, 0x2b567d2b67cea967);
AS_UL2(&sharedMemory[0x20C]) = make_ulonglong2(0xd7b562d7fee719fe, 0x76ec9a76ab4de6ab);
AS_UL2(&sharedMemory[0x210]) = make_ulonglong2(0x821f9d82ca8f45ca, 0x7dfa877dc98940c9);
AS_UL2(&sharedMemory[0x214]) = make_ulonglong2(0x59b2eb59faef15fa, 0xf0fb0bf0478ec947);
AS_UL2(&sharedMemory[0x218]) = make_ulonglong2(0xd4b367d4ad41ecad, 0xaf45eaafa25ffda2);
AS_UL2(&sharedMemory[0x21C]) = make_ulonglong2(0xa453f7a49c23bf9c, 0xc09b5bc072e49672);
break;
case 1:
AS_UL2(&sharedMemory[0x220]) = make_ulonglong2(0xfde11cfdb775c2b7, 0x264c6a26933dae93);
AS_UL2(&sharedMemory[0x224]) = make_ulonglong2(0x3f7e413f366c5a36, 0xcc834fccf7f502f7);
AS_UL2(&sharedMemory[0x228]) = make_ulonglong2(0xa551f4a534685c34, 0xf1f908f1e5d134e5);
AS_UL2(&sharedMemory[0x22C]) = make_ulonglong2(0xd8ab73d871e29371, 0x152a3f1531625331);
AS_UL2(&sharedMemory[0x230]) = make_ulonglong2(0xc79552c704080c04, 0xc39d5ec323466523);
AS_UL2(&sharedMemory[0x234]) = make_ulonglong2(0x9637a19618302818, 0x9a2fb59a050a0f05);
AS_UL2(&sharedMemory[0x238]) = make_ulonglong2(0x12243612070e0907, 0xe2df3de2801b9b80);
AS_UL2(&sharedMemory[0x23C]) = make_ulonglong2(0x274e6927ebcd26eb, 0x75ea9f75b27fcdb2);
break;
case 2:
AS_UL2(&sharedMemory[0x240]) = make_ulonglong2(0x831d9e8309121b09, 0x1a342e1a2c58742c);
AS_UL2(&sharedMemory[0x244]) = make_ulonglong2(0x6edcb26e1b362d1b, 0xa05bfba05ab4ee5a);
AS_UL2(&sharedMemory[0x248]) = make_ulonglong2(0x3b764d3b52a4f652, 0xb37dceb3d6b761d6);
AS_UL2(&sharedMemory[0x24C]) = make_ulonglong2(0xe3dd3ee329527b29, 0x841397842f5e712f);
AS_UL2(&sharedMemory[0x250]) = make_ulonglong2(0xd1b968d153a6f553, 0xedc12ced00000000);
AS_UL2(&sharedMemory[0x254]) = make_ulonglong2(0xfce31ffc20406020, 0x5bb6ed5bb179c8b1);
AS_UL2(&sharedMemory[0x258]) = make_ulonglong2(0xcb8d46cb6ad4be6a, 0x39724b39be67d9be);
AS_UL2(&sharedMemory[0x25C]) = make_ulonglong2(0x4c98d44c4a94de4a, 0xcf854acf58b0e858);
break;
case 3:
AS_UL2(&sharedMemory[0x260]) = make_ulonglong2(0xefc52aefd0bb6bd0, 0xfbed16fbaa4fe5aa);
AS_UL2(&sharedMemory[0x264]) = make_ulonglong2(0x4d9ad74d4386c543, 0x8511948533665533);
AS_UL2(&sharedMemory[0x268]) = make_ulonglong2(0xf9e910f9458acf45, 0x7ffe817f02040602);
AS_UL2(&sharedMemory[0x26C]) = make_ulonglong2(0x3c78443c50a0f050, 0xa84be3a89f25ba9f);
AS_UL2(&sharedMemory[0x270]) = make_ulonglong2(0xa35dfea351a2f351, 0x8f058a8f4080c040);
AS_UL2(&sharedMemory[0x274]) = make_ulonglong2(0x9d21bc9d923fad92, 0xf5f104f538704838);
AS_UL2(&sharedMemory[0x278]) = make_ulonglong2(0xb677c1b6bc63dfbc, 0x21426321daaf75da);
AS_UL2(&sharedMemory[0x27C]) = make_ulonglong2(0xffe51aff10203010, 0xd2bf6dd2f3fd0ef3);
break;
case 4:
AS_UL2(&sharedMemory[0x280]) = make_ulonglong2(0x0c18140ccd814ccd, 0xecc32fec13263513);
AS_UL2(&sharedMemory[0x284]) = make_ulonglong2(0x9735a2975fbee15f, 0x172e39174488cc44);
AS_UL2(&sharedMemory[0x288]) = make_ulonglong2(0xa755f2a7c49357c4, 0x3d7a473d7efc827e);
AS_UL2(&sharedMemory[0x28C]) = make_ulonglong2(0x5dbae75d64c8ac64, 0x73e6957319322b19);
AS_UL2(&sharedMemory[0x290]) = make_ulonglong2(0x8119988160c0a060, 0xdca37fdc4f9ed14f);
AS_UL2(&sharedMemory[0x294]) = make_ulonglong2(0x2a547e2a22446622, 0x880b8388903bab90);
AS_UL2(&sharedMemory[0x298]) = make_ulonglong2(0xeec729ee468cca46, 0x14283c14b86bd3b8);
AS_UL2(&sharedMemory[0x29C]) = make_ulonglong2(0x5ebce25edea779de, 0xdbad76db0b161d0b);
break;
case 5:
AS_UL2(&sharedMemory[0x2A0]) = make_ulonglong2(0x32645632e0db3be0, 0x0a141e0a3a744e3a);
AS_UL2(&sharedMemory[0x2A4]) = make_ulonglong2(0x060c0a064992db49, 0x5cb8e45c24486c24);
AS_UL2(&sharedMemory[0x2A8]) = make_ulonglong2(0xd3bd6ed3c29f5dc2, 0x62c4a662ac43efac);
AS_UL2(&sharedMemory[0x2AC]) = make_ulonglong2(0x9531a4959139a891, 0x79f28b79e4d337e4);
AS_UL2(&sharedMemory[0x2B0]) = make_ulonglong2(0xc88b43c8e7d532e7, 0x6ddab76d376e5937);
AS_UL2(&sharedMemory[0x2B4]) = make_ulonglong2(0xd5b164d58d018c8d, 0xa949e0a94e9cd24e);
AS_UL2(&sharedMemory[0x2B8]) = make_ulonglong2(0x56acfa566cd8b46c, 0xeacf25eaf4f307f4);
AS_UL2(&sharedMemory[0x2BC]) = make_ulonglong2(0x7af48e7a65caaf65, 0x08101808ae47e9ae);
break;
case 6:
AS_UL2(&sharedMemory[0x2C0]) = make_ulonglong2(0x78f08878ba6fd5ba, 0x2e5c722e254a6f25);
AS_UL2(&sharedMemory[0x2C4]) = make_ulonglong2(0xa657f1a61c38241c, 0xc69751c6b473c7b4);
AS_UL2(&sharedMemory[0x2C8]) = make_ulonglong2(0xdda17cdde8cb23e8, 0x1f3e211f74e89c74);
AS_UL2(&sharedMemory[0x2CC]) = make_ulonglong2(0xbd61dcbd4b96dd4b, 0x8a0f858a8b0d868b);
AS_UL2(&sharedMemory[0x2D0]) = make_ulonglong2(0x3e7c423e70e09070, 0x66ccaa66b571c4b5);
AS_UL2(&sharedMemory[0x2D4]) = make_ulonglong2(0x030605034890d848, 0x0e1c120ef6f701f6);
AS_UL2(&sharedMemory[0x2D8]) = make_ulonglong2(0x356a5f3561c2a361, 0xb969d0b957aef957);
AS_UL2(&sharedMemory[0x2DC]) = make_ulonglong2(0xc19958c186179186, 0x9e27b99e1d3a271d);
break;
case 7:
AS_UL2(&sharedMemory[0x2E0]) = make_ulonglong2(0xf8eb13f8e1d938e1, 0x11223311982bb398);
AS_UL2(&sharedMemory[0x2E4]) = make_ulonglong2(0xd9a970d969d2bb69, 0x9433a7948e07898e);
AS_UL2(&sharedMemory[0x2E8]) = make_ulonglong2(0x1e3c221e9b2db69b, 0xe9c920e987159287);
AS_UL2(&sharedMemory[0x2EC]) = make_ulonglong2(0x55aaff55ce8749ce, 0xdfa57adf28507828);
AS_UL2(&sharedMemory[0x2F0]) = make_ulonglong2(0xa159f8a18c038f8c, 0x0d1a170d89098089);
AS_UL2(&sharedMemory[0x2F4]) = make_ulonglong2(0xe6d731e6bf65dabf, 0x68d0b8684284c642);
AS_UL2(&sharedMemory[0x2F8]) = make_ulonglong2(0x9929b0994182c341, 0x0f1e110f2d5a772d);
AS_UL2(&sharedMemory[0x2FC]) = make_ulonglong2(0x54a8fc54b07bcbb0, 0x162c3a16bb6dd6bb);
break;
}
// AES 3
switch (threadIdx.x) {
case 0:
AS_UL2(&sharedMemory[0x300]) = make_ulonglong2(0xf8847c7cc6a56363, 0xf68d7b7bee997777);
AS_UL2(&sharedMemory[0x304]) = make_ulonglong2(0xd6bd6b6bff0df2f2, 0x9154c5c5deb16f6f);
AS_UL2(&sharedMemory[0x308]) = make_ulonglong2(0x0203010160503030, 0x567d2b2bcea96767);
AS_UL2(&sharedMemory[0x30C]) = make_ulonglong2(0xb562d7d7e719fefe, 0xec9a76764de6abab);
AS_UL2(&sharedMemory[0x310]) = make_ulonglong2(0x1f9d82828f45caca, 0xfa877d7d8940c9c9);
AS_UL2(&sharedMemory[0x314]) = make_ulonglong2(0xb2eb5959ef15fafa, 0xfb0bf0f08ec94747);
AS_UL2(&sharedMemory[0x318]) = make_ulonglong2(0xb367d4d441ecadad, 0x45eaafaf5ffda2a2);
AS_UL2(&sharedMemory[0x31C]) = make_ulonglong2(0x53f7a4a423bf9c9c, 0x9b5bc0c0e4967272);
break;
case 1:
AS_UL2(&sharedMemory[0x320]) = make_ulonglong2(0xe11cfdfd75c2b7b7, 0x4c6a26263dae9393);
AS_UL2(&sharedMemory[0x324]) = make_ulonglong2(0x7e413f3f6c5a3636, 0x834fccccf502f7f7);
AS_UL2(&sharedMemory[0x328]) = make_ulonglong2(0x51f4a5a5685c3434, 0xf908f1f1d134e5e5);
AS_UL2(&sharedMemory[0x32C]) = make_ulonglong2(0xab73d8d8e2937171, 0x2a3f151562533131);
AS_UL2(&sharedMemory[0x330]) = make_ulonglong2(0x9552c7c7080c0404, 0x9d5ec3c346652323);
AS_UL2(&sharedMemory[0x334]) = make_ulonglong2(0x37a1969630281818, 0x2fb59a9a0a0f0505);
AS_UL2(&sharedMemory[0x338]) = make_ulonglong2(0x243612120e090707, 0xdf3de2e21b9b8080);
AS_UL2(&sharedMemory[0x33C]) = make_ulonglong2(0x4e692727cd26ebeb, 0xea9f75757fcdb2b2);
break;
case 2:
AS_UL2(&sharedMemory[0x340]) = make_ulonglong2(0x1d9e8383121b0909, 0x342e1a1a58742c2c);
AS_UL2(&sharedMemory[0x344]) = make_ulonglong2(0xdcb26e6e362d1b1b, 0x5bfba0a0b4ee5a5a);
AS_UL2(&sharedMemory[0x348]) = make_ulonglong2(0x764d3b3ba4f65252, 0x7dceb3b3b761d6d6);
AS_UL2(&sharedMemory[0x34C]) = make_ulonglong2(0xdd3ee3e3527b2929, 0x139784845e712f2f);
AS_UL2(&sharedMemory[0x350]) = make_ulonglong2(0xb968d1d1a6f55353, 0xc12ceded00000000);
AS_UL2(&sharedMemory[0x354]) = make_ulonglong2(0xe31ffcfc40602020, 0xb6ed5b5b79c8b1b1);
AS_UL2(&sharedMemory[0x358]) = make_ulonglong2(0x8d46cbcbd4be6a6a, 0x724b393967d9bebe);
AS_UL2(&sharedMemory[0x35C]) = make_ulonglong2(0x98d44c4c94de4a4a, 0x854acfcfb0e85858);
break;
case 3:
AS_UL2(&sharedMemory[0x360]) = make_ulonglong2(0xc52aefefbb6bd0d0, 0xed16fbfb4fe5aaaa);
AS_UL2(&sharedMemory[0x364]) = make_ulonglong2(0x9ad74d4d86c54343, 0x1194858566553333);
AS_UL2(&sharedMemory[0x368]) = make_ulonglong2(0xe910f9f98acf4545, 0xfe817f7f04060202);
AS_UL2(&sharedMemory[0x36C]) = make_ulonglong2(0x78443c3ca0f05050, 0x4be3a8a825ba9f9f);
AS_UL2(&sharedMemory[0x370]) = make_ulonglong2(0x5dfea3a3a2f35151, 0x058a8f8f80c04040);
AS_UL2(&sharedMemory[0x374]) = make_ulonglong2(0x21bc9d9d3fad9292, 0xf104f5f570483838);
AS_UL2(&sharedMemory[0x378]) = make_ulonglong2(0x77c1b6b663dfbcbc, 0x42632121af75dada);
AS_UL2(&sharedMemory[0x37C]) = make_ulonglong2(0xe51affff20301010, 0xbf6dd2d2fd0ef3f3);
break;
case 4:
AS_UL2(&sharedMemory[0x380]) = make_ulonglong2(0x18140c0c814ccdcd, 0xc32fecec26351313);
AS_UL2(&sharedMemory[0x384]) = make_ulonglong2(0x35a29797bee15f5f, 0x2e39171788cc4444);
AS_UL2(&sharedMemory[0x388]) = make_ulonglong2(0x55f2a7a79357c4c4, 0x7a473d3dfc827e7e);
AS_UL2(&sharedMemory[0x38C]) = make_ulonglong2(0xbae75d5dc8ac6464, 0xe6957373322b1919);
AS_UL2(&sharedMemory[0x390]) = make_ulonglong2(0x19988181c0a06060, 0xa37fdcdc9ed14f4f);
AS_UL2(&sharedMemory[0x394]) = make_ulonglong2(0x547e2a2a44662222, 0x0b8388883bab9090);
AS_UL2(&sharedMemory[0x398]) = make_ulonglong2(0xc729eeee8cca4646, 0x283c14146bd3b8b8);
AS_UL2(&sharedMemory[0x39C]) = make_ulonglong2(0xbce25e5ea779dede, 0xad76dbdb161d0b0b);
break;
case 5:
AS_UL2(&sharedMemory[0x3A0]) = make_ulonglong2(0x64563232db3be0e0, 0x141e0a0a744e3a3a);
AS_UL2(&sharedMemory[0x3A4]) = make_ulonglong2(0x0c0a060692db4949, 0xb8e45c5c486c2424);
AS_UL2(&sharedMemory[0x3A8]) = make_ulonglong2(0xbd6ed3d39f5dc2c2, 0xc4a6626243efacac);
AS_UL2(&sharedMemory[0x3AC]) = make_ulonglong2(0x31a4959539a89191, 0xf28b7979d337e4e4);
AS_UL2(&sharedMemory[0x3B0]) = make_ulonglong2(0x8b43c8c8d532e7e7, 0xdab76d6d6e593737);
AS_UL2(&sharedMemory[0x3B4]) = make_ulonglong2(0xb164d5d5018c8d8d, 0x49e0a9a99cd24e4e);
AS_UL2(&sharedMemory[0x3B8]) = make_ulonglong2(0xacfa5656d8b46c6c, 0xcf25eaeaf307f4f4);
AS_UL2(&sharedMemory[0x3BC]) = make_ulonglong2(0xf48e7a7acaaf6565, 0x1018080847e9aeae);
break;
case 6:
AS_UL2(&sharedMemory[0x3C0]) = make_ulonglong2(0xf08878786fd5baba, 0x5c722e2e4a6f2525);
AS_UL2(&sharedMemory[0x3C4]) = make_ulonglong2(0x57f1a6a638241c1c, 0x9751c6c673c7b4b4);
AS_UL2(&sharedMemory[0x3C8]) = make_ulonglong2(0xa17cddddcb23e8e8, 0x3e211f1fe89c7474);
AS_UL2(&sharedMemory[0x3CC]) = make_ulonglong2(0x61dcbdbd96dd4b4b, 0x0f858a8a0d868b8b);
AS_UL2(&sharedMemory[0x3D0]) = make_ulonglong2(0x7c423e3ee0907070, 0xccaa666671c4b5b5);
AS_UL2(&sharedMemory[0x3D4]) = make_ulonglong2(0x0605030390d84848, 0x1c120e0ef701f6f6);
AS_UL2(&sharedMemory[0x3D8]) = make_ulonglong2(0x6a5f3535c2a36161, 0x69d0b9b9aef95757);
AS_UL2(&sharedMemory[0x3DC]) = make_ulonglong2(0x9958c1c117918686, 0x27b99e9e3a271d1d);
break;
case 7:
AS_UL2(&sharedMemory[0x3E0]) = make_ulonglong2(0xeb13f8f8d938e1e1, 0x223311112bb39898);
AS_UL2(&sharedMemory[0x3E4]) = make_ulonglong2(0xa970d9d9d2bb6969, 0x33a7949407898e8e);
AS_UL2(&sharedMemory[0x3E8]) = make_ulonglong2(0x3c221e1e2db69b9b, 0xc920e9e915928787);
AS_UL2(&sharedMemory[0x3EC]) = make_ulonglong2(0xaaff55558749cece, 0xa57adfdf50782828);
AS_UL2(&sharedMemory[0x3F0]) = make_ulonglong2(0x59f8a1a1038f8c8c, 0x1a170d0d09808989);
AS_UL2(&sharedMemory[0x3F4]) = make_ulonglong2(0xd731e6e665dabfbf, 0xd0b8686884c64242);
AS_UL2(&sharedMemory[0x3F8]) = make_ulonglong2(0x29b0999982c34141, 0x1e110f0f5a772d2d);
AS_UL2(&sharedMemory[0x3FC]) = make_ulonglong2(0xa8fc54547bcbb0b0, 0x2c3a16166dd6bbbb);
break;
}
}
__device__ __forceinline__
void cn_aes_gpu_init_u4(uint32_t* sharedMemory)
{
// AES 0
switch (threadIdx.x) {
case 0:
AS_UINT4(&sharedMemory[0x000]) = make_uint4(0xa56363c6, 0x847c7cf8, 0x997777ee, 0x8d7b7bf6);
AS_UINT4(&sharedMemory[0x004]) = make_uint4(0x0df2f2ff, 0xbd6b6bd6, 0xb16f6fde, 0x54c5c591);
AS_UINT4(&sharedMemory[0x008]) = make_uint4(0x50303060, 0x03010102, 0xa96767ce, 0x7d2b2b56);
AS_UINT4(&sharedMemory[0x00C]) = make_uint4(0x19fefee7, 0x62d7d7b5, 0xe6abab4d, 0x9a7676ec);
AS_UINT4(&sharedMemory[0x010]) = make_uint4(0x45caca8f, 0x9d82821f, 0x40c9c989, 0x877d7dfa);
AS_UINT4(&sharedMemory[0x014]) = make_uint4(0x15fafaef, 0xeb5959b2, 0xc947478e, 0x0bf0f0fb);
AS_UINT4(&sharedMemory[0x018]) = make_uint4(0xecadad41, 0x67d4d4b3, 0xfda2a25f, 0xeaafaf45);
AS_UINT4(&sharedMemory[0x01C]) = make_uint4(0xbf9c9c23, 0xf7a4a453, 0x967272e4, 0x5bc0c09b);
break;
case 1:
AS_UINT4(&sharedMemory[0x020]) = make_uint4(0xc2b7b775, 0x1cfdfde1, 0xae93933d, 0x6a26264c);
AS_UINT4(&sharedMemory[0x024]) = make_uint4(0x5a36366c, 0x413f3f7e, 0x02f7f7f5, 0x4fcccc83);
AS_UINT4(&sharedMemory[0x028]) = make_uint4(0x5c343468, 0xf4a5a551, 0x34e5e5d1, 0x08f1f1f9);
AS_UINT4(&sharedMemory[0x02C]) = make_uint4(0x937171e2, 0x73d8d8ab, 0x53313162, 0x3f15152a);
AS_UINT4(&sharedMemory[0x030]) = make_uint4(0x0c040408, 0x52c7c795, 0x65232346, 0x5ec3c39d);
AS_UINT4(&sharedMemory[0x034]) = make_uint4(0x28181830, 0xa1969637, 0x0f05050a, 0xb59a9a2f);
AS_UINT4(&sharedMemory[0x038]) = make_uint4(0x0907070e, 0x36121224, 0x9b80801b, 0x3de2e2df);
AS_UINT4(&sharedMemory[0x03C]) = make_uint4(0x26ebebcd, 0x6927274e, 0xcdb2b27f, 0x9f7575ea);
break;
case 2:
AS_UINT4(&sharedMemory[0x040]) = make_uint4(0x1b090912, 0x9e83831d, 0x742c2c58, 0x2e1a1a34);
AS_UINT4(&sharedMemory[0x044]) = make_uint4(0x2d1b1b36, 0xb26e6edc, 0xee5a5ab4, 0xfba0a05b);
AS_UINT4(&sharedMemory[0x048]) = make_uint4(0xf65252a4, 0x4d3b3b76, 0x61d6d6b7, 0xceb3b37d);
AS_UINT4(&sharedMemory[0x04C]) = make_uint4(0x7b292952, 0x3ee3e3dd, 0x712f2f5e, 0x97848413);
AS_UINT4(&sharedMemory[0x050]) = make_uint4(0xf55353a6, 0x68d1d1b9, 0x00000000, 0x2cededc1);
AS_UINT4(&sharedMemory[0x054]) = make_uint4(0x60202040, 0x1ffcfce3, 0xc8b1b179, 0xed5b5bb6);
AS_UINT4(&sharedMemory[0x058]) = make_uint4(0xbe6a6ad4, 0x46cbcb8d, 0xd9bebe67, 0x4b393972);
AS_UINT4(&sharedMemory[0x05C]) = make_uint4(0xde4a4a94, 0xd44c4c98, 0xe85858b0, 0x4acfcf85);
break;
case 3:
AS_UINT4(&sharedMemory[0x060]) = make_uint4(0x6bd0d0bb, 0x2aefefc5, 0xe5aaaa4f, 0x16fbfbed);
AS_UINT4(&sharedMemory[0x064]) = make_uint4(0xc5434386, 0xd74d4d9a, 0x55333366, 0x94858511);
AS_UINT4(&sharedMemory[0x068]) = make_uint4(0xcf45458a, 0x10f9f9e9, 0x06020204, 0x817f7ffe);
AS_UINT4(&sharedMemory[0x06C]) = make_uint4(0xf05050a0, 0x443c3c78, 0xba9f9f25, 0xe3a8a84b);
AS_UINT4(&sharedMemory[0x070]) = make_uint4(0xf35151a2, 0xfea3a35d, 0xc0404080, 0x8a8f8f05);
AS_UINT4(&sharedMemory[0x074]) = make_uint4(0xad92923f, 0xbc9d9d21, 0x48383870, 0x04f5f5f1);
AS_UINT4(&sharedMemory[0x078]) = make_uint4(0xdfbcbc63, 0xc1b6b677, 0x75dadaaf, 0x63212142);
AS_UINT4(&sharedMemory[0x07C]) = make_uint4(0x30101020, 0x1affffe5, 0x0ef3f3fd, 0x6dd2d2bf);
break;
case 4:
AS_UINT4(&sharedMemory[0x080]) = make_uint4(0x4ccdcd81, 0x140c0c18, 0x35131326, 0x2fececc3);
AS_UINT4(&sharedMemory[0x084]) = make_uint4(0xe15f5fbe, 0xa2979735, 0xcc444488, 0x3917172e);
AS_UINT4(&sharedMemory[0x088]) = make_uint4(0x57c4c493, 0xf2a7a755, 0x827e7efc, 0x473d3d7a);
AS_UINT4(&sharedMemory[0x08C]) = make_uint4(0xac6464c8, 0xe75d5dba, 0x2b191932, 0x957373e6);
AS_UINT4(&sharedMemory[0x090]) = make_uint4(0xa06060c0, 0x98818119, 0xd14f4f9e, 0x7fdcdca3);
AS_UINT4(&sharedMemory[0x094]) = make_uint4(0x66222244, 0x7e2a2a54, 0xab90903b, 0x8388880b);
AS_UINT4(&sharedMemory[0x098]) = make_uint4(0xca46468c, 0x29eeeec7, 0xd3b8b86b, 0x3c141428);
AS_UINT4(&sharedMemory[0x09C]) = make_uint4(0x79dedea7, 0xe25e5ebc, 0x1d0b0b16, 0x76dbdbad);
break;
case 5:
AS_UINT4(&sharedMemory[0x0A0]) = make_uint4(0x3be0e0db, 0x56323264, 0x4e3a3a74, 0x1e0a0a14);
AS_UINT4(&sharedMemory[0x0A4]) = make_uint4(0xdb494992, 0x0a06060c, 0x6c242448, 0xe45c5cb8);
AS_UINT4(&sharedMemory[0x0A8]) = make_uint4(0x5dc2c29f, 0x6ed3d3bd, 0xefacac43, 0xa66262c4);
AS_UINT4(&sharedMemory[0x0AC]) = make_uint4(0xa8919139, 0xa4959531, 0x37e4e4d3, 0x8b7979f2);
AS_UINT4(&sharedMemory[0x0B0]) = make_uint4(0x32e7e7d5, 0x43c8c88b, 0x5937376e, 0xb76d6dda);
AS_UINT4(&sharedMemory[0x0B4]) = make_uint4(0x8c8d8d01, 0x64d5d5b1, 0xd24e4e9c, 0xe0a9a949);
AS_UINT4(&sharedMemory[0x0B8]) = make_uint4(0xb46c6cd8, 0xfa5656ac, 0x07f4f4f3, 0x25eaeacf);
AS_UINT4(&sharedMemory[0x0BC]) = make_uint4(0xaf6565ca, 0x8e7a7af4, 0xe9aeae47, 0x18080810);
break;
case 6:
AS_UINT4(&sharedMemory[0x0C0]) = make_uint4(0xd5baba6f, 0x887878f0, 0x6f25254a, 0x722e2e5c);
AS_UINT4(&sharedMemory[0x0C4]) = make_uint4(0x241c1c38, 0xf1a6a657, 0xc7b4b473, 0x51c6c697);
AS_UINT4(&sharedMemory[0x0C8]) = make_uint4(0x23e8e8cb, 0x7cdddda1, 0x9c7474e8, 0x211f1f3e);
AS_UINT4(&sharedMemory[0x0CC]) = make_uint4(0xdd4b4b96, 0xdcbdbd61, 0x868b8b0d, 0x858a8a0f);
AS_UINT4(&sharedMemory[0x0D0]) = make_uint4(0x907070e0, 0x423e3e7c, 0xc4b5b571, 0xaa6666cc);
AS_UINT4(&sharedMemory[0x0D4]) = make_uint4(0xd8484890, 0x05030306, 0x01f6f6f7, 0x120e0e1c);
AS_UINT4(&sharedMemory[0x0D8]) = make_uint4(0xa36161c2, 0x5f35356a, 0xf95757ae, 0xd0b9b969);
AS_UINT4(&sharedMemory[0x0DC]) = make_uint4(0x91868617, 0x58c1c199, 0x271d1d3a, 0xb99e9e27);
break;
case 7:
AS_UINT4(&sharedMemory[0x0E0]) = make_uint4(0x38e1e1d9, 0x13f8f8eb, 0xb398982b, 0x33111122);
AS_UINT4(&sharedMemory[0x0E4]) = make_uint4(0xbb6969d2, 0x70d9d9a9, 0x898e8e07, 0xa7949433);
AS_UINT4(&sharedMemory[0x0E8]) = make_uint4(0xb69b9b2d, 0x221e1e3c, 0x92878715, 0x20e9e9c9);
AS_UINT4(&sharedMemory[0x0EC]) = make_uint4(0x49cece87, 0xff5555aa, 0x78282850, 0x7adfdfa5);
AS_UINT4(&sharedMemory[0x0F0]) = make_uint4(0x8f8c8c03, 0xf8a1a159, 0x80898909, 0x170d0d1a);
AS_UINT4(&sharedMemory[0x0F4]) = make_uint4(0xdabfbf65, 0x31e6e6d7, 0xc6424284, 0xb86868d0);
AS_UINT4(&sharedMemory[0x0F8]) = make_uint4(0xc3414182, 0xb0999929, 0x772d2d5a, 0x110f0f1e);
AS_UINT4(&sharedMemory[0x0FC]) = make_uint4(0xcbb0b07b, 0xfc5454a8, 0xd6bbbb6d, 0x3a16162c);
break;
}
// AES 1
switch (threadIdx.x) {
case 0:
AS_UINT4(&sharedMemory[0x100]) = make_uint4(0x6363c6a5, 0x7c7cf884, 0x7777ee99, 0x7b7bf68d);
AS_UINT4(&sharedMemory[0x104]) = make_uint4(0xf2f2ff0d, 0x6b6bd6bd, 0x6f6fdeb1, 0xc5c59154);
AS_UINT4(&sharedMemory[0x108]) = make_uint4(0x30306050, 0x01010203, 0x6767cea9, 0x2b2b567d);
AS_UINT4(&sharedMemory[0x10C]) = make_uint4(0xfefee719, 0xd7d7b562, 0xabab4de6, 0x7676ec9a);
AS_UINT4(&sharedMemory[0x110]) = make_uint4(0xcaca8f45, 0x82821f9d, 0xc9c98940, 0x7d7dfa87);
AS_UINT4(&sharedMemory[0x114]) = make_uint4(0xfafaef15, 0x5959b2eb, 0x47478ec9, 0xf0f0fb0b);
AS_UINT4(&sharedMemory[0x118]) = make_uint4(0xadad41ec, 0xd4d4b367, 0xa2a25ffd, 0xafaf45ea);
AS_UINT4(&sharedMemory[0x11C]) = make_uint4(0x9c9c23bf, 0xa4a453f7, 0x7272e496, 0xc0c09b5b);
break;
case 1:
AS_UINT4(&sharedMemory[0x120]) = make_uint4(0xb7b775c2, 0xfdfde11c, 0x93933dae, 0x26264c6a);
AS_UINT4(&sharedMemory[0x124]) = make_uint4(0x36366c5a, 0x3f3f7e41, 0xf7f7f502, 0xcccc834f);
AS_UINT4(&sharedMemory[0x128]) = make_uint4(0x3434685c, 0xa5a551f4, 0xe5e5d134, 0xf1f1f908);
AS_UINT4(&sharedMemory[0x12C]) = make_uint4(0x7171e293, 0xd8d8ab73, 0x31316253, 0x15152a3f);
AS_UINT4(&sharedMemory[0x130]) = make_uint4(0x0404080c, 0xc7c79552, 0x23234665, 0xc3c39d5e);
AS_UINT4(&sharedMemory[0x134]) = make_uint4(0x18183028, 0x969637a1, 0x05050a0f, 0x9a9a2fb5);
AS_UINT4(&sharedMemory[0x138]) = make_uint4(0x07070e09, 0x12122436, 0x80801b9b, 0xe2e2df3d);
AS_UINT4(&sharedMemory[0x13C]) = make_uint4(0xebebcd26, 0x27274e69, 0xb2b27fcd, 0x7575ea9f);
break;
case 2:
AS_UINT4(&sharedMemory[0x140]) = make_uint4(0x0909121b, 0x83831d9e, 0x2c2c5874, 0x1a1a342e);
AS_UINT4(&sharedMemory[0x144]) = make_uint4(0x1b1b362d, 0x6e6edcb2, 0x5a5ab4ee, 0xa0a05bfb);
AS_UINT4(&sharedMemory[0x148]) = make_uint4(0x5252a4f6, 0x3b3b764d, 0xd6d6b761, 0xb3b37dce);
AS_UINT4(&sharedMemory[0x14C]) = make_uint4(0x2929527b, 0xe3e3dd3e, 0x2f2f5e71, 0x84841397);
AS_UINT4(&sharedMemory[0x150]) = make_uint4(0x5353a6f5, 0xd1d1b968, 0x00000000, 0xededc12c);
AS_UINT4(&sharedMemory[0x154]) = make_uint4(0x20204060, 0xfcfce31f, 0xb1b179c8, 0x5b5bb6ed);
AS_UINT4(&sharedMemory[0x158]) = make_uint4(0x6a6ad4be, 0xcbcb8d46, 0xbebe67d9, 0x3939724b);
AS_UINT4(&sharedMemory[0x15C]) = make_uint4(0x4a4a94de, 0x4c4c98d4, 0x5858b0e8, 0xcfcf854a);
break;
case 3:
AS_UINT4(&sharedMemory[0x160]) = make_uint4(0xd0d0bb6b, 0xefefc52a, 0xaaaa4fe5, 0xfbfbed16);
AS_UINT4(&sharedMemory[0x164]) = make_uint4(0x434386c5, 0x4d4d9ad7, 0x33336655, 0x85851194);
AS_UINT4(&sharedMemory[0x168]) = make_uint4(0x45458acf, 0xf9f9e910, 0x02020406, 0x7f7ffe81);
AS_UINT4(&sharedMemory[0x16C]) = make_uint4(0x5050a0f0, 0x3c3c7844, 0x9f9f25ba, 0xa8a84be3);
AS_UINT4(&sharedMemory[0x170]) = make_uint4(0x5151a2f3, 0xa3a35dfe, 0x404080c0, 0x8f8f058a);
AS_UINT4(&sharedMemory[0x174]) = make_uint4(0x92923fad, 0x9d9d21bc, 0x38387048, 0xf5f5f104);
AS_UINT4(&sharedMemory[0x178]) = make_uint4(0xbcbc63df, 0xb6b677c1, 0xdadaaf75, 0x21214263);
AS_UINT4(&sharedMemory[0x17C]) = make_uint4(0x10102030, 0xffffe51a, 0xf3f3fd0e, 0xd2d2bf6d);
break;
case 4:
AS_UINT4(&sharedMemory[0x180]) = make_uint4(0xcdcd814c, 0x0c0c1814, 0x13132635, 0xececc32f);
AS_UINT4(&sharedMemory[0x184]) = make_uint4(0x5f5fbee1, 0x979735a2, 0x444488cc, 0x17172e39);
AS_UINT4(&sharedMemory[0x188]) = make_uint4(0xc4c49357, 0xa7a755f2, 0x7e7efc82, 0x3d3d7a47);
AS_UINT4(&sharedMemory[0x18C]) = make_uint4(0x6464c8ac, 0x5d5dbae7, 0x1919322b, 0x7373e695);
AS_UINT4(&sharedMemory[0x190]) = make_uint4(0x6060c0a0, 0x81811998, 0x4f4f9ed1, 0xdcdca37f);
AS_UINT4(&sharedMemory[0x194]) = make_uint4(0x22224466, 0x2a2a547e, 0x90903bab, 0x88880b83);
AS_UINT4(&sharedMemory[0x198]) = make_uint4(0x46468cca, 0xeeeec729, 0xb8b86bd3, 0x1414283c);
AS_UINT4(&sharedMemory[0x19C]) = make_uint4(0xdedea779, 0x5e5ebce2, 0x0b0b161d, 0xdbdbad76);
break;
case 5:
AS_UINT4(&sharedMemory[0x1A0]) = make_uint4(0xe0e0db3b, 0x32326456, 0x3a3a744e, 0x0a0a141e);
AS_UINT4(&sharedMemory[0x1A4]) = make_uint4(0x494992db, 0x06060c0a, 0x2424486c, 0x5c5cb8e4);
AS_UINT4(&sharedMemory[0x1A8]) = make_uint4(0xc2c29f5d, 0xd3d3bd6e, 0xacac43ef, 0x6262c4a6);
AS_UINT4(&sharedMemory[0x1AC]) = make_uint4(0x919139a8, 0x959531a4, 0xe4e4d337, 0x7979f28b);
AS_UINT4(&sharedMemory[0x1B0]) = make_uint4(0xe7e7d532, 0xc8c88b43, 0x37376e59, 0x6d6ddab7);
AS_UINT4(&sharedMemory[0x1B4]) = make_uint4(0x8d8d018c, 0xd5d5b164, 0x4e4e9cd2, 0xa9a949e0);
AS_UINT4(&sharedMemory[0x1B8]) = make_uint4(0x6c6cd8b4, 0x5656acfa, 0xf4f4f307, 0xeaeacf25);
AS_UINT4(&sharedMemory[0x1BC]) = make_uint4(0x6565caaf, 0x7a7af48e, 0xaeae47e9, 0x08081018);
break;
case 6:
AS_UINT4(&sharedMemory[0x1C0]) = make_uint4(0xbaba6fd5, 0x7878f088, 0x25254a6f, 0x2e2e5c72);
AS_UINT4(&sharedMemory[0x1C4]) = make_uint4(0x1c1c3824, 0xa6a657f1, 0xb4b473c7, 0xc6c69751);
AS_UINT4(&sharedMemory[0x1C8]) = make_uint4(0xe8e8cb23, 0xdddda17c, 0x7474e89c, 0x1f1f3e21);
AS_UINT4(&sharedMemory[0x1CC]) = make_uint4(0x4b4b96dd, 0xbdbd61dc, 0x8b8b0d86, 0x8a8a0f85);
AS_UINT4(&sharedMemory[0x1D0]) = make_uint4(0x7070e090, 0x3e3e7c42, 0xb5b571c4, 0x6666ccaa);
AS_UINT4(&sharedMemory[0x1D4]) = make_uint4(0x484890d8, 0x03030605, 0xf6f6f701, 0x0e0e1c12);
AS_UINT4(&sharedMemory[0x1D8]) = make_uint4(0x6161c2a3, 0x35356a5f, 0x5757aef9, 0xb9b969d0);
AS_UINT4(&sharedMemory[0x1DC]) = make_uint4(0x86861791, 0xc1c19958, 0x1d1d3a27, 0x9e9e27b9);
break;
case 7:
AS_UINT4(&sharedMemory[0x1E0]) = make_uint4(0xe1e1d938, 0xf8f8eb13, 0x98982bb3, 0x11112233);
AS_UINT4(&sharedMemory[0x1E4]) = make_uint4(0x6969d2bb, 0xd9d9a970, 0x8e8e0789, 0x949433a7);
AS_UINT4(&sharedMemory[0x1E8]) = make_uint4(0x9b9b2db6, 0x1e1e3c22, 0x87871592, 0xe9e9c920);
AS_UINT4(&sharedMemory[0x1EC]) = make_uint4(0xcece8749, 0x5555aaff, 0x28285078, 0xdfdfa57a);
AS_UINT4(&sharedMemory[0x1F0]) = make_uint4(0x8c8c038f, 0xa1a159f8, 0x89890980, 0x0d0d1a17);
AS_UINT4(&sharedMemory[0x1F4]) = make_uint4(0xbfbf65da, 0xe6e6d731, 0x424284c6, 0x6868d0b8);
AS_UINT4(&sharedMemory[0x1F8]) = make_uint4(0x414182c3, 0x999929b0, 0x2d2d5a77, 0x0f0f1e11);
AS_UINT4(&sharedMemory[0x1FC]) = make_uint4(0xb0b07bcb, 0x5454a8fc, 0xbbbb6dd6, 0x16162c3a);
break;
}
// AES 2
switch (threadIdx.x) {
case 0:
AS_UINT4(&sharedMemory[0x200]) = make_uint4(0x63c6a563, 0x7cf8847c, 0x77ee9977, 0x7bf68d7b);
AS_UINT4(&sharedMemory[0x204]) = make_uint4(0xf2ff0df2, 0x6bd6bd6b, 0x6fdeb16f, 0xc59154c5);
AS_UINT4(&sharedMemory[0x208]) = make_uint4(0x30605030, 0x01020301, 0x67cea967, 0x2b567d2b);
AS_UINT4(&sharedMemory[0x20C]) = make_uint4(0xfee719fe, 0xd7b562d7, 0xab4de6ab, 0x76ec9a76);
AS_UINT4(&sharedMemory[0x210]) = make_uint4(0xca8f45ca, 0x821f9d82, 0xc98940c9, 0x7dfa877d);
AS_UINT4(&sharedMemory[0x214]) = make_uint4(0xfaef15fa, 0x59b2eb59, 0x478ec947, 0xf0fb0bf0);
AS_UINT4(&sharedMemory[0x218]) = make_uint4(0xad41ecad, 0xd4b367d4, 0xa25ffda2, 0xaf45eaaf);
AS_UINT4(&sharedMemory[0x21C]) = make_uint4(0x9c23bf9c, 0xa453f7a4, 0x72e49672, 0xc09b5bc0);
break;
case 1:
AS_UINT4(&sharedMemory[0x220]) = make_uint4(0xb775c2b7, 0xfde11cfd, 0x933dae93, 0x264c6a26);
AS_UINT4(&sharedMemory[0x224]) = make_uint4(0x366c5a36, 0x3f7e413f, 0xf7f502f7, 0xcc834fcc);
AS_UINT4(&sharedMemory[0x228]) = make_uint4(0x34685c34, 0xa551f4a5, 0xe5d134e5, 0xf1f908f1);
AS_UINT4(&sharedMemory[0x22C]) = make_uint4(0x71e29371, 0xd8ab73d8, 0x31625331, 0x152a3f15);
AS_UINT4(&sharedMemory[0x230]) = make_uint4(0x04080c04, 0xc79552c7, 0x23466523, 0xc39d5ec3);
AS_UINT4(&sharedMemory[0x234]) = make_uint4(0x18302818, 0x9637a196, 0x050a0f05, 0x9a2fb59a);
AS_UINT4(&sharedMemory[0x238]) = make_uint4(0x070e0907, 0x12243612, 0x801b9b80, 0xe2df3de2);
AS_UINT4(&sharedMemory[0x23C]) = make_uint4(0xebcd26eb, 0x274e6927, 0xb27fcdb2, 0x75ea9f75);
break;
case 2:
AS_UINT4(&sharedMemory[0x240]) = make_uint4(0x09121b09, 0x831d9e83, 0x2c58742c, 0x1a342e1a);
AS_UINT4(&sharedMemory[0x244]) = make_uint4(0x1b362d1b, 0x6edcb26e, 0x5ab4ee5a, 0xa05bfba0);
AS_UINT4(&sharedMemory[0x248]) = make_uint4(0x52a4f652, 0x3b764d3b, 0xd6b761d6, 0xb37dceb3);
AS_UINT4(&sharedMemory[0x24C]) = make_uint4(0x29527b29, 0xe3dd3ee3, 0x2f5e712f, 0x84139784);
AS_UINT4(&sharedMemory[0x250]) = make_uint4(0x53a6f553, 0xd1b968d1, 0x00000000, 0xedc12ced);
AS_UINT4(&sharedMemory[0x254]) = make_uint4(0x20406020, 0xfce31ffc, 0xb179c8b1, 0x5bb6ed5b);
AS_UINT4(&sharedMemory[0x258]) = make_uint4(0x6ad4be6a, 0xcb8d46cb, 0xbe67d9be, 0x39724b39);
AS_UINT4(&sharedMemory[0x25C]) = make_uint4(0x4a94de4a, 0x4c98d44c, 0x58b0e858, 0xcf854acf);
break;
case 3:
AS_UINT4(&sharedMemory[0x260]) = make_uint4(0xd0bb6bd0, 0xefc52aef, 0xaa4fe5aa, 0xfbed16fb);
AS_UINT4(&sharedMemory[0x264]) = make_uint4(0x4386c543, 0x4d9ad74d, 0x33665533, 0x85119485);
AS_UINT4(&sharedMemory[0x268]) = make_uint4(0x458acf45, 0xf9e910f9, 0x02040602, 0x7ffe817f);
AS_UINT4(&sharedMemory[0x26C]) = make_uint4(0x50a0f050, 0x3c78443c, 0x9f25ba9f, 0xa84be3a8);
AS_UINT4(&sharedMemory[0x270]) = make_uint4(0x51a2f351, 0xa35dfea3, 0x4080c040, 0x8f058a8f);
AS_UINT4(&sharedMemory[0x274]) = make_uint4(0x923fad92, 0x9d21bc9d, 0x38704838, 0xf5f104f5);
AS_UINT4(&sharedMemory[0x278]) = make_uint4(0xbc63dfbc, 0xb677c1b6, 0xdaaf75da, 0x21426321);
AS_UINT4(&sharedMemory[0x27C]) = make_uint4(0x10203010, 0xffe51aff, 0xf3fd0ef3, 0xd2bf6dd2);
break;
case 4:
AS_UINT4(&sharedMemory[0x280]) = make_uint4(0xcd814ccd, 0x0c18140c, 0x13263513, 0xecc32fec);
AS_UINT4(&sharedMemory[0x284]) = make_uint4(0x5fbee15f, 0x9735a297, 0x4488cc44, 0x172e3917);
AS_UINT4(&sharedMemory[0x288]) = make_uint4(0xc49357c4, 0xa755f2a7, 0x7efc827e, 0x3d7a473d);
AS_UINT4(&sharedMemory[0x28C]) = make_uint4(0x64c8ac64, 0x5dbae75d, 0x19322b19, 0x73e69573);
AS_UINT4(&sharedMemory[0x290]) = make_uint4(0x60c0a060, 0x81199881, 0x4f9ed14f, 0xdca37fdc);
AS_UINT4(&sharedMemory[0x294]) = make_uint4(0x22446622, 0x2a547e2a, 0x903bab90, 0x880b8388);
AS_UINT4(&sharedMemory[0x298]) = make_uint4(0x468cca46, 0xeec729ee, 0xb86bd3b8, 0x14283c14);
AS_UINT4(&sharedMemory[0x29C]) = make_uint4(0xdea779de, 0x5ebce25e, 0x0b161d0b, 0xdbad76db);
break;
case 5:
AS_UINT4(&sharedMemory[0x2A0]) = make_uint4(0xe0db3be0, 0x32645632, 0x3a744e3a, 0x0a141e0a);
AS_UINT4(&sharedMemory[0x2A4]) = make_uint4(0x4992db49, 0x060c0a06, 0x24486c24, 0x5cb8e45c);
AS_UINT4(&sharedMemory[0x2A8]) = make_uint4(0xc29f5dc2, 0xd3bd6ed3, 0xac43efac, 0x62c4a662);
AS_UINT4(&sharedMemory[0x2AC]) = make_uint4(0x9139a891, 0x9531a495, 0xe4d337e4, 0x79f28b79);
AS_UINT4(&sharedMemory[0x2B0]) = make_uint4(0xe7d532e7, 0xc88b43c8, 0x376e5937, 0x6ddab76d);
AS_UINT4(&sharedMemory[0x2B4]) = make_uint4(0x8d018c8d, 0xd5b164d5, 0x4e9cd24e, 0xa949e0a9);
AS_UINT4(&sharedMemory[0x2B8]) = make_uint4(0x6cd8b46c, 0x56acfa56, 0xf4f307f4, 0xeacf25ea);
AS_UINT4(&sharedMemory[0x2BC]) = make_uint4(0x65caaf65, 0x7af48e7a, 0xae47e9ae, 0x08101808);
break;
case 6:
AS_UINT4(&sharedMemory[0x2C0]) = make_uint4(0xba6fd5ba, 0x78f08878, 0x254a6f25, 0x2e5c722e);
AS_UINT4(&sharedMemory[0x2C4]) = make_uint4(0x1c38241c, 0xa657f1a6, 0xb473c7b4, 0xc69751c6);
AS_UINT4(&sharedMemory[0x2C8]) = make_uint4(0xe8cb23e8, 0xdda17cdd, 0x74e89c74, 0x1f3e211f);
AS_UINT4(&sharedMemory[0x2CC]) = make_uint4(0x4b96dd4b, 0xbd61dcbd, 0x8b0d868b, 0x8a0f858a);
AS_UINT4(&sharedMemory[0x2D0]) = make_uint4(0x70e09070, 0x3e7c423e, 0xb571c4b5, 0x66ccaa66);
AS_UINT4(&sharedMemory[0x2D4]) = make_uint4(0x4890d848, 0x03060503, 0xf6f701f6, 0x0e1c120e);
AS_UINT4(&sharedMemory[0x2D8]) = make_uint4(0x61c2a361, 0x356a5f35, 0x57aef957, 0xb969d0b9);
AS_UINT4(&sharedMemory[0x2DC]) = make_uint4(0x86179186, 0xc19958c1, 0x1d3a271d, 0x9e27b99e);
break;
case 7:
AS_UINT4(&sharedMemory[0x2E0]) = make_uint4(0xe1d938e1, 0xf8eb13f8, 0x982bb398, 0x11223311);
AS_UINT4(&sharedMemory[0x2E4]) = make_uint4(0x69d2bb69, 0xd9a970d9, 0x8e07898e, 0x9433a794);
AS_UINT4(&sharedMemory[0x2E8]) = make_uint4(0x9b2db69b, 0x1e3c221e, 0x87159287, 0xe9c920e9);
AS_UINT4(&sharedMemory[0x2EC]) = make_uint4(0xce8749ce, 0x55aaff55, 0x28507828, 0xdfa57adf);
AS_UINT4(&sharedMemory[0x2F0]) = make_uint4(0x8c038f8c, 0xa159f8a1, 0x89098089, 0x0d1a170d);
AS_UINT4(&sharedMemory[0x2F4]) = make_uint4(0xbf65dabf, 0xe6d731e6, 0x4284c642, 0x68d0b868);
AS_UINT4(&sharedMemory[0x2F8]) = make_uint4(0x4182c341, 0x9929b099, 0x2d5a772d, 0x0f1e110f);
AS_UINT4(&sharedMemory[0x2FC]) = make_uint4(0xb07bcbb0, 0x54a8fc54, 0xbb6dd6bb, 0x162c3a16);
break;
}
// AES 3
switch (threadIdx.x) {
case 0:
AS_UINT4(&sharedMemory[0x300]) = make_uint4(0xc6a56363, 0xf8847c7c, 0xee997777, 0xf68d7b7b);
AS_UINT4(&sharedMemory[0x304]) = make_uint4(0xff0df2f2, 0xd6bd6b6b, 0xdeb16f6f, 0x9154c5c5);
AS_UINT4(&sharedMemory[0x308]) = make_uint4(0x60503030, 0x02030101, 0xcea96767, 0x567d2b2b);
AS_UINT4(&sharedMemory[0x30C]) = make_uint4(0xe719fefe, 0xb562d7d7, 0x4de6abab, 0xec9a7676);
AS_UINT4(&sharedMemory[0x310]) = make_uint4(0x8f45caca, 0x1f9d8282, 0x8940c9c9, 0xfa877d7d);
AS_UINT4(&sharedMemory[0x314]) = make_uint4(0xef15fafa, 0xb2eb5959, 0x8ec94747, 0xfb0bf0f0);
AS_UINT4(&sharedMemory[0x318]) = make_uint4(0x41ecadad, 0xb367d4d4, 0x5ffda2a2, 0x45eaafaf);
AS_UINT4(&sharedMemory[0x31C]) = make_uint4(0x23bf9c9c, 0x53f7a4a4, 0xe4967272, 0x9b5bc0c0);
break;
case 1:
AS_UINT4(&sharedMemory[0x320]) = make_uint4(0x75c2b7b7, 0xe11cfdfd, 0x3dae9393, 0x4c6a2626);
AS_UINT4(&sharedMemory[0x324]) = make_uint4(0x6c5a3636, 0x7e413f3f, 0xf502f7f7, 0x834fcccc);
