34 Commits
Author SHA1 Message Date
Tanguy Pruvot 52788a5962 Import new SpreadCoin algo from tsiv spreadminer
Note: stratum doesn't exists for spreadcoin yet...
and this version have a problem to fix (PoK) to allow solo mining

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2015-09-28 18:28:37 +02:00
Tanguy Pruvot 8fb92415fb stratum: prevent multiple bloc notifications on yaamp
yaamp always notify clean jobs,
was not correct for my "new bloc" detection system...
2015-09-26 10:15:52 +02:00
Tanguy Pruvot 2ee86e8f47 skein: fix share ratio / diff 2015-09-26 10:10:39 +02:00
Tanguy Pruvot e1c4b3042c algos: add functions to free allocated resources
Will be used later for algo switching

not really tested yet...
2015-09-25 07:51:57 +02:00
Tanguy Pruvot 8ec5dd7d8e Handle the getwork typo in neoscrypt wallets
allow solo mining neoscrypt coins... (FTC/UFO)
2015-09-24 20:53:17 +02:00
Tanguy Pruvot ec054db317 Show shares diff on longpoll pools too
and remove unused remains of old difficulty functions
2015-09-24 13:23:24 +02:00
Tanguy Pruvot 41707760f5 stratum: force exit on pool auth failure
the miner was stuck instead of exiting after the retries...
2015-09-24 11:31:20 +02:00
Tanguy Pruvot fade2b74b5 diff: cleanup after windows test 2015-09-24 03:57:12 +02:00
Tanguy Pruvot 5308898d1c start v1.7, apply new prototypes to all algos 2015-09-23 15:42:17 +02:00
Tanguy Pruvot 4babf37256 Add --show-diff parameter and fix pool net diff
which display submitted block and net difficulty and is able
to detect shares above net diff (solved blocs)

Note: only made on lyra2v2 and zr5 algos

TODO: compute the found diff on all algos...
require changes in all scan hash "kernel" function parameters

to be continued...
2015-09-23 14:44:59 +02:00
Tanguy Pruvot cd64f9b0e7 Add a new cuda-schedule parameter
0: cudaDeviceScheduleAuto
1: cudaDeviceScheduleSpin
2: cudaDeviceScheduleYield
4: cudaDeviceScheduleBlockingSync

Also set the best one (4) for luffa algo by default...
2015-09-17 23:45:03 +02:00
Tanguy Pruvot 8f98bde4fb lyra2v2: improve cubehash with uint2 2015-09-06 13:49:52 +02:00
Tanguy Pruvot 6e64383b32 klaus small fixes 2015-09-05 20:16:24 +02:00
Tanguy Pruvot ac1f6ad1bb try vstudio MinimalRebuild flag... 2015-09-04 21:14:58 +02:00
Tanguy Pruvot 4b40c16359 v1.6.6 release 2015-08-28 11:30:04 +02:00
Tanguy Pruvot bcbb959a34 applog: remove useless mutex 2015-08-24 13:41:12 +02:00
Tanguy Pruvot 9a13624c2e check abort_flag in threads loops
This flag was added recently (scrypt) from cudaminer code,
and i missed some...

and prevent multiple calls to proper_exit(), one is enough...
2015-08-24 13:33:03 +02:00
Tanguy Pruvot 629d987177 Allow to load remote config files with curl
different samples of use:

  ccminer -c http://127.0.0.1/ccminer.json

or dynamically from a web application :
  ccminer -c http://127.0.0.1/ccminer/config.php?h=linux

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2015-08-24 08:44:34 +02:00
Tanguy Pruvot 21f5435420 lyra2: improve skein256 component 2015-08-23 09:46:48 +02:00
Tanguy Pruvot 877f25aee5 lyra2v2: SM3.0 compat too
and strip most of the unused vectors defines...
2015-08-23 09:46:23 +02:00
Tanguy Pruvot 5be0a5c586 update autoconf scripts
made on ubuntu 14.04.2 with 'autoreconf -fvi'
2015-08-23 06:36:14 +02:00
Tanguy Pruvot 64e997e9d6 prepare release 1.6.6 2015-08-22 17:33:26 +02:00
Tanguy Pruvot 01f3183c31 bmw algo for MDT, with midstate
which could be extracted from json too

replace a satcoin by another one ;)

Signed-off-by: Tanguy Pruvot <tanguy.pruvot@gmail.com>
2015-08-22 15:01:51 +02:00
Tanguy Pruvot e4414752c4 vstudio: lyra warning fixes + project dups 2015-08-22 14:57:15 +02:00
Tanguy Pruvot e3548f46f3 drop animecoin support
no more really minable... just minable in french
2015-08-22 12:35:22 +02:00
Tanguy Pruvot 736f916da3 Restore default timeout (300s) for normal use
Was only useful to reduce pool switching problems but
some pools require a longer value.
2015-08-22 12:30:30 +02:00
Tanguy Pruvot c9ef0b6b39 affinity: allow more than 8 cores values (and hexa) 2015-08-22 12:30:30 +02:00
Tanguy Pruvot b256ca47a0 bmw256: reduce target array size 2015-08-22 12:30:07 +02:00
Tanguy Pruvot 6500e7401a lyra2v2: properly exclude SM 3.0 devices 2015-08-18 13:43:21 +02:00
Tanguy Pruvot da8a31c4e5 vstudio: restore missing x15 whirlpool 2015-08-18 12:26:52 +02:00
Tanguy Pruvot 53cd591956 lyra2v2, bmw256 and cubehash256 cleanup + diff fix 2015-08-18 11:10:58 +02:00
Tanguy Pruvot 4b04ca8027 lyra2v2: aliases and readme 2015-08-18 09:27:30 +02:00
Tanguy Pruvot c9a66b696f return to original lyra2 code
until i find the problem...
2015-08-18 09:27:30 +02:00
Tanguy Pruvot d4e191610e Import and adapt lyra2v2
not tested on windows and with SM <= 5
2015-08-18 09:27:11 +02:00
85 changed files with 7317 additions and 3678 deletions

No files matched your search

+21 -4
View File
@@ -379,17 +379,18 @@ void blake256_cpu_setBlock_16(uint32_t *penddata, const uint32_t *midstate, cons
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_blake256(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done, int8_t blakerounds=14)
extern "C" int scanhash_blake256(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done, int8_t blakerounds=14)
{
const uint32_t first_nonce = pdata[19];
uint64_t targetHigh = ((uint64_t*)ptarget)[3];
uint32_t _ALIGN(64) endiandata[20];
#if PRECALC64
uint32_t _ALIGN(64) midstate[8];
#else
uint32_t crcsum;
#endif
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint64_t targetHigh = ((uint64_t*)ptarget)[3];
int intensity = (device_sm[device_map[thr_id]] > 500) ? 22 : 20;
uint32_t throughput = device_intensity(thr_id, __func__, 1U << intensity);
throughput = min(throughput, max_nonce - first_nonce);
@@ -488,3 +489,19 @@ extern "C" int scanhash_blake256(int thr_id, uint32_t *pdata, const uint32_t *pt
return rc;
}
// cleanup
extern "C" void free_blake256(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFreeHost(h_resNonce[thr_id]);
cudaFree(d_resNonce[thr_id]);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+119
View File
@@ -0,0 +1,119 @@
/**
* bmw-256 MDT
* tpruvot - 2015
*/
extern "C" {
#include "sph/sph_bmw.h"
}
#include "miner.h"
#include "cuda_helper.h"
static uint32_t *d_hash[MAX_GPUS];
extern void bmw256_midstate_init(int thr_id, uint32_t threads);
extern void bmw256_setBlock_80(int thr_id, void *pdata);
extern void bmw256_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_outputHash, int swap);
extern uint32_t cuda_check_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash);
// CPU Hash
extern "C" void bmw_hash(void *state, const void *input)
{
uint32_t _ALIGN(64) hash[16];
sph_bmw256_context ctx;
sph_bmw256_init(&ctx);
sph_bmw256(&ctx, input, 80);
sph_bmw256_close(&ctx, (void*) hash);
memcpy(state, hash, 32);
}
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 = device_intensity(thr_id, __func__, 1U << 21);
throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x0005;
if (!init[thr_id]) {
cudaSetDevice(device_map[thr_id]);
cuda_check_cpu_init(thr_id, throughput);
bmw256_midstate_init(thr_id, throughput);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t)32 * throughput));
init[thr_id] = true;
}
for (int k=0; k < 20; k++) {
be32enc(&endiandata[k], ((uint32_t*)pdata)[k]);
}
bmw256_setBlock_80(thr_id, (void*)endiandata);
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);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
*hashes_done = foundNonce - first_nonce + 1;
pdata[19] = swab32_if(foundNonce,!swapnonce);
bn_store_hash_target_ratio(vhash64, ptarget, work);
return 1;
}
else {
applog(LOG_DEBUG, "GPU #%d: result for nounce %08x does not validate on CPU!", thr_id, foundNonce);
}
}
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_bmw(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+374
View File
@@ -0,0 +1,374 @@
/**
* BMW-256 CUDA Implementation - tpruvot 2015
*
* Not optimal but close to the sph version and easier to adapt.
*/
#include <stdio.h>
#include <memory.h>
#define SPH_64 1
#define USE_MIDSTATE
extern "C" {
#include "sph/sph_bmw.h"
}
#include "cuda_helper.h"
__constant__ uint64_t c_PaddedMessage80[16]; // padded message (80 bytes + padding)
#ifndef USE_MIDSTATE
__constant__ static sph_u32 IV256[16] = {
0x40414243, 0x44454647, 0x48494A4B, 0x4C4D4E4F,
0x50515253, 0x54555657, 0x58595A5B, 0x5C5D5E5F,
0x60616263, 0x64656667, 0x68696A6B, 0x6C6D6E6F,
0x70717273, 0x74757677, 0x78797A7B, 0x7C7D7E7F
};
#endif
__constant__ static sph_u32 final_s[16] = {
0xaaaaaaa0, 0xaaaaaaa1, 0xaaaaaaa2, 0xaaaaaaa3,
0xaaaaaaa4, 0xaaaaaaa5, 0xaaaaaaa6, 0xaaaaaaa7,
0xaaaaaaa8, 0xaaaaaaa9, 0xaaaaaaaa, 0xaaaaaaab,
0xaaaaaaac, 0xaaaaaaad, 0xaaaaaaae, 0xaaaaaaaf
};
static sph_bmw_small_context* d_midstate[MAX_GPUS];
#define I16_16 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15
#define I16_17 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16
#define I16_18 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17
#define I16_19 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18
#define I16_20 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19
#define I16_21 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20
#define I16_22 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21
#define I16_23 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22
#define I16_24 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23
#define I16_25 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24
#define I16_26 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25
#define I16_27 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26
#define I16_28 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27
#define I16_29 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28
#define I16_30 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29
#define I16_31 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30
//#define M16_16 0, 1, 3, 4, 7, 10, 11
//#define M16_17 1, 2, 4, 5, 8, 11, 12
#define M16_18 2, 3, 5, 6, 9, 12, 13
#define M16_19 3, 4, 6, 7, 10, 13, 14
#define M16_20 4, 5, 7, 8, 11, 14, 15
#define M16_21 5, 6, 8, 9, 12, 15, 16
#define M16_22 6, 7, 9, 10, 13, 0, 1
#define M16_23 7, 8, 10, 11, 14, 1, 2
#define M16_24 8, 9, 11, 12, 15, 2, 3
#define M16_25 9, 10, 12, 13, 0, 3, 4
#define M16_26 10, 11, 13, 14, 1, 4, 5
#define M16_27 11, 12, 14, 15, 2, 5, 6
#define M16_28 12, 13, 15, 16, 3, 6, 7
#define M16_29 13, 14, 0, 1, 4, 7, 8
#define M16_30 14, 15, 1, 2, 5, 8, 9
#define M16_31 15, 16, 2, 3, 6, 9, 10
#define ss0(x) (((x) >> 1) ^ ((x) << 3) ^ ROTL32(x, 4) ^ ROTL32(x, 19))
#define ss1(x) (((x) >> 1) ^ ((x) << 2) ^ ROTL32(x, 8) ^ ROTL32(x, 23))
#define ss2(x) (((x) >> 2) ^ ((x) << 1) ^ ROTL32(x, 12) ^ ROTL32(x, 25))
#define ss3(x) (((x) >> 2) ^ ((x) << 2) ^ ROTL32(x, 15) ^ ROTL32(x, 29))
#define ss4(x) (((x) >> 1) ^ (x))
#define ss5(x) (((x) >> 2) ^ (x))
#define rs1(x) ROTL32(x, 3)
#define rs2(x) ROTL32(x, 7)
#define rs3(x) ROTL32(x, 13)
#define rs4(x) ROTL32(x, 16)
#define rs5(x) ROTL32(x, 19)
#define rs6(x) ROTL32(x, 23)
#define rs7(x) ROTL32(x, 27)
#define MAKE_W(tt, i0, op01, i1, op12, i2, op23, i3, op34, i4) \
tt((data[i0] ^ h[i0]) op01 (data[i1] ^ h[i1]) op12 (data[i2] ^ h[i2]) op23 (data[i3] ^ h[i3]) op34 (data[i4] ^ h[i4]))
//#define Ws0 MAKE_W(SPH_T32, 5, -, 7, +, 10, +, 13, +, 14)
//#define Ws1 MAKE_W(SPH_T32, 6, -, 8, +, 11, +, 14, -, 15)
//#define Ws2 MAKE_W(SPH_T32, 0, +, 7, +, 9, -, 12, +, 15)
//#define Ws3 MAKE_W(SPH_T32, 0, -, 1, +, 8, -, 10, +, 13)
//#define Ws4 MAKE_W(SPH_T32, 1, +, 2, +, 9, -, 11, -, 14)
//#define Ws5 MAKE_W(SPH_T32, 3, -, 2, +, 10, -, 12, +, 15)
//#define Ws6 MAKE_W(SPH_T32, 4, -, 0, -, 3, -, 11, +, 13)
//#define Ws7 MAKE_W(SPH_T32, 1, -, 4, -, 5, -, 12, -, 14)
//#define Ws8 MAKE_W(SPH_T32, 2, -, 5, -, 6, +, 13, -, 15)
//#define Ws9 MAKE_W(SPH_T32, 0, -, 3, +, 6, -, 7, +, 14)
//#define Ws10 MAKE_W(SPH_T32, 8, -, 1, -, 4, -, 7, +, 15)
//#define Ws11 MAKE_W(SPH_T32, 8, -, 0, -, 2, -, 5, +, 9)
//#define Ws12 MAKE_W(SPH_T32, 1, +, 3, -, 6, -, 9, +, 10)
//#define Ws13 MAKE_W(SPH_T32, 2, +, 4, +, 7, +, 10, +, 11)
//#define Ws14 MAKE_W(SPH_T32, 3, -, 5, +, 8, -, 11, -, 12)
//#define Ws15 MAKE_W(SPH_T32, 12, -, 4, -, 6, -, 9, +, 13)
__device__
static void gpu_compress_small(const sph_u32 *data, const sph_u32 h[16], sph_u32 dh[16])
{
// FOLD MAKE_Qas;
sph_u32 dx[16];
for (int i=0; i<16; i++)
dx[i] = data[i] ^ h[i];
sph_u32 qt[32];
qt[ 0] = dx[ 5] - dx[7] + dx[10] + dx[13] + dx[14]; // Ws0
qt[ 1] = dx[ 6] - dx[8] + dx[11] + dx[14] - dx[15]; // Ws1
qt[ 2] = dx[ 0] + dx[7] + dx[ 9] - dx[12] + dx[15]; // Ws2
qt[ 3] = dx[ 0] - dx[1] + dx[ 8] - dx[10] + dx[13]; // Ws3
qt[ 4] = dx[ 1] + dx[2] + dx[ 9] - dx[11] - dx[14]; // Ws4;
qt[ 5] = dx[ 3] - dx[2] + dx[10] - dx[12] + dx[15]; // Ws5;
qt[ 6] = dx[ 4] - dx[0] - dx[ 3] - dx[11] + dx[13]; // Ws6;
qt[ 7] = dx[ 1] - dx[4] - dx[ 5] - dx[12] - dx[14]; // Ws7;
qt[ 8] = dx[ 2] - dx[5] - dx[ 6] + dx[13] - dx[15]; // Ws8;
qt[ 9] = dx[ 0] - dx[3] + dx[ 6] - dx[ 7] + dx[14]; // Ws9;
qt[10] = dx[ 8] - dx[1] - dx[ 4] - dx[ 7] + dx[15]; // Ws10;
qt[11] = dx[ 8] - dx[0] - dx[ 2] - dx[ 5] + dx[ 9]; // Ws11;
qt[12] = dx[ 1] + dx[3] - dx[ 6] - dx[ 9] + dx[10]; // Ws12;
qt[13] = dx[ 2] + dx[4] + dx[ 7] + dx[10] + dx[11]; // Ws13;
qt[14] = dx[ 3] - dx[5] + dx[ 8] - dx[11] - dx[12]; // Ws14;
qt[15] = dx[12] - dx[4] - dx[ 6] - dx[ 9] + dx[13]; // Ws15;
qt[ 0] = ss0(qt[ 0]) + h[ 1];
qt[ 1] = ss1(qt[ 1]) + h[ 2];
qt[ 2] = ss2(qt[ 2]) + h[ 3];
qt[ 3] = ss3(qt[ 3]) + h[ 4];
qt[ 4] = ss4(qt[ 4]) + h[ 5];
qt[ 5] = ss0(qt[ 5]) + h[ 6];
qt[ 6] = ss1(qt[ 6]) + h[ 7];
qt[ 7] = ss2(qt[ 7]) + h[ 8];
qt[ 8] = ss3(qt[ 8]) + h[ 9];
qt[ 9] = ss4(qt[ 9]) + h[10];
qt[10] = ss0(qt[10]) + h[11];
qt[11] = ss1(qt[11]) + h[12];
qt[12] = ss2(qt[12]) + h[13];
qt[13] = ss3(qt[13]) + h[14];
qt[14] = ss4(qt[14]) + h[15];
qt[15] = ss0(qt[15]) + h[ 0];
//MAKE_Qbs;
#define Ks(j) ((sph_u32)(0x05555555UL * j))
#define Qs(j) (qt[j])
#define expand1s_in(i16, \
i0, i1, i2, i3, i4, i5, i6, i7, i8, i9, i10, i11, i12, i13, i14, i15, \
i0m, i1m, i3m, i4m, i7m, i10m, i11m) \
(ss1(qt[i0]) + ss2(qt[i1]) + ss3(qt[i2]) + ss0(qt[i3]) + ss1(qt[i4]) + ss2(qt[i5]) + ss3(qt[i6]) + ss0(qt[i7]) \
+ ss1(qt[i8]) + ss2(qt[i9]) + ss3(qt[i10]) + ss0(qt[i11]) + ss1(qt[i12]) + ss2(qt[i13]) + ss3(qt[i14]) + ss0(qt[i15]) \
+ ((ROTL32(data[i0m], i1m) + ROTL32(data[i3m], i4m) - ROTL32(data[i10m], i11m) + Ks(i16)) ^ h[i7m]))
qt[16] = expand1s_in(16, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 0, 1, 3, 4, 7, 10, 11);
qt[17] = expand1s_in(17, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2, 4, 5, 8, 11, 12);
#define expand2s_inner(qf, i16, \
i0, i1, i2, i3, i4, i5, i6, i7, i8, i9, i10, i11, i12, i13, i14, i15, \
i0m, i1m, i3m, i4m, i7m, i10m, i11m) \
(qf(i0) + rs1(qf(i1)) + qf(i2) + rs2(qf(i3)) \
+ qf(i4) + rs3(qf(i5)) + qf(i6) + rs4(qf(i7)) + qf(i8) + rs5(qf(i9)) + qf(i10) + rs6(qf(i11)) \
+ qf(i12) + rs7(qf(i13)) + ss4(qf(i14)) + ss5(qf(i15)) \
+ ((ROTL32(data[i0m], i1m) + ROTL32(data[i3m], i4m) - ROTL32(data[i10m], i11m) + Ks(i16)) ^ h[i7m]))
#ifdef _MSC_VER
#define LPAR (
#define expand2s(i16) \
expand2s_(Qs, i16, I16_ ## i16, M16_ ## i16)
#define expand2s_(qf, i16, ix, iy) \
expand2s_inner LPAR qf, i16, ix, iy)
#else
#define expand2s_(i16, ix, iy) \
expand2s_inner(Qs, i16, ix, iy)
#define expand2s(i16) \
expand2s_(i16, I16_ ## i16, M16_ ## i16)
#endif
qt[18] = expand2s(18);
qt[19] = expand2s(19);
qt[20] = expand2s(20);
qt[21] = expand2s(21);
qt[22] = expand2s(22);
qt[23] = expand2s(23);
qt[24] = expand2s(24);
qt[25] = expand2s(25);
qt[26] = expand2s(26);
qt[27] = expand2s(27);
qt[28] = expand2s(28);
qt[29] = expand2s(29);
qt[30] = expand2s(30);
qt[31] = expand2s(31);
sph_u32 xl, xh;
xl = Qs(16) ^ Qs(17) ^ Qs(18) ^ Qs(19) ^ Qs(20) ^ Qs(21) ^ Qs(22) ^ Qs(23);
xh = xl ^ Qs(24) ^ Qs(25) ^ Qs(26) ^ Qs(27) ^ Qs(28) ^ Qs(29) ^ Qs(30) ^ Qs(31);
dh[ 0] = ((xh << 5) ^ (Qs(16) >> 5) ^ data[ 0]) + (xl ^ Qs(24) ^ Qs(0));
dh[ 1] = ((xh >> 7) ^ (Qs(17) << 8) ^ data[ 1]) + (xl ^ Qs(25) ^ Qs(1));
dh[ 2] = ((xh >> 5) ^ (Qs(18) << 5) ^ data[ 2]) + (xl ^ Qs(26) ^ Qs(2));
dh[ 3] = ((xh >> 1) ^ (Qs(19) << 5) ^ data[ 3]) + (xl ^ Qs(27) ^ Qs(3));
dh[ 4] = ((xh >> 3) ^ (Qs(20) << 0) ^ data[ 4]) + (xl ^ Qs(28) ^ Qs(4));
dh[ 5] = ((xh << 6) ^ (Qs(21) >> 6) ^ data[ 5]) + (xl ^ Qs(29) ^ Qs(5));
dh[ 6] = ((xh >> 4) ^ (Qs(22) << 6) ^ data[ 6]) + (xl ^ Qs(30) ^ Qs(6));
dh[ 7] = ((xh >> 11) ^ (Qs(23) << 2) ^ data[ 7]) + (xl ^ Qs(31) ^ Qs(7));
dh[ 8] = ROTL32(dh[4], 9) + (xh ^ Qs(24) ^ data[ 8]) + ((xl << 8) ^ Qs(23) ^ Qs( 8));
dh[ 9] = ROTL32(dh[5], 10) + (xh ^ Qs(25) ^ data[ 9]) + ((xl >> 6) ^ Qs(16) ^ Qs( 9));
dh[10] = ROTL32(dh[6], 11) + (xh ^ Qs(26) ^ data[10]) + ((xl << 6) ^ Qs(17) ^ Qs(10));
dh[11] = ROTL32(dh[7], 12) + (xh ^ Qs(27) ^ data[11]) + ((xl << 4) ^ Qs(18) ^ Qs(11));
dh[12] = ROTL32(dh[0], 13) + (xh ^ Qs(28) ^ data[12]) + ((xl >> 3) ^ Qs(19) ^ Qs(12));
dh[13] = ROTL32(dh[1], 14) + (xh ^ Qs(29) ^ data[13]) + ((xl >> 4) ^ Qs(20) ^ Qs(13));
dh[14] = ROTL32(dh[2], 15) + (xh ^ Qs(30) ^ data[14]) + ((xl >> 7) ^ Qs(21) ^ Qs(14));
dh[15] = ROTL32(dh[3], 16) + (xh ^ Qs(31) ^ data[15]) + ((xl >> 2) ^ Qs(22) ^ Qs(15));
}
#ifndef USE_MIDSTATE
__device__
static void gpu_bmw256_init(sph_bmw_small_context *sc)
{
memcpy(sc->H, IV256, sizeof sc->H);
sc->ptr = 0;
sc->bit_count = 0;
}
__device__
static void gpu_bmw256(sph_bmw_small_context *sc, const void *data, size_t len)
{
sph_u32 htmp[16];
sph_u32 *h1, *h2;
unsigned char *buf = sc->buf;
size_t ptr = sc->ptr;
sc->bit_count += (sph_u64)len << 3;
h1 = sc->H;
h2 = htmp;
while (len > 0) {
size_t clen;
clen = (sizeof sc->buf) - ptr;
if (clen > len)
clen = len;
memcpy(buf + ptr, data, clen);
data = (const unsigned char *)data + clen;
len -= clen;
ptr += clen;
if (ptr == sizeof sc->buf) {
sph_u32 *ht;
gpu_compress_small((sph_u32 *) buf, h1, h2);
ht = h1;
h1 = h2;
h2 = ht;
ptr = 0;
}
}
sc->ptr = ptr;
if (h1 != sc->H)
memcpy(sc->H, h1, sizeof sc->H);
}
#endif
#define sph_enc64le(ptr, x) \
*((uint64_t*)(ptr)) = x
#define sph_enc64le_aligned sph_enc64le
__device__
static void gpu_bmw256_close(sph_bmw_small_context *sc, uint2 *out)
{
unsigned char *buf = sc->buf;
size_t ptr = sc->ptr;
buf[ptr ++] = 0x80;
sph_u32 *h = sc->H;
sph_u32 h1[16];
if (ptr > (sizeof sc->buf) - 8) {
memset(buf + ptr, 0, (sizeof sc->buf) - ptr);
gpu_compress_small((sph_u32 *) buf, h, h1);
ptr = 0;
h = h1;
}
memset(buf + ptr, 0, sizeof(sc->buf) - 8 - ptr);
sph_enc64le_aligned(buf + sizeof(sc->buf) - 8, SPH_T64(sc->bit_count));
sph_u32 h2[16];
gpu_compress_small((sph_u32 *) buf, h, h2);
gpu_compress_small(h2, final_s, h1);
uint64_t* h64 = (uint64_t*) (&h1[8]);
#pragma unroll
for (int i = 0; i < 4; i++) {
out[i] = vectorize(h64[i]);
}
}
__global__ /* __launch_bounds__(256, 3) */
void bmw256_gpu_hash_80(uint32_t threads, uint32_t startNonce, uint64_t *g_hash, sph_bmw256_context *d_midstate, int swap)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t nonce = startNonce + thread;
nonce = swap ? cuda_swab32(nonce): nonce;
#ifndef USE_MIDSTATE
uint2 hash[10];
#pragma unroll
for(int i=0;i<9;i++)
hash[i] = vectorize(c_PaddedMessage80[i]);
hash[9] = make_uint2(c_PaddedMessage80[9], nonce);
sph_bmw256_context ctx;
gpu_bmw256_init(&ctx);
gpu_bmw256(&ctx, (void*) hash, 80);
#else
sph_bmw256_context ctx;
ctx.ptr = 16; ctx.bit_count = 640;
uint2 *buf = (uint2 *) ctx.buf;
buf[0] = vectorize(c_PaddedMessage80[8]);
buf[1] = make_uint2(c_PaddedMessage80[9], nonce);
#pragma unroll
for(int i=0;i<16;i++)
ctx.H[i] = d_midstate->H[i];
#endif
gpu_bmw256_close(&ctx, (uint2*) &g_hash[thread << 2]);
}
}
__host__
void bmw256_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_outputHash, int swap)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
bmw256_gpu_hash_80<<<grid, block>>>(threads, startNonce, (uint64_t*)d_outputHash, d_midstate[thr_id], swap);
}
__host__
void bmw256_setBlock_80(int thr_id, void *pdata)
{
uint64_t PaddedMessage[16];
memcpy(PaddedMessage, pdata, 80);
memset(&PaddedMessage[10], 0, 48);
CUDA_SAFE_CALL(cudaMemcpyToSymbol(c_PaddedMessage80, PaddedMessage, sizeof(PaddedMessage), 0, cudaMemcpyHostToDevice));
sph_bmw256_context ctx;
sph_bmw256_init(&ctx);
sph_bmw256(&ctx, (void*) PaddedMessage, 80);
CUDA_SAFE_CALL(cudaMemcpy(d_midstate[thr_id], &ctx, sizeof(sph_bmw256_context), cudaMemcpyHostToDevice));
}
__host__
void bmw256_midstate_init(int thr_id, uint32_t threads)
{
cudaMalloc(&d_midstate[thr_id], sizeof(sph_bmw256_context));
}
+283
View File
@@ -0,0 +1,283 @@
#include <stdio.h>
#include <memory.h>
#include "cuda_helper.h"
#undef SPH_ROTL32
#define SPH_ROTL32 ROTL32
static uint32_t *d_gnounce[MAX_GPUS];
static uint32_t *d_GNonce[MAX_GPUS];
__constant__ uint64_t pTarget[4];
#define shl(x, n) ((x) << (n))
#define shr(x, n) ((x) >> (n))
#define ss0(x) (shr((x), 1) ^ shl((x), 3) ^ SPH_ROTL32((x), 4) ^ SPH_ROTL32((x), 19))
#define ss1(x) (shr((x), 1) ^ shl((x), 2) ^ SPH_ROTL32((x), 8) ^ SPH_ROTL32((x), 23))
#define ss2(x) (shr((x), 2) ^ shl((x), 1) ^ SPH_ROTL32((x), 12) ^ SPH_ROTL32((x), 25))
#define ss3(x) (shr((x), 2) ^ shl((x), 2) ^ SPH_ROTL32((x), 15) ^ SPH_ROTL32((x), 29))
#define ss4(x) (shr((x), 1) ^ (x))
#define ss5(x) (shr((x), 2) ^ (x))
#define rs1(x) SPH_ROTL32((x), 3)
#define rs2(x) SPH_ROTL32((x), 7)
#define rs3(x) SPH_ROTL32((x), 13)
#define rs4(x) SPH_ROTL32((x), 16)
#define rs5(x) SPH_ROTL32((x), 19)
#define rs6(x) SPH_ROTL32((x), 23)
#define rs7(x) SPH_ROTL32((x), 27)
/* Message expansion function 1 */
__forceinline__ __device__
uint32_t expand32_1(int i, uint32_t *M32, const uint32_t *H, uint32_t *Q)
{
return (ss1(Q[i - 16]) + ss2(Q[i - 15]) + ss3(Q[i - 14]) + ss0(Q[i - 13])
+ ss1(Q[i - 12]) + ss2(Q[i - 11]) + ss3(Q[i - 10]) + ss0(Q[i - 9])
+ ss1(Q[i - 8]) + ss2(Q[i - 7]) + ss3(Q[i - 6]) + ss0(Q[i - 5])
+ ss1(Q[i - 4]) + ss2(Q[i - 3]) + ss3(Q[i - 2]) + ss0(Q[i - 1])
+ ((i*(0x05555555ul) + SPH_ROTL32(M32[(i - 16) % 16], ((i - 16) % 16) + 1)
+ SPH_ROTL32(M32[(i - 13) % 16], ((i - 13) % 16) + 1)
- SPH_ROTL32(M32[(i - 6) % 16], ((i - 6) % 16) + 1)) ^ H[(i - 16 + 7) % 16]));
}
/* Message expansion function 2 */
__forceinline__ __device__
uint32_t expand32_2(int i, uint32_t *M32, const uint32_t *H, uint32_t *Q)
{
return (Q[i - 16] + rs1(Q[i - 15]) + Q[i - 14] + rs2(Q[i - 13])
+ Q[i - 12] + rs3(Q[i - 11]) + Q[i - 10] + rs4(Q[i - 9])
+ Q[i - 8] + rs5(Q[i - 7]) + Q[i - 6] + rs6(Q[i - 5])
+ Q[i - 4] + rs7(Q[i - 3]) + ss4(Q[i - 2]) + ss5(Q[i - 1])
+ ((i*(0x05555555ul) + SPH_ROTL32(M32[(i - 16) % 16], ((i - 16) % 16) + 1)
+ SPH_ROTL32(M32[(i - 13) % 16], ((i - 13) % 16) + 1)
- SPH_ROTL32(M32[(i - 6) % 16], ((i - 6) % 16) + 1)) ^ H[(i - 16 + 7) % 16]));
}
__forceinline__ __device__
void Compression256(uint32_t * M32)
{
uint32_t Q[32], XL32, XH32;
const uint32_t H[16] = {
0x40414243, 0x44454647, 0x48494A4B, 0x4C4D4E4F,
0x50515253, 0x54555657, 0x58595A5B, 0x5C5D5E5F,
0x60616263, 0x64656667, 0x68696A6B, 0x6C6D6E6F,
0x70717273, 0x74757677, 0x78797A7B, 0x7C7D7E7F
};
Q[0] = (M32[5] ^ H[5]) - (M32[7] ^ H[7]) + (M32[10] ^ H[10]) + (M32[13] ^ H[13]) + (M32[14] ^ H[14]);
Q[1] = (M32[6] ^ H[6]) - (M32[8] ^ H[8]) + (M32[11] ^ H[11]) + (M32[14] ^ H[14]) - (M32[15] ^ H[15]);
Q[2] = (M32[0] ^ H[0]) + (M32[7] ^ H[7]) + (M32[9] ^ H[9]) - (M32[12] ^ H[12]) + (M32[15] ^ H[15]);
Q[3] = (M32[0] ^ H[0]) - (M32[1] ^ H[1]) + (M32[8] ^ H[8]) - (M32[10] ^ H[10]) + (M32[13] ^ H[13]);
Q[4] = (M32[1] ^ H[1]) + (M32[2] ^ H[2]) + (M32[9] ^ H[9]) - (M32[11] ^ H[11]) - (M32[14] ^ H[14]);
Q[5] = (M32[3] ^ H[3]) - (M32[2] ^ H[2]) + (M32[10] ^ H[10]) - (M32[12] ^ H[12]) + (M32[15] ^ H[15]);
Q[6] = (M32[4] ^ H[4]) - (M32[0] ^ H[0]) - (M32[3] ^ H[3]) - (M32[11] ^ H[11]) + (M32[13] ^ H[13]);
Q[7] = (M32[1] ^ H[1]) - (M32[4] ^ H[4]) - (M32[5] ^ H[5]) - (M32[12] ^ H[12]) - (M32[14] ^ H[14]);
Q[8] = (M32[2] ^ H[2]) - (M32[5] ^ H[5]) - (M32[6] ^ H[6]) + (M32[13] ^ H[13]) - (M32[15] ^ H[15]);
Q[9] = (M32[0] ^ H[0]) - (M32[3] ^ H[3]) + (M32[6] ^ H[6]) - (M32[7] ^ H[7]) + (M32[14] ^ H[14]);
Q[10] = (M32[8] ^ H[8]) - (M32[1] ^ H[1]) - (M32[4] ^ H[4]) - (M32[7] ^ H[7]) + (M32[15] ^ H[15]);
Q[11] = (M32[8] ^ H[8]) - (M32[0] ^ H[0]) - (M32[2] ^ H[2]) - (M32[5] ^ H[5]) + (M32[9] ^ H[9]);
Q[12] = (M32[1] ^ H[1]) + (M32[3] ^ H[3]) - (M32[6] ^ H[6]) - (M32[9] ^ H[9]) + (M32[10] ^ H[10]);
Q[13] = (M32[2] ^ H[2]) + (M32[4] ^ H[4]) + (M32[7] ^ H[7]) + (M32[10] ^ H[10]) + (M32[11] ^ H[11]);
Q[14] = (M32[3] ^ H[3]) - (M32[5] ^ H[5]) + (M32[8] ^ H[8]) - (M32[11] ^ H[11]) - (M32[12] ^ H[12]);
Q[15] = (M32[12] ^ H[12]) - (M32[4] ^ H[4]) - (M32[6] ^ H[6]) - (M32[9] ^ H[9]) + (M32[13] ^ H[13]);
/* Diffuse the differences in every word in a bijective manner with ssi, and then add the values of the previous double pipe. */
Q[0] = ss0(Q[0]) + H[1];
Q[1] = ss1(Q[1]) + H[2];
Q[2] = ss2(Q[2]) + H[3];
Q[3] = ss3(Q[3]) + H[4];
Q[4] = ss4(Q[4]) + H[5];
Q[5] = ss0(Q[5]) + H[6];
Q[6] = ss1(Q[6]) + H[7];
Q[7] = ss2(Q[7]) + H[8];
Q[8] = ss3(Q[8]) + H[9];
Q[9] = ss4(Q[9]) + H[10];
Q[10] = ss0(Q[10]) + H[11];
Q[11] = ss1(Q[11]) + H[12];
Q[12] = ss2(Q[12]) + H[13];
Q[13] = ss3(Q[13]) + H[14];
Q[14] = ss4(Q[14]) + H[15];
Q[15] = ss0(Q[15]) + H[0];
/* This is the Message expansion or f_1 in the documentation. */
/* It has 16 rounds. */
/* Blue Midnight Wish has two tunable security parameters. */
/* The parameters are named EXPAND_1_ROUNDS and EXPAND_2_ROUNDS. */
/* The following relation for these parameters should is satisfied: */
/* EXPAND_1_ROUNDS + EXPAND_2_ROUNDS = 16 */
#pragma unroll
for (int i=16; i<18; i++)
Q[i] = expand32_1(i, M32, H, Q);
#pragma nounroll
for (int i=18; i<32; i++)
Q[i] = expand32_2(i, M32, H, Q);
/* Blue Midnight Wish has two temporary cummulative variables that accumulate via XORing */
/* 16 new variables that are prooduced in the Message Expansion part. */
XL32 = Q[16] ^ Q[17] ^ Q[18] ^ Q[19] ^ Q[20] ^ Q[21] ^ Q[22] ^ Q[23];
XH32 = XL32^Q[24] ^ Q[25] ^ Q[26] ^ Q[27] ^ Q[28] ^ Q[29] ^ Q[30] ^ Q[31];
/* This part is the function f_2 - in the documentation */
/* Compute the double chaining pipe for the next message block. */
M32[0] = (shl(XH32, 5) ^ shr(Q[16], 5) ^ M32[0]) + (XL32 ^ Q[24] ^ Q[0]);
M32[1] = (shr(XH32, 7) ^ shl(Q[17], 8) ^ M32[1]) + (XL32 ^ Q[25] ^ Q[1]);
M32[2] = (shr(XH32, 5) ^ shl(Q[18], 5) ^ M32[2]) + (XL32 ^ Q[26] ^ Q[2]);
M32[3] = (shr(XH32, 1) ^ shl(Q[19], 5) ^ M32[3]) + (XL32 ^ Q[27] ^ Q[3]);
M32[4] = (shr(XH32, 3) ^ Q[20] ^ M32[4]) + (XL32 ^ Q[28] ^ Q[4]);
M32[5] = (shl(XH32, 6) ^ shr(Q[21], 6) ^ M32[5]) + (XL32 ^ Q[29] ^ Q[5]);
M32[6] = (shr(XH32, 4) ^ shl(Q[22], 6) ^ M32[6]) + (XL32 ^ Q[30] ^ Q[6]);
M32[7] = (shr(XH32, 11) ^ shl(Q[23], 2) ^ M32[7]) + (XL32 ^ Q[31] ^ Q[7]);
M32[8] = SPH_ROTL32(M32[4], 9) + (XH32 ^ Q[24] ^ M32[8]) + (shl(XL32, 8) ^ Q[23] ^ Q[8]);
M32[9] = SPH_ROTL32(M32[5], 10) + (XH32 ^ Q[25] ^ M32[9]) + (shr(XL32, 6) ^ Q[16] ^ Q[9]);
M32[10] = SPH_ROTL32(M32[6], 11) + (XH32 ^ Q[26] ^ M32[10]) + (shl(XL32, 6) ^ Q[17] ^ Q[10]);
M32[11] = SPH_ROTL32(M32[7], 12) + (XH32 ^ Q[27] ^ M32[11]) + (shl(XL32, 4) ^ Q[18] ^ Q[11]);
M32[12] = SPH_ROTL32(M32[0], 13) + (XH32 ^ Q[28] ^ M32[12]) + (shr(XL32, 3) ^ Q[19] ^ Q[12]);
M32[13] = SPH_ROTL32(M32[1], 14) + (XH32 ^ Q[29] ^ M32[13]) + (shr(XL32, 4) ^ Q[20] ^ Q[13]);
M32[14] = SPH_ROTL32(M32[2], 15) + (XH32 ^ Q[30] ^ M32[14]) + (shr(XL32, 7) ^ Q[21] ^ Q[14]);
M32[15] = SPH_ROTL32(M32[3], 16) + (XH32 ^ Q[31] ^ M32[15]) + (shr(XL32, 2) ^ Q[22] ^ Q[15]);
}
__forceinline__ __device__
void Compression256_2(uint32_t * M32)
{
uint32_t XL32, XH32, Q[32];
const uint32_t H[16] = {
0xaaaaaaa0, 0xaaaaaaa1, 0xaaaaaaa2, 0xaaaaaaa3,
0xaaaaaaa4, 0xaaaaaaa5, 0xaaaaaaa6, 0xaaaaaaa7,
0xaaaaaaa8, 0xaaaaaaa9, 0xaaaaaaaa, 0xaaaaaaab,
0xaaaaaaac, 0xaaaaaaad, 0xaaaaaaae, 0xaaaaaaaf
};
Q[0] = (M32[5] ^ H[5]) - (M32[7] ^ H[7]) + (M32[10] ^ H[10]) + (M32[13] ^ H[13]) + (M32[14] ^ H[14]);
Q[1] = (M32[6] ^ H[6]) - (M32[8] ^ H[8]) + (M32[11] ^ H[11]) + (M32[14] ^ H[14]) - (M32[15] ^ H[15]);
Q[2] = (M32[0] ^ H[0]) + (M32[7] ^ H[7]) + (M32[9] ^ H[9]) - (M32[12] ^ H[12]) + (M32[15] ^ H[15]);
Q[3] = (M32[0] ^ H[0]) - (M32[1] ^ H[1]) + (M32[8] ^ H[8]) - (M32[10] ^ H[10]) + (M32[13] ^ H[13]);
Q[4] = (M32[1] ^ H[1]) + (M32[2] ^ H[2]) + (M32[9] ^ H[9]) - (M32[11] ^ H[11]) - (M32[14] ^ H[14]);
Q[5] = (M32[3] ^ H[3]) - (M32[2] ^ H[2]) + (M32[10] ^ H[10]) - (M32[12] ^ H[12]) + (M32[15] ^ H[15]);
Q[6] = (M32[4] ^ H[4]) - (M32[0] ^ H[0]) - (M32[3] ^ H[3]) - (M32[11] ^ H[11]) + (M32[13] ^ H[13]);
Q[7] = (M32[1] ^ H[1]) - (M32[4] ^ H[4]) - (M32[5] ^ H[5]) - (M32[12] ^ H[12]) - (M32[14] ^ H[14]);
Q[8] = (M32[2] ^ H[2]) - (M32[5] ^ H[5]) - (M32[6] ^ H[6]) + (M32[13] ^ H[13]) - (M32[15] ^ H[15]);
Q[9] = (M32[0] ^ H[0]) - (M32[3] ^ H[3]) + (M32[6] ^ H[6]) - (M32[7] ^ H[7]) + (M32[14] ^ H[14]);
Q[10] = (M32[8] ^ H[8]) - (M32[1] ^ H[1]) - (M32[4] ^ H[4]) - (M32[7] ^ H[7]) + (M32[15] ^ H[15]);
Q[11] = (M32[8] ^ H[8]) - (M32[0] ^ H[0]) - (M32[2] ^ H[2]) - (M32[5] ^ H[5]) + (M32[9] ^ H[9]);
Q[12] = (M32[1] ^ H[1]) + (M32[3] ^ H[3]) - (M32[6] ^ H[6]) - (M32[9] ^ H[9]) + (M32[10] ^ H[10]);
Q[13] = (M32[2] ^ H[2]) + (M32[4] ^ H[4]) + (M32[7] ^ H[7]) + (M32[10] ^ H[10]) + (M32[11] ^ H[11]);
Q[14] = (M32[3] ^ H[3]) - (M32[5] ^ H[5]) + (M32[8] ^ H[8]) - (M32[11] ^ H[11]) - (M32[12] ^ H[12]);
Q[15] = (M32[12] ^ H[12]) - (M32[4] ^ H[4]) - (M32[6] ^ H[6]) - (M32[9] ^ H[9]) + (M32[13] ^ H[13]);
/* Diffuse the differences in every word in a bijective manner with ssi, and then add the values of the previous double pipe.*/
Q[0] = ss0(Q[0]) + H[1];
Q[1] = ss1(Q[1]) + H[2];
Q[2] = ss2(Q[2]) + H[3];
Q[3] = ss3(Q[3]) + H[4];
Q[4] = ss4(Q[4]) + H[5];
Q[5] = ss0(Q[5]) + H[6];
Q[6] = ss1(Q[6]) + H[7];
Q[7] = ss2(Q[7]) + H[8];
Q[8] = ss3(Q[8]) + H[9];
Q[9] = ss4(Q[9]) + H[10];
Q[10] = ss0(Q[10]) + H[11];
Q[11] = ss1(Q[11]) + H[12];
Q[12] = ss2(Q[12]) + H[13];
Q[13] = ss3(Q[13]) + H[14];
Q[14] = ss4(Q[14]) + H[15];
Q[15] = ss0(Q[15]) + H[0];
/* This is the Message expansion or f_1 in the documentation. */
/* It has 16 rounds. */
/* Blue Midnight Wish has two tunable security parameters. */
/* The parameters are named EXPAND_1_ROUNDS and EXPAND_2_ROUNDS. */
/* The following relation for these parameters should is satisfied: */
/* EXPAND_1_ROUNDS + EXPAND_2_ROUNDS = 16 */
#pragma unroll
for (int i = 16; i<18; i++)
Q[i] = expand32_1(i, M32, H, Q);
#pragma nounroll
for (int i = 18; i<32; i++)
Q[i] = expand32_2(i, M32, H, Q);
/* Blue Midnight Wish has two temporary cummulative variables that accumulate via XORing */
/* 16 new variables that are prooduced in the Message Expansion part. */
XL32 = Q[16] ^ Q[17] ^ Q[18] ^ Q[19] ^ Q[20] ^ Q[21] ^ Q[22] ^ Q[23];
XH32 = XL32 ^ Q[24] ^ Q[25] ^ Q[26] ^ Q[27] ^ Q[28] ^ Q[29] ^ Q[30] ^ Q[31];
M32[2] = (shr(XH32, 5) ^ shl(Q[18], 5) ^ M32[2]) + (XL32 ^ Q[26] ^ Q[2]);
M32[3] = (shr(XH32, 1) ^ shl(Q[19], 5) ^ M32[3]) + (XL32 ^ Q[27] ^ Q[3]);
M32[14] = SPH_ROTL32(M32[2], 15) + (XH32 ^ Q[30] ^ M32[14]) + (shr(XL32, 7) ^ Q[21] ^ Q[14]);
M32[15] = SPH_ROTL32(M32[3], 16) + (XH32 ^ Q[31] ^ M32[15]) + (shr(XL32, 2) ^ Q[22] ^ Q[15]);
}
#define TPB 512
__global__ __launch_bounds__(TPB, 2)
void bmw256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint64_t *g_hash, uint32_t *const __restrict__ nonceVector)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t message[16] = { 0 };
LOHI(message[0], message[1], __ldg(&g_hash[thread]));
LOHI(message[2], message[3], __ldg(&g_hash[thread + 1 * threads]));
LOHI(message[4], message[5], __ldg(&g_hash[thread + 2 * threads]));
LOHI(message[6], message[7], __ldg(&g_hash[thread + 3 * threads]));
message[8]=0x80;
message[14]=0x100;
Compression256(message);
Compression256_2(message);
if (((uint64_t*)message)[7] <= pTarget[3])
{
uint32_t tmp = atomicExch(&nonceVector[0], startNounce + thread);
if (tmp != 0)
nonceVector[1] = tmp;
}
}
}
__host__
void bmw256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *g_hash, uint32_t *resultnonces)
{
const uint32_t threadsperblock = TPB;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
cudaMemset(d_GNonce[thr_id], 0, 2 * sizeof(uint32_t));
bmw256_gpu_hash_32 << <grid, block >> >(threads, startNounce, g_hash, d_GNonce[thr_id]);
cudaMemcpy(d_gnounce[thr_id], d_GNonce[thr_id], 2 * sizeof(uint32_t), cudaMemcpyDeviceToHost);
resultnonces[0] = *(d_gnounce[thr_id]);
resultnonces[1] = *(d_gnounce[thr_id] + 1);
}
__host__
void bmw256_cpu_init(int thr_id, uint32_t threads)
{
cudaMalloc(&d_GNonce[thr_id], 2 * sizeof(uint32_t));
cudaMallocHost(&d_gnounce[thr_id], 2 * sizeof(uint32_t));
}
__host__
void bmw256_cpu_free(int thr_id)
{
cudaFree(d_GNonce[thr_id]);
cudaFreeHost(d_gnounce[thr_id]);
}
__host__
void bmw256_setTarget(const void *pTargetIn)
{
cudaMemcpyToSymbol(pTarget, pTargetIn, 32, 0, cudaMemcpyHostToDevice);
}
+305
View File
@@ -0,0 +1,305 @@
#include "cuda_helper.h"
#define CUBEHASH_ROUNDS 16 /* this is r for CubeHashr/b */
#define CUBEHASH_BLOCKBYTES 32 /* this is b for CubeHashr/b */
#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 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];
/* "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]);
/* "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])
/* "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];
/* "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]);
/* "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])
/* "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_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])
}
}
__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)
{
/* "the integer 1 is xored into the last state word x_11111" */
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);
}
#if __CUDA_ARCH__ >= 500
__global__ __launch_bounds__(TPB, 1)
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)
{
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]);
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;
x[0][0][0][0][1] ^= Hash[0].y;
x[0][0][0][1][0] ^= Hash[1].x;
x[0][0][0][1][1] ^= Hash[1].y;
x[0][0][1][0][0] ^= Hash[2].x;
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];
x[0][0][0][1][1] ^= Hash[3];
x[0][0][1][0][0] ^= Hash[4];
x[0][0][1][0][1] ^= Hash[5];
x[0][0][1][1][0] ^= Hash[6];
x[0][0][1][1][1] ^= Hash[7];
rrounds(x);
x[0][0][0][0][0] ^= 0x80U;
rrounds(x);
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];
}
}
#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;
dim3 grid((threads + tpb-1)/tpb);
dim3 block(tpb);
cubehash256_gpu_hash_32 <<<grid, block>>> (threads, startNounce, (uint2*) d_hash);
}
+7
View File
@@ -732,6 +732,13 @@ void fugue256_cpu_init(int thr_id, uint32_t threads)
cudaMalloc(&d_resultNonce[thr_id], sizeof(uint32_t));
}
__host__
void fugue256_cpu_free(int thr_id)
{
cudaFree(d_fugue256_hashoutput[thr_id]);
cudaFree(d_resultNonce[thr_id]);
}
__host__
void fugue256_cpu_setBlock(int thr_id, void *data, void *pTargetIn)
{
+7 -1
View File
@@ -262,7 +262,6 @@ void groestl256_gpu_hash32(uint32_t threads, uint32_t startNounce, uint64_t *out
__host__
void groestl256_cpu_init(int thr_id, uint32_t threads)
{
// Texturen mit obigem Makro initialisieren
texDef(t0up2, d_T0up, T0up_cpu, sizeof(uint32_t) * 256);
texDef(t0dn2, d_T0dn, T0dn_cpu, sizeof(uint32_t) * 256);
@@ -277,6 +276,13 @@ void groestl256_cpu_init(int thr_id, uint32_t threads)
cudaMallocHost(&h_GNonces[thr_id], 2*sizeof(uint32_t));
}
__host__
void groestl256_cpu_free(int thr_id)
{
cudaFree(d_GNonces[thr_id]);
cudaFreeHost(h_GNonces[thr_id]);
}
__host__
uint32_t groestl256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
{
+7
View File
@@ -300,3 +300,10 @@ void keccak256_cpu_init(int thr_id, uint32_t threads)
CUDA_SAFE_CALL(cudaMalloc(&d_KNonce[thr_id], sizeof(uint32_t)));
CUDA_SAFE_CALL(cudaMallocHost(&d_nounce[thr_id], 1*sizeof(uint32_t)));
}
__host__
void keccak256_cpu_free(int thr_id)
{
cudaFree(d_KNonce[thr_id]);
cudaFreeHost(d_nounce[thr_id]);
}
+134 -133
View File
@@ -2,79 +2,40 @@
#include "cuda_helper.h"
static __constant__ uint64_t SKEIN_IV512_256[8] = {
0xCCD044A12FDB3E13, 0xE83590301A79A9EB,
0x55AEA0614F816E6F, 0x2A2767A4AE9B94DB,
0xEC06025E74DD7683, 0xE7A436CDC4746251,
0xC36FBAF9393AD185, 0x3EEDBA1833EDFC13
};
static __constant__ uint2 vSKEIN_IV512_256[8] = {
{ 0x2FDB3E13, 0xCCD044A1 },
{ 0x1A79A9EB, 0xE8359030 },
{ 0x4F816E6F, 0x55AEA061 },
{ 0xAE9B94DB, 0x2A2767A4 },
{ 0x74DD7683, 0xEC06025E },
{ 0xC4746251, 0xE7A436CD },
{ 0x393AD185, 0xC36FBAF9 },
{ 0x33EDFC13, 0x3EEDBA18 }
};
static __constant__ int8_t ROT256[8][4] = {
46, 36, 19, 37,
33, 27, 14, 42,
17, 49, 36, 39,
44, 9, 54, 56,
39, 30, 34, 24,
13, 50, 10, 17,
25, 29, 39, 43,
8, 35, 56, 22,
};
static __constant__ uint2 t12[6] = {
{ 0x20, 0 },
{ 0, 0xf0000000 },
{ 0x20, 0xf0000000 },
{ 0x08, 0 },
{ 0, 0xff000000 },
{ 0x08, 0xff000000 }
};
static __constant__ uint2 skein_ks_parity = { 0xA9FC1A22, 0x1BD11BDA };
static __constant__ uint64_t skein_ks_parity64 = 0x1BD11BDAA9FC1A22ull;
static __forceinline__ __device__
void Round512v35(uint2 &p0, uint2 &p1, uint2 &p2, uint2 &p3, uint2 &p4, uint2 &p5, uint2 &p6, uint2 &p7, int ROT)
void Round512v35(uint2 &p0, uint2 &p1, uint2 &p2, uint2 &p3, uint2 &p4, uint2 &p5, uint2 &p6, uint2 &p7,
const int ROT0, const int ROT1, const int ROT2, const int ROT3)
{
p0 += p1; p1 = ROL2(p1, ROT256[ROT][0]) ^ p0;
p2 += p3; p3 = ROL2(p3, ROT256[ROT][1]) ^ p2;
p4 += p5; p5 = ROL2(p5, ROT256[ROT][2]) ^ p4;
p6 += p7; p7 = ROL2(p7, ROT256[ROT][3]) ^ p6;
p0 += p1; p1 = ROL2(p1, ROT0) ^ p0;
p2 += p3; p3 = ROL2(p3, ROT1) ^ p2;
p4 += p5; p5 = ROL2(p5, ROT2) ^ p4;
p6 += p7; p7 = ROL2(p7, ROT3) ^ p6;
}
static __forceinline__ __device__
void Round_8_512v35(uint2 *ks, uint2 *ts,
uint2 &p0, uint2 &p1, uint2 &p2, uint2 &p3,
uint2 &p4, uint2 &p5, uint2 &p6, uint2 &p7, int R)
__forceinline__ __device__
void Round_8_512v35(const uint2 *const __restrict__ ks, const uint2 *const __restrict__ ts,
uint2 &p0, uint2 &p1, uint2 &p2, uint2 &p3, uint2 &p4, uint2 &p5, uint2 &p6, uint2 &p7, int R)
{
Round512v35(p0, p1, p2, p3, p4, p5, p6, p7, 0);
Round512v35(p2, p1, p4, p7, p6, p5, p0, p3, 1);
Round512v35(p4, p1, p6, p3, p0, p5, p2, p7, 2);
Round512v35(p6, p1, p0, p7, p2, p5, p4, p3, 3);
p0 += ks[(R+0) % 9]; /* inject the key schedule value */
Round512v35(p0, p1, p2, p3, p4, p5, p6, p7, 46, 36, 19, 37);
Round512v35(p2, p1, p4, p7, p6, p5, p0, p3, 33, 27, 14, 42);
Round512v35(p4, p1, p6, p3, p0, p5, p2, p7, 17, 49, 36, 39);
Round512v35(p6, p1, p0, p7, p2, p5, p4, p3, 44, 9, 54, 56);
p0 += ks[(R+0) % 9];
p1 += ks[(R+1) % 9];
p2 += ks[(R+2) % 9];
p3 += ks[(R+3) % 9];
p4 += ks[(R+4) % 9];
p5 += ks[(R+5) % 9] + ts[(R+0) % 3];
p6 += ks[(R+6) % 9] + ts[(R+1) % 3];
p7 += ks[(R+7) % 9] + make_uint2((R),0);
Round512v35(p0, p1, p2, p3, p4, p5, p6, p7, 4);
Round512v35(p2, p1, p4, p7, p6, p5, p0, p3, 5);
Round512v35(p4, p1, p6, p3, p0, p5, p2, p7, 6);
Round512v35(p6, p1, p0, p7, p2, p5, p4, p3, 7);
p0 += ks[(R+1) % 9]; /* inject the key schedule value */
p7 += ks[(R+7) % 9] + make_uint2(R, 0);
Round512v35(p0, p1, p2, p3, p4, p5, p6, p7, 39, 30, 34, 24);
Round512v35(p2, p1, p4, p7, p6, p5, p0, p3, 13, 50, 10, 17);
Round512v35(p4, p1, p6, p3, p0, p5, p2, p7, 25, 29, 39, 43);
Round512v35(p6, p1, p0, p7, p2, p5, p4, p3, 8, 35, 56, 22);
p0 += ks[(R+1) % 9];
p1 += ks[(R+2) % 9];
p2 += ks[(R+3) % 9];
p3 += ks[(R+4) % 9];
@@ -84,38 +45,83 @@ void Round_8_512v35(uint2 *ks, uint2 *ts,
p7 += ks[(R+8) % 9] + make_uint2(R+1, 0);
}
__forceinline__ __device__
void Round_8_512v35_final(const uint2 *const __restrict__ ks, const uint2 *const __restrict__ ts,
uint2 &p0, uint2 &p1, uint2 &p2, uint2 &p3, uint2 &p4, uint2 &p5, uint2 &p6, uint2 &p7)
{
Round512v35(p0, p1, p2, p3, p4, p5, p6, p7, 46, 36, 19, 37);
Round512v35(p2, p1, p4, p7, p6, p5, p0, p3, 33, 27, 14, 42);
Round512v35(p4, p1, p6, p3, p0, p5, p2, p7, 17, 49, 36, 39);
Round512v35(p6, p1, p0, p7, p2, p5, p4, p3, 44, 9, 54, 56);
p0 += ks[8];
p1 += ks[0];
p2 += ks[1];
p3 += ks[2];
p4 += ks[3];
p5 += ks[4] + ts[2];
p6 += ks[5] + ts[0];
p7 += ks[6] + make_uint2(17, 0);
Round512v35(p0, p1, p2, p3, p4, p5, p6, p7, 39, 30, 34, 24);
Round512v35(p2, p1, p4, p7, p6, p5, p0, p3, 13, 50, 10, 17);
Round512v35(p4, p1, p6, p3, p0, p5, p2, p7, 25, 29, 39, 43);
Round512v35(p6, p1, p0, p7, p2, p5, p4, p3, 8, 35, 56, 22);
p0 += ks[0];
p1 += ks[1];
p2 += ks[2];
p3 += ks[3];
}
__global__ __launch_bounds__(256,3)
void skein256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint64_t *outputHash)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint2 h[9], t[3] = { t12[0], t12[1], t12[2] };
const uint2 skein_ks_parity = { 0xA9FC1A22, 0x1BD11BDA };
const uint2 t12[6] = {
{ 0x20, 0 },
{ 0, 0xf0000000 },
{ 0x20, 0xf0000000 },
{ 0x08, 0 },
{ 0, 0xff000000 },
{ 0x08, 0xff000000 }
};
h[8] = skein_ks_parity;
for (int i = 0; i<8; i++) {
h[i] = vSKEIN_IV512_256[i];
h[8] ^= h[i];
}
uint2 h[9] = {
{ 0x2FDB3E13, 0xCCD044A1 },
{ 0x1A79A9EB, 0xE8359030 },
{ 0x4F816E6F, 0x55AEA061 },
{ 0xAE9B94DB, 0x2A2767A4 },
{ 0x74DD7683, 0xEC06025E },
{ 0xC4746251, 0xE7A436CD },
{ 0x393AD185, 0xC36FBAF9 },
{ 0x33EDFC13, 0x3EEDBA18 },
{ 0xC73A4E2A, 0xB69D3CFC }
};
uint2 dt0,dt1,dt2,dt3;
uint2 p0, p1, p2, p3, p4, p5, p6, p7;
uint2 dt0 = vectorize(outputHash[thread]);
uint2 dt1 = vectorize(outputHash[threads + thread]);
uint2 dt2 = vectorize(outputHash[threads*2 + thread]);
uint2 dt3 = vectorize(outputHash[threads*3 + thread]);
LOHI(dt0.x,dt0.y,outputHash[thread]);
LOHI(dt1.x,dt1.y,outputHash[threads+thread]);
LOHI(dt2.x,dt2.y,outputHash[2*threads+thread]);
LOHI(dt3.x,dt3.y,outputHash[3*threads+thread]);
uint2 p0 = h[0] + dt0;
uint2 p1 = h[1] + dt1;
uint2 p2 = h[2] + dt2;
uint2 p3 = h[3] + dt3;
uint2 p4 = h[4];
uint2 p5 = h[5] + t[0];
uint2 p6 = h[6] + t[1];
uint2 p7 = h[7];
p0 = h[0] + dt0;
p1 = h[1] + dt1;
p2 = h[2] + dt2;
p3 = h[3] + dt3;
p4 = h[4];
p5 = h[5] + t12[0];
p6 = h[6] + t12[1];
p7 = h[7];
#pragma unroll 9
#pragma unroll
for (int i = 1; i<19; i+=2) {
Round_8_512v35(h, t, p0, p1, p2, p3, p4, p5, p6, p7, i);
Round_8_512v35(h,t12,p0,p1,p2,p3,p4,p5,p6,p7,i);
}
p0 ^= dt0;
@@ -131,51 +137,46 @@ void skein256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint64_t *outp
h[5] = p5;
h[6] = p6;
h[7] = p7;
h[8] = skein_ks_parity;
h[8] = skein_ks_parity ^ h[0] ^ h[1] ^ h[2] ^ h[3] ^ h[4] ^ h[5] ^ h[6] ^ h[7];
#pragma unroll 8
for (int i = 0; i<8; i++) {
h[8] ^= h[i];
}
const uint2 *t = t12+3;
p5 += t12[3]; //p5 already equal h[5]
p6 += t12[4];
t[0] = t12[3];
t[1] = t12[4];
t[2] = t12[5];
p5 += t[0];
p6 += t[1];
#pragma unroll 9
for (int i = 1; i<19; i+=2) {
#pragma unroll
for (int i = 1; i<17; i+=2) {
Round_8_512v35(h, t, p0, p1, p2, p3, p4, p5, p6, p7, i);
}
outputHash[thread] = devectorize(p0);
outputHash[threads + thread] = devectorize(p1);
outputHash[threads*2 + thread] = devectorize(p2);
outputHash[threads*3 + thread] = devectorize(p3);
Round_8_512v35_final(h, t, p0, p1, p2, p3, p4, p5, p6, p7);
outputHash[thread] = devectorize(p0);
outputHash[threads+thread] = devectorize(p1);
outputHash[2*threads+thread] = devectorize(p2);
outputHash[3*threads+thread] = devectorize(p3);
}
}
static __forceinline__ __device__
void Round512v30(uint64_t &p0, uint64_t &p1, uint64_t &p2, uint64_t &p3,
uint64_t &p4, uint64_t &p5, uint64_t &p6, uint64_t &p7, int ROT)
uint64_t &p4, uint64_t &p5, uint64_t &p6, uint64_t &p7,
const int ROT0, const int ROT1, const int ROT2, const int ROT3)
{
p0 += p1; p1 = ROTL64(p1, ROT256[ROT][0]) ^ p0;
p2 += p3; p3 = ROTL64(p3, ROT256[ROT][1]) ^ p2;
p4 += p5; p5 = ROTL64(p5, ROT256[ROT][2]) ^ p4;
p6 += p7; p7 = ROTL64(p7, ROT256[ROT][3]) ^ p6;
p0 += p1; p1 = ROTL64(p1, ROT0) ^ p0;
p2 += p3; p3 = ROTL64(p3, ROT1) ^ p2;
p4 += p5; p5 = ROTL64(p5, ROT2) ^ p4;
p6 += p7; p7 = ROTL64(p7, ROT3) ^ p6;
}
static __forceinline__ __device__
void Round_8_512v30(uint64_t *ks, uint64_t *ts, uint64_t &p0, uint64_t &p1, uint64_t &p2, uint64_t &p3,
void Round_8_512v30(uint64_t *ks, uint64_t *ts,
uint64_t &p0, uint64_t &p1, uint64_t &p2, uint64_t &p3,
uint64_t &p4, uint64_t &p5, uint64_t &p6, uint64_t &p7, int R)
{
Round512v30(p0, p1, p2, p3, p4, p5, p6, p7, 0);
Round512v30(p2, p1, p4, p7, p6, p5, p0, p3, 1);
Round512v30(p4, p1, p6, p3, p0, p5, p2, p7, 2);
Round512v30(p6, p1, p0, p7, p2, p5, p4, p3, 3);
p0 += ks[(R+0) % 9]; /* inject the key schedule value */
Round512v30(p0, p1, p2, p3, p4, p5, p6, p7, 46, 36, 19, 37);
Round512v30(p2, p1, p4, p7, p6, p5, p0, p3, 33, 27, 14, 42);
Round512v30(p4, p1, p6, p3, p0, p5, p2, p7, 17, 49, 36, 39);
Round512v30(p6, p1, p0, p7, p2, p5, p4, p3, 44, 9, 54, 56);
p0 += ks[(R+0) % 9];
p1 += ks[(R+1) % 9];
p2 += ks[(R+2) % 9];
p3 += ks[(R+3) % 9];
@@ -183,11 +184,13 @@ void Round_8_512v30(uint64_t *ks, uint64_t *ts, uint64_t &p0, uint64_t &p1, uint
p5 += ks[(R+5) % 9] + ts[(R+0) % 3];
p6 += ks[(R+6) % 9] + ts[(R+1) % 3];
p7 += ks[(R+7) % 9] + R;
Round512v30(p0, p1, p2, p3, p4, p5, p6, p7, 4);
Round512v30(p2, p1, p4, p7, p6, p5, p0, p3, 5);
Round512v30(p4, p1, p6, p3, p0, p5, p2, p7, 6);
Round512v30(p6, p1, p0, p7, p2, p5, p4, p3, 7);
p0 += ks[(R+1) % 9]; /* inject the key schedule value */
Round512v30(p0, p1, p2, p3, p4, p5, p6, p7, 39, 30, 34, 24);
Round512v30(p2, p1, p4, p7, p6, p5, p0, p3, 13, 50, 10, 17);
Round512v30(p4, p1, p6, p3, p0, p5, p2, p7, 25, 29, 39, 43);
Round512v30(p6, p1, p0, p7, p2, p5, p4, p3, 8, 35, 56, 22);
p0 += ks[(R+1) % 9];
p1 += ks[(R+2) % 9];
p2 += ks[(R+3) % 9];
p3 += ks[(R+4) % 9];
@@ -197,29 +200,28 @@ void Round_8_512v30(uint64_t *ks, uint64_t *ts, uint64_t &p0, uint64_t &p1, uint
p7 += ks[(R+8) % 9] + R+1;
}
#define skein_ks_parity64 0x1BD11BDAA9FC1A22ull
#include <stdio.h>
__global__ __launch_bounds__(256, 3)
void skein256_gpu_hash_32_v30(uint32_t threads, uint32_t startNounce, uint64_t *outputHash)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint64_t h[9], t[3];
h[8] = skein_ks_parity64;
for (int i = 0; i<8; i++) {
h[i] = SKEIN_IV512_256[i];
h[8] ^= h[i];
}
t[0] = devectorize(t12[0]);
t[1] = devectorize(t12[1]);
t[2] = devectorize(t12[2]);
uint64_t h[12] = { // SKEIN_IV512_256
0xCCD044A12FDB3E13, 0xE83590301A79A9EB,
0x55AEA0614F816E6F, 0x2A2767A4AE9B94DB,
0xEC06025E74DD7683, 0xE7A436CDC4746251,
0xC36FBAF9393AD185, 0x3EEDBA1833EDFC13,
0xb69d3cfcc73a4e2a, // skein_ks_parity64 ^ h[0..7]
0x20, 0xf000000000000000, 0xf000000000000020 // t0..2
};
uint64_t dt0 = outputHash[thread];
uint64_t dt1 = outputHash[threads + thread];
uint64_t dt2 = outputHash[threads*2 + thread];
uint64_t dt3 = outputHash[threads*3 + thread];
uint64_t *t = &h[9];
uint64_t p0 = h[0] + dt0;
uint64_t p1 = h[1] + dt1;
uint64_t p2 = h[2] + dt2;
@@ -254,9 +256,9 @@ void skein256_gpu_hash_32_v30(uint32_t threads, uint32_t startNounce, uint64_t *
h[8] ^= h[i];
}
t[0] = devectorize(t12[3]);
t[1] = devectorize(t12[4]);
t[2] = devectorize(t12[5]);
t[0] = 0x08;
t[1] = 0xff00000000000000;
t[2] = 0xff00000000000008;
p5 += t[0]; //p5 already equal h[5]
p6 += t[1];
@@ -270,7 +272,6 @@ void skein256_gpu_hash_32_v30(uint32_t threads, uint32_t startNounce, uint64_t *
outputHash[threads + thread] = p1;
outputHash[threads*2 + thread] = p2;
outputHash[threads*3 + thread] = p3;
} //thread
}
+22 -4
View File
@@ -17,6 +17,7 @@ extern "C"
static uint32_t *d_hash[MAX_GPUS];
extern void keccak256_cpu_init(int thr_id, uint32_t threads);
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);
@@ -35,10 +36,11 @@ extern "C" void keccak256_hash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_keccak256(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_keccak256(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];
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 21); // 256*256*8*4
throughput = min(throughput, max_nonce - first_nonce);
@@ -55,7 +57,6 @@ extern "C" int scanhash_keccak256(int thr_id, uint32_t *pdata,
init[thr_id] = true;
}
uint32_t endiandata[20];
for (int k=0; k < 20; k++) {
be32enc(&endiandata[k], pdata[k]);
}
@@ -74,6 +75,7 @@ extern "C" int scanhash_keccak256(int thr_id, uint32_t *pdata,
keccak256_hash(vhash64, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[19] = foundNonce;
return 1;
}
@@ -92,3 +94,19 @@ extern "C" int scanhash_keccak256(int thr_id, uint32_t *pdata,
return 0;
}
// cleanup
extern "C" void free_keccak256(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
keccak256_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+31 -4
View File
@@ -87,10 +87,11 @@ extern "C" unsigned int jackpothash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_jackpot(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_jackpot(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];
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 20);
@@ -123,7 +124,6 @@ extern "C" int scanhash_jackpot(int thr_id, uint32_t *pdata,
init[thr_id] = true;
}
uint32_t endiandata[22];
for (int k=0; k < 22; k++)
be32enc(&endiandata[k], pdata[k]);
@@ -221,7 +221,12 @@ extern "C" int scanhash_jackpot(int thr_id, uint32_t *pdata,
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
nist5hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -245,3 +250,25 @@ extern "C" int scanhash_jackpot(int thr_id, uint32_t *pdata,
return 0;
}
// cleanup
extern "C" void free_jackpot(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cudaFree(d_branch1Nonces[thr_id]);
cudaFree(d_branch2Nonces[thr_id]);
cudaFree(d_branch3Nonces[thr_id]);
cudaFree(d_jackpotNonces[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+9 -4
View File
@@ -21,7 +21,6 @@ ccminer_SOURCES = elist.h miner.h compat.h \
crc32.c hefty1.c \
ccminer.cpp pools.cpp util.cpp bignum.cpp \
api.cpp hashlog.cpp nvml.cpp stats.cpp sysinfos.cpp cuda.cpp \
ethash/etherpc.cpp ethash/ethash.cu \
heavy/heavy.cu \
heavy/cuda_blake512.cu heavy/cuda_blake512.h \
heavy/cuda_combine.cu heavy/cuda_combine.h \
@@ -34,14 +33,16 @@ ccminer_SOURCES = elist.h miner.h compat.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 \
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/blake256.cu Algo256/keccak256.cu \
Algo256/bmw.cu Algo256/cuda_bmw.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/quarkcoin.cu quark/animecoin.cu \
quark/cuda_quark_compactionTest.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 \
sph/bmw.c sph/blake.c sph/groestl.c sph/jh.c sph/keccak.c sph/skein.c \
@@ -56,7 +57,8 @@ ccminer_SOURCES = elist.h miner.h compat.h \
x15/x14.cu x15/x15.cu x15/cuda_x14_shabal512.cu x15/cuda_x15_whirlpool.cu \
x15/whirlpoolx.cu x15/cuda_whirlpoolx.cu \
x17/x17.cu x17/cuda_x17_haval512.cu x17/cuda_x17_sha512.cu \
x11/c11.cu x11/s3.cu
x11/c11.cu x11/s3.cu \
spreadx11/spreadx11.cu spreadx11/cuda_blake.cu spreadx11/cuda_spreadx11_sign.cu spreadx11/cuda_sha256.cu
# scrypt
ccminer_SOURCES += scrypt.cpp scrypt-jane.cpp \
@@ -95,6 +97,9 @@ nvcc_FLAGS += $(JANSSON_INCLUDES) --ptxas-options="-v"
Algo256/blake256.o: Algo256/blake256.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=64 -o $@ -c $<
Algo256/cuda_bmw.o: Algo256/cuda_bmw.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=76 -o $@ -c $<
heavy/cuda_hefty1.o: heavy/cuda_hefty1.cu
$(NVCC) $(nvcc_FLAGS) --maxrregcount=80 -o $@ -c $<
+18 -5
View File
@@ -1,5 +1,5 @@
ccMiner release 1.6.5-C11 (Jul 2015) "Special for Chaincoin C11"
ccMiner release 1.6.6 (Aug 2015) "Lyra2REv2, BMW And Vanilla"
---------------------------------------------------------------
***************************************************************
@@ -34,6 +34,7 @@ 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)
@@ -64,9 +65,9 @@ This code is based on the pooler cpuminer and inherits
its command line interface and options.
-a, --algo=ALGO specify the algorithm to use
anime use to mine Animecoin
blake use to mine Saffroncoin (Blake 256)
blakecoin use to mine Old Blake 256
bmw use to mine Midnight
c11/flax use to mine Chaincoin and Flax
deep use to mine Deepcoin
dmd-gr use to mine Diamond-Groestl
@@ -105,6 +106,7 @@ its command line interface and options.
-i, --intensity=N[,N] GPU threads per call 8-25 (2^N + F, default: 0=auto)
Decimals and multiple values are allowed for fine tuning
--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)
@@ -120,7 +122,7 @@ its command line interface and options.
(default: retry indefinitely)
-R, --retry-pause=N time to pause between retries, in seconds (default: 15)
--time-limit maximum time [s] to mine before exiting the program.
-T, --timeout=N network timeout, in seconds (default: 270)
-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)
-n, --ndevs list cuda devices
@@ -140,12 +142,14 @@ its command line interface and options.
--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
--keep-clocks prevent reset clocks and/or power limit on exit
--show-diff display submitted block and net difficulty
-B, --background run the miner in the background
--benchmark run in offline benchmark mode
--cputest debug hashes from cpu algorithms
--cpu-affinity set process affinity to specific cpu core(s) mask
--cpu-priority set process priority (default: 0 idle, 2 normal to 5 highest)
-c, --config=FILE load a JSON-format configuration file
can be from an url with the http:// prefix
-V, --version display version information and exit
-h, --help display this help text and exit
@@ -223,8 +227,17 @@ features.
>>> RELEASE HISTORY <<<
August 2015...
Restore WhirlpoolX algo
Under Dev... v1.7
Add --cuda-schedule parameter
Add --show-diff parameter, which display shares diff,
and is able to detect real solved blocks on pools.
Aug. 28th 2015 v1.6.6
Allow to load remote config with curl (-c http://...)