AS_UINT4(&sharedMemory[0x328]) = make_uint4(0x685c3434, 0x51f4a5a5, 0xd134e5e5, 0xf908f1f1);
AS_UINT4(&sharedMemory[0x32C]) = make_uint4(0xe2937171, 0xab73d8d8, 0x62533131, 0x2a3f1515);
AS_UINT4(&sharedMemory[0x330]) = make_uint4(0x080c0404, 0x9552c7c7, 0x46652323, 0x9d5ec3c3);
AS_UINT4(&sharedMemory[0x334]) = make_uint4(0x30281818, 0x37a19696, 0x0a0f0505, 0x2fb59a9a);
AS_UINT4(&sharedMemory[0x338]) = make_uint4(0x0e090707, 0x24361212, 0x1b9b8080, 0xdf3de2e2);
AS_UINT4(&sharedMemory[0x33C]) = make_uint4(0xcd26ebeb, 0x4e692727, 0x7fcdb2b2, 0xea9f7575);
break;
case 2:
AS_UINT4(&sharedMemory[0x340]) = make_uint4(0x121b0909, 0x1d9e8383, 0x58742c2c, 0x342e1a1a);
AS_UINT4(&sharedMemory[0x344]) = make_uint4(0x362d1b1b, 0xdcb26e6e, 0xb4ee5a5a, 0x5bfba0a0);
AS_UINT4(&sharedMemory[0x348]) = make_uint4(0xa4f65252, 0x764d3b3b, 0xb761d6d6, 0x7dceb3b3);
AS_UINT4(&sharedMemory[0x34C]) = make_uint4(0x527b2929, 0xdd3ee3e3, 0x5e712f2f, 0x13978484);
AS_UINT4(&sharedMemory[0x350]) = make_uint4(0xa6f55353, 0xb968d1d1, 0x00000000, 0xc12ceded);
AS_UINT4(&sharedMemory[0x354]) = make_uint4(0x40602020, 0xe31ffcfc, 0x79c8b1b1, 0xb6ed5b5b);
AS_UINT4(&sharedMemory[0x358]) = make_uint4(0xd4be6a6a, 0x8d46cbcb, 0x67d9bebe, 0x724b3939);
AS_UINT4(&sharedMemory[0x35C]) = make_uint4(0x94de4a4a, 0x98d44c4c, 0xb0e85858, 0x854acfcf);
break;
case 3:
AS_UINT4(&sharedMemory[0x360]) = make_uint4(0xbb6bd0d0, 0xc52aefef, 0x4fe5aaaa, 0xed16fbfb);
AS_UINT4(&sharedMemory[0x364]) = make_uint4(0x86c54343, 0x9ad74d4d, 0x66553333, 0x11948585);
AS_UINT4(&sharedMemory[0x368]) = make_uint4(0x8acf4545, 0xe910f9f9, 0x04060202, 0xfe817f7f);
AS_UINT4(&sharedMemory[0x36C]) = make_uint4(0xa0f05050, 0x78443c3c, 0x25ba9f9f, 0x4be3a8a8);
AS_UINT4(&sharedMemory[0x370]) = make_uint4(0xa2f35151, 0x5dfea3a3, 0x80c04040, 0x058a8f8f);
AS_UINT4(&sharedMemory[0x374]) = make_uint4(0x3fad9292, 0x21bc9d9d, 0x70483838, 0xf104f5f5);
AS_UINT4(&sharedMemory[0x378]) = make_uint4(0x63dfbcbc, 0x77c1b6b6, 0xaf75dada, 0x42632121);
AS_UINT4(&sharedMemory[0x37C]) = make_uint4(0x20301010, 0xe51affff, 0xfd0ef3f3, 0xbf6dd2d2);
break;
case 4:
AS_UINT4(&sharedMemory[0x380]) = make_uint4(0x814ccdcd, 0x18140c0c, 0x26351313, 0xc32fecec);
AS_UINT4(&sharedMemory[0x384]) = make_uint4(0xbee15f5f, 0x35a29797, 0x88cc4444, 0x2e391717);
AS_UINT4(&sharedMemory[0x388]) = make_uint4(0x9357c4c4, 0x55f2a7a7, 0xfc827e7e, 0x7a473d3d);
AS_UINT4(&sharedMemory[0x38C]) = make_uint4(0xc8ac6464, 0xbae75d5d, 0x322b1919, 0xe6957373);
AS_UINT4(&sharedMemory[0x390]) = make_uint4(0xc0a06060, 0x19988181, 0x9ed14f4f, 0xa37fdcdc);
AS_UINT4(&sharedMemory[0x394]) = make_uint4(0x44662222, 0x547e2a2a, 0x3bab9090, 0x0b838888);
AS_UINT4(&sharedMemory[0x398]) = make_uint4(0x8cca4646, 0xc729eeee, 0x6bd3b8b8, 0x283c1414);
AS_UINT4(&sharedMemory[0x39C]) = make_uint4(0xa779dede, 0xbce25e5e, 0x161d0b0b, 0xad76dbdb);
break;
case 5:
AS_UINT4(&sharedMemory[0x3A0]) = make_uint4(0xdb3be0e0, 0x64563232, 0x744e3a3a, 0x141e0a0a);
AS_UINT4(&sharedMemory[0x3A4]) = make_uint4(0x92db4949, 0x0c0a0606, 0x486c2424, 0xb8e45c5c);
AS_UINT4(&sharedMemory[0x3A8]) = make_uint4(0x9f5dc2c2, 0xbd6ed3d3, 0x43efacac, 0xc4a66262);
AS_UINT4(&sharedMemory[0x3AC]) = make_uint4(0x39a89191, 0x31a49595, 0xd337e4e4, 0xf28b7979);
AS_UINT4(&sharedMemory[0x3B0]) = make_uint4(0xd532e7e7, 0x8b43c8c8, 0x6e593737, 0xdab76d6d);
AS_UINT4(&sharedMemory[0x3B4]) = make_uint4(0x018c8d8d, 0xb164d5d5, 0x9cd24e4e, 0x49e0a9a9);
AS_UINT4(&sharedMemory[0x3B8]) = make_uint4(0xd8b46c6c, 0xacfa5656, 0xf307f4f4, 0xcf25eaea);
AS_UINT4(&sharedMemory[0x3BC]) = make_uint4(0xcaaf6565, 0xf48e7a7a, 0x47e9aeae, 0x10180808);
break;
case 6:
AS_UINT4(&sharedMemory[0x3C0]) = make_uint4(0x6fd5baba, 0xf0887878, 0x4a6f2525, 0x5c722e2e);
AS_UINT4(&sharedMemory[0x3C4]) = make_uint4(0x38241c1c, 0x57f1a6a6, 0x73c7b4b4, 0x9751c6c6);
AS_UINT4(&sharedMemory[0x3C8]) = make_uint4(0xcb23e8e8, 0xa17cdddd, 0xe89c7474, 0x3e211f1f);
AS_UINT4(&sharedMemory[0x3CC]) = make_uint4(0x96dd4b4b, 0x61dcbdbd, 0x0d868b8b, 0x0f858a8a);
AS_UINT4(&sharedMemory[0x3D0]) = make_uint4(0xe0907070, 0x7c423e3e, 0x71c4b5b5, 0xccaa6666);
AS_UINT4(&sharedMemory[0x3D4]) = make_uint4(0x90d84848, 0x06050303, 0xf701f6f6, 0x1c120e0e);
AS_UINT4(&sharedMemory[0x3D8]) = make_uint4(0xc2a36161, 0x6a5f3535, 0xaef95757, 0x69d0b9b9);
AS_UINT4(&sharedMemory[0x3DC]) = make_uint4(0x17918686, 0x9958c1c1, 0x3a271d1d, 0x27b99e9e);
break;
case 7:
AS_UINT4(&sharedMemory[0x3E0]) = make_uint4(0xd938e1e1, 0xeb13f8f8, 0x2bb39898, 0x22331111);
AS_UINT4(&sharedMemory[0x3E4]) = make_uint4(0xd2bb6969, 0xa970d9d9, 0x07898e8e, 0x33a79494);
AS_UINT4(&sharedMemory[0x3E8]) = make_uint4(0x2db69b9b, 0x3c221e1e, 0x15928787, 0xc920e9e9);
AS_UINT4(&sharedMemory[0x3EC]) = make_uint4(0x8749cece, 0xaaff5555, 0x50782828, 0xa57adfdf);
AS_UINT4(&sharedMemory[0x3F0]) = make_uint4(0x038f8c8c, 0x59f8a1a1, 0x09808989, 0x1a170d0d);
AS_UINT4(&sharedMemory[0x3F4]) = make_uint4(0x65dabfbf, 0xd731e6e6, 0x84c64242, 0xd0b86868);
AS_UINT4(&sharedMemory[0x3F8]) = make_uint4(0x82c34141, 0x29b09999, 0x5a772d2d, 0x1e110f0f);
AS_UINT4(&sharedMemory[0x3FC]) = make_uint4(0x7bcbb0b0, 0xa8fc5454, 0x6dd6bbbb, 0x2c3a1616);
break;
}
}
+180
View File
@@ -0,0 +1,180 @@
//#include <cuda_helper.h>
typedef struct {
uint32_t h[8], s[4], t[2];
int buflen, nullt;
uint8_t buf[64];
} blake_state;
#define U8TO32(p) \
(((uint32_t)((p)[0]) << 24) | ((uint32_t)((p)[1]) << 16) | \
((uint32_t)((p)[2]) << 8) | ((uint32_t)((p)[3]) ))
#define U32TO8(p, v) \
(p)[0] = (uint8_t)((v) >> 24); (p)[1] = (uint8_t)((v) >> 16); \
(p)[2] = (uint8_t)((v) >> 8); (p)[3] = (uint8_t)((v) );
#define BLAKE_ROT(x,n) ROTR32(x, n)
#define BLAKE_G(a,b,c,d,e) \
v[a] += (m[d_blake_sigma[i][e]] ^ d_blake_cst[d_blake_sigma[i][e+1]]) + v[b]; \
v[d] = BLAKE_ROT(v[d] ^ v[a],16); \
v[c] += v[d]; \
v[b] = BLAKE_ROT(v[b] ^ v[c],12); \
v[a] += (m[d_blake_sigma[i][e+1]] ^ d_blake_cst[d_blake_sigma[i][e]]) + v[b]; \
v[d] = BLAKE_ROT(v[d] ^ v[a], 8); \
v[c] += v[d]; \
v[b] = BLAKE_ROT(v[b] ^ v[c], 7);
__constant__ uint8_t d_blake_sigma[14][16] = {
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
{11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4},
{7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8},
{9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13},
{2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9},
{12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11},
{13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10},
{6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5},
{10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0},
{0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15},
{14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3},
{11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4},
{7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8}
};
__constant__ uint32_t d_blake_cst[16] = {
0x243F6A88, 0x85A308D3, 0x13198A2E, 0x03707344,
0xA4093822, 0x299F31D0, 0x082EFA98, 0xEC4E6C89,
0x452821E6, 0x38D01377, 0xBE5466CF, 0x34E90C6C,
0xC0AC29B7, 0xC97C50DD, 0x3F84D5B5, 0xB5470917
};
__device__
void cn_blake_compress(blake_state * __restrict__ S, const uint8_t * __restrict__ block)
{
uint32_t v[16], m[16], i;
for (i = 0; i < 16; ++i) m[i] = U8TO32(block + i * 4);
for (i = 0; i < 8; ++i) v[i] = S->h[i];
v[ 8] = S->s[0] ^ 0x243F6A88;
v[ 9] = S->s[1] ^ 0x85A308D3;
v[10] = S->s[2] ^ 0x13198A2E;
v[11] = S->s[3] ^ 0x03707344;
v[12] = 0xA4093822;
v[13] = 0x299F31D0;
v[14] = 0x082EFA98;
v[15] = 0xEC4E6C89;
if (S->nullt == 0) {
v[12] ^= S->t[0];
v[13] ^= S->t[0];
v[14] ^= S->t[1];
v[15] ^= S->t[1];
}
for (i = 0; i < 14; ++i) {
BLAKE_G(0, 4, 8, 12, 0);
BLAKE_G(1, 5, 9, 13, 2);
BLAKE_G(2, 6, 10, 14, 4);
BLAKE_G(3, 7, 11, 15, 6);
BLAKE_G(3, 4, 9, 14, 14);
BLAKE_G(2, 7, 8, 13, 12);
BLAKE_G(0, 5, 10, 15, 8);
BLAKE_G(1, 6, 11, 12, 10);
}
for (i = 0; i < 16; ++i) S->h[i % 8] ^= v[i];
for (i = 0; i < 8; ++i) S->h[i] ^= S->s[i % 4];
}
__device__ void cn_blake_update(blake_state * __restrict__ S, const uint8_t * __restrict__ data, uint64_t datalen)
{
int left = S->buflen >> 3;
int fill = 64 - left;
if (left && (((datalen >> 3) & 0x3F) >= (unsigned) fill)) {
memcpy((void *) (S->buf + left), (void *) data, fill);
S->t[0] += 512;
if (S->t[0] == 0) S->t[1]++;
cn_blake_compress(S, S->buf);
data += fill;
datalen -= (fill << 3);
left = 0;
}
while (datalen >= 512) {
S->t[0] += 512;
if (S->t[0] == 0) S->t[1]++;
cn_blake_compress(S, data);
data += 64;
datalen -= 512;
}
if (datalen > 0) {
memcpy((void *) (S->buf + left), (void *) data, datalen >> 3);
S->buflen = (left << 3) + datalen;
} else {
S->buflen = 0;
}
}
__device__
void cn_blake_final(blake_state * __restrict__ S, uint8_t * __restrict__ digest)
{
const uint8_t padding[] = {
0x80,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0
};
uint8_t pa = 0x81, pb = 0x01;
uint8_t msglen[8];
uint32_t lo = S->t[0] + S->buflen, hi = S->t[1];
if (lo < (unsigned) S->buflen) hi++;
U32TO8(msglen + 0, hi);
U32TO8(msglen + 4, lo);
if (S->buflen == 440) {
S->t[0] -= 8;
cn_blake_update(S, &pa, 8);
} else {
if (S->buflen < 440) {
if (S->buflen == 0) S->nullt = 1;
S->t[0] -= 440 - S->buflen;
cn_blake_update(S, padding, 440 - S->buflen);
} else {
S->t[0] -= 512 - S->buflen;
cn_blake_update(S, padding, 512 - S->buflen);
S->t[0] -= 440;
cn_blake_update(S, padding + 1, 440);
S->nullt = 1;
}
cn_blake_update(S, &pb, 8);
S->t[0] -= 8;
}
S->t[0] -= 64;
cn_blake_update(S, msglen, 64);
U32TO8(digest + 0, S->h[0]);
U32TO8(digest + 4, S->h[1]);
U32TO8(digest + 8, S->h[2]);
U32TO8(digest + 12, S->h[3]);
U32TO8(digest + 16, S->h[4]);
U32TO8(digest + 20, S->h[5]);
U32TO8(digest + 24, S->h[6]);
U32TO8(digest + 28, S->h[7]);
}
__device__
void cn_blake(const uint8_t * __restrict__ in, uint64_t inlen, uint32_t * out)
{
blake_state bs;
blake_state *S = (blake_state *)&bs;
S->h[0] = 0x6A09E667; S->h[1] = 0xBB67AE85; S->h[2] = 0x3C6EF372;
S->h[3] = 0xA54FF53A; S->h[4] = 0x510E527F; S->h[5] = 0x9B05688C;
S->h[6] = 0x1F83D9AB; S->h[7] = 0x5BE0CD19;
S->t[0] = S->t[1] = S->buflen = S->nullt = 0;
S->s[0] = S->s[1] = S->s[2] = S->s[3] = 0;
cn_blake_update(S, (uint8_t *)in, inlen * 8);
cn_blake_final(S, (uint8_t *)out);
}
+348
View File
@@ -0,0 +1,348 @@
#define GROESTL_ROWS 8
#define GROESTL_LENGTHFIELDLEN GROESTL_ROWS
#define GROESTL_COLS512 8
#define GROESTL_SIZE512 (GROESTL_ROWS*GROESTL_COLS512)
#define GROESTL_ROUNDS512 10
#define GROESTL_HASH_BIT_LEN 256
#define GROESTL_ROTL32(v, n) ROTL32(v, n)
#define li_32(h) 0x##h##u
#define GROESTL_EXT_BYTE(var,n) ((uint8_t)((uint32_t)(var) >> (8*n)))
#define u32BIG(a) \
((GROESTL_ROTL32(a,8) & li_32(00FF00FF)) | \
(GROESTL_ROTL32(a,24) & li_32(FF00FF00)))
typedef struct {
uint32_t chaining[GROESTL_SIZE512/sizeof(uint32_t)];
uint32_t block_counter1, block_counter2;
BitSequence buffer[GROESTL_SIZE512];
int buf_ptr;
int bits_in_last_byte;
} groestlHashState;
__constant__ uint32_t d_groestl_T[512] = {
0xa5f432c6, 0xc6a597f4, 0x84976ff8, 0xf884eb97, 0x99b05eee, 0xee99c7b0, 0x8d8c7af6, 0xf68df78c, 0x0d17e8ff, 0xff0de517, 0xbddc0ad6, 0xd6bdb7dc, 0xb1c816de, 0xdeb1a7c8, 0x54fc6d91, 0x915439fc,
0x50f09060, 0x6050c0f0, 0x03050702, 0x02030405, 0xa9e02ece, 0xcea987e0, 0x7d87d156, 0x567dac87, 0x192bcce7, 0xe719d52b, 0x62a613b5, 0xb56271a6, 0xe6317c4d, 0x4de69a31, 0x9ab559ec, 0xec9ac3b5,
0x45cf408f, 0x8f4505cf, 0x9dbca31f, 0x1f9d3ebc, 0x40c04989, 0x894009c0, 0x879268fa, 0xfa87ef92, 0x153fd0ef, 0xef15c53f, 0xeb2694b2, 0xb2eb7f26, 0xc940ce8e, 0x8ec90740, 0x0b1de6fb, 0xfb0bed1d,
0xec2f6e41, 0x41ec822f, 0x67a91ab3, 0xb3677da9, 0xfd1c435f, 0x5ffdbe1c, 0xea256045, 0x45ea8a25, 0xbfdaf923, 0x23bf46da, 0xf7025153, 0x53f7a602, 0x96a145e4, 0xe496d3a1, 0x5bed769b, 0x9b5b2ded,
0xc25d2875, 0x75c2ea5d, 0x1c24c5e1, 0xe11cd924, 0xaee9d43d, 0x3dae7ae9, 0x6abef24c, 0x4c6a98be, 0x5aee826c, 0x6c5ad8ee, 0x41c3bd7e, 0x7e41fcc3, 0x0206f3f5, 0xf502f106, 0x4fd15283, 0x834f1dd1,
0x5ce48c68, 0x685cd0e4, 0xf4075651, 0x51f4a207, 0x345c8dd1, 0xd134b95c, 0x0818e1f9, 0xf908e918, 0x93ae4ce2, 0xe293dfae, 0x73953eab, 0xab734d95, 0x53f59762, 0x6253c4f5, 0x3f416b2a, 0x2a3f5441,
0x0c141c08, 0x080c1014, 0x52f66395, 0x955231f6, 0x65afe946, 0x46658caf, 0x5ee27f9d, 0x9d5e21e2, 0x28784830, 0x30286078, 0xa1f8cf37, 0x37a16ef8, 0x0f111b0a, 0x0a0f1411, 0xb5c4eb2f, 0x2fb55ec4,
0x091b150e, 0x0e091c1b, 0x365a7e24, 0x2436485a, 0x9bb6ad1b, 0x1b9b36b6, 0x3d4798df, 0xdf3da547, 0x266aa7cd, 0xcd26816a, 0x69bbf54e, 0x4e699cbb, 0xcd4c337f, 0x7fcdfe4c, 0x9fba50ea, 0xea9fcfba,
0x1b2d3f12, 0x121b242d, 0x9eb9a41d, 0x1d9e3ab9, 0x749cc458, 0x5874b09c, 0x2e724634, 0x342e6872, 0x2d774136, 0x362d6c77, 0xb2cd11dc, 0xdcb2a3cd, 0xee299db4, 0xb4ee7329, 0xfb164d5b, 0x5bfbb616,
0xf601a5a4, 0xa4f65301, 0x4dd7a176, 0x764decd7, 0x61a314b7, 0xb76175a3, 0xce49347d, 0x7dcefa49, 0x7b8ddf52, 0x527ba48d, 0x3e429fdd, 0xdd3ea142, 0x7193cd5e, 0x5e71bc93, 0x97a2b113, 0x139726a2,
0xf504a2a6, 0xa6f55704, 0x68b801b9, 0xb96869b8, 0, 0, 0x2c74b5c1, 0xc12c9974, 0x60a0e040, 0x406080a0, 0x1f21c2e3, 0xe31fdd21, 0xc8433a79, 0x79c8f243, 0xed2c9ab6, 0xb6ed772c,
0xbed90dd4, 0xd4beb3d9, 0x46ca478d, 0x8d4601ca, 0xd9701767, 0x67d9ce70, 0x4bddaf72, 0x724be4dd, 0xde79ed94, 0x94de3379, 0xd467ff98, 0x98d42b67, 0xe82393b0, 0xb0e87b23, 0x4ade5b85, 0x854a11de,
0x6bbd06bb, 0xbb6b6dbd, 0x2a7ebbc5, 0xc52a917e, 0xe5347b4f, 0x4fe59e34, 0x163ad7ed, 0xed16c13a, 0xc554d286, 0x86c51754, 0xd762f89a, 0x9ad72f62, 0x55ff9966, 0x6655ccff, 0x94a7b611, 0x119422a7,
0xcf4ac08a, 0x8acf0f4a, 0x1030d9e9, 0xe910c930, 0x060a0e04, 0x0406080a, 0x819866fe, 0xfe81e798, 0xf00baba0, 0xa0f05b0b, 0x44ccb478, 0x7844f0cc, 0xbad5f025, 0x25ba4ad5, 0xe33e754b, 0x4be3963e,
0xf30eaca2, 0xa2f35f0e, 0xfe19445d, 0x5dfeba19, 0xc05bdb80, 0x80c01b5b, 0x8a858005, 0x058a0a85, 0xadecd33f, 0x3fad7eec, 0xbcdffe21, 0x21bc42df, 0x48d8a870, 0x7048e0d8, 0x040cfdf1, 0xf104f90c,
0xdf7a1963, 0x63dfc67a, 0xc1582f77, 0x77c1ee58, 0x759f30af, 0xaf75459f, 0x63a5e742, 0x426384a5, 0x30507020, 0x20304050, 0x1a2ecbe5, 0xe51ad12e, 0x0e12effd, 0xfd0ee112, 0x6db708bf, 0xbf6d65b7,
0x4cd45581, 0x814c19d4, 0x143c2418, 0x1814303c, 0x355f7926, 0x26354c5f, 0x2f71b2c3, 0xc32f9d71, 0xe13886be, 0xbee16738, 0xa2fdc835, 0x35a26afd, 0xcc4fc788, 0x88cc0b4f, 0x394b652e, 0x2e395c4b,
0x57f96a93, 0x93573df9, 0xf20d5855, 0x55f2aa0d, 0x829d61fc, 0xfc82e39d, 0x47c9b37a, 0x7a47f4c9, 0xacef27c8, 0xc8ac8bef, 0xe73288ba, 0xbae76f32, 0x2b7d4f32, 0x322b647d, 0x95a442e6, 0xe695d7a4,
0xa0fb3bc0, 0xc0a09bfb, 0x98b3aa19, 0x199832b3, 0xd168f69e, 0x9ed12768, 0x7f8122a3, 0xa37f5d81, 0x66aaee44, 0x446688aa, 0x7e82d654, 0x547ea882, 0xabe6dd3b, 0x3bab76e6, 0x839e950b, 0xb83169e,
0xca45c98c, 0x8cca0345, 0x297bbcc7, 0xc729957b, 0xd36e056b, 0x6bd3d66e, 0x3c446c28, 0x283c5044, 0x798b2ca7, 0xa779558b, 0xe23d81bc, 0xbce2633d, 0x1d273116, 0x161d2c27, 0x769a37ad, 0xad76419a,
0x3b4d96db, 0xdb3bad4d, 0x56fa9e64, 0x6456c8fa, 0x4ed2a674, 0x744ee8d2, 0x1e223614, 0x141e2822, 0xdb76e492, 0x92db3f76, 0x0a1e120c, 0x0c0a181e, 0x6cb4fc48, 0x486c90b4, 0xe4378fb8, 0xb8e46b37,
0x5de7789f, 0x9f5d25e7, 0x6eb20fbd, 0xbd6e61b2, 0xef2a6943, 0x43ef862a, 0xa6f135c4, 0xc4a693f1, 0xa8e3da39, 0x39a872e3, 0xa4f7c631, 0x31a462f7, 0x37598ad3, 0xd337bd59, 0x8b8674f2, 0xf28bff86,
0x325683d5, 0xd532b156, 0x43c54e8b, 0x8b430dc5, 0x59eb856e, 0x6e59dceb, 0xb7c218da, 0xdab7afc2, 0x8c8f8e01, 0x018c028f, 0x64ac1db1, 0xb16479ac, 0xd26df19c, 0x9cd2236d, 0xe03b7249, 0x49e0923b,
0xb4c71fd8, 0xd8b4abc7, 0xfa15b9ac, 0xacfa4315, 0x0709faf3, 0xf307fd09, 0x256fa0cf, 0xcf25856f, 0xafea20ca, 0xcaaf8fea, 0x8e897df4, 0xf48ef389, 0xe9206747, 0x47e98e20, 0x18283810, 0x10182028,
0xd5640b6f, 0x6fd5de64, 0x888373f0, 0xf088fb83, 0x6fb1fb4a, 0x4a6f94b1, 0x7296ca5c, 0x5c72b896, 0x246c5438, 0x3824706c, 0xf1085f57, 0x57f1ae08, 0xc7522173, 0x73c7e652, 0x51f36497, 0x975135f3,
0x2365aecb, 0xcb238d65, 0x7c8425a1, 0xa17c5984, 0x9cbf57e8, 0xe89ccbbf, 0x21635d3e, 0x3e217c63, 0xdd7cea96, 0x96dd377c, 0xdc7f1e61, 0x61dcc27f, 0x86919c0d, 0x0d861a91, 0x85949b0f, 0xf851e94,
0x90ab4be0, 0xe090dbab, 0x42c6ba7c, 0x7c42f8c6, 0xc4572671, 0x71c4e257, 0xaae529cc, 0xccaa83e5, 0xd873e390, 0x90d83b73, 0x050f0906, 0x06050c0f, 0x0103f4f7, 0xf701f503, 0x12362a1c, 0x1c123836,
0xa3fe3cc2, 0xc2a39ffe, 0x5fe18b6a, 0x6a5fd4e1, 0xf910beae, 0xaef94710, 0xd06b0269, 0x69d0d26b, 0x91a8bf17, 0x17912ea8, 0x58e87199, 0x995829e8, 0x2769533a, 0x3a277469, 0xb9d0f727, 0x27b94ed0,
0x384891d9, 0xd938a948, 0x1335deeb, 0xeb13cd35, 0xb3cee52b, 0x2bb356ce, 0x33557722, 0x22334455, 0xbbd604d2, 0xd2bbbfd6, 0x709039a9, 0xa9704990, 0x89808707, 0x07890e80, 0xa7f2c133, 0x33a766f2,
0xb6c1ec2d, 0x2db65ac1, 0x22665a3c, 0x3c227866, 0x92adb815, 0x15922aad, 0x2060a9c9, 0xc9208960, 0x49db5c87, 0x874915db, 0xff1ab0aa, 0xaaff4f1a, 0x7888d850, 0x5078a088, 0x7a8e2ba5, 0xa57a518e,
0x8f8a8903, 0x038f068a, 0xf8134a59, 0x59f8b213, 0x809b9209, 0x0980129b, 0x1739231a, 0x1a173439, 0xda751065, 0x65daca75, 0x315384d7, 0xd731b553, 0xc651d584, 0x84c61351, 0xb8d303d0, 0xd0b8bbd3,
0xc35edc82, 0x82c31f5e, 0xb0cbe229, 0x29b052cb, 0x7799c35a, 0x5a77b499, 0x11332d1e, 0x1e113c33, 0xcb463d7b, 0x7bcbf646, 0xfc1fb7a8, 0xa8fc4b1f, 0xd6610c6d, 0x6dd6da61, 0x3a4e622c, 0x2c3a584e
};
#define GROESTL_ROTATE_COLUMN_DOWN(v1, v2, amount_bytes, temp_var) { \
temp_var = (v1<<(8*amount_bytes))|(v2>>(8*(4-amount_bytes))); \
v2 = (v2<<(8*amount_bytes))|(v1>>(8*(4-amount_bytes))); \
v1 = temp_var; \
}
#define GROESTL_COLUMN(x,y,i, c0,c1,c2,c3,c4,c5,c6,c7, tv1,tv2,tu,tl,t) \
tu = d_groestl_T[2*(uint32_t)x[4*c0+0]]; \
tl = d_groestl_T[2*(uint32_t)x[4*c0+0]+1]; \
tv1 = d_groestl_T[2*(uint32_t)x[4*c1+1]]; \
tv2 = d_groestl_T[2*(uint32_t)x[4*c1+1]+1]; \
GROESTL_ROTATE_COLUMN_DOWN(tv1,tv2,1,t) \
tu ^= tv1; \
tl ^= tv2; \
tv1 = d_groestl_T[2*(uint32_t)x[4*c2+2]]; \
tv2 = d_groestl_T[2*(uint32_t)x[4*c2+2]+1]; \
GROESTL_ROTATE_COLUMN_DOWN(tv1,tv2,2,t) \
tu ^= tv1; \
tl ^= tv2; \
tv1 = d_groestl_T[2*(uint32_t)x[4*c3+3]]; \
tv2 = d_groestl_T[2*(uint32_t)x[4*c3+3]+1]; \
GROESTL_ROTATE_COLUMN_DOWN(tv1,tv2,3,t) \
tu ^= tv1; \
tl ^= tv2; \
tl ^= d_groestl_T[2*(uint32_t)x[4*c4+0]]; \
tu ^= d_groestl_T[2*(uint32_t)x[4*c4+0]+1]; \
tv1 = d_groestl_T[2*(uint32_t)x[4*c5+1]]; \
tv2 = d_groestl_T[2*(uint32_t)x[4*c5+1]+1]; \
GROESTL_ROTATE_COLUMN_DOWN(tv1,tv2,1,t) \
tl ^= tv1; \
tu ^= tv2; \
tv1 = d_groestl_T[2*(uint32_t)x[4*c6+2]]; \
tv2 = d_groestl_T[2*(uint32_t)x[4*c6+2]+1]; \
GROESTL_ROTATE_COLUMN_DOWN(tv1,tv2,2,t) \
tl ^= tv1; \
tu ^= tv2; \
tv1 = d_groestl_T[2*(uint32_t)x[4*c7+3]]; \
tv2 = d_groestl_T[2*(uint32_t)x[4*c7+3]+1]; \
GROESTL_ROTATE_COLUMN_DOWN(tv1,tv2,3,t) \
tl ^= tv1; \
tu ^= tv2; \
y[i] = tu; \
y[i+1] = tl;
__device__
void cn_groestl_RND512P(uint8_t * __restrict__ x, uint32_t * __restrict__ y, uint32_t r)
{
uint32_t temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp;
uint32_t* x32 = (uint32_t*)x;
x32[ 0] ^= 0x00000000^r;
x32[ 2] ^= 0x00000010^r;
x32[ 4] ^= 0x00000020^r;
x32[ 6] ^= 0x00000030^r;
x32[ 8] ^= 0x00000040^r;
x32[10] ^= 0x00000050^r;
x32[12] ^= 0x00000060^r;
x32[14] ^= 0x00000070^r;
GROESTL_COLUMN(x,y, 0, 0, 2, 4, 6, 9, 11, 13, 15, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 2, 2, 4, 6, 8, 11, 13, 15, 1, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 4, 4, 6, 8, 10, 13, 15, 1, 3, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 6, 6, 8, 10, 12, 15, 1, 3, 5, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 8, 8, 10, 12, 14, 1, 3, 5, 7, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y,10, 10, 12, 14, 0, 3, 5, 7, 9, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y,12, 12, 14, 0, 2, 5, 7, 9, 11, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y,14, 14, 0, 2, 4, 7, 9, 11, 13, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
}
__device__
void cn_groestl_RND512Q(uint8_t * __restrict__ x, uint32_t * __restrict__ y, uint32_t r)
{
uint32_t temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp;
uint32_t* x32 = (uint32_t*)x;
x32[ 0] = ~x32[ 0];
x32[ 1] ^= 0xffffffff^r;
x32[ 2] = ~x32[ 2];
x32[ 3] ^= 0xefffffff^r;
x32[ 4] = ~x32[ 4];
x32[ 5] ^= 0xdfffffff^r;
x32[ 6] = ~x32[ 6];
x32[ 7] ^= 0xcfffffff^r;
x32[ 8] = ~x32[ 8];
x32[ 9] ^= 0xbfffffff^r;
x32[10] = ~x32[10];
x32[11] ^= 0xafffffff^r;
x32[12] = ~x32[12];
x32[13] ^= 0x9fffffff^r;
x32[14] = ~x32[14];
x32[15] ^= 0x8fffffff^r;
GROESTL_COLUMN(x,y, 0, 2, 6, 10, 14, 1, 5, 9, 13, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 2, 4, 8, 12, 0, 3, 7, 11, 15, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 4, 6, 10, 14, 2, 5, 9, 13, 1, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 6, 8, 12, 0, 4, 7, 11, 15, 3, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y, 8, 10, 14, 2, 6, 9, 13, 1, 5, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y,10, 12, 0, 4, 8, 11, 15, 3, 7, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y,12, 14, 2, 6, 10, 13, 1, 5, 9, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
GROESTL_COLUMN(x,y,14, 0, 4, 8, 12, 15, 3, 7, 11, temp_v1, temp_v2, temp_upper_value, temp_lower_value, temp);
}
__device__
void cn_groestl_F512(uint32_t * __restrict__ h, const uint32_t * __restrict__ m)
{
int i;
uint32_t Ptmp[2*GROESTL_COLS512];
uint32_t Qtmp[2*GROESTL_COLS512];
uint32_t y[2*GROESTL_COLS512];
uint32_t z[2*GROESTL_COLS512];
for (i = 0; i < 2*GROESTL_COLS512; i++) {
z[i] = m[i];
Ptmp[i] = h[i]^m[i];
}
cn_groestl_RND512Q((uint8_t*)z, y, 0x00000000);
cn_groestl_RND512Q((uint8_t*)y, z, 0x01000000);
cn_groestl_RND512Q((uint8_t*)z, y, 0x02000000);
cn_groestl_RND512Q((uint8_t*)y, z, 0x03000000);
cn_groestl_RND512Q((uint8_t*)z, y, 0x04000000);
cn_groestl_RND512Q((uint8_t*)y, z, 0x05000000);
cn_groestl_RND512Q((uint8_t*)z, y, 0x06000000);
cn_groestl_RND512Q((uint8_t*)y, z, 0x07000000);
cn_groestl_RND512Q((uint8_t*)z, y, 0x08000000);
cn_groestl_RND512Q((uint8_t*)y, Qtmp, 0x09000000);
cn_groestl_RND512P((uint8_t*)Ptmp, y, 0x00000000);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000001);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000002);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000003);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000004);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000005);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000006);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000007);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000008);
cn_groestl_RND512P((uint8_t*)y, Ptmp, 0x00000009);
for (i = 0; i < 2*GROESTL_COLS512; i++)
h[i] ^= Ptmp[i]^Qtmp[i];
}
__device__
void cn_groestl_outputtransformation(groestlHashState *ctx)
{
int j;
uint32_t temp[2*GROESTL_COLS512];
uint32_t y[2*GROESTL_COLS512];
uint32_t z[2*GROESTL_COLS512];
for (j = 0; j < 2*GROESTL_COLS512; j++)
temp[j] = ctx->chaining[j];
cn_groestl_RND512P((uint8_t*)temp, y, 0x00000000);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000001);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000002);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000003);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000004);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000005);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000006);
cn_groestl_RND512P((uint8_t*)y, z, 0x00000007);
cn_groestl_RND512P((uint8_t*)z, y, 0x00000008);
cn_groestl_RND512P((uint8_t*)y, temp, 0x00000009);
for (j = 0; j < 2*GROESTL_COLS512; j++)
ctx->chaining[j] ^= temp[j];
}
__device__
void cn_groestl_transform(groestlHashState * __restrict__ ctx,
const uint8_t * __restrict__ input, int msglen)
{
for (; msglen >= GROESTL_SIZE512; msglen -= GROESTL_SIZE512, input += GROESTL_SIZE512) {
cn_groestl_F512(ctx->chaining,(uint32_t*)input);
ctx->block_counter1++;
if (ctx->block_counter1 == 0) ctx->block_counter2++;
}
}
__device__
void cn_groestl_final(groestlHashState* __restrict__ ctx, BitSequence* __restrict__ output)
{
int i, j = 0, hashbytelen = GROESTL_HASH_BIT_LEN/8;
uint8_t *s = (BitSequence*)ctx->chaining;
if (ctx->bits_in_last_byte) {
ctx->buffer[(int)ctx->buf_ptr-1] &= ((1<<ctx->bits_in_last_byte)-1)<<(8-ctx->bits_in_last_byte);
ctx->buffer[(int)ctx->buf_ptr-1] ^= 0x1<<(7-ctx->bits_in_last_byte);
ctx->bits_in_last_byte = 0;
}
else ctx->buffer[(int)ctx->buf_ptr++] = 0x80;
if (ctx->buf_ptr > GROESTL_SIZE512-GROESTL_LENGTHFIELDLEN) {
while (ctx->buf_ptr < GROESTL_SIZE512) {
ctx->buffer[(int)ctx->buf_ptr++] = 0;
}
cn_groestl_transform(ctx, ctx->buffer, GROESTL_SIZE512);
ctx->buf_ptr = 0;
}
while (ctx->buf_ptr < GROESTL_SIZE512-GROESTL_LENGTHFIELDLEN) {
ctx->buffer[(int)ctx->buf_ptr++] = 0;
}
ctx->block_counter1++;
if (ctx->block_counter1 == 0) ctx->block_counter2++;
ctx->buf_ptr = GROESTL_SIZE512;
while (ctx->buf_ptr > GROESTL_SIZE512-(int)sizeof(uint32_t)) {
ctx->buffer[(int)--ctx->buf_ptr] = (uint8_t)ctx->block_counter1;
ctx->block_counter1 >>= 8;
}
while (ctx->buf_ptr > GROESTL_SIZE512-GROESTL_LENGTHFIELDLEN) {
ctx->buffer[(int)--ctx->buf_ptr] = (uint8_t)ctx->block_counter2;
ctx->block_counter2 >>= 8;
}
cn_groestl_transform(ctx, ctx->buffer, GROESTL_SIZE512);
cn_groestl_outputtransformation(ctx);
for (i = GROESTL_SIZE512-hashbytelen; i < GROESTL_SIZE512; i++,j++) {
output[j] = s[i];
}
#if 0
for (i = 0; i < GROESTL_COLS512; i++) {
ctx->chaining[i] = 0;
}
for (i = 0; i < GROESTL_SIZE512; i++) {
ctx->buffer[i] = 0;
}
#endif
}
__device__
void cn_groestl_update(groestlHashState* __restrict__ ctx,
const BitSequence* __restrict__ input, DataLength databitlen)
{
int index = 0;
int msglen = (int)(databitlen/8);
int rem = (int)(databitlen%8);
if (ctx->buf_ptr) {
while (ctx->buf_ptr < GROESTL_SIZE512 && index < msglen) {
ctx->buffer[(int)ctx->buf_ptr++] = input[index++];
}
if (ctx->buf_ptr < GROESTL_SIZE512) {
if (rem) {
ctx->bits_in_last_byte = rem;
ctx->buffer[(int)ctx->buf_ptr++] = input[index];
}
return;
}
ctx->buf_ptr = 0;
cn_groestl_transform(ctx, ctx->buffer, GROESTL_SIZE512);
}
cn_groestl_transform(ctx, input+index, msglen-index);
index += ((msglen-index)/GROESTL_SIZE512)*GROESTL_SIZE512;
while (index < msglen) {
ctx->buffer[(int)ctx->buf_ptr++] = input[index++];
}
if (rem) {
ctx->bits_in_last_byte = rem;
ctx->buffer[(int)ctx->buf_ptr++] = input[index];
}
}
__device__
void cn_groestl_init(groestlHashState* ctx)
{
int i = 0;
for(;i<(GROESTL_SIZE512/sizeof(uint32_t));i++)
ctx->chaining[i] = 0;
ctx->chaining[2*GROESTL_COLS512-1] = u32BIG((uint32_t)GROESTL_HASH_BIT_LEN);
ctx->buf_ptr = 0;
ctx->block_counter1 = 0;
ctx->block_counter2 = 0;
ctx->bits_in_last_byte = 0;
}
__device__
void cn_groestl(const uint8_t * __restrict__ data, DataLength len, uint32_t * hashval)
{
DataLength databitlen = len << 3;
groestlHashState context;
cn_groestl_init(&context);
cn_groestl_update(&context, (BitSequence*) data, databitlen);
cn_groestl_final(&context, (BitSequence*) hashval);
}
+296
View File
@@ -0,0 +1,296 @@
typedef struct {
int hashbitlen;
unsigned long long databitlen;
unsigned long long datasize_in_buffer;
uint64_t x[8][2];
unsigned char buffer[64];
} jhHashState;
__constant__ unsigned char d_JH256_H0[512] = {
0xeb, 0x98, 0xa3, 0x41, 0x2c, 0x20, 0xd3, 0xeb, 0x92, 0xcd, 0xbe, 0x7b, 0x9c, 0xb2, 0x45, 0xc1,
0x1c, 0x93, 0x51, 0x91, 0x60, 0xd4, 0xc7, 0xfa, 0x26, 0x0, 0x82, 0xd6, 0x7e, 0x50, 0x8a, 0x3,
0xa4, 0x23, 0x9e, 0x26, 0x77, 0x26, 0xb9, 0x45, 0xe0, 0xfb, 0x1a, 0x48, 0xd4, 0x1a, 0x94, 0x77,
0xcd, 0xb5, 0xab, 0x26, 0x2, 0x6b, 0x17, 0x7a, 0x56, 0xf0, 0x24, 0x42, 0xf, 0xff, 0x2f, 0xa8,
0x71, 0xa3, 0x96, 0x89, 0x7f, 0x2e, 0x4d, 0x75, 0x1d, 0x14, 0x49, 0x8, 0xf7, 0x7d, 0xe2, 0x62,
0x27, 0x76, 0x95, 0xf7, 0x76, 0x24, 0x8f, 0x94, 0x87, 0xd5, 0xb6, 0x57, 0x47, 0x80, 0x29, 0x6c,
0x5c, 0x5e, 0x27, 0x2d, 0xac, 0x8e, 0xd, 0x6c, 0x51, 0x84, 0x50, 0xc6, 0x57, 0x5, 0x7a, 0xf,
0x7b, 0xe4, 0xd3, 0x67, 0x70, 0x24, 0x12, 0xea, 0x89, 0xe3, 0xab, 0x13, 0xd3, 0x1c, 0xd7, 0x69
};
__constant__ unsigned char d_E8_rc[42][32] = {
{0x72, 0xd5, 0xde, 0xa2, 0xdf, 0x15, 0xf8, 0x67, 0x7b, 0x84, 0x15, 0xa, 0xb7, 0x23, 0x15, 0x57, 0x81, 0xab, 0xd6, 0x90, 0x4d, 0x5a, 0x87, 0xf6, 0x4e, 0x9f, 0x4f, 0xc5, 0xc3, 0xd1, 0x2b, 0x40},
{0xea, 0x98, 0x3a, 0xe0, 0x5c, 0x45, 0xfa, 0x9c, 0x3, 0xc5, 0xd2, 0x99, 0x66, 0xb2, 0x99, 0x9a, 0x66, 0x2, 0x96, 0xb4, 0xf2, 0xbb, 0x53, 0x8a, 0xb5, 0x56, 0x14, 0x1a, 0x88, 0xdb, 0xa2, 0x31},
{0x3, 0xa3, 0x5a, 0x5c, 0x9a, 0x19, 0xe, 0xdb, 0x40, 0x3f, 0xb2, 0xa, 0x87, 0xc1, 0x44, 0x10, 0x1c, 0x5, 0x19, 0x80, 0x84, 0x9e, 0x95, 0x1d, 0x6f, 0x33, 0xeb, 0xad, 0x5e, 0xe7, 0xcd, 0xdc},
{0x10, 0xba, 0x13, 0x92, 0x2, 0xbf, 0x6b, 0x41, 0xdc, 0x78, 0x65, 0x15, 0xf7, 0xbb, 0x27, 0xd0, 0xa, 0x2c, 0x81, 0x39, 0x37, 0xaa, 0x78, 0x50, 0x3f, 0x1a, 0xbf, 0xd2, 0x41, 0x0, 0x91, 0xd3},
{0x42, 0x2d, 0x5a, 0xd, 0xf6, 0xcc, 0x7e, 0x90, 0xdd, 0x62, 0x9f, 0x9c, 0x92, 0xc0, 0x97, 0xce, 0x18, 0x5c, 0xa7, 0xb, 0xc7, 0x2b, 0x44, 0xac, 0xd1, 0xdf, 0x65, 0xd6, 0x63, 0xc6, 0xfc, 0x23},
{0x97, 0x6e, 0x6c, 0x3, 0x9e, 0xe0, 0xb8, 0x1a, 0x21, 0x5, 0x45, 0x7e, 0x44, 0x6c, 0xec, 0xa8, 0xee, 0xf1, 0x3, 0xbb, 0x5d, 0x8e, 0x61, 0xfa, 0xfd, 0x96, 0x97, 0xb2, 0x94, 0x83, 0x81, 0x97},
{0x4a, 0x8e, 0x85, 0x37, 0xdb, 0x3, 0x30, 0x2f, 0x2a, 0x67, 0x8d, 0x2d, 0xfb, 0x9f, 0x6a, 0x95, 0x8a, 0xfe, 0x73, 0x81, 0xf8, 0xb8, 0x69, 0x6c, 0x8a, 0xc7, 0x72, 0x46, 0xc0, 0x7f, 0x42, 0x14},
{0xc5, 0xf4, 0x15, 0x8f, 0xbd, 0xc7, 0x5e, 0xc4, 0x75, 0x44, 0x6f, 0xa7, 0x8f, 0x11, 0xbb, 0x80, 0x52, 0xde, 0x75, 0xb7, 0xae, 0xe4, 0x88, 0xbc, 0x82, 0xb8, 0x0, 0x1e, 0x98, 0xa6, 0xa3, 0xf4},
{0x8e, 0xf4, 0x8f, 0x33, 0xa9, 0xa3, 0x63, 0x15, 0xaa, 0x5f, 0x56, 0x24, 0xd5, 0xb7, 0xf9, 0x89, 0xb6, 0xf1, 0xed, 0x20, 0x7c, 0x5a, 0xe0, 0xfd, 0x36, 0xca, 0xe9, 0x5a, 0x6, 0x42, 0x2c, 0x36},
{0xce, 0x29, 0x35, 0x43, 0x4e, 0xfe, 0x98, 0x3d, 0x53, 0x3a, 0xf9, 0x74, 0x73, 0x9a, 0x4b, 0xa7, 0xd0, 0xf5, 0x1f, 0x59, 0x6f, 0x4e, 0x81, 0x86, 0xe, 0x9d, 0xad, 0x81, 0xaf, 0xd8, 0x5a, 0x9f},
{0xa7, 0x5, 0x6, 0x67, 0xee, 0x34, 0x62, 0x6a, 0x8b, 0xb, 0x28, 0xbe, 0x6e, 0xb9, 0x17, 0x27, 0x47, 0x74, 0x7, 0x26, 0xc6, 0x80, 0x10, 0x3f, 0xe0, 0xa0, 0x7e, 0x6f, 0xc6, 0x7e, 0x48, 0x7b},
{0xd, 0x55, 0xa, 0xa5, 0x4a, 0xf8, 0xa4, 0xc0, 0x91, 0xe3, 0xe7, 0x9f, 0x97, 0x8e, 0xf1, 0x9e, 0x86, 0x76, 0x72, 0x81, 0x50, 0x60, 0x8d, 0xd4, 0x7e, 0x9e, 0x5a, 0x41, 0xf3, 0xe5, 0xb0, 0x62},
{0xfc, 0x9f, 0x1f, 0xec, 0x40, 0x54, 0x20, 0x7a, 0xe3, 0xe4, 0x1a, 0x0, 0xce, 0xf4, 0xc9, 0x84, 0x4f, 0xd7, 0x94, 0xf5, 0x9d, 0xfa, 0x95, 0xd8, 0x55, 0x2e, 0x7e, 0x11, 0x24, 0xc3, 0x54, 0xa5},
{0x5b, 0xdf, 0x72, 0x28, 0xbd, 0xfe, 0x6e, 0x28, 0x78, 0xf5, 0x7f, 0xe2, 0xf, 0xa5, 0xc4, 0xb2, 0x5, 0x89, 0x7c, 0xef, 0xee, 0x49, 0xd3, 0x2e, 0x44, 0x7e, 0x93, 0x85, 0xeb, 0x28, 0x59, 0x7f},
{0x70, 0x5f, 0x69, 0x37, 0xb3, 0x24, 0x31, 0x4a, 0x5e, 0x86, 0x28, 0xf1, 0x1d, 0xd6, 0xe4, 0x65, 0xc7, 0x1b, 0x77, 0x4, 0x51, 0xb9, 0x20, 0xe7, 0x74, 0xfe, 0x43, 0xe8, 0x23, 0xd4, 0x87, 0x8a},
{0x7d, 0x29, 0xe8, 0xa3, 0x92, 0x76, 0x94, 0xf2, 0xdd, 0xcb, 0x7a, 0x9, 0x9b, 0x30, 0xd9, 0xc1, 0x1d, 0x1b, 0x30, 0xfb, 0x5b, 0xdc, 0x1b, 0xe0, 0xda, 0x24, 0x49, 0x4f, 0xf2, 0x9c, 0x82, 0xbf},
{0xa4, 0xe7, 0xba, 0x31, 0xb4, 0x70, 0xbf, 0xff, 0xd, 0x32, 0x44, 0x5, 0xde, 0xf8, 0xbc, 0x48, 0x3b, 0xae, 0xfc, 0x32, 0x53, 0xbb, 0xd3, 0x39, 0x45, 0x9f, 0xc3, 0xc1, 0xe0, 0x29, 0x8b, 0xa0},
{0xe5, 0xc9, 0x5, 0xfd, 0xf7, 0xae, 0x9, 0xf, 0x94, 0x70, 0x34, 0x12, 0x42, 0x90, 0xf1, 0x34, 0xa2, 0x71, 0xb7, 0x1, 0xe3, 0x44, 0xed, 0x95, 0xe9, 0x3b, 0x8e, 0x36, 0x4f, 0x2f, 0x98, 0x4a},
{0x88, 0x40, 0x1d, 0x63, 0xa0, 0x6c, 0xf6, 0x15, 0x47, 0xc1, 0x44, 0x4b, 0x87, 0x52, 0xaf, 0xff, 0x7e, 0xbb, 0x4a, 0xf1, 0xe2, 0xa, 0xc6, 0x30, 0x46, 0x70, 0xb6, 0xc5, 0xcc, 0x6e, 0x8c, 0xe6},
{0xa4, 0xd5, 0xa4, 0x56, 0xbd, 0x4f, 0xca, 0x0, 0xda, 0x9d, 0x84, 0x4b, 0xc8, 0x3e, 0x18, 0xae, 0x73, 0x57, 0xce, 0x45, 0x30, 0x64, 0xd1, 0xad, 0xe8, 0xa6, 0xce, 0x68, 0x14, 0x5c, 0x25, 0x67},
{0xa3, 0xda, 0x8c, 0xf2, 0xcb, 0xe, 0xe1, 0x16, 0x33, 0xe9, 0x6, 0x58, 0x9a, 0x94, 0x99, 0x9a, 0x1f, 0x60, 0xb2, 0x20, 0xc2, 0x6f, 0x84, 0x7b, 0xd1, 0xce, 0xac, 0x7f, 0xa0, 0xd1, 0x85, 0x18},
{0x32, 0x59, 0x5b, 0xa1, 0x8d, 0xdd, 0x19, 0xd3, 0x50, 0x9a, 0x1c, 0xc0, 0xaa, 0xa5, 0xb4, 0x46, 0x9f, 0x3d, 0x63, 0x67, 0xe4, 0x4, 0x6b, 0xba, 0xf6, 0xca, 0x19, 0xab, 0xb, 0x56, 0xee, 0x7e},
{0x1f, 0xb1, 0x79, 0xea, 0xa9, 0x28, 0x21, 0x74, 0xe9, 0xbd, 0xf7, 0x35, 0x3b, 0x36, 0x51, 0xee, 0x1d, 0x57, 0xac, 0x5a, 0x75, 0x50, 0xd3, 0x76, 0x3a, 0x46, 0xc2, 0xfe, 0xa3, 0x7d, 0x70, 0x1},
{0xf7, 0x35, 0xc1, 0xaf, 0x98, 0xa4, 0xd8, 0x42, 0x78, 0xed, 0xec, 0x20, 0x9e, 0x6b, 0x67, 0x79, 0x41, 0x83, 0x63, 0x15, 0xea, 0x3a, 0xdb, 0xa8, 0xfa, 0xc3, 0x3b, 0x4d, 0x32, 0x83, 0x2c, 0x83},
{0xa7, 0x40, 0x3b, 0x1f, 0x1c, 0x27, 0x47, 0xf3, 0x59, 0x40, 0xf0, 0x34, 0xb7, 0x2d, 0x76, 0x9a, 0xe7, 0x3e, 0x4e, 0x6c, 0xd2, 0x21, 0x4f, 0xfd, 0xb8, 0xfd, 0x8d, 0x39, 0xdc, 0x57, 0x59, 0xef},
{0x8d, 0x9b, 0xc, 0x49, 0x2b, 0x49, 0xeb, 0xda, 0x5b, 0xa2, 0xd7, 0x49, 0x68, 0xf3, 0x70, 0xd, 0x7d, 0x3b, 0xae, 0xd0, 0x7a, 0x8d, 0x55, 0x84, 0xf5, 0xa5, 0xe9, 0xf0, 0xe4, 0xf8, 0x8e, 0x65},
{0xa0, 0xb8, 0xa2, 0xf4, 0x36, 0x10, 0x3b, 0x53, 0xc, 0xa8, 0x7, 0x9e, 0x75, 0x3e, 0xec, 0x5a, 0x91, 0x68, 0x94, 0x92, 0x56, 0xe8, 0x88, 0x4f, 0x5b, 0xb0, 0x5c, 0x55, 0xf8, 0xba, 0xbc, 0x4c},
{0xe3, 0xbb, 0x3b, 0x99, 0xf3, 0x87, 0x94, 0x7b, 0x75, 0xda, 0xf4, 0xd6, 0x72, 0x6b, 0x1c, 0x5d, 0x64, 0xae, 0xac, 0x28, 0xdc, 0x34, 0xb3, 0x6d, 0x6c, 0x34, 0xa5, 0x50, 0xb8, 0x28, 0xdb, 0x71},
{0xf8, 0x61, 0xe2, 0xf2, 0x10, 0x8d, 0x51, 0x2a, 0xe3, 0xdb, 0x64, 0x33, 0x59, 0xdd, 0x75, 0xfc, 0x1c, 0xac, 0xbc, 0xf1, 0x43, 0xce, 0x3f, 0xa2, 0x67, 0xbb, 0xd1, 0x3c, 0x2, 0xe8, 0x43, 0xb0},
{0x33, 0xa, 0x5b, 0xca, 0x88, 0x29, 0xa1, 0x75, 0x7f, 0x34, 0x19, 0x4d, 0xb4, 0x16, 0x53, 0x5c, 0x92, 0x3b, 0x94, 0xc3, 0xe, 0x79, 0x4d, 0x1e, 0x79, 0x74, 0x75, 0xd7, 0xb6, 0xee, 0xaf, 0x3f},
{0xea, 0xa8, 0xd4, 0xf7, 0xbe, 0x1a, 0x39, 0x21, 0x5c, 0xf4, 0x7e, 0x9, 0x4c, 0x23, 0x27, 0x51, 0x26, 0xa3, 0x24, 0x53, 0xba, 0x32, 0x3c, 0xd2, 0x44, 0xa3, 0x17, 0x4a, 0x6d, 0xa6, 0xd5, 0xad},
{0xb5, 0x1d, 0x3e, 0xa6, 0xaf, 0xf2, 0xc9, 0x8, 0x83, 0x59, 0x3d, 0x98, 0x91, 0x6b, 0x3c, 0x56, 0x4c, 0xf8, 0x7c, 0xa1, 0x72, 0x86, 0x60, 0x4d, 0x46, 0xe2, 0x3e, 0xcc, 0x8, 0x6e, 0xc7, 0xf6},
{0x2f, 0x98, 0x33, 0xb3, 0xb1, 0xbc, 0x76, 0x5e, 0x2b, 0xd6, 0x66, 0xa5, 0xef, 0xc4, 0xe6, 0x2a, 0x6, 0xf4, 0xb6, 0xe8, 0xbe, 0xc1, 0xd4, 0x36, 0x74, 0xee, 0x82, 0x15, 0xbc, 0xef, 0x21, 0x63},
{0xfd, 0xc1, 0x4e, 0xd, 0xf4, 0x53, 0xc9, 0x69, 0xa7, 0x7d, 0x5a, 0xc4, 0x6, 0x58, 0x58, 0x26, 0x7e, 0xc1, 0x14, 0x16, 0x6, 0xe0, 0xfa, 0x16, 0x7e, 0x90, 0xaf, 0x3d, 0x28, 0x63, 0x9d, 0x3f},
{0xd2, 0xc9, 0xf2, 0xe3, 0x0, 0x9b, 0xd2, 0xc, 0x5f, 0xaa, 0xce, 0x30, 0xb7, 0xd4, 0xc, 0x30, 0x74, 0x2a, 0x51, 0x16, 0xf2, 0xe0, 0x32, 0x98, 0xd, 0xeb, 0x30, 0xd8, 0xe3, 0xce, 0xf8, 0x9a},
{0x4b, 0xc5, 0x9e, 0x7b, 0xb5, 0xf1, 0x79, 0x92, 0xff, 0x51, 0xe6, 0x6e, 0x4, 0x86, 0x68, 0xd3, 0x9b, 0x23, 0x4d, 0x57, 0xe6, 0x96, 0x67, 0x31, 0xcc, 0xe6, 0xa6, 0xf3, 0x17, 0xa, 0x75, 0x5},
{0xb1, 0x76, 0x81, 0xd9, 0x13, 0x32, 0x6c, 0xce, 0x3c, 0x17, 0x52, 0x84, 0xf8, 0x5, 0xa2, 0x62, 0xf4, 0x2b, 0xcb, 0xb3, 0x78, 0x47, 0x15, 0x47, 0xff, 0x46, 0x54, 0x82, 0x23, 0x93, 0x6a, 0x48},
{0x38, 0xdf, 0x58, 0x7, 0x4e, 0x5e, 0x65, 0x65, 0xf2, 0xfc, 0x7c, 0x89, 0xfc, 0x86, 0x50, 0x8e, 0x31, 0x70, 0x2e, 0x44, 0xd0, 0xb, 0xca, 0x86, 0xf0, 0x40, 0x9, 0xa2, 0x30, 0x78, 0x47, 0x4e},
{0x65, 0xa0, 0xee, 0x39, 0xd1, 0xf7, 0x38, 0x83, 0xf7, 0x5e, 0xe9, 0x37, 0xe4, 0x2c, 0x3a, 0xbd, 0x21, 0x97, 0xb2, 0x26, 0x1, 0x13, 0xf8, 0x6f, 0xa3, 0x44, 0xed, 0xd1, 0xef, 0x9f, 0xde, 0xe7},
{0x8b, 0xa0, 0xdf, 0x15, 0x76, 0x25, 0x92, 0xd9, 0x3c, 0x85, 0xf7, 0xf6, 0x12, 0xdc, 0x42, 0xbe, 0xd8, 0xa7, 0xec, 0x7c, 0xab, 0x27, 0xb0, 0x7e, 0x53, 0x8d, 0x7d, 0xda, 0xaa, 0x3e, 0xa8, 0xde},
{0xaa, 0x25, 0xce, 0x93, 0xbd, 0x2, 0x69, 0xd8, 0x5a, 0xf6, 0x43, 0xfd, 0x1a, 0x73, 0x8, 0xf9, 0xc0, 0x5f, 0xef, 0xda, 0x17, 0x4a, 0x19, 0xa5, 0x97, 0x4d, 0x66, 0x33, 0x4c, 0xfd, 0x21, 0x6a},
{0x35, 0xb4, 0x98, 0x31, 0xdb, 0x41, 0x15, 0x70, 0xea, 0x1e, 0xf, 0xbb, 0xed, 0xcd, 0x54, 0x9b, 0x9a, 0xd0, 0x63, 0xa1, 0x51, 0x97, 0x40, 0x72, 0xf6, 0x75, 0x9d, 0xbf, 0x91, 0x47, 0x6f, 0xe2}
};
#define JH_SWAP1(x) (x) = ((((x) & 0x5555555555555555ULL) << 1) | (((x) & 0xaaaaaaaaaaaaaaaaULL) >> 1));
#define JH_SWAP2(x) (x) = ((((x) & 0x3333333333333333ULL) << 2) | (((x) & 0xccccccccccccccccULL) >> 2));
#define JH_SWAP4(x) (x) = ((((x) & 0x0f0f0f0f0f0f0f0fULL) << 4) | (((x) & 0xf0f0f0f0f0f0f0f0ULL) >> 4));
#define JH_SWAP8(x) (x) = ((((x) & 0x00ff00ff00ff00ffULL) << 8) | (((x) & 0xff00ff00ff00ff00ULL) >> 8));
#define JH_SWAP16(x) (x) = ((((x) & 0x0000ffff0000ffffULL) << 16) | (((x) & 0xffff0000ffff0000ULL) >> 16));
#define JH_SWAP32(x) (x) = (((x) << 32) | ((x) >> 32));
#define JH_L(m0,m1,m2,m3,m4,m5,m6,m7) \
(m4) ^= (m1); \
(m5) ^= (m2); \
(m6) ^= (m0) ^ (m3); \
(m7) ^= (m0); \
(m0) ^= (m5); \
(m1) ^= (m6); \
(m2) ^= (m4) ^ (m7); \
(m3) ^= (m4);
#define JH_SS(m0,m1,m2,m3,m4,m5,m6,m7,cc0,cc1) \
m3 = ~(m3); \
m7 = ~(m7); \
m0 ^= ((~(m2)) & (cc0)); \
m4 ^= ((~(m6)) & (cc1)); \
temp0 = (cc0) ^ ((m0) & (m1));\
temp1 = (cc1) ^ ((m4) & (m5));\
m0 ^= ((m2) & (m3)); \
m4 ^= ((m6) & (m7)); \
m3 ^= ((~(m1)) & (m2)); \
m7 ^= ((~(m5)) & (m6)); \
m1 ^= ((m0) & (m2)); \
m5 ^= ((m4) & (m6)); \
m2 ^= ((m0) & (~(m3))); \
m6 ^= ((m4) & (~(m7))); \
m0 ^= ((m1) | (m3)); \
m4 ^= ((m5) | (m7)); \
m3 ^= ((m1) & (m2)); \
m7 ^= ((m5) & (m6)); \
m1 ^= (temp0 & (m0)); \
m5 ^= (temp1 & (m4)); \
m2 ^= temp0; \
m6 ^= temp1;
__device__
void cn_jh_E8(jhHashState *state)
{
uint64_t i,roundnumber,temp0,temp1;
for (roundnumber = 0; roundnumber < 42; roundnumber = roundnumber+7)
{
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+0])[i],((uint64_t *)d_E8_rc[roundnumber+0])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
JH_SWAP1(state->x[1][i]); JH_SWAP1(state->x[3][i]); JH_SWAP1(state->x[5][i]); JH_SWAP1(state->x[7][i]);
}
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+1])[i],((uint64_t *)d_E8_rc[roundnumber+1])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
JH_SWAP2(state->x[1][i]); JH_SWAP2(state->x[3][i]); JH_SWAP2(state->x[5][i]); JH_SWAP2(state->x[7][i]);
}
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+2])[i],((uint64_t *)d_E8_rc[roundnumber+2])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
JH_SWAP4(state->x[1][i]); JH_SWAP4(state->x[3][i]); JH_SWAP4(state->x[5][i]); JH_SWAP4(state->x[7][i]);
}
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+3])[i],((uint64_t *)d_E8_rc[roundnumber+3])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
JH_SWAP8(state->x[1][i]); JH_SWAP8(state->x[3][i]); JH_SWAP8(state->x[5][i]); JH_SWAP8(state->x[7][i]);
}
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+4])[i],((uint64_t *)d_E8_rc[roundnumber+4])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
JH_SWAP16(state->x[1][i]); JH_SWAP16(state->x[3][i]); JH_SWAP16(state->x[5][i]); JH_SWAP16(state->x[7][i]);
}
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+5])[i],((uint64_t *)d_E8_rc[roundnumber+5])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
JH_SWAP32(state->x[1][i]); JH_SWAP32(state->x[3][i]); JH_SWAP32(state->x[5][i]); JH_SWAP32(state->x[7][i]);
}
for (i = 0; i < 2; i++) {
JH_SS(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i],
((uint64_t *)d_E8_rc[roundnumber+6])[i],((uint64_t *)d_E8_rc[roundnumber+6])[i+2] );
JH_L(state->x[0][i],state->x[2][i],state->x[4][i],state->x[6][i],state->x[1][i],state->x[3][i],state->x[5][i],state->x[7][i]);
}
for (i = 1; i < 8; i = i+2) {
temp0 = state->x[i][0]; state->x[i][0] = state->x[i][1]; state->x[i][1] = temp0;
}
}
}
__device__
void cn_jh_F8(jhHashState *state)
{
uint64_t i;
for (i = 0; i < 8; i++) {
state->x[i >> 1][i & 1] ^= ((uint64_t *)state->buffer)[i];
}
cn_jh_E8(state);
for (i = 0; i < 8; i++) {
state->x[(8+i) >> 1][(8+i) & 1] ^= ((uint64_t *)state->buffer)[i];
}
}
__device__
void cn_jh_update(jhHashState * __restrict__ state, const uint8_t * __restrict__ data, DataLength databitlen)
{
DataLength index;
state->databitlen += databitlen;
index = 0;
if ( (state->datasize_in_buffer > 0 ) && (( state->datasize_in_buffer + databitlen) < 512) )
{
if ( (databitlen & 7) == 0 ) {
memcpy(state->buffer + (state->datasize_in_buffer >> 3), data, 64-(state->datasize_in_buffer >> 3));
}
else memcpy(state->buffer + (state->datasize_in_buffer >> 3), data, 64-(state->datasize_in_buffer >> 3)+1);
state->datasize_in_buffer += databitlen;
databitlen = 0;
}
if ( (state->datasize_in_buffer > 0 ) && (( state->datasize_in_buffer + databitlen) >= 512) )
{
memcpy( state->buffer + (state->datasize_in_buffer >> 3), data, 64-(state->datasize_in_buffer >> 3) );
index = 64-(state->datasize_in_buffer >> 3);
databitlen = databitlen - (512 - state->datasize_in_buffer);
cn_jh_F8(state);
state->datasize_in_buffer = 0;
}
for ( ; databitlen >= 512; index = index+64, databitlen = databitlen - 512) {
memcpy(state->buffer, data+index, 64);
cn_jh_F8(state);
}
if ( databitlen > 0) {
if ((databitlen & 7) == 0)
memcpy(state->buffer, data+index, (databitlen & 0x1ff) >> 3);
else
memcpy(state->buffer, data+index, ((databitlen & 0x1ff) >> 3)+1);
state->datasize_in_buffer = databitlen;
}
}
/* pad the message, process the padded block(s), truncate the hash value H to obtain the message digest */
__device__
void cn_jh_final(jhHashState * __restrict__ state, uint8_t * __restrict__ hashval)
{
unsigned int i;
//uint32_t *bufptr = (uint32_t *)state->buffer;
if ( (state->databitlen & 0x1ff) == 0 ) {
/* pad the message when databitlen is multiple of 512 bits, then process the padded block */
memset(state->buffer, 0, 64);
//for( i = 0; i < 16; i++ ) *(bufptr+i) = 0x00000000;
state->buffer[0] = 0x80;
state->buffer[63] = state->databitlen & 0xff;
state->buffer[62] = (state->databitlen >> 8) & 0xff;
state->buffer[61] = (state->databitlen >> 16) & 0xff;
state->buffer[60] = (state->databitlen >> 24) & 0xff;
state->buffer[59] = (state->databitlen >> 32) & 0xff;
state->buffer[58] = (state->databitlen >> 40) & 0xff;
state->buffer[57] = (state->databitlen >> 48) & 0xff;
state->buffer[56] = (state->databitlen >> 56) & 0xff;
cn_jh_F8(state);
} else {
/* set the rest of the bytes in the buffer to 0 */
if ( (state->datasize_in_buffer & 7) == 0) {
for (i = (state->databitlen & 0x1ff) >> 3; i < 64; i++) state->buffer[i] = 0;
} else {
for (i = ((state->databitlen & 0x1ff) >> 3)+1; i < 64; i++) state->buffer[i] = 0;
}
/*pad and process the partial block when databitlen is not multiple of 512 bits, then hash the padded blocks*/
state->buffer[((state->databitlen & 0x1ff) >> 3)] |= 1 << (7- (state->databitlen & 7));
cn_jh_F8(state);
memset(state->buffer, 0, 64);
//for( i = 0; i < 16; i++ ) *(bufptr+i) = 0x00000000;
state->buffer[63] = state->databitlen & 0xff;
state->buffer[62] = (state->databitlen >> 8) & 0xff;
state->buffer[61] = (state->databitlen >> 16) & 0xff;
state->buffer[60] = (state->databitlen >> 24) & 0xff;
state->buffer[59] = (state->databitlen >> 32) & 0xff;
state->buffer[58] = (state->databitlen >> 40) & 0xff;
state->buffer[57] = (state->databitlen >> 48) & 0xff;
state->buffer[56] = (state->databitlen >> 56) & 0xff;
cn_jh_F8(state);
}
memcpy(hashval, ((unsigned char*)state->x) + 64 + 32, 32);
//MEMCPY4(hashval, ((unsigned char*)state->x) + 64 + 32, 8);
}
__device__
void cn_jh_init(jhHashState *state, int hashbitlen)
{
state->databitlen = 0;
state->datasize_in_buffer = 0;
state->hashbitlen = hashbitlen;
memcpy(state->x, d_JH256_H0, 128);
//MEMCPY8(state->x, d_JH256_H0, 128 / 8);
}
__device__
void cn_jh(const uint8_t * __restrict__ data, DataLength len, uint32_t * hashval)
{
const int hashbitlen = 256;
DataLength databitlen = len << 3;
jhHashState state;
cn_jh_init(&state, hashbitlen);
cn_jh_update(&state, data, databitlen);
cn_jh_final(&state, (uint8_t*) hashval);
}
+211
View File
@@ -0,0 +1,211 @@
__constant__ uint64_t keccakf_rndc[24] = {
0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
0x8000000000008003, 0x8000000000008002, 0x8000000000000080,
0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
0x8000000000008080, 0x0000000080000001, 0x8000000080008008
};
#if __CUDA_ARCH__ >= 350
__forceinline__ __device__ uint64_t cuda_rotl64(const uint64_t value, const int offset)
{
uint2 result;
if(offset >= 32)
{
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
}
else
{
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
}
return __double_as_longlong(__hiloint2double(result.y, result.x));
}
#define rotl64_1(x, y) (cuda_rotl64((x), (y)))
#else
#define rotl64_1(x, y) ((x) << (y) | ((x) >> (64 - (y))))
#endif
#define rotl64_2(x, y) rotl64_1(((x) >> 32) | ((x) << 32), (y))
#define bitselect(a, b, c) ((a) ^ ((c) & ((b) ^ (a))))
__device__ __forceinline__
void cn_keccakf2(uint64_t *s)
{
uint8_t i;
for(i = 0; i < 24; ++i)
{
uint64_t bc[5], tmpxor[5], tmp1, tmp2;
tmpxor[0] = s[0] ^ s[5] ^ s[10] ^ s[15] ^ s[20];
tmpxor[1] = s[1] ^ s[6] ^ s[11] ^ s[16] ^ s[21];
tmpxor[2] = s[2] ^ s[7] ^ s[12] ^ s[17] ^ s[22];
tmpxor[3] = s[3] ^ s[8] ^ s[13] ^ s[18] ^ s[23];
tmpxor[4] = s[4] ^ s[9] ^ s[14] ^ s[19] ^ s[24];
bc[0] = tmpxor[0] ^ rotl64_1(tmpxor[2], 1);
bc[1] = tmpxor[1] ^ rotl64_1(tmpxor[3], 1);
bc[2] = tmpxor[2] ^ rotl64_1(tmpxor[4], 1);
bc[3] = tmpxor[3] ^ rotl64_1(tmpxor[0], 1);
bc[4] = tmpxor[4] ^ rotl64_1(tmpxor[1], 1);
tmp1 = s[1] ^ bc[0];
s[0] ^= bc[4];
s[1] = rotl64_2(s[6] ^ bc[0], 12);
s[6] = rotl64_1(s[9] ^ bc[3], 20);
s[9] = rotl64_2(s[22] ^ bc[1], 29);
s[22] = rotl64_2(s[14] ^ bc[3], 7);
s[14] = rotl64_1(s[20] ^ bc[4], 18);
s[20] = rotl64_2(s[2] ^ bc[1], 30);
s[2] = rotl64_2(s[12] ^ bc[1], 11);
s[12] = rotl64_1(s[13] ^ bc[2], 25);
s[13] = rotl64_1(s[19] ^ bc[3], 8);
s[19] = rotl64_2(s[23] ^ bc[2], 24);
s[23] = rotl64_2(s[15] ^ bc[4], 9);
s[15] = rotl64_1(s[4] ^ bc[3], 27);
s[4] = rotl64_1(s[24] ^ bc[3], 14);
s[24] = rotl64_1(s[21] ^ bc[0], 2);
s[21] = rotl64_2(s[8] ^ bc[2], 23);
s[8] = rotl64_2(s[16] ^ bc[0], 13);
s[16] = rotl64_2(s[5] ^ bc[4], 4);
s[5] = rotl64_1(s[3] ^ bc[2], 28);
s[3] = rotl64_1(s[18] ^ bc[2], 21);
s[18] = rotl64_1(s[17] ^ bc[1], 15);
s[17] = rotl64_1(s[11] ^ bc[0], 10);
s[11] = rotl64_1(s[7] ^ bc[1], 6);
s[7] = rotl64_1(s[10] ^ bc[4], 3);
s[10] = rotl64_1(tmp1, 1);
tmp1 = s[0]; tmp2 = s[1]; s[0] = bitselect(s[0] ^ s[2], s[0], s[1]); s[1] = bitselect(s[1] ^ s[3], s[1], s[2]); s[2] = bitselect(s[2] ^ s[4], s[2], s[3]); s[3] = bitselect(s[3] ^ tmp1, s[3], s[4]); s[4] = bitselect(s[4] ^ tmp2, s[4], tmp1);
tmp1 = s[5]; tmp2 = s[6]; s[5] = bitselect(s[5] ^ s[7], s[5], s[6]); s[6] = bitselect(s[6] ^ s[8], s[6], s[7]); s[7] = bitselect(s[7] ^ s[9], s[7], s[8]); s[8] = bitselect(s[8] ^ tmp1, s[8], s[9]); s[9] = bitselect(s[9] ^ tmp2, s[9], tmp1);
tmp1 = s[10]; tmp2 = s[11]; s[10] = bitselect(s[10] ^ s[12], s[10], s[11]); s[11] = bitselect(s[11] ^ s[13], s[11], s[12]); s[12] = bitselect(s[12] ^ s[14], s[12], s[13]); s[13] = bitselect(s[13] ^ tmp1, s[13], s[14]); s[14] = bitselect(s[14] ^ tmp2, s[14], tmp1);
tmp1 = s[15]; tmp2 = s[16]; s[15] = bitselect(s[15] ^ s[17], s[15], s[16]); s[16] = bitselect(s[16] ^ s[18], s[16], s[17]); s[17] = bitselect(s[17] ^ s[19], s[17], s[18]); s[18] = bitselect(s[18] ^ tmp1, s[18], s[19]); s[19] = bitselect(s[19] ^ tmp2, s[19], tmp1);
tmp1 = s[20]; tmp2 = s[21]; s[20] = bitselect(s[20] ^ s[22], s[20], s[21]); s[21] = bitselect(s[21] ^ s[23], s[21], s[22]); s[22] = bitselect(s[22] ^ s[24], s[22], s[23]); s[23] = bitselect(s[23] ^ tmp1, s[23], s[24]); s[24] = bitselect(s[24] ^ tmp2, s[24], tmp1);
s[0] ^= keccakf_rndc[i];
}
}
__device__ __forceinline__
void cn_keccakf(uint64_t *s)
{
uint64_t bc[5], tmpxor[5], tmp1, tmp2;
tmpxor[0] = s[0] ^ s[5];
tmpxor[1] = s[1] ^ s[6] ^ 0x8000000000000000ULL;
tmpxor[2] = s[2] ^ s[7];
tmpxor[3] = s[3] ^ s[8];
tmpxor[4] = s[4] ^ s[9];
bc[0] = tmpxor[0] ^ rotl64_1(tmpxor[2], 1);
bc[1] = tmpxor[1] ^ rotl64_1(tmpxor[3], 1);
bc[2] = tmpxor[2] ^ rotl64_1(tmpxor[4], 1);
bc[3] = tmpxor[3] ^ rotl64_1(tmpxor[0], 1);
bc[4] = tmpxor[4] ^ rotl64_1(tmpxor[1], 1);
tmp1 = s[1] ^ bc[0];
s[0] ^= bc[4];
s[1] = rotl64_2(s[6] ^ bc[0], 12);
s[6] = rotl64_1(s[9] ^ bc[3], 20);
s[9] = rotl64_2(bc[1], 29);
s[22] = rotl64_2(bc[3], 7);
s[14] = rotl64_1(bc[4], 18);
s[20] = rotl64_2(s[2] ^ bc[1], 30);
s[2] = rotl64_2(bc[1], 11);
s[12] = rotl64_1(bc[2], 25);
s[13] = rotl64_1(bc[3], 8);
s[19] = rotl64_2(bc[2], 24);
s[23] = rotl64_2(bc[4], 9);
s[15] = rotl64_1(s[4] ^ bc[3], 27);
s[4] = rotl64_1(bc[3], 14);
s[24] = rotl64_1(bc[0], 2);
s[21] = rotl64_2(s[8] ^ bc[2], 23);
s[8] = rotl64_2(0x8000000000000000ULL ^ bc[0], 13);
s[16] = rotl64_2(s[5] ^ bc[4], 4);
s[5] = rotl64_1(s[3] ^ bc[2], 28);
s[3] = rotl64_1(bc[2], 21);
s[18] = rotl64_1(bc[1], 15);
s[17] = rotl64_1(bc[0], 10);
s[11] = rotl64_1(s[7] ^ bc[1], 6);
s[7] = rotl64_1(bc[4], 3);
s[10] = rotl64_1(tmp1, 1);
tmp1 = s[0]; tmp2 = s[1]; s[0] = bitselect(s[0] ^ s[2], s[0], s[1]); s[1] = bitselect(s[1] ^ s[3], s[1], s[2]); s[2] = bitselect(s[2] ^ s[4], s[2], s[3]); s[3] = bitselect(s[3] ^ tmp1, s[3], s[4]); s[4] = bitselect(s[4] ^ tmp2, s[4], tmp1);
tmp1 = s[5]; tmp2 = s[6]; s[5] = bitselect(s[5] ^ s[7], s[5], s[6]); s[6] = bitselect(s[6] ^ s[8], s[6], s[7]); s[7] = bitselect(s[7] ^ s[9], s[7], s[8]); s[8] = bitselect(s[8] ^ tmp1, s[8], s[9]); s[9] = bitselect(s[9] ^ tmp2, s[9], tmp1);
tmp1 = s[10]; tmp2 = s[11]; s[10] = bitselect(s[10] ^ s[12], s[10], s[11]); s[11] = bitselect(s[11] ^ s[13], s[11], s[12]); s[12] = bitselect(s[12] ^ s[14], s[12], s[13]); s[13] = bitselect(s[13] ^ tmp1, s[13], s[14]); s[14] = bitselect(s[14] ^ tmp2, s[14], tmp1);
tmp1 = s[15]; tmp2 = s[16]; s[15] = bitselect(s[15] ^ s[17], s[15], s[16]); s[16] = bitselect(s[16] ^ s[18], s[16], s[17]); s[17] = bitselect(s[17] ^ s[19], s[17], s[18]); s[18] = bitselect(s[18] ^ tmp1, s[18], s[19]); s[19] = bitselect(s[19] ^ tmp2, s[19], tmp1);
tmp1 = s[20]; tmp2 = s[21]; s[20] = bitselect(s[20] ^ s[22], s[20], s[21]); s[21] = bitselect(s[21] ^ s[23], s[21], s[22]); s[22] = bitselect(s[22] ^ s[24], s[22], s[23]); s[23] = bitselect(s[23] ^ tmp1, s[23], s[24]); s[24] = bitselect(s[24] ^ tmp2, s[24], tmp1);
s[0] ^= 0x0000000000000001;
for(int i = 1; i < 24; ++i)
{
tmpxor[0] = s[0] ^ s[5] ^ s[10] ^ s[15] ^ s[20];
tmpxor[1] = s[1] ^ s[6] ^ s[11] ^ s[16] ^ s[21];
tmpxor[2] = s[2] ^ s[7] ^ s[12] ^ s[17] ^ s[22];
tmpxor[3] = s[3] ^ s[8] ^ s[13] ^ s[18] ^ s[23];
tmpxor[4] = s[4] ^ s[9] ^ s[14] ^ s[19] ^ s[24];
bc[0] = tmpxor[0] ^ rotl64_1(tmpxor[2], 1);
bc[1] = tmpxor[1] ^ rotl64_1(tmpxor[3], 1);
bc[2] = tmpxor[2] ^ rotl64_1(tmpxor[4], 1);
bc[3] = tmpxor[3] ^ rotl64_1(tmpxor[0], 1);
bc[4] = tmpxor[4] ^ rotl64_1(tmpxor[1], 1);
tmp1 = s[1] ^ bc[0];
s[0] ^= bc[4];
s[1] = rotl64_2(s[6] ^ bc[0], 12);
s[6] = rotl64_1(s[9] ^ bc[3], 20);
s[9] = rotl64_2(s[22] ^ bc[1], 29);
s[22] = rotl64_2(s[14] ^ bc[3], 7);
s[14] = rotl64_1(s[20] ^ bc[4], 18);
s[20] = rotl64_2(s[2] ^ bc[1], 30);
s[2] = rotl64_2(s[12] ^ bc[1], 11);
s[12] = rotl64_1(s[13] ^ bc[2], 25);
s[13] = rotl64_1(s[19] ^ bc[3], 8);
s[19] = rotl64_2(s[23] ^ bc[2], 24);
s[23] = rotl64_2(s[15] ^ bc[4], 9);
s[15] = rotl64_1(s[4] ^ bc[3], 27);
s[4] = rotl64_1(s[24] ^ bc[3], 14);
s[24] = rotl64_1(s[21] ^ bc[0], 2);
s[21] = rotl64_2(s[8] ^ bc[2], 23);
s[8] = rotl64_2(s[16] ^ bc[0], 13);
s[16] = rotl64_2(s[5] ^ bc[4], 4);
s[5] = rotl64_1(s[3] ^ bc[2], 28);
s[3] = rotl64_1(s[18] ^ bc[2], 21);
s[18] = rotl64_1(s[17] ^ bc[1], 15);
s[17] = rotl64_1(s[11] ^ bc[0], 10);
s[11] = rotl64_1(s[7] ^ bc[1], 6);
s[7] = rotl64_1(s[10] ^ bc[4], 3);
s[10] = rotl64_1(tmp1, 1);
tmp1 = s[0]; tmp2 = s[1]; s[0] = bitselect(s[0] ^ s[2], s[0], s[1]); s[1] = bitselect(s[1] ^ s[3], s[1], s[2]); s[2] = bitselect(s[2] ^ s[4], s[2], s[3]); s[3] = bitselect(s[3] ^ tmp1, s[3], s[4]); s[4] = bitselect(s[4] ^ tmp2, s[4], tmp1);
tmp1 = s[5]; tmp2 = s[6]; s[5] = bitselect(s[5] ^ s[7], s[5], s[6]); s[6] = bitselect(s[6] ^ s[8], s[6], s[7]); s[7] = bitselect(s[7] ^ s[9], s[7], s[8]); s[8] = bitselect(s[8] ^ tmp1, s[8], s[9]); s[9] = bitselect(s[9] ^ tmp2, s[9], tmp1);
tmp1 = s[10]; tmp2 = s[11]; s[10] = bitselect(s[10] ^ s[12], s[10], s[11]); s[11] = bitselect(s[11] ^ s[13], s[11], s[12]); s[12] = bitselect(s[12] ^ s[14], s[12], s[13]); s[13] = bitselect(s[13] ^ tmp1, s[13], s[14]); s[14] = bitselect(s[14] ^ tmp2, s[14], tmp1);
tmp1 = s[15]; tmp2 = s[16]; s[15] = bitselect(s[15] ^ s[17], s[15], s[16]); s[16] = bitselect(s[16] ^ s[18], s[16], s[17]); s[17] = bitselect(s[17] ^ s[19], s[17], s[18]); s[18] = bitselect(s[18] ^ tmp1, s[18], s[19]); s[19] = bitselect(s[19] ^ tmp2, s[19], tmp1);
tmp1 = s[20]; tmp2 = s[21]; s[20] = bitselect(s[20] ^ s[22], s[20], s[21]); s[21] = bitselect(s[21] ^ s[23], s[21], s[22]); s[22] = bitselect(s[22] ^ s[24], s[22], s[23]); s[23] = bitselect(s[23] ^ tmp1, s[23], s[24]); s[24] = bitselect(s[24] ^ tmp2, s[24], tmp1);
s[0] ^= keccakf_rndc[i];
}
}
__device__ __forceinline__
void cn_keccak(const uint32_t * __restrict__ in, uint64_t * __restrict__ md)
{
uint64_t st[25];
MEMCPY4(st, in, 19);
MEMSET8(&st[10], 0x00, 15);
st[9] = (st[9] & 0x00000000FFFFFFFFULL) | 0x0000000100000000ULL;
st[16] = 0x8000000000000000ULL;
cn_keccakf(st);
MEMCPY8(md, st, 25);
return;
}
+319
View File
@@ -0,0 +1,319 @@
typedef unsigned int uint_t; /* native unsigned integer */
#define SKEIN_MODIFIER_WORDS ( 2) /* number of modifier (tweak) words */
#define SKEIN_256_STATE_WORDS ( 4)
#define SKEIN_512_STATE_WORDS ( 8)
#define SKEIN_256_STATE_BYTES ( 8*SKEIN_256_STATE_WORDS)
#define SKEIN_512_STATE_BYTES ( 8*SKEIN_512_STATE_WORDS)
#define SKEIN_256_STATE_BITS (64*SKEIN_256_STATE_WORDS)
#define SKEIN_512_STATE_BITS (64*SKEIN_512_STATE_WORDS)
#define SKEIN_256_BLOCK_BYTES ( 8*SKEIN_256_STATE_WORDS)
#define SKEIN_512_BLOCK_BYTES ( 8*SKEIN_512_STATE_WORDS)
#define SKEIN_MK_64(hi32,lo32) ((lo32) + (((uint64_t) (hi32)) << 32))
#define SKEIN_KS_PARITY SKEIN_MK_64(0x1BD11BDA,0xA9FC1A22)
#define SKEIN_T1_BIT(BIT) ((BIT) - 64) /* offset 64 because it's the second word */
#define SKEIN_T1_POS_FIRST SKEIN_T1_BIT(126) /* bits 126 : first block flag */
#define SKEIN_T1_POS_BIT_PAD SKEIN_T1_BIT(119) /* bit 119 : partial final input byte */
#define SKEIN_T1_POS_FINAL SKEIN_T1_BIT(127) /* bit 127 : final block flag */
#define SKEIN_T1_POS_BLK_TYPE SKEIN_T1_BIT(120) /* bits 120..125: type field */
#define SKEIN_T1_FLAG_FIRST (((uint64_t) 1 ) << SKEIN_T1_POS_FIRST)
#define SKEIN_T1_FLAG_BIT_PAD (((uint64_t) 1 ) << SKEIN_T1_POS_BIT_PAD)
#define SKEIN_T1_FLAG_FINAL (((uint64_t) 1 ) << SKEIN_T1_POS_FINAL)
#define SKEIN_BLK_TYPE_MSG (48) /* message processing */
#define SKEIN_BLK_TYPE_OUT (63) /* output stage */
#define SKEIN_T1_BLK_TYPE(T) (((uint64_t) (SKEIN_BLK_TYPE_##T)) << SKEIN_T1_POS_BLK_TYPE)
#define SKEIN_T1_BLK_TYPE_MSG SKEIN_T1_BLK_TYPE(MSG) /* message processing */
#define SKEIN_T1_BLK_TYPE_OUT SKEIN_T1_BLK_TYPE(OUT) /* output stage */
#define SKEIN_T1_BLK_TYPE_OUT_FINAL (SKEIN_T1_BLK_TYPE_OUT | SKEIN_T1_FLAG_FINAL)
#define Skein_Set_Tweak(ctxPtr,TWK_NUM,tVal) {(ctxPtr)->h.T[TWK_NUM] = (tVal);}
#define Skein_Set_T0(ctxPtr,T0) Skein_Set_Tweak(ctxPtr,0,T0)
#define Skein_Set_T1(ctxPtr,T1) Skein_Set_Tweak(ctxPtr,1,T1)
#define Skein_Set_T0_T1(ctxPtr,T0,T1) { \
Skein_Set_T0(ctxPtr,(T0)); \
Skein_Set_T1(ctxPtr,(T1)); }
#define Skein_Start_New_Type(ctxPtr,BLK_TYPE) \
{ Skein_Set_T0_T1(ctxPtr,0,SKEIN_T1_FLAG_FIRST | SKEIN_T1_BLK_TYPE_##BLK_TYPE); (ctxPtr)->h.bCnt=0; }
#define Skein_Set_Bit_Pad_Flag(hdr) { (hdr).T[1] |= SKEIN_T1_FLAG_BIT_PAD; }
#define KW_TWK_BASE (0)
#define KW_KEY_BASE (3)
#define ks (kw + KW_KEY_BASE)
#define ts (kw + KW_TWK_BASE)
#define R512(p0,p1,p2,p3,p4,p5,p6,p7,R512ROT,rNum) \
X##p0 += X##p1; X##p1 = ROTL64(X##p1,R512ROT##_0); X##p1 ^= X##p0; \
X##p2 += X##p3; X##p3 = ROTL64(X##p3,R512ROT##_1); X##p3 ^= X##p2; \
X##p4 += X##p5; X##p5 = ROTL64(X##p5,R512ROT##_2); X##p5 ^= X##p4; \
X##p6 += X##p7; X##p7 = ROTL64(X##p7,R512ROT##_3); X##p7 ^= X##p6;
#define I512(R) \
X0 += ks[((R)+1) % 9]; \
X1 += ks[((R)+2) % 9]; \
X2 += ks[((R)+3) % 9]; \
X3 += ks[((R)+4) % 9]; \
X4 += ks[((R)+5) % 9]; \
X5 += ks[((R)+6) % 9] + ts[((R)+1) % 3]; \
X6 += ks[((R)+7) % 9] + ts[((R)+2) % 3]; \
X7 += ks[((R)+8) % 9] + (R)+1;
#define R512_8_rounds(R) \
R512(0,1,2,3,4,5,6,7,R_512_0,8*(R)+ 1); \
R512(2,1,4,7,6,5,0,3,R_512_1,8*(R)+ 2); \
R512(4,1,6,3,0,5,2,7,R_512_2,8*(R)+ 3); \
R512(6,1,0,7,2,5,4,3,R_512_3,8*(R)+ 4); \
I512(2*(R)); \
R512(0,1,2,3,4,5,6,7,R_512_4,8*(R)+ 5); \
R512(2,1,4,7,6,5,0,3,R_512_5,8*(R)+ 6); \
R512(4,1,6,3,0,5,2,7,R_512_6,8*(R)+ 7); \
R512(6,1,0,7,2,5,4,3,R_512_7,8*(R)+ 8); \
I512(2*(R)+1);
typedef struct
{
size_t hashBitLen;
size_t bCnt;
uint64_t T[SKEIN_MODIFIER_WORDS];
} Skein_Ctxt_Hdr_t;
typedef struct {
Skein_Ctxt_Hdr_t h;
uint64_t X[SKEIN_256_STATE_WORDS];
uint8_t b[SKEIN_256_BLOCK_BYTES];
} Skein_256_Ctxt_t;
typedef struct {
Skein_Ctxt_Hdr_t h;
uint64_t X[SKEIN_512_STATE_WORDS];
uint8_t b[SKEIN_512_BLOCK_BYTES];
} Skein_512_Ctxt_t;
typedef struct {
uint_t statebits;
union {
Skein_Ctxt_Hdr_t h;
Skein_256_Ctxt_t ctx_256;
Skein_512_Ctxt_t ctx_512;
} u;
} skeinHashState;
__device__
void cn_skein_init(skeinHashState *state, size_t hashBitLen)
{
const uint64_t SKEIN_512_IV_256[] =
{
SKEIN_MK_64(0xCCD044A1,0x2FDB3E13),
SKEIN_MK_64(0xE8359030,0x1A79A9EB),
SKEIN_MK_64(0x55AEA061,0x4F816E6F),
SKEIN_MK_64(0x2A2767A4,0xAE9B94DB),
SKEIN_MK_64(0xEC06025E,0x74DD7683),
SKEIN_MK_64(0xE7A436CD,0xC4746251),
SKEIN_MK_64(0xC36FBAF9,0x393AD185),
SKEIN_MK_64(0x3EEDBA18,0x33EDFC13)
};
Skein_512_Ctxt_t *ctx = &state->u.ctx_512;
ctx->h.hashBitLen = hashBitLen;
memcpy(ctx->X, SKEIN_512_IV_256, sizeof(ctx->X));
Skein_Start_New_Type(ctx, MSG);
}
__device__
void cn_skein_processblock(Skein_512_Ctxt_t * __restrict__ ctx, const uint8_t * __restrict__ blkPtr, size_t blkCnt, size_t byteCntAdd)
{
enum {
R_512_0_0=46, R_512_0_1=36, R_512_0_2=19, R_512_0_3=37,
R_512_1_0=33, R_512_1_1=27, R_512_1_2=14, R_512_1_3=42,
R_512_2_0=17, R_512_2_1=49, R_512_2_2=36, R_512_2_3=39,
R_512_3_0=44, R_512_3_1= 9, R_512_3_2=54, R_512_3_3=56,
R_512_4_0=39, R_512_4_1=30, R_512_4_2=34, R_512_4_3=24,
R_512_5_0=13, R_512_5_1=50, R_512_5_2=10, R_512_5_3=17,
R_512_6_0=25, R_512_6_1=29, R_512_6_2=39, R_512_6_3=43,
R_512_7_0= 8, R_512_7_1=35, R_512_7_2=56, R_512_7_3=22
};
uint64_t X0,X1,X2,X3,X4,X5,X6,X7;
uint64_t w[SKEIN_512_STATE_WORDS];
uint64_t kw[SKEIN_512_STATE_WORDS+4];
ts[0] = ctx->h.T[0];
ts[1] = ctx->h.T[1];
do {
ts[0] += byteCntAdd;
ks[0] = ctx->X[0];
ks[1] = ctx->X[1];
ks[2] = ctx->X[2];
ks[3] = ctx->X[3];
ks[4] = ctx->X[4];
ks[5] = ctx->X[5];
ks[6] = ctx->X[6];
ks[7] = ctx->X[7];
ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;
ts[2] = ts[0] ^ ts[1];
memcpy(w, blkPtr, SKEIN_512_STATE_WORDS << 3);
X0 = w[0] + ks[0];
X1 = w[1] + ks[1];
X2 = w[2] + ks[2];
X3 = w[3] + ks[3];
X4 = w[4] + ks[4];
X5 = w[5] + ks[5] + ts[0];
X6 = w[6] + ks[6] + ts[1];
X7 = w[7] + ks[7];
blkPtr += SKEIN_512_BLOCK_BYTES;
R512_8_rounds( 0);
R512_8_rounds( 1);
R512_8_rounds( 2);
R512_8_rounds( 3);
R512_8_rounds( 4);
R512_8_rounds( 5);
R512_8_rounds( 6);
R512_8_rounds( 7);
R512_8_rounds( 8);
ctx->X[0] = X0 ^ w[0];
ctx->X[1] = X1 ^ w[1];
ctx->X[2] = X2 ^ w[2];
ctx->X[3] = X3 ^ w[3];
ctx->X[4] = X4 ^ w[4];
ctx->X[5] = X5 ^ w[5];
ctx->X[6] = X6 ^ w[6];
ctx->X[7] = X7 ^ w[7];
ts[1] &= ~SKEIN_T1_FLAG_FIRST;
} while (--blkCnt);
ctx->h.T[0] = ts[0];
ctx->h.T[1] = ts[1];
}
__device__
void cn_skein_block(Skein_512_Ctxt_t * __restrict__ ctx, const uint8_t * __restrict__ msg, size_t msgByteCnt)
{
size_t n;
if (msgByteCnt + ctx->h.bCnt > SKEIN_512_BLOCK_BYTES)
{
if (ctx->h.bCnt) {
n = SKEIN_512_BLOCK_BYTES - ctx->h.bCnt;
if (n) {
memcpy(&ctx->b[ctx->h.bCnt],msg,n);
msgByteCnt -= n;
msg += n;
ctx->h.bCnt += n;
}
cn_skein_processblock(ctx, ctx->b, 1, SKEIN_512_BLOCK_BYTES);
ctx->h.bCnt = 0;
}
if (msgByteCnt > SKEIN_512_BLOCK_BYTES) {
n = (msgByteCnt - 1) / SKEIN_512_BLOCK_BYTES;
cn_skein_processblock(ctx, msg, n, SKEIN_512_BLOCK_BYTES);
msgByteCnt -= n * SKEIN_512_BLOCK_BYTES;
msg += n * SKEIN_512_BLOCK_BYTES;
}
}
if (msgByteCnt) {
memcpy(&ctx->b[ctx->h.bCnt],msg,msgByteCnt);
ctx->h.bCnt += msgByteCnt;
}
}
__device__
void cn_skein_update(skeinHashState * __restrict__ state, const uint8_t * __restrict__ data, DataLength databitlen)
{
if ((databitlen & 7) == 0) {
cn_skein_block(&state->u.ctx_512, data, databitlen >> 3);
}
else {
size_t bCnt = (databitlen >> 3) + 1;
uint8_t b,mask;
mask = (uint8_t) (1u << (7 - (databitlen & 7)));
b = (uint8_t) ((data[bCnt-1] & (0-mask)) | mask);
cn_skein_block(&state->u.ctx_512, data, bCnt - 1);
cn_skein_block(&state->u.ctx_512, &b, 1);
Skein_Set_Bit_Pad_Flag(state->u.h);
}
}
__device__
void cn_skein_final(skeinHashState * __restrict__ state, uint8_t * __restrict__ hashVal)
{
uint64_t X[SKEIN_512_STATE_WORDS];
Skein_512_Ctxt_t *ctx = (Skein_512_Ctxt_t *)&state->u.ctx_512;
const int byteCnt = (ctx->h.hashBitLen + 7) >> 3;
ctx->h.T[1] |= SKEIN_T1_FLAG_FINAL;
if (ctx->h.bCnt < SKEIN_512_BLOCK_BYTES)
{
memset(&ctx->b[ctx->h.bCnt], 0, SKEIN_512_BLOCK_BYTES - ctx->h.bCnt);
}
cn_skein_processblock(ctx, ctx->b, 1, ctx->h.bCnt);
memset(ctx->b, 0, sizeof(ctx->b));
memcpy(X, ctx->X, sizeof(X));
for (int i = 0; i*SKEIN_512_BLOCK_BYTES < byteCnt; i++)
{
int n = byteCnt - i*SKEIN_512_BLOCK_BYTES;
if (n > SKEIN_512_BLOCK_BYTES) n = SKEIN_512_BLOCK_BYTES;
((uint64_t *)ctx->b)[0] = (uint64_t)i;
Skein_Start_New_Type(ctx, OUT_FINAL);
cn_skein_processblock(ctx, ctx->b, 1, sizeof(uint64_t));
memcpy(hashVal + (i*SKEIN_512_BLOCK_BYTES), ctx->X, n);
memcpy(ctx->X, X, sizeof(X)); // restore the counter mode key for next time
}
}
__device__
void cn_skein(const uint8_t * __restrict__ data, DataLength len, uint32_t * hashval)
{
int hashbitlen = 256;
DataLength databitlen = len << 3;
skeinHashState state;
state.statebits = 64*SKEIN_512_STATE_WORDS;
cn_skein_init(&state, hashbitlen);
cn_skein_update(&state, data, databitlen);
cn_skein_final(&state, (uint8_t*) hashval);
}
+122
View File
@@ -0,0 +1,122 @@
// keccak.c
// 19-Nov-11 Markku-Juhani O. Saarinen <mjos@iki.fi>
// A baseline Keccak (3rd round) implementation.