Add Lyra2REv2 algo (Vertcoin/Zoom)
Restore WhirlpoolX algo (VNL)
Drop Animecoin support
Add bmw (Midnight) algo
July 06th 2015 v1.6.5-C11
Nvml api power limits
+1 -1
View File
@@ -878,7 +878,7 @@ static void api()
buffer = (char *) calloc(1, MYBUFSIZ + 1);
counter = 0;
while (bye == 0) {
while (bye == 0 && !abort_flag) {
counter++;
clisiz = sizeof(cli);
+54
View File
@@ -24,3 +24,57 @@ extern "C" void bn_nbits_to_uchar(const uint32_t nBits, unsigned char *target)
snprintf(buff, 65, "%s\n", bn.ToString().c_str()); buff[64] = '\0';
hex2bin(target, buff, 32);
}
// unused, but should allow more than 256bits targets
#if 0
extern "C" double bn_hash_target_ratio(uint32_t* hash, uint32_t* target)
{
double dhash;
if (!opt_showdiff)
return 0.0;
CBigNum h(0), t(0);
std::vector<unsigned char> vch(32);
memcpy(&vch[0], (void*) target, 32);
t.setvch(vch);
memcpy(&vch[0], (void*) hash, 32);
h.setvch(vch);
dhash = h.getuint256().getdouble();
if (dhash > 0.)
return t.getuint256().getdouble() / dhash;
else
return dhash;
}
#endif
// compute the diff ratio between a found hash and the target
extern "C" double bn_hash_target_ratio(uint32_t* hash, uint32_t* target)
{
uint256 h, t;
double dhash;
if (!opt_showdiff)
return 0.0;
memcpy(&t, (void*) target, 32);
memcpy(&h, (void*) hash, 32);
dhash = h.getdouble();
if (dhash > 0.)
return t.getdouble() / dhash;
else
return dhash;
}
// store ratio in work struct
extern "C" void bn_store_hash_target_ratio(uint32_t* hash, uint32_t* target, struct work* work)
{
// 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;
}
}
+401 -305
View File
File diff suppressed because it is too large. Load diff
+19 -12
View File
@@ -135,6 +135,7 @@
<IntrinsicFunctions>true</IntrinsicFunctions>
<PreprocessorDefinitions>WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;CURL_STATICLIB;USE_WRAPNVML;SCRYPT_KECCAK512;SCRYPT_CHACHA;SCRYPT_CHOOSE_COMPILETIME;%(PreprocessorDefinitions)</PreprocessorDefinitions>
<AdditionalIncludeDirectories>.;compat;compat\curl-for-windows\curl\include;compat\jansson;compat\getopt;compat\pthreads;compat\curl-for-windows\openssl\openssl\include;compat\curl-for-windows\zlib;%(AdditionalIncludeDirectories);$(CudaToolkitIncludeDir)</AdditionalIncludeDirectories>
<MinimalRebuild>true</MinimalRebuild>
<EnableParallelCodeGeneration>true</EnableParallelCodeGeneration>
<MultiProcessorCompilation>true</MultiProcessorCompilation>
<OpenMPSupport>true</OpenMPSupport>
@@ -180,6 +181,7 @@
<IntrinsicFunctions>true</IntrinsicFunctions>
<PreprocessorDefinitions>WIN32;NDEBUG;_CONSOLE;_CRT_SECURE_NO_WARNINGS;CURL_STATICLIB;USE_WRAPNVML;SCRYPT_KECCAK512;SCRYPT_CHACHA;SCRYPT_CHOOSE_COMPILETIME;%(PreprocessorDefinitions)</PreprocessorDefinitions>
<AdditionalIncludeDirectories>.;compat;compat\curl-for-windows\curl\include;compat\jansson;compat\getopt;compat\pthreads;compat\curl-for-windows\openssl\openssl\include;compat\curl-for-windows\zlib;%(AdditionalIncludeDirectories);$(CudaToolkitIncludeDir)</AdditionalIncludeDirectories>
<MinimalRebuild>true</MinimalRebuild>
<MultiProcessorCompilation>true</MultiProcessorCompilation>
</ClCompile>
<Link>
@@ -247,6 +249,10 @@
<ClCompile Include="neoscrypt\neoscrypt.cpp" />
<ClCompile Include="neoscrypt\neoscrypt-cpu.c" />
<ClInclude Include="neoscrypt\cuda_vectors.h" />
<CudaCompile Include="Algo256\bmw.cu" />
<CudaCompile Include="Algo256\cuda_bmw.cu">
<MaxRegCount>76</MaxRegCount>
</CudaCompile>
<CudaCompile Include="neoscrypt\cuda_neoscrypt.cu" />
<ClCompile Include="scrypt-jane.cpp" />
<ClCompile Include="scrypt.cpp" />
@@ -330,6 +336,8 @@
<ClInclude Include="lyra2\Sponge.h" />
<ClInclude Include="quark\groestl_transf_quad.h" />
<ClInclude Include="quark\groestl_functions_quad.h" />
<ClInclude Include="spreadx11\bignum_fp.h" />
<ClInclude Include="spreadx11\spreadx11.h" />
</ItemGroup>
<ItemGroup>
<CudaCompile Include="cuda.cpp" />
@@ -392,6 +400,8 @@
</CudaCompile>
<CudaCompile Include="Algo256\keccak256.cu" />
<CudaCompile Include="Algo256\cuda_blake256.cu" />
<CudaCompile Include="Algo256\cuda_bmw256.cu" />
<CudaCompile Include="Algo256\cuda_cubehash256.cu" />
<CudaCompile Include="Algo256\cuda_fugue256.cu" />
<CudaCompile Include="Algo256\cuda_groestl256.cu" />
<CudaCompile Include="Algo256\cuda_keccak256.cu">
@@ -403,8 +413,6 @@
<AdditionalOptions Condition="'$(Configuration)'=='Release'">--ptxas-options="-dlcm=cg" %(AdditionalOptions)</AdditionalOptions>
<FastMath>true</FastMath>
</CudaCompile>
<CudaCompile Include="quark\animecoin.cu">
</CudaCompile>
<CudaCompile Include="quark\cuda_bmw512.cu">
<MaxRegCount>128</MaxRegCount>
</CudaCompile>
@@ -440,12 +448,18 @@
</CudaCompile>
<CudaCompile Include="lyra2\lyra2RE.cu" />
<CudaCompile Include="lyra2\cuda_lyra2.cu" />
<CudaCompile Include="lyra2\lyra2REv2.cu" />
<CudaCompile Include="lyra2\cuda_lyra2v2.cu" />
<CudaCompile Include="skein.cu">
<MaxRegCount>64</MaxRegCount>
</CudaCompile>
<CudaCompile Include="cuda_skeincoin.cu">
<MaxRegCount>48</MaxRegCount>
</CudaCompile>
<CudaCompile Include="spreadx11\spreadx11.cu" />
<CudaCompile Include="spreadx11\cuda_blake.cu" />
<CudaCompile Include="spreadx11\cuda_spreadx11_sign.cu" />
<CudaCompile Include="spreadx11\cuda_sha256.cu" />
<CudaCompile Include="x11\cuda_x11_aes.cu">
<ExcludedFromBuild>true</ExcludedFromBuild>
</CudaCompile>
@@ -479,10 +493,9 @@
</CudaCompile>
<CudaCompile Include="x15\x14.cu">
</CudaCompile>
<CudaCompile Include="x15\cuda_x14_shabal512.cu">
</CudaCompile>
<CudaCompile Include="x15\x15.cu">
</CudaCompile>
<CudaCompile Include="x15\cuda_x14_shabal512.cu" />
<CudaCompile Include="x15\cuda_x15_whirlpool.cu" />
<CudaCompile Include="x15\x15.cu" />
<CudaCompile Include="x15\cuda_whirlpoolx.cu" />
<CudaCompile Include="x15\whirlpoolx.cu" />
<CudaCompile Include="x17\x17.cu">
@@ -492,12 +505,6 @@
<CudaCompile Include="x17\cuda_x17_sha512.cu">
<MaxRegCount>80</MaxRegCount>
</CudaCompile>
<CudaCompile Include="x15\x14.cu">
</CudaCompile>
<CudaCompile Include="x15\x15.cu">
</CudaCompile>
<CudaCompile Include="x17\x17.cu">
</CudaCompile>
</ItemGroup>
<ItemGroup>
<Image Include="res\ccminer.ico" />
+19 -4
View File
@@ -466,9 +466,6 @@
<CudaCompile Include="heavy\heavy.cu">
<Filter>Source Files\CUDA\heavy</Filter>
</CudaCompile>
<CudaCompile Include="quark\animecoin.cu">
<Filter>Source Files\CUDA\quark</Filter>
</CudaCompile>
<CudaCompile Include="quark\cuda_bmw512.cu">
<Filter>Source Files\CUDA\quark</Filter>
</CudaCompile>
@@ -583,6 +580,12 @@
<CudaCompile Include="Algo256\cuda_blake256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_bmw256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_cubehash256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_fugue256.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
@@ -601,6 +604,12 @@
<CudaCompile Include="lyra2\lyra2RE.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\cuda_lyra2v2.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<CudaCompile Include="lyra2\lyra2REv2.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
<CudaCompile Include="zr5.cu">
<Filter>Source Files\CUDA</Filter>
</CudaCompile>
@@ -643,6 +652,12 @@
<CudaCompile Include="neoscrypt\cuda_neoscrypt.cu">
<Filter>Source Files\neoscrypt</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\bmw.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
<CudaCompile Include="Algo256\cuda_bmw.cu">
<Filter>Source Files\CUDA\Algo256</Filter>
</CudaCompile>
</ItemGroup>
<ItemGroup>
<Image Include="res\ccminer.ico">
@@ -659,4 +674,4 @@
<Filter>Ressources</Filter>
</Text>
</ItemGroup>
</Project>
</Project>
+4 -1
View File
@@ -4,8 +4,9 @@
#ifdef WIN32
#include <windows.h>
#include <time.h>
extern int opt_priority;
#define localtime_r(src, dst) localtime_s(dst, src)
static __inline void sleep(int secs)
{
@@ -16,6 +17,8 @@ enum {
PRIO_PROCESS = 0,
};
extern int opt_priority;
static __inline int setpriority(int which, int who, int prio)
{
switch (opt_priority) {
Regular → Executable
+347 -310
View File
@@ -1,310 +1,347 @@
#! /bin/sh
# Wrapper for compilers which do not understand '-c -o'.
scriptversion=2012-01-04.17; # UTC
# Copyright (C) 1999, 2000, 2003, 2004, 2005, 2009, 2010, 2012 Free
# Software Foundation, Inc.
# Written by Tom Tromey <tromey@cygnus.com>.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2, or (at your option)
# any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
# As a special exception to the GNU General Public License, if you
# distribute this file as part of a program that contains a
# configuration script generated by Autoconf, you may include it under
# the same distribution terms that you use for the rest of that program.
# This file is maintained in Automake, please report
# bugs to <bug-automake@gnu.org> or send patches to
# <automake-patches@gnu.org>.
nl='
'
# We need space, tab and new line, in precisely that order. Quoting is
# there to prevent tools from complaining about whitespace usage.
IFS=" "" $nl"
file_conv=
# func_file_conv build_file lazy
# Convert a $build file to $host form and store it in $file
# Currently only supports Windows hosts. If the determined conversion
# type is listed in (the comma separated) LAZY, no conversion will
# take place.
func_file_conv ()
{
file=$1
case $file in
/ | /[!/]*) # absolute file, and not a UNC file
if test -z "$file_conv"; then
# lazily determine how to convert abs files
case `uname -s` in
MINGW*)
file_conv=mingw
;;
CYGWIN*)
file_conv=cygwin
;;
*)
file_conv=wine
;;
esac
fi
case $file_conv/,$2, in
*,$file_conv,*)
;;
mingw/*)
file=`cmd //C echo "$file " | sed -e 's/"\(.*\) " *$/\1/'`
;;
cygwin/*)
file=`cygpath -m "$file" || echo "$file"`
;;
wine/*)
file=`winepath -w "$file" || echo "$file"`
;;
esac
;;
esac
}
# func_cl_wrapper cl arg...
# Adjust compile command to suit cl
func_cl_wrapper ()
{
# Assume a capable shell
lib_path=
shared=:
linker_opts=
for arg
do
if test -n "$eat"; then
eat=
else
case $1 in
-o)
# configure might choose to run compile as 'compile cc -o foo foo.c'.
eat=1
case $2 in
*.o | *.[oO][bB][jJ])
func_file_conv "$2"
set x "$@" -Fo"$file"
shift
;;
*)
func_file_conv "$2"
set x "$@" -Fe"$file"
shift
;;
esac
;;
-I*)
func_file_conv "${1#-I}" mingw
set x "$@" -I"$file"
shift
;;
-l*)
lib=${1#-l}
found=no
save_IFS=$IFS
IFS=';'
for dir in $lib_path $LIB
do
IFS=$save_IFS
if $shared && test -f "$dir/$lib.dll.lib"; then
found=yes
set x "$@" "$dir/$lib.dll.lib"
break
fi
if test -f "$dir/$lib.lib"; then
found=yes
set x "$@" "$dir/$lib.lib"
break
fi
done
IFS=$save_IFS
test "$found" != yes && set x "$@" "$lib.lib"
shift
;;
-L*)
func_file_conv "${1#-L}"
if test -z "$lib_path"; then
lib_path=$file
else
lib_path="$lib_path;$file"
fi
linker_opts="$linker_opts -LIBPATH:$file"
;;
-static)
shared=false
;;
-Wl,*)
arg=${1#-Wl,}
save_ifs="$IFS"; IFS=','
for flag in $arg; do
IFS="$save_ifs"
linker_opts="$linker_opts $flag"
done
IFS="$save_ifs"
;;
-Xlinker)
eat=1
linker_opts="$linker_opts $2"
;;
-*)
set x "$@" "$1"
shift
;;
*.cc | *.CC | *.cxx | *.CXX | *.[cC]++)
func_file_conv "$1"
set x "$@" -Tp"$file"
shift
;;
*.c | *.cpp | *.CPP | *.lib | *.LIB | *.Lib | *.OBJ | *.obj | *.[oO])
func_file_conv "$1" mingw
set x "$@" "$file"
shift
;;
*)
set x "$@" "$1"
shift
;;
esac
fi
shift
done
if test -n "$linker_opts"; then
linker_opts="-link$linker_opts"
fi
exec "$@" $linker_opts
exit 1
}
eat=
case $1 in
'')
echo "$0: No command. Try '$0 --help' for more information." 1>&2
exit 1;
;;
-h | --h*)
cat <<\EOF
Usage: compile [--help] [--version] PROGRAM [ARGS]
Wrapper for compilers which do not understand '-c -o'.
Remove '-o dest.o' from ARGS, run PROGRAM with the remaining
arguments, and rename the output as expected.
If you are trying to build a whole package this is not the
right script to run: please start by reading the file 'INSTALL'.
Report bugs to <bug-automake@gnu.org>.
EOF
exit $?
;;
-v | --v*)
echo "compile $scriptversion"
exit $?
;;
cl | *[/\\]cl | cl.exe | *[/\\]cl.exe )
func_cl_wrapper "$@" # Doesn't return...
;;
esac
ofile=
cfile=
for arg
do
if test -n "$eat"; then
eat=
else
case $1 in
-o)
# configure might choose to run compile as 'compile cc -o foo foo.c'.
# So we strip '-o arg' only if arg is an object.
eat=1
case $2 in
*.o | *.obj)
ofile=$2
;;
*)
set x "$@" -o "$2"
shift
;;
esac
;;
*.c)
cfile=$1
set x "$@" "$1"
shift
;;
*)
set x "$@" "$1"
shift
;;
esac
fi
shift
done
if test -z "$ofile" || test -z "$cfile"; then
# If no '-o' option was seen then we might have been invoked from a
# pattern rule where we don't need one. That is ok -- this is a
# normal compilation that the losing compiler can handle. If no
# '.c' file was seen then we are probably linking. That is also
# ok.
exec "$@"
fi
# Name of file we expect compiler to create.
cofile=`echo "$cfile" | sed 's|^.*[\\/]||; s|^[a-zA-Z]:||; s/\.c$/.o/'`
# Create the lock directory.
# Note: use '[/\\:.-]' here to ensure that we don't use the same name
# that we are using for the .o file. Also, base the name on the expected
# object file name, since that is what matters with a parallel build.
lockdir=`echo "$cofile" | sed -e 's|[/\\:.-]|_|g'`.d
while true; do
if mkdir "$lockdir" >/dev/null 2>&1; then
break
fi
sleep 1
done
# FIXME: race condition here if user kills between mkdir and trap.
trap "rmdir '$lockdir'; exit 1" 1 2 15
# Run the compile.
"$@"
ret=$?
if test -f "$cofile"; then
test "$cofile" = "$ofile" || mv "$cofile" "$ofile"
elif test -f "${cofile}bj"; then
test "${cofile}bj" = "$ofile" || mv "${cofile}bj" "$ofile"
fi
rmdir "$lockdir"
exit $ret
# Local Variables:
# mode: shell-script
# sh-indentation: 2
# eval: (add-hook 'write-file-hooks 'time-stamp)
# time-stamp-start: "scriptversion="
# time-stamp-format: "%:y-%02m-%02d.%02H"
# time-stamp-time-zone: "UTC"
# time-stamp-end: "; # UTC"
# End:
#! /bin/sh
# Wrapper for compilers which do not understand '-c -o'.
scriptversion=2012-10-14.11; # UTC
# Copyright (C) 1999-2013 Free Software Foundation, Inc.
# Written by Tom Tromey <tromey@cygnus.com>.
#
# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2, or (at your option)
# any later version.
#
# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program. If not, see <http://www.gnu.org/licenses/>.
# As a special exception to the GNU General Public License, if you
# distribute this file as part of a program that contains a
# configuration script generated by Autoconf, you may include it under
# the same distribution terms that you use for the rest of that program.
# This file is maintained in Automake, please report
# bugs to <bug-automake@gnu.org> or send patches to
# <automake-patches@gnu.org>.
nl='
'
# We need space, tab and new line, in precisely that order. Quoting is
# there to prevent tools from complaining about whitespace usage.
IFS=" "" $nl"
file_conv=
# func_file_conv build_file lazy
# Convert a $build file to $host form and store it in $file
# Currently only supports Windows hosts. If the determined conversion
# type is listed in (the comma separated) LAZY, no conversion will
# take place.
func_file_conv ()
{
file=$1
case $file in
/ | /[!/]*) # absolute file, and not a UNC file
if test -z "$file_conv"; then
# lazily determine how to convert abs files
case `uname -s` in
MINGW*)
file_conv=mingw
;;
CYGWIN*)
file_conv=cygwin
;;
*)
file_conv=wine
;;
esac
fi
case $file_conv/,$2, in
*,$file_conv,*)
;;
mingw/*)
file=`cmd //C echo "$file " | sed -e 's/"\(.*\) " *$/\1/'`
;;
cygwin/*)
file=`cygpath -m "$file" || echo "$file"`
;;
wine/*)
file=`winepath -w "$file" || echo "$file"`
;;
esac
;;
esac
}
# func_cl_dashL linkdir
# Make cl look for libraries in LINKDIR
func_cl_dashL ()
{
func_file_conv "$1"
if test -z "$lib_path"; then
lib_path=$file
else
lib_path="$lib_path;$file"
fi
linker_opts="$linker_opts -LIBPATH:$file"
}
# func_cl_dashl library
# Do a library search-path lookup for cl
func_cl_dashl ()
{
lib=$1
found=no
save_IFS=$IFS
IFS=';'
for dir in $lib_path $LIB
do
IFS=$save_IFS
if $shared && test -f "$dir/$lib.dll.lib"; then
found=yes
lib=$dir/$lib.dll.lib
break
fi
if test -f "$dir/$lib.lib"; then
found=yes
lib=$dir/$lib.lib
break
fi
if test -f "$dir/lib$lib.a"; then
found=yes
lib=$dir/lib$lib.a
break
fi
done
IFS=$save_IFS
if test "$found" != yes; then
lib=$lib.lib
fi
}
# func_cl_wrapper cl arg...
# Adjust compile command to suit cl
func_cl_wrapper ()
{
# Assume a capable shell
lib_path=
shared=:
linker_opts=
for arg
do
if test -n "$eat"; then
eat=
else
case $1 in
-o)
# configure might choose to run compile as 'compile cc -o foo foo.c'.
eat=1
case $2 in
*.o | *.[oO][bB][jJ])
func_file_conv "$2"
set x "$@" -Fo"$file"
shift
;;
*)
func_file_conv "$2"
set x "$@" -Fe"$file"
shift
;;
esac
;;
-I)
eat=1
func_file_conv "$2" mingw
set x "$@" -I"$file"
shift
;;
-I*)
func_file_conv "${1#-I}" mingw
set x "$@" -I"$file"
shift
;;
-l)
eat=1
func_cl_dashl "$2"
set x "$@" "$lib"
shift
;;
-l*)
func_cl_dashl "${1#-l}"
set x "$@" "$lib"
shift
;;
-L)
eat=1
func_cl_dashL "$2"
;;
-L*)
func_cl_dashL "${1#-L}"
;;
-static)
shared=false
;;
-Wl,*)
arg=${1#-Wl,}
save_ifs="$IFS"; IFS=','
for flag in $arg; do
IFS="$save_ifs"
linker_opts="$linker_opts $flag"
done
IFS="$save_ifs"
;;
-Xlinker)
eat=1
linker_opts="$linker_opts $2"
;;
-*)
set x "$@" "$1"
shift
;;
*.cc | *.CC | *.cxx | *.CXX | *.[cC]++)
func_file_conv "$1"
set x "$@" -Tp"$file"
shift
;;
*.c | *.cpp | *.CPP | *.lib | *.LIB | *.Lib | *.OBJ | *.obj | *.[oO])
func_file_conv "$1" mingw
set x "$@" "$file"
shift
;;
*)
set x "$@" "$1"
shift
;;
esac
fi
shift
done
if test -n "$linker_opts"; then
linker_opts="-link$linker_opts"
fi
exec "$@" $linker_opts
exit 1
}
eat=
case $1 in
'')
echo "$0: No command. Try '$0 --help' for more information." 1>&2
exit 1;
;;
-h | --h*)
cat <<\EOF
Usage: compile [--help] [--version] PROGRAM [ARGS]
Wrapper for compilers which do not understand '-c -o'.
Remove '-o dest.o' from ARGS, run PROGRAM with the remaining
arguments, and rename the output as expected.
If you are trying to build a whole package this is not the
right script to run: please start by reading the file 'INSTALL'.
Report bugs to <bug-automake@gnu.org>.
EOF
exit $?
;;
-v | --v*)
echo "compile $scriptversion"
exit $?
;;
cl | *[/\\]cl | cl.exe | *[/\\]cl.exe )
func_cl_wrapper "$@" # Doesn't return...
;;
esac
ofile=
cfile=
for arg
do
if test -n "$eat"; then
eat=
else
case $1 in
-o)
# configure might choose to run compile as 'compile cc -o foo foo.c'.
# So we strip '-o arg' only if arg is an object.
eat=1
case $2 in
*.o | *.obj)
ofile=$2
;;
*)
set x "$@" -o "$2"
shift
;;
esac
;;
*.c)
cfile=$1
set x "$@" "$1"
shift
;;
*)
set x "$@" "$1"
shift
;;
esac
fi
shift
done
if test -z "$ofile" || test -z "$cfile"; then
# If no '-o' option was seen then we might have been invoked from a
# pattern rule where we don't need one. That is ok -- this is a
# normal compilation that the losing compiler can handle. If no
# '.c' file was seen then we are probably linking. That is also
# ok.
exec "$@"
fi
# Name of file we expect compiler to create.
cofile=`echo "$cfile" | sed 's|^.*[\\/]||; s|^[a-zA-Z]:||; s/\.c$/.o/'`
# Create the lock directory.
# Note: use '[/\\:.-]' here to ensure that we don't use the same name
# that we are using for the .o file. Also, base the name on the expected
# object file name, since that is what matters with a parallel build.
lockdir=`echo "$cofile" | sed -e 's|[/\\:.-]|_|g'`.d
while true; do
if mkdir "$lockdir" >/dev/null 2>&1; then
break
fi
sleep 1
done
# FIXME: race condition here if user kills between mkdir and trap.
trap "rmdir '$lockdir'; exit 1" 1 2 15
# Run the compile.
"$@"
ret=$?
if test -f "$cofile"; then
test "$cofile" = "$ofile" || mv "$cofile" "$ofile"
elif test -f "${cofile}bj"; then
test "${cofile}bj" = "$ofile" || mv "${cofile}bj" "$ofile"
fi
rmdir "$lockdir"
exit $ret
# Local Variables:
# mode: shell-script
# sh-indentation: 2
# eval: (add-hook 'write-file-hooks 'time-stamp)
# time-stamp-start: "scriptversion="
# time-stamp-format: "%:y-%02m-%02d.%02H"
# time-stamp-time-zone: "UTC"
# time-stamp-end: "; # UTC"
# End:
Vendored Regular → Executable
+332 -300
View File
@@ -1,14 +1,12 @@
#! /bin/sh
# Attempt to guess a canonical system name.
# Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
# 2000, 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008
# Free Software Foundation, Inc.
# Copyright 1992-2013 Free Software Foundation, Inc.
timestamp='2008-01-23'
timestamp='2013-06-10'
# This file is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2 of the License, or
# the Free Software Foundation; either version 3 of the License, or
# (at your option) any later version.
#
# This program is distributed in the hope that it will be useful, but
@@ -17,26 +15,22 @@ timestamp='2008-01-23'
# General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 51 Franklin Street - Fifth Floor, Boston, MA
# 02110-1301, USA.
# along with this program; if not, see <http://www.gnu.org/licenses/>.
#
# As a special exception to the GNU General Public License, if you
# distribute this file as part of a program that contains a
# configuration script generated by Autoconf, you may include it under
# the same distribution terms that you use for the rest of that program.
# Originally written by Per Bothner <per@bothner.com>.
# Please send patches to <config-patches@gnu.org>. Submit a context
# diff and a properly formatted ChangeLog entry.
# the same distribution terms that you use for the rest of that
# program. This Exception is an additional permission under section 7
# of the GNU General Public License, version 3 ("GPLv3").
#
# This script attempts to guess a canonical system name similar to
# config.sub. If it succeeds, it prints the system name on stdout, and
# exits with 0. Otherwise, it exits with 1.
# Originally written by Per Bothner.
#
# The plan is that this can be called by configure scripts if you
# don't specify an explicit build system type.
# You can get the latest version of this script from:
# http://git.savannah.gnu.org/gitweb/?p=config.git;a=blob_plain;f=config.guess;hb=HEAD
#
# Please send patches with a ChangeLog entry to config-patches@gnu.org.
me=`echo "$0" | sed -e 's,.*/,,'`
@@ -56,8 +50,7 @@ version="\
GNU config.guess ($timestamp)
Originally written by Per Bothner.
Copyright (C) 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001,
2002, 2003, 2004, 2005, 2006, 2007, 2008 Free Software Foundation, Inc.
Copyright 1992-2013 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE."
@@ -139,12 +132,33 @@ UNAME_RELEASE=`(uname -r) 2>/dev/null` || UNAME_RELEASE=unknown
UNAME_SYSTEM=`(uname -s) 2>/dev/null` || UNAME_SYSTEM=unknown
UNAME_VERSION=`(uname -v) 2>/dev/null` || UNAME_VERSION=unknown
case "${UNAME_SYSTEM}" in
Linux|GNU|GNU/*)
# If the system lacks a compiler, then just pick glibc.
# We could probably try harder.
LIBC=gnu
eval $set_cc_for_build
cat <<-EOF > $dummy.c
#include <features.h>
#if defined(__UCLIBC__)
LIBC=uclibc
#elif defined(__dietlibc__)
LIBC=dietlibc
#else
LIBC=gnu
#endif
EOF
eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^LIBC'`
;;
esac
# Note: order is significant - the case branches are not exclusive.
case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
*:NetBSD:*:*)
# NetBSD (nbsd) targets should (where applicable) match one or
# more of the tupples: *-*-netbsdelf*, *-*-netbsdaout*,
# more of the tuples: *-*-netbsdelf*, *-*-netbsdaout*,
# *-*-netbsdecoff* and *-*-netbsd*. For targets that recently
# switched to ELF, *-*-netbsd* would select the old
# object file format. This provides both forward
@@ -170,7 +184,7 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
arm*|i386|m68k|ns32k|sh3*|sparc|vax)
eval $set_cc_for_build
if echo __ELF__ | $CC_FOR_BUILD -E - 2>/dev/null \
| grep __ELF__ >/dev/null
| grep -q __ELF__
then
# Once all utilities can be ECOFF (netbsdecoff) or a.out (netbsdaout).
# Return netbsd for either. FIX?
@@ -180,7 +194,7 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
fi
;;
*)
os=netbsd
os=netbsd
;;
esac
# The OS release
@@ -201,6 +215,10 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
# CPU_TYPE-MANUFACTURER-OPERATING_SYSTEM is used.
echo "${machine}-${os}${release}"
exit ;;
*:Bitrig:*:*)
UNAME_MACHINE_ARCH=`arch | sed 's/Bitrig.//'`
echo ${UNAME_MACHINE_ARCH}-unknown-bitrig${UNAME_RELEASE}
exit ;;
*:OpenBSD:*:*)
UNAME_MACHINE_ARCH=`arch | sed 's/OpenBSD.//'`
echo ${UNAME_MACHINE_ARCH}-unknown-openbsd${UNAME_RELEASE}
@@ -223,7 +241,7 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $3}'`
;;
*5.*)
UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'`
UNAME_RELEASE=`/usr/sbin/sizer -v | awk '{print $4}'`
;;
esac
# According to Compaq, /usr/sbin/psrinfo has been available on
@@ -269,7 +287,10 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
# A Xn.n version is an unreleased experimental baselevel.
# 1.2 uses "1.2" for uname -r.
echo ${UNAME_MACHINE}-dec-osf`echo ${UNAME_RELEASE} | sed -e 's/^[PVTX]//' | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'`
exit ;;
# Reset EXIT trap before exiting to avoid spurious non-zero exit code.
exitcode=$?
trap '' 0
exit $exitcode ;;
Alpha\ *:Windows_NT*:*)
# How do we know it's Interix rather than the generic POSIX subsystem?
# Should we change UNAME_MACHINE based on the output of uname instead
@@ -295,12 +316,12 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
echo s390-ibm-zvmoe
exit ;;
*:OS400:*:*)
echo powerpc-ibm-os400
echo powerpc-ibm-os400
exit ;;
arm:RISC*:1.[012]*:*|arm:riscix:1.[012]*:*)
echo arm-acorn-riscix${UNAME_RELEASE}
exit ;;
arm:riscos:*:*|arm:RISCOS:*:*)
arm*:riscos:*:*|arm*:RISCOS:*:*)
echo arm-unknown-riscos
exit ;;
SR2?01:HI-UX/MPP:*:* | SR8000:HI-UX/MPP:*:*)
@@ -324,14 +345,33 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
case `/usr/bin/uname -p` in
sparc) echo sparc-icl-nx7; exit ;;
esac ;;
s390x:SunOS:*:*)
echo ${UNAME_MACHINE}-ibm-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
exit ;;
sun4H:SunOS:5.*:*)
echo sparc-hal-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
exit ;;
sun4*:SunOS:5.*:* | tadpole*:SunOS:5.*:*)
echo sparc-sun-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
exit ;;
i86pc:AuroraUX:5.*:* | i86xen:AuroraUX:5.*:*)
echo i386-pc-auroraux${UNAME_RELEASE}
exit ;;
i86pc:SunOS:5.*:* | i86xen:SunOS:5.*:*)
echo i386-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
eval $set_cc_for_build
SUN_ARCH="i386"
# If there is a compiler, see if it is configured for 64-bit objects.
# Note that the Sun cc does not turn __LP64__ into 1 like gcc does.
# This test works for both compilers.
if [ "$CC_FOR_BUILD" != 'no_compiler_found' ]; then
if (echo '#ifdef __amd64'; echo IS_64BIT_ARCH; echo '#endif') | \
(CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \
grep IS_64BIT_ARCH >/dev/null
then
SUN_ARCH="x86_64"
fi
fi
echo ${SUN_ARCH}-pc-solaris2`echo ${UNAME_RELEASE}|sed -e 's/[^.]*//'`
exit ;;
sun4*:SunOS:6*:*)
# According to config.sub, this is the proper way to canonicalize
@@ -375,23 +415,23 @@ case "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" in
# MiNT. But MiNT is downward compatible to TOS, so this should
# be no problem.
atarist[e]:*MiNT:*:* | atarist[e]:*mint:*:* | atarist[e]:*TOS:*:*)
echo m68k-atari-mint${UNAME_RELEASE}
echo m68k-atari-mint${UNAME_RELEASE}
exit ;;
atari*:*MiNT:*:* | atari*:*mint:*:* | atarist[e]:*TOS:*:*)
echo m68k-atari-mint${UNAME_RELEASE}
exit ;;
exit ;;
*falcon*:*MiNT:*:* | *falcon*:*mint:*:* | *falcon*:*TOS:*:*)
echo m68k-atari-mint${UNAME_RELEASE}
echo m68k-atari-mint${UNAME_RELEASE}
exit ;;
milan*:*MiNT:*:* | milan*:*mint:*:* | *milan*:*TOS:*:*)
echo m68k-milan-mint${UNAME_RELEASE}
exit ;;
echo m68k-milan-mint${UNAME_RELEASE}
exit ;;
hades*:*MiNT:*:* | hades*:*mint:*:* | *hades*:*TOS:*:*)
echo m68k-hades-mint${UNAME_RELEASE}
exit ;;
echo m68k-hades-mint${UNAME_RELEASE}
exit ;;
*:*MiNT:*:* | *:*mint:*:* | *:*TOS:*:*)
echo m68k-unknown-mint${UNAME_RELEASE}
exit ;;
echo m68k-unknown-mint${UNAME_RELEASE}
exit ;;
m68k:machten:*:*)
echo m68k-apple-machten${UNAME_RELEASE}
exit ;;
@@ -461,8 +501,8 @@ EOF
echo m88k-motorola-sysv3
exit ;;
AViiON:dgux:*:*)
# DG/UX returns AViiON for all architectures
UNAME_PROCESSOR=`/usr/bin/uname -p`
# DG/UX returns AViiON for all architectures
UNAME_PROCESSOR=`/usr/bin/uname -p`
if [ $UNAME_PROCESSOR = mc88100 ] || [ $UNAME_PROCESSOR = mc88110 ]
then
if [ ${TARGET_BINARY_INTERFACE}x = m88kdguxelfx ] || \
@@ -475,7 +515,7 @@ EOF
else
echo i586-dg-dgux${UNAME_RELEASE}
fi
exit ;;
exit ;;
M88*:DolphinOS:*:*) # DolphinOS (SVR3)
echo m88k-dolphin-sysv3
exit ;;
@@ -532,7 +572,7 @@ EOF
echo rs6000-ibm-aix3.2
fi
exit ;;
*:AIX:*:[456])
*:AIX:*:[4567])
IBM_CPU_ID=`/usr/sbin/lsdev -C -c processor -S available | sed 1q | awk '{ print $1 }'`
if /usr/sbin/lsattr -El ${IBM_CPU_ID} | grep ' POWER' >/dev/null 2>&1; then
IBM_ARCH=rs6000
@@ -575,52 +615,52 @@ EOF
9000/[678][0-9][0-9])
if [ -x /usr/bin/getconf ]; then
sc_cpu_version=`/usr/bin/getconf SC_CPU_VERSION 2>/dev/null`
sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null`
case "${sc_cpu_version}" in
523) HP_ARCH="hppa1.0" ;; # CPU_PA_RISC1_0
528) HP_ARCH="hppa1.1" ;; # CPU_PA_RISC1_1
532) # CPU_PA_RISC2_0
case "${sc_kernel_bits}" in
32) HP_ARCH="hppa2.0n" ;;
64) HP_ARCH="hppa2.0w" ;;
sc_kernel_bits=`/usr/bin/getconf SC_KERNEL_BITS 2>/dev/null`
case "${sc_cpu_version}" in
523) HP_ARCH="hppa1.0" ;; # CPU_PA_RISC1_0
528) HP_ARCH="hppa1.1" ;; # CPU_PA_RISC1_1
532) # CPU_PA_RISC2_0
case "${sc_kernel_bits}" in
32) HP_ARCH="hppa2.0n" ;;
64) HP_ARCH="hppa2.0w" ;;
'') HP_ARCH="hppa2.0" ;; # HP-UX 10.20
esac ;;
esac
esac ;;
esac
fi
if [ "${HP_ARCH}" = "" ]; then
eval $set_cc_for_build
sed 's/^ //' << EOF >$dummy.c
sed 's/^ //' << EOF >$dummy.c
#define _HPUX_SOURCE
#include <stdlib.h>
#include <unistd.h>
#define _HPUX_SOURCE
#include <stdlib.h>
#include <unistd.h>
int main ()
{
#if defined(_SC_KERNEL_BITS)
long bits = sysconf(_SC_KERNEL_BITS);
#endif
long cpu = sysconf (_SC_CPU_VERSION);
int main ()
{
#if defined(_SC_KERNEL_BITS)
long bits = sysconf(_SC_KERNEL_BITS);
#endif
long cpu = sysconf (_SC_CPU_VERSION);
switch (cpu)
{
case CPU_PA_RISC1_0: puts ("hppa1.0"); break;
case CPU_PA_RISC1_1: puts ("hppa1.1"); break;
case CPU_PA_RISC2_0:
#if defined(_SC_KERNEL_BITS)
switch (bits)
{
case 64: puts ("hppa2.0w"); break;
case 32: puts ("hppa2.0n"); break;
default: puts ("hppa2.0"); break;
} break;
#else /* !defined(_SC_KERNEL_BITS) */
puts ("hppa2.0"); break;
#endif
default: puts ("hppa1.0"); break;
}
exit (0);
}
switch (cpu)
{
case CPU_PA_RISC1_0: puts ("hppa1.0"); break;
case CPU_PA_RISC1_1: puts ("hppa1.1"); break;
case CPU_PA_RISC2_0:
#if defined(_SC_KERNEL_BITS)
switch (bits)
{
case 64: puts ("hppa2.0w"); break;
case 32: puts ("hppa2.0n"); break;
default: puts ("hppa2.0"); break;
} break;
#else /* !defined(_SC_KERNEL_BITS) */
puts ("hppa2.0"); break;
#endif
default: puts ("hppa1.0"); break;
}
exit (0);
}
EOF
(CCOPTS= $CC_FOR_BUILD -o $dummy $dummy.c 2>/dev/null) && HP_ARCH=`$dummy`
test -z "$HP_ARCH" && HP_ARCH=hppa
@@ -640,7 +680,7 @@ EOF
# => hppa64-hp-hpux11.23
if echo __LP64__ | (CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) |
grep __LP64__ >/dev/null
grep -q __LP64__
then
HP_ARCH="hppa2.0w"
else
@@ -711,22 +751,22 @@ EOF
exit ;;
C1*:ConvexOS:*:* | convex:ConvexOS:C1*:*)
echo c1-convex-bsd
exit ;;
exit ;;
C2*:ConvexOS:*:* | convex:ConvexOS:C2*:*)
if getsysinfo -f scalar_acc
then echo c32-convex-bsd
else echo c2-convex-bsd
fi
exit ;;
exit ;;
C34*:ConvexOS:*:* | convex:ConvexOS:C34*:*)
echo c34-convex-bsd
exit ;;
exit ;;
C38*:ConvexOS:*:* | convex:ConvexOS:C38*:*)
echo c38-convex-bsd
exit ;;
exit ;;
C4*:ConvexOS:*:* | convex:ConvexOS:C4*:*)
echo c4-convex-bsd
exit ;;
exit ;;
CRAY*Y-MP:*:*:*)
echo ymp-cray-unicos${UNAME_RELEASE} | sed -e 's/\.[^.]*$/.X/'
exit ;;
@@ -750,14 +790,14 @@ EOF
exit ;;
F30[01]:UNIX_System_V:*:* | F700:UNIX_System_V:*:*)
FUJITSU_PROC=`uname -m | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz'`
FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'`
FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'`
echo "${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
exit ;;
FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'`
FUJITSU_REL=`echo ${UNAME_RELEASE} | sed -e 's/ /_/'`
echo "${FUJITSU_PROC}-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
exit ;;
5000:UNIX_System_V:4.*:*)
FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'`
FUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'`
echo "sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
FUJITSU_SYS=`uname -p | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/\///'`
FUJITSU_REL=`echo ${UNAME_RELEASE} | tr 'ABCDEFGHIJKLMNOPQRSTUVWXYZ' 'abcdefghijklmnopqrstuvwxyz' | sed -e 's/ /_/'`
echo "sparc-fujitsu-${FUJITSU_SYS}${FUJITSU_REL}"
exit ;;
i*86:BSD/386:*:* | i*86:BSD/OS:*:* | *:Ascend\ Embedded/OS:*:*)
echo ${UNAME_MACHINE}-pc-bsdi${UNAME_RELEASE}
@@ -769,34 +809,39 @@ EOF
echo ${UNAME_MACHINE}-unknown-bsdi${UNAME_RELEASE}
exit ;;
*:FreeBSD:*:*)
case ${UNAME_MACHINE} in
pc98)
echo i386-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;;
UNAME_PROCESSOR=`/usr/bin/uname -p`
case ${UNAME_PROCESSOR} in
amd64)
echo x86_64-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;;
*)
echo ${UNAME_MACHINE}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;;
echo ${UNAME_PROCESSOR}-unknown-freebsd`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'` ;;
esac
exit ;;
i*:CYGWIN*:*)
echo ${UNAME_MACHINE}-pc-cygwin
exit ;;
*:MINGW64*:*)
echo ${UNAME_MACHINE}-pc-mingw64
exit ;;
*:MINGW*:*)
echo ${UNAME_MACHINE}-pc-mingw32
exit ;;
i*:MSYS*:*)
echo ${UNAME_MACHINE}-pc-msys
exit ;;
i*:windows32*:*)
# uname -m includes "-pc" on this system.
echo ${UNAME_MACHINE}-mingw32
# uname -m includes "-pc" on this system.
echo ${UNAME_MACHINE}-mingw32
exit ;;
i*:PW*:*)
echo ${UNAME_MACHINE}-pc-pw32
exit ;;
*:Interix*:[3456]*)
case ${UNAME_MACHINE} in
*:Interix*:*)
case ${UNAME_MACHINE} in
x86)
echo i586-pc-interix${UNAME_RELEASE}
exit ;;
EM64T | authenticamd)
authenticamd | genuineintel | EM64T)
echo x86_64-unknown-interix${UNAME_RELEASE}
exit ;;
IA64)
@@ -806,6 +851,9 @@ EOF
[345]86:Windows_95:* | [345]86:Windows_98:* | [345]86:Windows_NT:*)
echo i${UNAME_MACHINE}-pc-mks
exit ;;
8664:Windows_NT:*)
echo x86_64-pc-mks
exit ;;
i*:Windows_NT*:* | Pentium*:Windows_NT*:*)