#include "c_keccak.h"
const uint64_t keccakf_rndc[24] =
{
0x0000000000000001, 0x0000000000008082, 0x800000000000808a,
0x8000000080008000, 0x000000000000808b, 0x0000000080000001,
0x8000000080008081, 0x8000000000008009, 0x000000000000008a,
0x0000000000000088, 0x0000000080008009, 0x000000008000000a,
0x000000008000808b, 0x800000000000008b, 0x8000000000008089,
0x8000000000008003, 0x8000000000008002, 0x8000000000000080,
0x000000000000800a, 0x800000008000000a, 0x8000000080008081,
0x8000000000008080, 0x0000000080000001, 0x8000000080008008
};
const int keccakf_rotc[24] =
{
1, 3, 6, 10, 15, 21, 28, 36, 45, 55, 2, 14,
27, 41, 56, 8, 25, 43, 62, 18, 39, 61, 20, 44
};
const int keccakf_piln[24] =
{
10, 7, 11, 17, 18, 3, 5, 16, 8, 21, 24, 4,
15, 23, 19, 13, 12, 2, 20, 14, 22, 9, 6, 1
};
// update the state with given number of rounds
void keccakf(uint64_t st[25], int rounds)
{
int i, j, round;
uint64_t t, bc[5];
for (round = 0; round < rounds; ++round) {
// Theta
bc[0] = st[0] ^ st[5] ^ st[10] ^ st[15] ^ st[20];
bc[1] = st[1] ^ st[6] ^ st[11] ^ st[16] ^ st[21];
bc[2] = st[2] ^ st[7] ^ st[12] ^ st[17] ^ st[22];
bc[3] = st[3] ^ st[8] ^ st[13] ^ st[18] ^ st[23];
bc[4] = st[4] ^ st[9] ^ st[14] ^ st[19] ^ st[24];
for (i = 0; i < 5; ++i) {
t = bc[(i + 4) % 5] ^ ROTL64(bc[(i + 1) % 5], 1);
st[i ] ^= t;
st[i + 5] ^= t;
st[i + 10] ^= t;
st[i + 15] ^= t;
st[i + 20] ^= t;
}
// Rho Pi
t = st[1];
for (i = 0; i < 24; ++i) {
bc[0] = st[keccakf_piln[i]];
st[keccakf_piln[i]] = ROTL64(t, keccakf_rotc[i]);
t = bc[0];
}
// Chi
for (j = 0; j < 25; j += 5) {
bc[0] = st[j ];
bc[1] = st[j + 1];
bc[2] = st[j + 2];
bc[3] = st[j + 3];
bc[4] = st[j + 4];
st[j ] ^= (~bc[1]) & bc[2];
st[j + 1] ^= (~bc[2]) & bc[3];
st[j + 2] ^= (~bc[3]) & bc[4];
st[j + 3] ^= (~bc[4]) & bc[0];
st[j + 4] ^= (~bc[0]) & bc[1];
}
// Iota
st[0] ^= keccakf_rndc[round];
}
}
// compute a keccak hash (md) of given byte length from "in"
typedef uint64_t state_t[25];
int keccak(const uint8_t *in, int inlen, uint8_t *md, int mdlen)
{
state_t st;
uint8_t temp[144];
int i, rsiz, rsizw;
rsiz = sizeof(state_t) == mdlen ? HASH_DATA_AREA : 200 - 2 * mdlen;
rsizw = rsiz / 8;
memset(st, 0, sizeof(st));
for ( ; inlen >= rsiz; inlen -= rsiz, in += rsiz) {
for (i = 0; i < rsizw; i++)
st[i] ^= ((uint64_t *) in)[i];
keccakf(st, KECCAK_ROUNDS);
}
// last block and padding
memcpy(temp, in, inlen);
temp[inlen++] = 1;
memset(temp + inlen, 0, rsiz - inlen);
temp[rsiz - 1] |= 0x80;
for (i = 0; i < rsizw; i++)
st[i] ^= ((uint64_t *) temp)[i];
keccakf(st, KECCAK_ROUNDS);
memcpy(md, st, mdlen);
return 0;
}
void keccak1600(const uint8_t *in, int inlen, uint8_t *md)
{
keccak(in, inlen, md, sizeof(state_t));
}
+34
View File
@@ -0,0 +1,34 @@
// keccak.h
// 19-Nov-11 Markku-Juhani O. Saarinen <mjos@iki.fi>
#ifndef KECCAK_H
#define KECCAK_H
#include <stdint.h>
#include <string.h>
#ifndef KECCAK_ROUNDS
#define KECCAK_ROUNDS 24
#endif
#ifndef ROTL64
#define ROTL64(x, y) (((x) << (y)) | ((x) >> (64 - (y))))
#endif
#ifndef HASH_SIZE
#define HASH_SIZE 32
#endif
#ifndef HASH_DATA_AREA
#define HASH_DATA_AREA 136
#endif
// compute a keccak hash (md) of given byte length from "in"
int keccak(const uint8_t *in, int inlen, uint8_t *md, int mdlen);
// update the state
void keccakf(uint64_t st[25], int norounds);
void keccak1600(const uint8_t *in, int inlen, uint8_t *md);
#endif
+351
View File
@@ -0,0 +1,351 @@
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <sys/time.h>
#include <unistd.h>
#include "cryptolight.h"
#define LONG_SHL_IDX 18
#define LONG_LOOPS32 0x40000
#include "cn_aes.cuh"
#define MUL_SUM_XOR_DST(a,c,dst) { \
uint64_t hi, lo = cuda_mul128(((uint64_t *)a)[0], ((uint64_t *)dst)[0], &hi) + ((uint64_t *)c)[1]; \
hi += ((uint64_t *)c)[0]; \
((uint64_t *)c)[0] = ((uint64_t *)dst)[0] ^ hi; \
((uint64_t *)c)[1] = ((uint64_t *)dst)[1] ^ lo; \
((uint64_t *)dst)[0] = hi; \
((uint64_t *)dst)[1] = lo; }
__device__ __forceinline__ uint64_t cuda_mul128(uint64_t multiplier, uint64_t multiplicand, uint64_t* product_hi)
{
*product_hi = __umul64hi(multiplier, multiplicand);
return(multiplier * multiplicand);
}
__global__
void cryptolight_core_gpu_phase1(int threads, uint32_t * long_state, uint32_t * ctx_state, uint32_t * ctx_key1)
{
__shared__ uint32_t __align__(16) sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
const int thread = (blockDim.x * blockIdx.x + threadIdx.x) >> 3;
const int sub = (threadIdx.x & 7) << 2;
if(thread < threads)
{
const int oft = thread * 50 + sub + 16; // not aligned 16!
const int long_oft = (thread << LONG_SHL_IDX) + sub;
uint32_t __align__(16) key[40];
uint32_t __align__(16) text[4];
// copy 160 bytes
#pragma unroll
for (int i = 0; i < 40; i += 4)
AS_UINT4(&key[i]) = AS_UINT4(ctx_key1 + thread * 40 + i);
AS_UINT2(&text[0]) = AS_UINT2(&ctx_state[oft]);
AS_UINT2(&text[2]) = AS_UINT2(&ctx_state[oft + 2]);
__syncthreads();
for(int i = 0; i < LONG_LOOPS32; i += 32) {
cn_aes_pseudo_round_mut(sharedMemory, text, key);
AS_UINT4(&long_state[long_oft + i]) = AS_UINT4(text);
}
}
}
// --------------------------------------------------------------------------------------------------------------
__global__
void cryptolight_old_gpu_phase2(const int threads, const int bfactor, const int partidx, uint32_t * d_long_state, uint32_t * d_ctx_a, uint32_t * d_ctx_b)
{
__shared__ uint32_t __align__(16) sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
__syncthreads();
#if 0 && __CUDA_ARCH__ >= 300
const int thread = (blockDim.x * blockIdx.x + threadIdx.x) >> 2;
const int sub = threadIdx.x & 3;
if(thread < threads)
{
const int batchsize = ITER >> (2 + bfactor);
const int start = partidx * batchsize;
const int end = start + batchsize;
uint32_t * __restrict__ long_state = &d_long_state[thread << LONG_SHL_IDX];
uint32_t * __restrict__ ctx_a = d_ctx_a + thread * 4;
uint32_t * __restrict__ ctx_b = d_ctx_b + thread * 4;
uint32_t a, b, c, x[4];
uint32_t t1[4], t2[4], res;
uint64_t reshi, reslo;
int j;
a = ctx_a[sub];
b = ctx_b[sub];
#pragma unroll 8
for(int i = start; i < end; ++i)
{
//j = ((uint32_t *)a)[0] & 0xFFFF0;
j = (__shfl((int)a, 0, 4) & E2I_MASK1) >> 2;
//cn_aes_single_round(sharedMemory, &long_state[j], c, a);
x[0] = long_state[j + sub];
x[1] = __shfl((int)x[0], sub + 1, 4);
x[2] = __shfl((int)x[0], sub + 2, 4);
x[3] = __shfl((int)x[0], sub + 3, 4);
c = a ^
t_fn0(x[0] & 0xff) ^
t_fn1((x[1] >> 8) & 0xff) ^
t_fn2((x[2] >> 16) & 0xff) ^
t_fn3((x[3] >> 24) & 0xff);
//XOR_BLOCKS_DST(c, b, &long_state[j]);
long_state[j + sub] = c ^ b;
//MUL_SUM_XOR_DST(c, a, &long_state[((uint32_t *)c)[0] & 0xFFFF0]);
j = (__shfl((int)c, 0, 4) & E2I_MASK1) >> 2;
#pragma unroll
for(int k = 0; k < 2; k++)
t1[k] = __shfl((int)c, k, 4);
#pragma unroll
for(int k = 0; k < 4; k++)
t2[k] = __shfl((int)a, k, 4);
asm(
"mad.lo.u64 %0, %2, %3, %4;\n\t"
"mad.hi.u64 %1, %2, %3, %5;\n\t"
: "=l"(reslo), "=l"(reshi)
: "l"(((uint64_t *)t1)[0]), "l"(((uint64_t *)long_state)[j >> 1]), "l"(((uint64_t *)t2)[1]), "l"(((uint64_t *)t2)[0]));
res = (sub & 2 ? reslo : reshi) >> (sub & 1 ? 32 : 0);
a = long_state[j + sub] ^ res;
long_state[j + sub] = res;
//j = ((uint32_t *)a)[0] & 0xFFFF0;
j = (__shfl((int)a, 0, 4) & E2I_MASK1) >> 2;
//cn_aes_single_round(sharedMemory, &long_state[j], b, a);
x[0] = long_state[j + sub];
x[1] = __shfl((int)x[0], sub + 1, 4);
x[2] = __shfl((int)x[0], sub + 2, 4);
x[3] = __shfl((int)x[0], sub + 3, 4);
b = a ^
t_fn0(x[0] & 0xff) ^
t_fn1((x[1] >> 8) & 0xff) ^
t_fn2((x[2] >> 16) & 0xff) ^
t_fn3((x[3] >> 24) & 0xff);
//XOR_BLOCKS_DST(b, c, &long_state[j]);
long_state[j + sub] = c ^ b;
//MUL_SUM_XOR_DST(b, a, &long_state[((uint32_t *)b)[0] & 0xFFFF0]);
j = (__shfl((int)b, 0, 4) & E2I_MASK1) >> 2;
#pragma unroll
for(int k = 0; k < 2; k++)
t1[k] = __shfl((int)b, k, 4);
#pragma unroll
for(int k = 0; k < 4; k++)
t2[k] = __shfl((int)a, k, 4);
asm(
"mad.lo.u64 %0, %2, %3, %4;\n\t"
"mad.hi.u64 %1, %2, %3, %5;\n\t"
: "=l"(reslo), "=l"(reshi)
: "l"(((uint64_t *)t1)[0]), "l"(((uint64_t *)long_state)[j >> 1]), "l"(((uint64_t *)t2)[1]), "l"(((uint64_t *)t2)[0]));
res = (sub & 2 ? reslo : reshi) >> (sub & 1 ? 32 : 0);
a = long_state[j + sub] ^ res;
long_state[j + sub] = res;
}
if(bfactor > 0)
{
ctx_a[sub] = a;
ctx_b[sub] = b;
}
}
#else // __CUDA_ARCH__ < 300
const int thread = blockDim.x * blockIdx.x + threadIdx.x;
if (thread < threads)
{
const int batchsize = ITER >> (2 + bfactor);
const int start = partidx * batchsize;
const int end = start + batchsize;
const int longptr = thread << LONG_SHL_IDX;
uint32_t * long_state = &d_long_state[longptr];
uint64_t * ctx_a = (uint64_t*)(&d_ctx_a[thread * 4]);
uint64_t * ctx_b = (uint64_t*)(&d_ctx_b[thread * 4]);
uint4 A = AS_UINT4(ctx_a);
uint4 B = AS_UINT4(ctx_b);
uint32_t* a = (uint32_t*)&A;
uint32_t* b = (uint32_t*)&B;
for (int i = start; i < end; i++) // end = 262144
{
uint32_t c[4];
uint32_t j = (a[0] >> 2) & E2I_MASK2;
cn_aes_single_round(sharedMemory, &long_state[j], c, a);
XOR_BLOCKS_DST(c, b, &long_state[j]);
MUL_SUM_XOR_DST(c, a, &long_state[(c[0] >> 2) & E2I_MASK2]);
j = (a[0] >> 2) & E2I_MASK2;
cn_aes_single_round(sharedMemory, &long_state[j], b, a);
XOR_BLOCKS_DST(b, c, &long_state[j]);
MUL_SUM_XOR_DST(b, a, &long_state[(b[0] >> 2) & E2I_MASK2]);
}
if (bfactor > 0) {
AS_UINT4(ctx_a) = A;
AS_UINT4(ctx_b) = B;
}
}
#endif // __CUDA_ARCH__ >= 300
}
__device__ __forceinline__ void store_variant1(uint32_t* long_state)
{
uint4* Z = (uint4*) long_state;
const uint32_t tmp = (Z->z >> 24); // __byte_perm(src, 0, 0x7773);
const uint32_t index = (((tmp >> 3) & 6u) | (tmp & 1u)) << 1;
Z->z = (Z->z & 0x00ffffffu) | ((tmp ^ ((0x75310u >> index) & 0x30u)) << 24);
}
#define MUL_SUM_XOR_DST_1(a,c,dst,tweak) { \
uint64_t hi, lo = cuda_mul128(((uint64_t *)a)[0], ((uint64_t *)dst)[0], &hi) + ((uint64_t *)c)[1]; \
hi += ((uint64_t *)c)[0]; \
((uint64_t *)c)[0] = ((uint64_t *)dst)[0] ^ hi; \
((uint64_t *)c)[1] = ((uint64_t *)dst)[1] ^ lo; \
((uint64_t *)dst)[0] = hi; \
((uint64_t *)dst)[1] = lo ^ tweak; }
__global__
void cryptolight_gpu_phase2(const uint32_t threads, const uint16_t bfactor, const uint32_t partidx,
uint32_t * __restrict__ d_long_state, uint32_t * __restrict__ d_ctx_a, uint32_t * __restrict__ d_ctx_b,
uint64_t * __restrict__ d_tweak)
{
__shared__ __align__(16) uint32_t sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
__syncthreads();
const uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
if (thread < threads)
{
const uint32_t batchsize = ITER >> (2 + bfactor);
const uint32_t start = partidx * batchsize;
const uint32_t end = start + batchsize;
const uint32_t longptr = thread << LONG_SHL_IDX;
uint32_t * long_state = &d_long_state[longptr];
uint64_t tweak = d_tweak[thread];
void * ctx_a = (void*)(&d_ctx_a[thread << 2]);
void * ctx_b = (void*)(&d_ctx_b[thread << 2]);
uint4 A = AS_UINT4(ctx_a); // ld.global.u32.v4
uint4 B = AS_UINT4(ctx_b);
uint32_t* a = (uint32_t*)&A;
uint32_t* b = (uint32_t*)&B;
for (int i = start; i < end; i++)
{
uint32_t c[4];
uint32_t j = (A.x >> 2) & E2I_MASK2;
cn_aes_single_round(sharedMemory, &long_state[j], c, a);
XOR_BLOCKS_DST(c, b, &long_state[j]);
store_variant1(&long_state[j]);
MUL_SUM_XOR_DST_1(c, a, &long_state[(c[0] >> 2) & E2I_MASK2], tweak);
j = (A.x >> 2) & E2I_MASK2;
cn_aes_single_round(sharedMemory, &long_state[j], b, a);
XOR_BLOCKS_DST(b, c, &long_state[j]);
store_variant1(&long_state[j]);
MUL_SUM_XOR_DST_1(b, a, &long_state[(b[0] >> 2) & E2I_MASK2], tweak);
}
if (bfactor) {
AS_UINT4(ctx_a) = A;
AS_UINT4(ctx_b) = B;
}
}
}
__global__
void cryptolight_core_gpu_phase3(int threads, const uint32_t * long_state, uint32_t * ctx_state, uint32_t * ctx_key2)
{
__shared__ uint32_t __align__(16) sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
const int thread = (blockDim.x * blockIdx.x + threadIdx.x) >> 3;
const int sub = (threadIdx.x & 7) << 2;
if(thread < threads)
{
const int long_oft = (thread << LONG_SHL_IDX) + sub;
const int oft = thread * 50 + sub + 16;
uint32_t __align__(16) key[40];
uint32_t __align__(16) text[4];
#pragma unroll
for (int i = 0; i < 40; i += 4)
AS_UINT4(&key[i]) = AS_UINT4(ctx_key2 + thread * 40 + i);
AS_UINT2(&text[0]) = AS_UINT2(&ctx_state[oft + 0]);
AS_UINT2(&text[2]) = AS_UINT2(&ctx_state[oft + 2]);
__syncthreads();
for(int i = 0; i < LONG_LOOPS32; i += 32)
{
#pragma unroll
for(int j = 0; j < 4; j++)
text[j] ^= long_state[long_oft + i + j];
cn_aes_pseudo_round_mut(sharedMemory, text, key);
}
AS_UINT2(&ctx_state[oft + 0]) = AS_UINT2(&text[0]);
AS_UINT2(&ctx_state[oft + 2]) = AS_UINT2(&text[2]);
}
}
extern int device_bfactor[MAX_GPUS];
__host__
void cryptolight_core_hash(int thr_id, int blocks, int threads, uint32_t *d_long_state, uint32_t *d_ctx_state,
uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak)
{
dim3 grid(blocks);
dim3 block(threads);
dim3 block4(threads << 2);
dim3 block8(threads << 3);
const int bfactor = device_bfactor[thr_id];
const int bsleep = bfactor ? 100 : 0;
int i, partcount = 1 << bfactor;
int dev_id = device_map[thr_id];
cryptolight_core_gpu_phase1 <<<grid, block8 >>>(blocks*threads, d_long_state, d_ctx_state, d_ctx_key1);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
if(partcount > 1) usleep(bsleep);
for(i = 0; i < partcount; i++)
{
dim3 b = device_sm[dev_id] >= 300 ? block4 : block;
if (variant == 0)
cryptolight_old_gpu_phase2 <<<grid, b>>> (blocks*threads, bfactor, i, d_long_state, d_ctx_a, d_ctx_b);
else
cryptolight_gpu_phase2 <<<grid, b>>> (blocks*threads, bfactor, i, d_long_state, d_ctx_a, d_ctx_b, d_ctx_tweak);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
if(partcount > 1) usleep(bsleep);
}
cryptolight_core_gpu_phase3 <<<grid, block8 >>>(blocks*threads, d_long_state, d_ctx_state, d_ctx_key2);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
}
+246
View File
@@ -0,0 +1,246 @@
#include <miner.h>
#include <memory.h>
#include "oaes_lib.h"
#include "cryptolight.h"
extern "C" {
#include <sph/sph_blake.h>
#include <sph/sph_groestl.h>
#include <sph/sph_jh.h>
#include <sph/sph_skein.h>
#include "cpu/c_keccak.h"
}
struct cryptonight_ctx {
uint8_t long_state[MEMORY];
union cn_slow_hash_state state;
uint8_t text[INIT_SIZE_BYTE];
uint8_t a[AES_BLOCK_SIZE];
uint8_t b[AES_BLOCK_SIZE];
uint8_t c[AES_BLOCK_SIZE];
oaes_ctx* aes_ctx;
};
static void do_blake_hash(const void* input, int len, void* output)
{
uchar hash[32];
sph_blake256_context ctx;
sph_blake256_set_rounds(14);
sph_blake256_init(&ctx);
sph_blake256(&ctx, input, len);
sph_blake256_close(&ctx, hash);
memcpy(output, hash, 32);
}
static void do_groestl_hash(const void* input, int len, void* output)
{
uchar hash[32];
sph_groestl256_context ctx;
sph_groestl256_init(&ctx);
sph_groestl256(&ctx, input, len);
sph_groestl256_close(&ctx, hash);
memcpy(output, hash, 32);
}
static void do_jh_hash(const void* input, int len, void* output)
{
uchar hash[64];
sph_jh256_context ctx;
sph_jh256_init(&ctx);
sph_jh256(&ctx, input, len);
sph_jh256_close(&ctx, hash);
memcpy(output, hash, 32);
}
static void do_skein_hash(const void* input, int len, void* output)
{
uchar hash[32];
sph_skein256_context ctx;
sph_skein256_init(&ctx);
sph_skein256(&ctx, input, len);
sph_skein256_close(&ctx, hash);
memcpy(output, hash, 32);
}
// todo: use sph if possible
static void keccak_hash_permutation(union hash_state *state) {
keccakf((uint64_t*)state, 24);
}
static void keccak_hash_process(union hash_state *state, const uint8_t *buf, int count) {
keccak1600(buf, (int)count, (uint8_t*)state);
}
extern "C" int fast_aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey);
extern "C" int aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey);
extern "C" int aesb_pseudo_round_mut(uint8_t *val, uint8_t *expandedKey);
extern "C" int fast_aesb_pseudo_round_mut(uint8_t *val, uint8_t *expandedKey);
static void (* const extra_hashes[4])(const void*, int, void *) = {
do_blake_hash, do_groestl_hash, do_jh_hash, do_skein_hash
};
static uint64_t mul128(uint64_t multiplier, uint64_t multiplicand, uint64_t* product_hi)
{
// multiplier = ab = a * 2^32 + b
// multiplicand = cd = c * 2^32 + d
// ab * cd = a * c * 2^64 + (a * d + b * c) * 2^32 + b * d
uint64_t a = hi_dword(multiplier);
uint64_t b = lo_dword(multiplier);
uint64_t c = hi_dword(multiplicand);
uint64_t d = lo_dword(multiplicand);
uint64_t ac = a * c;
uint64_t ad = a * d;
uint64_t bc = b * c;
uint64_t bd = b * d;
uint64_t adbc = ad + bc;
uint64_t adbc_carry = adbc < ad ? 1 : 0;
// multiplier * multiplicand = product_hi * 2^64 + product_lo
uint64_t product_lo = bd + (adbc << 32);
uint64_t product_lo_carry = product_lo < bd ? 1 : 0;
*product_hi = ac + (adbc >> 32) + (adbc_carry << 32) + product_lo_carry;
return product_lo;
}
static size_t e2i(const uint8_t* a) {
//const uint32_t mask = (MEMORY / AES_BLOCK_SIZE - 1);
//return (*((uint64_t*) a) / AES_BLOCK_SIZE) & mask;
return *((uint64_t*) a) & 0xFFFF0; /* mask * AES_BLOCK_SIZE */
}
static void mul(const uint8_t* a, const uint8_t* b, uint8_t* res) {
((uint64_t*) res)[1] = mul128(((uint64_t*) a)[0], ((uint64_t*) b)[0], (uint64_t*) res);
}
static void sum_half_blocks(uint8_t* a, const uint8_t* b) {
((uint64_t*) a)[0] += ((uint64_t*) b)[0];
((uint64_t*) a)[1] += ((uint64_t*) b)[1];
}
static void sum_half_blocks_dst(const uint8_t* a, const uint8_t* b, uint8_t* dst) {
((uint64_t*) dst)[0] = ((uint64_t*) a)[0] + ((uint64_t*) b)[0];
((uint64_t*) dst)[1] = ((uint64_t*) a)[1] + ((uint64_t*) b)[1];
}
static void mul_sum_dst(const uint8_t* a, const uint8_t* b, const uint8_t* c, uint8_t* dst) {
((uint64_t*) dst)[1] = mul128(((uint64_t*) a)[0], ((uint64_t*) b)[0], (uint64_t*) dst) + ((uint64_t*) c)[1];
((uint64_t*) dst)[0] += ((uint64_t*) c)[0];
}
static void mul_sum_xor_dst(const uint8_t* a, uint8_t* c, uint8_t* dst, const int variant, const uint64_t tweak) {
uint64_t hi, lo = mul128(((uint64_t*) a)[0], ((uint64_t*) dst)[0], &hi) + ((uint64_t*) c)[1];
hi += ((uint64_t*) c)[0];
((uint64_t*) c)[0] = ((uint64_t*) dst)[0] ^ hi;
((uint64_t*) c)[1] = ((uint64_t*) dst)[1] ^ lo;
((uint64_t*) dst)[0] = hi;
((uint64_t*) dst)[1] = variant ? lo ^ tweak : lo;
}
static void copy_block(uint8_t* dst, const uint8_t* src) {
((uint64_t*) dst)[0] = ((uint64_t*) src)[0];
((uint64_t*) dst)[1] = ((uint64_t*) src)[1];
}
static void xor_blocks(uint8_t* a, const uint8_t* b) {
((uint64_t*) a)[0] ^= ((uint64_t*) b)[0];
((uint64_t*) a)[1] ^= ((uint64_t*) b)[1];
}
static void xor_blocks_dst(const uint8_t* a, const uint8_t* b, uint8_t* dst) {
((uint64_t*) dst)[0] = ((uint64_t*) a)[0] ^ ((uint64_t*) b)[0];
((uint64_t*) dst)[1] = ((uint64_t*) a)[1] ^ ((uint64_t*) b)[1];
}
static void cryptolight_store_variant(void* state, int variant) {
if (variant == 1) {
// use variant 1 like monero since june 2018
const uint8_t tmp = ((const uint8_t*)(state))[11];
const uint8_t index = (((tmp >> 3) & 6) | (tmp & 1)) << 1;
((uint8_t*)(state))[11] = tmp ^ ((0x75310 >> index) & 0x30);
}
}
static void cryptolight_hash_ctx(void* output, const void* input, const int len, struct cryptonight_ctx* ctx, const int variant)
{
size_t i, j;
keccak_hash_process(&ctx->state.hs, (const uint8_t*) input, len);
ctx->aes_ctx = (oaes_ctx*) oaes_alloc();
memcpy(ctx->text, ctx->state.init, INIT_SIZE_BYTE);
const uint64_t tweak = variant ? *((uint64_t*) (((uint8_t*)input) + 35)) ^ ctx->state.hs.w[24] : 0;
oaes_key_import_data(ctx->aes_ctx, ctx->state.hs.b, AES_KEY_SIZE);
for (i = 0; likely(i < MEMORY); i += INIT_SIZE_BYTE) {
#undef RND
#define RND(p) aesb_pseudo_round_mut(&ctx->text[AES_BLOCK_SIZE * p], ctx->aes_ctx->key->exp_data);
RND(0);
RND(1);
RND(2);
RND(3);
RND(4);
RND(5);
RND(6);
RND(7);
memcpy(&ctx->long_state[i], ctx->text, INIT_SIZE_BYTE);
}
xor_blocks_dst(&ctx->state.k[0], &ctx->state.k[32], ctx->a);
xor_blocks_dst(&ctx->state.k[16], &ctx->state.k[48], ctx->b);
for (i = 0; likely(i < ITER / 4); ++i) {
j = e2i(ctx->a);
aesb_single_round(&ctx->long_state[j], ctx->c, ctx->a);
xor_blocks_dst(ctx->c, ctx->b, &ctx->long_state[j]);
cryptolight_store_variant(&ctx->long_state[j], variant);
mul_sum_xor_dst(ctx->c, ctx->a, &ctx->long_state[e2i(ctx->c)], variant, tweak);
j = e2i(ctx->a);
aesb_single_round(&ctx->long_state[j], ctx->b, ctx->a);
xor_blocks_dst(ctx->b, ctx->c, &ctx->long_state[j]);
cryptolight_store_variant(&ctx->long_state[j], variant);
mul_sum_xor_dst(ctx->b, ctx->a, &ctx->long_state[e2i(ctx->b)], variant, tweak);
}
memcpy(ctx->text, ctx->state.init, INIT_SIZE_BYTE);
oaes_key_import_data(ctx->aes_ctx, &ctx->state.hs.b[32], AES_KEY_SIZE);
for (i = 0; likely(i < MEMORY); i += INIT_SIZE_BYTE) {
#undef RND
#define RND(p) xor_blocks(&ctx->text[p * AES_BLOCK_SIZE], &ctx->long_state[i + p * AES_BLOCK_SIZE]); \
aesb_pseudo_round_mut(&ctx->text[p * AES_BLOCK_SIZE], ctx->aes_ctx->key->exp_data);
RND(0);
RND(1);
RND(2);
RND(3);
RND(4);
RND(5);
RND(6);
RND(7);
}
memcpy(ctx->state.init, ctx->text, INIT_SIZE_BYTE);
keccak_hash_permutation(&ctx->state.hs);
int extra_algo = ctx->state.hs.b[0] & 3;
extra_hashes[extra_algo](&ctx->state, 200, output);
if (opt_debug) applog(LOG_DEBUG, "extra algo=%d", extra_algo);
oaes_free((OAES_CTX **) &ctx->aes_ctx);
}
void cryptolight_hash_variant(void* output, const void* input, int len, int variant)
{
struct cryptonight_ctx *ctx = (struct cryptonight_ctx*)malloc(sizeof(struct cryptonight_ctx));
cryptolight_hash_ctx(output, input, len, ctx, variant);
free(ctx);
}
void cryptolight_hash(void* output, const void* input)
{
cryptolight_hash_variant(output, input, 76, 1);
}
+176
View File
@@ -0,0 +1,176 @@
#include "cryptolight.h"
extern char *device_config[MAX_GPUS]; // -l 32x16
static __thread uint32_t cn_blocks = 32;
static __thread uint32_t cn_threads = 16;
static uint32_t *d_long_state[MAX_GPUS];
static uint32_t *d_ctx_state[MAX_GPUS];
static uint32_t *d_ctx_key1[MAX_GPUS];
static uint32_t *d_ctx_key2[MAX_GPUS];
static uint32_t *d_ctx_text[MAX_GPUS];
static uint64_t *d_ctx_tweak[MAX_GPUS];
static uint32_t *d_ctx_a[MAX_GPUS];
static uint32_t *d_ctx_b[MAX_GPUS];
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_cryptolight(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done, int variant)
{
int res = 0;
uint32_t throughput = 0;
uint32_t *ptarget = work->target;
uint8_t *pdata = (uint8_t*) work->data;
uint32_t *nonceptr = (uint32_t*) (&pdata[39]);
const uint32_t first_nonce = *nonceptr;
uint32_t nonce = first_nonce;
int dev_id = device_map[thr_id];
if(opt_benchmark) {
ptarget[7] = 0x00ff;
}
if(!init[thr_id])
{
if (!device_config[thr_id] && strcmp(device_name[dev_id], "TITAN V") == 0) {
device_config[thr_id] = strdup("80x32");
}
if (device_config[thr_id]) {
sscanf(device_config[thr_id], "%ux%u", &cn_blocks, &cn_threads);
throughput = cuda_default_throughput(thr_id, cn_blocks*cn_threads);
gpulog(LOG_INFO, thr_id, "Using %u x %u kernel launch config, %u threads",
cn_blocks, cn_threads, throughput);
} else {
throughput = cuda_default_throughput(thr_id, cn_blocks*cn_threads);
if (throughput != cn_blocks*cn_threads && cn_threads) {
cn_blocks = throughput / cn_threads;
throughput = cn_threads * cn_blocks;
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u threads (%ux%u)",
throughput2intensity(throughput), throughput, cn_blocks, cn_threads);
}
if(sizeof(size_t) == 4 && throughput > UINT32_MAX / MEMORY) {
gpulog(LOG_ERR, thr_id, "THE 32bit VERSION CAN'T ALLOCATE MORE THAN 4GB OF MEMORY!");
gpulog(LOG_ERR, thr_id, "PLEASE REDUCE THE NUMBER OF THREADS OR BLOCKS");
exit(1);
}
cudaSetDevice(device_map[thr_id]);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
const size_t alloc = MEMORY * throughput;
cryptonight_extra_init(thr_id);
cudaMalloc(&d_long_state[thr_id], alloc);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_state[thr_id], 25 * sizeof(uint64_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_key1[thr_id], 40 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_key2[thr_id], 40 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_text[thr_id], 32 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_a[thr_id], 4 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_b[thr_id], 4 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_tweak[thr_id], sizeof(uint64_t) * throughput);
init[thr_id] = true;
}
throughput = cn_blocks*cn_threads;
do
{
const uint32_t Htarg = ptarget[7];
uint32_t resNonces[2] = { UINT32_MAX, UINT32_MAX };
cryptonight_extra_setData(thr_id, pdata, ptarget);
cryptonight_extra_prepare(thr_id, throughput, nonce, d_ctx_state[thr_id], d_ctx_a[thr_id], d_ctx_b[thr_id], d_ctx_key1[thr_id], d_ctx_key2[thr_id], variant, d_ctx_tweak[thr_id]);
cryptolight_core_hash(thr_id, cn_blocks, cn_threads, d_long_state[thr_id], d_ctx_state[thr_id], d_ctx_a[thr_id], d_ctx_b[thr_id], d_ctx_key1[thr_id], d_ctx_key2[thr_id], variant, d_ctx_tweak[thr_id]);
cryptonight_extra_final(thr_id, throughput, nonce, resNonces, d_ctx_state[thr_id]);
*hashes_done = nonce - first_nonce + throughput;
if(resNonces[0] != UINT32_MAX)
{
uint32_t vhash[8];
uint32_t tempdata[19];
uint32_t *tempnonceptr = (uint32_t*)(((char*)tempdata) + 39);
memcpy(tempdata, pdata, 76);
*tempnonceptr = resNonces[0];
cryptolight_hash_variant(vhash, tempdata, 76, variant);
if(vhash[7] <= Htarg && fulltest(vhash, ptarget))
{
res = 1;
work->nonces[0] = resNonces[0];
work_set_target_ratio(work, vhash);
// second nonce
if(resNonces[1] != UINT32_MAX)
{
*tempnonceptr = resNonces[1];
cryptolight_hash_variant(vhash, tempdata, 76, variant);
if(vhash[7] <= Htarg && fulltest(vhash, ptarget)) {
res++;
work->nonces[1] = resNonces[1];
} else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
}
}
goto done;
} else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for nonce %08x does not validate on CPU!", resNonces[0]);
}
}
if ((uint64_t) throughput + nonce >= max_nonce - 127) {
nonce = max_nonce;
break;
}
nonce += throughput;
gpulog(LOG_DEBUG, thr_id, "nonce %08x", nonce);
} while (!work_restart[thr_id].restart && max_nonce > (uint64_t)throughput + nonce);
done:
gpulog(LOG_DEBUG, thr_id, "nonce %08x exit", nonce);
work->valid_nonces = res;
*nonceptr = nonce;
return res;
}
void free_cryptolight(int thr_id)
{
if (!init[thr_id])
return;
cudaFree(d_long_state[thr_id]);
cudaFree(d_ctx_state[thr_id]);
cudaFree(d_ctx_key1[thr_id]);
cudaFree(d_ctx_key2[thr_id]);
cudaFree(d_ctx_text[thr_id]);
cudaFree(d_ctx_tweak[thr_id]);
cudaFree(d_ctx_a[thr_id]);
cudaFree(d_ctx_b[thr_id]);
cryptonight_extra_free(thr_id);
cudaDeviceSynchronize();
init[thr_id] = false;
}
+144
View File
@@ -0,0 +1,144 @@
#pragma once
#include <cuda_runtime.h>
#include <miner.h>
#ifdef __INTELLISENSE__
/* avoid red underlining */
#define __CUDA_ARCH__ 520
struct uint3 {
unsigned int x, y, z;
};
struct uint3 threadIdx;
struct uint3 blockIdx;
struct uint3 blockDim;
#define atomicExch(p,y) (*p) = y
#define __funnelshift_r(a,b,c) 1
#define __syncthreads()
#define __threadfence_block()
#define asm(x)
#define __shfl(a,b,c) 1
#define __umul64hi(a,b) a*b
#endif
#define MEMORY (1UL << 20) /* 1 MiB - 1048576 */
#define ITER (1UL << 19) /* 512k */
#define E2I_MASK1 0xFFFF0 /* MEMORY / AES_BLOCK_SIZE - 1 = 0xFFFF */
#define E2I_MASK2 0x3FFFC /* 0xFFFF0 >> 2 */
#define AES_BLOCK_SIZE 16
#define AES_KEY_SIZE 32
#define INIT_SIZE_BLK 8
#define INIT_SIZE_BYTE (INIT_SIZE_BLK * AES_BLOCK_SIZE) // 128 B
#define AES_RKEY_LEN 4
#define AES_COL_LEN 4
#define AES_ROUND_BASE 7
#ifndef HASH_SIZE
#define HASH_SIZE 32
#endif
#ifndef HASH_DATA_AREA
#define HASH_DATA_AREA 136
#endif
#define hi_dword(x) (x >> 32)
#define lo_dword(x) (x & 0xFFFFFFFF)
#define C32(x) ((uint32_t)(x ## U))
#define T32(x) ((x) & C32(0xFFFFFFFF))
#ifndef ROTL64
#if __CUDA_ARCH__ >= 350
__forceinline__ __device__ uint64_t cuda_ROTL64(const uint64_t value, const int offset) {
uint2 result;
if(offset >= 32) {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
} else {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
}
return __double_as_longlong(__hiloint2double(result.y, result.x));
}
#define ROTL64(x, n) (cuda_ROTL64(x, n))
#else
#define ROTL64(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
#endif
#endif
#ifndef ROTL32
#if __CUDA_ARCH__ < 350
#define ROTL32(x, n) T32(((x) << (n)) | ((x) >> (32 - (n))))
#else
#define ROTL32(x, n) __funnelshift_l( (x), (x), (n) )
#endif
#endif
#ifndef ROTR32
#if __CUDA_ARCH__ < 350
#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#else
#define ROTR32(x, n) __funnelshift_r( (x), (x), (n) )
#endif
#endif
#define MEMSET8(dst,what,cnt) { \
int i_memset8; \
uint64_t *out_memset8 = (uint64_t *)(dst); \
for( i_memset8 = 0; i_memset8 < cnt; i_memset8++ ) \
out_memset8[i_memset8] = (what); }
#define MEMSET4(dst,what,cnt) { \
int i_memset4; \
uint32_t *out_memset4 = (uint32_t *)(dst); \
for( i_memset4 = 0; i_memset4 < cnt; i_memset4++ ) \
out_memset4[i_memset4] = (what); }
#define MEMCPY8(dst,src,cnt) { \
int i_memcpy8; \
uint64_t *in_memcpy8 = (uint64_t *)(src); \
uint64_t *out_memcpy8 = (uint64_t *)(dst); \
for( i_memcpy8 = 0; i_memcpy8 < cnt; i_memcpy8++ ) \
out_memcpy8[i_memcpy8] = in_memcpy8[i_memcpy8]; }
#define MEMCPY4(dst,src,cnt) { \
int i_memcpy4; \
uint32_t *in_memcpy4 = (uint32_t *)(src); \
uint32_t *out_memcpy4 = (uint32_t *)(dst); \
for( i_memcpy4 = 0; i_memcpy4 < cnt; i_memcpy4++ ) \
out_memcpy4[i_memcpy4] = in_memcpy4[i_memcpy4]; }
#define XOR_BLOCKS_DST(x,y,z) { \
((uint64_t *)z)[0] = ((uint64_t *)(x))[0] ^ ((uint64_t *)(y))[0]; \
((uint64_t *)z)[1] = ((uint64_t *)(x))[1] ^ ((uint64_t *)(y))[1]; }
union hash_state {
uint8_t b[200];
uint64_t w[25];
};
union cn_slow_hash_state {
union hash_state hs;
struct {
uint8_t k[64];
uint8_t init[INIT_SIZE_BYTE];
};
};
static inline void exit_if_cudaerror(int thr_id, const char *src, int line)
{
cudaError_t err = cudaGetLastError();
if(err != cudaSuccess) {
gpulog(LOG_ERR, thr_id, "%s %s line %d", cudaGetErrorString(err), src, line);
exit(1);
}
}
void cryptolight_core_hash(int thr_id, int blocks, int threads, uint32_t *d_long_state, uint32_t *d_ctx_state, uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak);
void cryptonight_extra_setData(int thr_id, const void *data, const void *ptarget);
void cryptonight_extra_init(int thr_id/*, uint32_t threads*/);
void cryptonight_extra_free(int thr_id);
void cryptonight_extra_prepare(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_ctx_state, uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak);
void cryptonight_extra_final(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *resNonces, uint32_t *d_ctx_state);
+316
View File
@@ -0,0 +1,316 @@
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#include <sys/time.h>
#ifndef _WIN32
#include <unistd.h>
#endif
#include <cuda.h>
#include <cuda_runtime.h>
#if CUDA_VERSION >= 9000 && __CUDA_ARCH__ >= 300
#undef __shfl
#define __shfl(var, srcLane, width) __shfl_sync(0xFFFFFFFFu, var, srcLane, width)
#endif
#include "cryptonight.h"
#define LONG_SHL32 19 // 1<<19 (uint32_t* index)
#define LONG_SHL64 18 // 1<<18 (uint64_t* index)
#define LONG_LOOPS32 0x80000U
#include "cn_aes.cuh"
__global__
void cryptonight_gpu_phase1(const uint32_t threads, uint32_t * __restrict__ d_long_state,
uint32_t * __restrict__ ctx_state, uint32_t * __restrict__ ctx_key1)
{
__shared__ uint32_t sharedMemory[1024];
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x) >> 3;
if(thread < threads)
{
cn_aes_gpu_init(sharedMemory);
__syncthreads();
const uint32_t sub = (threadIdx.x & 0x7U) << 2;
uint32_t *longstate = &d_long_state[(thread << LONG_SHL32) + sub];
uint32_t __align__(8) key[40];
MEMCPY8(key, &ctx_key1[thread * 40U], 20);
uint32_t __align__(8) text[4];
MEMCPY8(text, &ctx_state[thread * 50U + sub + 16U], 2);
for(int i = 0; i < LONG_LOOPS32; i += 32)
{
cn_aes_pseudo_round_mut(sharedMemory, text, key);
MEMCPY8(&longstate[i], text, 2);
}
}
}
// --------------------------------------------------------------------------------------------------------------
__device__ __forceinline__ ulonglong2 cuda_mul128(const uint64_t multiplier, const uint64_t multiplicand)
{
ulonglong2 product;
product.x = __umul64hi(multiplier, multiplicand);
product.y = multiplier * multiplicand;
return product;
}
static __forceinline__ __device__ void operator += (ulonglong2 &a, const ulonglong2 b) {
a.x += b.x; a.y += b.y;
}
static __forceinline__ __device__ ulonglong2 operator ^ (const ulonglong2 &a, const ulonglong2 &b) {
return make_ulonglong2(a.x ^ b.x, a.y ^ b.y);
}
__device__ __forceinline__ void MUL_SUM_XOR_DST_0(const uint64_t m, uint4 &a, void* far_dst)
{
ulonglong2 d = AS_UL2(far_dst);
ulonglong2 p = cuda_mul128(m, d.x);
p += AS_UL2(&a);
AS_UL2(&a) = p ^ d;
AS_UL2(far_dst) = p;
}
__global__
#if __CUDA_ARCH__ >= 500
//__launch_bounds__(128,12) /* force 40 regs to allow -l ...x32 */
#endif
void cryptonight_gpu_phase2(const uint32_t threads, const uint16_t bfactor, const uint32_t partidx,
uint64_t * __restrict__ d_long_state, uint32_t * __restrict__ d_ctx_a, uint32_t * __restrict__ d_ctx_b)
{
__shared__ __align__(16) uint32_t sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
__syncthreads();
const uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
if (thread < threads)
{
const uint32_t batchsize = ITER >> (2 + bfactor);
const uint32_t start = partidx * batchsize;
const uint32_t end = start + batchsize;
void * ctx_a = (void*)(&d_ctx_a[thread << 2U]);
void * ctx_b = (void*)(&d_ctx_b[thread << 2U]);
uint4 A = AS_UINT4(ctx_a); // ld.global.u32.v4
uint4 B = AS_UINT4(ctx_b);
uint64_t * long_state = &d_long_state[thread << LONG_SHL64];
for (int i = start; i < end; i++) // end = 262144
{
uint4 C;
uint32_t j = (A.x & E2I_MASK) >> 3;
cn_aes_single_round_b((uint8_t*)sharedMemory, &long_state[j], A, &C);
AS_UINT4(&long_state[j]) = C ^ B; // st.global.u32.v4
MUL_SUM_XOR_DST_0((AS_UL2(&C)).x, A, &long_state[(C.x & E2I_MASK) >> 3]);
j = (A.x & E2I_MASK) >> 3;
cn_aes_single_round_b((uint8_t*)sharedMemory, &long_state[j], A, &B);
AS_UINT4(&long_state[j]) = C ^ B;
MUL_SUM_XOR_DST_0((AS_UL2(&B)).x, A, &long_state[(B.x & E2I_MASK) >> 3]);
}
if (bfactor) {
AS_UINT4(ctx_a) = A;
AS_UINT4(ctx_b) = B;
}
}
}
// --------------------------------------------------------------------------------------------------------------
__device__ __forceinline__ void store_variant1(uint64_t* long_state, uint4 Z)
{
const uint32_t tmp = (Z.z >> 24); // __byte_perm(src, 0, 0x7773);
const uint32_t index = (((tmp >> 3) & 6u) | (tmp & 1u)) << 1;
Z.z = (Z.z & 0x00ffffffu) | ((tmp ^ ((0x75310u >> index) & 0x30u)) << 24);
AS_UINT4(long_state) = Z;
}
__device__ __forceinline__ void store_variant2(uint64_t* long_state, uint4 Z)
{
const uint32_t tmp = (Z.z >> 24); // __byte_perm(src, 0, 0x7773);
const uint32_t index = (((tmp >> 4) & 6u) | (tmp & 1u)) << 1;
Z.z = (Z.z & 0x00ffffffu) | ((tmp ^ ((0x75312u >> index) & 0x30u)) << 24);
AS_UINT4(long_state) = Z;
}
__device__ __forceinline__ void MUL_SUM_XOR_DST_1(const uint64_t m, uint4 &a, void* far_dst, uint64_t tweak)
{
ulonglong2 d = AS_UL2(far_dst);
ulonglong2 p = cuda_mul128(m, d.x);
p += AS_UL2(&a);
AS_UL2(&a) = p ^ d;
p.y = p.y ^ tweak;
AS_UL2(far_dst) = p;
}
__global__
void monero_gpu_phase2(const uint32_t threads, const uint16_t bfactor, const uint32_t partidx,
uint64_t * __restrict__ d_long_state, uint32_t * __restrict__ d_ctx_a, uint32_t * __restrict__ d_ctx_b,
uint64_t * __restrict__ d_tweak)
{
__shared__ __align__(16) uint32_t sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
__syncthreads();
const uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
if (thread < threads)
{
const uint32_t batchsize = ITER >> (2 + bfactor);
const uint32_t start = partidx * batchsize;
const uint32_t end = start + batchsize;
uint64_t tweak = d_tweak[thread];
void * ctx_a = (void*)(&d_ctx_a[thread << 2]);
void * ctx_b = (void*)(&d_ctx_b[thread << 2]);
uint4 A = AS_UINT4(ctx_a); // ld.global.u32.v4
uint4 B = AS_UINT4(ctx_b);
uint64_t * long_state = &d_long_state[thread << LONG_SHL64];
for (int i = start; i < end; i++) // end = 262144
{
uint4 C;
uint32_t j = (A.x & E2I_MASK) >> 3;
cn_aes_single_round_b((uint8_t*)sharedMemory, &long_state[j], A, &C);
store_variant1(&long_state[j], C ^ B); // st.global
MUL_SUM_XOR_DST_1((AS_UL2(&C)).x, A, &long_state[(C.x & E2I_MASK) >> 3], tweak);
j = (A.x & E2I_MASK) >> 3;
cn_aes_single_round_b((uint8_t*)sharedMemory, &long_state[j], A, &B);
store_variant1(&long_state[j], C ^ B);
MUL_SUM_XOR_DST_1((AS_UL2(&B)).x, A, &long_state[(B.x & E2I_MASK) >> 3], tweak);
}
if (bfactor) {
AS_UINT4(ctx_a) = A;
AS_UINT4(ctx_b) = B;
}
}
}
// --------------------------------------------------------------------------------------------------------------
__global__
void stellite_gpu_phase2(const uint32_t threads, const uint16_t bfactor, const uint32_t partidx,
uint64_t * __restrict__ d_long_state, uint32_t * __restrict__ d_ctx_a, uint32_t * __restrict__ d_ctx_b,
uint64_t * __restrict__ d_tweak)
{
__shared__ __align__(16) uint32_t sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
__syncthreads();
const uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
if (thread < threads)
{
const uint32_t batchsize = ITER >> (2 + bfactor);
const uint32_t start = partidx * batchsize;
const uint32_t end = start + batchsize;
uint64_t tweak = d_tweak[thread];
void * ctx_a = (void*)(&d_ctx_a[thread << 2]);
void * ctx_b = (void*)(&d_ctx_b[thread << 2]);
uint4 A = AS_UINT4(ctx_a); // ld.global.u32.v4
uint4 B = AS_UINT4(ctx_b);
uint64_t * long_state = &d_long_state[thread << LONG_SHL64];
for (int i = start; i < end; i++) // end = 262144
{
uint4 C;
uint32_t j = (A.x & E2I_MASK) >> 3;
cn_aes_single_round_b((uint8_t*)sharedMemory, &long_state[j], A, &C);
store_variant2(&long_state[j], C ^ B); // st.global
MUL_SUM_XOR_DST_1((AS_UL2(&C)).x, A, &long_state[(C.x & E2I_MASK) >> 3], tweak);
j = (A.x & E2I_MASK) >> 3;
cn_aes_single_round_b((uint8_t*)sharedMemory, &long_state[j], A, &B);
store_variant2(&long_state[j], C ^ B);
MUL_SUM_XOR_DST_1((AS_UL2(&B)).x, A, &long_state[(B.x & E2I_MASK) >> 3], tweak);
}
if (bfactor) {
AS_UINT4(ctx_a) = A;
AS_UINT4(ctx_b) = B;
}
}
}
// --------------------------------------------------------------------------------------------------------------
__global__
void cryptonight_gpu_phase3(const uint32_t threads, const uint32_t * __restrict__ d_long_state,
uint32_t * __restrict__ d_ctx_state, const uint32_t * __restrict__ d_ctx_key2)
{
__shared__ uint32_t sharedMemory[1024];
cn_aes_gpu_init(sharedMemory);
__syncthreads();
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x) >> 3;
if(thread < threads)
{
const int sub = (threadIdx.x & 7) << 2;
const uint32_t *longstate = &d_long_state[(thread << LONG_SHL32) + sub];
uint32_t key[40], text[4];
MEMCPY8(key, d_ctx_key2 + thread * 40, 20);
MEMCPY8(text, d_ctx_state + thread * 50 + sub + 16, 2);
for(int i = 0; i < LONG_LOOPS32; i += 32)
{
#pragma unroll
for(int j = 0; j < 4; ++j)
text[j] ^= longstate[i + j];
cn_aes_pseudo_round_mut(sharedMemory, text, key);
}
MEMCPY8(d_ctx_state + thread * 50 + sub + 16, text, 2);
}
}
// --------------------------------------------------------------------------------------------------------------
extern int device_bfactor[MAX_GPUS];
__host__
void cryptonight_core_cuda(int thr_id, uint32_t blocks, uint32_t threads, uint64_t *d_long_state, uint32_t *d_ctx_state,
uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak)
{
dim3 grid(blocks);
dim3 block(threads);
dim3 block4(threads << 2);
dim3 block8(threads << 3);
const uint16_t bfactor = (uint16_t) device_bfactor[thr_id];
const uint32_t partcount = 1U << bfactor;