# How do we know it's Interix rather than the generic POSIX subsystem?
# It also conflicts with pre-2.0 versions of AT&T UWIN. Should we
@@ -826,100 +874,21 @@ EOF
exit ;;
*:GNU:*:*)
# the GNU system
echo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-gnu`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'`
echo `echo ${UNAME_MACHINE}|sed -e 's,[-/].*$,,'`-unknown-${LIBC}`echo ${UNAME_RELEASE}|sed -e 's,/.*$,,'`
exit ;;
*:GNU/*:*:*)
# other systems with GNU libc and userland
echo ${UNAME_MACHINE}-unknown-`echo ${UNAME_SYSTEM} | sed 's,^[^/]*/,,' | tr '[A-Z]' '[a-z]'``echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`-gnu
echo ${UNAME_MACHINE}-unknown-`echo ${UNAME_SYSTEM} | sed 's,^[^/]*/,,' | tr '[A-Z]' '[a-z]'``echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`-${LIBC}
exit ;;
i*86:Minix:*:*)
echo ${UNAME_MACHINE}-pc-minix
exit ;;
arm*:Linux:*:*)
eval $set_cc_for_build
if echo __ARM_EABI__ | $CC_FOR_BUILD -E - 2>/dev/null \
| grep -q __ARM_EABI__
then
echo ${UNAME_MACHINE}-unknown-linux-gnu
else
echo ${UNAME_MACHINE}-unknown-linux-gnueabi
fi
aarch64:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
avr32*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
exit ;;
cris:Linux:*:*)
echo cris-axis-linux-gnu
exit ;;
crisv32:Linux:*:*)
echo crisv32-axis-linux-gnu
exit ;;
frv:Linux:*:*)
echo frv-unknown-linux-gnu
exit ;;
ia64:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
exit ;;
m32r*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
exit ;;
m68*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
exit ;;
mips:Linux:*:*)
eval $set_cc_for_build
sed 's/^ //' << EOF >$dummy.c
#undef CPU
#undef mips
#undef mipsel
#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL)
CPU=mipsel
#else
#if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB)
CPU=mips
#else
CPU=
#endif
#endif
EOF
eval "`$CC_FOR_BUILD -E $dummy.c 2>/dev/null | sed -n '
/^CPU/{
s: ::g
p
}'`"
test x"${CPU}" != x && { echo "${CPU}-unknown-linux-gnu"; exit; }
;;
mips64:Linux:*:*)
eval $set_cc_for_build
sed 's/^ //' << EOF >$dummy.c
#undef CPU
#undef mips64
#undef mips64el
#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL)
CPU=mips64el
#else
#if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB)
CPU=mips64
#else
CPU=
#endif
#endif
EOF
eval "`$CC_FOR_BUILD -E $dummy.c 2>/dev/null | sed -n '
/^CPU/{
s: ::g
p
}'`"
test x"${CPU}" != x && { echo "${CPU}-unknown-linux-gnu"; exit; }
;;
or32:Linux:*:*)
echo or32-unknown-linux-gnu
exit ;;
ppc:Linux:*:*)
echo powerpc-unknown-linux-gnu
exit ;;
ppc64:Linux:*:*)
echo powerpc64-unknown-linux-gnu
aarch64_be:Linux:*:*)
UNAME_MACHINE=aarch64_be
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
alpha:Linux:*:*)
case `sed -n '/^cpu model/s/^.*: \(.*\)/\1/p' < /proc/cpuinfo` in
@@ -930,106 +899,132 @@ EOF
EV6) UNAME_MACHINE=alphaev6 ;;
EV67) UNAME_MACHINE=alphaev67 ;;
EV68*) UNAME_MACHINE=alphaev68 ;;
esac
objdump --private-headers /bin/sh | grep ld.so.1 >/dev/null
if test "$?" = 0 ; then LIBC="libc1" ; else LIBC="" ; fi
echo ${UNAME_MACHINE}-unknown-linux-gnu${LIBC}
esac
objdump --private-headers /bin/sh | grep -q ld.so.1
if test "$?" = 0 ; then LIBC="gnulibc1" ; fi
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
arc:Linux:*:* | arceb:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
arm*:Linux:*:*)
eval $set_cc_for_build
if echo __ARM_EABI__ | $CC_FOR_BUILD -E - 2>/dev/null \
| grep -q __ARM_EABI__
then
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
else
if echo __ARM_PCS_VFP | $CC_FOR_BUILD -E - 2>/dev/null \
| grep -q __ARM_PCS_VFP
then
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}eabi
else
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}eabihf
fi
fi
exit ;;
avr32*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
cris:Linux:*:*)
echo ${UNAME_MACHINE}-axis-linux-${LIBC}
exit ;;
crisv32:Linux:*:*)
echo ${UNAME_MACHINE}-axis-linux-${LIBC}
exit ;;
frv:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
hexagon:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
i*86:Linux:*:*)
echo ${UNAME_MACHINE}-pc-linux-${LIBC}
exit ;;
ia64:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
m32r*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
m68*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
mips:Linux:*:* | mips64:Linux:*:*)
eval $set_cc_for_build
sed 's/^ //' << EOF >$dummy.c
#undef CPU
#undef ${UNAME_MACHINE}
#undef ${UNAME_MACHINE}el
#if defined(__MIPSEL__) || defined(__MIPSEL) || defined(_MIPSEL) || defined(MIPSEL)
CPU=${UNAME_MACHINE}el
#else
#if defined(__MIPSEB__) || defined(__MIPSEB) || defined(_MIPSEB) || defined(MIPSEB)
CPU=${UNAME_MACHINE}
#else
CPU=
#endif
#endif
EOF
eval `$CC_FOR_BUILD -E $dummy.c 2>/dev/null | grep '^CPU'`
test x"${CPU}" != x && { echo "${CPU}-unknown-linux-${LIBC}"; exit; }
;;
or1k:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
or32:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
padre:Linux:*:*)
echo sparc-unknown-linux-${LIBC}
exit ;;
parisc64:Linux:*:* | hppa64:Linux:*:*)
echo hppa64-unknown-linux-${LIBC}
exit ;;
parisc:Linux:*:* | hppa:Linux:*:*)
# Look for CPU level
case `grep '^cpu[^a-z]*:' /proc/cpuinfo 2>/dev/null | cut -d' ' -f2` in
PA7*) echo hppa1.1-unknown-linux-gnu ;;
PA8*) echo hppa2.0-unknown-linux-gnu ;;
*) echo hppa-unknown-linux-gnu ;;
PA7*) echo hppa1.1-unknown-linux-${LIBC} ;;
PA8*) echo hppa2.0-unknown-linux-${LIBC} ;;
*) echo hppa-unknown-linux-${LIBC} ;;
esac
exit ;;
parisc64:Linux:*:* | hppa64:Linux:*:*)
echo hppa64-unknown-linux-gnu
ppc64:Linux:*:*)
echo powerpc64-unknown-linux-${LIBC}
exit ;;
ppc:Linux:*:*)
echo powerpc-unknown-linux-${LIBC}
exit ;;
ppc64le:Linux:*:*)
echo powerpc64le-unknown-linux-${LIBC}
exit ;;
ppcle:Linux:*:*)
echo powerpcle-unknown-linux-${LIBC}
exit ;;
s390:Linux:*:* | s390x:Linux:*:*)
echo ${UNAME_MACHINE}-ibm-linux
echo ${UNAME_MACHINE}-ibm-linux-${LIBC}
exit ;;
sh64*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
sh*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
sparc:Linux:*:* | sparc64:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
tile*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
vax:Linux:*:*)
echo ${UNAME_MACHINE}-dec-linux-gnu
echo ${UNAME_MACHINE}-dec-linux-${LIBC}
exit ;;
x86_64:Linux:*:*)
echo x86_64-unknown-linux-gnu
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
xtensa*:Linux:*:*)
echo ${UNAME_MACHINE}-unknown-linux-gnu
echo ${UNAME_MACHINE}-unknown-linux-${LIBC}
exit ;;
i*86:Linux:*:*)
# The BFD linker knows what the default object file format is, so
# first see if it will tell us. cd to the root directory to prevent
# problems with other programs or directories called `ld' in the path.
# Set LC_ALL=C to ensure ld outputs messages in English.
ld_supported_targets=`cd /; LC_ALL=C ld --help 2>&1 \
| sed -ne '/supported targets:/!d
s/[ ][ ]*/ /g
s/.*supported targets: *//
s/ .*//
p'`
case "$ld_supported_targets" in
elf32-i386)
TENTATIVE="${UNAME_MACHINE}-pc-linux-gnu"
;;
a.out-i386-linux)
echo "${UNAME_MACHINE}-pc-linux-gnuaout"
exit ;;
coff-i386)
echo "${UNAME_MACHINE}-pc-linux-gnucoff"
exit ;;
"")
# Either a pre-BFD a.out linker (linux-gnuoldld) or
# one that does not give us useful --help.
echo "${UNAME_MACHINE}-pc-linux-gnuoldld"
exit ;;
esac
# Determine whether the default compiler is a.out or elf
eval $set_cc_for_build
sed 's/^ //' << EOF >$dummy.c
#include <features.h>
#ifdef __ELF__
# ifdef __GLIBC__
# if __GLIBC__ >= 2
LIBC=gnu
# else
LIBC=gnulibc1
# endif
# else
LIBC=gnulibc1
# endif
#else
#if defined(__INTEL_COMPILER) || defined(__PGI) || defined(__SUNPRO_C) || defined(__SUNPRO_CC)
LIBC=gnu
#else
LIBC=gnuaout
#endif
#endif
#ifdef __dietlibc__
LIBC=dietlibc
#endif
EOF
eval "`$CC_FOR_BUILD -E $dummy.c 2>/dev/null | sed -n '
/^LIBC/{
s: ::g
p
}'`"
test x"${LIBC}" != x && {
echo "${UNAME_MACHINE}-pc-linux-${LIBC}"
exit
}
test x"${TENTATIVE}" != x && { echo "${TENTATIVE}"; exit; }
;;
i*86:DYNIX/ptx:4*:*)
# ptx 4.0 does uname -s correctly, with DYNIX/ptx in there.
# earlier versions are messed up and put the nodename in both
@@ -1037,11 +1032,11 @@ EOF
echo i386-sequent-sysv4
exit ;;
i*86:UNIX_SV:4.2MP:2.*)
# Unixware is an offshoot of SVR4, but it has its own version
# number series starting with 2...
# I am not positive that other SVR4 systems won't match this,
# Unixware is an offshoot of SVR4, but it has its own version
# number series starting with 2...
# I am not positive that other SVR4 systems won't match this,
# I just have to hope. -- rms.
# Use sysv4.2uw... so that sysv4* matches it.
# Use sysv4.2uw... so that sysv4* matches it.
echo ${UNAME_MACHINE}-pc-sysv4.2uw${UNAME_VERSION}
exit ;;
i*86:OS/2:*:*)
@@ -1058,7 +1053,7 @@ EOF
i*86:syllable:*:*)
echo ${UNAME_MACHINE}-pc-syllable
exit ;;
i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.0*:*)
i*86:LynxOS:2.*:* | i*86:LynxOS:3.[01]*:* | i*86:LynxOS:4.[02]*:*)
echo i386-unknown-lynxos${UNAME_RELEASE}
exit ;;
i*86:*DOS:*:*)
@@ -1073,7 +1068,7 @@ EOF
fi
exit ;;
i*86:*:5:[678]*)
# UnixWare 7.x, OpenUNIX and OpenServer 6.
# UnixWare 7.x, OpenUNIX and OpenServer 6.
case `/bin/uname -X | grep "^Machine"` in
*486*) UNAME_MACHINE=i486 ;;
*Pentium) UNAME_MACHINE=i586 ;;
@@ -1101,10 +1096,13 @@ EOF
exit ;;
pc:*:*:*)
# Left here for compatibility:
# uname -m prints for DJGPP always 'pc', but it prints nothing about
# the processor, so we play safe by assuming i386.
echo i386-pc-msdosdjgpp
exit ;;
# uname -m prints for DJGPP always 'pc', but it prints nothing about
# the processor, so we play safe by assuming i586.
# Note: whatever this is, it MUST be the same as what config.sub
# prints for the "djgpp" host, or else GDB configury will decide that
# this is a cross-build.
echo i586-pc-msdosdjgpp
exit ;;
Intel:Mach:3*:*)
echo i386-pc-mach3
exit ;;
@@ -1139,8 +1137,18 @@ EOF
/bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \
&& { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;;
3[34]??:*:4.0:* | 3[34]??,*:*:4.0:*)
/bin/uname -p 2>/dev/null | grep 86 >/dev/null \
&& { echo i486-ncr-sysv4; exit; } ;;
/bin/uname -p 2>/dev/null | grep 86 >/dev/null \
&& { echo i486-ncr-sysv4; exit; } ;;
NCR*:*:4.2:* | MPRAS*:*:4.2:*)
OS_REL='.3'
test -r /etc/.relid \
&& OS_REL=.`sed -n 's/[^ ]* [^ ]* \([0-9][0-9]\).*/\1/p' < /etc/.relid`
/bin/uname -p 2>/dev/null | grep 86 >/dev/null \
&& { echo i486-ncr-sysv4.3${OS_REL}; exit; }
/bin/uname -p 2>/dev/null | /bin/grep entium >/dev/null \
&& { echo i586-ncr-sysv4.3${OS_REL}; exit; }
/bin/uname -p 2>/dev/null | /bin/grep pteron >/dev/null \
&& { echo i586-ncr-sysv4.3${OS_REL}; exit; } ;;
m68*:LynxOS:2.*:* | m68*:LynxOS:3.0*:*)
echo m68k-unknown-lynxos${UNAME_RELEASE}
exit ;;
@@ -1153,7 +1161,7 @@ EOF
rs6000:LynxOS:2.*:*)
echo rs6000-unknown-lynxos${UNAME_RELEASE}
exit ;;
PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.0*:*)
PowerPC:LynxOS:2.*:* | PowerPC:LynxOS:3.[01]*:* | PowerPC:LynxOS:4.[02]*:*)
echo powerpc-unknown-lynxos${UNAME_RELEASE}
exit ;;
SM[BE]S:UNIX_SV:*:*)
@@ -1173,10 +1181,10 @@ EOF
echo ns32k-sni-sysv
fi
exit ;;
PENTIUM:*:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort
# says <Richard.M.Bartel@ccMail.Census.GOV>
echo i586-unisys-sysv4
exit ;;
PENTIUM:*:4.0*:*) # Unisys `ClearPath HMP IX 4000' SVR4/MP effort
# says <Richard.M.Bartel@ccMail.Census.GOV>
echo i586-unisys-sysv4
exit ;;
*:UNIX_System_V:4*:FTX*)
# From Gerald Hewes <hewes@openmarket.com>.
# How about differentiating between stratus architectures? -djm
@@ -1202,11 +1210,11 @@ EOF
exit ;;
R[34]000:*System_V*:*:* | R4000:UNIX_SYSV:*:* | R*000:UNIX_SV:*:*)
if [ -d /usr/nec ]; then
echo mips-nec-sysv${UNAME_RELEASE}
echo mips-nec-sysv${UNAME_RELEASE}
else
echo mips-unknown-sysv${UNAME_RELEASE}
echo mips-unknown-sysv${UNAME_RELEASE}
fi
exit ;;
exit ;;
BeBox:BeOS:*:*) # BeOS running on hardware made by Be, PPC only.
echo powerpc-be-beos
exit ;;
@@ -1216,6 +1224,12 @@ EOF
BePC:BeOS:*:*) # BeOS running on Intel PC compatible.
echo i586-pc-beos
exit ;;
BePC:Haiku:*:*) # Haiku running on Intel PC compatible.
echo i586-pc-haiku
exit ;;
x86_64:Haiku:*:*)
echo x86_64-unknown-haiku
exit ;;
SX-4:SUPER-UX:*:*)
echo sx4-nec-superux${UNAME_RELEASE}
exit ;;
@@ -1242,9 +1256,21 @@ EOF
exit ;;
*:Darwin:*:*)
UNAME_PROCESSOR=`uname -p` || UNAME_PROCESSOR=unknown
case $UNAME_PROCESSOR in
unknown) UNAME_PROCESSOR=powerpc ;;
esac
eval $set_cc_for_build
if test "$UNAME_PROCESSOR" = unknown ; then
UNAME_PROCESSOR=powerpc
fi
if [ "$CC_FOR_BUILD" != 'no_compiler_found' ]; then
if (echo '#ifdef __LP64__'; echo IS_64BIT_ARCH; echo '#endif') | \
(CCOPTS= $CC_FOR_BUILD -E - 2>/dev/null) | \
grep IS_64BIT_ARCH >/dev/null
then
case $UNAME_PROCESSOR in
i386) UNAME_PROCESSOR=x86_64 ;;
powerpc) UNAME_PROCESSOR=powerpc64 ;;
esac
fi
fi
echo ${UNAME_PROCESSOR}-apple-darwin${UNAME_RELEASE}
exit ;;
*:procnto*:*:* | *:QNX:[0123456789]*:*)
@@ -1258,7 +1284,10 @@ EOF
*:QNX:*:4*)
echo i386-pc-qnx
exit ;;
NSE-?:NONSTOP_KERNEL:*:*)
NEO-?:NONSTOP_KERNEL:*:*)
echo neo-tandem-nsk${UNAME_RELEASE}
exit ;;
NSE-*:NONSTOP_KERNEL:*:*)
echo nse-tandem-nsk${UNAME_RELEASE}
exit ;;
NSR-?:NONSTOP_KERNEL:*:*)
@@ -1303,13 +1332,13 @@ EOF
echo pdp10-unknown-its
exit ;;
SEI:*:*:SEIUX)
echo mips-sei-seiux${UNAME_RELEASE}
echo mips-sei-seiux${UNAME_RELEASE}
exit ;;
*:DragonFly:*:*)
echo ${UNAME_MACHINE}-unknown-dragonfly`echo ${UNAME_RELEASE}|sed -e 's/[-(].*//'`
exit ;;
*:*VMS:*:*)
UNAME_MACHINE=`(uname -p) 2>/dev/null`
UNAME_MACHINE=`(uname -p) 2>/dev/null`
case "${UNAME_MACHINE}" in
A*) echo alpha-dec-vms ; exit ;;
I*) echo ia64-dec-vms ; exit ;;
@@ -1324,11 +1353,14 @@ EOF
i*86:rdos:*:*)
echo ${UNAME_MACHINE}-pc-rdos
exit ;;
i*86:AROS:*:*)
echo ${UNAME_MACHINE}-pc-aros
exit ;;
x86_64:VMkernel:*:*)
echo ${UNAME_MACHINE}-unknown-esx
exit ;;
esac
#echo '(No uname command or uname output not recognized.)' 1>&2
#echo "${UNAME_MACHINE}:${UNAME_SYSTEM}:${UNAME_RELEASE}:${UNAME_VERSION}" 1>&2
eval $set_cc_for_build
cat >$dummy.c <<EOF
#ifdef _SEQUENT_
@@ -1346,11 +1378,11 @@ main ()
#include <sys/param.h>
printf ("m68k-sony-newsos%s\n",
#ifdef NEWSOS4
"4"
"4"
#else
""
""
#endif
); exit (0);
); exit (0);
#endif
#endif
+1 -1
View File
@@ -1,4 +1,4 @@
AC_INIT([ccminer], [1.6.6-dev])
AC_INIT([ccminer], [1.7-dev])
AC_PREREQ([2.59c])
AC_CANONICAL_SYSTEM
+3 -3
View File
@@ -162,7 +162,7 @@
#define PACKAGE_NAME "ccminer"
/* Define to the full name and version of this package. */
#define PACKAGE_STRING "ccminer 1.6.6-dev"
#define PACKAGE_STRING "ccminer 1.7-dev"
/* Define to the one symbol short name of this package. */
#define PACKAGE_TARNAME "ccminer"
@@ -171,7 +171,7 @@
#define PACKAGE_URL "http://github.com/tpruvot/ccminer"
/* Define to the version of this package. */
#define PACKAGE_VERSION "1.6.6-dev"
#define PACKAGE_VERSION "1.7-dev"
/* If using the C implementation of alloca, define if you know the
direction of stack growth for your system; otherwise it will be
@@ -185,7 +185,7 @@
#define STDC_HEADERS 1
/* Version number of package */
#define VERSION "1.6.6-dev"
#define VERSION "1.7-dev"
/* Define curl_free() as free() if our version of curl lacks curl_free. */
/* #undef curl_free */
+2
View File
@@ -212,6 +212,8 @@ void cuda_reset_device(int thr_id, bool *init)
usleep(1000);
}
cudaDeviceReset();
if (opt_cudaschedule >= 0)
cudaSetDeviceFlags((unsigned)(opt_cudaschedule & cudaDeviceScheduleMask));
}
void cudaReportHardwareFailure(int thr_id, cudaError_t err, const char* func)
+44
View File
@@ -23,6 +23,13 @@ void cuda_check_cpu_init(int thr_id, uint32_t threads)
init_done = true;
}
__host__
void cuda_check_cpu_free(int thr_id)
{
cudaFree(d_resNonces[thr_id]);
cudaFreeHost(h_resNonces[thr_id]);
}
// Target Difficulty
__host__
void cuda_check_cpu_setTarget(const void *ptarget)
@@ -89,6 +96,21 @@ void cuda_checkhash_64(uint32_t threads, uint32_t startNounce, uint32_t *hash, u
}
}
__global__ __launch_bounds__(512, 4)
void cuda_checkhash_32(uint32_t threads, uint32_t startNounce, uint32_t *hash, uint32_t *resNonces)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t *inpHash = &hash[thread << 3];
if (resNonces[0] == UINT32_MAX) {
if (hashbelowtarget(inpHash, pTarget))
resNonces[0] = (startNounce + thread);
}
}
}
__host__
uint32_t cuda_check_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash)
{
@@ -111,6 +133,28 @@ uint32_t cuda_check_hash(int thr_id, uint32_t threads, uint32_t startNounce, uin
return h_resNonces[thr_id][0];
}
__host__
uint32_t cuda_check_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash)
{
cudaMemset(d_resNonces[thr_id], 0xff, sizeof(uint32_t));
const uint32_t threadsperblock = 512;
dim3 grid((threads + threadsperblock - 1) / threadsperblock);
dim3 block(threadsperblock);
if (!init_done) {
applog(LOG_ERR, "missing call to cuda_check_cpu_init");
return UINT32_MAX;
}
cuda_checkhash_32 <<<grid, block>>> (threads, startNounce, d_inputHash, d_resNonces[thr_id]);
cudaThreadSynchronize();
cudaMemcpy(h_resNonces[thr_id], d_resNonces[thr_id], sizeof(uint32_t), cudaMemcpyDeviceToHost);
return h_resNonces[thr_id][0];
}
/* --------------------------------------------------------------------------------------------- */
__global__ __launch_bounds__(512, 4)
+1
View File
@@ -4,5 +4,6 @@
void fugue256_cpu_hash(int thr_id, uint32_t threads, int startNounce, void *outputHashes, uint32_t *nounce);
void fugue256_cpu_setBlock(int thr_id, void *data, void *pTargetIn);
void fugue256_cpu_init(int thr_id, uint32_t threads);
void fugue256_cpu_free(int thr_id);
#endif
+6
View File
@@ -102,6 +102,12 @@ void groestlcoin_cpu_init(int thr_id, uint32_t threads)
cudaMalloc(&d_resultNonce[thr_id], sizeof(uint32_t));
}
__host__
void groestlcoin_cpu_free(int thr_id)
{
cudaFree(d_resultNonce[thr_id]);
}
__host__
void groestlcoin_cpu_setBlock(int thr_id, void *data, void *pTargetIn)
{
+1
View File
@@ -2,6 +2,7 @@
#define _CUDA_GROESTLCOIN_H
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);
+32
View File
@@ -31,6 +31,7 @@ extern int cuda_arch[MAX_GPUS];
extern int cuda_get_arch(int thr_id);
extern void cuda_reset_device(int thr_id, bool *init);
extern void cuda_check_cpu_init(int thr_id, uint32_t threads);
extern void cuda_check_cpu_free(int thr_id);
extern void cuda_check_cpu_setTarget(const void *ptarget);
extern uint32_t cuda_check_hash(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash);
extern uint32_t cuda_check_hash_suppl(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_inputHash, uint8_t numNonce);
@@ -606,6 +607,37 @@ uint2 SWAPUINT2(uint2 value)
return make_uint2(value.y, value.x);
}
/* Byte aligned Rotations (lyra2) */
#ifdef __CUDA_ARCH__
__device__ __inline__ uint2 ROL8(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x6543);
result.y = __byte_perm(a.y, a.x, 0x2107);
return result;
}
__device__ __inline__ uint2 ROR16(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x1076);
result.y = __byte_perm(a.y, a.x, 0x5432);
return result;
}
__device__ __inline__ uint2 ROR24(const uint2 a)
{
uint2 result;
result.x = __byte_perm(a.y, a.x, 0x2107);
result.y = __byte_perm(a.y, a.x, 0x6543);
return result;
}
#else
#define ROL8(u) ((u) << 8)
#define ROR16(u) ((u) >> 16)
#define ROR24(u) ((u) >> 24)
#endif
/* uint2 for bmw512 - to double check later */
__device__ __forceinline__
+7
View File
@@ -327,6 +327,13 @@ void myriadgroestl_cpu_init(int thr_id, uint32_t threads)
cudaMalloc(&d_outputHashes[thr_id], 16*sizeof(uint32_t)*threads);
}
__host__
void myriadgroestl_cpu_free(int thr_id)
{
cudaFree(d_resultNonce[thr_id]);
cudaFree(d_outputHashes[thr_id]);
}
__host__
void myriadgroestl_cpu_setBlock(int thr_id, void *data, void *pTargetIn)
{
+46 -11
View File
@@ -66,11 +66,13 @@ extern "C" void nist5hash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_nist5(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_nist5(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];
int res = 0;
uint32_t throughput = device_intensity(thr_id, __func__, 1 << 20); // 256*256*16
throughput = min(throughput, (max_nonce - first_nonce));
@@ -82,20 +84,26 @@ extern "C" int scanhash_nist5(int thr_id, uint32_t *pdata,
{
cudaSetDevice(device_map[thr_id]);
// Konstanten kopieren, Speicher belegen
// Constants copy/init (no device alloc in these algos)
quark_blake512_cpu_init(thr_id, throughput);
quark_groestl512_cpu_init(thr_id, throughput);
quark_jh512_cpu_init(thr_id, throughput);
quark_keccak512_cpu_init(thr_id, throughput);
quark_skein512_cpu_init(thr_id, throughput);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], 16 * sizeof(uint32_t) * throughput));
// char[64] work space for hashes results
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t)64 * throughput));
cuda_check_cpu_init(thr_id, throughput);
init[thr_id] = true;
}
uint32_t endiandata[20];
#ifdef USE_STREAMS
cudaStream_t stream[5];
for (int i = 0; i < 5; i++)
cudaStreamCreate(&stream[i]);
#endif
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
@@ -112,6 +120,8 @@ extern "C" int scanhash_nist5(int thr_id, uint32_t *pdata,
quark_keccak512_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[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -121,15 +131,19 @@ extern "C" int scanhash_nist5(int thr_id, uint32_t *pdata,
nist5hash(vhash64, endiandata);
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
nist5hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
pdata[19] = foundNonce;
return res;
goto out;
}
else {
applog(LOG_WARNING, "GPU #%d: result for nonce %08x does not validate on CPU!", device_map[thr_id], foundNonce);
@@ -140,6 +154,27 @@ extern "C" int scanhash_nist5(int thr_id, uint32_t *pdata,
} while (pdata[19] < max_nonce && !work_restart[thr_id].restart);
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
out:
#ifdef USE_STREAMS
for (int i = 0; i < 5; i++)
cudaStreamDestroy(stream[i]);
#endif
return res;
}
// ressources cleanup
extern "C" void free_nist5(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
-531
View File
@@ -1,531 +0,0 @@
/*
* Genoil's CUDA mining kernel for Ethereum
* based on Tim Hughes' opencl kernel.
* thanks to tpruvot,djm34,sp,cbuchner for things i took from ccminer.
*/
#define SHUFFLE_MIN_VER 9350
#include "miner.h"
#include "cuda_helper.h"
#include "ethash_cu_miner_kernel.h"
#include <stdio.h>
#include <unistd.h>
#include <memory.h>
#include <fcntl.h>
#include <sys/stat.h>
#include <sys/mman.h>
#include "vector_types.h"
#include "rotl64.cuh"
#include "keccak.cuh"
//#include "ethash_cu_miner_kernel_globals.h"
__constant__ uint32_t d_dag_size;
__constant__ uint32_t d_max_outputs;
#define ACCESSES 64
#define THREADS_PER_HASH (128 / 16) /* 8 */
#define FNV_PRIME 0x01000193
#define SWAP64(v) \
((ROTL64L(v, 8) & 0x000000FF000000FF) | \
(ROTL64L(v, 24) & 0x0000FF000000FF00) | \
(ROTL64H(v, 40) & 0x00FF000000FF0000) | \
(ROTL64H(v, 56) & 0xFF000000FF000000))
#undef SWAP64
#define SWAP64(v) cuda_swab64(v)
#define PACK64(result, lo, hi) asm("mov.b64 %0, {%1,%2};//pack64" : "=l"(result) : "r"(lo), "r"(hi));
#define UNPACK64(lo, hi, input) asm("mov.b64 {%0, %1}, %2;//unpack64" : "=r"(lo),"=r"(hi) : "l"(input));
#define copy(dst, src, count) for (uint32_t i = 0; i < count; i++) { (dst)[i] = (src)[i]; }
#define countof(x) (sizeof(x) / sizeof(x[0]))
#define fnv(x,y) ((x) * FNV_PRIME ^(y))
__device__ uint4 fnv4(uint4 a, uint4 b)
{
uint4 c;
c.x = a.x * FNV_PRIME ^ b.x;
c.y = a.y * FNV_PRIME ^ b.y;
c.z = a.z * FNV_PRIME ^ b.z;
c.w = a.w * FNV_PRIME ^ b.w;
return c;
}
__device__ uint32_t fnv_reduce(uint4 v)
{
return fnv(fnv(fnv(v.x, v.y), v.z), v.w);
}
__device__ void init_hash(hash32_t const* header, uint64_t nonce, hash64_t &init)
{
// sha3_512(header .. nonce)
uint64_t state[25] = { 0 };
//copy(state, header->uint64s, 4);
#pragma unroll
for (int i = 0; i < 4; i++) state[i] = header->uint64s[i];
state[4] = nonce;
state[5] = 0x0000000000000001;
//state[6] = 0;
//state[7] = 0;
state[8] = 0x8000000000000000;
//for (uint32_t i = 9; i < 25; i++) {
// state[i] = 0;
//}
keccak_f1600_block(state, 8);
//copy(init.uint64s, state, 8);
#pragma unroll
for (int i = 0; i < 8; i++) init.uint64s[i] = state[i];
}
__device__ uint32_t inner_loop(uint4 mix, uint32_t thread_id, uint32_t* share, hash128_t const* g_dag)
{
// share init0
if (thread_id == 0)
*share = mix.x;
uint32_t init0 = *share;
uint32_t a = 0;
do
{
bool update_share = thread_id == ((a >> 2) & (THREADS_PER_HASH-1));
//#pragma unroll 4
for (uint32_t i = 0; i < 4; i++)
{
if (update_share)
{
uint32_t m[4] = { mix.x, mix.y, mix.z, mix.w };
*share = fnv(init0 ^ (a + i), m[i]) % d_dag_size;
}
__threadfence_block();
#if __CUDA_ARCH__ >= 350
mix = fnv4(mix, __ldg(&g_dag[*share].uint4s[thread_id]));
#else
mix = fnv4(mix, g_dag[*share].uint4s[thread_id]);
#endif
}
} while ((a += 4) != ACCESSES);
return fnv_reduce(mix);
}
__device__ void final_hash(hash64_t const* init, hash32_t const* mix, hash32_t &hash)
{
uint64_t state[25] = { 0 };
// keccak_256(keccak_512(header..nonce) .. mix);
copy(state, init->uint64s, 8);
copy(state + 8, mix->uint64s, 4);
state[12] = 0x0000000000000001;
//#pragma unroll
//for (uint32_t i = 13; i < 16; i++)
//{
//state[i] = 0;
//}
state[16] = 0x8000000000000000;
//#pragma unroll
//for (uint32_t i = 17; i < 25; i++)
//{
//state[i] = 0;
//}
keccak_f1600_block(state,1);
// copy out
copy(hash.uint64s, state, 4);
}
typedef union
{
hash64_t init;
hash32_t mix;
} compute_hash_share;
#if __CUDA_ARCH__ >= SHUFFLE_MIN_VER
__device__ uint64_t compute_hash_shuffle(
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t nonce
)
{
// sha3_512(header .. nonce)
uint64_t state[25];
copy(state, g_header->uint64s, 4);
state[4] = nonce;
state[5] = 0x0000000000000001ULL;
for (uint32_t i = 6; i < 25; i++)
{
state[i] = 0;
}
state[8] = 0x8000000000000000ULL;
keccak_f1600_block(state, 8);
// Threads work together in this phase in groups of 8.
const int thread_id = threadIdx.x & (THREADS_PER_HASH - 1);
const int start_lane = threadIdx.x & ~(THREADS_PER_HASH - 1);
const int mix_idx = (thread_id & 3);
uint4 mix;
uint32_t shuffle[16];
//uint32_t * init = (uint32_t *)state;
uint32_t init[16];
UNPACK64(init[0], init[1], state[0]);
UNPACK64(init[2], init[3], state[1]);
UNPACK64(init[4], init[5], state[2]);
UNPACK64(init[6], init[7], state[3]);
UNPACK64(init[8], init[9], state[4]);
UNPACK64(init[10], init[11], state[5]);
UNPACK64(init[12], init[13], state[6]);
UNPACK64(init[14], init[15], state[7]);
for (int i = 0; i < THREADS_PER_HASH; i++)
{
// share init among threads
for (int j = 0; j < 16; j++)
shuffle[j] = __shfl(init[j], start_lane + i);
// ugly but avoids local reads/writes
if (mix_idx == 0) {
mix = make_uint4(shuffle[0], shuffle[1], shuffle[2], shuffle[3]);
}
else if (mix_idx == 1) {
mix = make_uint4(shuffle[4], shuffle[5], shuffle[6], shuffle[7]);
}
else if (mix_idx == 2) {
mix = make_uint4(shuffle[8], shuffle[9], shuffle[10], shuffle[11]);
}
else {
mix = make_uint4(shuffle[12], shuffle[13], shuffle[14], shuffle[15]);
}
uint32_t init0 = __shfl(shuffle[0], start_lane);
for (uint32_t a = 0; a < ACCESSES; a+=4)
{
int t = ((a >> 2) & (THREADS_PER_HASH - 1));
for (uint32_t b = 0; b < 4; b++)
{
if (thread_id == t)
{
shuffle[0] = fnv(init0 ^ (a + b), ((uint32_t *)&mix)[b]) % d_dag_size;;
}
shuffle[0] = __shfl(shuffle[0], start_lane + t);
mix = fnv4(mix, g_dag[shuffle[0]].uint4s[thread_id]);
}
}
uint32_t thread_mix = fnv_reduce(mix);
// update mix accross threads
for (int j = 0; j < 8; j++)
shuffle[j] = __shfl(thread_mix, start_lane + j);
if (i == thread_id) {
//move mix into state:
PACK64(state[8], shuffle[0], shuffle[1]);
PACK64(state[9], shuffle[2], shuffle[3]);
PACK64(state[10], shuffle[4], shuffle[5]);
PACK64(state[11], shuffle[6], shuffle[7]);
}
}
// keccak_256(keccak_512(header..nonce) .. mix);
state[12] = 0x0000000000000001ULL;
for (uint32_t i = 13; i < 25; i++)
{
state[i] = 0ULL;
}
state[16] = 0x8000000000000000;
keccak_f1600_block(state, 1);
return state[0];
}
#endif
__device__ void compute_hash(
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t nonce,
hash32_t &out
)
{
extern __shared__ compute_hash_share share[];
// Compute one init hash per work item.
hash64_t init;
init_hash(g_header, nonce, init);
// Threads work together in this phase in groups of 8.