const uint32_t throughput = (uint32_t) (blocks*threads);
const int bsleep = bfactor ? 100 : 0;
const int dev_id = device_map[thr_id];
cryptonight_gpu_phase1 <<<grid, block8>>> (throughput, (uint32_t*) d_long_state, d_ctx_state, d_ctx_key1);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
if(partcount > 1) usleep(bsleep);
for (uint32_t i = 0; i < partcount; i++)
{
dim3 b = device_sm[dev_id] >= 300 ? block4 : block;
if (variant == 0)
cryptonight_gpu_phase2 <<<grid, b>>> (throughput, bfactor, i, d_long_state, d_ctx_a, d_ctx_b);
else if (variant == 1 || cryptonight_fork == 8)
monero_gpu_phase2 <<<grid, b>>> (throughput, bfactor, i, d_long_state, d_ctx_a, d_ctx_b, d_ctx_tweak);
else if (variant == 2 && cryptonight_fork == 3)
stellite_gpu_phase2 <<<grid, b>>> (throughput, bfactor, i, d_long_state, d_ctx_a, d_ctx_b, d_ctx_tweak);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
if(partcount > 1) usleep(bsleep);
}
//cudaDeviceSynchronize();
//exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cryptonight_gpu_phase3 <<<grid, block8>>> (throughput, (uint32_t*) d_long_state, d_ctx_state, d_ctx_key2);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
}
+269
View File
@@ -0,0 +1,269 @@
#include <miner.h>
#include <memory.h>
#include "oaes_lib.h"
#include "cryptonight.h"
extern "C" {
#include <sph/sph_blake.h>
#include <sph/sph_groestl.h>
#include <sph/sph_jh.h>
#include <sph/sph_skein.h>
#include "cpu/c_keccak.h"
}
struct cryptonight_ctx {
uint8_t long_state[MEMORY];
union cn_slow_hash_state state;
uint8_t text[INIT_SIZE_BYTE];
uint8_t a[AES_BLOCK_SIZE];
uint8_t b[AES_BLOCK_SIZE];
uint8_t c[AES_BLOCK_SIZE];
oaes_ctx* aes_ctx;
};
static void do_blake_hash(const void* input, size_t len, void* output)
{
uchar hash[32];
sph_blake256_context ctx;
sph_blake256_set_rounds(14);
sph_blake256_init(&ctx);
sph_blake256(&ctx, input, len);
sph_blake256_close(&ctx, hash);
memcpy(output, hash, 32);
}
static void do_groestl_hash(const void* input, size_t len, void* output)
{
uchar hash[32];
sph_groestl256_context ctx;
sph_groestl256_init(&ctx);
sph_groestl256(&ctx, input, len);
sph_groestl256_close(&ctx, hash);
memcpy(output, hash, 32);
}
static void do_jh_hash(const void* input, size_t len, void* output)
{
uchar hash[64];
sph_jh256_context ctx;
sph_jh256_init(&ctx);
sph_jh256(&ctx, input, len);
sph_jh256_close(&ctx, hash);
memcpy(output, hash, 32);
}
static void do_skein_hash(const void* input, size_t len, void* output)
{
uchar hash[32];
sph_skein256_context ctx;
sph_skein256_init(&ctx);
sph_skein256(&ctx, input, len);
sph_skein256_close(&ctx, hash);
memcpy(output, hash, 32);
}
// todo: use sph if possible
static void keccak_hash_permutation(union hash_state *state) {
keccakf((uint64_t*)state, 24);
}
static void keccak_hash_process(union hash_state *state, const uint8_t *buf, size_t count) {
keccak1600(buf, (int)count, (uint8_t*)state);
}
extern "C" int fast_aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey);
extern "C" int aesb_single_round(const uint8_t *in, uint8_t*out, const uint8_t *expandedKey);
extern "C" int aesb_pseudo_round_mut(uint8_t *val, uint8_t *expandedKey);
extern "C" int fast_aesb_pseudo_round_mut(uint8_t *val, uint8_t *expandedKey);
static void (* const extra_hashes[4])(const void*, size_t, void *) = {
do_blake_hash, do_groestl_hash, do_jh_hash, do_skein_hash
};
static uint64_t mul128(uint64_t multiplier, uint64_t multiplicand, uint64_t* product_hi)
{
// multiplier = ab = a * 2^32 + b
// multiplicand = cd = c * 2^32 + d
// ab * cd = a * c * 2^64 + (a * d + b * c) * 2^32 + b * d
uint64_t a = hi_dword(multiplier);
uint64_t b = lo_dword(multiplier);
uint64_t c = hi_dword(multiplicand);
uint64_t d = lo_dword(multiplicand);
uint64_t ac = a * c;
uint64_t ad = a * d;
uint64_t bc = b * c;
uint64_t bd = b * d;
uint64_t adbc = ad + bc;
uint64_t adbc_carry = adbc < ad ? 1 : 0;
// multiplier * multiplicand = product_hi * 2^64 + product_lo
uint64_t product_lo = bd + (adbc << 32);
uint64_t product_lo_carry = product_lo < bd ? 1 : 0;
*product_hi = ac + (adbc >> 32) + (adbc_carry << 32) + product_lo_carry;
return product_lo;
}
static size_t e2i(const uint8_t* a) {
return (*((uint64_t*) a) / AES_BLOCK_SIZE) & (MEMORY / AES_BLOCK_SIZE - 1);
}
static void mul(const uint8_t* a, const uint8_t* b, uint8_t* res) {
((uint64_t*) res)[1] = mul128(((uint64_t*) a)[0], ((uint64_t*) b)[0], (uint64_t*) res);
}
static void sum_half_blocks(uint8_t* a, const uint8_t* b) {
((uint64_t*) a)[0] += ((uint64_t*) b)[0];
((uint64_t*) a)[1] += ((uint64_t*) b)[1];
}
static void sum_half_blocks_dst(const uint8_t* a, const uint8_t* b, uint8_t* dst) {
((uint64_t*) dst)[0] = ((uint64_t*) a)[0] + ((uint64_t*) b)[0];
((uint64_t*) dst)[1] = ((uint64_t*) a)[1] + ((uint64_t*) b)[1];
}
static void mul_sum_dst(const uint8_t* a, const uint8_t* b, const uint8_t* c, uint8_t* dst) {
((uint64_t*) dst)[1] = mul128(((uint64_t*) a)[0], ((uint64_t*) b)[0], (uint64_t*) dst) + ((uint64_t*) c)[1];
((uint64_t*) dst)[0] += ((uint64_t*) c)[0];
}
static void mul_sum_xor_dst(const uint8_t* a, uint8_t* c, uint8_t* dst, const int variant, const uint64_t tweak) {
uint64_t hi, lo = mul128(((uint64_t*) a)[0], ((uint64_t*) dst)[0], &hi) + ((uint64_t*) c)[1];
hi += ((uint64_t*) c)[0];
((uint64_t*) c)[0] = ((uint64_t*) dst)[0] ^ hi;
((uint64_t*) c)[1] = ((uint64_t*) dst)[1] ^ lo;
((uint64_t*) dst)[0] = hi;
((uint64_t*) dst)[1] = variant ? lo ^ tweak : lo;
}
static void copy_block(uint8_t* dst, const uint8_t* src) {
((uint64_t*) dst)[0] = ((uint64_t*) src)[0];
((uint64_t*) dst)[1] = ((uint64_t*) src)[1];
}
static void xor_blocks(uint8_t* a, const uint8_t* b) {
((uint64_t*) a)[0] ^= ((uint64_t*) b)[0];
((uint64_t*) a)[1] ^= ((uint64_t*) b)[1];
}
static void xor_blocks_dst(const uint8_t* a, const uint8_t* b, uint8_t* dst) {
((uint64_t*) dst)[0] = ((uint64_t*) a)[0] ^ ((uint64_t*) b)[0];
((uint64_t*) dst)[1] = ((uint64_t*) a)[1] ^ ((uint64_t*) b)[1];
}
static void cryptonight_store_variant(void* state, int variant) {
if (variant == 1 || cryptonight_fork == 8) {
// monero and graft
const uint8_t tmp = ((const uint8_t*)(state))[11];
const uint8_t index = (((tmp >> 3) & 6) | (tmp & 1)) << 1;
((uint8_t*)(state))[11] = tmp ^ ((0x75310 >> index) & 0x30);
} else if (variant == 2 && cryptonight_fork == 3) {
// stellite
const uint8_t tmp = ((const uint8_t*)(state))[11];
const uint8_t index = (((tmp >> 4) & 6) | (tmp & 1)) << 1;
((uint8_t*)(state))[11] = tmp ^ ((0x75312 >> index) & 0x30);
}
}
static void cryptonight_hash_ctx(void* output, const void* input, const size_t len, struct cryptonight_ctx* ctx, const int variant)
{
size_t i, j;
keccak_hash_process(&ctx->state.hs, (const uint8_t*) input, len);
ctx->aes_ctx = (oaes_ctx*) oaes_alloc();
memcpy(ctx->text, ctx->state.init, INIT_SIZE_BYTE);
const uint64_t tweak = variant ? *((uint64_t*) (((uint8_t*)input) + 35)) ^ ctx->state.hs.w[24] : 0;
oaes_key_import_data(ctx->aes_ctx, ctx->state.hs.b, AES_KEY_SIZE);
for (i = 0; likely(i < MEMORY); i += INIT_SIZE_BYTE) {
#undef RND
#define RND(p) aesb_pseudo_round_mut(&ctx->text[AES_BLOCK_SIZE * p], ctx->aes_ctx->key->exp_data);
RND(0);
RND(1);
RND(2);
RND(3);
RND(4);
RND(5);
RND(6);
RND(7);
memcpy(&ctx->long_state[i], ctx->text, INIT_SIZE_BYTE);
}
xor_blocks_dst(&ctx->state.k[0], &ctx->state.k[32], ctx->a);
xor_blocks_dst(&ctx->state.k[16], &ctx->state.k[48], ctx->b);
for (i = 0; likely(i < ITER / 4); ++i) {
j = e2i(ctx->a) * AES_BLOCK_SIZE;
aesb_single_round(&ctx->long_state[j], ctx->c, ctx->a);
xor_blocks_dst(ctx->c, ctx->b, &ctx->long_state[j]);
cryptonight_store_variant(&ctx->long_state[j], variant);
mul_sum_xor_dst(ctx->c, ctx->a, &ctx->long_state[e2i(ctx->c) * AES_BLOCK_SIZE], variant, tweak);
j = e2i(ctx->a) * AES_BLOCK_SIZE;
aesb_single_round(&ctx->long_state[j], ctx->b, ctx->a);
xor_blocks_dst(ctx->b, ctx->c, &ctx->long_state[j]);
cryptonight_store_variant(&ctx->long_state[j], variant);
mul_sum_xor_dst(ctx->b, ctx->a, &ctx->long_state[e2i(ctx->b) * AES_BLOCK_SIZE], variant, tweak);
}
memcpy(ctx->text, ctx->state.init, INIT_SIZE_BYTE);
oaes_key_import_data(ctx->aes_ctx, &ctx->state.hs.b[32], AES_KEY_SIZE);
for (i = 0; likely(i < MEMORY); i += INIT_SIZE_BYTE) {
#undef RND
#define RND(p) xor_blocks(&ctx->text[p * AES_BLOCK_SIZE], &ctx->long_state[i + p * AES_BLOCK_SIZE]); \
aesb_pseudo_round_mut(&ctx->text[p * AES_BLOCK_SIZE], ctx->aes_ctx->key->exp_data);
RND(0);
RND(1);
RND(2);
RND(3);
RND(4);
RND(5);
RND(6);
RND(7);
}
memcpy(ctx->state.init, ctx->text, INIT_SIZE_BYTE);
keccak_hash_permutation(&ctx->state.hs);
int extra_algo = ctx->state.hs.b[0] & 3;
extra_hashes[extra_algo](&ctx->state, 200, output);
if (opt_debug) applog(LOG_DEBUG, "extra algo=%d", extra_algo);
oaes_free((OAES_CTX **) &ctx->aes_ctx);
}
void cryptonight_hash_variant(void* output, const void* input, size_t len, int variant)
{
struct cryptonight_ctx *ctx = (struct cryptonight_ctx*)malloc(sizeof(struct cryptonight_ctx));
cryptonight_hash_ctx(output, input, len, ctx, variant);
free(ctx);
}
void cryptonight_hash(void* output, const void* input)
{
cryptonight_fork = 1;
cryptonight_hash_variant(output, input, 76, 0);
}
void graft_hash(void* output, const void* input)
{
cryptonight_fork = 8;
cryptonight_hash_variant(output, input, 76, 1);
}
void monero_hash(void* output, const void* input)
{
cryptonight_fork = 7;
cryptonight_hash_variant(output, input, 76, 1);
}
void stellite_hash(void* output, const void* input)
{
cryptonight_fork = 3;
cryptonight_hash_variant(output, input, 76, 2);
}
+212
View File
@@ -0,0 +1,212 @@
#include <stdio.h>
#include <stdint.h>
#include <string.h>
#ifndef _WIN32
#include <unistd.h>
#endif
#include <miner.h>
#include <cuda_helper.h>
#include "cryptonight.h"
static uint32_t *d_input[MAX_GPUS];
static uint32_t *d_target[MAX_GPUS];
static uint32_t *d_result[MAX_GPUS];
typedef uint8_t BitSequence;
typedef uint32_t DataLength;
#include "cn_keccak.cuh"
#include "cn_blake.cuh"
#include "cn_groestl.cuh"
#include "cn_jh.cuh"
#include "cn_skein.cuh"
__constant__ uint8_t d_sub_byte[16][16] = {
{0x63, 0x7c, 0x77, 0x7b, 0xf2, 0x6b, 0x6f, 0xc5, 0x30, 0x01, 0x67, 0x2b, 0xfe, 0xd7, 0xab, 0x76},
{0xca, 0x82, 0xc9, 0x7d, 0xfa, 0x59, 0x47, 0xf0, 0xad, 0xd4, 0xa2, 0xaf, 0x9c, 0xa4, 0x72, 0xc0},
{0xb7, 0xfd, 0x93, 0x26, 0x36, 0x3f, 0xf7, 0xcc, 0x34, 0xa5, 0xe5, 0xf1, 0x71, 0xd8, 0x31, 0x15},
{0x04, 0xc7, 0x23, 0xc3, 0x18, 0x96, 0x05, 0x9a, 0x07, 0x12, 0x80, 0xe2, 0xeb, 0x27, 0xb2, 0x75},
{0x09, 0x83, 0x2c, 0x1a, 0x1b, 0x6e, 0x5a, 0xa0, 0x52, 0x3b, 0xd6, 0xb3, 0x29, 0xe3, 0x2f, 0x84},
{0x53, 0xd1, 0x00, 0xed, 0x20, 0xfc, 0xb1, 0x5b, 0x6a, 0xcb, 0xbe, 0x39, 0x4a, 0x4c, 0x58, 0xcf},
{0xd0, 0xef, 0xaa, 0xfb, 0x43, 0x4d, 0x33, 0x85, 0x45, 0xf9, 0x02, 0x7f, 0x50, 0x3c, 0x9f, 0xa8},
{0x51, 0xa3, 0x40, 0x8f, 0x92, 0x9d, 0x38, 0xf5, 0xbc, 0xb6, 0xda, 0x21, 0x10, 0xff, 0xf3, 0xd2},
{0xcd, 0x0c, 0x13, 0xec, 0x5f, 0x97, 0x44, 0x17, 0xc4, 0xa7, 0x7e, 0x3d, 0x64, 0x5d, 0x19, 0x73},
{0x60, 0x81, 0x4f, 0xdc, 0x22, 0x2a, 0x90, 0x88, 0x46, 0xee, 0xb8, 0x14, 0xde, 0x5e, 0x0b, 0xdb},
{0xe0, 0x32, 0x3a, 0x0a, 0x49, 0x06, 0x24, 0x5c, 0xc2, 0xd3, 0xac, 0x62, 0x91, 0x95, 0xe4, 0x79},
{0xe7, 0xc8, 0x37, 0x6d, 0x8d, 0xd5, 0x4e, 0xa9, 0x6c, 0x56, 0xf4, 0xea, 0x65, 0x7a, 0xae, 0x08},
{0xba, 0x78, 0x25, 0x2e, 0x1c, 0xa6, 0xb4, 0xc6, 0xe8, 0xdd, 0x74, 0x1f, 0x4b, 0xbd, 0x8b, 0x8a},
{0x70, 0x3e, 0xb5, 0x66, 0x48, 0x03, 0xf6, 0x0e, 0x61, 0x35, 0x57, 0xb9, 0x86, 0xc1, 0x1d, 0x9e},
{0xe1, 0xf8, 0x98, 0x11, 0x69, 0xd9, 0x8e, 0x94, 0x9b, 0x1e, 0x87, 0xe9, 0xce, 0x55, 0x28, 0xdf},
{0x8c, 0xa1, 0x89, 0x0d, 0xbf, 0xe6, 0x42, 0x68, 0x41, 0x99, 0x2d, 0x0f, 0xb0, 0x54, 0xbb, 0x16}
};
__device__ __forceinline__
void cryptonight_aes_set_key(uint32_t * __restrict__ key, const uint32_t * __restrict__ data)
{
const uint32_t aes_gf[10] = {
0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1b, 0x36
};
MEMCPY4(key, data, 8);
#pragma unroll
for(int i = 8; i < 40; i++)
{
uint8_t temp[4];
*(uint32_t *)temp = key[i - 1];
if(i % 8 == 0) {
*(uint32_t *)temp = ROTR32(*(uint32_t *)temp, 8);
for(int j = 0; j < 4; j++)
temp[j] = d_sub_byte[(temp[j] >> 4) & 0x0f][temp[j] & 0x0f];
*(uint32_t *)temp ^= aes_gf[i / 8 - 1];
}
else if(i % 8 == 4) {
#pragma unroll
for(int j = 0; j < 4; j++)
temp[j] = d_sub_byte[(temp[j] >> 4) & 0x0f][temp[j] & 0x0f];
}
key[i] = key[(i - 8)] ^ *(uint32_t *)temp;
}
}
__global__
void cryptonight_extra_gpu_prepare(const uint32_t threads, const uint32_t * __restrict__ d_input, uint32_t startNonce,
uint32_t * __restrict__ d_ctx_state, uint32_t * __restrict__ d_ctx_a, uint32_t * __restrict__ d_ctx_b,
uint32_t * __restrict__ d_ctx_key1, uint32_t * __restrict__ d_ctx_key2, int variant, uint64_t * d_ctx_tweak)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if(thread < threads)
{
uint64_t ctx_state[25];
uint32_t ctx_a[4];
uint32_t ctx_b[4];
uint32_t ctx_key1[40];
uint32_t ctx_key2[40];
uint32_t input[19];
MEMCPY4(input, d_input, 19);
uint32_t nonce = startNonce + thread;
*(((uint8_t *)input) + 39) = nonce & 0xff;
*(((uint8_t *)input) + 40) = (nonce >> 8) & 0xff;
*(((uint8_t *)input) + 41) = (nonce >> 16) & 0xff;
*(((uint8_t *)input) + 42) = (nonce >> 24) & 0xff;
cn_keccak(input, ctx_state);
MEMCPY4(&d_ctx_state[thread * 50U], ctx_state, 50);
cryptonight_aes_set_key(ctx_key1, (uint32_t*)(&ctx_state[0]));
cryptonight_aes_set_key(ctx_key2, (uint32_t*)(&ctx_state[4]));
MEMCPY4(&d_ctx_key1[thread * 40U], ctx_key1, 40);
MEMCPY4(&d_ctx_key2[thread * 40U], ctx_key2, 40);
XOR_BLOCKS_DST(&ctx_state[0], &ctx_state[4], ctx_a);
XOR_BLOCKS_DST(&ctx_state[2], &ctx_state[6], ctx_b);
MEMCPY4(&d_ctx_a[thread * 4U], ctx_a, 4);
MEMCPY4(&d_ctx_b[thread * 4U], ctx_b, 4);
if (variant) {
uint2 tweak = AS_UINT2(&ctx_state[24]);
//tweak.x ^= (input[8] >> 24) | (input[9] << 8);
tweak.x ^= __byte_perm(input[8], input[ 9], 0x6543);
tweak.y ^= __byte_perm(input[9], input[10], 0x6543);
MEMCPY4(&d_ctx_tweak[thread], &tweak, 2);
}
}
}
__global__
void cryptonight_extra_gpu_final(const uint32_t threads, uint32_t startNonce, const uint32_t * __restrict__ d_target,
uint32_t * __restrict__ resNonces, uint32_t * __restrict__ d_ctx_state)
{
const uint32_t thread = blockDim.x * blockIdx.x + threadIdx.x;
if(thread < threads)
{
uint32_t *ctx_state = &d_ctx_state[thread * 50U];
uint32_t hash[8];
uint32_t state[50];
#pragma unroll 25
for(int i = 0; i < 50; i+=2)
AS_UINT2(&state[i]) = AS_UINT2(&ctx_state[i]);
cn_keccakf2((uint64_t *)state);
int branch = ((uint8_t *)state)[0] & 0x03;
if(branch == 0)
cn_blake((const uint8_t *)state, 200, hash);
if(branch == 1)
cn_groestl((const uint8_t *)state, 200, hash);
if(branch == 2)
cn_jh((const uint8_t *)state, 200, hash);
if(branch == 3)
cn_skein((const uint8_t *)state, 200, hash);
if(hash[7] <= d_target[1] && hash[6] <= d_target[0])
{
const uint32_t nonce = startNonce + thread;
uint32_t tmp = atomicExch(resNonces, nonce);
if(tmp != UINT32_MAX)
resNonces[1] = tmp;
}
}
}
__host__
void cryptonight_extra_setData(int thr_id, const void *data, const void *ptarget)
{
uint32_t *pTargetIn = (uint32_t*) ptarget;
cudaMemcpy(d_input[thr_id], data, 20 * sizeof(uint32_t), cudaMemcpyHostToDevice);
cudaMemcpy(d_target[thr_id], &pTargetIn[6], 2 * sizeof(uint32_t), cudaMemcpyHostToDevice);
cudaMemset(d_result[thr_id], 0xFF, 2 * sizeof(uint32_t));
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
}
__host__
void cryptonight_extra_init(int thr_id)
{
cudaMalloc(&d_input[thr_id], 20 * sizeof(uint32_t));
cudaMalloc(&d_target[thr_id], 2 * sizeof(uint32_t));
cudaMalloc(&d_result[thr_id], 2 * sizeof(uint32_t));
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
}
__host__
void cryptonight_extra_prepare(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_ctx_state, uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak)
{
uint32_t threadsperblock = 128;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
cryptonight_extra_gpu_prepare <<<grid, block>>> (threads, d_input[thr_id], startNonce, d_ctx_state, d_ctx_a, d_ctx_b, d_ctx_key1, d_ctx_key2, variant, d_ctx_tweak);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
}
__host__
void cryptonight_extra_final(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *resNonces, uint32_t *d_ctx_state)
{
uint32_t threadsperblock = 128;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cryptonight_extra_gpu_final <<<grid, block>>> (threads, startNonce, d_target[thr_id], d_result[thr_id], d_ctx_state);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMemcpy(resNonces, d_result[thr_id], 2 * sizeof(uint32_t), cudaMemcpyDeviceToHost);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
}
__host__
void cryptonight_extra_free(int thr_id)
{
if (d_input[thr_id]) {
cudaFree(d_input[thr_id]);
cudaFree(d_target[thr_id]);
cudaFree(d_result[thr_id]);
d_input[thr_id] = NULL;
}
}
+196
View File
@@ -0,0 +1,196 @@
#include "cryptonight.h"
extern char *device_config[MAX_GPUS]; // -l 32x16
static __thread uint32_t cn_blocks;
static __thread uint32_t cn_threads;
// used for gpu intensity on algo init
static __thread bool gpu_init_shown = false;
#define gpulog_init(p,thr,fmt, ...) if (!gpu_init_shown) \
gpulog(p, thr, fmt, ##__VA_ARGS__)
static uint64_t *d_long_state[MAX_GPUS];
static uint32_t *d_ctx_state[MAX_GPUS];
static uint32_t *d_ctx_key1[MAX_GPUS];
static uint32_t *d_ctx_key2[MAX_GPUS];
static uint32_t *d_ctx_text[MAX_GPUS];
static uint64_t *d_ctx_tweak[MAX_GPUS];
static uint32_t *d_ctx_a[MAX_GPUS];
static uint32_t *d_ctx_b[MAX_GPUS];
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_cryptonight(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done, int variant)
{
int res = 0;
uint32_t throughput = 0;
uint32_t *ptarget = work->target;
uint8_t *pdata = (uint8_t*) work->data;
uint32_t *nonceptr = (uint32_t*) (&pdata[39]);
const uint32_t first_nonce = *nonceptr;
uint32_t nonce = first_nonce;
int dev_id = device_map[thr_id];
if(opt_benchmark) {
ptarget[7] = 0x00ff;
}
if(!init[thr_id])
{
int mem = cuda_available_memory(thr_id);
int mul = device_sm[dev_id] >= 300 ? 4 : 1; // see cryptonight-core.cu
cn_threads = device_sm[dev_id] >= 600 ? 16 : 8; // real TPB is x4 on SM3+
cn_blocks = device_mpcount[dev_id] * 4;
if (cn_blocks*cn_threads*2.2 > mem) cn_blocks = device_mpcount[dev_id] * 2;
if (!opt_quiet)
gpulog_init(LOG_INFO, thr_id, "%s, %d MB available, %hd SMX", device_name[dev_id],
mem, device_mpcount[dev_id]);
if (!device_config[thr_id]) {
if(strcmp(device_name[dev_id], "TITAN V") == 0)
device_config[thr_id] = strdup("80x24");
if(strstr(device_name[dev_id], "V100"))
device_config[thr_id] = strdup("80x24");
}
if (device_config[thr_id]) {
int res = sscanf(device_config[thr_id], "%ux%u", &cn_blocks, &cn_threads);
throughput = cuda_default_throughput(thr_id, cn_blocks*cn_threads);
gpulog_init(LOG_INFO, thr_id, "Using %ux%u(x%d) kernel launch config, %u threads",
cn_blocks, cn_threads, mul, throughput);
} else {
throughput = cuda_default_throughput(thr_id, cn_blocks*cn_threads);
if (throughput != cn_blocks*cn_threads && cn_threads) {
cn_blocks = throughput / cn_threads;
throughput = cn_threads * cn_blocks;
}
gpulog_init(LOG_INFO, thr_id, "%u threads (%g) with %u blocks",// of %ux%d",
throughput, throughput2intensity(throughput), cn_blocks);//, cn_threads, mul);
}
if(sizeof(size_t) == 4 && throughput > UINT32_MAX / MEMORY) {
gpulog(LOG_ERR, thr_id, "THE 32bit VERSION CAN'T ALLOCATE MORE THAN 4GB OF MEMORY!");
gpulog(LOG_ERR, thr_id, "PLEASE REDUCE THE NUMBER OF THREADS OR BLOCKS");
exit(1);
}
cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
const size_t alloc = MEMORY * size_t(throughput);
cryptonight_extra_init(thr_id);
cudaMalloc(&d_long_state[thr_id], alloc);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_state[thr_id], 50 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_key1[thr_id], 40 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_key2[thr_id], 40 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_text[thr_id], 32 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_a[thr_id], 4 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_b[thr_id], 4 * sizeof(uint32_t) * throughput);
exit_if_cudaerror(thr_id, __FUNCTION__, __LINE__);
cudaMalloc(&d_ctx_tweak[thr_id], sizeof(uint64_t) * throughput);
exit_if_cudaerror(thr_id, __FILE__, __LINE__);
gpu_init_shown = true;
init[thr_id] = true;
}
throughput = cn_blocks*cn_threads;
do
{
const uint32_t Htarg = ptarget[7];
uint32_t resNonces[2] = { UINT32_MAX, UINT32_MAX };
cryptonight_extra_setData(thr_id, pdata, ptarget);
cryptonight_extra_prepare(thr_id, throughput, nonce, d_ctx_state[thr_id], d_ctx_a[thr_id], d_ctx_b[thr_id], d_ctx_key1[thr_id], d_ctx_key2[thr_id], variant, d_ctx_tweak[thr_id]);
cryptonight_core_cuda(thr_id, cn_blocks, cn_threads, d_long_state[thr_id], d_ctx_state[thr_id], d_ctx_a[thr_id], d_ctx_b[thr_id], d_ctx_key1[thr_id], d_ctx_key2[thr_id], variant, d_ctx_tweak[thr_id]);
cryptonight_extra_final(thr_id, throughput, nonce, resNonces, d_ctx_state[thr_id]);
*hashes_done = nonce - first_nonce + throughput;
if(resNonces[0] != UINT32_MAX)
{
uint32_t vhash[8];
uint32_t tempdata[19];
uint32_t *tempnonceptr = (uint32_t*)(((char*)tempdata) + 39);
memcpy(tempdata, pdata, 76);
*tempnonceptr = resNonces[0];
cryptonight_hash_variant(vhash, tempdata, 76, variant);
if(vhash[7] <= Htarg && fulltest(vhash, ptarget))
{
res = 1;
work->nonces[0] = resNonces[0];
work_set_target_ratio(work, vhash);
// second nonce
if(resNonces[1] != UINT32_MAX)
{
*tempnonceptr = resNonces[1];
cryptonight_hash_variant(vhash, tempdata, 76, variant);
if(vhash[7] <= Htarg && fulltest(vhash, ptarget)) {
res++;
work->nonces[1] = resNonces[1];
} else {
gpu_increment_reject(thr_id);
}
}
goto done;
} else if (vhash[7] > Htarg) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for nonce %08x does not validate on CPU!", resNonces[0]);
}
}
if ((uint64_t) throughput + nonce >= max_nonce - 127) {
nonce = max_nonce;
break;
}
nonce += throughput;
gpulog(LOG_DEBUG, thr_id, "nonce %08x", nonce);
} while (!work_restart[thr_id].restart && max_nonce > (uint64_t)throughput + nonce);
done:
gpulog(LOG_DEBUG, thr_id, "nonce %08x exit", nonce);
work->valid_nonces = res;
*nonceptr = nonce;
return res;
}
void free_cryptonight(int thr_id)
{
if (!init[thr_id])
return;
cudaFree(d_long_state[thr_id]);
cudaFree(d_ctx_state[thr_id]);
cudaFree(d_ctx_key1[thr_id]);
cudaFree(d_ctx_key2[thr_id]);
cudaFree(d_ctx_text[thr_id]);
cudaFree(d_ctx_tweak[thr_id]);
cudaFree(d_ctx_a[thr_id]);
cudaFree(d_ctx_b[thr_id]);
cryptonight_extra_free(thr_id);
cudaDeviceSynchronize();
init[thr_id] = false;
}
+144
View File
@@ -0,0 +1,144 @@
#pragma once
#include <cuda_runtime.h>
#include <miner.h>
#ifdef __INTELLISENSE__
/* avoid red underlining */
#define __CUDA_ARCH__ 520
struct uint3 {
unsigned int x, y, z;
};
struct uint3 threadIdx;
struct uint3 blockIdx;
struct uint3 blockDim;
#define atomicExch(p,y) (*p) = y
#define __funnelshift_r(a,b,c) 1
#define __syncthreads()
#define __threadfence_block()
#define asm(x)
#define __shfl(a,b,c) 1
#define __umul64hi(a,b) a*b
#endif
#define MEMORY (1U << 21) // 2 MiB / 2097152 B
#define ITER (1U << 20) // 1048576
#define E2I_MASK 0x1FFFF0u
#define AES_BLOCK_SIZE 16U
#define AES_KEY_SIZE 32
#define INIT_SIZE_BLK 8
#define INIT_SIZE_BYTE (INIT_SIZE_BLK * AES_BLOCK_SIZE) // 128 B
#define AES_RKEY_LEN 4
#define AES_COL_LEN 4
#define AES_ROUND_BASE 7
#ifndef HASH_SIZE
#define HASH_SIZE 32
#endif
#ifndef HASH_DATA_AREA
#define HASH_DATA_AREA 136
#endif
#define hi_dword(x) (x >> 32)
#define lo_dword(x) (x & 0xFFFFFFFF)
#define C32(x) ((uint32_t)(x ## U))
#define T32(x) ((x) & C32(0xFFFFFFFF))
#ifndef ROTL64
#if __CUDA_ARCH__ >= 350
__forceinline__ __device__ uint64_t cuda_ROTL64(const uint64_t value, const int offset) {
uint2 result;
if(offset >= 32) {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
} else {
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.x) : "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))), "r"(offset));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(result.y) : "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))), "r"(offset));
}
return __double_as_longlong(__hiloint2double(result.y, result.x));
}
#define ROTL64(x, n) (cuda_ROTL64(x, n))
#else
#define ROTL64(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
#endif
#endif
#ifndef ROTL32
#if __CUDA_ARCH__ < 350
#define ROTL32(x, n) T32(((x) << (n)) | ((x) >> (32 - (n))))
#else
#define ROTL32(x, n) __funnelshift_l( (x), (x), (n) )
#endif
#endif
#ifndef ROTR32
#if __CUDA_ARCH__ < 350
#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#else
#define ROTR32(x, n) __funnelshift_r( (x), (x), (n) )
#endif
#endif
#define MEMSET8(dst,what,cnt) { \
int i_memset8; \
uint64_t *out_memset8 = (uint64_t *)(dst); \
for( i_memset8 = 0; i_memset8 < cnt; i_memset8++ ) \
out_memset8[i_memset8] = (what); }
#define MEMSET4(dst,what,cnt) { \
int i_memset4; \
uint32_t *out_memset4 = (uint32_t *)(dst); \
for( i_memset4 = 0; i_memset4 < cnt; i_memset4++ ) \
out_memset4[i_memset4] = (what); }
#define MEMCPY8(dst,src,cnt) { \
int i_memcpy8; \
uint64_t *in_memcpy8 = (uint64_t *)(src); \
uint64_t *out_memcpy8 = (uint64_t *)(dst); \
for( i_memcpy8 = 0; i_memcpy8 < cnt; i_memcpy8++ ) \
out_memcpy8[i_memcpy8] = in_memcpy8[i_memcpy8]; }
#define MEMCPY4(dst,src,cnt) { \
int i_memcpy4; \
uint32_t *in_memcpy4 = (uint32_t *)(src); \
uint32_t *out_memcpy4 = (uint32_t *)(dst); \
for( i_memcpy4 = 0; i_memcpy4 < cnt; i_memcpy4++ ) \
out_memcpy4[i_memcpy4] = in_memcpy4[i_memcpy4]; }
#define XOR_BLOCKS_DST(x,y,z) { \
((uint64_t *)z)[0] = ((uint64_t *)(x))[0] ^ ((uint64_t *)(y))[0]; \
((uint64_t *)z)[1] = ((uint64_t *)(x))[1] ^ ((uint64_t *)(y))[1]; }
#define E2I(x) ((size_t)(((*((uint64_t*)(x)) >> 4) & 0x1ffff)))
union hash_state {
uint8_t b[200];
uint64_t w[25];
};
union cn_slow_hash_state {
union hash_state hs;
struct {
uint8_t k[64];
uint8_t init[INIT_SIZE_BYTE];
};
};
static inline void exit_if_cudaerror(int thr_id, const char *src, int line)
{
cudaError_t err = cudaGetLastError();
if(err != cudaSuccess) {
gpulog(LOG_ERR, thr_id, "%s %s line %d", cudaGetErrorString(err), src, line);
exit(1);
}
}
void cryptonight_core_cuda(int thr_id, uint32_t blocks, uint32_t threads, uint64_t *d_long_state, uint32_t *d_ctx_state, uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak);
void cryptonight_extra_setData(int thr_id, const void *data, const void *ptarget);
void cryptonight_extra_init(int thr_id);
void cryptonight_extra_free(int thr_id);
void cryptonight_extra_prepare(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_ctx_state, uint32_t *d_ctx_a, uint32_t *d_ctx_b, uint32_t *d_ctx_key1, uint32_t *d_ctx_key2, int variant, uint64_t *d_ctx_tweak);
void cryptonight_extra_final(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *resnonce, uint32_t *d_ctx_state);
+573
View File
@@ -0,0 +1,573 @@
#ifndef UINT128_C_H
#define UINT128_C_H
struct __uint128 {
uint64_t Hi;
uint64_t Lo;
};
typedef struct __uint128 uint128;
void Increment(uint128 * N)
{
uint64_t T = (N->Lo + 1);
N->Hi += ((N->Lo ^T) & N->Lo) >> 63;
N->Lo = T;
}
void Decrement(uint128 * N)
{
uint64_t T = (N->Lo - 1);
N->Hi -= ((T ^ N->Lo) & T) >> 63;
N->Lo = T;
}
void Add(uint128 * Ans, uint128 N, uint128 M)
{
uint64_t C = (((N.Lo & M.Lo) & 1) + (N.Lo >> 1) + (M.Lo >> 1)) >> 63;
Ans->Hi = N.Hi + M.Hi + C;
Ans->Lo = N.Lo + M.Lo;
}
void Subtract(uint128 * Ans, uint128 N, uint128 M)
{
Ans->Lo = N.Lo - M.Lo;
uint64_t C = (((Ans->Lo & M.Lo) & 1) + (M.Lo >> 1) + (Ans->Lo >> 1)) >> 63;
Ans->Hi = N.Hi - (M.Hi + C);
}
void inc128(uint128 N, uint128* A)
{
A->Lo = (N.Lo + 1);
A->Hi = N.Hi + (((N.Lo ^ A->Lo) & N.Lo) >> 63);
}
void dec128(uint128 N, uint128* A)
{
A->Lo = N.Lo - 1;
A->Hi = N.Hi - (((A->Lo ^ N.Lo) & A->Lo) >> 63);
}
void add128(uint128 N, uint128 M, uint128* A)
{
uint64_t C = (((N.Lo & M.Lo) & 1) + (N.Lo >> 1) + (M.Lo >> 1)) >> 63;
A->Hi = N.Hi + M.Hi + C;
A->Lo = N.Lo + M.Lo;
}
void sub128(uint128 N, uint128 M, uint128* A)
{
A->Lo = N.Lo - M.Lo;
uint64_t C = (((A->Lo & M.Lo) & 1) + (M.Lo >> 1) + (A->Lo >> 1)) >> 63;
A->Hi = N.Hi - (M.Hi + C);
}
void mult64to128(uint64_t u, uint64_t v, uint64_t * h, uint64_t *l)
{
uint64_t u1 = (u & 0xffffffff);
uint64_t v1 = (v & 0xffffffff);
uint64_t t = (u1 * v1);
uint64_t w3 = (t & 0xffffffff);
uint64_t k = (t >> 32);
u >>= 32;
t = (u * v1) + k;
k = (t & 0xffffffff);
uint64_t w1 = (t >> 32);
v >>= 32;
t = (u1 * v) + k;
k = (t >> 32);
*h = (u * v) + w1 + k;
*l = (t << 32) + w3;
}
void mult128(uint128 N, uint128 M, uint128 * Ans)
{
mult64to128(N.Lo, M.Lo, &Ans->Hi, &Ans->Lo);
Ans->Hi += (N.Hi * M.Lo) + (N.Lo * M.Hi);
}
void mult128to256(uint128 N, uint128 M, uint128 * H, uint128 * L)
{
mult64to128(N.Hi, M.Hi, &H->Hi, &H->Lo);
mult64to128(N.Lo, M.Lo, &L->Hi, &L->Lo);
uint128 T;
mult64to128(N.Hi, M.Lo, &T.Hi, &T.Lo);
L->Hi += T.Lo;
if(L->Hi < T.Lo) // if L->Hi overflowed
{
Increment(H);
}
H->Lo += T.Hi;
if(H->Lo < T.Hi) // if H->Lo overflowed
{
++H->Hi;
}
mult64to128(N.Lo, M.Hi, &T.Hi, &T.Lo);
L->Hi += T.Lo;
if(L->Hi < T.Lo) // if L->Hi overflowed
{
Increment(H);
}
H->Lo += T.Hi;
if(H->Lo < T.Hi) // if H->Lo overflowed
{
++H->Hi;
}
}
void sqr64to128(uint64_t r, uint64_t * h, uint64_t *l)
{
uint64_t r1 = (r & 0xffffffff);
uint64_t t = (r1 * r1);
uint64_t w3 = (t & 0xffffffff);
uint64_t k = (t >> 32);
r >>= 32;
uint64_t m = (r * r1);
t = m + k;
uint64_t w2 = (t & 0xffffffff);
uint64_t w1 = (t >> 32);
t = m + w2;
k = (t >> 32);
*h = (r * r) + w1 + k;
*l = (t << 32) + w3;
}
void sqr128(uint128 R, uint128 * Ans)
{
sqr64to128(R.Lo, &Ans->Hi, &Ans->Lo);
Ans->Hi += (R.Hi * R.Lo) << 1;
}
void sqr128to256(uint128 R, uint128 * H, uint128 * L)
{
sqr64to128(R.Hi, &H->Hi, &H->Lo);
sqr64to128(R.Lo, &L->Hi, &L->Lo);
uint128 T;
mult64to128(R.Hi, R.Lo, &T.Hi, &T.Lo);
H->Hi += (T.Hi >> 63);
T.Hi = (T.Hi << 1) | (T.Lo >> 63); // Shift Left 1 bit
T.Lo <<= 1;
L->Hi += T.Lo;
if(L->Hi < T.Lo) // if L->Hi overflowed
{
Increment(H);
}
H->Lo += T.Hi;
if(H->Lo < T.Hi) // if H->Lo overflowed
{
++H->Hi;
}
}
void shiftleft128(uint128 N, size_t S, uint128 * A)
{
uint64_t M1, M2;
S &= 127;
M1 = ((((S + 127) | S) & 64) >> 6) - 1llu;
M2 = (S >> 6) - 1llu;
S &= 63;
A->Hi = (N.Lo << S) & (~M2);
A->Lo = (N.Lo << S) & M2;
A->Hi |= ((N.Hi << S) | ((N.Lo >> (64 - S)) & M1)) & M2;
/*
S &= 127;
if(S != 0)
{
if(S > 64)
{
A.Hi = N.Lo << (S - 64);
A.Lo = 0;
}
else if(S < 64)
{
A.Hi = (N.Hi << S) | (N.Lo >> (64 - S));
A.Lo = N.Lo << S;
}
else
{
A.Hi = N.Lo;
A.Lo = 0;
}
}
else
{
A.Hi = N.Hi;
A.Lo = N.Lo;
}
//*/
}
void shiftright128(uint128 N, size_t S, uint128 * A)
{
uint64_t M1, M2;
S &= 127;
M1 = ((((S + 127) | S) & 64) >> 6) - 1llu;
M2 = (S >> 6) - 1llu;
S &= 63;
A->Lo = (N.Hi >> S) & (~M2);
A->Hi = (N.Hi >> S) & M2;
A->Lo |= ((N.Lo >> S) | ((N.Hi << (64 - S)) & M1)) & M2;
/*
S &= 127;
if(S != 0)
{
if(S > 64)
{
A.Hi = N.Hi >> (S - 64);
A.Lo = 0;
}
else if(S < 64)
{
A.Lo = (N.Lo >> S) | (N.Hi << (64 - S));
A.Hi = N.Hi >> S;
}
else
{
A.Lo = N.Hi;
A.Hi = 0;
}
}
else
{
A.Hi = N.Hi;
A.Lo = N.Lo;
}
//*/
}
void not128(uint128 N, uint128 * A)
{
A->Hi = ~N.Hi;
A->Lo = ~N.Lo;
}
void or128(uint128 N1, uint128 N2, uint128 * A)
{
A->Hi = N1.Hi | N2.Hi;
A->Lo = N1.Lo | N2.Lo;
}
void and128(uint128 N1, uint128 N2, uint128 * A)
{
A->Hi = N1.Hi & N2.Hi;
A->Lo = N1.Lo & N2.Lo;
}
void xor128(uint128 N1, uint128 N2, uint128 * A)
{
A->Hi = N1.Hi ^ N2.Hi;
A->Lo = N1.Lo ^ N2.Lo;
}
size_t nlz64(uint64_t N)
{
uint64_t I;
size_t C;
I = ~N;
C = ((I ^ (I + 1)) & I) >> 63;
I = (N >> 32) + 0xffffffff;
I = ((I & 0x100000000) ^ 0x100000000) >> 27;
C += I; N <<= I;
I = (N >> 48) + 0xffff;
I = ((I & 0x10000) ^ 0x10000) >> 12;
C += I; N <<= I;
I = (N >> 56) + 0xff;
I = ((I & 0x100) ^ 0x100) >> 5;
C += I; N <<= I;
I = (N >> 60) + 0xf;
I = ((I & 0x10) ^ 0x10) >> 2;
C += I; N <<= I;
I = (N >> 62) + 3;
I = ((I & 4) ^ 4) >> 1;
C += I; N <<= I;
C += (N >> 63) ^ 1;
return C;
}
size_t ntz64(uint64_t N)
{
uint64_t I = ~N;
size_t C = ((I ^ (I + 1)) & I) >> 63;
I = (N & 0xffffffff) + 0xffffffff;
I = ((I & 0x100000000) ^ 0x100000000) >> 27;
C += I; N >>= I;
I = (N & 0xffff) + 0xffff;
I = ((I & 0x10000) ^ 0x10000) >> 12;
C += I; N >>= I;
I = (N & 0xff) + 0xff;
I = ((I & 0x100) ^ 0x100) >> 5;
C += I; N >>= I;
I = (N & 0xf) + 0xf;
I = ((I & 0x10) ^ 0x10) >> 2;
C += I; N >>= I;
I = (N & 3) + 3;
I = ((I & 4) ^ 4) >> 1;
C += I; N >>= I;
C += ((N & 1) ^ 1);
return C;
}
size_t popcnt64(uint64_t V)
{
// http://graphics.stanford.edu/~seander/bithacks.html#CountBitsSetParallel
V -= ((V >> 1) & 0x5555555555555555);
V = (V & 0x3333333333333333) + ((V >> 2) & 0x3333333333333333);
return ((V + (V >> 4) & 0xF0F0F0F0F0F0F0F) * 0x101010101010101) >> 56;
}
size_t popcnt128(uint128 N)
{
return popcnt64(N.Hi) + popcnt64(N.Lo);
}
size_t nlz128(uint128 N)
{
return (N.Hi == 0) ? nlz64(N.Lo) + 64 : nlz64(N.Hi);
}
size_t ntz128(uint128 N)
{
return (N.Lo == 0) ? ntz64(N.Hi) + 64 : ntz64(N.Lo);
}
int compare128(uint128 N1, uint128 N2)
{
return (((N1.Hi > N2.Hi) || ((N1.Hi == N2.Hi) && (N1.Lo > N2.Lo))) ? 1 : 0)
- (((N1.Hi < N2.Hi) || ((N1.Hi == N2.Hi) && (N1.Lo < N2.Lo))) ? 1 : 0);
}
void bindivmod128(uint128 M, uint128 N, uint128 * Q, uint128 *R)
{
Q->Hi = Q->Lo = 0;
size_t Shift = nlz128(N) - nlz128(M);
shiftleft128(N, Shift, &N);
do
{
shiftleft128(*Q, (size_t)1, Q);
if(compare128(M, N) >= 0)
{
sub128(M, N, &M);
Q->Lo |= 1;
}
shiftright128(N, 1, &N);
}while(Shift-- != 0);
R->Hi = M.Hi;
R->Lo = M.Lo;
}
void divmod128by64(const uint64_t u1, const uint64_t u0, uint64_t v, uint64_t * q, uint64_t * r)
{
const uint64_t b = 1ll << 32;
uint64_t un1, un0, vn1, vn0, q1, q0, un32, un21, un10, rhat, left, right;
size_t s;
s = nlz64(v);
v <<= s;
vn1 = v >> 32;
vn0 = v & 0xffffffff;
if (s > 0)
{
un32 = (u1 << s) | (u0 >> (64 - s));
un10 = u0 << s;
}
else
{
un32 = u1;
un10 = u0;
}
un1 = un10 >> 32;
un0 = un10 & 0xffffffff;
q1 = un32 / vn1;
rhat = un32 % vn1;
left = q1 * vn0;
right = (rhat << 32) + un1;
again1:
if ((q1 >= b) || (left > right))
{
--q1;
rhat += vn1;
if (rhat < b)
{
left -= vn0;
right = (rhat << 32) | un1;
goto again1;
}
}
un21 = (un32 << 32) + (un1 - (q1 * v));
q0 = un21 / vn1;
rhat = un21 % vn1;
left = q0 * vn0;
right = (rhat << 32) | un0;
again2:
if ((q0 >= b) || (left > right))
{
--q0;
rhat += vn1;
if (rhat < b)
{
left -= vn0;
right = (rhat << 32) | un0;
goto again2;
}
}
*r = ((un21 << 32) + (un0 - (q0 * v))) >> s;
*q = (q1 << 32) | q0;
}
static void divmod128by128(uint128 M, uint128 N, uint128 * Q, uint128 * R)
{
if (N.Hi == 0)
{
if (M.Hi < N.Lo)
{
divmod128by64(M.Hi, M.Lo, N.Lo, &Q->Lo, &R->Lo);
Q->Hi = 0;
R->Hi = 0;
return;
}
else
{
Q->Hi = M.Hi / N.Lo;
R->Hi = M.Hi % N.Lo;
divmod128by64(R->Hi, M.Lo, N.Lo, &Q->Lo, &R->Lo);
R->Hi = 0;
return;
}
}
else
{
size_t n = nlz64(N.Hi);
uint128 v1;
shiftleft128(N, n, &v1);
uint128 u1;
shiftright128(M, 1, &u1);
uint128 q1;
divmod128by64(u1.Hi, u1.Lo, v1.Hi, &q1.Hi, &q1.Lo);
q1.Hi = 0;
shiftright128(q1, 63 - n, &q1);
if ((q1.Hi | q1.Lo) != 0)
{
dec128(q1, &q1);
}
Q->Hi = q1.Hi;
Q->Lo = q1.Lo;
mult128(q1, N, &q1);
sub128(M, q1, R);
if (compare128(*R, N) >= 0)
{
inc128(*Q, Q);
sub128(*R, N, R);
}
return;
}
}
void divmod128(uint128 M, uint128 N, uint128 * Q, uint128 * R)
{
size_t Nlz, Mlz, Ntz;
int C;
Nlz = nlz128(N);
Mlz = nlz128(M);
Ntz = ntz128(N);
if(Nlz == 128)
{
return;
}
else if((M.Hi | N.Hi) == 0)
{
Q->Hi = R->Hi = 0;
Q->Lo = M.Lo / N.Lo;
R->Lo = M.Lo % N.Lo;
return;
}
else if(Nlz == 127)
{
*Q = M;
R->Hi = R->Lo = 0;
return;
}
else if((Ntz + Nlz) == 127)
{
shiftright128(M, Ntz, Q);
dec128(N, &N);
and128(N, M, R);
return;
}
C = compare128(M, N);
if(C < 0)
{
Q->Hi = Q->Lo = 0;
*R = M;
return;
}
else if(C == 0)
{
Q->Hi = R->Hi = R->Lo = 0;
Q->Lo = 1;
return;
}
if((Nlz - Mlz) > 5)
{
divmod128by128(M, N, Q, R);
}
else
{
bindivmod128(M, N, Q, R);
}
}
#endif
+180
View File
@@ -0,0 +1,180 @@
#include <windows.h>
#include <errno.h>
#include <io.h>
#include "mman.h"
#ifndef FILE_MAP_EXECUTE
#define FILE_MAP_EXECUTE 0x0020
#endif /* FILE_MAP_EXECUTE */
static int __map_mman_error(const DWORD err, const int deferr)
{
if (err == 0)
return 0;
//TODO: implement
return err;
}
static DWORD __map_mmap_prot_page(const int prot)
{
DWORD protect = 0;
if (prot == PROT_NONE)
return protect;
if ((prot & PROT_EXEC) != 0)
{
protect = ((prot & PROT_WRITE) != 0) ?
PAGE_EXECUTE_READWRITE : PAGE_EXECUTE_READ;
}
else
{
protect = ((prot & PROT_WRITE) != 0) ?
PAGE_READWRITE : PAGE_READONLY;
}
return protect;
}
static DWORD __map_mmap_prot_file(const int prot)
{
DWORD desiredAccess = 0;
if (prot == PROT_NONE)
return desiredAccess;
if ((prot & PROT_READ) != 0)
desiredAccess |= FILE_MAP_READ;
if ((prot & PROT_WRITE) != 0)
desiredAccess |= FILE_MAP_WRITE;
if ((prot & PROT_EXEC) != 0)
desiredAccess |= FILE_MAP_EXECUTE;
return desiredAccess;
}
void* mmap(void *addr, size_t len, int prot, int flags, int fildes, OffsetType off)
{
HANDLE fm, h;
void * map = MAP_FAILED;
#ifdef _MSC_VER
#pragma warning(push)
#pragma warning(disable: 4293)
#endif
const DWORD dwFileOffsetLow = (sizeof(OffsetType) <= sizeof(DWORD)) ?
(DWORD)off : (DWORD)(off & 0xFFFFFFFFL);
const DWORD dwFileOffsetHigh = (sizeof(OffsetType) <= sizeof(DWORD)) ?
(DWORD)0 : (DWORD)((off >> 32) & 0xFFFFFFFFL);
const DWORD protect = __map_mmap_prot_page(prot);
const DWORD desiredAccess = __map_mmap_prot_file(prot);
const OffsetType maxSize = off + (OffsetType)len;
const DWORD dwMaxSizeLow = (sizeof(OffsetType) <= sizeof(DWORD)) ?
(DWORD)maxSize : (DWORD)(maxSize & 0xFFFFFFFFL);
const DWORD dwMaxSizeHigh = (sizeof(OffsetType) <= sizeof(DWORD)) ?
(DWORD)0 : (DWORD)((maxSize >> 32) & 0xFFFFFFFFL);
#ifdef _MSC_VER
#pragma warning(pop)
#endif
errno = 0;
if (len == 0
/* Unsupported flag combinations */
|| (flags & MAP_FIXED) != 0
/* Usupported protection combinations */
|| prot == PROT_EXEC)
{
errno = EINVAL;
return MAP_FAILED;
}
h = ((flags & MAP_ANONYMOUS) == 0) ?
(HANDLE)_get_osfhandle(fildes) : INVALID_HANDLE_VALUE;
if ((flags & MAP_ANONYMOUS) == 0 && h == INVALID_HANDLE_VALUE)
{
errno = EBADF;
return MAP_FAILED;
}
fm = CreateFileMapping(h, NULL, protect, dwMaxSizeHigh, dwMaxSizeLow, NULL);
if (fm == NULL)
{
errno = __map_mman_error(GetLastError(), EPERM);
return MAP_FAILED;
}
map = MapViewOfFile(fm, desiredAccess, dwFileOffsetHigh, dwFileOffsetLow, len);
CloseHandle(fm);
if (map == NULL)
{
errno = __map_mman_error(GetLastError(), EPERM);
return MAP_FAILED;
}
return map;
}
int munmap(void *addr, size_t len)
{
if (UnmapViewOfFile(addr))
return 0;
errno = __map_mman_error(GetLastError(), EPERM);
return -1;
}
int _mprotect(void *addr, size_t len, int prot)
{
DWORD newProtect = __map_mmap_prot_page(prot);
DWORD oldProtect = 0;
if (VirtualProtect(addr, len, newProtect, &oldProtect))
return 0;
errno = __map_mman_error(GetLastError(), EPERM);
return -1;
}
int msync(void *addr, size_t len, int flags)
{
if (FlushViewOfFile(addr, len))
return 0;
errno = __map_mman_error(GetLastError(), EPERM);
return -1;
}
int mlock(const void *addr, size_t len)
{
if (VirtualLock((LPVOID)addr, len))
return 0;
errno = __map_mman_error(GetLastError(), EPERM);
return -1;
}
int munlock(const void *addr, size_t len)
{
if (VirtualUnlock((LPVOID)addr, len))
return 0;
errno = __map_mman_error(GetLastError(), EPERM);
return -1;
}
+63
View File
@@ -0,0 +1,63 @@
/*
* sys/mman.h
* mman-win32
*/
#ifndef _SYS_MMAN_H_
#define _SYS_MMAN_H_
#ifndef _WIN32_WINNT // Allow use of features specific to Windows XP or later.