uint32_t const thread_id = threadIdx.x & (THREADS_PER_HASH-1);
uint32_t const hash_id = threadIdx.x >> 3;
hash32_t mix;
for (int i = 0; i < THREADS_PER_HASH; i++)
{
// share init with other threads
if (i == thread_id)
share[hash_id].init = init;
uint4 thread_init = share[hash_id].init.uint4s[thread_id & 3];
uint32_t thread_mix = inner_loop(thread_init, thread_id, share[hash_id].mix.uint32s, g_dag);
share[hash_id].mix.uint32s[thread_id] = thread_mix;
if (i == thread_id)
mix = share[hash_id].mix;
}
final_hash(&init, &mix, out);
}
__global__ void
__launch_bounds__(128, 7)
ethash_search(
uint32_t* g_output,
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t start_nonce,
uint64_t target
)
{
uint32_t const gid = blockIdx.x * blockDim.x + threadIdx.x;
#if 0 // _CUDA_ARCH__ >= SHUFFLE_MIN_VER
uint64_t hash = compute_hash_shuffle(g_header, g_dag, start_nonce + gid);
if (SWAP64(hash) < target)
{
atomicInc(g_output, d_max_outputs);
g_output[g_output[0]] = gid;
}
#else
hash32_t hash;
compute_hash(g_header, g_dag, start_nonce + gid, hash);
if (SWAP64(hash.uint64s[0]) <= target)
{
atomicInc(g_output,d_max_outputs);
g_output[g_output[0]] = gid;
}
#endif
}
void run_ethash_hash(
hash32_t* g_hashes,
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t start_nonce
)
{
}
void run_ethash_search(
uint32_t blocks,
uint32_t threads,
cudaStream_t stream,
uint32_t* g_output,
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t start_nonce,
uint64_t target
)
{
#if __CUDA_ARCH__ >= SHUFFLE_MIN_VER
ethash_search <<<blocks, threads, 0, stream>>>(g_output, g_header, g_dag, start_nonce, target);
#else
ethash_search <<<blocks, threads, (sizeof(compute_hash_share) * threads) / THREADS_PER_HASH, stream>>>(g_output, g_header, g_dag, start_nonce, target);
#endif
}
cudaError set_constants(
uint32_t * dag_size,
uint32_t * max_outputs
)
{
cudaError result;
result = cudaMemcpyToSymbol(d_dag_size, dag_size, sizeof(uint32_t));
result = cudaMemcpyToSymbol(d_max_outputs, max_outputs, sizeof(uint32_t));
return result;
}
static unsigned m_search_batch_size = 262144;
static unsigned m_workgroup_size = 64;
static __thread hash128_t * m_dag_ptr;
static __thread hash32_t * m_header;
//static void* m_hash_buf[MAX_GPUS] = { 0 };
static uint32_t* m_search_buf[MAX_GPUS] = { 0 };
static cudaStream_t m_streams[MAX_GPUS] = { 0 };
static uint64_t dagSize = 0;
static int fdDAG = -1;
static uint8_t const *dag = NULL;
uint32_t const c_zero = 0;
#define ETHASH_MIX_BYTES 128
#define MIX_WORDS (ETHASH_MIX_BYTES/4)
uint8_t* alloc_dag_file(uint32_t rev, uint32_t* seedhash)
{
uint8_t* buffer = NULL;
char fn[256];
struct stat stat_buf;
int rc;
sprintf(fn, "%s/.ethash/full-R%u-%08x%08x", getenv("HOME"), rev, seedhash[7], seedhash[6]);
rc = stat(fn, &stat_buf);
fdDAG = open(fn, O_RDONLY);
if (rc != 0) {
applog(LOG_ERR, "Unable to open DAG file %s (%d)!", fn, rc);
return NULL;
}
dagSize = stat_buf.st_size - 8;
#ifdef __linux__
buffer = (uint8_t*) (mmap(NULL, dagSize, PROT_READ, MAP_FILE | MAP_SHARED | MAP_POPULATE , fdDAG, 0));
#else
buffer = (uint8_t*) (mmap(NULL, dagSize, PROT_READ, MAP_FILE | MAP_SHARED), fdDAG, 0));
#endif
if (buffer == MAP_FAILED) {
applog(LOG_ERR, "Unable to map DAG file %s!", fn);
return NULL;
}
if (buffer[0] != 0xfe) {
applog(LOG_ERR, "DAG file header is bad %s!", fn);
close(fdDAG);
return NULL;
}
return (buffer+8);
}
void close_dag_file(uint8_t const *dag)
{
uint8_t* buffer = dag ? (uint8_t*) dag - 8 : NULL;
munmap(buffer, dagSize);
dag = NULL;
close(fdDAG);
fdDAG = -1;
}
static bool init[MAX_GPUS] = { 0 };
int scanhash_ether(int thr_id, struct work *work, uint64_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t *pseed = &work->data[8];
const uint64_t first_nonce = (((uint64_t)pdata[20]) << 32) + pdata[19];
uint64_t nonce = first_nonce;
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 18);
throughput = min(throughput, max_nonce - first_nonce);
if (!dagSize) {
dag = alloc_dag_file(23, pseed);
applog(LOG_NOTICE, "DAG size=%ld", (long int) dagSize);
}
//0000.0000 0050.4c39c 047378ad262a995be73cc504643443728e1ca415bae9208
uint64_t target64 = ptarget[7];
target64 = (target64 << 32) + ptarget[6];
if (!init[thr_id])
{
uint32_t dag_size = (unsigned)(dagSize / ETHASH_MIX_BYTES);
uint32_t max_outputs=63;
cudaSetDevice(device_map[thr_id]);
// create buffer for header256
cudaMalloc(&m_header, 32);
// create mining buffer
cudaMalloc(&m_search_buf[thr_id], 8 * sizeof(uint32_t));
//cudaMalloc(&m_hash_buf[thr_id], c_hash_batch_size * 32);
//cudaStreamCreate(&m_streams[thr_id]);
m_streams[thr_id] = 0;
// create buffer for dag
CUDA_SAFE_CALL(cudaMalloc(&m_dag_ptr, dagSize));
CUDA_SAFE_CALL(cudaMemcpy(m_dag_ptr, dag, dagSize, cudaMemcpyHostToDevice));
CUDA_SAFE_CALL(set_constants(&dag_size, &max_outputs));
init[thr_id] = true;
}
else if (dag) {
close_dag_file(dag);
}
CUDA_SAFE_CALL(cudaMemcpy(m_header, work->data, 32, cudaMemcpyHostToDevice));
CUDA_SAFE_CALL(cudaMemcpy(m_search_buf[thr_id], &c_zero, 4, cudaMemcpyHostToDevice));
uint32_t results[8] = { 0 };
do {
results[0] = 0;
CUDA_SAFE_CALL(cudaMemset(m_search_buf[thr_id], 0, sizeof(uint32_t)));
run_ethash_search(m_search_batch_size / m_workgroup_size, m_workgroup_size, m_streams[thr_id], m_search_buf[thr_id], m_header, m_dag_ptr, nonce, target64);
CUDA_SAFE_CALL(cudaThreadSynchronize());
CUDA_SAFE_CALL(cudaMemcpyAsync(results, m_search_buf[thr_id], 2*sizeof(uint32_t), cudaMemcpyDeviceToHost, m_streams[thr_id]));
unsigned num_found = results[0];
if (num_found) {
nonce = nonce + results[1];
pdata[19] = (uint32_t) nonce;
pdata[20] = (uint32_t) (nonce >> 32);
return 1;
}
nonce += throughput;
} while (nonce < max_nonce && !work_restart[thr_id].restart);
*hashes_done = nonce - first_nonce + 1;
return 0;
}
-59
View File
@@ -1,59 +0,0 @@
#ifndef _ETHASH_CU_MINER_KERNEL_H_
#define _ETHASH_CU_MINER_KERNEL_H_
#include <stdint.h>
typedef union
{
uint64_t uint64s[16 / sizeof(uint64_t)];
uint32_t uint32s[16 / sizeof(uint32_t)];
} hash16_t;
typedef union
{
uint32_t uint32s[32 / sizeof(uint32_t)];
uint64_t uint64s[32 / sizeof(uint64_t)];
} hash32_t;
typedef union
{
uint32_t uint32s[64 / sizeof(uint32_t)];
uint64_t uint64s[64 / sizeof(uint64_t)];
uint4 uint4s[64 / sizeof(uint4)];
} hash64_t;
typedef union
{
uint32_t uint32s[128 / sizeof(uint32_t)];
uint4 uint4s[128 / sizeof(uint4)];
} hash128_t;
//typedef uint32_t hash128_t;
cudaError set_constants(
uint32_t * dag_size,
uint32_t * max_outputs
);
void run_ethash_hash(
hash32_t* g_hashes,
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t start_nonce
);
void run_ethash_search(
uint32_t search_batch_size,
uint32_t workgroup_size,
cudaStream_t stream,
uint32_t* g_output,
hash32_t const* g_header,
hash128_t const* g_dag,
uint64_t start_nonce,
uint64_t target
);
#endif
-510
View File
@@ -1,510 +0,0 @@
#include "cpuminer-config.h"
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
#include <stdarg.h>
#include <unistd.h>
//#include <jansson.h>
#include <curl/curl.h>
#ifdef WIN32
#include <winsock2.h>
#include <mstcpip.h>
#else
#include <errno.h>
#include <sys/socket.h>
#include <netinet/in.h>
#include <netinet/tcp.h>
#endif
#include "miner.h"
struct data_buffer {
void *buf;
size_t len;
};
struct upload_buffer {
const void *buf;
size_t len;
size_t pos;
};
struct header_info {
char *lp_path;
char *reason;
char *stratum_url;
};
static void databuf_free(struct data_buffer *db)
{
if (!db)
return;
free(db->buf);
memset(db, 0, sizeof(*db));
}
static size_t all_data_cb(const void *ptr, size_t size, size_t nmemb,
void *user_data)
{
struct data_buffer *db = (struct data_buffer *)user_data;
size_t len = size * nmemb;
size_t oldlen, newlen;
void *newmem;
static const uchar zero = 0;
oldlen = db->len;
newlen = oldlen + len;
newmem = realloc(db->buf, newlen + 1);
if (!newmem)
return 0;
db->buf = newmem;
db->len = newlen;
memcpy((char*)db->buf + oldlen, ptr, len);
memcpy((char*)db->buf + newlen, &zero, 1); /* null terminate */
return len;
}
static size_t upload_data_cb(void *ptr, size_t size, size_t nmemb,
void *user_data)
{
struct upload_buffer *ub = (struct upload_buffer *)user_data;
unsigned int len = (unsigned int)(size * nmemb);
if (len > ub->len - ub->pos)
len = (unsigned int)(ub->len - ub->pos);
if (len) {
memcpy(ptr, (char*)ub->buf + ub->pos, len);
ub->pos += len;
}
return len;
}
#if LIBCURL_VERSION_NUM >= 0x071200
static int seek_data_cb(void *user_data, curl_off_t offset, int origin)
{
struct upload_buffer *ub = (struct upload_buffer *)user_data;
switch (origin) {
case SEEK_SET:
ub->pos = (size_t)offset;
break;
case SEEK_CUR:
ub->pos += (size_t)offset;
break;
case SEEK_END:
ub->pos = ub->len + (size_t)offset;
break;
default:
return 1; /* CURL_SEEKFUNC_FAIL */
}
return 0; /* CURL_SEEKFUNC_OK */
}
#endif
static size_t resp_hdr_cb(void *ptr, size_t size, size_t nmemb, void *user_data)
{
struct header_info *hi = (struct header_info *)user_data;
size_t remlen, slen, ptrlen = size * nmemb;
char *rem, *val = NULL, *key = NULL;
void *tmp;
val = (char*)calloc(1, ptrlen);
key = (char*)calloc(1, ptrlen);
if (!key || !val)
goto out;
tmp = memchr(ptr, ':', ptrlen);
if (!tmp || (tmp == ptr)) /* skip empty keys / blanks */
goto out;
slen = (size_t)((char*)tmp - (char*)ptr);
if ((slen + 1) == ptrlen) /* skip key w/ no value */
goto out;
memcpy(key, ptr, slen); /* store & nul term key */
key[slen] = 0;
rem = (char*)ptr + slen + 1; /* trim value's leading whitespace */
remlen = ptrlen - slen - 1;
while ((remlen > 0) && (isspace(*rem))) {
remlen--;
rem++;
}
memcpy(val, rem, remlen); /* store value, trim trailing ws */
val[remlen] = 0;
while ((*val) && (isspace(val[strlen(val) - 1]))) {
val[strlen(val) - 1] = 0;
}
if (!*val) /* skip blank value */
goto out;
if (!strcasecmp("X-Long-Polling", key)) {
hi->lp_path = val; /* X-Mining-Extensions: longpoll */
val = NULL;
}
if (!strcasecmp("X-Reject-Reason", key)) {
hi->reason = val; /* X-Mining-Extensions: reject-reason */
//applog(LOG_WARNING, "%s:%s", key, val);
val = NULL;
}
if (!strcasecmp("X-Stratum", key)) {
hi->stratum_url = val; /* steal memory reference */
val = NULL;
}
if (!strcasecmp("X-Nonce-Range", key)) {
/* todo when available: X-Mining-Extensions: noncerange */
}
out:
free(key);
free(val);
return ptrlen;
}
#if LIBCURL_VERSION_NUM >= 0x070f06
static int sockopt_keepalive_cb(void *userdata, curl_socket_t fd,
curlsocktype purpose)
{
int keepalive = 1;
int tcp_keepcnt = 3;
int tcp_keepidle = 50;
int tcp_keepintvl = 50;
#ifdef WIN32
DWORD outputBytes;
#endif
#ifndef WIN32
if (unlikely(setsockopt(fd, SOL_SOCKET, SO_KEEPALIVE, &keepalive,
sizeof(keepalive))))
return 1;
#ifdef __linux
if (unlikely(setsockopt(fd, SOL_TCP, TCP_KEEPCNT,
&tcp_keepcnt, sizeof(tcp_keepcnt))))
return 1;
if (unlikely(setsockopt(fd, SOL_TCP, TCP_KEEPIDLE,
&tcp_keepidle, sizeof(tcp_keepidle))))
return 1;
if (unlikely(setsockopt(fd, SOL_TCP, TCP_KEEPINTVL,
&tcp_keepintvl, sizeof(tcp_keepintvl))))
return 1;
#endif /* __linux */
#ifdef __APPLE_CC__
if (unlikely(setsockopt(fd, IPPROTO_TCP, TCP_KEEPALIVE,
&tcp_keepintvl, sizeof(tcp_keepintvl))))
return 1;
#endif /* __APPLE_CC__ */
#else /* WIN32 */
struct tcp_keepalive vals;
vals.onoff = 1;
vals.keepalivetime = tcp_keepidle * 1000;
vals.keepaliveinterval = tcp_keepintvl * 1000;
if (unlikely(WSAIoctl(fd, SIO_KEEPALIVE_VALS, &vals, sizeof(vals),
NULL, 0, &outputBytes, NULL, NULL)))
return 1;
#endif /* WIN32 */
return 0;
}
#endif
static json_t *ether_json_rpc(CURL *curl, const char *url, const char *rpc_req, int *curl_err)
{
char len_hdr[64], hashrate_hdr[64];
char curl_err_str[CURL_ERROR_SIZE] = { 0 };
struct data_buffer all_data = { 0 };
struct upload_buffer upload_data;
struct curl_slist *headers = NULL;
struct header_info hi = { 0 };
char *httpdata;
json_t *val, *err_val, *res_val;
json_error_t err;
int rc;
long timeout = opt_timeout;
bool keepalive = false;
/* it is assumed that 'curl' is freshly [re]initialized at this pt */
if (opt_protocol)
curl_easy_setopt(curl, CURLOPT_VERBOSE, 1);
curl_easy_setopt(curl, CURLOPT_URL, url);
if (opt_cert)
curl_easy_setopt(curl, CURLOPT_CAINFO, opt_cert);
curl_easy_setopt(curl, CURLOPT_ENCODING, "");
curl_easy_setopt(curl, CURLOPT_FAILONERROR, 0);
curl_easy_setopt(curl, CURLOPT_NOSIGNAL, 1);
curl_easy_setopt(curl, CURLOPT_TCP_NODELAY, 1);
curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, all_data_cb);
curl_easy_setopt(curl, CURLOPT_WRITEDATA, &all_data);
curl_easy_setopt(curl, CURLOPT_READFUNCTION, upload_data_cb);
curl_easy_setopt(curl, CURLOPT_READDATA, &upload_data);
#if LIBCURL_VERSION_NUM >= 0x071200
curl_easy_setopt(curl, CURLOPT_SEEKFUNCTION, &seek_data_cb);
curl_easy_setopt(curl, CURLOPT_SEEKDATA, &upload_data);
#endif
curl_easy_setopt(curl, CURLOPT_ERRORBUFFER, curl_err_str);
curl_easy_setopt(curl, CURLOPT_FOLLOWLOCATION, 1);
curl_easy_setopt(curl, CURLOPT_TIMEOUT, timeout);
curl_easy_setopt(curl, CURLOPT_HEADERFUNCTION, resp_hdr_cb);
curl_easy_setopt(curl, CURLOPT_HEADERDATA, &hi);
if (opt_proxy) {
curl_easy_setopt(curl, CURLOPT_PROXY, opt_proxy);
curl_easy_setopt(curl, CURLOPT_PROXYTYPE, opt_proxy_type);
}
#if 0
if (userpass) {
curl_easy_setopt(curl, CURLOPT_USERPWD, userpass);
curl_easy_setopt(curl, CURLOPT_HTTPAUTH, CURLAUTH_BASIC);
}
#endif
#if LIBCURL_VERSION_NUM >= 0x070f06
if (keepalive)
curl_easy_setopt(curl, CURLOPT_SOCKOPTFUNCTION, sockopt_keepalive_cb);
#endif
curl_easy_setopt(curl, CURLOPT_POST, 1);
if (opt_protocol)
applog(LOG_DEBUG, "JSON protocol request:\n%s", rpc_req);
upload_data.buf = rpc_req;
upload_data.len = strlen(rpc_req);
upload_data.pos = 0;
sprintf(len_hdr, "Content-Length: %lu", (unsigned long) upload_data.len);
sprintf(hashrate_hdr, "X-Mining-Hashrate: %llu", (unsigned long long) global_hashrate);
headers = curl_slist_append(headers, "Content-Type: application/json");
headers = curl_slist_append(headers, len_hdr);
headers = curl_slist_append(headers, "User-Agent: " USER_AGENT);
headers = curl_slist_append(headers, "X-Mining-Extensions: longpoll noncerange reject-reason");
headers = curl_slist_append(headers, hashrate_hdr);
headers = curl_slist_append(headers, "Accept:"); /* disable Accept hdr*/
headers = curl_slist_append(headers, "Expect:"); /* disable Expect hdr*/
curl_easy_setopt(curl, CURLOPT_HTTPHEADER, headers);
rc = curl_easy_perform(curl);
if (curl_err != NULL)
*curl_err = rc;
if (rc) {
if (rc != CURLE_OPERATION_TIMEDOUT) {
applog(LOG_ERR, "HTTP request failed: %s", curl_err_str);
goto err_out;
}
}
if (!all_data.buf || !all_data.len) {
applog(LOG_ERR, "Empty data received in json_rpc_call.");
goto err_out;
}
httpdata = (char*) all_data.buf;
if (*httpdata != '{' && *httpdata != '[') {
long errcode = 0;
CURLcode c = curl_easy_getinfo(curl, CURLINFO_RESPONSE_CODE, &errcode);
if (c == CURLE_OK && errcode == 401) {
applog(LOG_ERR, "You are not authorized, check your login and password.");
goto err_out;
}
}
val = JSON_LOADS(httpdata, &err);
if (!val) {
applog(LOG_ERR, "JSON decode failed(%d): %s", err.line, err.text);
if (opt_protocol)
applog(LOG_DEBUG, "%s", httpdata);
goto err_out;
}
if (opt_protocol) {
char *s = json_dumps(val, JSON_INDENT(3));
applog(LOG_DEBUG, "JSON protocol response:\n%s\n", s);
free(s);
}
/* JSON-RPC valid response returns a non-null 'result',
* and a null 'error'. */
res_val = json_object_get(val, "result");
err_val = json_object_get(val, "error");
if (!res_val || json_is_null(res_val) ||
(err_val && !json_is_null(err_val))) {
char *s = NULL;
if (err_val) {
s = json_dumps(err_val, 0);
json_t *msg = json_object_get(err_val, "message");
json_t *err_code = json_object_get(err_val, "code");
if (curl_err && json_integer_value(err_code))
*curl_err = (int) json_integer_value(err_code);
if (json_is_string(msg)) {
free(s);
s = strdup(json_string_value(msg));
if (have_longpoll && s && !strcmp(s, "method not getwork")) {
json_decref(err_val);
free(s);
goto err_out;
}
}
json_decref(err_val);
}
else
s = strdup("(unknown reason)");
if (!curl_err || opt_debug)
applog(LOG_ERR, "JSON-RPC call failed: %s", s);
free(s);
goto err_out;
}
if (hi.reason)
json_object_set_new(val, "reject-reason", json_string(hi.reason));
databuf_free(&all_data);
curl_slist_free_all(headers);
curl_easy_reset(curl);
return val;
err_out:
free(hi.lp_path);
free(hi.reason);
free(hi.stratum_url);
databuf_free(&all_data);
curl_slist_free_all(headers);
curl_easy_reset(curl);
return NULL;
}
// Made for len 32
bool hex2uint(uint32_t *target, const char *hexstr, int len)
{
char buf[32];
char pad[80];
char hex_byte[3];
char *ep;
char *p = buf;
char *hex = pad;
hex_byte[2] = '\0';
// pad, we can receive non padded numbers (0x44 => 0x0044)
size_t hexlen = strlen(hexstr);
if (hexlen < len*2)
memset(hex, '0', len*2 - hexlen);
sprintf(&hex[len*2 - hexlen],"%s", hexstr);
while (*hex && len) {
if (!hex[1]) {
applog(LOG_ERR, "hex2uint str truncated");
return false;
}
hex_byte[0] = hex[0];
hex_byte[1] = hex[1];
*p = (uchar) strtol(hex_byte, &ep, 16);
if (*ep) {
applog(LOG_ERR, "hex2uint failed on '%s'", hex_byte);
return false;
}
p++;
hex += 2;
len--;
if (len % 8 == 0) {
//..
}
}
for (int i = 0; i < 8; i++) {
be32enc(&target[7 - i], ((uint32_t*)buf)[i]);
}
return (len == 0 && *hexstr == 0) ? true : false;
}
void uint2hex(char *out, uint32_t *in, int len)
{
if (out) {
for (int i = 0; i < len/sizeof(uint32_t); i++)
sprintf(out + (i * 8), "%08x", in[7-i]);
}
}
bool ether_getwork(CURL *curl, struct pool_infos *pool, struct work *work)
{
char req[512];
bool ret = false;
sprintf(req, "{ \"jsonrpc\": \"2.0\", \"method\": \"eth_getWork\", \"params\": [], \"id\": 4 }\r\n");
json_t *result = ether_json_rpc(curl, pool->url, req, NULL);
if (result)
ret = true;
json_t *arr = json_object_get(result, "result");
if (!json_is_array(arr)) {
return false;
}
const char *header = json_string_value(json_array_get(arr, 0));
const char *seed = json_string_value(json_array_get(arr, 1));
const char *target = json_string_value(json_array_get(arr, 2));
//0x2b800e3ffcafdf7ffdd5b2ea1c0ed79c534e92b3459331eae61597943d0743bc
//0x356e5a2cc1eba076e650ac7473fccc37952b46bc2e419a200cec0c451dce2336
//0x000000000050ec7dea051867e987e3eed0066a128f161b5776a8f5da326b06ec
applog(LOG_NOTICE, "%s %s %s", header, seed, target);
hex2uint(work->data, &header[2], 32);
hex2uint(&work->data[8], &seed[2], 32);
hex2uint(work->target, &target[2], 32);
//applog(LOG_BLUE, "%08x %08x", work->target[7], work->target[6]);
return ret;
}
bool ether_submitwork(CURL *curl, struct pool_infos *pool, struct work *work)
{
char req[512];
bool ret = false;
char noncehex[32]={0}, header[80]={0}, mixhash[80]={0};
sprintf(noncehex, "0x%08x%08x", work->data[20], work->data[19]);
uint2hex(header, work->data, 32);
uint2hex(mixhash, &work->data[24], 32);
applog(LOG_NOTICE, "%s mixhash:", __func__);
applog_hash((uchar*)&work->data[24]);
sprintf(req, "{ \"jsonrpc\": \"2.0\", \"method\": \"eth_submitWork\","
"\"params\": [\"%s\", \"0x%s\", \"0x%s\"], \"id\": 4 }\r\n",
noncehex, header, mixhash
);
json_t *result = ether_json_rpc(curl, pool->url, req, NULL);
if (result)
ret = true;
json_t *res = json_object_get(result, "result");
if (!json_is_true(res)) {
applog(LOG_ERR, "not true! %s", json_dumps(result, JSON_INDENT(3)));
exit(0);
return false;
}
return ret;
}
-83
View File
@@ -1,83 +0,0 @@
__device__ __constant__ uint64_t const 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
};
__device__ static void keccak_f1600_block(uint64_t* s, uint32_t out_size)//, uint32_t in_size, uint32_t out_size)
{
uint64_t t[5], u, v;
for (size_t 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) */
/* theta: a[0,i], a[1,i], .. a[4,i] ^= d[i] */
u = t[4] ^ ROTL64L(t[1], 1);
s[0] ^= u; s[5] ^= u; s[10] ^= u; s[15] ^= u; s[20] ^= u;
u = t[0] ^ ROTL64L(t[2], 1);
s[1] ^= u; s[6] ^= u; s[11] ^= u; s[16] ^= u; s[21] ^= u;
u = t[1] ^ ROTL64L(t[3], 1);
s[2] ^= u; s[7] ^= u; s[12] ^= u; s[17] ^= u; s[22] ^= u;
u = t[2] ^ ROTL64L(t[4], 1);
s[3] ^= u; s[8] ^= u; s[13] ^= u; s[18] ^= u; s[23] ^= u;
u = t[3] ^ ROTL64L(t[0], 1);
s[4] ^= u; s[9] ^= u; s[14] ^= u; s[19] ^= u; s[24] ^= u;
/* rho pi: b[..] = rotl(a[..], ..) */
u = s[1];
s[1] = ROTL64H(s[6], 44);
s[6] = ROTL64L(s[9], 20);
s[9] = ROTL64H(s[22], 61);
s[22] = ROTL64H(s[14], 39);
s[14] = ROTL64L(s[20], 18);
s[20] = ROTL64H(s[2], 62);
s[2] = ROTL64H(s[12], 43);
s[12] = ROTL64L(s[13], 25);
s[13] = ROTL64L(s[19], 8);
s[19] = ROTL64H(s[23], 56);
s[23] = ROTL64H(s[15], 41);
s[15] = ROTL64L(s[4], 27);
s[4] = ROTL64L(s[24], 14);
s[24] = ROTL64L(s[21], 2);
s[21] = ROTL64H(s[8], 55);
s[8] = ROTL64H(s[16], 45);
s[16] = ROTL64H(s[5], 36);
s[5] = ROTL64L(s[3], 28);
s[3] = ROTL64L(s[18], 21);
s[18] = ROTL64L(s[17], 15);
s[17] = ROTL64L(s[11], 10);
s[11] = ROTL64L(s[7], 6);
s[7] = ROTL64L(s[10], 3);
s[10] = ROTL64L(u, 1);
// squeeze this in here
/* chi: a[i,j] ^= ~b[i,j+1] & b[i,j+2] */
u = s[0]; v = s[1]; s[0] ^= (~v) & s[2];
/* iota: a[0,0] ^= round constant */
s[0] ^= keccak_round_constants[i];
if (i == 23 && out_size == 1) return;
// continue chi
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];
if (i == 23) return;
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;
}
}
-64
View File
@@ -1,64 +0,0 @@
#include <cuda.h>
#include <cuda_runtime.h>
#include <stdint.h>
// 64-bit ROTATE LEFT
#if __CUDA_ARCH__ >= 320
__device__ uint64_t ROTL64H(const uint64_t x, const int offset)
{
uint64_t res;
asm("{\n\t"
".reg .u32 tl,th,vl,vh;\n\t"
"mov.b64 {tl,th}, %1;\n\t"
"shf.l.wrap.b32 vl, tl, th, %2;\n\t"
"shf.l.wrap.b32 vh, th, tl, %2;\n\t"
"mov.b64 %0, {vl,vh};\n\t"
"}"
: "=l"(res) : "l"(x), "r"(offset)
);
return res;
}
__device__ uint64_t ROTL64L(const uint64_t x, const int offset)
{
uint64_t res;
asm("{\n\t"
".reg .u32 tl,th,vl,vh;\n\t"
"mov.b64 {tl,th}, %1;\n\t"
"shf.l.wrap.b32 vl, tl, th, %2;\n\t"
"shf.l.wrap.b32 vh, th, tl, %2;\n\t"
"mov.b64 %0, {vh,vl};\n\t"
"}"
: "=l"(res) : "l"(x), "r"(offset)
);
return res;
}
#elif __CUDA_ARCH__ >= 120
#define ROTL64H(x, n) ROTL64(x,n)
#define ROTL64L(x, n) ROTL64(x,n)
__device__ __forceinline__
uint64_t ROTL64(const uint64_t x, const int offset)
{
uint64_t result;
asm("{\n\t"
".reg .b64 lhs;\n\t"
".reg .u32 roff;\n\t"
"shl.b64 lhs, %1, %2;\n\t"
"sub.u32 roff, 64, %2;\n\t"
"shr.b64 %0, %1, roff;\n\t"
"add.u64 %0, lhs, %0;\n\t"
"}\n"
: "=l"(result) : "l"(x), "r"(offset));
return result;
}
#else
#define ROTL64H(x, n) ROTL64(x,n)
#define ROTL64L(x, n) ROTL64(x,n)
/* host */
#define ROTL64(x, n) (((x) << (n)) | ((x) >> (64 - (n))))
#endif
+32 -16
View File
@@ -20,11 +20,22 @@ extern "C" void my_fugue256_addbits_and_close(void *cc, unsigned ub, unsigned n,
((((x) << 24) & 0xff000000u) | (((x) << 8) & 0x00ff0000u) | \
(((x) >> 8) & 0x0000ff00u) | (((x) >> 24) & 0x000000ffu))
void fugue256_hash(unsigned char* output, const unsigned char* input, int len)
{
sph_fugue256_context ctx;
sph_fugue256_init(&ctx);
sph_fugue256(&ctx, input, len);
sph_fugue256_close(&ctx, (void *)output);
}
static bool init[MAX_GPUS] = { 0 };
int scanhash_fugue256(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
int scanhash_fugue256(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;
uint32_t start_nonce = pdata[19]++;
int intensity = (device_sm[device_map[thr_id]] > 500) ? 22 : 19;
uint32_t throughput = device_intensity(thr_id, __func__, 1 << intensity); // 256*256*8
@@ -42,8 +53,7 @@ int scanhash_fugue256(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
init[thr_id] = true;
}
// Endian Drehung ist notwendig
uint32_t endiandata[20];
// Endian
for (int kk=0; kk < 20; kk++)
be32enc(&endiandata[kk], pdata[kk]);
@@ -57,17 +67,17 @@ int scanhash_fugue256(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
if (foundNounce < UINT32_MAX)
{
uint32_t hash[8];
const uint32_t Htarg = ptarget[7];
endiandata[19] = SWAP32(foundNounce);
uint32_t vhash[8];
sph_fugue256_context ctx_fugue;
endiandata[19] = SWAP32(foundNounce);
sph_fugue256_init(&ctx_fugue);
sph_fugue256 (&ctx_fugue, endiandata, 80);
sph_fugue256_close(&ctx_fugue, &hash);
sph_fugue256_close(&ctx_fugue, &vhash);
if (hash[7] <= Htarg && fulltest(hash, ptarget))
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget))
{
bn_store_hash_target_ratio(vhash, ptarget, work);
pdata[19] = foundNounce;
*hashes_done = foundNounce - start_nonce + 1;
return 1;
@@ -90,11 +100,17 @@ int scanhash_fugue256(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
return 0;
}
void fugue256_hash(unsigned char* output, const unsigned char* input, int len)
// cleanup
void free_fugue256(int thr_id)
{
sph_fugue256_context ctx;
if (!init[thr_id])
return;
sph_fugue256_init(&ctx);
sph_fugue256(&ctx, input, len);
sph_fugue256_close(&ctx, (void *)output);
}
cudaSetDevice(device_map[thr_id]);
fugue256_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+21 -6
View File
@@ -27,10 +27,11 @@ void groestlhash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
int scanhash_groestlcoin(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
int scanhash_groestlcoin(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[20];
uint32_t _ALIGN(64) endiandata[32];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t start_nonce = pdata[19];
uint32_t throughput = device_intensity(thr_id, __func__, 1 << 19); // 256*256*8
throughput = min(throughput, max_nonce - start_nonce);
@@ -62,11 +63,12 @@ int scanhash_groestlcoin(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
if (foundNounce < UINT32_MAX)
{
uint32_t _ALIGN(64) tmpHash[8];
uint32_t _ALIGN(64) vhash[8];
endiandata[19] = swab32(foundNounce);
groestlhash(tmpHash, endiandata);
groestlhash(vhash, endiandata);
if (tmpHash[7] <= ptarget[7] && fulltest(tmpHash, ptarget)) {
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
bn_store_hash_target_ratio(vhash, ptarget, work);
pdata[19] = foundNounce;
free(outputHash);
return true;
@@ -89,3 +91,16 @@ int scanhash_groestlcoin(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
return 0;
}
// cleanup
void free_groestlcoin(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
groestlcoin_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+30 -15
View File
@@ -130,10 +130,10 @@ struct check_nonce_for_remove
static bool init[MAX_GPUS] = { 0 };
__host__
int scanhash_heavy(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done, uint32_t maxvote, int blocklen)
int scanhash_heavy(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done, uint32_t maxvote, int blocklen)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
// CUDA will process thousands of threads.
uint32_t throughput = device_intensity(thr_id, __func__, (1U << 19) - 256);
@@ -271,18 +271,17 @@ int scanhash_heavy(int thr_id, uint32_t *pdata,
{
uint32_t nonce = cpu_nonceVector[i];
uint32_t *foundhash = &hash[8*i];
if (foundhash[7] <= ptarget[7]) {
if (fulltest(foundhash, ptarget)) {
uint32_t verification[8];
pdata[19] += nonce - pdata[19];
heavycoin_hash((uchar*)verification, (uchar*)pdata, blocklen);
if (memcmp(verification, foundhash, 8*sizeof(uint32_t))) {
applog(LOG_ERR, "hash for nonce=$%08X does not validate on CPU!\n", nonce);
} else {
*hashes_done = pdata[19] - first_nonce;
rc = 1;
goto exit;
}
if (foundhash[7] <= ptarget[7] && fulltest(foundhash, ptarget)) {
uint32_t vhash[8];
pdata[19] += nonce - pdata[19];
heavycoin_hash((uchar*)vhash, (uchar*)pdata, blocklen);
if (memcmp(vhash, foundhash, 8*sizeof(uint32_t))) {
applog(LOG_ERR, "hash for nonce %08x does not validate on CPU!\n", nonce);
} else {
*hashes_done = pdata[19] - first_nonce;
bn_store_hash_target_ratio(vhash, ptarget, work);
rc = 1;
goto exit;
}
}
}
@@ -301,6 +300,22 @@ exit:
return rc;
}
// cleanup
extern "C" void free_heavy(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(heavy_nonceVector[thr_id]);
// todo: free sub algos vectors
init[thr_id] = false;
cudaDeviceSynchronize();
}
__host__
void heavycoin_hash(uchar* output, const uchar* input, int len)
{
Regular → Executable
+7 -7
View File
@@ -1,7 +1,7 @@
#!/bin/sh
# install - install a program, script, or datafile
scriptversion=2011-01-19.21; # UTC
scriptversion=2011-11-20.07; # UTC
# This originates from X11R5 (mit/util/scripts/install.sh), which was
# later released in X11R6 (xc/config/util/install.sh) with the
@@ -35,7 +35,7 @@ scriptversion=2011-01-19.21; # UTC
# FSF changes to this file are in the public domain.
#
# Calling this script install-sh is preferred over install.sh, to prevent
# `make' implicit rules from creating a file called install from it
# 'make' implicit rules from creating a file called install from it
# when there is no Makefile.
#
# This script is compatible with the BSD install script, but was written
@@ -156,7 +156,7 @@ while test $# -ne 0; do
-s) stripcmd=$stripprog;;
-t) dst_arg=$2
# Protect names problematic for `test' and other utilities.
# Protect names problematic for 'test' and other utilities.
case $dst_arg in
-* | [=\(\)!]) dst_arg=./$dst_arg;;
esac
@@ -190,7 +190,7 @@ if test $# -ne 0 && test -z "$dir_arg$dst_arg"; then
fi
shift # arg
dst_arg=$arg
# Protect names problematic for `test' and other utilities.
# Protect names problematic for 'test' and other utilities.
case $dst_arg in
-* | [=\(\)!]) dst_arg=./$dst_arg;;
esac
@@ -202,7 +202,7 @@ if test $# -eq 0; then
echo "$0: no input file specified." >&2
exit 1
fi
# It's OK to call `install-sh -d' without argument.
# It's OK to call 'install-sh -d' without argument.
# This can happen when creating conditional directories.
exit 0
fi
@@ -240,7 +240,7 @@ fi
for src
do
# Protect names problematic for `test' and other utilities.
# Protect names problematic for 'test' and other utilities.
case $src in
-* | [=\(\)!]) src=./$src;;