#define _WIN32_WINNT 0x0501 // Change this to the appropriate value to target other versions of Windows.
#endif
/* All the headers include this file. */
#ifndef _MSC_VER
#include <_mingw.h>
#endif
/* Determine offset type */
#include <stdint.h>
#if defined(_WIN64)
typedef int64_t OffsetType;
#else
typedef uint32_t OffsetType;
#endif
#include <sys/types.h>
#ifdef __cplusplus
extern "C" {
#endif
#define PROT_NONE 0
#define PROT_READ 1
#define PROT_WRITE 2
#define PROT_EXEC 4
#define MAP_FILE 0
#define MAP_SHARED 1
#define MAP_PRIVATE 2
#define MAP_TYPE 0xf
#define MAP_FIXED 0x10
#define MAP_ANONYMOUS 0x20
#define MAP_ANON MAP_ANONYMOUS
#define MAP_FAILED ((void *)-1)
/* Flags for msync. */
#define MS_ASYNC 1
#define MS_SYNC 2
#define MS_INVALIDATE 4
void* mmap(void *addr, size_t len, int prot, int flags, int fildes, OffsetType off);
int munmap(void *addr, size_t len);
int _mprotect(void *addr, size_t len, int prot);
int msync(void *addr, size_t len, int flags);
int mlock(const void *addr, size_t len);
int munlock(const void *addr, size_t len);
#ifdef __cplusplus
}
#endif
#endif /* _SYS_MMAN_H_ */
+51
View File
@@ -0,0 +1,51 @@
/*
* ---------------------------------------------------------------------------
* OpenAES License
* ---------------------------------------------------------------------------
* Copyright (c) 2012, Nabil S. Al Ramli, www.nalramli.com
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
* ---------------------------------------------------------------------------
*/
#ifndef _OAES_CONFIG_H
#define _OAES_CONFIG_H
#ifdef __cplusplus
extern "C" {
#endif
//#ifndef OAES_HAVE_ISAAC
//#define OAES_HAVE_ISAAC 1
//#endif // OAES_HAVE_ISAAC
//#ifndef OAES_DEBUG
//#define OAES_DEBUG 0
//#endif // OAES_DEBUG
//#define OAES_DEBUG 1
#ifdef __cplusplus
}
#endif
#endif // _OAES_CONFIG_H
+1447
View File
File diff suppressed because it is too large. Load diff
+214
View File
@@ -0,0 +1,214 @@
/*
* ---------------------------------------------------------------------------
* OpenAES License
* ---------------------------------------------------------------------------
* Copyright (c) 2012, Nabil S. Al Ramli, www.nalramli.com
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* - Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* - Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
* ---------------------------------------------------------------------------
*/
#ifndef _OAES_LIB_H
#define _OAES_LIB_H
#include <stdint.h>
#ifdef __cplusplus
extern "C" {
#endif
#ifdef _WIN32
# ifdef OAES_SHARED
# ifdef oaes_lib_EXPORTS
# define OAES_API __declspec(dllexport)
# else
# define OAES_API __declspec(dllimport)
# endif
# else
# define OAES_API
# endif
#else
# define OAES_API
#endif // WIN32
#define OAES_VERSION "0.8.1"
#define OAES_BLOCK_SIZE 16
typedef void OAES_CTX;
typedef enum
{
OAES_RET_FIRST = 0,
OAES_RET_SUCCESS = 0,
OAES_RET_UNKNOWN,
OAES_RET_ARG1,
OAES_RET_ARG2,
OAES_RET_ARG3,
OAES_RET_ARG4,
OAES_RET_ARG5,
OAES_RET_NOKEY,
OAES_RET_MEM,
OAES_RET_BUF,
OAES_RET_HEADER,
OAES_RET_COUNT
} OAES_RET;
/*
* oaes_set_option() takes one of these values for its [option] parameter
* some options accept either an optional or a required [value] parameter
*/
// no option
#define OAES_OPTION_NONE 0
// enable ECB mode, disable CBC mode
#define OAES_OPTION_ECB 1
// enable CBC mode, disable ECB mode
// value is optional, may pass uint8_t iv[OAES_BLOCK_SIZE] to specify
// the value of the initialization vector, iv
#define OAES_OPTION_CBC 2
#ifdef OAES_DEBUG
typedef int ( * oaes_step_cb ) (
const uint8_t state[OAES_BLOCK_SIZE],
const char * step_name,
int step_count,
void * user_data );
// enable state stepping mode
// value is required, must pass oaes_step_cb to receive the state at each step
#define OAES_OPTION_STEP_ON 4
// disable state stepping mode
#define OAES_OPTION_STEP_OFF 8
#endif // OAES_DEBUG
typedef uint16_t OAES_OPTION;
typedef struct _oaes_key
{
size_t data_len;
uint8_t *data;
size_t exp_data_len;
uint8_t *exp_data;
size_t num_keys;
size_t key_base;
} oaes_key;
typedef struct _oaes_ctx
{
#ifdef OAES_HAVE_ISAAC
randctx * rctx;
#endif // OAES_HAVE_ISAAC
#ifdef OAES_DEBUG
oaes_step_cb step_cb;
#endif // OAES_DEBUG
oaes_key * key;
OAES_OPTION options;
uint8_t iv[OAES_BLOCK_SIZE];
} oaes_ctx;
/*
* // usage:
*
* OAES_CTX * ctx = oaes_alloc();
* .
* .
* .
* {
* oaes_gen_key_xxx( ctx );
* {
* oaes_key_export( ctx, _buf, &_buf_len );
* // or
* oaes_key_export_data( ctx, _buf, &_buf_len );\
* }
* }
* // or
* {
* oaes_key_import( ctx, _buf, _buf_len );
* // or
* oaes_key_import_data( ctx, _buf, _buf_len );
* }
* .
* .
* .
* oaes_encrypt( ctx, m, m_len, c, &c_len );
* .
* .
* .
* oaes_decrypt( ctx, c, c_len, m, &m_len );
* .
* .
* .
* oaes_free( &ctx );
*/
OAES_API OAES_CTX * oaes_alloc(void);
OAES_API OAES_RET oaes_free( OAES_CTX ** ctx );
OAES_API OAES_RET oaes_set_option( OAES_CTX * ctx,
OAES_OPTION option, const void * value );
OAES_API OAES_RET oaes_key_gen_128( OAES_CTX * ctx );
OAES_API OAES_RET oaes_key_gen_192( OAES_CTX * ctx );
OAES_API OAES_RET oaes_key_gen_256( OAES_CTX * ctx );
// export key with header information
// set data == NULL to get the required data_len
OAES_API OAES_RET oaes_key_export( OAES_CTX * ctx,
uint8_t * data, size_t * data_len );
// directly export the data from key
// set data == NULL to get the required data_len
OAES_API OAES_RET oaes_key_export_data( OAES_CTX * ctx,
uint8_t * data, size_t * data_len );
// import key with header information
OAES_API OAES_RET oaes_key_import( OAES_CTX * ctx,
const uint8_t * data, size_t data_len );
// directly import data into key
OAES_API OAES_RET oaes_key_import_data( OAES_CTX * ctx,
const uint8_t * data, size_t data_len );
// set c == NULL to get the required c_len
OAES_API OAES_RET oaes_encrypt( OAES_CTX * ctx,
const uint8_t * m, size_t m_len, uint8_t * c, size_t * c_len );
// set m == NULL to get the required m_len
OAES_API OAES_RET oaes_decrypt( OAES_CTX * ctx,
const uint8_t * c, size_t c_len, uint8_t * m, size_t * m_len );
// set buf == NULL to get the required buf_len
OAES_API OAES_RET oaes_sprintf(
char * buf, size_t * buf_len, const uint8_t * data, size_t data_len );
OAES_API OAES_RET oaes_encryption_round( const uint8_t * key, uint8_t * c );
OAES_API OAES_RET oaes_pseudo_encrypt_ecb( OAES_CTX * ctx, uint8_t * c );
#ifdef __cplusplus
}
#endif
#endif // _OAES_LIB_H
+376
View File
@@ -0,0 +1,376 @@
// Memory-hard extension of keccak for PoW
// Copyright (c) 2012-2013 The Cryptonote developers
// Copyright (c) 2014 The Boolberry developers
// Distributed under the MIT/X11 software license, see the accompanying
// file COPYING or http://www.opensource.org/licenses/mit-license.php.
// Modified for CPUminer by Lucas Jones
// Adapted for ccminer by Tanguy Pruvot - 2016
#include <errno.h>
#include <unistd.h>
#include <stdlib.h>
#include <string.h>
#ifdef _MSC_VER
#include <emmintrin.h>
#include <openssl/opensslv.h>
#if OPENSSL_VERSION_NUMBER < 0x10100000L
#include "compat/bignum_ssl10.hpp"
#else
#include "bignum.hpp"
#endif
#include "int128_c.h"
#else
#include <x86intrin.h>
#endif
#include <miner.h>
#include "xmr-rpc.h"
extern uint64_t* pscratchpad_buff;
struct reciprocal_value64 {
uint64_t m;
uint8_t sh1, sh2;
};
static inline int fls64(uint64_t x)
{
#if defined(_WIN64)
unsigned long bitpos = 0;
_BitScanReverse64(&bitpos, x);
return (int) (bitpos + 1);
#elif defined(WIN32)
unsigned long hipos = 0, bitpos = 0;
uint32_t hi = x >> 32;
_BitScanReverse(&hipos, hi);
if (!hipos) {
_BitScanReverse(&bitpos, (uint32_t) x);
}
return (int) hipos ? hipos + 33 : bitpos + 1;
#else
/*
* AMD64 says BSRQ won't clobber the dest reg if x==0; Intel64 says the
* dest reg is undefined if x==0, but their CPU architect says its
* value is written to set it to the same as before.
*/
register long bitpos = -1;
asm("bsrq %1,%0" : "+r" (bitpos) : "rm" (x));
return bitpos + 1;
#endif
}
static inline struct reciprocal_value64 reciprocal_val64(uint64_t d)
{
struct reciprocal_value64 R;
int l;
l = fls64(d - 1);
#ifdef _MSC_VER
uint128 v1;
v1.Lo = (1ULL << l) - d;v1.Hi=0;
uint128 v2;
v2.Hi = 1; v2.Lo = 0;
uint128 v;
mult128(v1,v2,&v);
divmod128by64(v.Hi,v.Lo,d,&v.Hi,&v.Lo);
Increment(&v);
R.m = (uint64_t)v.Hi;
#else
__uint128_t m;
m = (((__uint128_t)1 << 64) * ((1ULL << l) - d));
m /= d;
++m;
R.m = (uint64_t)m;
#endif
R.sh1 = min(l, 1);
R.sh2 = max(l - 1, 0);
return R;
}
static inline uint64_t reciprocal_divide64(uint64_t a, struct reciprocal_value64 R)
{
#ifdef _MSC_VER
uint128 v;
mult64to128(a,R.m,&v.Hi,&v.Lo);
uint64_t t = v.Hi;
#else
uint64_t t = (uint64_t)(((__uint128_t)a * R.m) >> 64);
#endif
return (t + ((a - t) >> R.sh1)) >> R.sh2;
}
static inline uint64_t reciprocal_remainder64(uint64_t A, uint64_t B, struct reciprocal_value64 R)
{
uint64_t div, mod;
div = reciprocal_divide64(A, R);
mod = A - (uint64_t) (div * B);
if (mod >= B) mod -= B;
return mod;
}
//#define UNROLL_SCR_MIX
static inline uint64_t rotl641(uint64_t x) { return((x << 1) | (x >> 63)); }
static inline uint64_t rotl64_1(uint64_t x, uint64_t y) { return((x << y) | (x >> (64 - y))); }
static inline uint64_t rotl64_2(uint64_t x, uint64_t y) { return(rotl64_1((x >> 32) | (x << 32), y)); }
static inline uint64_t bitselect(uint64_t a, uint64_t b, uint64_t c) { return(a ^ (c & (b ^ a))); }
static inline void keccakf_mul(uint64_t *s)
{
uint64_t bc[5], t[5];
uint64_t tmp1, tmp2;
int i;
for(i = 0; i < 5; i++)
t[i] = s[i + 0] ^ s[i + 5] ^ s[i + 10] * s[i + 15] * s[i + 20];
bc[0] = t[0] ^ rotl641(t[2]);
bc[1] = t[1] ^ rotl641(t[3]);
bc[2] = t[2] ^ rotl641(t[4]);
bc[3] = t[3] ^ rotl641(t[0]);
bc[4] = t[4] ^ rotl641(t[1]);
tmp1 = s[1] ^ bc[0];
s[ 0] ^= bc[4];
s[ 1] = rotl64_1(s[ 6] ^ bc[0], 44);
s[ 6] = rotl64_1(s[ 9] ^ bc[3], 20);
s[ 9] = rotl64_1(s[22] ^ bc[1], 61);
s[22] = rotl64_1(s[14] ^ bc[3], 39);
s[14] = rotl64_1(s[20] ^ bc[4], 18);
s[20] = rotl64_1(s[ 2] ^ bc[1], 62);
s[ 2] = rotl64_1(s[12] ^ bc[1], 43);
s[12] = rotl64_1(s[13] ^ bc[2], 25);
s[13] = rotl64_1(s[19] ^ bc[3], 8);
s[19] = rotl64_1(s[23] ^ bc[2], 56);
s[23] = rotl64_1(s[15] ^ bc[4], 41);
s[15] = rotl64_1(s[ 4] ^ bc[3], 27);
s[ 4] = rotl64_1(s[24] ^ bc[3], 14);
s[24] = rotl64_1(s[21] ^ bc[0], 2);
s[21] = rotl64_1(s[ 8] ^ bc[2], 55);
s[ 8] = rotl64_1(s[16] ^ bc[0], 45);
s[16] = rotl64_1(s[ 5] ^ bc[4], 36);
s[ 5] = rotl64_1(s[ 3] ^ bc[2], 28);
s[ 3] = rotl64_1(s[18] ^ bc[2], 21);
s[18] = rotl64_1(s[17] ^ bc[1], 15);
s[17] = rotl64_1(s[11] ^ bc[0], 10);
s[11] = rotl64_1(s[ 7] ^ bc[1], 6);
s[ 7] = rotl64_1(s[10] ^ bc[4], 3);
s[10] = rotl64_1(tmp1, 1);
tmp1 = s[0]; tmp2 = s[1]; s[0] = bitselect(s[0] ^ s[2], s[0], s[1]); s[1] = bitselect(s[1] ^ s[3], s[1], s[2]); s[2] = bitselect(s[2] ^ s[4], s[2], s[3]); s[3] = bitselect(s[3] ^ tmp1, s[3], s[4]); s[4] = bitselect(s[4] ^ tmp2, s[4], tmp1);
tmp1 = s[5]; tmp2 = s[6]; s[5] = bitselect(s[5] ^ s[7], s[5], s[6]); s[6] = bitselect(s[6] ^ s[8], s[6], s[7]); s[7] = bitselect(s[7] ^ s[9], s[7], s[8]); s[8] = bitselect(s[8] ^ tmp1, s[8], s[9]); s[9] = bitselect(s[9] ^ tmp2, s[9], tmp1);
tmp1 = s[10]; tmp2 = s[11]; s[10] = bitselect(s[10] ^ s[12], s[10], s[11]); s[11] = bitselect(s[11] ^ s[13], s[11], s[12]); s[12] = bitselect(s[12] ^ s[14], s[12], s[13]); s[13] = bitselect(s[13] ^ tmp1, s[13], s[14]); s[14] = bitselect(s[14] ^ tmp2, s[14], tmp1);
tmp1 = s[15]; tmp2 = s[16]; s[15] = bitselect(s[15] ^ s[17], s[15], s[16]); s[16] = bitselect(s[16] ^ s[18], s[16], s[17]); s[17] = bitselect(s[17] ^ s[19], s[17], s[18]); s[18] = bitselect(s[18] ^ tmp1, s[18], s[19]); s[19] = bitselect(s[19] ^ tmp2, s[19], tmp1);
tmp1 = s[20]; tmp2 = s[21]; s[20] = bitselect(s[20] ^ s[22], s[20], s[21]); s[21] = bitselect(s[21] ^ s[23], s[21], s[22]); s[22] = bitselect(s[22] ^ s[24], s[22], s[23]); s[23] = bitselect(s[23] ^ tmp1, s[23], s[24]); s[24] = bitselect(s[24] ^ tmp2, s[24], tmp1);
s[0] ^= 0x0000000000000001ULL;
}
static inline void keccakf_mul_last(uint64_t *s)
{
uint64_t bc[5], xormul[5];
uint64_t tmp1, tmp2;
int i;
for(i = 0; i < 5; i++)
xormul[i] = s[i + 0] ^ s[i + 5] ^ s[i + 10] * s[i + 15] * s[i + 20];
bc[0] = xormul[0] ^ rotl641(xormul[2]);
bc[1] = xormul[1] ^ rotl641(xormul[3]);
bc[2] = xormul[2] ^ rotl641(xormul[4]);
bc[3] = xormul[3] ^ rotl641(xormul[0]);
bc[4] = xormul[4] ^ rotl641(xormul[1]);
s[0] ^= bc[4];
s[1] = rotl64_2(s[6] ^ bc[0], 12);
s[2] = rotl64_2(s[12] ^ bc[1], 11);
s[4] = rotl64_1(s[24] ^ bc[3], 14);
s[3] = rotl64_1(s[18] ^ bc[2], 21);
tmp1 = s[0]; tmp2 = s[1]; s[0] = bitselect(s[0] ^ s[2], s[0], s[1]); s[1] = bitselect(s[1] ^ s[3], s[1], s[2]); s[2] = bitselect(s[2] ^ s[4], s[2], s[3]); s[3] = bitselect(s[3] ^ tmp1, s[3], s[4]);
s[0] ^= 0x0000000000000001ULL;
}
struct reciprocal_value64 cached_recip;
static uint64_t cached_scr_size = 0;
static inline void scr_mix(uint64_t *st, uint64_t scr_size, struct reciprocal_value64 recip)
{
#define KK_MIXIN_SIZE 24
uint64_t _ALIGN(128) idx[KK_MIXIN_SIZE];
#ifdef _MSC_VER
#define pscr pscratchpad_buff
int x;
// non-optimized 64bit operations
for (x = 0; x < KK_MIXIN_SIZE; x++) {
idx[x] = reciprocal_remainder64(st[x], scr_size, recip) << 2;
}
if (idx[7] > scr_size*4) {
applog(LOG_WARNING, "Wrong remainder64 returned by the cpu hash %016llx > %016llx",
(unsigned long long) idx[7], (unsigned long long) scr_size*4);
return;
}
for(x = 0; x < KK_MIXIN_SIZE; x += 4) {
st[x + 0] ^= pscr[idx[x] + 0] ^ pscr[idx[x + 1] + 0] ^ pscr[idx[x + 2] + 0] ^ pscr[idx[x + 3] + 0];
st[x + 1] ^= pscr[idx[x] + 1] ^ pscr[idx[x + 1] + 1] ^ pscr[idx[x + 2] + 1] ^ pscr[idx[x + 3] + 1];
st[x + 2] ^= pscr[idx[x] + 2] ^ pscr[idx[x + 1] + 2] ^ pscr[idx[x + 2] + 2] ^ pscr[idx[x + 3] + 2];
st[x + 3] ^= pscr[idx[x] + 3] ^ pscr[idx[x + 1] + 3] ^ pscr[idx[x + 2] + 3] ^ pscr[idx[x + 3] + 3];
}
return;
#elif !defined(UNROLL_SCR_MIX)
#pragma GCC ivdep
for(int x = 0; x < 3; ++x)
{
__m128i *st0, *st1, *st2, *st3;
idx[0] = reciprocal_remainder64(st[(x << 3) + 0], scr_size, recip) << 2;
idx[1] = reciprocal_remainder64(st[(x << 3) + 1], scr_size, recip) << 2;
idx[2] = reciprocal_remainder64(st[(x << 3) + 2], scr_size, recip) << 2;
idx[3] = reciprocal_remainder64(st[(x << 3) + 3], scr_size, recip) << 2;
idx[4] = reciprocal_remainder64(st[(x << 3) + 4], scr_size, recip) << 2;
idx[5] = reciprocal_remainder64(st[(x << 3) + 5], scr_size, recip) << 2;
idx[6] = reciprocal_remainder64(st[(x << 3) + 6], scr_size, recip) << 2;
idx[7] = reciprocal_remainder64(st[(x << 3) + 7], scr_size, recip) << 2;
for(int y = 0; y < 8; y++) _mm_prefetch((const char*) (&pscratchpad_buff[idx[y]]), _MM_HINT_T1);
st0 = (__m128i *)&st[(x << 3) + 0];
st1 = (__m128i *)&st[(x << 3) + 2];
st2 = (__m128i *)&st[(x << 3) + 4];
st3 = (__m128i *)&st[(x << 3) + 6];
*st0 = _mm_xor_si128(*st0, *((__m128i *)&pscratchpad_buff[idx[0]]));
*st0 = _mm_xor_si128(*st0, *((__m128i *)&pscratchpad_buff[idx[1]]));
*st0 = _mm_xor_si128(*st0, *((__m128i *)&pscratchpad_buff[idx[2]]));
*st0 = _mm_xor_si128(*st0, *((__m128i *)&pscratchpad_buff[idx[3]]));
*st1 = _mm_xor_si128(*st1, *((__m128i *)&pscratchpad_buff[idx[0] + 2]));
*st1 = _mm_xor_si128(*st1, *((__m128i *)&pscratchpad_buff[idx[1] + 2]));
*st1 = _mm_xor_si128(*st1, *((__m128i *)&pscratchpad_buff[idx[2] + 2]));
*st1 = _mm_xor_si128(*st1, *((__m128i *)&pscratchpad_buff[idx[3] + 2]));
*st2 = _mm_xor_si128(*st2, *((__m128i *)&pscratchpad_buff[idx[4]]));
*st2 = _mm_xor_si128(*st2, *((__m128i *)&pscratchpad_buff[idx[5]]));
*st2 = _mm_xor_si128(*st2, *((__m128i *)&pscratchpad_buff[idx[6]]));
*st2 = _mm_xor_si128(*st2, *((__m128i *)&pscratchpad_buff[idx[7]]));
*st3 = _mm_xor_si128(*st3, *((__m128i *)&pscratchpad_buff[idx[4] + 2]));
*st3 = _mm_xor_si128(*st3, *((__m128i *)&pscratchpad_buff[idx[5] + 2]));
*st3 = _mm_xor_si128(*st3, *((__m128i *)&pscratchpad_buff[idx[6] + 2]));
*st3 = _mm_xor_si128(*st3, *((__m128i *)&pscratchpad_buff[idx[7] + 2]));
}
#else
#warning using AVX2 optimizations
idx[ 0] = reciprocal_remainder64(st[0], scr_size, recip) << 2;
idx[ 1] = reciprocal_remainder64(st[1], scr_size, recip) << 2;
idx[ 2] = reciprocal_remainder64(st[2], scr_size, recip) << 2;
idx[ 3] = reciprocal_remainder64(st[3], scr_size, recip) << 2;
idx[ 4] = reciprocal_remainder64(st[4], scr_size, recip) << 2;
idx[ 5] = reciprocal_remainder64(st[5], scr_size, recip) << 2;
idx[ 6] = reciprocal_remainder64(st[6], scr_size, recip) << 2;
idx[ 7] = reciprocal_remainder64(st[7], scr_size, recip) << 2;
for(int y = 0; y < 8; y++) _mm_prefetch(&pscratchpad_buff[idx[y]], _MM_HINT_T1);
idx[ 8] = reciprocal_remainder64(st[8], scr_size, recip) << 2;
idx[ 9] = reciprocal_remainder64(st[9], scr_size, recip) << 2;
idx[10] = reciprocal_remainder64(st[10], scr_size, recip) << 2;
idx[11] = reciprocal_remainder64(st[11], scr_size, recip) << 2;
idx[12] = reciprocal_remainder64(st[12], scr_size, recip) << 2;
idx[13] = reciprocal_remainder64(st[13], scr_size, recip) << 2;
idx[14] = reciprocal_remainder64(st[14], scr_size, recip) << 2;
idx[15] = reciprocal_remainder64(st[15], scr_size, recip) << 2;
for(int y = 8; y < 16; ++y) _mm_prefetch(&pscratchpad_buff[idx[y]], _MM_HINT_T1);
idx[16] = reciprocal_remainder64(st[16], scr_size, recip) << 2;
idx[17] = reciprocal_remainder64(st[17], scr_size, recip) << 2;
idx[18] = reciprocal_remainder64(st[18], scr_size, recip) << 2;
idx[19] = reciprocal_remainder64(st[19], scr_size, recip) << 2;
idx[20] = reciprocal_remainder64(st[20], scr_size, recip) << 2;
idx[21] = reciprocal_remainder64(st[21], scr_size, recip) << 2;
idx[22] = reciprocal_remainder64(st[22], scr_size, recip) << 2;
idx[23] = reciprocal_remainder64(st[23], scr_size, recip) << 2;
for(int y = 16; y < 24; ++y) _mm_prefetch(&pscratchpad_buff[idx[y]], _MM_HINT_T1);
__m256i *st0 = (__m256i *)&st[0];
for(int x = 0; x < 6; ++x)
{
*st0 = _mm256_xor_si256(*st0, *((__m256i *)&pscratchpad_buff[idx[(x << 2) + 0]]));
*st0 = _mm256_xor_si256(*st0, *((__m256i *)&pscratchpad_buff[idx[(x << 2) + 1]]));
*st0 = _mm256_xor_si256(*st0, *((__m256i *)&pscratchpad_buff[idx[(x << 2) + 2]]));
*st0 = _mm256_xor_si256(*st0, *((__m256i *)&pscratchpad_buff[idx[(x << 2) + 3]]));
++st0;
}
#endif
return;
}
static void wild_keccak_hash_dbl(uint8_t * __restrict md, const uint8_t * __restrict in)
{
uint64_t _ALIGN(32) st[25];
uint64_t scr_size, i;
struct reciprocal_value64 recip;
scr_size = scratchpad_size >> 2;
if (scr_size == cached_scr_size)
recip = cached_recip;
else {
cached_recip = recip = reciprocal_val64(scr_size);
cached_scr_size = scr_size;
}
// Wild Keccak #1
memcpy(st, in, 88);
st[10] = (st[10] & 0x00000000000000FFULL) | 0x0000000000000100ULL;
memset(&st[11], 0, 112);
st[16] |= 0x8000000000000000ULL;
for(i = 0; i < 23; i++) {
keccakf_mul(st);
scr_mix(st, scr_size, recip);
}
keccakf_mul_last(st);
// Wild Keccak #2
memset(&st[4], 0x00, 168);
st[ 4] = 0x0000000000000001ULL;
st[16] = 0x8000000000000000ULL;
for(i = 0; i < 23; i++) {
keccakf_mul(st);
scr_mix(st, scr_size, recip);
}
keccakf_mul_last(st);
memcpy(md, st, 32);
return;
}
void wildkeccak_hash(void* output, const void* input, uint64_t* scratchpad, uint64_t ssize)
{
if (scratchpad) pscratchpad_buff = scratchpad;
if (!scratchpad_size) scratchpad_size = ssize;
wild_keccak_hash_dbl((uint8_t*)output, (uint8_t*)input);
}
+381
View File
@@ -0,0 +1,381 @@
// WildKeccak CUDA Kernel, Code based on Linux Wolf0 bbr-miner implementation from 2014
// Adapted to ccminer 2.0 - tpruvot 2016-2017
//
// NOTE FOR SP: this ccminer version is licensed under GPLv3 Licence
extern "C" {
#include <errno.h>
#include <stdio.h>
#include <unistd.h>
}
#include <miner.h>
#include <cuda_helper.h>
#include <cuda_vector_uint2x4.h> // todo
#include "wildkeccak.h"
extern char *device_config[MAX_GPUS]; // -l
extern uint64_t* pscratchpad_buff;
static uint64_t* d_input[MAX_GPUS];
static uint32_t* d_retnonce[MAX_GPUS];
static ulonglong4* d_scratchpad[MAX_GPUS];
static uint64_t* h_scratchpad[MAX_GPUS] = { 0 };
static cudaStream_t bufpad_stream[MAX_GPUS] = { 0 };
static cudaStream_t kernel_stream[MAX_GPUS] = { 0 };
uint64_t scratchpad_size = 0;
uint32_t WK_CUDABlocks = 64;
uint32_t WK_CUDAThreads = 256;
#define st0 vst0.x
#define st1 vst0.y
#define st2 vst0.z
#define st3 vst0.w
#define st4 vst4.x
#define st5 vst4.y
#define st6 vst4.z
#define st7 vst4.w
#define st8 vst8.x
#define st9 vst8.y
#define st10 vst8.z
#define st11 vst8.w
#define st12 vst12.x
#define st13 vst12.y
#define st14 vst12.z
#define st15 vst12.w
#define st16 vst16.x
#define st17 vst16.y
#define st18 vst16.z
#define st19 vst16.w
#define st20 vst20.x
#define st21 vst20.y
#define st22 vst20.z
#define st23 vst20.w
#if __CUDA_ARCH__ >= 320
__device__ __forceinline__ uint64_t cuda_rotl641(const uint64_t value)
{
uint2 result;
asm("shf.l.wrap.b32 %0, %1, %2, 1U;" : "=r"(result.x)
: "r"(__double2hiint(__longlong_as_double(value))), "r"(__double2loint(__longlong_as_double(value))));
asm("shf.l.wrap.b32 %0, %1, %2, 1U;" : "=r"(result.y)
: "r"(__double2loint(__longlong_as_double(value))), "r"(__double2hiint(__longlong_as_double(value))));
return __double_as_longlong(__hiloint2double(result.y, result.x));
}
#else
__noinline__ __device__ uint64_t cuda_rotl641(const uint64_t x) { return((x << 1) | (x >> 63)); }
#endif
__noinline__ __device__ uint64_t bitselect(const uint64_t a, const uint64_t b, const uint64_t c) { return(a ^ (c & (b ^ a))); }
#define ROTL641(x) (cuda_rotl641(x))
#define RND() \
bc[0] = st0 ^ st5 ^ st10 * st15 * st20 ^ ROTL641(st2 ^ st7 ^ st12 * st17 * st22); \
bc[1] = st1 ^ st6 ^ st11 * st16 * st21 ^ ROTL641(st3 ^ st8 ^ st13 * st18 * st23); \
bc[2] = st2 ^ st7 ^ st12 * st17 * st22 ^ ROTL641(st4 ^ st9 ^ st14 * st19 * st24); \
bc[3] = st3 ^ st8 ^ st13 * st18 * st23 ^ ROTL641(st0 ^ st5 ^ st10 * st15 * st20); \
bc[4] = st4 ^ st9 ^ st14 * st19 * st24 ^ ROTL641(st1 ^ st6 ^ st11 * st16 * st21); \
tmp1 = st1 ^ bc[0]; \
\
st0 ^= bc[4]; \
st1 = ROTL64(st6 ^ bc[0], 44); \
st6 = ROTL64(st9 ^ bc[3], 20); \
st9 = ROTL64(st22 ^ bc[1], 61); \
st22 = ROTL64(st14 ^ bc[3], 39); \
st14 = ROTL64(st20 ^ bc[4], 18); \
st20 = ROTL64(st2 ^ bc[1], 62); \
st2 = ROTL64(st12 ^ bc[1], 43); \
st12 = ROTL64(st13 ^ bc[2], 25); \
st13 = ROTL64(st19 ^ bc[3], 8); \
st19 = ROTL64(st23 ^ bc[2], 56); \
st23 = ROTL64(st15 ^ bc[4], 41); \
st15 = ROTL64(st4 ^ bc[3], 27); \
st4 = ROTL64(st24 ^ bc[3], 14); \
st24 = ROTL64(st21 ^ bc[0], 2); \
st21 = ROTL64(st8 ^ bc[2], 55); \
st8 = ROTL64(st16 ^ bc[0], 45); \
st16 = ROTL64(st5 ^ bc[4], 36); \
st5 = ROTL64(st3 ^ bc[2], 28); \
st3 = ROTL64(st18 ^ bc[2], 21); \
st18 = ROTL64(st17 ^ bc[1], 15); \
st17 = ROTL64(st11 ^ bc[0], 10); \
st11 = ROTL64(st7 ^ bc[1], 6); \
st7 = ROTL64(st10 ^ bc[4], 3); \
st10 = ROTL641(tmp1); \
\
tmp1 = st0; tmp2 = st1; st0 = bitselect(st0 ^ st2, st0, st1); st1 = bitselect(st1 ^ st3, st1, st2); \
st2 = bitselect(st2 ^ st4, st2, st3); st3 = bitselect(st3 ^ tmp1, st3, st4); st4 = bitselect(st4 ^ tmp2, st4, tmp1); \
tmp1 = st5; tmp2 = st6; st5 = bitselect(st5 ^ st7, st5, st6); st6 = bitselect(st6 ^ st8, st6, st7); \
st7 = bitselect(st7 ^ st9, st7, st8); st8 = bitselect(st8 ^ tmp1, st8, st9); st9 = bitselect(st9 ^ tmp2, st9, tmp1); \
tmp1 = st10; tmp2 = st11; st10 = bitselect(st10 ^ st12, st10, st11); st11 = bitselect(st11 ^ st13, st11, st12); \
st12 = bitselect(st12 ^ st14, st12, st13); st13 = bitselect(st13 ^ tmp1, st13, st14); st14 = bitselect(st14 ^ tmp2, st14, tmp1); \
tmp1 = st15; tmp2 = st16; st15 = bitselect(st15 ^ st17, st15, st16); st16 = bitselect(st16 ^ st18, st16, st17); \
st17 = bitselect(st17 ^ st19, st17, st18); st18 = bitselect(st18 ^ tmp1, st18, st19); st19 = bitselect(st19 ^ tmp2, st19, tmp1); \
tmp1 = st20; tmp2 = st21; st20 = bitselect(st20 ^ st22, st20, st21); st21 = bitselect(st21 ^ st23, st21, st22); \
st22 = bitselect(st22 ^ st24, st22, st23); st23 = bitselect(st23 ^ tmp1, st23, st24); st24 = bitselect(st24 ^ tmp2, st24, tmp1); \
st0 ^= 1;
#define LASTRND1() \
bc[0] = st0 ^ st5 ^ st10 * st15 * st20 ^ ROTL64(st2 ^ st7 ^ st12 * st17 * st22, 1); \
bc[1] = st1 ^ st6 ^ st11 * st16 * st21 ^ ROTL64(st3 ^ st8 ^ st13 * st18 * st23, 1); \
bc[2] = st2 ^ st7 ^ st12 * st17 * st22 ^ ROTL64(st4 ^ st9 ^ st14 * st19 * st24, 1); \
bc[3] = st3 ^ st8 ^ st13 * st18 * st23 ^ ROTL64(st0 ^ st5 ^ st10 * st15 * st20, 1); \
bc[4] = st4 ^ st9 ^ st14 * st19 * st24 ^ ROTL64(st1 ^ st6 ^ st11 * st16 * st21, 1); \
\
st0 ^= bc[4]; \
st1 = ROTL64(st6 ^ bc[0], 44); \
st2 = ROTL64(st12 ^ bc[1], 43); \
st4 = ROTL64(st24 ^ bc[3], 14); \
st3 = ROTL64(st18 ^ bc[2], 21); \
\
tmp1 = st0; st0 = bitselect(st0 ^ st2, st0, st1); st1 = bitselect(st1 ^ st3, st1, st2); st2 = bitselect(st2 ^ st4, st2, st3); st3 = bitselect(st3 ^ tmp1, st3, st4); \
st0 ^= 1;
#define LASTRND2() \
bc[2] = st2 ^ st7 ^ st12 * st17 * st22 ^ ROTL64(st4 ^ st9 ^ st14 * st19 * st24, 1); \
bc[3] = st3 ^ st8 ^ st13 * st18 * st23 ^ ROTL64(st0 ^ st5 ^ st10 * st15 * st20, 1); \
bc[4] = st4 ^ st9 ^ st14 * st19 * st24 ^ ROTL64(st1 ^ st6 ^ st11 * st16 * st21, 1); \
\
st0 ^= bc[4]; \
st4 = ROTL64(st24 ^ bc[3], 14); \
st3 = ROTL64(st18 ^ bc[2], 21); \
st3 = bitselect(st3 ^ st0, st3, st4);
__device__ ulonglong4 operator^(const ulonglong4 &a, const ulonglong4 &b)
{
return(make_ulonglong4(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w));
}
#define MIX(vst) vst = vst ^ scratchpad[vst.x % scr_size] ^ scratchpad[vst.y % scr_size] ^ scratchpad[vst.z % scr_size] ^ scratchpad[vst.w % scr_size];
#define MIX_ALL MIX(vst0); MIX(vst4); MIX(vst8); MIX(vst12); MIX(vst16); MIX(vst20);
__global__
void wk(uint32_t* __restrict__ retnonce, const uint64_t* __restrict__ input, const ulonglong4* __restrict__ scratchpad,
const uint32_t scr_size, const uint32_t target, uint64_t startNonce)
{
ulonglong4 vst0, vst4, vst8, vst12, vst16, vst20;
uint64_t bc[5];
uint64_t st24, tmp1, tmp2;
const uint64_t nonce = startNonce + (blockDim.x * blockIdx.x) + threadIdx.x;
vst0 = make_ulonglong4((nonce << 8) + (input[0] & 0xFF), input[1] & 0xFFFFFFFFFFFFFF00ULL, input[2], input[3]);
vst4 = make_ulonglong4(input[4], input[5], input[6], input[7]);
vst8 = make_ulonglong4(input[8], input[9], (input[10] & 0xFF) | 0x100, 0);
vst12 = make_ulonglong4(0, 0, 0, 0);
vst16 = make_ulonglong4(0x8000000000000000ULL, 0, 0, 0);
vst20 = make_ulonglong4(0, 0, 0, 0);
st24 = 0;
RND();
MIX_ALL;
for(int i = 0; i < 22; i++) {
RND();
MIX_ALL;
}
LASTRND1();
vst4 = make_ulonglong4(1, 0, 0, 0);
vst8 = make_ulonglong4(0, 0, 0, 0);
vst12 = make_ulonglong4(0, 0, 0, 0);
vst16 = make_ulonglong4(0x8000000000000000ULL, 0, 0, 0);
vst20 = make_ulonglong4(0, 0, 0, 0);
st24 = 0;
RND();
MIX_ALL;
#pragma unroll
for(int i = 0; i < 22; i++) {
RND();
MIX_ALL;
}
LASTRND2();
if((st3 >> 32) <= target) {
retnonce[0] = (uint32_t) nonce;
retnonce[1] = retnonce[0];
}
}
__host__
void wildkeccak_kernel(const int thr_id, const uint32_t threads, const uint32_t startNounce, const uint2 target, uint32_t *resNonces)
{
CUDA_SAFE_CALL(cudaMemsetAsync(d_retnonce[thr_id], 0xff, 2 * sizeof(uint32_t), kernel_stream[thr_id]));
const uint32_t threadsperblock = WK_CUDAThreads;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
wk <<<grid, block, 0, kernel_stream[thr_id]>>> (d_retnonce[thr_id], d_input[thr_id], d_scratchpad[thr_id],
(uint32_t)(scratchpad_size >> 2), target.y, startNounce);
cudaMemcpyAsync(resNonces, d_retnonce[thr_id], 2*sizeof(uint32_t), cudaMemcpyDeviceToHost, kernel_stream[thr_id]);
}
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_wildkeccak(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *ptarget = work->target;
uint32_t throughput = 0;
uint64_t n, nonce, first;
uint8_t *pdata = (uint8_t*) work->data;
memcpy(&first, &pdata[1], 8);
n = nonce = first;
if (!scratchpad_size || !h_scratchpad[thr_id]) {
if (h_scratchpad[thr_id])
applog(LOG_ERR, "Scratchpad size is not set!");
work->data[0] = 0; // invalidate
sleep(1);
return -EBUSY;
}
if (!init[thr_id]) {
if (device_config[thr_id]) {
sscanf(device_config[thr_id], "%ux%u", &WK_CUDABlocks, &WK_CUDAThreads);
gpulog(LOG_INFO, thr_id, "Using %u x %u kernel launch config, %u threads",
WK_CUDABlocks, WK_CUDAThreads, throughput);
} else {
throughput = cuda_default_throughput(thr_id, WK_CUDABlocks*WK_CUDAThreads);
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
}
cudaSetDevice(device_map[thr_id]);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage (linux)
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
CUDA_LOG_ERROR();
}
CUDA_SAFE_CALL(cudaMalloc(&d_input[thr_id], 88));
CUDA_SAFE_CALL(cudaMalloc(&d_retnonce[thr_id], 2*sizeof(uint32_t)));
int status = (int) cudaMalloc(&d_scratchpad[thr_id], WILD_KECCAK_SCRATCHPAD_BUFFSIZE);
if (status != cudaSuccess) {
gpulog(LOG_ERR, thr_id, "Unable to allocate device memory, %u MB, err %d",
(uint32_t) (WILD_KECCAK_SCRATCHPAD_BUFFSIZE/(1024*1024)), status);
exit(-ENOMEM);
}
cudaStreamCreate(&bufpad_stream[thr_id]);
cudaStreamCreate(&kernel_stream[thr_id]);
CUDA_SAFE_CALL(cudaMemcpyAsync(d_scratchpad[thr_id], h_scratchpad[thr_id], scratchpad_size << 3, cudaMemcpyHostToDevice, bufpad_stream[thr_id]));
init[thr_id] = true;
}
throughput = WK_CUDABlocks * WK_CUDAThreads;
cudaMemcpy(d_input[thr_id], pdata, 88, cudaMemcpyHostToDevice);
// cudaMemset(d_retnonce[thr_id], 0xFF, 2*sizeof(uint32_t));
if (h_scratchpad[thr_id]) {
cudaStreamSynchronize(bufpad_stream[thr_id]);
}
do {
// const uint32_t blocks = WK_CUDABlocks, threads = WK_CUDAThreads;
// const dim3 block(blocks);
// const dim3 thread(threads);
uint32_t h_retnonce[2] = { UINT32_MAX, UINT32_MAX };
uint2 target = make_uint2(ptarget[6], ptarget[7]);
wildkeccak_kernel(thr_id, throughput, (uint32_t) nonce, target, h_retnonce);
/*
wk <<<block, thread, 0, kernel_stream[thr_id]>>> (d_retnonce[thr_id], d_input[thr_id], d_scratchpad[thr_id],
(uint32_t)(scratchpad_size >> 2), nonce, ptarget[7]);
*/
*hashes_done = (unsigned long) (n - first + throughput);
cudaStreamSynchronize(kernel_stream[thr_id]);
if(h_retnonce[0] != UINT32_MAX) {
uint8_t _ALIGN(64) cpuhash[32];
uint32_t* vhash = (uint32_t*) cpuhash;
uint64_t nonce64;
memcpy(&pdata[1], &h_retnonce[0], sizeof(uint32_t));
memcpy(&nonce64, &pdata[1], 8);
wildkeccak_hash(cpuhash, pdata, pscratchpad_buff, scratchpad_size);
if (!cpuhash[31] && vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work_set_target_ratio(work, vhash);
//applog_hex(pdata, 84);
//applog_hex(cpuhash, 32);
//applog_hex(ptarget, 32);
memcpy(work->nonces, &nonce64, 8);
if (n + throughput > max_nonce) {
*hashes_done = (unsigned long) (max_nonce - first);
}
work->valid_nonces = 1;
return 1;
} else if (vhash[7] > ptarget[7]) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for nonce %08x does not validate on CPU!", h_retnonce[0]);
}
}
if (n + throughput >= max_nonce) {
n = max_nonce;
break;
}
n += throughput;
nonce += throughput;
} while(!work_restart[thr_id].restart);
*hashes_done = (unsigned long) (n - first + 1);
return 0;
}
void wildkeccak_scratchpad_need_update(uint64_t* pscratchpad_buff)
{
for(int i = 0; i < opt_n_threads; i++) {
h_scratchpad[i] = pscratchpad_buff;
if (init[i]) {
gpulog(LOG_DEBUG, i, "Starting scratchpad update...");
cudaMemcpyAsync(d_scratchpad[i], h_scratchpad[i], scratchpad_size << 3, cudaMemcpyHostToDevice, bufpad_stream[i]);
work_restart[i].restart = true;
}
}
}
void free_wildkeccak(int thr_id)
{
if (!init[thr_id])
return;
cudaThreadSynchronize();
cudaFree(d_scratchpad[thr_id]);
cudaFree(d_input[thr_id]);
cudaFree(d_retnonce[thr_id]);
cudaStreamDestroy(bufpad_stream[thr_id]);
cudaStreamDestroy(kernel_stream[thr_id]);
cudaDeviceSynchronize();
init[thr_id] = false;
}
+10
View File
@@ -0,0 +1,10 @@
#define WILD_KECCAK_SCRATCHPAD_BUFFSIZE 1ULL << 29
#define WILD_KECCAK_ADDENDUMS_ARRAY_SIZE 10
extern uint64_t scratchpad_size;
extern uint32_t WK_CUDABlocks, WK_CUDAThreads;
void wildkeccak_scratchpad_need_update(uint64_t* pscratchpad_buff);
+1334
View File
File diff suppressed because it is too large. Load diff
+41
View File
@@ -0,0 +1,41 @@
#include <jansson.h>
#include "wildkeccak.h"
#ifdef WIN32
#define _PACKED _ALIGN(4)
#else
#define _PACKED __attribute__((__packed__))
#endif
struct _PACKED scratchpad_hi {
unsigned char prevhash[32];
uint64_t height;
};
struct _PACKED addendums_array_entry {
struct scratchpad_hi prev_hi;
uint64_t add_size;
};
struct _PACKED scratchpad_file_header {
struct scratchpad_hi current_hi;
struct addendums_array_entry add_arr[WILD_KECCAK_ADDENDUMS_ARRAY_SIZE];
uint64_t scratchpad_size;
};
bool rpc2_job_decode(const json_t *job, struct work *work);
bool rpc2_stratum_job(struct stratum_ctx *sctx, json_t *id, json_t *params);
bool rpc2_stratum_gen_work(struct stratum_ctx *sctx, struct work *work);
bool rpc2_stratum_submit(struct pool_infos *pool, struct work *work);
int rpc2_stratum_thread_stuff(struct pool_infos* pool);
bool rpc2_login_decode(const json_t *val);
void rpc2_init();
void GetScratchpad();
+56 -17
View File
@@ -27,25 +27,21 @@ extern "C" {
// CUDA Devices on the System
int cuda_num_devices()
{
int version;
int version = 0, GPU_N = 0;
cudaError_t err = cudaDriverGetVersion(&version);
if (err != cudaSuccess)
{
if (err != cudaSuccess) {
applog(LOG_ERR, "Unable to query CUDA driver version! Is an nVidia driver installed?");
exit(1);
}
int maj = version / 1000, min = version % 100; // same as in deviceQuery sample
if (maj < 5 || (maj == 5 && min < 5))
{
applog(LOG_ERR, "Driver does not support CUDA %d.%d API! Update your nVidia driver!", 5, 5);
if (version < CUDART_VERSION) {
applog(LOG_ERR, "Your system does not support CUDA %d.%d API!",
CUDART_VERSION / 1000, (CUDART_VERSION % 1000) / 10);
exit(1);
}
int GPU_N;
err = cudaGetDeviceCount(&GPU_N);
if (err != cudaSuccess)
{
if (err != cudaSuccess) {
applog(LOG_ERR, "Unable to query number of CUDA devices! Is an nVidia driver installed?");
exit(1);
}
@@ -78,6 +74,7 @@ void cuda_devicenames()
cudaGetDeviceProperties(&props, dev_id);
device_sm[dev_id] = (props.major * 100 + props.minor * 10);
device_mpcount[dev_id] = (short) props.multiProcessorCount;
if (device_name[dev_id]) {
free(device_name[dev_id]);
@@ -105,14 +102,28 @@ void cuda_print_devices()
cudaDeviceProp props;
cudaGetDeviceProperties(&props, dev_id);
if (!opt_n_threads || n < opt_n_threads) {
fprintf(stderr, "GPU #%d: SM %d.%d %s\n", dev_id, props.major, props.minor, device_name[dev_id]);
fprintf(stderr, "GPU #%d: SM %d.%d %s @ %.0f MHz (MEM %.0f)\n", dev_id,
props.major, props.minor, device_name[dev_id],
(double) props.clockRate/1000,
(double) props.memoryClockRate/1000);
#ifdef USE_WRAPNVML
if (opt_debug) nvml_print_device_info(dev_id);
#ifdef WIN32
if (opt_debug) {
unsigned int devNum = nvapi_devnum(dev_id);
nvapi_pstateinfo(devNum);
}
#endif
#endif
}
}
}
void cuda_shutdown()
{
cudaDeviceSynchronize();
// require gpu init first
//if (thr_info != NULL)
// cudaDeviceSynchronize();
cudaDeviceReset();
}
@@ -162,6 +173,21 @@ uint32_t cuda_default_throughput(int thr_id, uint32_t defcount)
return throughput;
}
// since 1.8.3
double throughput2intensity(uint32_t throughput)
{
double intensity = 0.;
uint32_t ws = throughput;
uint8_t i = 0;
while (ws > 1 && i++ < 32)
ws = ws >> 1;
intensity = (double) i;
if (i && ((1U << i) < throughput)) {
intensity += ((double) (throughput-(1U << i)) / (1U << i));
}
return intensity;
}
// if we use 2 threads on the same gpu, we need to reinit the threads
void cuda_reset_device(int thr_id, bool *init)
{
@@ -193,10 +219,18 @@ void cuda_reset_device(int thr_id, bool *init)
int cuda_available_memory(int thr_id)
{
int dev_id = device_map[thr_id % MAX_GPUS];
size_t mtotal, mfree = 0;
#if defined(_WIN32) && defined(USE_WRAPNVML)
uint64_t tot64 = 0, free64 = 0;
// cuda (6.5) one can crash on pascal and dont handle 8GB
nvapiMemGetInfo(dev_id, &free64, &tot64);
return (int) (free64 / (1024));
#else
size_t mtotal = 0, mfree = 0;
cudaSetDevice(dev_id);
cudaDeviceSynchronize();
cudaMemGetInfo(&mfree, &mtotal);
return (int) (mfree / (1024 * 1024));
#endif
}
// Check (and reset) last cuda error, and report it in logs
@@ -217,13 +251,18 @@ void cuda_clear_lasterror()
} /* extern "C" */
#endif
int cuda_gpu_clocks(struct cgpu_info *gpu)
int cuda_gpu_info(struct cgpu_info *gpu)
{
cudaDeviceProp props;
if (cudaGetDeviceProperties(&props, gpu->gpu_id) == cudaSuccess) {
gpu->gpu_clock = props.clockRate;
gpu->gpu_memclock = props.memoryClockRate;
gpu->gpu_mem = props.totalGlobalMem;
gpu->gpu_clock = (uint32_t) props.clockRate;
gpu->gpu_memclock = (uint32_t) props.memoryClockRate;
gpu->gpu_mem = (uint64_t) (props.totalGlobalMem / 1024); // kB
#if defined(_WIN32) && defined(USE_WRAPNVML)
// required to get mem size > 4GB (size_t too small for bytes on 32bit)
nvapiMemGetInfo(gpu->gpu_id, &gpu->gpu_memfree, &gpu->gpu_mem); // kB
#endif
gpu->gpu_mem = gpu->gpu_mem / 1024; // MB
return 0;
}
return -1;
+15 -19
View File
@@ -5,6 +5,11 @@
#include "cuda_helper.h"
#ifdef __INTELLISENSE__
#define __CUDA_ARCH__ 500
#define __byte_perm(x,y,n) x
#endif
#include "miner.h"
__constant__ uint32_t pTarget[8]; // Single GPU
@@ -85,9 +90,8 @@ void groestlcoin_gpu_hash_quad(uint32_t threads, uint32_t startNounce, uint32_t
}
}
if(rc == true)
if(resNounce[0] > nounce)
resNounce[0] = nounce;
if(rc && resNounce[0] > nounce)
resNounce[0] = nounce;
}
}
#endif
@@ -111,9 +115,8 @@ void groestlcoin_cpu_free(int thr_id)
__host__
void groestlcoin_cpu_setBlock(int thr_id, void *data, void *pTargetIn)
{
uint32_t msgBlock[32];
uint32_t msgBlock[32] = { 0 };
memset(msgBlock, 0, sizeof(uint32_t) * 32);
memcpy(&msgBlock[0], data, 80);
// Erweitere die Nachricht auf den Nachrichtenblock (padding)
@@ -125,18 +128,14 @@ void groestlcoin_cpu_setBlock(int thr_id, void *data, void *pTargetIn)
// auf der GPU ausgeführt)
// Blockheader setzen (korrekte Nonce und Hefty Hash fehlen da drin noch)
cudaMemcpyToSymbol( groestlcoin_gpu_msg,
msgBlock,
128);
cudaMemcpyToSymbol(groestlcoin_gpu_msg, msgBlock, 128);
cudaMemset(d_resultNonce[thr_id], 0xFF, sizeof(uint32_t));
cudaMemcpyToSymbol( pTarget,
pTargetIn,
sizeof(uint32_t) * 8 );
cudaMemcpyToSymbol(pTarget, pTargetIn, 32);
}
__host__
void groestlcoin_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, void *outputHashes, uint32_t *nounce)
void groestlcoin_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *resNonce)
{
uint32_t threadsperblock = 256;
@@ -144,13 +143,10 @@ void groestlcoin_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, vo
// mit den Quad Funktionen brauchen wir jetzt 4 threads pro Hash, daher Faktor 4 bei der Blockzahl
int factor = 4;
// berechne wie viele Thread Blocks wir brauchen
// berechne wie viele Thread Blocks wir brauchen
dim3 grid(factor*((threads + threadsperblock-1)/threadsperblock));
dim3 block(threadsperblock);
// Größe des dynamischen Shared Memory Bereichs
size_t shared_size = 0;
int dev_id = device_map[thr_id];
if (device_sm[dev_id] < 300 || cuda_arch[dev_id] < 300) {
gpulog(LOG_ERR, thr_id, "Sorry, This algo is not supported by this GPU arch (SM 3.0 required)");
@@ -158,10 +154,10 @@ void groestlcoin_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, vo
}
cudaMemset(d_resultNonce[thr_id], 0xFF, sizeof(uint32_t));
groestlcoin_gpu_hash_quad<<<grid, block, shared_size>>>(threads, startNounce, d_resultNonce[thr_id]);
groestlcoin_gpu_hash_quad <<<grid, block>>> (threads, startNounce, d_resultNonce[thr_id]);
// Strategisches Sleep Kommando zur Senkung der CPU Last
MyStreamSynchronize(NULL, 0, thr_id);
// MyStreamSynchronize(NULL, 0, thr_id);
cudaMemcpy(nounce, d_resultNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
cudaMemcpy(resNonce, d_resultNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
}
+1 -1
View File
@@ -4,6 +4,6 @@
void groestlcoin_cpu_init(int thr_id, uint32_t threads);
void groestlcoin_cpu_free(int thr_id);
void groestlcoin_cpu_setBlock(int thr_id, void *data, void *pTargetIn);
void groestlcoin_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, void *outputHashes, uint32_t *nounce);
void groestlcoin_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *resNonce);
#endif
+17 -20
View File
@@ -25,7 +25,7 @@
extern "C" short device_map[MAX_GPUS];
extern "C" long device_sm[MAX_GPUS];
extern "C" short device_mpcount[MAX_GPUS];
extern int cuda_arch[MAX_GPUS];
// common functions
@@ -77,6 +77,12 @@ extern const uint3 threadIdx;
#define ROTR32(x, n) __funnelshift_r( (x), (x), (n) )
#endif
#define AS_U32(addr) *((uint32_t*)(addr))
#define AS_U64(addr) *((uint64_t*)(addr))
#define AS_UINT2(addr) *((uint2*)(addr))
#define AS_UINT4(addr) *((uint4*)(addr))
#define AS_UL2(addr) *((ulonglong2*)(addr))
__device__ __forceinline__ uint64_t MAKE_ULONGLONG(uint32_t LO, uint32_t HI)
{
#if __CUDA_ARCH__ >= 130
@@ -481,31 +487,12 @@ static __device__ __forceinline__ uint2 operator~ (uint2 a) { return make_uint2(
static __device__ __forceinline__ void operator^= (uint2 &a, uint2 b) { a = a ^ b; }
static __device__ __forceinline__ uint2 operator+ (uint2 a, uint2 b) {
#ifdef __CUDA_ARCH__
uint2 result;
asm("{ // uint2 a+b \n\t"
"add.cc.u32 %0, %2, %4; \n\t"
"addc.u32 %1, %3, %5; \n\t"
"}\n" : "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b.x), "r"(b.y));
return result;
#else
return vectorize(devectorize(a) + devectorize(b));
#endif
}
static __device__ __forceinline__ void operator+= (uint2 &a, uint2 b) { a = a + b; }
static __device__ __forceinline__ uint2 operator- (uint2 a, uint2 b) {
#if defined(__CUDA_ARCH__) && CUDA_VERSION < 7000
uint2 result;
asm("{ // uint2 a-b \n\t"
"sub.cc.u32 %0, %2, %4; \n\t"
"subc.u32 %1, %3, %5; \n\t"
"}\n" : "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b.x), "r"(b.y));
return result;
#else
return vectorize(devectorize(a) - devectorize(b));
#endif
}
static __device__ __forceinline__ void operator-= (uint2 &a, uint2 b) { a = a - b; }
@@ -682,4 +669,14 @@ static uint2 SHR2(uint2 a, int offset)
#endif
}
// CUDA 9+ deprecated functions warnings (new mask param)
#if CUDA_VERSION >= 9000 && __CUDA_ARCH__ >= 300
#undef __shfl
#define __shfl(var, srcLane, width) __shfl_sync(0xFFFFFFFFu, var, srcLane, width)
#undef __shfl_up
#define __shfl_up(var, delta, width) __shfl_up_sync(0xFFFFFFFF, var, delta, width)
#undef __any
#define __any(p) __any_sync(0xFFFFFFFFu, p)
#endif
#endif // #ifndef CUDA_HELPER_H
+258 -293
View File
@@ -5,6 +5,12 @@
#include "cuda_helper.h"
#ifdef __INTELLISENSE__
#define __CUDA_ARCH__ 500
#define __funnelshift_r(x,y,n) (x >> n)
#define atomicExch(p,x) x
#endif
#if __CUDA_ARCH__ >= 300
// 64 Registers Variant for Compute 3.0
#include "quark/groestl_functions_quad.h"
@@ -12,290 +18,276 @@
#endif
// globaler Speicher für alle HeftyHashes aller Threads
__constant__ uint32_t pTarget[8]; // Single GPU
uint32_t *d_outputHashes[MAX_GPUS];
static uint32_t *d_resultNonce[MAX_GPUS];
static uint32_t *d_outputHashes[MAX_GPUS];
static uint32_t *d_resultNonces[MAX_GPUS];
__constant__ uint32_t pTarget[2]; // Same for all GPU
__constant__ uint32_t myriadgroestl_gpu_msg[32];
// muss expandiert werden
__constant__ uint32_t myr_sha256_gpu_constantTable[64];
__constant__ uint32_t myr_sha256_gpu_constantTable2[64];
__constant__ uint32_t myr_sha256_gpu_hashTable[8];
uint32_t myr_sha256_cpu_hashTable[] = {
0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a, 0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19 };
uint32_t myr_sha256_cpu_constantTable[] = {
0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5, 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3, 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc, 0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7, 0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13, 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3, 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5, 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208, 0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2,
const uint32_t myr_sha256_cpu_constantTable[] = {
0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5, 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3, 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc, 0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7, 0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13, 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3, 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5, 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208, 0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2,
};
uint32_t myr_sha256_cpu_w2Table[] = {
0x80000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000200,
0x80000000, 0x01400000, 0x00205000, 0x00005088, 0x22000800, 0x22550014, 0x05089742, 0xa0000020,
0x5a880000, 0x005c9400, 0x0016d49d, 0xfa801f00, 0xd33225d0, 0x11675959, 0xf6e6bfda, 0xb30c1549,
0x08b2b050, 0x9d7c4c27, 0x0ce2a393, 0x88e6e1ea, 0xa52b4335, 0x67a16f49, 0xd732016f, 0x4eeb2e91,
0x5dbf55e5, 0x8eee2335, 0xe2bc5ec2, 0xa83f4394, 0x45ad78f7, 0x36f3d0cd, 0xd99c05e8, 0xb0511dc7,
0x69bc7ac4, 0xbd11375b, 0xe3ba71e5, 0x3b209ff2, 0x18feee17, 0xe25ad9e7, 0x13375046, 0x0515089d,
0x4f0d0f04, 0x2627484e, 0x310128d2, 0xc668b434, 0x420841cc, 0x62d311b8, 0xe59ba771, 0x85a7a484 };
const uint32_t myr_sha256_cpu_w2Table[] = {
0x80000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x00000200,
0x80000000, 0x01400000, 0x00205000, 0x00005088, 0x22000800, 0x22550014, 0x05089742, 0xa0000020,
0x5a880000, 0x005c9400, 0x0016d49d, 0xfa801f00, 0xd33225d0, 0x11675959, 0xf6e6bfda, 0xb30c1549,
0x08b2b050, 0x9d7c4c27, 0x0ce2a393, 0x88e6e1ea, 0xa52b4335, 0x67a16f49, 0xd732016f, 0x4eeb2e91,
0x5dbf55e5, 0x8eee2335, 0xe2bc5ec2, 0xa83f4394, 0x45ad78f7, 0x36f3d0cd, 0xd99c05e8, 0xb0511dc7,
0x69bc7ac4, 0xbd11375b, 0xe3ba71e5, 0x3b209ff2, 0x18feee17, 0xe25ad9e7, 0x13375046, 0x0515089d,
0x4f0d0f04, 0x2627484e, 0x310128d2, 0xc668b434, 0x420841cc, 0x62d311b8, 0xe59ba771, 0x85a7a484
};
#define SWAB32(x) ( ((x & 0x000000FF) << 24) | ((x & 0x0000FF00) << 8) | ((x & 0x00FF0000) >> 8) | ((x & 0xFF000000) >> 24) )
#define SWAB32(x) cuda_swab32(x)
#if __CUDA_ARCH__ < 320
// Kepler (Compute 3.0)
#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
// Kepler (Compute 3.0)
#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#else
// Kepler (Compute 3.5)
#define ROTR32(x, n) __funnelshift_r( (x), (x), (n) )
// Kepler (Compute 3.5)
#define ROTR32(x, n) __funnelshift_r( (x), (x), (n) )
#endif
#define R(x, n) ((x) >> (n))
#define Ch(x, y, z) ((x & (y ^ z)) ^ z)
#define R(x, n) ((x) >> (n))
#define Ch(x, y, z) ((x & (y ^ z)) ^ z)
#define Maj(x, y, z) ((x & (y | z)) | (y & z))
#define S0(x) (ROTR32(x, 2) ^ ROTR32(x, 13) ^ ROTR32(x, 22))
#define S1(x) (ROTR32(x, 6) ^ ROTR32(x, 11) ^ ROTR32(x, 25))
#define s0(x) (ROTR32(x, 7) ^ ROTR32(x, 18) ^ R(x, 3))
#define s1(x) (ROTR32(x, 17) ^ ROTR32(x, 19) ^ R(x, 10))
#define S0(x) (ROTR32(x, 2) ^ ROTR32(x, 13) ^ ROTR32(x, 22))
#define S1(x) (ROTR32(x, 6) ^ ROTR32(x, 11) ^ ROTR32(x, 25))
#define s0(x) (ROTR32(x, 7) ^ ROTR32(x, 18) ^ R(x, 3))
#define s1(x) (ROTR32(x, 17) ^ ROTR32(x, 19) ^ R(x, 10))
__device__ void myriadgroestl_gpu_sha256(uint32_t *message)
__device__ __forceinline__
void myriadgroestl_gpu_sha256(uint32_t *message)
{
uint32_t W1[16];
uint32_t W2[16];
uint32_t W1[16];
#pragma unroll
for(int k=0; k<16; k++)
W1[k] = SWAB32(message[k]);
// Initialisiere die register a bis h mit der Hash-Tabelle
uint32_t regs[8];
uint32_t hash[8];
uint32_t regs[8] = {
0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19
};
// pre
#pragma unroll 8
for (int k=0; k < 8; k++)
{
regs[k] = myr_sha256_gpu_hashTable[k];
hash[k] = regs[k];
}
// Progress W1
#pragma unroll
for(int j=0; j<16; j++)
{
uint32_t T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j] + W1[j];
uint32_t T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
#pragma unroll 16
for(int k=0;k<16;k++)
W1[k] = SWAB32(message[k]);
#pragma unroll 7
for (int k=6; k >= 0; k--) regs[k+1] = regs[k];
regs[0] = T1 + T2;
regs[4] += T1;
}
// Progress W1
#pragma unroll 16
for(int j=0;j<16;j++)
{
uint32_t T1, T2;
T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j] + W1[j];
T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
// Progress W2...W3
uint32_t W2[16];
#pragma unroll 7
for (int k=6; k >= 0; k--) regs[k+1] = regs[k];
regs[0] = T1 + T2;
regs[4] += T1;
}
////// PART 1
#pragma unroll
for(int j=0; j<2; j++)
W2[j] = s1(W1[14+j]) + W1[9+j] + s0(W1[1+j]) + W1[j];
// Progress W2...W3
////// PART 1
#pragma unroll 2
for(int j=0;j<2;j++)
W2[j] = s1(W1[14+j]) + W1[9+j] + s0(W1[1+j]) + W1[j];
#pragma unroll 5
for(int j=2;j<7;j++)
W2[j] = s1(W2[j-2]) + W1[9+j] + s0(W1[1+j]) + W1[j];
#pragma unroll 5
for(int j=2; j<7;j++)
W2[j] = s1(W2[j-2]) + W1[9+j] + s0(W1[1+j]) + W1[j];
#pragma unroll 8
for(int j=7;j<15;j++)
W2[j] = s1(W2[j-2]) + W2[j-7] + s0(W1[1+j]) + W1[j];
#pragma unroll
for(int j=7; j<15; j++)
W2[j] = s1(W2[j-2]) + W2[j-7] + s0(W1[1+j]) + W1[j];
W2[15] = s1(W2[13]) + W2[8] + s0(W2[0]) + W1[15];
W2[15] = s1(W2[13]) + W2[8] + s0(W2[0]) + W1[15];
// Rundenfunktion
#pragma unroll 16
for(int j=0;j<16;j++)
{
uint32_t T1, T2;
T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j + 16] + W2[j];
T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
// Round function
#pragma unroll
for(int j=0; j<16; j++)
{
uint32_t T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j + 16] + W2[j];
uint32_t T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
#pragma unroll 7
for (int l=6; l >= 0; l--) regs[l+1] = regs[l];
regs[0] = T1 + T2;
regs[4] += T1;
}
#pragma unroll 7
for (int l=6; l >= 0; l--) regs[l+1] = regs[l];
regs[0] = T1 + T2;
regs[4] += T1;
}
////// PART 2
#pragma unroll 2
for(int j=0;j<2;j++)
W1[j] = s1(W2[14+j]) + W2[9+j] + s0(W2[1+j]) + W2[j];
#pragma unroll 5
for(int j=2;j<7;j++)
W1[j] = s1(W1[j-2]) + W2[9+j] + s0(W2[1+j]) + W2[j];
////// PART 2
#pragma unroll
for(int j=0; j<2; j++)
W1[j] = s1(W2[14+j]) + W2[9+j] + s0(W2[1+j]) + W2[j];
#pragma unroll 5
for(int j=2; j<7; j++)
W1[j] = s1(W1[j-2]) + W2[9+j] + s0(W2[1+j]) + W2[j];
#pragma unroll 8
for(int j=7;j<15;j++)
W1[j] = s1(W1[j-2]) + W1[j-7] + s0(W2[1+j]) + W2[j];
#pragma unroll
for(int j=7; j<15; j++)
W1[j] = s1(W1[j-2]) + W1[j-7] + s0(W2[1+j]) + W2[j];
W1[15] = s1(W1[13]) + W1[8] + s0(W1[0]) + W2[15];
W1[15] = s1(W1[13]) + W1[8] + s0(W1[0]) + W2[15];
// Rundenfunktion
#pragma unroll 16
for(int j=0;j<16;j++)
{
uint32_t T1, T2;
T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j + 32] + W1[j];
T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
// Round function
#pragma unroll
for(int j=0; j<16; j++)
{
uint32_t T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j + 32] + W1[j];
uint32_t T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
#pragma unroll 7
for (int l=6; l >= 0; l--) regs[l+1] = regs[l];
regs[0] = T1 + T2;
regs[4] += T1;
}
#pragma unroll 7
for (int l=6; l >= 0; l--) regs[l+1] = regs[l];
regs[0] = T1 + T2;
regs[4] += T1;
}
////// PART 3
#pragma unroll 2
for(int j=0;j<2;j++)
W2[j] = s1(W1[14+j]) + W1[9+j] + s0(W1[1+j]) + W1[j];
#pragma unroll 5
for(int j=2;j<7;j++)
W2[j] = s1(W2[j-2]) + W1[9+j] + s0(W1[1+j]) + W1[j];
////// PART 3
#pragma unroll
for(int j=0; j<2; j++)
W2[j] = s1(W1[14+j]) + W1[9+j] + s0(W1[1+j]) + W1[j];
#pragma unroll 8
for(int j=7;j<15;j++)
W2[j] = s1(W2[j-2]) + W2[j-7] + s0(W1[1+j]) + W1[j];
#pragma unroll 5
for(int j=2; j<7; j++)
W2[j] = s1(W2[j-2]) + W1[9+j] + s0(W1[1+j]) + W1[j];
W2[15] = s1(W2[13]) + W2[8] + s0(W2[0]) + W1[15];
#pragma unroll
for(int j=7; j<15; j++)
W2[j] = s1(W2[j-2]) + W2[j-7] + s0(W1[1+j]) + W1[j];
// Rundenfunktion
#pragma unroll 16
for(int j=0;j<16;j++)
{
uint32_t T1, T2;
T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j + 48] + W2[j];
T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
W2[15] = s1(W2[13]) + W2[8] + s0(W2[0]) + W1[15];
#pragma unroll 7
for (int l=6; l >= 0; l--) regs[l+1] = regs[l];
regs[0] = T1 + T2;
regs[4] += T1;
}
// Round function
#pragma unroll
for(int j=0; j<16; j++)
{
uint32_t T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable[j + 48] + W2[j];
uint32_t T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
#pragma unroll 8
for(int k=0;k<8;k++)
hash[k] += regs[k];
#pragma unroll 7
for (int l=6; l >= 0; l--) regs[l+1] = regs[l];
regs[0] = T1 + T2;
regs[4] += T1;
}
/////
///// Zweite Runde (wegen Msg-Padding)
/////
#pragma unroll 8
for(int k=0;k<8;k++)
regs[k] = hash[k];
uint32_t hash[8] = {
0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19
};
// Progress W1
#pragma unroll 64
for(int j=0;j<64;j++)
{
uint32_t T1, T2;
T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable2[j];
T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
#pragma unroll 8
for(int k=0; k<8; k++)
hash[k] += regs[k];
#pragma unroll 7
for (int k=6; k >= 0; k--) regs[k+1] = regs[k];
regs[0] = T1 + T2;
regs[4] += T1;
}
/////
///// 2nd Round (wegen Msg-Padding)
/////
#pragma unroll
for(int k=0; k<8; k++)
regs[k] = hash[k];
#pragma unroll 8
for(int k=0;k<8;k++)
hash[k] += regs[k];
// Progress W1
#pragma unroll
for(int j=0; j<64; j++)
{
uint32_t T1 = regs[7] + S1(regs[4]) + Ch(regs[4], regs[5], regs[6]) + myr_sha256_gpu_constantTable2[j];
uint32_t T2 = S0(regs[0]) + Maj(regs[0], regs[1], regs[2]);
//// FERTIG
#pragma unroll 7
for (int k=6; k >= 0; k--) regs[k+1] = regs[k];
regs[0] = T1 + T2;
regs[4] += T1;
}
#pragma unroll 8
for(int k=0;k<8;k++)
message[k] = SWAB32(hash[k]);
}
#if 0
// Full sha hash
#pragma unroll
for(int k=0; k<8; k++)
hash[k] += regs[k];
__global__ void __launch_bounds__(256, 4)
myriadgroestl_gpu_hash_quad(uint32_t threads, uint32_t startNounce, uint32_t *hashBuffer)
{
#if __CUDA_ARCH__ >= 300
// durch 4 dividieren, weil jeweils 4 Threads zusammen ein Hash berechnen
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x) / 4;
if (thread < threads)
{
// GROESTL
uint32_t paddedInput[8];
#pragma unroll 8
for(int k=0;k<8;k++) paddedInput[k] = myriadgroestl_gpu_msg[4*k+threadIdx.x%4];
uint32_t nounce = startNounce + thread;
if ((threadIdx.x % 4) == 3)
paddedInput[4] = SWAB32(nounce); // 4*4+3 = 19
uint32_t msgBitsliced[8];
to_bitslice_quad(paddedInput, msgBitsliced);
uint32_t state[8];
groestl512_progressMessage_quad(state, msgBitsliced);
uint32_t out_state[16];
from_bitslice_quad(state, out_state);
if ((threadIdx.x & 0x03) == 0)
{
uint32_t *outpHash = &hashBuffer[16 * thread];
#pragma unroll 16
for(int k=0;k<16;k++) outpHash[k] = out_state[k];
}
}
#pragma unroll
for(int k=0; k<8; k++)
message[k] = SWAB32(hash[k]);
#else
message[6] = SWAB32(hash[6] + regs[6]);
message[7] = SWAB32(hash[7] + regs[7]);
#endif
}
__global__ void
myriadgroestl_gpu_hash_quad2(uint32_t threads, uint32_t startNounce, uint32_t *resNounce, uint32_t *hashBuffer)
__global__
//__launch_bounds__(256, 6) // we want <= 40 regs
void myriadgroestl_gpu_hash_sha(uint32_t threads, uint32_t startNounce, uint32_t *hashBuffer, uint32_t *resNonces)
{
#if __CUDA_ARCH__ >= 300
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t nounce = startNounce + thread;
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
const uint32_t nonce = startNounce + thread;
uint32_t out_state[16];
uint32_t *inpHash = &hashBuffer[16 * thread];
#pragma unroll 16
for (int i=0; i < 16; i++)
out_state[i] = inpHash[i];
uint32_t out_state[16];
uint32_t *inpHash = &hashBuffer[16 * thread];
myriadgroestl_gpu_sha256(out_state);
#pragma unroll 16
for (int i=0; i < 16; i++)
out_state[i] = inpHash[i];
int i, position = -1;
bool rc = true;
myriadgroestl_gpu_sha256(out_state);
#pragma unroll 8
for (i = 7; i >= 0; i--) {
if (out_state[i] > pTarget[i]) {
if(position < i) {
position = i;
rc = false;
}
}
if (out_state[i] < pTarget[i]) {
if(position < i) {
position = i;
rc = true;
}
}
}
if (out_state[7] <= pTarget[1] && out_state[6] <= pTarget[0])
{
uint32_t tmp = atomicExch(&resNonces[0], nonce);
if (tmp != UINT32_MAX)
resNonces[1] = tmp;
}
}
#endif
}
if(rc == true)
if(resNounce[0] > nounce)
resNounce[0] = nounce;
}
__global__
__launch_bounds__(256, 4)
void myriadgroestl_gpu_hash_quad(uint32_t threads, uint32_t startNounce, uint32_t *hashBuffer)
{
#if __CUDA_ARCH__ >= 300
// durch 4 dividieren, weil jeweils 4 Threads zusammen ein Hash berechnen
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x) / 4;
if (thread < threads)
{
// GROESTL
uint32_t paddedInput[8];
#pragma unroll 8
for(int k=0; k<8; k++)
paddedInput[k] = myriadgroestl_gpu_msg[4*k+threadIdx.x%4];
uint32_t nounce = startNounce + thread;
if ((threadIdx.x % 4) == 3)
paddedInput[4] = SWAB32(nounce); // 4*4+3 = 19
uint32_t msgBitsliced[8];
to_bitslice_quad(paddedInput, msgBitsliced);
uint32_t state[8];
groestl512_progressMessage_quad(state, msgBitsliced);
uint32_t out_state[16];
from_bitslice_quad(state, out_state);
if ((threadIdx.x & 0x03) == 0)
{
uint32_t *outpHash = &hashBuffer[16 * thread];
#pragma unroll 16
for(int k=0; k<16; k++) outpHash[k] = out_state[k];
}
}
#endif
}
@@ -303,93 +295,66 @@ __global__ void
__host__
void myriadgroestl_cpu_init(int thr_id, uint32_t threads)
{
cudaMemcpyToSymbol( myr_sha256_gpu_hashTable,
myr_sha256_cpu_hashTable,
sizeof(uint32_t) * 8 );
uint32_t temp[64];
for(int i=0; i<64; i++)
temp[i] = myr_sha256_cpu_w2Table[i] + myr_sha256_cpu_constantTable[i];
cudaMemcpyToSymbol( myr_sha256_gpu_constantTable,
myr_sha256_cpu_constantTable,
sizeof(uint32_t) * 64 );
cudaMemcpyToSymbol( myr_sha256_gpu_constantTable2, temp, sizeof(uint32_t) * 64 );
// zweite CPU-Tabelle bauen und auf die GPU laden
uint32_t temp[64];
for(int i=0;i<64;i++)
temp[i] = myr_sha256_cpu_w2Table[i] + myr_sha256_cpu_constantTable[i];
cudaMemcpyToSymbol( myr_sha256_gpu_constantTable,
myr_sha256_cpu_constantTable,
sizeof(uint32_t) * 64 );
cudaMemcpyToSymbol( myr_sha256_gpu_constantTable2,
temp,
sizeof(uint32_t) * 64 );
// to check if the binary supports SM3+
cuda_get_arch(thr_id);
// Speicher für Gewinner-Nonce belegen
cudaMalloc(&d_resultNonce[thr_id], sizeof(uint32_t));
// Speicher für temporäreHashes
cudaMalloc(&d_outputHashes[thr_id], 16*sizeof(uint32_t)*threads);
cudaMalloc(&d_outputHashes[thr_id], (size_t) 64 * threads);
cudaMalloc(&d_resultNonces[thr_id], 2 * sizeof(uint32_t));
}
__host__
void myriadgroestl_cpu_free(int thr_id)
{
cudaFree(d_resultNonce[thr_id]);
cudaFree(d_outputHashes[thr_id]);
cudaFree(d_outputHashes[thr_id]);
cudaFree(d_resultNonces[thr_id]);
}
__host__
void myriadgroestl_cpu_setBlock(int thr_id, void *data, void *pTargetIn)
void myriadgroestl_cpu_setBlock(int thr_id, void *data, uint32_t *pTargetIn)
{
// Nachricht expandieren und setzen
uint32_t msgBlock[32];
uint32_t msgBlock[32] = { 0 };
memcpy(&msgBlock[0], data, 80);
msgBlock[20] = 0x80;
msgBlock[31] = 0x01000000;
memset(msgBlock, 0, sizeof(uint32_t) * 32);
memcpy(&msgBlock[0], data, 80);
// Erweitere die Nachricht auf den Nachrichtenblock (padding)
// Unsere Nachricht hat 80 Byte
msgBlock[20] = 0x80;
msgBlock[31] = 0x01000000;
// groestl512 braucht hierfür keinen CPU-Code (die einzige Runde wird
// auf der GPU ausgeführt)
// Blockheader setzen (korrekte Nonce und Hefty Hash fehlen da drin noch)
cudaMemcpyToSymbol( myriadgroestl_gpu_msg,
msgBlock,
128);
cudaMemset(d_resultNonce[thr_id], 0xFF, sizeof(uint32_t));
cudaMemcpyToSymbol( pTarget,
pTargetIn,
sizeof(uint32_t) * 8 );
cudaMemcpyToSymbol(myriadgroestl_gpu_msg, msgBlock, 128);
cudaMemcpyToSymbol(pTarget, &pTargetIn[6], 2 * sizeof(uint32_t));
}
__host__
void myriadgroestl_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, void *outputHashes, uint32_t *nounce)
void myriadgroestl_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *resNounce)
{
uint32_t threadsperblock = 256;
uint32_t threadsperblock = 256;
// Compute 3.0 benutzt die registeroptimierte Quad Variante mit Warp Shuffle
// mit den Quad Funktionen brauchen wir jetzt 4 threads pro Hash, daher Faktor 4 bei der Blockzahl
const int factor=4;
cudaMemset(d_resultNonces[thr_id], 0xFF, 2 * sizeof(uint32_t));
// Größe des dynamischen Shared Memory Bereichs
size_t shared_size = 0;
// Compute 3.0 benutzt die registeroptimierte Quad Variante mit Warp Shuffle
// mit den Quad Funktionen brauchen wir jetzt 4 threads pro Hash, daher Faktor 4 bei der Blockzahl
const int factor = 4;
cudaMemset(d_resultNonce[thr_id], 0xFF, sizeof(uint32_t));
// berechne wie viele Thread Blocks wir brauchen
dim3 grid(factor*((threads + threadsperblock-1)/threadsperblock));
dim3 block(threadsperblock);
dim3 grid(factor*((threads + threadsperblock-1)/threadsperblock));
dim3 block(threadsperblock);
if (device_sm[device_map[thr_id]] < 300) {
printf("Sorry, This algo is not supported by this GPU arch (SM 3.0 required)");
return;
}
int dev_id = device_map[thr_id];
if (device_sm[dev_id] < 300 || cuda_arch[dev_id] < 300) {
printf("Sorry, This algo is not supported by this GPU arch (SM 3.0 required)");
return;
}
myriadgroestl_gpu_hash_quad<<<grid, block, shared_size>>>(threads, startNounce, d_outputHashes[thr_id]);
dim3 grid2((threads + threadsperblock-1)/threadsperblock);
myriadgroestl_gpu_hash_quad2<<<grid2, block, shared_size>>>(threads, startNounce, d_resultNonce[thr_id], d_outputHashes[thr_id]);
myriadgroestl_gpu_hash_quad <<< grid, block >>> (threads, startNounce, d_outputHashes[thr_id]);
// Strategisches Sleep Kommando zur Senkung der CPU Last
MyStreamSynchronize(NULL, 0, thr_id);
dim3 grid2((threads + threadsperblock-1)/threadsperblock);
myriadgroestl_gpu_hash_sha <<< grid2, block >>> (threads, startNounce, d_outputHashes[thr_id], d_resultNonces[thr_id]);
cudaMemcpy(nounce, d_resultNonce[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
cudaMemcpy(resNounce, d_resultNonces[thr_id], 2 * sizeof(uint32_t), cudaMemcpyDeviceToHost);
}
+47
View File
@@ -0,0 +1,47 @@
#include "cuda_helper.h"
/* Macros for uint2 operations (used by skein) */
__device__ __forceinline__
uint2 ROR8(const uint2 a) {
uint2 result;
result.x = __byte_perm(a.x, a.y, 0x4321);
result.y = __byte_perm(a.y, a.x, 0x4321);
return result;
}
__device__ __forceinline__
uint2 ROL24(const uint2 a) {
uint2 result;
result.x = __byte_perm(a.x, a.y, 0x0765);
result.y = __byte_perm(a.y, a.x, 0x0765);
return result;
}
static __device__ __forceinline__ uint2 operator+ (const uint2 a, const uint32_t b)
{
#if 0 && defined(__CUDA_ARCH__) && CUDA_VERSION < 7000
uint2 result;
asm(
"add.cc.u32 %0,%2,%4; \n\t"
"addc.u32 %1,%3,%5; \n\t"
: "=r"(result.x), "=r"(result.y) : "r"(a.x), "r"(a.y), "r"(b), "r"(0));
return result;
#else
return vectorize(devectorize(a) + b);
#endif
}
/* whirlpool ones */
#ifdef __CUDA_ARCH__
__device__ __forceinline__
uint2 ROL16(const uint2 a) {
uint2 result;
result.x = __byte_perm(a.x, a.y, 0x1076);
result.y = __byte_perm(a.y, a.x, 0x1076);
return result;
}
#else
#define ROL16(a) make_uint2(a.x, a.y) /* bad, just to define it */
#endif
+81
View File
@@ -0,0 +1,81 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2_CONFIG_H__
#define __BLAKE2_CONFIG_H__
// These don't work everywhere
#if (defined(__SSE2__) || defined(_M_AMD_64) || defined(_M_X64))
#define HAVE_SSE2
#endif
#if defined(__SSSE3__)
#define HAVE_SSSE3
#endif
#if defined(__SSE4_1__)
#define HAVE_SSE41
#endif
#if defined(__AVX__)
#define HAVE_AVX
#endif
#if defined(__XOP__)
#define HAVE_XOP
#endif
#ifdef HAVE_AVX2
#ifndef HAVE_AVX
#define HAVE_AVX
#endif
#endif
#ifdef HAVE_XOP
#ifndef HAVE_AVX
#define HAVE_AVX
#endif
#endif
#ifdef HAVE_AVX
#ifndef HAVE_SSE41
#define HAVE_SSE41
#endif
#endif
#ifdef HAVE_SSE41
#ifndef HAVE_SSSE3
#define HAVE_SSSE3
#endif
#endif
#ifdef HAVE_SSSE3
#define HAVE_SSE2
#endif
#if !defined(HAVE_SSE2)
#ifdef _MSC_VER
// enforce required stuff for now
#define HAVE_SSE2
//#define HAVE_SSSE3
#define HAVE_SSE41
#else
# error "This code requires at least SSE 4.1"
#endif
#endif
#endif
+136
View File
@@ -0,0 +1,136 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2_IMPL_H__
#define __BLAKE2_IMPL_H__
#include <stdint.h>
static inline uint32_t load32( const void *src )
{
#if defined(NATIVE_LITTLE_ENDIAN)
uint32_t w;
memcpy(&w, src, sizeof w);
return w;
#else
const uint8_t *p = ( const uint8_t * )src;
uint32_t w = *p++;
w |= ( uint32_t )( *p++ ) << 8;
w |= ( uint32_t )( *p++ ) << 16;
w |= ( uint32_t )( *p++ ) << 24;
return w;
#endif
}
static inline uint64_t load64( const void *src )
{
#if defined(NATIVE_LITTLE_ENDIAN)
uint64_t w;
memcpy(&w, src, sizeof w);
return w;
#else
const uint8_t *p = ( const uint8_t * )src;
uint64_t w = *p++;
w |= ( uint64_t )( *p++ ) << 8;
w |= ( uint64_t )( *p++ ) << 16;
w |= ( uint64_t )( *p++ ) << 24;
w |= ( uint64_t )( *p++ ) << 32;
w |= ( uint64_t )( *p++ ) << 40;
w |= ( uint64_t )( *p++ ) << 48;
w |= ( uint64_t )( *p++ ) << 56;
return w;
#endif
}
static inline void store32( void *dst, uint32_t w )
{
#if defined(NATIVE_LITTLE_ENDIAN)
memcpy(dst, &w, sizeof w);
#else
uint8_t *p = ( uint8_t * )dst;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w;
#endif
}
static inline void store64( void *dst, uint64_t w )
{
#if defined(NATIVE_LITTLE_ENDIAN)
memcpy(dst, &w, sizeof w);
#else
uint8_t *p = ( uint8_t * )dst;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w;
#endif
}
static inline uint64_t load48( const void *src )
{
const uint8_t *p = ( const uint8_t * )src;
uint64_t w = *p++;
w |= ( uint64_t )( *p++ ) << 8;
w |= ( uint64_t )( *p++ ) << 16;
w |= ( uint64_t )( *p++ ) << 24;
w |= ( uint64_t )( *p++ ) << 32;
w |= ( uint64_t )( *p++ ) << 40;
return w;
}
static inline void store48( void *dst, uint64_t w )
{
uint8_t *p = ( uint8_t * )dst;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w; w >>= 8;
*p++ = ( uint8_t )w;
}
static inline uint32_t rotl32( const uint32_t w, const unsigned c )
{
return ( w << c ) | ( w >> ( 32 - c ) );
}
static inline uint64_t rotl64( const uint64_t w, const unsigned c )
{
return ( w << c ) | ( w >> ( 64 - c ) );
}
static inline uint32_t rotr32( const uint32_t w, const unsigned c )
{
return ( w >> c ) | ( w << ( 32 - c ) );
}
static inline uint64_t rotr64( const uint64_t w, const unsigned c )
{
return ( w >> c ) | ( w << ( 64 - c ) );
}
/* prevents compiler optimizing out memset() */
static inline void secure_zero_memory( void *v, size_t n )
{
volatile uint8_t *p = ( volatile uint8_t * )v;
while( n-- ) *p++ = 0;
}
#endif
+85
View File
@@ -0,0 +1,85 @@
#define _mm_roti_epi64(x, c) \
(-(c) == 32) ? _mm_shuffle_epi32((x), _MM_SHUFFLE(2,3,0,1)) \
: (-(c) == 24) ? _mm_shuffle_epi8((x), r24) \
: (-(c) == 16) ? _mm_shuffle_epi8((x), r16) \
: (-(c) == 63) ? _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_add_epi64((x), (x))) \
: _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_slli_epi64((x), 64-(-(c))))
#define G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
row1l = _mm_add_epi64(row1l, row2l); \
row1h = _mm_add_epi64(row1h, row2h); \
\
row4l = _mm_xor_si128(row4l, row1l); \
row4h = _mm_xor_si128(row4h, row1h); \
\
row4l = _mm_roti_epi64(row4l, -32); \
row4h = _mm_roti_epi64(row4h, -32); \
\
row3l = _mm_add_epi64(row3l, row4l); \
row3h = _mm_add_epi64(row3h, row4h); \
\
row2l = _mm_xor_si128(row2l, row3l); \
row2h = _mm_xor_si128(row2h, row3h); \
\
row2l = _mm_roti_epi64(row2l, -24); \
row2h = _mm_roti_epi64(row2h, -24); \
#define G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
row1l = _mm_add_epi64(row1l, row2l); \
row1h = _mm_add_epi64(row1h, row2h); \
\
row4l = _mm_xor_si128(row4l, row1l); \
row4h = _mm_xor_si128(row4h, row1h); \
\
row4l = _mm_roti_epi64(row4l, -16); \
row4h = _mm_roti_epi64(row4h, -16); \
\
row3l = _mm_add_epi64(row3l, row4l); \
row3h = _mm_add_epi64(row3h, row4h); \
\
row2l = _mm_xor_si128(row2l, row3l); \
row2h = _mm_xor_si128(row2h, row3h); \
\
row2l = _mm_roti_epi64(row2l, -63); \
row2h = _mm_roti_epi64(row2h, -63); \
#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = _mm_alignr_epi8(row2h, row2l, 8); \
t1 = _mm_alignr_epi8(row2l, row2h, 8); \
row2l = t0; \
row2h = t1; \
\
t0 = row3l; \
row3l = row3h; \
row3h = t0; \
\
t0 = _mm_alignr_epi8(row4h, row4l, 8); \
t1 = _mm_alignr_epi8(row4l, row4h, 8); \
row4l = t1; \
row4h = t0;
#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = _mm_alignr_epi8(row2l, row2h, 8); \
t1 = _mm_alignr_epi8(row2h, row2l, 8); \
row2l = t0; \
row2h = t1; \
\
t0 = row3l; \
row3l = row3h; \
row3h = t0; \
\
t0 = _mm_alignr_epi8(row4l, row4h, 8); \
t1 = _mm_alignr_epi8(row4h, row4l, 8); \
row4l = t1; \
row4h = t0;
#define BLAKE2_ROUND(row1l,row1h,row2l,row2h,row3l,row3h,row4l,row4h) \
G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
\
DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
\
G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
\
UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
+85
View File
@@ -0,0 +1,85 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2_H__
#define __BLAKE2_H__
#include <stddef.h>
#include <stdint.h>
#if defined(_MSC_VER)
#define ALIGN(x) __declspec(align(x))
#else
#define ALIGN(x) __attribute__ ((__aligned__(x)))
#endif
#if defined(__cplusplus)
extern "C" {
#endif
enum blake2b_constant
{
BLAKE2B_BLOCKBYTES = 128,
BLAKE2B_OUTBYTES = 64,
BLAKE2B_KEYBYTES = 64,
BLAKE2B_SALTBYTES = 16,
BLAKE2B_PERSONALBYTES = 16
};
#pragma pack(push, 1)
typedef struct __blake2b_param
{
uint8_t digest_length; // 1
uint8_t key_length; // 2
uint8_t fanout; // 3
uint8_t depth; // 4
uint32_t leaf_length; // 8
uint64_t node_offset; // 16
uint8_t node_depth; // 17
uint8_t inner_length; // 18
uint8_t reserved[14]; // 32
uint8_t salt[BLAKE2B_SALTBYTES]; // 48
uint8_t personal[BLAKE2B_PERSONALBYTES]; // 64
} blake2b_param;
ALIGN( 64 ) typedef struct __blake2b_state
{
uint64_t h[8];
uint8_t buf[BLAKE2B_BLOCKBYTES];
uint16_t counter;
uint8_t buflen;
uint8_t lastblock;
} blake2b_state;
#pragma pack(pop)
int eq_blake2b_init( blake2b_state *S, const uint8_t outlen );
int eq_blake2b_init_key( blake2b_state *S, const uint8_t outlen, const void *key, const uint8_t keylen );
int eq_blake2b_init_param( blake2b_state *S, const blake2b_param *P );
int eq_blake2b_update( blake2b_state *S, const uint8_t *in, uint64_t inlen );
int eq_blake2b_final( blake2b_state *S, uint8_t *out, uint8_t outlen );
// Simple API
int eq_blake2b( uint8_t *out, const void *in, const void *key, const uint8_t outlen, const uint64_t inlen, uint8_t keylen );
static inline int eq_blake2( uint8_t *out, const void *in, const void *key, const uint8_t outlen, const uint64_t inlen, uint8_t keylen )
{
return eq_blake2b( out, in, key, outlen, inlen, keylen );
}
#if defined(__cplusplus)
}
#endif
#endif
+68
View File
@@ -0,0 +1,68 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2B_LOAD_SSE2_H__
#define __BLAKE2B_LOAD_SSE2_H__
#define LOAD_MSG_0_1(b0, b1) b0 = _mm_set_epi64x(m2, m0); b1 = _mm_set_epi64x(m6, m4)
#define LOAD_MSG_0_2(b0, b1) b0 = _mm_set_epi64x(m3, m1); b1 = _mm_set_epi64x(m7, m5)
#define LOAD_MSG_0_3(b0, b1) b0 = _mm_set_epi64x(m10, m8); b1 = _mm_set_epi64x(m14, m12)
#define LOAD_MSG_0_4(b0, b1) b0 = _mm_set_epi64x(m11, m9); b1 = _mm_set_epi64x(m15, m13)
#define LOAD_MSG_1_1(b0, b1) b0 = _mm_set_epi64x(m4, m14); b1 = _mm_set_epi64x(m13, m9)
#define LOAD_MSG_1_2(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 = _mm_set_epi64x(m6, m15)
#define LOAD_MSG_1_3(b0, b1) b0 = _mm_set_epi64x(m0, m1); b1 = _mm_set_epi64x(m5, m11)
#define LOAD_MSG_1_4(b0, b1) b0 = _mm_set_epi64x(m2, m12); b1 = _mm_set_epi64x(m3, m7)
#define LOAD_MSG_2_1(b0, b1) b0 = _mm_set_epi64x(m12, m11); b1 = _mm_set_epi64x(m15, m5)
#define LOAD_MSG_2_2(b0, b1) b0 = _mm_set_epi64x(m0, m8); b1 = _mm_set_epi64x(m13, m2)
#define LOAD_MSG_2_3(b0, b1) b0 = _mm_set_epi64x(m3, m10); b1 = _mm_set_epi64x(m9, m7)
#define LOAD_MSG_2_4(b0, b1) b0 = _mm_set_epi64x(m6, m14); b1 = _mm_set_epi64x(m4, m1)
#define LOAD_MSG_3_1(b0, b1) b0 = _mm_set_epi64x(m3, m7); b1 = _mm_set_epi64x(m11, m13)
#define LOAD_MSG_3_2(b0, b1) b0 = _mm_set_epi64x(m1, m9); b1 = _mm_set_epi64x(m14, m12)
#define LOAD_MSG_3_3(b0, b1) b0 = _mm_set_epi64x(m5, m2); b1 = _mm_set_epi64x(m15, m4)
#define LOAD_MSG_3_4(b0, b1) b0 = _mm_set_epi64x(m10, m6); b1 = _mm_set_epi64x(m8, m0)
#define LOAD_MSG_4_1(b0, b1) b0 = _mm_set_epi64x(m5, m9); b1 = _mm_set_epi64x(m10, m2)
#define LOAD_MSG_4_2(b0, b1) b0 = _mm_set_epi64x(m7, m0); b1 = _mm_set_epi64x(m15, m4)
#define LOAD_MSG_4_3(b0, b1) b0 = _mm_set_epi64x(m11, m14); b1 = _mm_set_epi64x(m3, m6)
#define LOAD_MSG_4_4(b0, b1) b0 = _mm_set_epi64x(m12, m1); b1 = _mm_set_epi64x(m13, m8)
#define LOAD_MSG_5_1(b0, b1) b0 = _mm_set_epi64x(m6, m2); b1 = _mm_set_epi64x(m8, m0)
#define LOAD_MSG_5_2(b0, b1) b0 = _mm_set_epi64x(m10, m12); b1 = _mm_set_epi64x(m3, m11)
#define LOAD_MSG_5_3(b0, b1) b0 = _mm_set_epi64x(m7, m4); b1 = _mm_set_epi64x(m1, m15)
#define LOAD_MSG_5_4(b0, b1) b0 = _mm_set_epi64x(m5, m13); b1 = _mm_set_epi64x(m9, m14)
#define LOAD_MSG_6_1(b0, b1) b0 = _mm_set_epi64x(m1, m12); b1 = _mm_set_epi64x(m4, m14)
#define LOAD_MSG_6_2(b0, b1) b0 = _mm_set_epi64x(m15, m5); b1 = _mm_set_epi64x(m10, m13)
#define LOAD_MSG_6_3(b0, b1) b0 = _mm_set_epi64x(m6, m0); b1 = _mm_set_epi64x(m8, m9)
#define LOAD_MSG_6_4(b0, b1) b0 = _mm_set_epi64x(m3, m7); b1 = _mm_set_epi64x(m11, m2)
#define LOAD_MSG_7_1(b0, b1) b0 = _mm_set_epi64x(m7, m13); b1 = _mm_set_epi64x(m3, m12)
#define LOAD_MSG_7_2(b0, b1) b0 = _mm_set_epi64x(m14, m11); b1 = _mm_set_epi64x(m9, m1)
#define LOAD_MSG_7_3(b0, b1) b0 = _mm_set_epi64x(m15, m5); b1 = _mm_set_epi64x(m2, m8)
#define LOAD_MSG_7_4(b0, b1) b0 = _mm_set_epi64x(m4, m0); b1 = _mm_set_epi64x(m10, m6)
#define LOAD_MSG_8_1(b0, b1) b0 = _mm_set_epi64x(m14, m6); b1 = _mm_set_epi64x(m0, m11)
#define LOAD_MSG_8_2(b0, b1) b0 = _mm_set_epi64x(m9, m15); b1 = _mm_set_epi64x(m8, m3)
#define LOAD_MSG_8_3(b0, b1) b0 = _mm_set_epi64x(m13, m12); b1 = _mm_set_epi64x(m10, m1)
#define LOAD_MSG_8_4(b0, b1) b0 = _mm_set_epi64x(m7, m2); b1 = _mm_set_epi64x(m5, m4)
#define LOAD_MSG_9_1(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 = _mm_set_epi64x(m1, m7)
#define LOAD_MSG_9_2(b0, b1) b0 = _mm_set_epi64x(m4, m2); b1 = _mm_set_epi64x(m5, m6)
#define LOAD_MSG_9_3(b0, b1) b0 = _mm_set_epi64x(m9, m15); b1 = _mm_set_epi64x(m13, m3)
#define LOAD_MSG_9_4(b0, b1) b0 = _mm_set_epi64x(m14, m11); b1 = _mm_set_epi64x(m0, m12)
#define LOAD_MSG_10_1(b0, b1) b0 = _mm_set_epi64x(m2, m0); b1 = _mm_set_epi64x(m6, m4)
#define LOAD_MSG_10_2(b0, b1) b0 = _mm_set_epi64x(m3, m1); b1 = _mm_set_epi64x(m7, m5)
#define LOAD_MSG_10_3(b0, b1) b0 = _mm_set_epi64x(m10, m8); b1 = _mm_set_epi64x(m14, m12)
#define LOAD_MSG_10_4(b0, b1) b0 = _mm_set_epi64x(m11, m9); b1 = _mm_set_epi64x(m15, m13)
#define LOAD_MSG_11_1(b0, b1) b0 = _mm_set_epi64x(m4, m14); b1 = _mm_set_epi64x(m13, m9)
#define LOAD_MSG_11_2(b0, b1) b0 = _mm_set_epi64x(m8, m10); b1 = _mm_set_epi64x(m6, m15)
#define LOAD_MSG_11_3(b0, b1) b0 = _mm_set_epi64x(m0, m1); b1 = _mm_set_epi64x(m5, m11)
#define LOAD_MSG_11_4(b0, b1) b0 = _mm_set_epi64x(m2, m12); b1 = _mm_set_epi64x(m3, m7)
#endif
+402
View File
@@ -0,0 +1,402 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2B_LOAD_SSE41_H__
#define __BLAKE2B_LOAD_SSE41_H__
#define LOAD_MSG_0_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m0, m1); \
b1 = _mm_unpacklo_epi64(m2, m3); \
} while(0)
#define LOAD_MSG_0_2(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m0, m1); \
b1 = _mm_unpackhi_epi64(m2, m3); \
} while(0)
#define LOAD_MSG_0_3(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m4, m5); \
b1 = _mm_unpacklo_epi64(m6, m7); \
} while(0)
#define LOAD_MSG_0_4(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m4, m5); \
b1 = _mm_unpackhi_epi64(m6, m7); \
} while(0)
#define LOAD_MSG_1_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m7, m2); \
b1 = _mm_unpackhi_epi64(m4, m6); \
} while(0)
#define LOAD_MSG_1_2(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m5, m4); \
b1 = _mm_alignr_epi8(m3, m7, 8); \
} while(0)
#define LOAD_MSG_1_3(b0, b1) \
do \
{ \
b0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
b1 = _mm_unpackhi_epi64(m5, m2); \
} while(0)
#define LOAD_MSG_1_4(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m6, m1); \
b1 = _mm_unpackhi_epi64(m3, m1); \
} while(0)
#define LOAD_MSG_2_1(b0, b1) \
do \
{ \
b0 = _mm_alignr_epi8(m6, m5, 8); \
b1 = _mm_unpackhi_epi64(m2, m7); \
} while(0)
#define LOAD_MSG_2_2(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m4, m0); \
b1 = _mm_blend_epi16(m1, m6, 0xF0); \
} while(0)
#define LOAD_MSG_2_3(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m5, m1, 0xF0); \
b1 = _mm_unpackhi_epi64(m3, m4); \
} while(0)
#define LOAD_MSG_2_4(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m7, m3); \
b1 = _mm_alignr_epi8(m2, m0, 8); \
} while(0)
#define LOAD_MSG_3_1(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m3, m1); \
b1 = _mm_unpackhi_epi64(m6, m5); \
} while(0)
#define LOAD_MSG_3_2(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m4, m0); \
b1 = _mm_unpacklo_epi64(m6, m7); \
} while(0)
#define LOAD_MSG_3_3(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m1, m2, 0xF0); \
b1 = _mm_blend_epi16(m2, m7, 0xF0); \
} while(0)
#define LOAD_MSG_3_4(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m3, m5); \
b1 = _mm_unpacklo_epi64(m0, m4); \
} while(0)
#define LOAD_MSG_4_1(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m4, m2); \
b1 = _mm_unpacklo_epi64(m1, m5); \
} while(0)
#define LOAD_MSG_4_2(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m0, m3, 0xF0); \
b1 = _mm_blend_epi16(m2, m7, 0xF0); \
} while(0)
#define LOAD_MSG_4_3(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m7, m5, 0xF0); \
b1 = _mm_blend_epi16(m3, m1, 0xF0); \
} while(0)
#define LOAD_MSG_4_4(b0, b1) \
do \
{ \
b0 = _mm_alignr_epi8(m6, m0, 8); \
b1 = _mm_blend_epi16(m4, m6, 0xF0); \
} while(0)
#define LOAD_MSG_5_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m1, m3); \
b1 = _mm_unpacklo_epi64(m0, m4); \
} while(0)
#define LOAD_MSG_5_2(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m6, m5); \
b1 = _mm_unpackhi_epi64(m5, m1); \
} while(0)
#define LOAD_MSG_5_3(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m2, m3, 0xF0); \
b1 = _mm_unpackhi_epi64(m7, m0); \
} while(0)
#define LOAD_MSG_5_4(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m6, m2); \
b1 = _mm_blend_epi16(m7, m4, 0xF0); \
} while(0)
#define LOAD_MSG_6_1(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m6, m0, 0xF0); \
b1 = _mm_unpacklo_epi64(m7, m2); \
} while(0)
#define LOAD_MSG_6_2(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m2, m7); \
b1 = _mm_alignr_epi8(m5, m6, 8); \
} while(0)
#define LOAD_MSG_6_3(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m0, m3); \
b1 = _mm_shuffle_epi32(m4, _MM_SHUFFLE(1,0,3,2)); \
} while(0)
#define LOAD_MSG_6_4(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m3, m1); \
b1 = _mm_blend_epi16(m1, m5, 0xF0); \
} while(0)
#define LOAD_MSG_7_1(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m6, m3); \
b1 = _mm_blend_epi16(m6, m1, 0xF0); \
} while(0)
#define LOAD_MSG_7_2(b0, b1) \
do \
{ \
b0 = _mm_alignr_epi8(m7, m5, 8); \
b1 = _mm_unpackhi_epi64(m0, m4); \
} while(0)
#define LOAD_MSG_7_3(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m2, m7); \
b1 = _mm_unpacklo_epi64(m4, m1); \
} while(0)
#define LOAD_MSG_7_4(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m0, m2); \
b1 = _mm_unpacklo_epi64(m3, m5); \
} while(0)
#define LOAD_MSG_8_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m3, m7); \
b1 = _mm_alignr_epi8(m0, m5, 8); \
} while(0)
#define LOAD_MSG_8_2(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m7, m4); \
b1 = _mm_alignr_epi8(m4, m1, 8); \
} while(0)
#define LOAD_MSG_8_3(b0, b1) \
do \
{ \
b0 = m6; \
b1 = _mm_alignr_epi8(m5, m0, 8); \
} while(0)
#define LOAD_MSG_8_4(b0, b1) \
do \
{ \
b0 = _mm_blend_epi16(m1, m3, 0xF0); \
b1 = m2; \
} while(0)
#define LOAD_MSG_9_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m5, m4); \
b1 = _mm_unpackhi_epi64(m3, m0); \
} while(0)
#define LOAD_MSG_9_2(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m1, m2); \
b1 = _mm_blend_epi16(m3, m2, 0xF0); \
} while(0)
#define LOAD_MSG_9_3(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m7, m4); \
b1 = _mm_unpackhi_epi64(m1, m6); \
} while(0)
#define LOAD_MSG_9_4(b0, b1) \
do \
{ \
b0 = _mm_alignr_epi8(m7, m5, 8); \
b1 = _mm_unpacklo_epi64(m6, m0); \
} while(0)
#define LOAD_MSG_10_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m0, m1); \
b1 = _mm_unpacklo_epi64(m2, m3); \
} while(0)
#define LOAD_MSG_10_2(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m0, m1); \
b1 = _mm_unpackhi_epi64(m2, m3); \
} while(0)
#define LOAD_MSG_10_3(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m4, m5); \
b1 = _mm_unpacklo_epi64(m6, m7); \
} while(0)
#define LOAD_MSG_10_4(b0, b1) \
do \
{ \
b0 = _mm_unpackhi_epi64(m4, m5); \
b1 = _mm_unpackhi_epi64(m6, m7); \
} while(0)
#define LOAD_MSG_11_1(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m7, m2); \
b1 = _mm_unpackhi_epi64(m4, m6); \
} while(0)
#define LOAD_MSG_11_2(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m5, m4); \
b1 = _mm_alignr_epi8(m3, m7, 8); \
} while(0)
#define LOAD_MSG_11_3(b0, b1) \
do \
{ \
b0 = _mm_shuffle_epi32(m0, _MM_SHUFFLE(1,0,3,2)); \
b1 = _mm_unpackhi_epi64(m5, m2); \
} while(0)
#define LOAD_MSG_11_4(b0, b1) \
do \
{ \
b0 = _mm_unpacklo_epi64(m6, m1); \
b1 = _mm_unpackhi_epi64(m3, m1); \
} while(0)
#endif
+170
View File
@@ -0,0 +1,170 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#pragma once
#ifndef __BLAKE2B_ROUND_H__
#define __BLAKE2B_ROUND_H__
#define LOAD(p) _mm_load_si128( (const __m128i *)(p) )
#define STORE(p,r) _mm_store_si128((__m128i *)(p), r)
#define LOADU(p) _mm_loadu_si128( (const __m128i *)(p) )
#define STOREU(p,r) _mm_storeu_si128((__m128i *)(p), r)
#define TOF(reg) _mm_castsi128_ps((reg))
#define TOI(reg) _mm_castps_si128((reg))
#define LIKELY(x) __builtin_expect((x),1)
/* Microarchitecture-specific macros */
#ifndef HAVE_XOP
#ifdef HAVE_SSSE3
#define _mm_roti_epi64(x, c) \
(-(c) == 32) ? _mm_shuffle_epi32((x), _MM_SHUFFLE(2,3,0,1)) \
: (-(c) == 24) ? _mm_shuffle_epi8((x), r24) \
: (-(c) == 16) ? _mm_shuffle_epi8((x), r16) \
: (-(c) == 63) ? _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_add_epi64((x), (x))) \
: _mm_xor_si128(_mm_srli_epi64((x), -(c)), _mm_slli_epi64((x), 64-(-(c))))
#else
#define _mm_roti_epi64(r, c) _mm_xor_si128(_mm_srli_epi64( (r), -(c) ),_mm_slli_epi64( (r), 64-(-c) ))
#endif
#else
/* ... */
#endif
#define G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
\
row4l = _mm_xor_si128(row4l, row1l); \
row4h = _mm_xor_si128(row4h, row1h); \
\
row4l = _mm_roti_epi64(row4l, (-32)); \
row4h = _mm_roti_epi64(row4h, (-32)); \
\
row3l = _mm_add_epi64(row3l, row4l); \
row3h = _mm_add_epi64(row3h, row4h); \
\
row2l = _mm_xor_si128(row2l, row3l); \
row2h = _mm_xor_si128(row2h, row3h); \
\
row2l = _mm_roti_epi64(row2l, (-24)); \
row2h = _mm_roti_epi64(row2h, (-24)); \
#define G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1) \
row1l = _mm_add_epi64(_mm_add_epi64(row1l, b0), row2l); \
row1h = _mm_add_epi64(_mm_add_epi64(row1h, b1), row2h); \
\
row4l = _mm_xor_si128(row4l, row1l); \
row4h = _mm_xor_si128(row4h, row1h); \
\
row4l = _mm_roti_epi64(row4l, (-16)); \
row4h = _mm_roti_epi64(row4h, (-16)); \
\
row3l = _mm_add_epi64(row3l, row4l); \
row3h = _mm_add_epi64(row3h, row4h); \
\
row2l = _mm_xor_si128(row2l, row3l); \
row2h = _mm_xor_si128(row2h, row3h); \
\
row2l = _mm_roti_epi64(row2l, (-63)); \
row2h = _mm_roti_epi64(row2h, (-63)); \
#if defined(HAVE_SSSE3)
#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = _mm_alignr_epi8(row2h, row2l, 8); \
t1 = _mm_alignr_epi8(row2l, row2h, 8); \
row2l = t0; \
row2h = t1; \
\
t0 = row3l; \
row3l = row3h; \
row3h = t0; \
\
t0 = _mm_alignr_epi8(row4h, row4l, 8); \
t1 = _mm_alignr_epi8(row4l, row4h, 8); \
row4l = t1; \
row4h = t0;
#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = _mm_alignr_epi8(row2l, row2h, 8); \
t1 = _mm_alignr_epi8(row2h, row2l, 8); \
row2l = t0; \
row2h = t1; \
\
t0 = row3l; \
row3l = row3h; \
row3h = t0; \
\
t0 = _mm_alignr_epi8(row4l, row4h, 8); \
t1 = _mm_alignr_epi8(row4h, row4l, 8); \
row4l = t1; \
row4h = t0;
#else
#define DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = row4l;\
t1 = row2l;\
row4l = row3l;\
row3l = row3h;\
row3h = row4l;\
row4l = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t0, t0)); \
row4h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row4h, row4h)); \
row2l = _mm_unpackhi_epi64(row2l, _mm_unpacklo_epi64(row2h, row2h)); \
row2h = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(t1, t1))
#define UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h) \
t0 = row3l;\
row3l = row3h;\
row3h = t0;\
t0 = row2l;\
t1 = row4l;\
row2l = _mm_unpackhi_epi64(row2h, _mm_unpacklo_epi64(row2l, row2l)); \
row2h = _mm_unpackhi_epi64(t0, _mm_unpacklo_epi64(row2h, row2h)); \
row4l = _mm_unpackhi_epi64(row4l, _mm_unpacklo_epi64(row4h, row4h)); \
row4h = _mm_unpackhi_epi64(row4h, _mm_unpacklo_epi64(t1, t1))
#endif
#if defined(HAVE_SSE41)
#include "blake2b-load-sse41.h"
#else
#include "blake2b-load-sse2.h"
#endif
#define ROUND(r) \
LOAD_MSG_ ##r ##_1(b0, b1); \
G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
LOAD_MSG_ ##r ##_2(b0, b1); \
G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
DIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h); \
LOAD_MSG_ ##r ##_3(b0, b1); \
G1(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
LOAD_MSG_ ##r ##_4(b0, b1); \
G2(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h,b0,b1); \
UNDIAGONALIZE(row1l,row2l,row3l,row4l,row1h,row2h,row3h,row4h);
#endif
#define BLAKE2_ROUND(row1l,row1h,row2l,row2h,row3l,row3h,row4l,row4h) \
G1(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h); \
G2(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h); \
\
DIAGONALIZE(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h); \
\
G1(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h); \
G2(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h); \
\
UNDIAGONALIZE(row1l, row2l, row3l, row4l, row1h, row2h, row3h, row4h);
+262
View File
@@ -0,0 +1,262 @@
/*
BLAKE2 reference source code package - optimized C implementations
Written in 2012 by Samuel Neves <sneves@dei.uc.pt>
To the extent possible under law, the author(s) have dedicated all copyright
and related and neighboring rights to this software to the public domain
worldwide. This software is distributed without any warranty.