esac
@@ -354,7 +354,7 @@ do
if test -z "$dir_arg" || {
# Check for POSIX incompatibilities with -m.
# HP-UX 11.23 and IRIX 6.5 mkdir -m -p sets group- or
# other-writeable bit of parent directory when it shouldn't.
# other-writable bit of parent directory when it shouldn't.
# FreeBSD 6.1 mkdir -m -p sets mode of existing directory.
ls_ld_tmpdir=`ls -ld "$tmpdir"`
case $ls_ld_tmpdir in
+49 -46
View File
@@ -44,7 +44,7 @@
*
* @return 0 if the key is generated correctly; -1 if there is an error (usually due to lack of memory for allocation)
*/
int LYRA2(void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen, const void *salt, uint64_t saltlen, uint64_t timeCost, uint64_t nRows, uint64_t nCols)
int LYRA2(void *K, int64_t kLen, const void *pwd, int32_t pwdlen, const void *salt, int32_t saltlen, int64_t timeCost, const int16_t nRows, const int16_t nCols)
{
//============================= Basic variables ============================//
int64_t row = 2; //index of row to be processed
@@ -55,25 +55,32 @@ int LYRA2(void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen, const void *
int64_t window = 2; //Visitation window (used to define which rows can be revisited during Setup)
int64_t gap = 1; //Modifier to the step, assuming the values 1 or -1
int64_t i; //auxiliary iteration counter
int64_t v64; // 64bit var for memcpy
//==========================================================================/
//========== Initializing the Memory Matrix and pointers to it =============//
//Tries to allocate enough space for the whole memory matrix
i = (int64_t) ((int64_t) nRows * (int64_t) ROW_LEN_BYTES);
uint64_t *wholeMatrix = (uint64_t*) malloc((size_t) i);
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * nCols;
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
// for Lyra2REv2, nCols = 4, v1 was using 8
const int64_t BLOCK_LEN = (nCols == 4) ? BLOCK_LEN_BLAKE2_SAFE_INT64 : BLOCK_LEN_BLAKE2_SAFE_BYTES;
size_t sz = (size_t)ROW_LEN_BYTES * nRows;
uint64_t *wholeMatrix = malloc(sz);
if (wholeMatrix == NULL) {
return -1;
}
memset(wholeMatrix, 0, (size_t) i);
memset(wholeMatrix, 0, sz);
//Allocates pointers to each row of the matrix
uint64_t **memMatrix = malloc((size_t) nRows * sizeof(uint64_t*));
uint64_t **memMatrix = malloc(sizeof(uint64_t*) * nRows);
if (memMatrix == NULL) {
return -1;
}
//Places the pointers in the correct positions
uint64_t *ptrWord = wholeMatrix;
for (i = 0; i < (int64_t) nRows; i++) {
for (i = 0; i < nRows; i++) {
memMatrix[i] = ptrWord;
ptrWord += ROW_LEN_INT64;
}
@@ -84,32 +91,38 @@ int LYRA2(void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen, const void *
//but this ensures that the password copied locally will be overwritten as soon as possible
//First, we clean enough blocks for the password, salt, basil and padding
uint64_t nBlocksInput = ((saltlen + pwdlen + 6 * sizeof (uint64_t)) / BLOCK_LEN_BLAKE2_SAFE_BYTES) + 1;
int64_t nBlocksInput = ((saltlen + pwdlen + 6 * sizeof(uint64_t)) / BLOCK_LEN_BLAKE2_SAFE_BYTES) + 1;
byte *ptrByte = (byte*) wholeMatrix;
memset(ptrByte, 0, (size_t) nBlocksInput * BLOCK_LEN_BLAKE2_SAFE_BYTES);
//Prepends the password
memcpy(ptrByte, pwd, (size_t) pwdlen);
memcpy(ptrByte, pwd, pwdlen);
ptrByte += pwdlen;
//Concatenates the salt
memcpy(ptrByte, salt, (size_t) saltlen);
memcpy(ptrByte, salt, saltlen);
ptrByte += saltlen;
memset(ptrByte, 0, (size_t) (nBlocksInput * BLOCK_LEN_BLAKE2_SAFE_BYTES - (saltlen + pwdlen)));
//Concatenates the basil: every integer passed as parameter, in the order they are provided by the interface
memcpy(ptrByte, &kLen, sizeof (uint64_t));
ptrByte += sizeof (uint64_t);
memcpy(ptrByte, &pwdlen, sizeof (uint64_t));
ptrByte += sizeof (uint64_t);
memcpy(ptrByte, &saltlen, sizeof (uint64_t));
ptrByte += sizeof (uint64_t);
memcpy(ptrByte, &timeCost, sizeof (uint64_t));
ptrByte += sizeof (uint64_t);
memcpy(ptrByte, &nRows, sizeof (uint64_t));
ptrByte += sizeof (uint64_t);
memcpy(ptrByte, &nCols, sizeof (uint64_t));
ptrByte += sizeof (uint64_t);
memcpy(ptrByte, &kLen, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = pwdlen;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = saltlen;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = timeCost;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = nRows;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = nCols;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
//Now comes the padding
*ptrByte = 0x80; //first byte of padding: right after the password
@@ -120,30 +133,27 @@ int LYRA2(void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen, const void *
//======================= Initializing the Sponge State ====================//
//Sponge state: 16 uint64_t, BLOCK_LEN_INT64 words of them for the bitrate (b) and the remainder for the capacity (c)
uint64_t *state = malloc(16 * sizeof (uint64_t));
if (state == NULL) {
return -1;
}
uint64_t state[16];
initState(state);
//==========================================================================/
//================================ Setup Phase =============================//
//Absorbing salt, password and basil: this is the only place in which the block length is hard-coded to 512 bits
ptrWord = wholeMatrix;
for (i = 0; i < (int64_t) nBlocksInput; i++) {
for (i = 0; i < nBlocksInput; i++) {
absorbBlockBlake2Safe(state, ptrWord); //absorbs each block of pad(pwd || salt || basil)
ptrWord += BLOCK_LEN_BLAKE2_SAFE_BYTES; //goes to next block of pad(pwd || salt || basil)
ptrWord += BLOCK_LEN; //goes to next block of pad(pwd || salt || basil)
}
//Initializes M[0] and M[1]
reducedSqueezeRow0(state, memMatrix[0]); //The locally copied password is most likely overwritten here
reducedSqueezeRow0(state, memMatrix[0], nCols); //The locally copied password is most likely overwritten here
reducedDuplexRow1(state, memMatrix[0], memMatrix[1]);
reducedDuplexRow1(state, memMatrix[0], memMatrix[1], nCols);
do {
//M[row] = rand; //M[row*] = M[row*] XOR rotW(rand)
reducedDuplexRowSetup(state, memMatrix[prev], memMatrix[rowa], memMatrix[row]);
reducedDuplexRowSetup(state, memMatrix[prev], memMatrix[rowa], memMatrix[row], nCols);
//updates the value of row* (deterministically picked during Setup))
rowa = (rowa + step) & (window - 1);
@@ -159,53 +169,46 @@ int LYRA2(void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen, const void *
gap = -gap; //inverts the modifier to the step
}
} while (row < (int64_t) nRows);
} while (row < nRows);
//==========================================================================/
//============================ Wandering Phase =============================//
row = 0; //Resets the visitation to the first row of the memory matrix
for (tau = 1; tau <= (int64_t) timeCost; tau++) {
for (tau = 1; tau <= timeCost; tau++) {
//Step is approximately half the number of all rows of the memory matrix for an odd tau; otherwise, it is -1
step = (tau % 2 == 0) ? -1 : nRows / 2 - 1;
do {
//Selects a pseudorandom index row*
//------------------------------------------------------------------------------------------
//rowa = ((unsigned int)state[0]) & (nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
rowa = ((uint64_t) (state[0])) % nRows; //(USE THIS FOR THE "GENERIC" CASE)
rowa = state[0] & (unsigned int)(nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
//rowa = state[0] % nRows; //(USE THIS FOR THE "GENERIC" CASE)
//------------------------------------------------------------------------------------------
//Performs a reduced-round duplexing operation over M[row*] XOR M[prev], updating both M[row*] and M[row]
reducedDuplexRow(state, memMatrix[prev], memMatrix[rowa], memMatrix[row]);
reducedDuplexRow(state, memMatrix[prev], memMatrix[rowa], memMatrix[row], nCols);
//update prev: it now points to the last row ever computed
prev = row;
//updates row: goes to the next row to be computed
//------------------------------------------------------------------------------------------
//row = (row + step) & (nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
row = (row + step) % nRows; //(USE THIS FOR THE "GENERIC" CASE)
row = (row + step) & (unsigned int)(nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
//row = (row + step) % nRows; //(USE THIS FOR THE "GENERIC" CASE)
//------------------------------------------------------------------------------------------
} while (row != 0);
}
//==========================================================================/
//============================ Wrap-up Phase ===============================//
//Absorbs the last block of the memory matrix
absorbBlock(state, memMatrix[rowa]);
//Squeezes the key
squeeze(state, K, (size_t) kLen);
//==========================================================================/
squeeze(state, K, (unsigned int) kLen);
//========================= Freeing the memory =============================//
free(memMatrix);
free(wholeMatrix);
//Wiping out the sponge's internal state before freeing it
memset(state, 0, 16 * sizeof (uint64_t));
free(state);
//==========================================================================/
return 0;
}
+1 -9
View File
@@ -37,14 +37,6 @@ typedef unsigned char byte;
#define BLOCK_LEN_BYTES (BLOCK_LEN_INT64 * 8) //Block length, in bytes
#endif
#ifndef N_COLS
#define N_COLS 8 //Number of columns in the memory matrix: fixed to 64 by default
#endif
#define ROW_LEN_INT64 (BLOCK_LEN_INT64 * N_COLS) //Total length of a row: N_COLS blocks
#define ROW_LEN_BYTES (ROW_LEN_INT64 * 8) //Number of bytes per row
int LYRA2(void *K, uint64_t kLen, const void *pwd, uint64_t pwdlen, const void *salt, uint64_t saltlen, uint64_t timeCost, uint64_t nRows, uint64_t nCols);
int LYRA2(void *K, int64_t kLen, const void *pwd, int32_t pwdlen, const void *salt, int32_t saltlen, int64_t timeCost, const int16_t nRows, const int16_t nCols);
#endif /* LYRA2_H_ */
+243 -588
View File
@@ -25,7 +25,6 @@
#include "Lyra2.h"
/**
* Initializes the Sponge State. The first 512 bits are set to zeros and the remainder
* receive Blake2b's IV as per Blake2b's specification. <b>Note:</b> Even though sponges
@@ -37,20 +36,18 @@
*
* @param state The 1024-bit array to be initialized
*/
void initState(uint64_t state[/*16*/]) {
//First 512 bis are zeros
memset(state, 0, 64);
//Remainder BLOCK_LEN_BLAKE2_SAFE_BYTES are reserved to the IV
state[8] = blake2b_IV[0];
state[9] = blake2b_IV[1];
state[10] = blake2b_IV[2];
state[11] = blake2b_IV[3];
state[12] = blake2b_IV[4];
state[13] = blake2b_IV[5];
state[14] = blake2b_IV[6];
state[15] = blake2b_IV[7];
void initState(uint64_t state[/*16*/]) {
//First 512 bis are zeros
memset(state, 0, 64);
//Remainder BLOCK_LEN_BLAKE2_SAFE_BYTES are reserved to the IV
state[8] = blake2b_IV[0];
state[9] = blake2b_IV[1];
state[10] = blake2b_IV[2];
state[11] = blake2b_IV[3];
state[12] = blake2b_IV[4];
state[13] = blake2b_IV[5];
state[14] = blake2b_IV[6];
state[15] = blake2b_IV[7];
}
/**
@@ -59,18 +56,18 @@
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's G function
*/
__inline static void blake2bLyra(uint64_t *v) {
ROUND_LYRA(0);
ROUND_LYRA(1);
ROUND_LYRA(2);
ROUND_LYRA(3);
ROUND_LYRA(4);
ROUND_LYRA(5);
ROUND_LYRA(6);
ROUND_LYRA(7);
ROUND_LYRA(8);
ROUND_LYRA(9);
ROUND_LYRA(10);
ROUND_LYRA(11);
ROUND_LYRA(0);
ROUND_LYRA(1);
ROUND_LYRA(2);
ROUND_LYRA(3);
ROUND_LYRA(4);
ROUND_LYRA(5);
ROUND_LYRA(6);
ROUND_LYRA(7);
ROUND_LYRA(8);
ROUND_LYRA(9);
ROUND_LYRA(10);
ROUND_LYRA(11);
}
/**
@@ -78,7 +75,7 @@ __inline static void blake2bLyra(uint64_t *v) {
* @param v A 1024-bit (16 uint64_t) array to be processed by Blake2b's G function
*/
__inline static void reducedBlake2bLyra(uint64_t *v) {
ROUND_LYRA(0);
ROUND_LYRA(0);
}
/**
@@ -89,19 +86,20 @@ __inline static void reducedBlake2bLyra(uint64_t *v) {
* @param out Array that will receive the data squeezed
* @param len The number of bytes to be squeezed into the "out" array
*/
void squeeze(uint64_t *state, byte *out, unsigned int len) {
int fullBlocks = len / BLOCK_LEN_BYTES;
byte *ptr = out;
int i;
//Squeezes full blocks
for (i = 0; i < fullBlocks; i++) {
memcpy(ptr, state, BLOCK_LEN_BYTES);
blake2bLyra(state);
ptr += BLOCK_LEN_BYTES;
}
void squeeze(uint64_t *state, byte *out, unsigned int len)
{
int fullBlocks = len / BLOCK_LEN_BYTES;
byte *ptr = out;
int i;
//Squeezes full blocks
for (i = 0; i < fullBlocks; i++) {
memcpy(ptr, state, BLOCK_LEN_BYTES);
blake2bLyra(state);
ptr += BLOCK_LEN_BYTES;
}
//Squeezes remaining bytes
memcpy(ptr, state, (len % BLOCK_LEN_BYTES));
//Squeezes remaining bytes
memcpy(ptr, state, (len % BLOCK_LEN_BYTES));
}
/**
@@ -111,23 +109,24 @@ __inline static void reducedBlake2bLyra(uint64_t *v) {
* @param state The current state of the sponge
* @param in The block to be absorbed (BLOCK_LEN_INT64 words)
*/
void absorbBlock(uint64_t *state, const uint64_t *in) {
//XORs the first BLOCK_LEN_INT64 words of "in" with the current state
state[0] ^= in[0];
state[1] ^= in[1];
state[2] ^= in[2];
state[3] ^= in[3];
state[4] ^= in[4];
state[5] ^= in[5];
state[6] ^= in[6];
state[7] ^= in[7];
state[8] ^= in[8];
state[9] ^= in[9];
state[10] ^= in[10];
state[11] ^= in[11];
void absorbBlock(uint64_t *state, const uint64_t *in)
{
//XORs the first BLOCK_LEN_INT64 words of "in" with the current state
state[0] ^= in[0];
state[1] ^= in[1];
state[2] ^= in[2];
state[3] ^= in[3];
state[4] ^= in[4];
state[5] ^= in[5];
state[6] ^= in[6];
state[7] ^= in[7];
state[8] ^= in[8];
state[9] ^= in[9];
state[10] ^= in[10];
state[11] ^= in[11];
//Applies the transformation f to the sponge's state
blake2bLyra(state);
//Applies the transformation f to the sponge's state
blake2bLyra(state);
}
/**
@@ -137,25 +136,21 @@ void absorbBlock(uint64_t *state, const uint64_t *in) {
* @param state The current state of the sponge
* @param in The block to be absorbed (BLOCK_LEN_BLAKE2_SAFE_INT64 words)
*/
void absorbBlockBlake2Safe(uint64_t *state, const uint64_t *in) {
//XORs the first BLOCK_LEN_BLAKE2_SAFE_INT64 words of "in" with the current state
state[0] ^= in[0];
state[1] ^= in[1];
state[2] ^= in[2];
state[3] ^= in[3];
state[4] ^= in[4];
state[5] ^= in[5];
state[6] ^= in[6];
state[7] ^= in[7];
void absorbBlockBlake2Safe(uint64_t *state, const uint64_t *in)
{
//XORs the first BLOCK_LEN_BLAKE2_SAFE_INT64 words of "in" with the current state
//Applies the transformation f to the sponge's state
blake2bLyra(state);
/*
for(int i = 0; i<16; i++) {
printf(" final state %d %08x %08x in %08x %08x\n", i, (uint32_t)(state[i] & 0xFFFFFFFFULL), (uint32_t)(state[i] >> 32),
(uint32_t)(in[i] & 0xFFFFFFFFULL), (uint32_t)(in[i] >> 32));
}
*/
state[0] ^= in[0];
state[1] ^= in[1];
state[2] ^= in[2];
state[3] ^= in[3];
state[4] ^= in[4];
state[5] ^= in[5];
state[6] ^= in[6];
state[7] ^= in[7];
//Applies the transformation f to the sponge's state
blake2bLyra(state);
}
/**
@@ -166,36 +161,31 @@ void absorbBlockBlake2Safe(uint64_t *state, const uint64_t *in) {
* @param state The current state of the sponge
* @param rowOut Row to receive the data squeezed
*/
void reducedSqueezeRow0(uint64_t* state, uint64_t* rowOut) {
uint64_t* ptrWord = rowOut + (N_COLS-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to M[0][C-1]
int i;
//M[row][C-1-col] = H.reduced_squeeze()
for (i = 0; i < N_COLS; i++) {
void reducedSqueezeRow0(uint64_t* state, uint64_t* rowOut, const uint32_t nCols)
{
uint64_t* ptrWord = rowOut + (nCols-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to M[0][C-1]
unsigned int i;
//M[row][C-1-col] = H.reduced_squeeze()
for (i = 0; i < nCols; i++) {
ptrWord[0] = state[0];
ptrWord[1] = state[1];
ptrWord[2] = state[2];
ptrWord[3] = state[3];
ptrWord[4] = state[4];
ptrWord[5] = state[5];
ptrWord[6] = state[6];
ptrWord[7] = state[7];
ptrWord[8] = state[8];
ptrWord[9] = state[9];
ptrWord[10] = state[10];
ptrWord[11] = state[11];
ptrWord[0] = state[0];
ptrWord[1] = state[1];
ptrWord[2] = state[2];
ptrWord[3] = state[3];
ptrWord[4] = state[4];
ptrWord[5] = state[5];
ptrWord[6] = state[6];
ptrWord[7] = state[7];
ptrWord[8] = state[8];
ptrWord[9] = state[9];
ptrWord[10] = state[10];
ptrWord[11] = state[11];
/*
for (int i = 0; i<12; i++) {
printf(" after reducedSqueezeRow0 %d %08x %08x in %08x %08x\n", i, (uint32_t)(ptrWord[i] & 0xFFFFFFFFULL), (uint32_t)(ptrWord[i] >> 32),
(uint32_t)(state[i] & 0xFFFFFFFFULL), (uint32_t)(state[i] >> 32));
//Goes to next block (column) that will receive the squeezed data
ptrWord -= BLOCK_LEN_INT64;
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
}
*/
//Goes to next block (column) that will receive the squeezed data
ptrWord -= BLOCK_LEN_INT64;
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
}
}
/**
@@ -207,50 +197,50 @@ for (int i = 0; i<12; i++) {
* @param rowIn Row to feed the sponge
* @param rowOut Row to receive the sponge's output
*/
void reducedDuplexRow1(uint64_t *state, uint64_t *rowIn, uint64_t *rowOut) {
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordOut = rowOut + (N_COLS-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to row
int i;
void reducedDuplexRow1(uint64_t *state, uint64_t *rowIn, uint64_t *rowOut, const uint32_t nCols)
{
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordOut = rowOut + (nCols-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to row
unsigned int i;
for (i = 0; i < N_COLS; i++) {
for (i = 0; i < nCols; i++) {
//Absorbing "M[prev][col]"
state[0] ^= (ptrWordIn[0]);
state[1] ^= (ptrWordIn[1]);
state[2] ^= (ptrWordIn[2]);
state[3] ^= (ptrWordIn[3]);
state[4] ^= (ptrWordIn[4]);
state[5] ^= (ptrWordIn[5]);
state[6] ^= (ptrWordIn[6]);
state[7] ^= (ptrWordIn[7]);
state[8] ^= (ptrWordIn[8]);
state[9] ^= (ptrWordIn[9]);
state[10] ^= (ptrWordIn[10]);
state[11] ^= (ptrWordIn[11]);
//Absorbing "M[prev][col]"
state[0] ^= (ptrWordIn[0]);
state[1] ^= (ptrWordIn[1]);
state[2] ^= (ptrWordIn[2]);
state[3] ^= (ptrWordIn[3]);
state[4] ^= (ptrWordIn[4]);
state[5] ^= (ptrWordIn[5]);
state[6] ^= (ptrWordIn[6]);
state[7] ^= (ptrWordIn[7]);
state[8] ^= (ptrWordIn[8]);
state[9] ^= (ptrWordIn[9]);
state[10] ^= (ptrWordIn[10]);
state[11] ^= (ptrWordIn[11]);
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[row][C-1-col] = M[prev][col] XOR rand
ptrWordOut[0] = ptrWordIn[0] ^ state[0];
ptrWordOut[1] = ptrWordIn[1] ^ state[1];
ptrWordOut[2] = ptrWordIn[2] ^ state[2];
ptrWordOut[3] = ptrWordIn[3] ^ state[3];
ptrWordOut[4] = ptrWordIn[4] ^ state[4];
ptrWordOut[5] = ptrWordIn[5] ^ state[5];
ptrWordOut[6] = ptrWordIn[6] ^ state[6];
ptrWordOut[7] = ptrWordIn[7] ^ state[7];
ptrWordOut[8] = ptrWordIn[8] ^ state[8];
ptrWordOut[9] = ptrWordIn[9] ^ state[9];
ptrWordOut[10] = ptrWordIn[10] ^ state[10];
ptrWordOut[11] = ptrWordIn[11] ^ state[11];
//M[row][C-1-col] = M[prev][col] XOR rand
ptrWordOut[0] = ptrWordIn[0] ^ state[0];
ptrWordOut[1] = ptrWordIn[1] ^ state[1];
ptrWordOut[2] = ptrWordIn[2] ^ state[2];
ptrWordOut[3] = ptrWordIn[3] ^ state[3];
ptrWordOut[4] = ptrWordIn[4] ^ state[4];
ptrWordOut[5] = ptrWordIn[5] ^ state[5];
ptrWordOut[6] = ptrWordIn[6] ^ state[6];
ptrWordOut[7] = ptrWordIn[7] ^ state[7];
ptrWordOut[8] = ptrWordIn[8] ^ state[8];
ptrWordOut[9] = ptrWordIn[9] ^ state[9];
ptrWordOut[10] = ptrWordIn[10] ^ state[10];
ptrWordOut[11] = ptrWordIn[11] ^ state[11];
//Input: next column (i.e., next block in sequence)
ptrWordIn += BLOCK_LEN_INT64;
//Output: goes to previous column
ptrWordOut -= BLOCK_LEN_INT64;
}
//Input: next column (i.e., next block in sequence)
ptrWordIn += BLOCK_LEN_INT64;
//Output: goes to previous column
ptrWordOut -= BLOCK_LEN_INT64;
}
}
/**
@@ -267,63 +257,66 @@ for (int i = 0; i<12; i++) {
* @param rowOut Row receiving the output
*
*/
void reducedDuplexRowSetup(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut) {
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordOut = rowOut + (N_COLS-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to row
int i;
for (i = 0; i < N_COLS; i++) {
//Absorbing "M[prev] [+] M[row*]"
state[0] ^= (ptrWordIn[0] + ptrWordInOut[0]);
state[1] ^= (ptrWordIn[1] + ptrWordInOut[1]);
state[2] ^= (ptrWordIn[2] + ptrWordInOut[2]);
state[3] ^= (ptrWordIn[3] + ptrWordInOut[3]);
state[4] ^= (ptrWordIn[4] + ptrWordInOut[4]);
state[5] ^= (ptrWordIn[5] + ptrWordInOut[5]);
state[6] ^= (ptrWordIn[6] + ptrWordInOut[6]);
state[7] ^= (ptrWordIn[7] + ptrWordInOut[7]);
state[8] ^= (ptrWordIn[8] + ptrWordInOut[8]);
state[9] ^= (ptrWordIn[9] + ptrWordInOut[9]);
state[10] ^= (ptrWordIn[10] + ptrWordInOut[10]);
state[11] ^= (ptrWordIn[11] + ptrWordInOut[11]);
void reducedDuplexRowSetup(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut, const uint32_t nCols)
{
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordOut = rowOut + (nCols-1)*BLOCK_LEN_INT64; //In Lyra2: pointer to row
unsigned int i;
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
for (i = 0; i < nCols; i++) {
//M[row][col] = M[prev][col] XOR rand
ptrWordOut[0] = ptrWordIn[0] ^ state[0];
ptrWordOut[1] = ptrWordIn[1] ^ state[1];
ptrWordOut[2] = ptrWordIn[2] ^ state[2];
ptrWordOut[3] = ptrWordIn[3] ^ state[3];
ptrWordOut[4] = ptrWordIn[4] ^ state[4];
ptrWordOut[5] = ptrWordIn[5] ^ state[5];
ptrWordOut[6] = ptrWordIn[6] ^ state[6];
ptrWordOut[7] = ptrWordIn[7] ^ state[7];
ptrWordOut[8] = ptrWordIn[8] ^ state[8];
ptrWordOut[9] = ptrWordIn[9] ^ state[9];
ptrWordOut[10] = ptrWordIn[10] ^ state[10];
ptrWordOut[11] = ptrWordIn[11] ^ state[11];
//Absorbing "M[prev] [+] M[row*]"
state[0] ^= (ptrWordIn[0] + ptrWordInOut[0]);
state[1] ^= (ptrWordIn[1] + ptrWordInOut[1]);
state[2] ^= (ptrWordIn[2] + ptrWordInOut[2]);
state[3] ^= (ptrWordIn[3] + ptrWordInOut[3]);
state[4] ^= (ptrWordIn[4] + ptrWordInOut[4]);
state[5] ^= (ptrWordIn[5] + ptrWordInOut[5]);
state[6] ^= (ptrWordIn[6] + ptrWordInOut[6]);
state[7] ^= (ptrWordIn[7] + ptrWordInOut[7]);
state[8] ^= (ptrWordIn[8] + ptrWordInOut[8]);
state[9] ^= (ptrWordIn[9] + ptrWordInOut[9]);
state[10] ^= (ptrWordIn[10] + ptrWordInOut[10]);
state[11] ^= (ptrWordIn[11] + ptrWordInOut[11]);
//M[row*][col] = M[row*][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[11];
ptrWordInOut[1] ^= state[0];
ptrWordInOut[2] ^= state[1];
ptrWordInOut[3] ^= state[2];
ptrWordInOut[4] ^= state[3];
ptrWordInOut[5] ^= state[4];
ptrWordInOut[6] ^= state[5];
ptrWordInOut[7] ^= state[6];
ptrWordInOut[8] ^= state[7];
ptrWordInOut[9] ^= state[8];
ptrWordInOut[10] ^= state[9];
ptrWordInOut[11] ^= state[10];
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//Inputs: next column (i.e., next block in sequence)
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
//Output: goes to previous column
ptrWordOut -= BLOCK_LEN_INT64;
}
//M[row][col] = M[prev][col] XOR rand
ptrWordOut[0] = ptrWordIn[0] ^ state[0];
ptrWordOut[1] = ptrWordIn[1] ^ state[1];
ptrWordOut[2] = ptrWordIn[2] ^ state[2];
ptrWordOut[3] = ptrWordIn[3] ^ state[3];
ptrWordOut[4] = ptrWordIn[4] ^ state[4];
ptrWordOut[5] = ptrWordIn[5] ^ state[5];
ptrWordOut[6] = ptrWordIn[6] ^ state[6];
ptrWordOut[7] = ptrWordIn[7] ^ state[7];
ptrWordOut[8] = ptrWordIn[8] ^ state[8];
ptrWordOut[9] = ptrWordIn[9] ^ state[9];
ptrWordOut[10] = ptrWordIn[10] ^ state[10];
ptrWordOut[11] = ptrWordIn[11] ^ state[11];
//M[row*][col] = M[row*][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[11];
ptrWordInOut[1] ^= state[0];
ptrWordInOut[2] ^= state[1];
ptrWordInOut[3] ^= state[2];
ptrWordInOut[4] ^= state[3];
ptrWordInOut[5] ^= state[4];
ptrWordInOut[6] ^= state[5];
ptrWordInOut[7] ^= state[6];
ptrWordInOut[8] ^= state[7];
ptrWordInOut[9] ^= state[8];
ptrWordInOut[10] ^= state[9];
ptrWordInOut[11] ^= state[10];
//Inputs: next column (i.e., next block in sequence)
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
//Output: goes to previous column
ptrWordOut -= BLOCK_LEN_INT64;
}
}
/**
@@ -340,410 +333,72 @@ for (int i = 0; i<12; i++) {
* @param rowOut Row receiving the output
*
*/
void reducedDuplexRow(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut) {
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordOut = rowOut; //In Lyra2: pointer to row
int i;
void reducedDuplexRow(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut, const uint32_t nCols)
{
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordOut = rowOut; //In Lyra2: pointer to row
unsigned int i;
for (i = 0; i < N_COLS; i++) {
for (i = 0; i < nCols; i++) {
//Absorbing "M[prev] [+] M[row*]"
state[0] ^= (ptrWordIn[0] + ptrWordInOut[0]);
state[1] ^= (ptrWordIn[1] + ptrWordInOut[1]);
state[2] ^= (ptrWordIn[2] + ptrWordInOut[2]);
state[3] ^= (ptrWordIn[3] + ptrWordInOut[3]);
state[4] ^= (ptrWordIn[4] + ptrWordInOut[4]);
state[5] ^= (ptrWordIn[5] + ptrWordInOut[5]);
state[6] ^= (ptrWordIn[6] + ptrWordInOut[6]);
state[7] ^= (ptrWordIn[7] + ptrWordInOut[7]);
state[8] ^= (ptrWordIn[8] + ptrWordInOut[8]);
state[9] ^= (ptrWordIn[9] + ptrWordInOut[9]);
state[10] ^= (ptrWordIn[10] + ptrWordInOut[10]);
state[11] ^= (ptrWordIn[11] + ptrWordInOut[11]);
//Absorbing "M[prev] [+] M[row*]"
state[0] ^= (ptrWordIn[0] + ptrWordInOut[0]);
state[1] ^= (ptrWordIn[1] + ptrWordInOut[1]);
state[2] ^= (ptrWordIn[2] + ptrWordInOut[2]);
state[3] ^= (ptrWordIn[3] + ptrWordInOut[3]);
state[4] ^= (ptrWordIn[4] + ptrWordInOut[4]);
state[5] ^= (ptrWordIn[5] + ptrWordInOut[5]);
state[6] ^= (ptrWordIn[6] + ptrWordInOut[6]);
state[7] ^= (ptrWordIn[7] + ptrWordInOut[7]);
state[8] ^= (ptrWordIn[8] + ptrWordInOut[8]);
state[9] ^= (ptrWordIn[9] + ptrWordInOut[9]);
state[10] ^= (ptrWordIn[10] + ptrWordInOut[10]);
state[11] ^= (ptrWordIn[11] + ptrWordInOut[11]);
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[rowOut][col] = M[rowOut][col] XOR rand
ptrWordOut[0] ^= state[0];
ptrWordOut[1] ^= state[1];
ptrWordOut[2] ^= state[2];
ptrWordOut[3] ^= state[3];
ptrWordOut[4] ^= state[4];
ptrWordOut[5] ^= state[5];
ptrWordOut[6] ^= state[6];
ptrWordOut[7] ^= state[7];
ptrWordOut[8] ^= state[8];
ptrWordOut[9] ^= state[9];
ptrWordOut[10] ^= state[10];
ptrWordOut[11] ^= state[11];
//M[rowOut][col] = M[rowOut][col] XOR rand
ptrWordOut[0] ^= state[0];
ptrWordOut[1] ^= state[1];
ptrWordOut[2] ^= state[2];
ptrWordOut[3] ^= state[3];
ptrWordOut[4] ^= state[4];
ptrWordOut[5] ^= state[5];
ptrWordOut[6] ^= state[6];
ptrWordOut[7] ^= state[7];
ptrWordOut[8] ^= state[8];
ptrWordOut[9] ^= state[9];
ptrWordOut[10] ^= state[10];
ptrWordOut[11] ^= state[11];
//M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[11];
ptrWordInOut[1] ^= state[0];
ptrWordInOut[2] ^= state[1];
ptrWordInOut[3] ^= state[2];
ptrWordInOut[4] ^= state[3];
ptrWordInOut[5] ^= state[4];
ptrWordInOut[6] ^= state[5];
ptrWordInOut[7] ^= state[6];
ptrWordInOut[8] ^= state[7];
ptrWordInOut[9] ^= state[8];
ptrWordInOut[10] ^= state[9];
ptrWordInOut[11] ^= state[10];
//M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[11];
ptrWordInOut[1] ^= state[0];
ptrWordInOut[2] ^= state[1];
ptrWordInOut[3] ^= state[2];
ptrWordInOut[4] ^= state[3];
ptrWordInOut[5] ^= state[4];
ptrWordInOut[6] ^= state[5];
ptrWordInOut[7] ^= state[6];
ptrWordInOut[8] ^= state[7];
ptrWordInOut[9] ^= state[8];
ptrWordInOut[10] ^= state[9];
ptrWordInOut[11] ^= state[10];
//Goes to next block
ptrWordOut += BLOCK_LEN_INT64;
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
}
//Goes to next block
ptrWordOut += BLOCK_LEN_INT64;
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
}
}
////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
/**
* Performs a duplex operation over "M[rowInOut] [+] M[rowIn]", writing the output "rand"
* on M[rowOut] and making "M[rowInOut] = M[rowInOut] XOR rotW(rand)", where rotW is a 64-bit
* rotation to the left.
*
* @param state The current state of the sponge
* @param rowIn Row used only as input
* @param rowInOut Row used as input and to receive output after rotation
* @param rowOut Row receiving the output
*
* Prints an array of unsigned chars
*/
/*
inline void reducedDuplexRowSetupOLD(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut) {
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordOut = rowOut; //In Lyra2: pointer to row
int i;
for (i = 0; i < N_COLS; i++) {
//Absorbing "M[rowInOut] XOR M[rowIn]"
state[0] ^= ptrWordInOut[0] ^ ptrWordIn[0];
state[1] ^= ptrWordInOut[1] ^ ptrWordIn[1];
state[2] ^= ptrWordInOut[2] ^ ptrWordIn[2];
state[3] ^= ptrWordInOut[3] ^ ptrWordIn[3];
state[4] ^= ptrWordInOut[4] ^ ptrWordIn[4];
state[5] ^= ptrWordInOut[5] ^ ptrWordIn[5];
state[6] ^= ptrWordInOut[6] ^ ptrWordIn[6];
state[7] ^= ptrWordInOut[7] ^ ptrWordIn[7];
state[8] ^= ptrWordInOut[8] ^ ptrWordIn[8];
state[9] ^= ptrWordInOut[9] ^ ptrWordIn[9];
state[10] ^= ptrWordInOut[10] ^ ptrWordIn[10];
state[11] ^= ptrWordInOut[11] ^ ptrWordIn[11];
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[row][col] = rand
ptrWordOut[0] = state[0];
ptrWordOut[1] = state[1];
ptrWordOut[2] = state[2];
ptrWordOut[3] = state[3];
ptrWordOut[4] = state[4];
ptrWordOut[5] = state[5];
ptrWordOut[6] = state[6];
ptrWordOut[7] = state[7];
ptrWordOut[8] = state[8];
ptrWordOut[9] = state[9];
ptrWordOut[10] = state[10];
ptrWordOut[11] = state[11];
//M[row*][col] = M[row*][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[10];
ptrWordInOut[1] ^= state[11];
ptrWordInOut[2] ^= state[0];
ptrWordInOut[3] ^= state[1];
ptrWordInOut[4] ^= state[2];
ptrWordInOut[5] ^= state[3];
ptrWordInOut[6] ^= state[4];
ptrWordInOut[7] ^= state[5];
ptrWordInOut[8] ^= state[6];
ptrWordInOut[9] ^= state[7];
ptrWordInOut[10] ^= state[8];
ptrWordInOut[11] ^= state[9];
//Goes to next column (i.e., next block in sequence)
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
ptrWordOut += BLOCK_LEN_INT64;
}
}
*/
/**
* Performs a duplex operation over "M[rowInOut] XOR M[rowIn]", writing the output "rand"
* on M[rowOut] and making "M[rowInOut] = M[rowInOut] XOR rotW(rand)", where rotW is a 64-bit
* rotation to the left.
*
* @param state The current state of the sponge
* @param rowIn Row used only as input
* @param rowInOut Row used as input and to receive output after rotation
* @param rowOut Row receiving the output
*
*/
/*
inline void reducedDuplexRowSetupv5(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut) {
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordOut = rowOut; //In Lyra2: pointer to row
int i;
for (i = 0; i < N_COLS; i++) {
//Absorbing "M[rowInOut] XOR M[rowIn]"
state[0] ^= ptrWordInOut[0] + ptrWordIn[0];
state[1] ^= ptrWordInOut[1] + ptrWordIn[1];
state[2] ^= ptrWordInOut[2] + ptrWordIn[2];
state[3] ^= ptrWordInOut[3] + ptrWordIn[3];
state[4] ^= ptrWordInOut[4] + ptrWordIn[4];
state[5] ^= ptrWordInOut[5] + ptrWordIn[5];
state[6] ^= ptrWordInOut[6] + ptrWordIn[6];
state[7] ^= ptrWordInOut[7] + ptrWordIn[7];
state[8] ^= ptrWordInOut[8] + ptrWordIn[8];
state[9] ^= ptrWordInOut[9] + ptrWordIn[9];
state[10] ^= ptrWordInOut[10] + ptrWordIn[10];
state[11] ^= ptrWordInOut[11] + ptrWordIn[11];
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[row*][col] = M[row*][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[10];
ptrWordInOut[1] ^= state[11];
ptrWordInOut[2] ^= state[0];
ptrWordInOut[3] ^= state[1];
ptrWordInOut[4] ^= state[2];
ptrWordInOut[5] ^= state[3];
ptrWordInOut[6] ^= state[4];
ptrWordInOut[7] ^= state[5];
ptrWordInOut[8] ^= state[6];
ptrWordInOut[9] ^= state[7];
ptrWordInOut[10] ^= state[8];
ptrWordInOut[11] ^= state[9];
//M[row][col] = rand
ptrWordOut[0] = state[0] ^ ptrWordIn[0];
ptrWordOut[1] = state[1] ^ ptrWordIn[1];
ptrWordOut[2] = state[2] ^ ptrWordIn[2];
ptrWordOut[3] = state[3] ^ ptrWordIn[3];
ptrWordOut[4] = state[4] ^ ptrWordIn[4];
ptrWordOut[5] = state[5] ^ ptrWordIn[5];
ptrWordOut[6] = state[6] ^ ptrWordIn[6];
ptrWordOut[7] = state[7] ^ ptrWordIn[7];
ptrWordOut[8] = state[8] ^ ptrWordIn[8];
ptrWordOut[9] = state[9] ^ ptrWordIn[9];
ptrWordOut[10] = state[10] ^ ptrWordIn[10];
ptrWordOut[11] = state[11] ^ ptrWordIn[11];
//Goes to next column (i.e., next block in sequence)
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
ptrWordOut += BLOCK_LEN_INT64;
}
}
*/
/**
* Performs a duplex operation over "M[rowInOut] XOR M[rowIn]", writing the output "rand"
* on M[rowOut] and making "M[rowInOut] = M[rowInOut] XOR rotW(rand)", where rotW is a 64-bit
* rotation to the left.
*
* @param state The current state of the sponge
* @param rowIn Row used only as input
* @param rowInOut Row used as input and to receive output after rotation
* @param rowOut Row receiving the output
*
*/
/*
inline void reducedDuplexRowSetupv5c(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut) {
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordOut = rowOut;
int i;
for (i = 0; i < N_COLS / 2; i++) {
//Absorbing "M[rowInOut] XOR M[rowIn]"
state[0] ^= ptrWordInOut[0] + ptrWordIn[0];
state[1] ^= ptrWordInOut[1] + ptrWordIn[1];
state[2] ^= ptrWordInOut[2] + ptrWordIn[2];
state[3] ^= ptrWordInOut[3] + ptrWordIn[3];
state[4] ^= ptrWordInOut[4] + ptrWordIn[4];
state[5] ^= ptrWordInOut[5] + ptrWordIn[5];
state[6] ^= ptrWordInOut[6] + ptrWordIn[6];
state[7] ^= ptrWordInOut[7] + ptrWordIn[7];
state[8] ^= ptrWordInOut[8] + ptrWordIn[8];
state[9] ^= ptrWordInOut[9] + ptrWordIn[9];
state[10] ^= ptrWordInOut[10] + ptrWordIn[10];
state[11] ^= ptrWordInOut[11] + ptrWordIn[11];
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[row*][col] = M[row*][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[10];
ptrWordInOut[1] ^= state[11];
ptrWordInOut[2] ^= state[0];
ptrWordInOut[3] ^= state[1];
ptrWordInOut[4] ^= state[2];
ptrWordInOut[5] ^= state[3];
ptrWordInOut[6] ^= state[4];
ptrWordInOut[7] ^= state[5];
ptrWordInOut[8] ^= state[6];
ptrWordInOut[9] ^= state[7];
ptrWordInOut[10] ^= state[8];
ptrWordInOut[11] ^= state[9];
//M[row][col] = rand
ptrWordOut[0] = state[0] ^ ptrWordIn[0];
ptrWordOut[1] = state[1] ^ ptrWordIn[1];
ptrWordOut[2] = state[2] ^ ptrWordIn[2];
ptrWordOut[3] = state[3] ^ ptrWordIn[3];
ptrWordOut[4] = state[4] ^ ptrWordIn[4];
ptrWordOut[5] = state[5] ^ ptrWordIn[5];
ptrWordOut[6] = state[6] ^ ptrWordIn[6];
ptrWordOut[7] = state[7] ^ ptrWordIn[7];
ptrWordOut[8] = state[8] ^ ptrWordIn[8];
ptrWordOut[9] = state[9] ^ ptrWordIn[9];
ptrWordOut[10] = state[10] ^ ptrWordIn[10];
ptrWordOut[11] = state[11] ^ ptrWordIn[11];
//Goes to next column (i.e., next block in sequence)
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
ptrWordOut += 2 * BLOCK_LEN_INT64;
}
ptrWordOut = rowOut + BLOCK_LEN_INT64;
for (i = 0; i < N_COLS / 2; i++) {
//Absorbing "M[rowInOut] XOR M[rowIn]"
state[0] ^= ptrWordInOut[0] + ptrWordIn[0];
state[1] ^= ptrWordInOut[1] + ptrWordIn[1];
state[2] ^= ptrWordInOut[2] + ptrWordIn[2];
state[3] ^= ptrWordInOut[3] + ptrWordIn[3];
state[4] ^= ptrWordInOut[4] + ptrWordIn[4];
state[5] ^= ptrWordInOut[5] + ptrWordIn[5];
state[6] ^= ptrWordInOut[6] + ptrWordIn[6];
state[7] ^= ptrWordInOut[7] + ptrWordIn[7];
state[8] ^= ptrWordInOut[8] + ptrWordIn[8];
state[9] ^= ptrWordInOut[9] + ptrWordIn[9];
state[10] ^= ptrWordInOut[10] + ptrWordIn[10];
state[11] ^= ptrWordInOut[11] + ptrWordIn[11];
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[row*][col] = M[row*][col] XOR rotW(rand)
ptrWordInOut[0] ^= state[10];
ptrWordInOut[1] ^= state[11];
ptrWordInOut[2] ^= state[0];
ptrWordInOut[3] ^= state[1];
ptrWordInOut[4] ^= state[2];
ptrWordInOut[5] ^= state[3];
ptrWordInOut[6] ^= state[4];
ptrWordInOut[7] ^= state[5];
ptrWordInOut[8] ^= state[6];
ptrWordInOut[9] ^= state[7];
ptrWordInOut[10] ^= state[8];
ptrWordInOut[11] ^= state[9];
//M[row][col] = rand
ptrWordOut[0] = state[0] ^ ptrWordIn[0];
ptrWordOut[1] = state[1] ^ ptrWordIn[1];
ptrWordOut[2] = state[2] ^ ptrWordIn[2];
ptrWordOut[3] = state[3] ^ ptrWordIn[3];
ptrWordOut[4] = state[4] ^ ptrWordIn[4];
ptrWordOut[5] = state[5] ^ ptrWordIn[5];
ptrWordOut[6] = state[6] ^ ptrWordIn[6];
ptrWordOut[7] = state[7] ^ ptrWordIn[7];
ptrWordOut[8] = state[8] ^ ptrWordIn[8];
ptrWordOut[9] = state[9] ^ ptrWordIn[9];
ptrWordOut[10] = state[10] ^ ptrWordIn[10];
ptrWordOut[11] = state[11] ^ ptrWordIn[11];
//Goes to next column (i.e., next block in sequence)
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
ptrWordOut += 2 * BLOCK_LEN_INT64;
}
}
*/
/**
* Performs a duplex operation over "M[rowInOut] XOR M[rowIn]", using the output "rand"
* to make "M[rowOut][col] = M[rowOut][col] XOR rand" and "M[rowInOut] = M[rowInOut] XOR rotW(rand)",
* where rotW is a 64-bit rotation to the left.
*
* @param state The current state of the sponge
* @param rowIn Row used only as input
* @param rowInOut Row used as input and to receive output after rotation
* @param rowOut Row receiving the output
*
*/
/*
inline void reducedDuplexRowd(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut) {
uint64_t* ptrWordInOut = rowInOut; //In Lyra2: pointer to row*
uint64_t* ptrWordIn = rowIn; //In Lyra2: pointer to prev
uint64_t* ptrWordOut = rowOut; //In Lyra2: pointer to row
int i;
for (i = 0; i < N_COLS; i++) {
//Absorbing "M[rowInOut] XOR M[rowIn]"
state[0] ^= ptrWordInOut[0] + ptrWordIn[0];
state[1] ^= ptrWordInOut[1] + ptrWordIn[1];
state[2] ^= ptrWordInOut[2] + ptrWordIn[2];
state[3] ^= ptrWordInOut[3] + ptrWordIn[3];
state[4] ^= ptrWordInOut[4] + ptrWordIn[4];
state[5] ^= ptrWordInOut[5] + ptrWordIn[5];
state[6] ^= ptrWordInOut[6] + ptrWordIn[6];
state[7] ^= ptrWordInOut[7] + ptrWordIn[7];
state[8] ^= ptrWordInOut[8] + ptrWordIn[8];
state[9] ^= ptrWordInOut[9] + ptrWordIn[9];
state[10] ^= ptrWordInOut[10] + ptrWordIn[10];
state[11] ^= ptrWordInOut[11] + ptrWordIn[11];
//Applies the reduced-round transformation f to the sponge's state
reducedBlake2bLyra(state);
//M[rowOut][col] = M[rowOut][col] XOR rand
ptrWordOut[0] ^= state[0];
ptrWordOut[1] ^= state[1];
ptrWordOut[2] ^= state[2];
ptrWordOut[3] ^= state[3];
ptrWordOut[4] ^= state[4];
ptrWordOut[5] ^= state[5];
ptrWordOut[6] ^= state[6];
ptrWordOut[7] ^= state[7];
ptrWordOut[8] ^= state[8];
ptrWordOut[9] ^= state[9];
ptrWordOut[10] ^= state[10];
ptrWordOut[11] ^= state[11];
//M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
//Goes to next block
ptrWordOut += BLOCK_LEN_INT64;
ptrWordInOut += BLOCK_LEN_INT64;
ptrWordIn += BLOCK_LEN_INT64;
}
}
*/
/**
Prints an array of unsigned chars
*/
void printArray(unsigned char *array, unsigned int size, char *name) {
void printArray(unsigned char *array, unsigned int size, char *name)
{
unsigned int i;
printf("%s: ", name);
for (i = 0; i < size; i++) {
+35 -55
View File
@@ -24,85 +24,65 @@
#include <stdint.h>
#if defined(__GNUC__)
#define ALIGN __attribute__ ((aligned(32)))
#elif defined(_MSC_VER)
#define ALIGN __declspec(align(32))
#else
#define ALIGN
#endif
/*Blake2b IV Array*/
/* Blake2b IV Array */
static const uint64_t blake2b_IV[8] =
{
0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
0x6a09e667f3bcc908ULL, 0xbb67ae8584caa73bULL,
0x3c6ef372fe94f82bULL, 0xa54ff53a5f1d36f1ULL,
0x510e527fade682d1ULL, 0x9b05688c2b3e6c1fULL,
0x1f83d9abfb41bd6bULL, 0x5be0cd19137e2179ULL
};
/*Blake2b's rotation*/
static __inline uint64_t rotr64( const uint64_t w, const unsigned c ){
return ( w >> c ) | ( w << ( 64 - c ) );
/* Blake2b's rotation */
static __inline uint64_t rotr64(const uint64_t w, const unsigned c) {
#ifdef _MSC_VER
return _rotr64(w, c);
#else
return ( w >> c ) | ( w << ( 64 - c ) );
#endif
}
/*Blake2b's G function*/
#define G(r,i,a,b,c,d) \
do { \
a = a + b; \
d = rotr64(d ^ a, 32); \
c = c + d; \
b = rotr64(b ^ c, 24); \
a = a + b; \
d = rotr64(d ^ a, 16); \
c = c + d; \
b = rotr64(b ^ c, 63); \
/* Blake2b's G function */
#define G(r,i,a,b,c,d) do { \
a = a + b; \
d = rotr64(d ^ a, 32); \
c = c + d; \
b = rotr64(b ^ c, 24); \
a = a + b; \
d = rotr64(d ^ a, 16); \
c = c + d; \
b = rotr64(b ^ c, 63); \
} while(0)
/*One Round of the Blake2b's compression function*/
#define ROUND_LYRA(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]);
#define ROUND_LYRA(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]);
//---- Housekeeping
void initState(uint64_t state[/*16*/]);
//---- Squeezes
void squeeze(uint64_t *state, unsigned char *out, unsigned int len);
void reducedSqueezeRow0(uint64_t* state, uint64_t* row);
void reducedSqueezeRow0(uint64_t* state, uint64_t* row, const uint32_t nCols);
//---- Absorbs
void absorbBlock(uint64_t *state, const uint64_t *in);
void absorbBlockBlake2Safe(uint64_t *state, const uint64_t *in);
//---- Duplexes
void reducedDuplexRow1(uint64_t *state, uint64_t *rowIn, uint64_t *rowOut);
void reducedDuplexRowSetup(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
void reducedDuplexRow(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
void reducedDuplexRow1(uint64_t *state, uint64_t *rowIn, uint64_t *rowOut, const uint32_t nCols);
void reducedDuplexRowSetup(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut, const uint32_t nCols);
void reducedDuplexRow(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut, const uint32_t nCols);
//---- Misc
void printArray(unsigned char *array, unsigned int size, char *name);
////////////////////////////////////////////////////////////////////////////////////////////////
////TESTS////
//void reducedDuplexRowc(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
//void reducedDuplexRowd(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
//void reducedDuplexRowSetupv4(uint64_t *state, uint64_t *rowIn1, uint64_t *rowIn2, uint64_t *rowOut1, uint64_t *rowOut2);
//void reducedDuplexRowSetupv5(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
//void reducedDuplexRowSetupv5c(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
//void reducedDuplexRowSetupv5d(uint64_t *state, uint64_t *rowIn, uint64_t *rowInOut, uint64_t *rowOut);
/////////////
#endif /* SPONGE_H_ */
+582
View File
@@ -0,0 +1,582 @@
/* DJM CRAP to strip (again) made for SM 3.2+ */
#ifndef CUDA_LYRA_VECTOR_H
#define CUDA_LYRA_VECTOR_H
///////////////////////////////////////////////////////////////////////////////////
#if (defined(_MSC_VER) && defined(_WIN64)) || defined(__LP64__)
#define __LDG_PTR "l"
#else
#define __LDG_PTR "r"
#endif
#include "cuda_helper.h"
#if __CUDA_ARCH__ < 320 && !defined(__ldg4)
#define __ldg4(x) (*(x))
#endif
typedef struct __align__(32) uint8 {
unsigned int s0, s1, s2, s3, s4, s5, s6, s7;
} uint8;
typedef struct __align__(64) uint2_8 {
uint2 s0, s1, s2, s3, s4, s5, s6, s7;
} uint2_8;
typedef struct __align__(64) ulonglong2to8 {
ulonglong2 l0,l1,l2,l3;
} ulonglong2to8;
typedef struct __align__(128) ulonglong8to16 {
ulonglong2to8 lo, hi;
} ulonglong8to16;
typedef struct __align__(256) ulonglong16to32 {
ulonglong8to16 lo, hi;
} ulonglong16to32;
typedef struct __align__(512) ulonglong32to64 {
ulonglong16to32 lo, hi;
} ulonglong32to64;
typedef struct __align__(128) ulonglonglong {
ulonglong2 s0,s1,s2,s3,s4,s5,s6,s7;
} ulonglonglong;
typedef struct __align__(64) uint16 {
union {
struct {unsigned int s0, s1, s2, s3, s4, s5, s6, s7;};
uint8 lo;
};
union {
struct {unsigned int s8, s9, sa, sb, sc, sd, se, sf;};
uint8 hi;
};
} uint16;
typedef struct __align__(128) uint2_16 {
union {
struct { uint2 s0, s1, s2, s3, s4, s5, s6, s7; };
uint2_8 lo;
};
union {
struct { uint2 s8, s9, sa, sb, sc, sd, se, sf; };
uint2_8 hi;
};
} uint2_16;
typedef struct __align__(128) uint32 {
uint16 lo,hi;
} uint32;
struct __align__(128) ulong8 {
ulonglong4 s0, s1, s2, s3;
};
typedef __device_builtin__ struct ulong8 ulong8;
typedef struct __align__(256) ulonglong16 {
ulonglong4 s0, s1, s2, s3, s4, s5, s6, s7;
} ulonglong16;
typedef struct __align__(16) uint28 {
uint2 x, y, z, w;
} uint28;
typedef struct __builtin_align__(32) uint48 {
uint4 s0,s1;
} uint48;
typedef struct __align__(256) uint4x16 {
uint4 s0, s1, s2, s3, s4, s5, s6, s7, s8, s9, s10, s11, s12, s13, s14, s15;
} uint4x16;
static __inline__ __device__ ulonglong2to8 make_ulonglong2to8(ulonglong2 s0, ulonglong2 s1, ulonglong2 s2, ulonglong2 s3)
{
ulonglong2to8 t; t.l0=s0; t.l1=s1; t.l2=s2; t.l3=s3;
return t;
}
static __inline__ __device__ ulonglong8to16 make_ulonglong8to16(const ulonglong2to8 &s0, const ulonglong2to8 &s1)
{
ulonglong8to16 t; t.lo = s0; t.hi = s1;
return t;
}
static __inline__ __device__ ulonglong16to32 make_ulonglong16to32(const ulonglong8to16 &s0, const ulonglong8to16 &s1)
{
ulonglong16to32 t; t.lo = s0; t.hi = s1;
return t;
}
static __inline__ __device__ ulonglong32to64 make_ulonglong32to64(const ulonglong16to32 &s0, const ulonglong16to32 &s1)
{
ulonglong32to64 t; t.lo = s0; t.hi = s1;
return t;
}
static __inline__ __host__ __device__ ulonglonglong make_ulonglonglong(
const ulonglong2 &s0, const ulonglong2 &s1, const ulonglong2 &s2, const ulonglong2 &s3,
const ulonglong2 &s4, const ulonglong2 &s5)
{
ulonglonglong t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5;
return t;
}
static __inline__ __device__ uint48 make_uint48(uint4 s0, uint4 s1)
{
uint48 t; t.s0 = s0; t.s1 = s1;
return t;
}
static __inline__ __device__ uint28 make_uint28(uint2 s0, uint2 s1, uint2 s2, uint2 s3)
{
uint28 t; t.x = s0; t.y = s1; t.z = s2; t.w = s3;
return t;
}
static __inline__ __host__ __device__ uint4x16 make_uint4x16(
uint4 s0, uint4 s1, uint4 s2, uint4 s3, uint4 s4, uint4 s5, uint4 s6, uint4 s7,
uint4 s8, uint4 s9, uint4 sa, uint4 sb, uint4 sc, uint4 sd, uint4 se, uint4 sf)
{
uint4x16 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
t.s8 = s8; t.s9 = s9; t.s10 = sa; t.s11 = sb; t.s12 = sc; t.s13 = sd; t.s14 = se; t.s15 = sf;
return t;
}
static __inline__ __device__ uint2_16 make_uint2_16(
uint2 s0, uint2 s1, uint2 s2, uint2 s3, uint2 s4, uint2 s5, uint2 s6, uint2 s7,
uint2 s8, uint2 s9, uint2 sa, uint2 sb, uint2 sc, uint2 sd, uint2 se, uint2 sf)
{
uint2_16 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
t.s8 = s8; t.s9 = s9; t.sa = sa; t.sb = sb; t.sc = sc; t.sd = sd; t.se = se; t.sf = sf;
return t;
}
static __inline__ __host__ __device__ uint16 make_uint16(
unsigned int s0, unsigned int s1, unsigned int s2, unsigned int s3, unsigned int s4, unsigned int s5, unsigned int s6, unsigned int s7,
unsigned int s8, unsigned int s9, unsigned int sa, unsigned int sb, unsigned int sc, unsigned int sd, unsigned int se, unsigned int sf)
{
uint16 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
t.s8 = s8; t.s9 = s9; t.sa = sa; t.sb = sb; t.sc = sc; t.sd = sd; t.se = se; t.sf = sf;
return t;
}
static __inline__ __host__ __device__ uint16 make_uint16(const uint8 &a, const uint8 &b)
{
uint16 t; t.lo=a; t.hi=b; return t;
}
static __inline__ __host__ __device__ uint32 make_uint32(const uint16 &a, const uint16 &b)
{
uint32 t; t.lo = a; t.hi = b; return t;
}
static __inline__ __host__ __device__ uint8 make_uint8(
unsigned int s0, unsigned int s1, unsigned int s2, unsigned int s3, unsigned int s4, unsigned int s5, unsigned int s6, unsigned int s7)
{
uint8 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
return t;
}
static __inline__ __host__ __device__ uint2_8 make_uint2_8(
uint2 s0, uint2 s1, uint2 s2, uint2 s3, uint2 s4, uint2 s5, uint2 s6, uint2 s7)
{
uint2_8 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
return t;
}
static __inline__ __host__ __device__ ulonglong16 make_ulonglong16(const ulonglong4 &s0, const ulonglong4 &s1,
const ulonglong4 &s2, const ulonglong4 &s3, const ulonglong4 &s4, const ulonglong4 &s5, const ulonglong4 &s6, const ulonglong4 &s7)
{
ulonglong16 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3; t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
return t;
}
static __inline__ __host__ __device__ ulong8 make_ulong8(
ulonglong4 s0, ulonglong4 s1, ulonglong4 s2, ulonglong4 s3)
{
ulong8 t; t.s0 = s0; t.s1 = s1; t.s2 = s2; t.s3 = s3;// t.s4 = s4; t.s5 = s5; t.s6 = s6; t.s7 = s7;
return t;
}
static __forceinline__ __device__ uchar4 operator^ (uchar4 a, uchar4 b) { return make_uchar4(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w); }
static __forceinline__ __device__ uchar4 operator+ (uchar4 a, uchar4 b) { return make_uchar4(a.x + b.x, a.y + b.y, a.z + b.z, a.w + b.w); }
static __forceinline__ __device__ uint4 operator+ (uint4 a, uint4 b) { return make_uint4(a.x + b.x, a.y + b.y, a.z + b.z, a.w + b.w); }
static __forceinline__ __device__ ulonglong4 operator^ (ulonglong4 a, ulonglong4 b) { return make_ulonglong4(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w); }
static __forceinline__ __device__ ulonglong4 operator+ (ulonglong4 a, ulonglong4 b) { return make_ulonglong4(a.x + b.x, a.y + b.y, a.z + b.z, a.w + b.w); }
static __forceinline__ __device__ ulonglong2 operator^ (ulonglong2 a, ulonglong2 b) { return make_ulonglong2(a.x ^ b.x, a.y ^ b.y); }
static __forceinline__ __device__ ulonglong2 operator+ (ulonglong2 a, ulonglong2 b) { return make_ulonglong2(a.x + b.x, a.y + b.y); }
static __forceinline__ __device__ ulong8 operator^ (const ulong8 &a, const ulong8 &b) {
return make_ulong8(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3);
}
static __forceinline__ __device__ ulong8 operator+ (const ulong8 &a, const ulong8 &b) {
return make_ulong8(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3);
}
static __forceinline__ __device__ __host__ uint8 operator^ (const uint8 &a, const uint8 &b) { return make_uint8(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3, a.s4 ^ b.s4, a.s5 ^ b.s5, a.s6 ^ b.s6, a.s7 ^ b.s7); }
static __forceinline__ __device__ __host__ uint8 operator+ (const uint8 &a, const uint8 &b) { return make_uint8(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3, a.s4 + b.s4, a.s5 + b.s5, a.s6 + b.s6, a.s7 + b.s7); }
static __forceinline__ __device__ uint2_8 operator^ (const uint2_8 &a, const uint2_8 &b) { return make_uint2_8(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3, a.s4 ^ b.s4, a.s5 ^ b.s5, a.s6 ^ b.s6, a.s7 ^ b.s7); }
static __forceinline__ __device__ uint2_8 operator+ (const uint2_8 &a, const uint2_8 &b) { return make_uint2_8(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3, a.s4 + b.s4, a.s5 + b.s5, a.s6 + b.s6, a.s7 + b.s7); }
////////////// mess++ //////
static __forceinline__ __device__ uint28 operator^ (const uint28 &a, const uint28 &b) {
return make_uint28(a.x ^ b.x, a.y ^ b.y, a.z ^ b.z, a.w ^ b.w);
}
static __forceinline__ __device__ uint28 operator+ (const uint28 &a, const uint28 &b) {
return make_uint28(a.x + b.x, a.y + b.y, a.z + b.z, a.w + b.w);
}
static __forceinline__ __device__ uint48 operator+ (const uint48 &a, const uint48 &b) {
return make_uint48(a.s0 + b.s0, a.s1 + b.s1);
}
/////////////////////////
static __forceinline__ __device__ __host__ uint16 operator^ (const uint16 &a, const uint16 &b) {
return make_uint16(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3, a.s4 ^ b.s4, a.s5 ^ b.s5, a.s6 ^ b.s6, a.s7 ^ b.s7,
a.s8 ^ b.s8, a.s9 ^ b.s9, a.sa ^ b.sa, a.sb ^ b.sb, a.sc ^ b.sc, a.sd ^ b.sd, a.se ^ b.se, a.sf ^ b.sf);
}
static __forceinline__ __device__ __host__ uint16 operator+ (const uint16 &a, const uint16 &b) {
return make_uint16(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3, a.s4 + b.s4, a.s5 + b.s5, a.s6 + b.s6, a.s7 + b.s7,
a.s8 + b.s8, a.s9 + b.s9, a.sa + b.sa, a.sb + b.sb, a.sc + b.sc, a.sd + b.sd, a.se + b.se, a.sf + b.sf);
}
static __forceinline__ __device__ uint2_16 operator^ (const uint2_16 &a, const uint2_16 &b) {
return make_uint2_16(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3, a.s4 ^ b.s4, a.s5 ^ b.s5, a.s6 ^ b.s6, a.s7 ^ b.s7,
a.s8 ^ b.s8, a.s9 ^ b.s9, a.sa ^ b.sa, a.sb ^ b.sb, a.sc ^ b.sc, a.sd ^ b.sd, a.se ^ b.se, a.sf ^ b.sf);
}
static __forceinline__ __device__ uint2_16 operator+ (const uint2_16 &a, const uint2_16 &b) {
return make_uint2_16(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3, a.s4 + b.s4, a.s5 + b.s5, a.s6 + b.s6, a.s7 + b.s7,
a.s8 + b.s8, a.s9 + b.s9, a.sa + b.sa, a.sb + b.sb, a.sc + b.sc, a.sd + b.sd, a.se + b.se, a.sf + b.sf);
}
static __forceinline__ __device__ uint32 operator^ (const uint32 &a, const uint32 &b) {
return make_uint32(a.lo ^ b.lo, a.hi ^ b.hi);
}
static __forceinline__ __device__ uint32 operator+ (const uint32 &a, const uint32 &b) {
return make_uint32(a.lo + b.lo, a.hi + b.hi);
}
static __forceinline__ __device__ ulonglong16 operator^ (const ulonglong16 &a, const ulonglong16 &b) {
return make_ulonglong16(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3, a.s4 ^ b.s4, a.s5 ^ b.s5, a.s6 ^ b.s6, a.s7 ^ b.s7);
}
static __forceinline__ __device__ ulonglong16 operator+ (const ulonglong16 &a, const ulonglong16 &b) {
return make_ulonglong16(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3, a.s4 + b.s4, a.s5 + b.s5, a.s6 + b.s6, a.s7 + b.s7);
}
static __forceinline__ __device__ void operator^= (ulong8 &a, const ulong8 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator^= (uint28 &a, const uint28 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (uint28 &a, const uint28 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (uint2_8 &a, const uint2_8 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (uint2_8 &a, const uint2_8 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (uint32 &a, const uint32 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (uint32 &a, const uint32 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (uchar4 &a, uchar4 b) { a = a ^ b; }
static __forceinline__ __device__ __host__ void operator^= (uint8 &a, const uint8 &b) { a = a ^ b; }
static __forceinline__ __device__ __host__ void operator^= (uint16 &a, const uint16 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator^= (ulonglong16 &a, const ulonglong16 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator^= (ulonglong4 &a, const ulonglong4 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (ulonglong4 &a, const ulonglong4 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (ulonglong2 &a, const ulonglong2 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (ulonglong2 &a, const ulonglong2 &b) { a = a + b; }
static __forceinline__ __device__
ulonglong2to8 operator^ (const ulonglong2to8 &a, const ulonglong2to8 &b)
{
return make_ulonglong2to8(a.l0 ^ b.l0, a.l1 ^ b.l1, a.l2 ^ b.l2, a.l3 ^ b.l3);
}
static __forceinline__ __device__
ulonglong2to8 operator+ (const ulonglong2to8 &a, const ulonglong2to8 &b)
{
return make_ulonglong2to8(a.l0 + b.l0, a.l1 + b.l1, a.l2 + b.l2, a.l3 + b.l3);
}
static __forceinline__ __device__
ulonglong8to16 operator^ (const ulonglong8to16 &a, const ulonglong8to16 &b)
{
return make_ulonglong8to16(a.lo ^ b.lo, a.hi ^ b.hi);
}
static __forceinline__ __device__
ulonglong8to16 operator+ (const ulonglong8to16 &a, const ulonglong8to16 &b)
{
return make_ulonglong8to16(a.lo + b.lo, a.hi + b.hi);
}
static __forceinline__ __device__
ulonglong16to32 operator^ (const ulonglong16to32 &a, const ulonglong16to32 &b)
{
return make_ulonglong16to32(a.lo ^ b.lo, a.hi ^ b.hi);
}
static __forceinline__ __device__
ulonglong16to32 operator+ (const ulonglong16to32 &a, const ulonglong16to32 &b)
{
return make_ulonglong16to32(a.lo + b.lo, a.hi + b.hi);
}
static __forceinline__ __device__
ulonglong32to64 operator^ (const ulonglong32to64 &a, const ulonglong32to64 &b)
{
return make_ulonglong32to64(a.lo ^ b.lo, a.hi ^ b.hi);
}
static __forceinline__ __device__
ulonglong32to64 operator+ (const ulonglong32to64 &a, const ulonglong32to64 &b)
{
return make_ulonglong32to64(a.lo + b.lo, a.hi + b.hi);
}
static __forceinline__ __device__ ulonglonglong operator^ (const ulonglonglong &a, const ulonglonglong &b) {
return make_ulonglonglong(a.s0 ^ b.s0, a.s1 ^ b.s1, a.s2 ^ b.s2, a.s3 ^ b.s3, a.s4 ^ b.s4, a.s5 ^ b.s5);
}
static __forceinline__ __device__ ulonglonglong operator+ (const ulonglonglong &a, const ulonglonglong &b) {
return make_ulonglonglong(a.s0 + b.s0, a.s1 + b.s1, a.s2 + b.s2, a.s3 + b.s3, a.s4 + b.s4, a.s5 + b.s5);
}
static __forceinline__ __device__ void operator^= (ulonglong2to8 &a, const ulonglong2to8 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (uint4 &a, uint4 b) { a = a + b; }
static __forceinline__ __device__ void operator+= (uchar4 &a, uchar4 b) { a = a + b; }
static __forceinline__ __device__ __host__ void operator+= (uint8 &a, const uint8 &b) { a = a + b; }
static __forceinline__ __device__ __host__ void operator+= (uint16 &a, const uint16 &b) { a = a + b; }
static __forceinline__ __device__ void operator+= (uint2_16 &a, const uint2_16 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (uint2_16 &a, const uint2_16 &b) { a = a + b; }
static __forceinline__ __device__ void operator+= (ulong8 &a, const ulong8 &b) { a = a + b; }
static __forceinline__ __device__ void operator+= (ulonglong16 &a, const ulonglong16 &b) { a = a + b; }
static __forceinline__ __device__ void operator+= (ulonglong8to16 &a, const ulonglong8to16 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (ulonglong8to16 &a, const ulonglong8to16 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (ulonglong16to32 &a, const ulonglong16to32 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (ulonglong16to32 &a, const ulonglong16to32 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (ulonglong32to64 &a, const ulonglong32to64 &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (ulonglong32to64 &a, const ulonglong32to64 &b) { a = a ^ b; }
static __forceinline__ __device__ void operator+= (ulonglonglong &a, const ulonglonglong &b) { a = a + b; }
static __forceinline__ __device__ void operator^= (ulonglonglong &a, const ulonglonglong &b) { a = a ^ b; }
#if __CUDA_ARCH__ < 320
#define rotate ROTL32
#define rotateR ROTR32
#else
static __forceinline__ __device__ uint4 rotate4(uint4 vec4, uint32_t shift)
{
uint4 ret;
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.x) : "r"(vec4.x), "r"(vec4.x), "r"(shift));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.y) : "r"(vec4.y), "r"(vec4.y), "r"(shift));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.z) : "r"(vec4.z), "r"(vec4.z), "r"(shift));
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.w) : "r"(vec4.w), "r"(vec4.w), "r"(shift));
return ret;
}
static __forceinline__ __device__ uint4 rotate4R(uint4 vec4, uint32_t shift)
{
uint4 ret;
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.x) : "r"(vec4.x), "r"(vec4.x), "r"(shift));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.y) : "r"(vec4.y), "r"(vec4.y), "r"(shift));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.z) : "r"(vec4.z), "r"(vec4.z), "r"(shift));
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(ret.w) : "r"(vec4.w), "r"(vec4.w), "r"(shift));
return ret;
}
static __forceinline__ __device__ uint32_t rotate(uint32_t vec4, uint32_t shift)
{
uint32_t ret;
asm("shf.l.wrap.b32 %0, %1, %2, %3;" : "=r"(ret) : "r"(vec4), "r"(vec4), "r"(shift));
return ret;
}
static __forceinline__ __device__ uint32_t rotateR(uint32_t vec4, uint32_t shift)
{
uint32_t ret;
asm("shf.r.wrap.b32 %0, %1, %2, %3;" : "=r"(ret) : "r"(vec4), "r"(vec4), "r"(shift));
return ret;
}
static __device__ __inline__ ulonglong4 __ldg4(const ulonglong4 *ptr)
{
ulonglong4 ret;
asm("ld.global.nc.v2.u64 {%0,%1}, [%2];" : "=l"(ret.x), "=l"(ret.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v2.u64 {%0,%1}, [%2+16];" : "=l"(ret.z), "=l"(ret.w) : __LDG_PTR(ptr));
return ret;
}
static __device__ __inline__ void ldg4(const ulonglong4 *ptr,ulonglong4 *ret)
{
asm("ld.global.nc.v2.u64 {%0,%1}, [%2];" : "=l"(ret[0].x), "=l"(ret[0].y) : __LDG_PTR(ptr));
asm("ld.global.nc.v2.u64 {%0,%1}, [%2+16];" : "=l"(ret[0].z), "=l"(ret[0].w) : __LDG_PTR(ptr));
asm("ld.global.nc.v2.u64 {%0,%1}, [%2+32];" : "=l"(ret[1].x), "=l"(ret[1].y) : __LDG_PTR(ptr));
asm("ld.global.nc.v2.u64 {%0,%1}, [%2+48];" : "=l"(ret[1].z), "=l"(ret[1].w) : __LDG_PTR(ptr));
asm("ld.global.nc.v2.u64 {%0,%1}, [%2+64];" : "=l"(ret[2].x), "=l"(ret[2].y) : __LDG_PTR(ptr));
asm("ld.global.nc.v2.u64 {%0,%1}, [%2+80];" : "=l"(ret[2].z), "=l"(ret[2].w) : __LDG_PTR(ptr));
}
static __device__ __inline__ uint28 __ldg4(const uint28 *ptr)
{
uint28 ret;
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4];" : "=r"(ret.x.x), "=r"(ret.x.y), "=r"(ret.y.x), "=r"(ret.y.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+16];" : "=r"(ret.z.x), "=r"(ret.z.y), "=r"(ret.w.x), "=r"(ret.w.y) : __LDG_PTR(ptr));
return ret;
}
static __device__ __inline__ uint48 __ldg4(const uint48 *ptr)
{
uint48 ret;
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4];" : "=r"(ret.s0.x), "=r"(ret.s0.y), "=r"(ret.s0.z), "=r"(ret.s0.w) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+16];" : "=r"(ret.s1.x), "=r"(ret.s1.y), "=r"(ret.s1.z), "=r"(ret.s1.w) : __LDG_PTR(ptr));
return ret;
}
static __device__ __inline__ void ldg4(const uint28 *ptr, uint28 *ret)
{
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4];" : "=r"(ret[0].x.x), "=r"(ret[0].x.y), "=r"(ret[0].y.x), "=r"(ret[0].y.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+16];" : "=r"(ret[0].z.x), "=r"(ret[0].z.y), "=r"(ret[0].w.x), "=r"(ret[0].w.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+32];" : "=r"(ret[1].x.x), "=r"(ret[1].x.y), "=r"(ret[1].y.x), "=r"(ret[1].y.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+48];" : "=r"(ret[1].z.x), "=r"(ret[1].z.y), "=r"(ret[1].w.x), "=r"(ret[1].w.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+64];" : "=r"(ret[2].x.x), "=r"(ret[2].x.y), "=r"(ret[2].y.x), "=r"(ret[2].y.y) : __LDG_PTR(ptr));
asm("ld.global.nc.v4.u32 {%0,%1,%2,%3}, [%4+80];" : "=r"(ret[2].z.x), "=r"(ret[2].z.y), "=r"(ret[2].w.x), "=r"(ret[2].w.y) : __LDG_PTR(ptr));
}
#endif /* __CUDA_ARCH__ < 320 */
static __forceinline__ __device__ uint8 swapvec(const uint8 &buf)
{
uint8 vec;
vec.s0 = cuda_swab32(buf.s0);
vec.s1 = cuda_swab32(buf.s1);
vec.s2 = cuda_swab32(buf.s2);
vec.s3 = cuda_swab32(buf.s3);
vec.s4 = cuda_swab32(buf.s4);
vec.s5 = cuda_swab32(buf.s5);
vec.s6 = cuda_swab32(buf.s6);
vec.s7 = cuda_swab32(buf.s7);
return vec;
}
static __forceinline__ __device__ uint8 swapvec(const uint8 *buf)
{
uint8 vec;
vec.s0 = cuda_swab32(buf[0].s0);
vec.s1 = cuda_swab32(buf[0].s1);
vec.s2 = cuda_swab32(buf[0].s2);
vec.s3 = cuda_swab32(buf[0].s3);
vec.s4 = cuda_swab32(buf[0].s4);
vec.s5 = cuda_swab32(buf[0].s5);
vec.s6 = cuda_swab32(buf[0].s6);
vec.s7 = cuda_swab32(buf[0].s7);
return vec;
}
static __forceinline__ __device__ uint16 swapvec(const uint16 *buf)
{
uint16 vec;
vec.s0 = cuda_swab32(buf[0].s0);
vec.s1 = cuda_swab32(buf[0].s1);
vec.s2 = cuda_swab32(buf[0].s2);
vec.s3 = cuda_swab32(buf[0].s3);
vec.s4 = cuda_swab32(buf[0].s4);
vec.s5 = cuda_swab32(buf[0].s5);
vec.s6 = cuda_swab32(buf[0].s6);
vec.s7 = cuda_swab32(buf[0].s7);
vec.s8 = cuda_swab32(buf[0].s8);
vec.s9 = cuda_swab32(buf[0].s9);
vec.sa = cuda_swab32(buf[0].sa);
vec.sb = cuda_swab32(buf[0].sb);
vec.sc = cuda_swab32(buf[0].sc);
vec.sd = cuda_swab32(buf[0].sd);
vec.se = cuda_swab32(buf[0].se);
vec.sf = cuda_swab32(buf[0].sf);
return vec;
}
static __forceinline__ __device__ uint16 swapvec(const uint16 &buf)
{
uint16 vec;
vec.s0 = cuda_swab32(buf.s0);
vec.s1 = cuda_swab32(buf.s1);
vec.s2 = cuda_swab32(buf.s2);
vec.s3 = cuda_swab32(buf.s3);
vec.s4 = cuda_swab32(buf.s4);
vec.s5 = cuda_swab32(buf.s5);
vec.s6 = cuda_swab32(buf.s6);
vec.s7 = cuda_swab32(buf.s7);
vec.s8 = cuda_swab32(buf.s8);
vec.s9 = cuda_swab32(buf.s9);
vec.sa = cuda_swab32(buf.sa);
vec.sb = cuda_swab32(buf.sb);
vec.sc = cuda_swab32(buf.sc);
vec.sd = cuda_swab32(buf.sd);
vec.se = cuda_swab32(buf.se);
vec.sf = cuda_swab32(buf.sf);
return vec;
}
static __device__ __forceinline__ uint28 shuffle4(const uint28 &var, int lane)
{
uint28 res;
res.x.x = __shfl(var.x.x, lane);
res.x.y = __shfl(var.x.y, lane);
res.y.x = __shfl(var.y.x, lane);
res.y.y = __shfl(var.y.y, lane);
res.z.x = __shfl(var.z.x, lane);
res.z.y = __shfl(var.z.y, lane);
res.w.x = __shfl(var.w.x, lane);
res.w.y = __shfl(var.w.y, lane);
return res;
}
static __device__ __forceinline__ ulonglong4 shuffle4(ulonglong4 var, int lane)
{
ulonglong4 res;
uint2 temp;
temp = vectorize(var.x);
temp.x = __shfl(temp.x, lane);
temp.y = __shfl(temp.y, lane);
res.x = devectorize(temp);
temp = vectorize(var.y);
temp.x = __shfl(temp.x, lane);
temp.y = __shfl(temp.y, lane);
res.y = devectorize(temp);
temp = vectorize(var.z);
temp.x = __shfl(temp.x, lane);
temp.y = __shfl(temp.y, lane);
res.z = devectorize(temp);
temp = vectorize(var.w);
temp.x = __shfl(temp.x, lane);
temp.y = __shfl(temp.y, lane);
res.w = devectorize(temp);
return res;
}
#endif // #ifndef CUDA_LYRA_VECTOR_H
+532
View File
@@ -0,0 +1,532 @@
#include <stdio.h>
#include <memory.h>
#ifdef __INTELLISENSE__
/* just for vstudio code colors */
#define __CUDA_ARCH__ 500
#endif
#include "cuda_lyra2_vectors.h"
#define TPB 16
#define Nrow 4
#define Ncol 4
#if __CUDA_ARCH__ < 500
#define vectype ulonglong4
#define u64type uint64_t
#define memshift 4
#elif __CUDA_ARCH__ == 500
#define u64type uint2
#define vectype uint28
#define memshift 3
#else
#define u64type uint2
#define vectype uint28
#define memshift 3
#endif
__device__ vectype *DMatrix;
#if __CUDA_ARCH__ >= 300
#if __CUDA_ARCH__ >= 500
static __device__ __forceinline__
void Gfunc_v35(uint2 &a, uint2 &b, uint2 &c, uint2 &d)
{
a += b; d ^= a; d = SWAPUINT2(d);
c += d; b ^= c; b = ROR24(b);
a += b; d ^= a; d = ROR16(d);
c += d; b ^= c; b = ROR2(b, 63);
}
#else
static __device__ __forceinline__
void Gfunc_v35(unsigned long long &a, unsigned long long &b, unsigned long long &c, unsigned long long &d)
{
a += b; d ^= a; d = ROTR64(d, 32);
c += d; b ^= c; b = ROTR64(b, 24);
a += b; d ^= a; d = ROTR64(d, 16);
c += d; b ^= c; b = ROTR64(b, 63);
}
#endif
static __device__ __forceinline__
void round_lyra_v35(vectype* s)
{
Gfunc_v35(s[0].x, s[1].x, s[2].x, s[3].x);
Gfunc_v35(s[0].y, s[1].y, s[2].y, s[3].y);
Gfunc_v35(s[0].z, s[1].z, s[2].z, s[3].z);
Gfunc_v35(s[0].w, s[1].w, s[2].w, s[3].w);
Gfunc_v35(s[0].x, s[1].y, s[2].z, s[3].w);
Gfunc_v35(s[0].y, s[1].z, s[2].w, s[3].x);
Gfunc_v35(s[0].z, s[1].w, s[2].x, s[3].y);
Gfunc_v35(s[0].w, s[1].x, s[2].y, s[3].z);
}
static __device__ __forceinline__
void reduceDuplex(vectype state[4], uint32_t thread)
{
vectype state1[3];
uint32_t ps1 = (Nrow * Ncol * memshift * thread);
uint32_t ps2 = (memshift * (Ncol-1) + memshift * Ncol + Nrow * Ncol * memshift * thread);
#pragma unroll 4
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 + i*memshift;
uint32_t s2 = ps2 - i*memshift;
#pragma unroll
for (int j = 0; j < 3; j++)
state1[j] = __ldg4(&(DMatrix+s1)[j]);
for (int j = 0; j < 3; j++)
state[j] ^= state1[j];
round_lyra_v35(state);
#pragma unroll
for (int j = 0; j < 3; j++)
state1[j] ^= state[j];
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state1[j];
}
}
static __device__ __forceinline__
void reduceDuplexV3(vectype state[4], uint32_t thread)
{
vectype state1[3];
uint32_t ps1 = (Nrow * Ncol * memshift * thread);
uint32_t ps2 = (memshift * (Ncol - 1) * Nrow + memshift * 1 + Nrow * Ncol * memshift * thread);
#pragma unroll 4
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 + Nrow * i *memshift;
uint32_t s2 = ps2 - Nrow * i *memshift;
for (int j = 0; j < 3; j++)
state1[j] = __ldg4(&(DMatrix + s1)[j]);
for (int j = 0; j < 3; j++)
state[j] ^= state1[j];
round_lyra_v35(state);
for (int j = 0; j < 3; j++)
state1[j] ^= state[j];
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state1[j];
}
}
static __device__ __forceinline__
void reduceDuplexRowSetupV2(const int rowIn, const int rowInOut, const int rowOut, vectype state[4], uint32_t thread)
{
vectype state2[3],state1[3];
uint32_t ps1 = (memshift * Ncol * rowIn + Nrow * Ncol * memshift * thread);
uint32_t ps2 = (memshift * Ncol * rowInOut + Nrow * Ncol * memshift * thread);
uint32_t ps3 = (memshift * (Ncol-1) + memshift * Ncol * rowOut + Nrow * Ncol * memshift * thread);
//#pragma unroll 1
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 + i*memshift;
uint32_t s2 = ps2 + i*memshift;
uint32_t s3 = ps3 - i*memshift;
for (int j = 0; j < 3; j++)
state1[j]= __ldg4(&(DMatrix + s1)[j]);
for (int j = 0; j < 3; j++)
state2[j]= __ldg4(&(DMatrix + s2)[j]);
for (int j = 0; j < 3; j++) {
vectype tmp = state1[j] + state2[j];
state[j] ^= tmp;
}
round_lyra_v35(state);
for (int j = 0; j < 3; j++) {
state1[j] ^= state[j];
(DMatrix + s3)[j] = state1[j];
}
((uint2*)state2)[0] ^= ((uint2*)state)[11];
for (int j = 0; j < 11; j++)
((uint2*)state2)[j+1] ^= ((uint2*)state)[j];
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state2[j];
}
}
static __device__ __forceinline__
void reduceDuplexRowSetupV3(const int rowIn, const int rowInOut, const int rowOut, vectype state[4], uint32_t thread)
{
vectype state2[3], state1[3];
uint32_t ps1 = (memshift * rowIn + Nrow * Ncol * memshift * thread);
uint32_t ps2 = (memshift * rowInOut + Nrow * Ncol * memshift * thread);
uint32_t ps3 = (Nrow * memshift * (Ncol - 1) + memshift * rowOut + Nrow * Ncol * memshift * thread);
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 + Nrow*i*memshift;
uint32_t s2 = ps2 + Nrow*i*memshift;
uint32_t s3 = ps3 - Nrow*i*memshift;
for (int j = 0; j < 3; j++)
state1[j] = __ldg4(&(DMatrix + s1 )[j]);
for (int j = 0; j < 3; j++)
state2[j] = __ldg4(&(DMatrix + s2 )[j]);
for (int j = 0; j < 3; j++) {
vectype tmp = state1[j] + state2[j];
state[j] ^= tmp;
}
round_lyra_v35(state);
for (int j = 0; j < 3; j++) {
state1[j] ^= state[j];
(DMatrix + s3)[j] = state1[j];
}
((uint2*)state2)[0] ^= ((uint2*)state)[11];
for (int j = 0; j < 11; j++)
((uint2*)state2)[j + 1] ^= ((uint2*)state)[j];
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state2[j];
}
}
static __device__ __forceinline__
void reduceDuplexRowtV2(const int rowIn, const int rowInOut, const int rowOut, vectype* state, uint32_t thread)
{
vectype state1[3],state2[3];
uint32_t ps1 = (memshift * Ncol * rowIn + Nrow * Ncol * memshift * thread);
uint32_t ps2 = (memshift * Ncol * rowInOut + Nrow * Ncol * memshift * thread);
uint32_t ps3 = (memshift * Ncol * rowOut + Nrow * Ncol * memshift * thread);
//#pragma unroll 1
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 + i*memshift;
uint32_t s2 = ps2 + i*memshift;
uint32_t s3 = ps3 + i*memshift;
for (int j = 0; j < 3; j++)
state1[j] = __ldg4(&(DMatrix + s1)[j]);
for (int j = 0; j < 3; j++)
state2[j] = __ldg4(&(DMatrix + s2)[j]);
for (int j = 0; j < 3; j++)
state1[j] += state2[j];
for (int j = 0; j < 3; j++)
state[j] ^= state1[j];
round_lyra_v35(state);
((uint2*)state2)[0] ^= ((uint2*)state)[11];
for (int j = 0; j < 11; j++)
((uint2*)state2)[j + 1] ^= ((uint2*)state)[j];
if (rowInOut != rowOut) {
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state2[j];
for (int j = 0; j < 3; j++)
(DMatrix + s3)[j] ^= state[j];
} else {
for (int j = 0; j < 3; j++)
state2[j] ^= state[j];
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j]=state2[j];
}
}
}
static __device__ __forceinline__
void reduceDuplexRowtV3(const int rowIn, const int rowInOut, const int rowOut, vectype* state, uint32_t thread)
{
vectype state1[3], state2[3];
uint32_t ps1 = (memshift * rowIn + Nrow * Ncol * memshift * thread);
uint32_t ps2 = (memshift * rowInOut + Nrow * Ncol * memshift * thread);
uint32_t ps3 = (memshift * rowOut + Nrow * Ncol * memshift * thread);
#pragma nounroll
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 + Nrow * i*memshift;
uint32_t s2 = ps2 + Nrow * i*memshift;
uint32_t s3 = ps3 + Nrow * i*memshift;
for (int j = 0; j < 3; j++)
state1[j] = __ldg4(&(DMatrix + s1)[j]);
for (int j = 0; j < 3; j++)
state2[j] = __ldg4(&(DMatrix + s2)[j]);
for (int j = 0; j < 3; j++)
state1[j] += state2[j];
for (int j = 0; j < 3; j++)
state[j] ^= state1[j];
round_lyra_v35(state);
((uint2*)state2)[0] ^= ((uint2*)state)[11];
for (int j = 0; j < 11; j++)
((uint2*)state2)[j + 1] ^= ((uint2*)state)[j];
if (rowInOut != rowOut) {
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state2[j];
for (int j = 0; j < 3; j++)