You should have received a copy of the CC0 Public Domain Dedication along with
this software. If not, see <http://creativecommons.org/publicdomain/zero/1.0/>.
*/
#include <stdint.h>
#include <string.h>
#include <stdio.h>
#include "blake2.h"
#include "blake2-impl.h"
#include "blake2-config.h"
#ifdef WIN32
#include <intrin.h>
#endif
#include <emmintrin.h>
#if defined(HAVE_SSSE3)
#include <tmmintrin.h>
#endif
#if defined(HAVE_SSE41)
#include <smmintrin.h>
#endif
#if defined(HAVE_AVX)
#include <immintrin.h>
#endif
#if defined(HAVE_XOP)
#include <x86intrin.h>
#endif
#include "blake2b-round.h"
ALIGN(64) static const uint64_t blake2b_IV[8] = {
0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
};
/* init xors IV with input parameter block */
int eq_blake2b_init_param(blake2b_state *S, const blake2b_param *P)
{
//blake2b_init0( S );
const uint8_t * v = (const uint8_t *)(blake2b_IV);
const uint8_t * p = (const uint8_t *)(P);
uint8_t * h = (uint8_t *)(S->h);
/* IV XOR ParamBlock */
memset(S, 0, sizeof(blake2b_state));
for (int i = 0; i < BLAKE2B_OUTBYTES; ++i) h[i] = v[i] ^ p[i];
return 0;
}
/* Some sort of default parameter block initialization, for sequential blake2b */
int eq_blake2b_init(blake2b_state *S, const uint8_t outlen)
{
if ((!outlen) || (outlen > BLAKE2B_OUTBYTES)) return -1;
const blake2b_param P =
{
outlen,
0,
1,
1,
0,
0,
0,
0,
{ 0 },
{ 0 },
{ 0 }
};
return eq_blake2b_init_param(S, &P);
}
int eq_blake2b_init_key(blake2b_state *S, const uint8_t outlen, const void *key, const uint8_t keylen)
{
if ((!outlen) || (outlen > BLAKE2B_OUTBYTES)) return -1;
if ((!keylen) || keylen > BLAKE2B_KEYBYTES) return -1;
const blake2b_param P =
{
outlen,
keylen,
1,
1,
0,
0,
0,
0,
{ 0 },
{ 0 },
{ 0 }
};
if (eq_blake2b_init_param(S, &P) < 0)
return 0;
{
uint8_t block[BLAKE2B_BLOCKBYTES];
memset(block, 0, BLAKE2B_BLOCKBYTES);
memcpy(block, key, keylen);
eq_blake2b_update(S, block, BLAKE2B_BLOCKBYTES);
secure_zero_memory(block, BLAKE2B_BLOCKBYTES); /* Burn the key from stack */
}
return 0;
}
static inline int blake2b_compress(blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES])
{
__m128i row1l, row1h;
__m128i row2l, row2h;
__m128i row3l, row3h;
__m128i row4l, row4h;
__m128i b0, b1;
__m128i t0, t1;
#if defined(HAVE_SSSE3) && !defined(HAVE_XOP)
const __m128i r16 = _mm_setr_epi8(2, 3, 4, 5, 6, 7, 0, 1, 10, 11, 12, 13, 14, 15, 8, 9);
const __m128i r24 = _mm_setr_epi8(3, 4, 5, 6, 7, 0, 1, 2, 11, 12, 13, 14, 15, 8, 9, 10);
#endif
#if defined(HAVE_SSE41)
const __m128i m0 = LOADU(block + 00);
const __m128i m1 = LOADU(block + 16);
const __m128i m2 = LOADU(block + 32);
const __m128i m3 = LOADU(block + 48);
const __m128i m4 = LOADU(block + 64);
const __m128i m5 = LOADU(block + 80);
const __m128i m6 = LOADU(block + 96);
const __m128i m7 = LOADU(block + 112);
#else
const uint64_t m0 = ( ( uint64_t * )block )[ 0];
const uint64_t m1 = ( ( uint64_t * )block )[ 1];
const uint64_t m2 = ( ( uint64_t * )block )[ 2];
const uint64_t m3 = ( ( uint64_t * )block )[ 3];
const uint64_t m4 = ( ( uint64_t * )block )[ 4];
const uint64_t m5 = ( ( uint64_t * )block )[ 5];
const uint64_t m6 = ( ( uint64_t * )block )[ 6];
const uint64_t m7 = ( ( uint64_t * )block )[ 7];
const uint64_t m8 = ( ( uint64_t * )block )[ 8];
const uint64_t m9 = ( ( uint64_t * )block )[ 9];
const uint64_t m10 = ( ( uint64_t * )block )[10];
const uint64_t m11 = ( ( uint64_t * )block )[11];
const uint64_t m12 = ( ( uint64_t * )block )[12];
const uint64_t m13 = ( ( uint64_t * )block )[13];
const uint64_t m14 = ( ( uint64_t * )block )[14];
const uint64_t m15 = ( ( uint64_t * )block )[15];
#endif
row1l = LOADU(&S->h[0]);
row1h = LOADU(&S->h[2]);
row2l = LOADU(&S->h[4]);
row2h = LOADU(&S->h[6]);
row3l = LOADU(&blake2b_IV[0]);
row3h = LOADU(&blake2b_IV[2]);
row4l = _mm_xor_si128(LOADU(&blake2b_IV[4]), _mm_set_epi32(0, 0, 0, S->counter));
row4h = _mm_xor_si128(LOADU(&blake2b_IV[6]), _mm_set_epi32(0, 0, 0L - S->lastblock, 0L - S->lastblock));
ROUND(0);
ROUND(1);
ROUND(2);
ROUND(3);
ROUND(4);
ROUND(5);
ROUND(6);
ROUND(7);
ROUND(8);
ROUND(9);
ROUND(10);
ROUND(11);
row1l = _mm_xor_si128(row3l, row1l);
row1h = _mm_xor_si128(row3h, row1h);
STOREU(&S->h[0], _mm_xor_si128(LOADU(&S->h[0]), row1l));
STOREU(&S->h[2], _mm_xor_si128(LOADU(&S->h[2]), row1h));
row2l = _mm_xor_si128(row4l, row2l);
row2h = _mm_xor_si128(row4h, row2h);
STOREU(&S->h[4], _mm_xor_si128(LOADU(&S->h[4]), row2l));
STOREU(&S->h[6], _mm_xor_si128(LOADU(&S->h[6]), row2h));
return 0;
}
int eq_blake2b_update(blake2b_state *S, const uint8_t *in, uint64_t inlen)
{
while (inlen > 0)
{
size_t left = S->buflen;
size_t fill = BLAKE2B_BLOCKBYTES - left;
if (inlen > fill)
{
memcpy(S->buf + left, in, fill); // Fill buffer
in += fill;
inlen -= fill;
S->counter += BLAKE2B_BLOCKBYTES;
blake2b_compress(S, S->buf); // Compress
S->buflen = 0;
}
else // inlen <= fill
{
memcpy(S->buf + left, in, inlen);
S->buflen += (uint8_t) inlen; // not enough to compress
in += inlen;
inlen = 0;
}
}
return 0;
}
int eq_blake2b_final(blake2b_state *S, uint8_t *out, uint8_t outlen)
{
if (outlen > BLAKE2B_OUTBYTES)
return -1;
if (S->buflen > BLAKE2B_BLOCKBYTES)
{
S->counter += BLAKE2B_BLOCKBYTES;
blake2b_compress(S, S->buf);
S->buflen -= BLAKE2B_BLOCKBYTES;
memcpy(S->buf, S->buf + BLAKE2B_BLOCKBYTES, S->buflen);
}
S->counter += S->buflen;
S->lastblock = 1;
memset(S->buf + S->buflen, 0, BLAKE2B_BLOCKBYTES - S->buflen); /* Padding */
blake2b_compress(S, S->buf);
memcpy(out, &S->h[0], outlen);
S->lastblock = 0;
return 0;
}
int eq_blake2b(uint8_t *out, const void *in, const void *key, const uint8_t outlen, const uint64_t inlen, uint8_t keylen)
{
blake2b_state S[1];
/* Verify parameters */
if (!in || !out) return -1;
if (NULL == key) keylen = 0;
if (keylen)
{
if (eq_blake2b_init_key(S, outlen, key, keylen) < 0) return -1;
}
else
{
if (eq_blake2b_init(S, outlen) < 0) return -1;
}
eq_blake2b_update(S, (const uint8_t *)in, inlen);
eq_blake2b_final(S, out, outlen);
return 0;
}
+2130
View File
File diff suppressed because it is too large. Load diff
+136
View File
@@ -0,0 +1,136 @@
#pragma once
#include <stdint.h>
#include <string.h>
#include <stdlib.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdexcept>
#include <vector>
#include <cuda.h>
#include <cuda_runtime.h>
#include <device_launch_parameters.h>
#ifdef WIN32
#define _SNPRINTF _snprintf
#else
#define _SNPRINTF snprintf
#endif
#ifndef nullptr
#define nullptr NULL
#endif
#ifdef WIN32
#define rt_error std::runtime_error
#else
class rt_error : public std::runtime_error
{
public:
explicit rt_error(const std::string& str) : std::runtime_error(str) {}
};
#endif
#define checkCudaErrors(call) \
do { \
cudaError_t err = call; \
if (cudaSuccess != err) { \
char errorBuff[512]; \
_SNPRINTF(errorBuff, sizeof(errorBuff) - 1, \
"CUDA error '%s' in func '%s' line %d", \
cudaGetErrorString(err), __FUNCTION__, __LINE__); \
throw rt_error(errorBuff); \
} \
} while (0)
#define checkCudaDriverErrors(call) \
do { \
CUresult err = call; \
if (CUDA_SUCCESS != err) { \
char errorBuff[512]; \
_SNPRINTF(errorBuff, sizeof(errorBuff) - 1, \
"CUDA error DRIVER: '%d' in func '%s' line %d", \
err, __FUNCTION__, __LINE__); \
throw rt_error(errorBuff); \
} \
} while (0)
typedef uint64_t u64;
typedef uint32_t u32;
typedef uint16_t u16;
typedef uint8_t u8;
typedef unsigned char uchar;
struct packer_default;
struct packer_cantor;
#define MAXREALSOLS 9
struct scontainerreal {
u32 sols[MAXREALSOLS][512];
u32 nsols;
};
#if 0
#include <functional>
#define fn_solution std::function<void(int thr_id, const std::vector<uint32_t>&, size_t, const unsigned char*)>
#define fn_hashdone std::function<void(int thr_id)>
#define fn_cancel std::function<bool(int thr_id)>
#else
typedef void (*fn_solution)(int thr_id, const std::vector<uint32_t>&, size_t, const unsigned char*);
typedef void (*fn_hashdone)(int thr_id);
typedef bool (*fn_cancel)(int thr_id);
#endif
template <u32 RB, u32 SM> struct equi;
// ---------------------------------------------------------------------------------------------------
struct eq_cuda_context_interface
{
//virtual ~eq_cuda_context_interface();
virtual void solve(const char *tequihash_header,
unsigned int tequihash_header_len,
const char* nonce,
unsigned int nonce_len,
fn_cancel cancelf,
fn_solution solutionf,
fn_hashdone hashdonef);
public:
int thread_id;
int device_id;
int throughput;
int totalblocks;
int threadsperblock;
int threadsperblock_digits;
size_t equi_mem_sz;
};
// ---------------------------------------------------------------------------------------------------
template <u32 RB, u32 SM, u32 SSM, u32 THREADS, typename PACKER>
class eq_cuda_context : public eq_cuda_context_interface
{
equi<RB, SM>* device_eq;
scontainerreal* solutions;
CUcontext pctx;
void solve(const char *tequihash_header,
unsigned int tequihash_header_len,
const char* nonce,
unsigned int nonce_len,
fn_cancel cancelf,
fn_solution solutionf,
fn_hashdone hashdonef);
public:
eq_cuda_context(int thr_id, int dev_id);
void freemem();
~eq_cuda_context();
};
// RB, SM, SSM, TPB, PACKER... but any change only here will fail..
#define CONFIG_MODE_1 9, 1248, 12, 640, packer_cantor
//#define CONFIG_MODE_2 8, 640, 12, 512, packer_default
+275
View File
@@ -0,0 +1,275 @@
/**
* Equihash specific stratum protocol
* tpruvot@github - 2017 - Part under GPLv3 Licence
*/
#include <errno.h>
#include <string.h>
#include <unistd.h>
#include <miner.h>
#include "equihash.h"
extern struct stratum_ctx stratum;
extern pthread_mutex_t stratum_work_lock;
// ZEC uses a different scale to compute diff...
// sample targets to diff (stored in the reverse byte order in work->target)
// 0007fff800000000000000000000000000000000000000000000000000000000 is stratum diff 32
// 003fffc000000000000000000000000000000000000000000000000000000000 is stratum diff 4
// 00ffff0000000000000000000000000000000000000000000000000000000000 is stratum diff 1
double target_to_diff_equi(uint32_t* target)
{
uchar* tgt = (uchar*) target;
uint64_t m =
(uint64_t)tgt[30] << 24 |
(uint64_t)tgt[29] << 16 |
(uint64_t)tgt[28] << 8 |
(uint64_t)tgt[27] << 0;
if (!m)
return 0.;
else
return (double)0xffff0000UL/m;
}
void diff_to_target_equi(uint32_t *target, double diff)
{
uint64_t m;
int k;
for (k = 6; k > 0 && diff > 1.0; k--)
diff /= 4294967296.0;
m = (uint64_t)(4294901760.0 / diff);
if (m == 0 && k == 6)
memset(target, 0xff, 32);
else {
memset(target, 0, 32);
target[k + 1] = (uint32_t)(m >> 8);
target[k + 2] = (uint32_t)(m >> 40);
//memset(target, 0xff, 6*sizeof(uint32_t));
for (k = 0; k < 28 && ((uint8_t*)target)[k] == 0; k++)
((uint8_t*)target)[k] = 0xff;
}
}
/* compute nbits to get the network diff */
double equi_network_diff(struct work *work)
{
//KMD bits: "1e 015971",
//KMD target: "00 00 015971000000000000000000000000000000000000000000000000000000",
//KMD bits: "1d 686aaf",
//KMD target: "00 0000 686aaf0000000000000000000000000000000000000000000000000000",
uint32_t nbits = work->data[26];
uint32_t bits = (nbits & 0xffffff);
int16_t shift = (swab32(nbits) & 0xff);
shift = (31 - shift) * 8; // 8 bits shift for 0x1e, 16 for 0x1d
uint64_t tgt64 = swab32(bits);
tgt64 = tgt64 << shift;
// applog_hex(&tgt64, 8);
uint8_t net_target[32] = { 0 };
for (int b=0; b<8; b++)
net_target[31-b] = ((uint8_t*)&tgt64)[b];
// applog_hex(net_target, 32);
double d = target_to_diff_equi((uint32_t*)net_target);
return d;
}
void equi_work_set_target(struct work* work, double diff)
{
// target is given as data by the equihash stratum
// memcpy(work->target, stratum.job.claim, 32); // claim field is only used for lbry
diff_to_target_equi(work->target, diff);
//applog(LOG_BLUE, "diff %f to target :", diff);
//applog_hex(work->target, 32);
work->targetdiff = diff;
}
bool equi_stratum_set_target(struct stratum_ctx *sctx, json_t *params)
{
uint8_t target_bin[32], target_be[32];
const char *target_hex = json_string_value(json_array_get(params, 0));
if (!target_hex || strlen(target_hex) == 0)
return false;
hex2bin(target_bin, target_hex, 32);
memset(target_be, 0xff, 32);
int filled = 0;
for (int i=0; i<32; i++) {
if (filled == 3) break;
target_be[31-i] = target_bin[i];
if (target_bin[i]) filled++;
}
memcpy(sctx->job.extra, target_be, 32);
pthread_mutex_lock(&stratum_work_lock);
sctx->next_diff = target_to_diff_equi((uint32_t*) &target_be);
pthread_mutex_unlock(&stratum_work_lock);
//applog(LOG_BLUE, "low diff %f", sctx->next_diff);
//applog_hex(target_be, 32);
return true;
}
bool equi_stratum_notify(struct stratum_ctx *sctx, json_t *params)
{
const char *job_id, *version, *prevhash, *coinb1, *coinb2, *nbits, *stime;
size_t coinb1_size, coinb2_size;
bool clean, ret = false;
int ntime, i, p=0;
job_id = json_string_value(json_array_get(params, p++));
version = json_string_value(json_array_get(params, p++));
prevhash = json_string_value(json_array_get(params, p++));
coinb1 = json_string_value(json_array_get(params, p++)); //merkle
coinb2 = json_string_value(json_array_get(params, p++)); //blank (reserved)
stime = json_string_value(json_array_get(params, p++));
nbits = json_string_value(json_array_get(params, p++));
clean = json_is_true(json_array_get(params, p)); p++;
if (!job_id || !prevhash || !coinb1 || !coinb2 || !version || !nbits || !stime ||
strlen(prevhash) != 64 || strlen(version) != 8 ||
strlen(coinb1) != 64 || strlen(coinb2) != 64 ||
strlen(nbits) != 8 || strlen(stime) != 8) {
applog(LOG_ERR, "Stratum notify: invalid parameters");
goto out;
}
/* store stratum server time diff */
hex2bin((uchar *)&ntime, stime, 4);
ntime = ntime - (int) time(0);
if (ntime > sctx->srvtime_diff) {
sctx->srvtime_diff = ntime;
if (opt_protocol && ntime > 20)
applog(LOG_DEBUG, "stratum time is at least %ds in the future", ntime);
}
pthread_mutex_lock(&stratum_work_lock);
hex2bin(sctx->job.version, version, 4);
hex2bin(sctx->job.prevhash, prevhash, 32);
coinb1_size = strlen(coinb1) / 2;
coinb2_size = strlen(coinb2) / 2;
sctx->job.coinbase_size = coinb1_size + coinb2_size + // merkle + reserved
sctx->xnonce1_size + sctx->xnonce2_size; // extranonce and...
sctx->job.coinbase = (uchar*) realloc(sctx->job.coinbase, sctx->job.coinbase_size);
hex2bin(sctx->job.coinbase, coinb1, coinb1_size);
hex2bin(sctx->job.coinbase + coinb1_size, coinb2, coinb2_size);
sctx->job.xnonce2 = sctx->job.coinbase + coinb1_size + coinb2_size + sctx->xnonce1_size;
if (!sctx->job.job_id || strcmp(sctx->job.job_id, job_id))
memset(sctx->job.xnonce2, 0, sctx->xnonce2_size);
memcpy(sctx->job.coinbase + coinb1_size + coinb2_size, sctx->xnonce1, sctx->xnonce1_size);
for (i = 0; i < sctx->job.merkle_count; i++)
free(sctx->job.merkle[i]);
free(sctx->job.merkle);
sctx->job.merkle = NULL;
sctx->job.merkle_count = 0;
free(sctx->job.job_id);
sctx->job.job_id = strdup(job_id);
hex2bin(sctx->job.nbits, nbits, 4);
hex2bin(sctx->job.ntime, stime, 4);
sctx->job.clean = clean;
sctx->job.diff = sctx->next_diff;
pthread_mutex_unlock(&stratum_work_lock);
ret = true;
out:
return ret;
}
// equihash stratum protocol is not standard, use client.show_message to pass block height
bool equi_stratum_show_message(struct stratum_ctx *sctx, json_t *id, json_t *params)
{
char *s;
json_t *val;
bool ret;
val = json_array_get(params, 0);
if (val) {
const char* data = json_string_value(val);
if (data && strlen(data)) {
char symbol[32] = { 0 };
uint32_t height = 0;
int ss = sscanf(data, "equihash %s block %u", symbol, &height);
if (height && ss > 1) sctx->job.height = height;
if (opt_debug && ss > 1) applog(LOG_DEBUG, "%s", data);
}
}
if (!id || json_is_null(id))
return true;
val = json_object();
json_object_set(val, "id", id);
json_object_set_new(val, "error", json_null());
json_object_set_new(val, "result", json_true());
s = json_dumps(val, 0);
ret = stratum_send_line(sctx, s);
json_decref(val);
free(s);
return ret;
}
void equi_store_work_solution(struct work* work, uint32_t* hash, void* sol_data)
{
int nonce = work->valid_nonces-1;
memcpy(work->extra, sol_data, 1347);
bn_store_hash_target_ratio(hash, work->target, work, nonce);
//work->sharediff[nonce] = target_to_diff_equi(hash);
}
#define JSON_SUBMIT_BUF_LEN (4*1024)
// called by submit_upstream_work()
bool equi_stratum_submit(struct pool_infos *pool, struct work *work)
{
char _ALIGN(64) s[JSON_SUBMIT_BUF_LEN];
char _ALIGN(64) timehex[16] = { 0 };
char *jobid, *noncestr, *solhex;
int idnonce = work->submit_nonce_id;
// scanned nonce
work->data[EQNONCE_OFFSET] = work->nonces[idnonce];
unsigned char * nonce = (unsigned char*) (&work->data[27]);
size_t nonce_len = 32 - stratum.xnonce1_size;
// long nonce without pool prefix (extranonce)
noncestr = bin2hex(&nonce[stratum.xnonce1_size], nonce_len);
solhex = (char*) calloc(1, 1344*2 + 64);
if (!solhex || !noncestr) {
applog(LOG_ERR, "unable to alloc share memory");
return false;
}
cbin2hex(solhex, (const char*) work->extra, 1347);
jobid = work->job_id + 8;
sprintf(timehex, "%08x", swab32(work->data[25]));
snprintf(s, sizeof(s), "{\"method\":\"mining.submit\",\"params\":"
"[\"%s\",\"%s\",\"%s\",\"%s\",\"%s\"], \"id\":%u}",
pool->user, jobid, timehex, noncestr, solhex,
stratum.job.shares_count + 10);
free(solhex);
free(noncestr);
gettimeofday(&stratum.tv_submit, NULL);
if(!stratum_send_line(&stratum, s)) {
applog(LOG_ERR, "%s stratum_send_line failed", __func__);
return false;
}
stratum.sharediff = work->sharediff[idnonce];
stratum.job.shares_count++;
return true;
}
+171
View File
@@ -0,0 +1,171 @@
/*
* Port to Generic C of C++ implementation of the Equihash Proof-of-Work
* algorithm from zcashd.
*
* Copyright (c) 2016 abc at openwall dot com
* Copyright (c) 2016 Jack Grigg
* Copyright (c) 2016 The Zcash developers
* Copyright (c) 2017 tpruvot
*
* Distributed under the MIT software license, see the accompanying
* file COPYING or http://www.opensource.org/licenses/mit-license.php.
*/
#include <string.h>
#include <stdint.h>
#include <stdbool.h>
#include <assert.h>
#include "equihash.h"
//#define USE_LIBSODIUM
#ifdef USE_LIBSODIUM
#include "sodium.h"
#define blake2b_state crypto_generichash_blake2b_state
#else
#include "blake2/blake2.h"
#define be32toh(x) swab32(x)
#define htole32(x) (x)
#define HASHOUT 50
#endif
#include <miner.h>
static void digestInit(blake2b_state *S, const uint32_t n, const uint32_t k)
{
uint32_t le_N = htole32(n);
uint32_t le_K = htole32(k);
#ifdef USE_LIBSODIUM
uint8_t personalization[crypto_generichash_blake2b_PERSONALBYTES] = { 0 };
memcpy(personalization, "ZcashPoW", 8);
memcpy(personalization + 8, &le_N, 4);
memcpy(personalization + 12, &le_K, 4);
crypto_generichash_blake2b_init_salt_personal(S,
NULL, 0, (512 / n) * n / 8, NULL, personalization);
#else
unsigned char personal[] = "ZcashPoW01230123";
memcpy(personal + 8, &le_N, 4);
memcpy(personal + 12, &le_K, 4);
blake2b_param P[1];
P->digest_length = HASHOUT;
P->key_length = 0;
P->fanout = 1;
P->depth = 1;
P->leaf_length = 0;
P->node_offset = 0;
P->node_depth = 0;
P->inner_length = 0;
memset(P->reserved, 0, sizeof(P->reserved));
memset(P->salt, 0, sizeof(P->salt));
memcpy(P->personal, (const uint8_t *)personal, 16);
eq_blake2b_init_param(S, P);
#endif
}
static void expandArray(const unsigned char *in, const uint32_t in_len,
unsigned char *out, const uint32_t out_len,
const uint32_t bit_len, const uint32_t byte_pad)
{
assert(bit_len >= 8);
assert(8 * sizeof(uint32_t) >= 7 + bit_len);
const uint32_t out_width = (bit_len + 7) / 8 + byte_pad;
assert(out_len == 8 * out_width * in_len / bit_len);
const uint32_t bit_len_mask = ((uint32_t)1 << bit_len) - 1;
// The acc_bits least-significant bits of acc_value represent a bit sequence
// in big-endian order.
uint32_t acc_bits = 0;
uint32_t acc_value = 0;
uint32_t j = 0;
for (uint32_t i = 0; i < in_len; i++)
{
acc_value = (acc_value << 8) | in[i];
acc_bits += 8;
// When we have bit_len or more bits in the accumulator, write the next
// output element.
if (acc_bits >= bit_len) {
acc_bits -= bit_len;
for (uint32_t x = 0; x < byte_pad; x++) {
out[j + x] = 0;
}
for (uint32_t x = byte_pad; x < out_width; x++) {
out[j + x] = (
// Big-endian
acc_value >> (acc_bits + (8 * (out_width - x - 1)))
) & (
// Apply bit_len_mask across byte boundaries
(bit_len_mask >> (8 * (out_width - x - 1))) & 0xFF
);
}
j += out_width;
}
}
}
static void generateHash(blake2b_state *S, const uint32_t g, uint8_t *hash, const size_t hashLen)
{
const uint32_t le_g = htole32(g);
blake2b_state digest = *S; /* copy */
#ifdef USE_LIBSODIUM
crypto_generichash_blake2b_update(&digest, (uint8_t *)&le_g, sizeof(le_g));
crypto_generichash_blake2b_final(&digest, hash, hashLen);
#else
eq_blake2b_update(&digest, (const uint8_t*) &le_g, sizeof(le_g));
eq_blake2b_final(&digest, hash, (uint8_t) (hashLen & 0xFF));
#endif
}
static int isZero(const uint8_t *hash, size_t len)
{
// This doesn't need to be constant time.
for (size_t i = 0; i < len; i++) {
if (hash[i] != 0) return 0;
}
return 1;
}
// hdr -> header including nonce (140 bytes)
// soln -> equihash solution (excluding 3 bytes with size, so 1344 bytes length)
bool equi_verify(uint8_t* const hdr, uint8_t* const soln)
{
const uint32_t n = WN; // 200
const uint32_t k = WK; // 9
const uint32_t collisionBitLength = n / (k + 1);
const uint32_t collisionByteLength = (collisionBitLength + 7) / 8;
const uint32_t hashLength = (k + 1) * collisionByteLength;
const uint32_t indicesPerHashOutput = 512 / n;
const uint32_t hashOutput = indicesPerHashOutput * n / 8;
const uint32_t equihashSolutionSize = (1 << k) * (n / (k + 1) + 1) / 8;
const uint32_t solnr = 1 << k;
uint32_t indices[512] = { 0 };
uint8_t vHash[hashLength] = { 0 };
blake2b_state state;
digestInit(&state, n, k);
#ifdef USE_LIBSODIUM
crypto_generichash_blake2b_update(&state, hdr, 140);
#else
eq_blake2b_update(&state, hdr, 140);
#endif
expandArray(soln, equihashSolutionSize, (uint8_t*) &indices, sizeof(indices), collisionBitLength + 1, 1);
for (uint32_t j = 0; j < solnr; j++) {
uint8_t tmpHash[hashOutput];
uint8_t hash[hashLength];
uint32_t i = be32toh(indices[j]);
generateHash(&state, i / indicesPerHashOutput, tmpHash, hashOutput);
expandArray(tmpHash + (i % indicesPerHashOutput * n / 8), n / 8, hash, hashLength, collisionBitLength, 0);
for (uint32_t k = 0; k < hashLength; k++)
vHash[k] ^= hash[k];
}
return isZero(vHash, sizeof(vHash));
}
+296
View File
@@ -0,0 +1,296 @@
/**
* Equihash solver interface for ccminer (compatible with linux and windows)
* Solver taken from nheqminer, by djeZo (and NiceHash)
* tpruvot - 2017 (GPL v3)
*/
#include <stdio.h>
#include <unistd.h>
#include <assert.h>
#include <stdexcept>
#include <vector>
#include <sph/sph_sha2.h>
#include "eqcuda.hpp"
#include "equihash.h" // equi_verify()
#include <miner.h>
// All solutions (BLOCK_HEADER_LEN + SOLSIZE_LEN + SOL_LEN) sha256d should be under the target
extern "C" void equi_hash(const void* input, void* output, int len)
{
uint8_t _ALIGN(64) hash0[32], hash1[32];
sph_sha256_context ctx_sha256;
sph_sha256_init(&ctx_sha256);
sph_sha256(&ctx_sha256, input, len);
sph_sha256_close(&ctx_sha256, hash0);
sph_sha256(&ctx_sha256, hash0, 32);
sph_sha256_close(&ctx_sha256, hash1);
memcpy(output, hash1, 32);
}
// input here is 140 for the header and 1344 for the solution (equi.cpp)
extern "C" int equi_verify_sol(void * const hdr, void * const sol)
{
bool res = equi_verify((uint8_t*) hdr, (uint8_t*) sol);
//applog_hex((void*)hdr, 140);
//applog_hex((void*)sol, 1344);
return res ? 1 : 0;
}
#include <cuda_helper.h>
//#define EQNONCE_OFFSET 30 /* 27:34 */
#define NONCE_OFT EQNONCE_OFFSET
static bool init[MAX_GPUS] = { 0 };
static int valid_sols[MAX_GPUS] = { 0 };
static uint8_t _ALIGN(64) data_sols[MAX_GPUS][MAXREALSOLS][1536] = { 0 }; // 140+3+1344 required
static eq_cuda_context_interface* solvers[MAX_GPUS] = { NULL };
static void CompressArray(const unsigned char* in, size_t in_len,
unsigned char* out, size_t out_len, size_t bit_len, size_t byte_pad)
{
assert(bit_len >= 8);
assert(8 * sizeof(uint32_t) >= 7 + bit_len);
size_t in_width = (bit_len + 7) / 8 + byte_pad;
assert(out_len == bit_len*in_len / (8 * in_width));
uint32_t bit_len_mask = (1UL << bit_len) - 1;
// The acc_bits least-significant bits of acc_value represent a bit sequence
// in big-endian order.
size_t acc_bits = 0;
uint32_t acc_value = 0;
size_t j = 0;
for (size_t i = 0; i < out_len; i++) {
// When we have fewer than 8 bits left in the accumulator, read the next
// input element.
if (acc_bits < 8) {
acc_value = acc_value << bit_len;
for (size_t x = byte_pad; x < in_width; x++) {
acc_value = acc_value | (
(
// Apply bit_len_mask across byte boundaries
in[j + x] & ((bit_len_mask >> (8 * (in_width - x - 1))) & 0xFF)
) << (8 * (in_width - x - 1))); // Big-endian
}
j += in_width;
acc_bits += bit_len;
}
acc_bits -= 8;
out[i] = (acc_value >> acc_bits) & 0xFF;
}
}
#ifndef htobe32
#define htobe32(x) swab32(x)
#endif
static void EhIndexToArray(const u32 i, unsigned char* arr)
{
u32 bei = htobe32(i);
memcpy(arr, &bei, sizeof(u32));
}
static std::vector<unsigned char> GetMinimalFromIndices(std::vector<u32> indices, size_t cBitLen)
{
assert(((cBitLen + 1) + 7) / 8 <= sizeof(u32));
size_t lenIndices = indices.size()*sizeof(u32);
size_t minLen = (cBitLen + 1)*lenIndices / (8 * sizeof(u32));
size_t bytePad = sizeof(u32) - ((cBitLen + 1) + 7) / 8;
std::vector<unsigned char> array(lenIndices);
for (size_t i = 0; i < indices.size(); i++) {
EhIndexToArray(indices[i], array.data() + (i*sizeof(u32)));
}
std::vector<unsigned char> ret(minLen);
CompressArray(array.data(), lenIndices, ret.data(), minLen, cBitLen + 1, bytePad);
return ret;
}
// solver callbacks
static void cb_solution(int thr_id, const std::vector<uint32_t>& solutions, size_t cbitlen, const unsigned char *compressed_sol)
{
std::vector<unsigned char> nSolution;
if (!compressed_sol) {
nSolution = GetMinimalFromIndices(solutions, cbitlen);
} else {
gpulog(LOG_INFO, thr_id, "compressed_sol");
nSolution = std::vector<unsigned char>(1344);
for (size_t i = 0; i < cbitlen; i++)
nSolution[i] = compressed_sol[i];
}
int nsol = valid_sols[thr_id];
if (nsol < 0) nsol = 0;
if(nSolution.size() == 1344) {
// todo, only store solution data here...
le32enc(&data_sols[thr_id][nsol][140], 0x000540fd); // sol sz header
memcpy(&data_sols[thr_id][nsol][143], nSolution.data(), 1344);
valid_sols[thr_id] = nsol + 1;
}
}
static void cb_hashdone(int thr_id) {
if (!valid_sols[thr_id]) valid_sols[thr_id] = -1;
}
static bool cb_cancel(int thr_id) {
if (work_restart[thr_id].restart)
valid_sols[thr_id] = -1;
return work_restart[thr_id].restart;
}
extern "C" int scanhash_equihash(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[35];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[NONCE_OFT];
uint32_t nonce_increment = rand() & 0xFF; // nonce randomizer
struct timeval tv_start, tv_end, diff;
double secs, solps;
uint32_t soluce_count = 0;
if (opt_benchmark)
ptarget[7] = 0xfffff;
if (!init[thr_id]) {
try {
int mode = 1;
switch (mode) {
case 1:
solvers[thr_id] = new eq_cuda_context<CONFIG_MODE_1>(thr_id, device_map[thr_id]);
break;
#ifdef CONFIG_MODE_2
case 2:
solvers[thr_id] = new eq_cuda_context<CONFIG_MODE_2>(thr_id, device_map[thr_id]);
break;
#endif
#ifdef CONFIG_MODE_3
case 3:
solvers[thr_id] = new eq_cuda_context<CONFIG_MODE_3>(thr_id, device_map[thr_id]);
break;
#endif
default:
proper_exit(EXIT_CODE_SW_INIT_ERROR);
return -1;
}
size_t memSz = solvers[thr_id]->equi_mem_sz / (1024*1024);
api_set_throughput(thr_id, (uint32_t) solvers[thr_id]->throughput);
gpulog(LOG_DEBUG, thr_id, "Allocated %u MB of context memory", (u32) memSz);
cuda_get_arch(thr_id);
init[thr_id] = true;
} catch (const std::exception & e) {
CUDA_LOG_ERROR();
gpulog(LOG_ERR, thr_id, "init: %s", e.what());
proper_exit(EXIT_CODE_CUDA_ERROR);
}
}
gettimeofday(&tv_start, NULL);
memcpy(endiandata, pdata, 140);
work->valid_nonces = 0;
do {
try {
valid_sols[thr_id] = 0;
solvers[thr_id]->solve(
(const char *) endiandata, (unsigned int) (140 - 32),
(const char *) &endiandata[27], (unsigned int) 32,
&cb_cancel, &cb_solution, &cb_hashdone
);
*hashes_done = soluce_count;
} catch (const std::exception & e) {
gpulog(LOG_WARNING, thr_id, "solver: %s", e.what());
free_equihash(thr_id);
sleep(1);
return -1;
}
if (valid_sols[thr_id] > 0)
{
const uint32_t Htarg = ptarget[7];
uint32_t _ALIGN(64) vhash[8];
uint8_t _ALIGN(64) full_data[140+3+1344] = { 0 };
uint8_t* sol_data = &full_data[140];
soluce_count += valid_sols[thr_id];
for (int nsol=0; nsol < valid_sols[thr_id]; nsol++)
{
memcpy(full_data, endiandata, 140);
memcpy(sol_data, &data_sols[thr_id][nsol][140], 1347);
equi_hash(full_data, vhash, 140+3+1344);
if (vhash[7] <= Htarg && fulltest(vhash, ptarget))
{
bool valid = equi_verify_sol(endiandata, &sol_data[3]);
if (valid && work->valid_nonces < MAX_NONCES) {
work->valid_nonces++;
memcpy(work->data, endiandata, 140);
equi_store_work_solution(work, vhash, sol_data);
work->nonces[work->valid_nonces-1] = endiandata[NONCE_OFT];
pdata[NONCE_OFT] = endiandata[NONCE_OFT] + 1;
//applog_hex(vhash, 32);
//applog_hex(&work->data[27], 32);
goto out; // second solution storage not handled..
}
}
if (work->valid_nonces == MAX_NONCES) goto out;
}
if (work->valid_nonces)
goto out;
valid_sols[thr_id] = 0;
}
endiandata[NONCE_OFT] += nonce_increment;
} while (!work_restart[thr_id].restart);
out:
gettimeofday(&tv_end, NULL);
timeval_subtract(&diff, &tv_end, &tv_start);
secs = (1.0 * diff.tv_sec) + (0.000001 * diff.tv_usec);
solps = (double)soluce_count / secs;
gpulog(LOG_DEBUG, thr_id, "%d solutions in %.2f s (%.2f Sol/s)", soluce_count, secs, solps);
// H/s
*hashes_done = soluce_count;
pdata[NONCE_OFT] = endiandata[NONCE_OFT] + 1;
return work->valid_nonces;
}
// cleanup
void free_equihash(int thr_id)
{
if (!init[thr_id])
return;
// assume config 1 was used... interface destructor seems bad
eq_cuda_context<CONFIG_MODE_1>* ptr = dynamic_cast<eq_cuda_context<CONFIG_MODE_1>*>(solvers[thr_id]);
ptr->freemem();
ptr = NULL;
solvers[thr_id] = NULL;
init[thr_id] = false;
}
// mmm... viva c++ junk
void eq_cuda_context_interface::solve(const char *tequihash_header, unsigned int tequihash_header_len,
const char* nonce, unsigned int nonce_len,
fn_cancel cancelf, fn_solution solutionf, fn_hashdone hashdonef) { }
+19
View File
@@ -0,0 +1,19 @@
#ifndef EQUIHASH_H
#define EQUIHASH_H
#include <stdint.h>
// miner nonce "cursor" unique for each thread
#define EQNONCE_OFFSET 30 /* 27:34 */
#define WK 9
#define WN 200
//#define CONFIG_MODE_1 9, 1248, 12, 640, packer_cantor /* eqcuda.hpp */
extern "C" {
void equi_hash(const void* input, void* output, int len);
int equi_verify_sol(void* const hdr, void* const soln);
bool equi_verify(uint8_t* const hdr, uint8_t* const soln);
}
#endif
+7
View File
@@ -40,6 +40,13 @@ int scanhash_fugue256(int thr_id, struct work* work, uint32_t max_nonce, unsigne
if(!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
fugue256_cpu_init(thr_id, throughput);
init[thr_id] = true;
+13 -13
View File
@@ -36,8 +36,6 @@ int scanhash_groestlcoin(int thr_id, struct work *work, uint32_t max_nonce, unsi
uint32_t throughput = cuda_default_throughput(thr_id, 1 << 19); // 256*256*8
if (init[thr_id]) throughput = min(throughput, max_nonce - start_nonce);
uint32_t *outputHash = (uint32_t*)malloc((size_t) 64* throughput);
if (opt_benchmark)
ptarget[7] = 0x001f;
@@ -50,6 +48,8 @@ int scanhash_groestlcoin(int thr_id, struct work *work, uint32_t max_nonce, unsi
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
CUDA_LOG_ERROR();
groestlcoin_cpu_init(thr_id, throughput);
init[thr_id] = true;
@@ -61,26 +61,28 @@ int scanhash_groestlcoin(int thr_id, struct work *work, uint32_t max_nonce, unsi
groestlcoin_cpu_setBlock(thr_id, endiandata, (void*)ptarget);
do {
uint32_t foundNounce = UINT32_MAX;
memset(work->nonces, 0xff, sizeof(work->nonces));
*hashes_done = pdata[19] - start_nonce + throughput;
// GPU hash
groestlcoin_cpu_hash(thr_id, throughput, pdata[19], outputHash, &foundNounce);
groestlcoin_cpu_hash(thr_id, throughput, pdata[19], &work->nonces[0]);
if (foundNounce < UINT32_MAX && bench_algo < 0)
if (work->nonces[0] < UINT32_MAX && bench_algo < 0)
{
uint32_t _ALIGN(64) vhash[8];
endiandata[19] = swab32(foundNounce);
endiandata[19] = swab32(work->nonces[0]);
groestlhash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work->valid_nonces = 1;
work_set_target_ratio(work, vhash);
pdata[19] = foundNounce;
free(outputHash);
return true;
} else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNounce);
pdata[19] = work->nonces[0] + 1; // cursor
return work->valid_nonces;
} else if (vhash[7] > ptarget[7]) {
gpu_increment_reject(thr_id);
if (!opt_quiet)
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]);
}
}
@@ -93,8 +95,6 @@ int scanhash_groestlcoin(int thr_id, struct work *work, uint32_t max_nonce, unsi
} while (!work_restart[thr_id].restart);
*hashes_done = pdata[19] - start_nonce;
free(outputHash);
return 0;
}
Loaded 100 of 244 files, more files were not shown because too many files have changed in this diff. Show more