(DMatrix + s3)[j] ^= state[j];
} else {
for (int j = 0; j < 3; j++)
state2[j] ^= state[j];
for (int j = 0; j < 3; j++)
(DMatrix + s2)[j] = state2[j];
}
}
}
#if __CUDA_ARCH__ < 500
__global__ __launch_bounds__(128, 1)
#elif __CUDA_ARCH__ == 500
__global__ __launch_bounds__(16, 1)
#else
__global__ __launch_bounds__(TPB, 1)
#endif
void lyra2v2_gpu_hash_32_v3(uint32_t threads, uint32_t startNounce, uint2 *outputHash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
vectype state[4];
uint28 blake2b_IV[2];
uint28 padding[2];
if (threadIdx.x == 0) {
((uint16*)blake2b_IV)[0] = make_uint16(
0xf3bcc908, 0x6a09e667 , 0x84caa73b, 0xbb67ae85 ,
0xfe94f82b, 0x3c6ef372 , 0x5f1d36f1, 0xa54ff53a ,
0xade682d1, 0x510e527f , 0x2b3e6c1f, 0x9b05688c ,
0xfb41bd6b, 0x1f83d9ab , 0x137e2179, 0x5be0cd19
);
((uint16*)padding)[0] = make_uint16(
0x20, 0x0 , 0x20, 0x0 , 0x20, 0x0 , 0x01, 0x0 ,
0x04, 0x0 , 0x04, 0x0 , 0x80, 0x0 , 0x0, 0x01000000
);
}
#if __CUDA_ARCH__ <= 350
if (thread < threads)
#endif
{
((uint2*)state)[0] = __ldg(&outputHash[thread]);
((uint2*)state)[1] = __ldg(&outputHash[thread + threads]);
((uint2*)state)[2] = __ldg(&outputHash[thread + 2 * threads]);
((uint2*)state)[3] = __ldg(&outputHash[thread + 3 * threads]);
state[1] = state[0];
state[2] = shuffle4(((vectype*)blake2b_IV)[0], 0);
state[3] = shuffle4(((vectype*)blake2b_IV)[1], 0);
for (int i = 0; i<12; i++)
round_lyra_v35(state);
state[0] ^= shuffle4(((vectype*)padding)[0], 0);
state[1] ^= shuffle4(((vectype*)padding)[1], 0);
for (int i = 0; i<12; i++)
round_lyra_v35(state);
uint32_t ps1 = (4 * memshift * 3 + 16 * memshift * thread);
//#pragma unroll 4
for (int i = 0; i < 4; i++)
{
uint32_t s1 = ps1 - 4 * memshift * i;
for (int j = 0; j < 3; j++)
(DMatrix + s1)[j] = (state)[j];
round_lyra_v35(state);
}
reduceDuplexV3(state, thread);
reduceDuplexRowSetupV3(1, 0, 2, state, thread);
reduceDuplexRowSetupV3(2, 1, 3, state, thread);
uint32_t rowa;
int prev = 3;
for (int i = 0; i < 4; i++)
{
rowa = ((uint2*)state)[0].x & 3; reduceDuplexRowtV3(prev, rowa, i, state, thread);
prev = i;
}
uint32_t shift = (memshift * rowa + 16 * memshift * thread);
for (int j = 0; j < 3; j++)
state[j] ^= __ldg4(&(DMatrix + shift)[j]);
for (int i = 0; i < 12; i++)
round_lyra_v35(state);
outputHash[thread] = ((uint2*)state)[0];
outputHash[thread + threads] = ((uint2*)state)[1];
outputHash[thread + 2 * threads] = ((uint2*)state)[2];
outputHash[thread + 3 * threads] = ((uint2*)state)[3];
//((vectype*)outputHash)[thread] = state[0];
} //thread
}
#if __CUDA_ARCH__ < 500
__global__ __launch_bounds__(64, 1)
#elif __CUDA_ARCH__ == 500
__global__ __launch_bounds__(32, 1)
#else
__global__ __launch_bounds__(TPB, 1)
#endif
void lyra2v2_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *outputHash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
vectype state[4];
uint28 blake2b_IV[2];
uint28 padding[2];
if (threadIdx.x == 0) {
((uint16*)blake2b_IV)[0] = make_uint16(
0xf3bcc908, 0x6a09e667, 0x84caa73b, 0xbb67ae85,
0xfe94f82b, 0x3c6ef372, 0x5f1d36f1, 0xa54ff53a,
0xade682d1, 0x510e527f, 0x2b3e6c1f, 0x9b05688c,
0xfb41bd6b, 0x1f83d9ab, 0x137e2179, 0x5be0cd19
);
((uint16*)padding)[0] = make_uint16(
0x20, 0x0, 0x20, 0x0, 0x20, 0x0, 0x01, 0x0,
0x04, 0x0, 0x04, 0x0, 0x80, 0x0, 0x0, 0x01000000
);
}
#if __CUDA_ARCH__ <= 350
if (thread < threads)
#endif
{
((uint2*)state)[0] = __ldg(&outputHash[thread]);
((uint2*)state)[1] = __ldg(&outputHash[thread + threads]);
((uint2*)state)[2] = __ldg(&outputHash[thread + 2 * threads]);
((uint2*)state)[3] = __ldg(&outputHash[thread + 3 * threads]);
state[1] = state[0];
state[2] = shuffle4(((vectype*)blake2b_IV)[0], 0);
state[3] = shuffle4(((vectype*)blake2b_IV)[1], 0);
for (int i = 0; i<12; i++)
round_lyra_v35(state);
state[0] ^= shuffle4(((vectype*)padding)[0], 0);
state[1] ^= shuffle4(((vectype*)padding)[1], 0);
for (int i = 0; i<12; i++)
round_lyra_v35(state);
uint32_t ps1 = (memshift * (Ncol - 1) + Nrow * Ncol * memshift * thread);
for (int i = 0; i < Ncol; i++)
{
uint32_t s1 = ps1 - memshift * i;
for (int j = 0; j < 3; j++)
(DMatrix + s1)[j] = (state)[j];
round_lyra_v35(state);
}
reduceDuplex(state, thread);
reduceDuplexRowSetupV2(1, 0, 2, state, thread);
reduceDuplexRowSetupV2(2, 1, 3, state, thread);
uint32_t rowa;
int prev=3;
for (int i = 0; i < 4; i++) {
rowa = ((uint2*)state)[0].x & 3;
reduceDuplexRowtV2(prev, rowa, i, state, thread);
prev=i;
}
uint32_t shift = (memshift * Ncol * rowa + Nrow * Ncol * memshift * thread);
for (int j = 0; j < 3; j++)
state[j] ^= __ldg4(&(DMatrix + shift)[j]);
for (int i = 0; i < 12; i++)
round_lyra_v35(state);
outputHash[thread] = ((uint2*)state)[0];
outputHash[thread + threads] = ((uint2*)state)[1];
outputHash[thread + 2 * threads] = ((uint2*)state)[2];
outputHash[thread + 3 * threads] = ((uint2*)state)[3];
}
}
#else
/* if __CUDA_ARCH__ < 300 .. */
__global__ void lyra2v2_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint2 *outputHash) {}
__global__ void lyra2v2_gpu_hash_32_v3(uint32_t threads, uint32_t startNounce, uint2 *outputHash) {}
#endif
__host__
void lyra2v2_cpu_init(int thr_id, uint32_t threads, uint64_t *d_hash2)
{
// just assign the device pointer allocated in main loop
cudaMemcpyToSymbol(DMatrix, &d_hash2, sizeof(uint64_t*), 0, cudaMemcpyHostToDevice);
}
__host__
void lyra2v2_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_outputHash, int order)
{
uint32_t tpb;
if (device_sm[device_map[thr_id]] == 350)
tpb = 64;
else if (device_sm[device_map[thr_id]] == 500)
tpb = 32;
else
tpb = TPB;
dim3 grid((threads + tpb - 1) / tpb);
dim3 block(tpb);
if (device_sm[device_map[thr_id]] >= 500)
lyra2v2_gpu_hash_32 <<<grid, block>>> (threads, startNounce, (uint2*)d_outputHash);
else
lyra2v2_gpu_hash_32_v3 <<<grid, block>>> (threads, startNounce, (uint2*)d_outputHash);
MyStreamSynchronize(NULL, order, thr_id);
}
+50 -8
View File
@@ -10,6 +10,7 @@ extern "C" {
#include "cuda_helper.h"
static uint64_t* d_hash[MAX_GPUS];
//static uint64_t* d_hash2[MAX_GPUS];
extern void blake256_cpu_init(int thr_id, uint32_t threads);
extern void blake256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32_t startNonce, uint64_t *Hash, int order);
@@ -19,14 +20,31 @@ extern void keccak256_cpu_init(int thr_id, uint32_t threads);
extern void skein256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNonce, uint64_t *d_outputHash, int order);
extern void skein256_cpu_init(int thr_id, uint32_t threads);
//extern void lyra2_cpu_init(int thr_id, uint32_t threads, uint64_t *hash);
extern void lyra2_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNonce, uint64_t *d_outputHash, int order);
extern void groestl256_cpu_init(int thr_id, uint32_t threads);
extern void groestl256_cpu_free(int thr_id);
extern void groestl256_setTarget(const void *ptarget);
extern uint32_t groestl256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_outputHash, int order);
extern uint32_t groestl256_getSecNonce(int thr_id, int num);
extern "C" void lyra2_hash(void *state, const void *input)
#ifdef _DEBUG
#define TRACE(algo) { \
if (max_nonce == 1 && pdata[19] <= 1) { \
uint32_t* debugbuf = NULL; \
cudaMallocHost(&debugbuf, 8*sizeof(uint32_t)); \
cudaMemcpy(debugbuf, d_hash[thr_id], 8*sizeof(uint32_t), cudaMemcpyDeviceToHost); \
printf("lyra %s %08x %08x %08x %08x...\n", algo, swab32(debugbuf[0]), swab32(debugbuf[1]), \
swab32(debugbuf[2]), swab32(debugbuf[3])); \
cudaFreeHost(debugbuf); \
} \
}
#else
#define TRACE(algo) {}
#endif
extern "C" void lyra2re_hash(void *state, const void *input)
{
sph_blake256_context ctx_blake;
sph_keccak256_context ctx_keccak;
@@ -58,17 +76,17 @@ extern "C" void lyra2_hash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_lyra2(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_lyra2(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int intensity = (device_sm[device_map[thr_id]] >= 500 && !is_windows()) ? 18 : 17;
uint32_t throughput = device_intensity(thr_id, __func__, 1U << intensity); // 18=256*256*4;
throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x0000ff;
((uint32_t*)ptarget)[7] = 0x00ff;
if (!init[thr_id])
{
@@ -79,7 +97,11 @@ extern "C" int scanhash_lyra2(int thr_id, uint32_t *pdata,
skein256_cpu_init(thr_id, throughput);
groestl256_cpu_init(thr_id, throughput);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], throughput * 64));
// DMatrix
// cudaMalloc(&d_hash2[thr_id], (size_t)16 * 8 * 8 * sizeof(uint64_t) * throughput);
// lyra2_cpu_init(thr_id, throughput, d_hash2[thr_id]);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t)32 * throughput));
init[thr_id] = true;
}
@@ -101,6 +123,7 @@ extern "C" int scanhash_lyra2(int thr_id, uint32_t *pdata,
keccak256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
lyra2_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
skein256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
TRACE("S")
foundNonce = groestl256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
if (foundNonce != UINT32_MAX)
@@ -108,18 +131,21 @@ extern "C" int scanhash_lyra2(int thr_id, uint32_t *pdata,
uint32_t _ALIGN(64) vhash64[8];
be32enc(&endiandata[19], foundNonce);
lyra2_hash(vhash64, endiandata);
lyra2re_hash(vhash64, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = groestl256_getSecNonce(thr_id, 1);
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != UINT32_MAX)
{
be32enc(&endiandata[19], secNonce);
lyra2_hash(vhash64, endiandata);
lyra2re_hash(vhash64, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
if (opt_debug)
applog(LOG_BLUE, "GPU #%d: found second nonce %08x", device_map[thr_id], secNonce);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -137,3 +163,19 @@ extern "C" int scanhash_lyra2(int thr_id, uint32_t *pdata,
return 0;
}
// cleanup
extern "C" void free_lyra2(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
groestl256_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+189
View File
@@ -0,0 +1,189 @@
extern "C" {
#include "sph/sph_blake.h"
#include "sph/sph_bmw.h"
#include "sph/sph_skein.h"
#include "sph/sph_keccak.h"
#include "sph/sph_cubehash.h"
#include "lyra2/Lyra2.h"
}
#include "miner.h"
#include "cuda_helper.h"
static _ALIGN(64) uint64_t *d_hash[MAX_GPUS];
static uint64_t *d_hash2[MAX_GPUS];
extern void blake256_cpu_init(int thr_id, uint32_t threads);
extern void blake256_cpu_hash_80(const int thr_id, const uint32_t threads, const uint32_t startNonce, uint64_t *Hash, int order);
extern void blake256_cpu_setBlock_80(uint32_t *pdata);
extern void keccak256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNonce, uint64_t *d_outputHash, int order);
extern void keccak256_cpu_init(int thr_id, uint32_t threads);
extern void skein256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNonce, uint64_t *d_outputHash, int order);
extern void skein256_cpu_init(int thr_id, uint32_t threads);
extern void cubehash256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *d_hash, int order);
extern void lyra2v2_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNonce, uint64_t *d_outputHash, int order);
extern void lyra2v2_cpu_init(int thr_id, uint32_t threads, uint64_t* matrix);
extern void bmw256_setTarget(const void *ptarget);
extern void bmw256_cpu_init(int thr_id, uint32_t threads);
extern void bmw256_cpu_free(int thr_id);
extern void bmw256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNounce, uint64_t *g_hash, uint32_t *resultnonces);
void lyra2v2_hash(void *state, const void *input)
{
uint32_t hashA[8], hashB[8];
sph_blake256_context ctx_blake;
sph_keccak256_context ctx_keccak;
sph_skein256_context ctx_skein;
sph_bmw256_context ctx_bmw;
sph_cubehash256_context ctx_cube;
sph_blake256_init(&ctx_blake);
sph_blake256(&ctx_blake, input, 80);
sph_blake256_close(&ctx_blake, hashA);
sph_keccak256_init(&ctx_keccak);
sph_keccak256(&ctx_keccak, hashA, 32);
sph_keccak256_close(&ctx_keccak, hashB);
sph_cubehash256_init(&ctx_cube);
sph_cubehash256(&ctx_cube, hashB, 32);
sph_cubehash256_close(&ctx_cube, hashA);
LYRA2(hashB, 32, hashA, 32, hashA, 32, 1, 4, 4);
sph_skein256_init(&ctx_skein);
sph_skein256(&ctx_skein, hashB, 32);
sph_skein256_close(&ctx_skein, hashA);
sph_cubehash256_init(&ctx_cube);
sph_cubehash256(&ctx_cube, hashA, 32);
sph_cubehash256_close(&ctx_cube, hashB);
sph_bmw256_init(&ctx_bmw);
sph_bmw256(&ctx_bmw, hashB, 32);
sph_bmw256_close(&ctx_bmw, hashA);
memcpy(state, hashA, 32);
}
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_lyra2v2(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int intensity = (device_sm[device_map[thr_id]] > 500 && !is_windows()) ? 18 : 17;
unsigned int defthr = 1U << intensity;
uint32_t throughput = device_intensity(device_map[thr_id], __func__, defthr);
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x00ff;
if (!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
//cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
//if (opt_n_gputhreads == 1)
// cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
blake256_cpu_init(thr_id, throughput);
keccak256_cpu_init(thr_id,throughput);
skein256_cpu_init(thr_id, throughput);
bmw256_cpu_init(thr_id, throughput);
if (device_sm[device_map[thr_id]] < 300) {
applog(LOG_ERR, "Device SM 3.0 or more recent required!");
proper_exit(1);
return -1;
}
// DMatrix
CUDA_SAFE_CALL(cudaMalloc(&d_hash2[thr_id], 16 * 4 * 4 * sizeof(uint64_t) * throughput));
lyra2v2_cpu_init(thr_id, throughput, d_hash2[thr_id]);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t)throughput * 32));
init[thr_id] = true;
}
uint32_t endiandata[20];
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], ((uint32_t*)pdata)[k]);
blake256_cpu_setBlock_80(pdata);
bmw256_setTarget(ptarget);
do {
int order = 0;
uint32_t foundNonces[2] = { 0, 0 };
blake256_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
keccak256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
cubehash256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
lyra2v2_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
skein256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
cubehash256_cpu_hash_32(thr_id, throughput,pdata[19], d_hash[thr_id], order++);
bmw256_cpu_hash_32(thr_id, throughput, pdata[19], d_hash[thr_id], foundNonces);
if (foundNonces[0] != 0)
{
const uint32_t Htarg = ptarget[7];
uint32_t vhash64[8];
be32enc(&endiandata[19], foundNonces[0]);
lyra2v2_hash(vhash64, endiandata);
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget))
{
int res = 1;
bn_store_hash_target_ratio(vhash64, ptarget, work);
// check if there was another one...
*hashes_done = pdata[19] - first_nonce + throughput;
if (foundNonces[1] != 0)
{
be32enc(&endiandata[19], foundNonces[1]);
lyra2v2_hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = foundNonces[1];
res++;
}
pdata[19] = foundNonces[0];
MyStreamSynchronize(NULL, 0, device_map[thr_id]);
return res;
}
else
{
if (vhash64[7] > Htarg) // don't show message if it is equal but fails fulltest
applog(LOG_WARNING, "GPU #%d: result does not validate on CPU!", thr_id);
}
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart && (max_nonce > ((uint64_t)(pdata[19]) + throughput)));
*hashes_done = pdata[19] - first_nonce + 1;
MyStreamSynchronize(NULL, 0, device_map[thr_id]);
return 0;
}
// cleanup
extern "C" void free_lyra2v2(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cudaFree(d_hash2[thr_id]);
bmw256_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+92 -127
View File
@@ -245,10 +245,14 @@ void aligned_free(void *ptr);
#if JANSSON_MAJOR_VERSION >= 2
#define JSON_LOADS(str, err_ptr) json_loads((str), 0, (err_ptr))
#define JSON_LOADF(str, err_ptr) json_load_file((str), 0, (err_ptr))
#else
#define JSON_LOADS(str, err_ptr) json_loads((str), (err_ptr))
#define JSON_LOADF(str, err_ptr) json_load_file((str), (err_ptr))
#endif
json_t * json_load_url(char* cfg_url, json_error_t *err);
#define USER_AGENT PACKAGE_NAME "/" PACKAGE_VERSION
void sha256_init(uint32_t *state);
@@ -260,129 +264,81 @@ void sha256d(unsigned char *hash, const unsigned char *data, int len);
struct work;
extern int scanhash_sha256d(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_blake256(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done, int8_t blakerounds);
extern int scanhash_bmw(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_c11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_deep(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_keccak256(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_fresh(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_fugue256(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_groestlcoin(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_heavy(int thr_id,struct work *work, uint32_t max_nonce, unsigned long *hashes_done, uint32_t maxvote, int blocklen);
extern int scanhash_jackpot(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_luffa(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_lyra2(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_lyra2v2(int thr_id,struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_myriad(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_neoscrypt(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_nist5(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_pentablake(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_quark(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_qubit(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_skeincoin(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_skein2(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_spreadx11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_s3(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_whirlx(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_x11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_x13(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_x14(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_x15(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_x17(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_zr5(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_deep(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_spreadx11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_luffa(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_sha256d(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_scrypt(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done,
unsigned char *scratchbuf, struct timeval *tv_start, struct timeval *tv_end);
extern int scanhash_scrypt_jane(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done,
unsigned char *scratchbuf, struct timeval *tv_start, struct timeval *tv_end);
extern int scanhash_fugue256(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
/* free device allocated memory per algo */
void miner_free_device(int thr_id);
extern int scanhash_groestlcoin(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_heavy(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done, uint32_t maxvote, int blocklen);
extern int scanhash_keccak256(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_myriad(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_jackpot(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_quark(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_anime(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_blake256(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done, int8_t blakerounds);
extern int scanhash_c11(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_fresh(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_lyra2(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_neoscrypt(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done);
extern int scanhash_nist5(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_pentablake(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_qubit(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_scrypt(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, unsigned char *scratchbuf, uint32_t max_nonce,
unsigned long *hashes_done, struct timeval *tv_start, struct timeval *tv_end);
extern int scanhash_scrypt_jane(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, unsigned char *scratchbuf, uint32_t max_nonce,
unsigned long *hashes_done, struct timeval *tv_start, struct timeval *tv_end);
extern int scanhash_skeincoin(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_skein2(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_s3(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_x11(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_x13(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_x14(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_x15(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_x17(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_whirlpoolx(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_zr5(int thr_id, struct work *work, uint32_t max_nonce,
unsigned long *hashes_done);
extern int scanhash_ether(int thr_id, struct work *work, uint64_t max_nonce,
unsigned long *hashes_done);
extern void free_blake256(int thr_id);
extern void free_bmw(int thr_id);
extern void free_c11(int thr_id);
extern void free_deep(int thr_id);
extern void free_keccak256(int thr_id);
extern void free_fresh(int thr_id);
extern void free_fugue256(int thr_id);
extern void free_groestlcoin(int thr_id);
extern void free_heavy(int thr_id);
extern void free_jackpot(int thr_id);
extern void free_luffa(int thr_id);
extern void free_lyra2(int thr_id);
extern void free_lyra2v2(int thr_id);
extern void free_myriad(int thr_id);
extern void free_neoscrypt(int thr_id);
extern void free_nist5(int thr_id);
extern void free_pentablake(int thr_id);
extern void free_quark(int thr_id);
extern void free_qubit(int thr_id);
extern void free_skeincoin(int thr_id);
extern void free_skein2(int thr_id);
extern void free_spreadx11(int thr_id);
extern void free_s3(int thr_id);
extern void free_whirlx(int thr_id);
extern void free_x11(int thr_id);
extern void free_x13(int thr_id);
extern void free_x14(int thr_id);
extern void free_x15(int thr_id);
extern void free_x17(int thr_id);
extern void free_zr5(int thr_id);
//extern void free_sha256d(int thr_id);
extern void free_scrypt(int thr_id);
extern void free_scrypt_jane(int thr_id);
/* api related */
void *api_thread(void *userdata);
@@ -492,6 +448,7 @@ extern bool opt_benchmark;
extern bool opt_debug;
extern bool opt_quiet;
extern bool opt_protocol;
extern bool opt_showdiff;
extern bool opt_tracegpu;
extern int opt_n_threads;
extern int active_gpus;
@@ -526,6 +483,7 @@ extern char* device_name[MAX_GPUS];
extern short device_map[MAX_GPUS];
extern long device_sm[MAX_GPUS];
extern uint32_t gpus_intensity[MAX_GPUS];
extern int opt_cudaschedule;
#define CL_N "\x1B[0m"
#define CL_RED "\x1B[31m"
@@ -567,13 +525,16 @@ extern bool hex2bin(unsigned char *p, const char *hexstr, size_t len);
extern int timeval_subtract(struct timeval *result, struct timeval *x,
struct timeval *y);
extern bool fulltest(const uint32_t *hash, const uint32_t *target);
extern void diff_to_target(uint32_t *target, double diff);
void diff_to_target(struct work* work, double diff);
double target_to_diff(uint32_t* target);
extern void get_currentalgo(char* buf, int sz);
extern uint32_t device_intensity(int thr_id, const char *func, uint32_t defcount);
// bignum
double bn_convert_nbits(const uint32_t nbits);
void bn_nbits_to_uchar(const uint32_t nBits, uchar *target);
double bn_hash_target_ratio(uint32_t* hash, uint32_t* target);
void bn_store_hash_target_ratio(uint32_t* hash, uint32_t* target, struct work* work);
struct stratum_job {
char *job_id;
@@ -603,6 +564,7 @@ struct stratum_ctx {
char *sockbuf;
double next_diff;
double sharediff;
char *session_id;
size_t xnonce1_size;
@@ -639,13 +601,18 @@ struct work {
uint64_t u64[1];
} noncerange;
double difficulty;
double targetdiff;
double shareratio;
double sharediff;
uint32_t height;
uint8_t pooln;
uint32_t scanned_from;
uint32_t scanned_to;
/* custom stuff */
void * extradata;
/* pok getwork txs */
uint32_t tx_count;
struct tx txs[POK_MAX_TXS];
@@ -662,7 +629,6 @@ struct pool_infos {
#define POOL_GETWORK 1
#define POOL_STRATUM 2
#define POOL_LONGPOLL 4
#define POOL_ETHER 128
uint8_t type;
#define POOL_ST_DEFINED 1
#define POOL_ST_VALID 2
@@ -715,9 +681,6 @@ json_t * json_rpc_call_pool(CURL *curl, struct pool_infos*,
json_t * json_rpc_longpoll(CURL *curl, char *lp_url, struct pool_infos*,
const char *req, int *err);
bool ether_getwork(CURL *curl, struct pool_infos *pool, struct work *work);
bool ether_submitwork(CURL *curl, struct pool_infos *pool, struct work *work);
bool stratum_socket_full(struct stratum_ctx *sctx, int timeout);
bool stratum_send_line(struct stratum_ctx *sctx, char *s);
char *stratum_recv_line(struct stratum_ctx *sctx);
@@ -777,8 +740,8 @@ void applog_hash(unsigned char *hash);
void applog_compare_hash(unsigned char *hash, unsigned char *hash2);
void print_hash_tests(void);
void animehash(void *state, const void *input);
void blake256hash(void *output, const void *input, int8_t rounds);
void bmw_hash(void *state, const void *input);
void c11hash(void *output, const void *input);
void deephash(void *state, const void *input);
void luffa_hash(void *state, const void *input);
@@ -788,7 +751,8 @@ void heavycoin_hash(unsigned char* output, const unsigned char* input, int len);
void keccak256_hash(void *state, const void *input);
unsigned int jackpothash(void *state, const void *input);
void groestlhash(void *state, const void *input);
void lyra2_hash(void *state, const void *input);
void lyra2re_hash(void *state, const void *input);
void lyra2v2_hash(void *state, const void *input);
void myriadhash(void *state, const void *input);
void neoscrypt(uchar *output, const uchar *input, uint32_t profile);
void nist5hash(void *state, const void *input);
@@ -799,6 +763,7 @@ void scrypthash(void* output, const void* input);
void scryptjane_hash(void* output, const void* input);
void skeincoinhash(void *output, const void *input);
void skein2hash(void *output, const void *input);
void spreadhash(void *output, const void *input);
void s3hash(void *output, const void *input);
void whirlxHash(void *state, const void *input);
void x11hash(void *output, const void *input);
+21 -5
View File
@@ -8,6 +8,7 @@
#include "miner.h"
void myriadgroestl_cpu_init(int thr_id, uint32_t threads);
void myriadgroestl_cpu_free(int thr_id);
void myriadgroestl_cpu_setBlock(int thr_id, void *data, void *pTargetIn);
void myriadgroestl_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce, void *outputHashes, uint32_t *nounce);
@@ -30,10 +31,11 @@ void myriadhash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
int scanhash_myriad(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
int scanhash_myriad(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t _ALIGN(64) endiandata[32];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t start_nonce = pdata[19];
uint32_t throughput = device_intensity(thr_id, __func__, 1 << 17);
throughput = min(throughput, max_nonce - start_nonce);
@@ -67,10 +69,11 @@ int scanhash_myriad(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
if (foundNounce < UINT32_MAX)
{
uint32_t _ALIGN(64) tmpHash[8];
uint32_t _ALIGN(64) vhash[8];
endiandata[19] = swab32(foundNounce);
myriadhash(tmpHash, endiandata);
if (tmpHash[7] <= ptarget[7] && fulltest(tmpHash, ptarget)) {
myriadhash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
bn_store_hash_target_ratio(vhash, ptarget, work);
pdata[19] = foundNounce;
free(outputHash);
return 1;
@@ -93,3 +96,16 @@ int scanhash_myriad(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
return 0;
}
// cleanup
void free_myriad(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
myriadgroestl_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+7
View File
@@ -735,6 +735,13 @@ void neoscrypt_cpu_init(int thr_id, uint32_t threads)
cudaMemcpyToSymbol(BLAKE2S_SIGMA, BLAKE2S_SIGMA_host, sizeof(BLAKE2S_SIGMA_host), 0, cudaMemcpyHostToDevice);
}
__host__
void neoscrypt_cpu_free(int thr_id)
{
cudaFree(d_NNonce[thr_id]);
cudaFree(d_buffer[thr_id]);
}
__host__
uint32_t neoscrypt_cpu_hash_k4(int thr_id, uint32_t threads, uint32_t startNounce, int have_stratum, int order)
{
+20 -2
View File
@@ -4,12 +4,16 @@
extern void neoscrypt_setBlockTarget(uint32_t * data, const void *ptarget);
extern void neoscrypt_cpu_init(int thr_id, uint32_t threads);
extern void neoscrypt_cpu_free(int thr_id);
extern uint32_t neoscrypt_cpu_hash_k4(int thr_id, uint32_t threads, uint32_t startNounce, int have_stratum, int order);
static bool init[MAX_GPUS] = { 0 };
int scanhash_neoscrypt(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done)
int scanhash_neoscrypt(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];
int intensity = is_windows() ? 18 : 19;
@@ -37,7 +41,6 @@ int scanhash_neoscrypt(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uin
init[thr_id] = true;
}
uint32_t endiandata[20];
if (have_stratum) {
for (int k = 0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
@@ -64,6 +67,7 @@ int scanhash_neoscrypt(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uin
neoscrypt((uchar*)vhash64, (uchar*) endiandata, 0x80000620U);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[19] = foundNonce;
return 1;
} else {
@@ -78,3 +82,17 @@ int scanhash_neoscrypt(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uin
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
void free_neoscrypt(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
neoscrypt_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+1 -1
View File
@@ -338,7 +338,7 @@ int nvml_set_clocks(nvml_handle *nvmlh, int dev_id)
nclocks = min(nclocks, 127);
if (nclocks)
nvmlh->nvmlDeviceGetSupportedMemoryClocks(nvmlh->devs[n], &nclocks, clocks);
for (int8_t u=0; u < nclocks; u++) {
for (uint8_t u=0; u < nclocks; u++) {
// ordered by pstate (so highest is first memory clock - P0)
if (clocks[u] <= mem_clk) {
mem_clk = clocks[u];
+31 -8
View File
@@ -364,11 +364,12 @@ void pentablake_cpu_setBlock_80(uint32_t *pdata, const uint32_t *ptarget)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_pentablake(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
extern "C" int scanhash_pentablake(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];
uint32_t endiandata[20];
int rc = 0;
uint32_t throughput = device_intensity(thr_id, __func__, 128U * 2560); // 18.5
throughput = min(throughput, max_nonce - first_nonce);
@@ -403,20 +404,25 @@ extern "C" int scanhash_pentablake(int thr_id, uint32_t *pdata, const uint32_t *
pentablake_cpu_hash_64(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
pentablake_cpu_hash_64(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = pentablake_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
if (foundNonce != UINT32_MAX)
{
const uint32_t Htarg = ptarget[7];
uint32_t vhashcpu[8];
uint32_t vhash[8];
be32enc(&endiandata[19], foundNonce);
pentablakehash(vhashcpu, endiandata);
pentablakehash(vhash, endiandata);
if (vhashcpu[7] <= Htarg && fulltest(vhashcpu, ptarget)) {
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
rc = 1;
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash, ptarget, work);
if (extra_results[0] != UINT32_MAX) {
// Rare but possible if the throughput is big
be32enc(&endiandata[19], extra_results[0]);
pentablakehash(vhash, endiandata);
if (bn_hash_target_ratio(vhash, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash, ptarget, work);
applog(LOG_NOTICE, "GPU found more than one result yippee!");
pdata[21] = extra_results[0];
extra_results[0] = UINT32_MAX;
@@ -436,3 +442,20 @@ extern "C" int scanhash_pentablake(int thr_id, uint32_t *pdata, const uint32_t *
*hashes_done = pdata[19] - first_nonce + 1;
return rc;
}
// cleanup
void free_pentablake(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cudaFree(h_resNounce[thr_id]);
cudaFree(d_resNounce[thr_id]);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+2
View File
@@ -21,6 +21,8 @@
"algo" : "zr5",
"timeout": 60,
"api-bind" : "0.0.0.0",
"api-remote" : true,
-292
View File
@@ -1,292 +0,0 @@
extern "C"
{
#include "sph/sph_blake.h"
#include "sph/sph_bmw.h"
#include "sph/sph_groestl.h"
#include "sph/sph_skein.h"
#include "sph/sph_jh.h"
#include "sph/sph_keccak.h"
}
#include "miner.h"
#include "cuda_helper.h"
static uint32_t *d_hash[MAX_GPUS];
// Speicher zur Generierung der Noncevektoren für die bedingten Hashes
static uint32_t *d_branch1Nonces[MAX_GPUS];
static uint32_t *d_branch2Nonces[MAX_GPUS];
static uint32_t *d_branch3Nonces[MAX_GPUS];
extern void quark_blake512_cpu_init(int thr_id, uint32_t threads);
extern void quark_blake512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_bmw512_cpu_init(int thr_id, uint32_t threads);
extern void quark_bmw512_cpu_setBlock_80(void *pdata);
extern void quark_bmw512_cpu_hash_80(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_outputHash, int order);
extern void quark_bmw512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_outputHash, int order);
extern void quark_groestl512_cpu_init(int thr_id, uint32_t threads);
extern void quark_groestl512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_doublegroestl512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_skein512_cpu_init(int thr_id, uint32_t threads);
extern void quark_skein512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_keccak512_cpu_init(int thr_id, uint32_t threads);
extern void quark_keccak512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_jh512_cpu_init(int thr_id, uint32_t threads);
extern void quark_jh512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes, uint32_t *d_validNonceTable,
uint32_t *d_nonces1, uint32_t *nrm1,
uint32_t *d_nonces2, uint32_t *nrm2,
int order);
extern void quark_compactTest_single_false_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes, uint32_t *d_validNonceTable,
uint32_t *d_nonces1, uint32_t *nrm1,
int order);
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);
/* CPU Hash */
extern "C" void animehash(void *state, const void *input)
{
sph_blake512_context ctx_blake;
sph_bmw512_context ctx_bmw;
sph_groestl512_context ctx_groestl;
sph_jh512_context ctx_jh;
sph_keccak512_context ctx_keccak;
sph_skein512_context ctx_skein;
unsigned char hash[64];
sph_bmw512_init(&ctx_bmw);
// ZBMW;
sph_bmw512 (&ctx_bmw, (const void*) input, 80);
sph_bmw512_close(&ctx_bmw, (void*) hash);
sph_blake512_init(&ctx_blake);
// ZBLAKE;
sph_blake512 (&ctx_blake, hash, 64);
sph_blake512_close(&ctx_blake, (void*) hash);
if (hash[0] & 0x8)
{
sph_groestl512_init(&ctx_groestl);
// ZGROESTL;
sph_groestl512 (&ctx_groestl, (const void*) hash, 64);
sph_groestl512_close(&ctx_groestl, (void*) hash);
}
else
{
sph_skein512_init(&ctx_skein);
// ZSKEIN;
sph_skein512 (&ctx_skein, (const void*) hash, 64);
sph_skein512_close(&ctx_skein, (void*) hash);
}
sph_groestl512_init(&ctx_groestl);
// ZGROESTL;
sph_groestl512 (&ctx_groestl, (const void*) hash, 64);
sph_groestl512_close(&ctx_groestl, (void*) hash);
sph_jh512_init(&ctx_jh);
// ZJH;
sph_jh512 (&ctx_jh, (const void*) hash, 64);
sph_jh512_close(&ctx_jh, (void*) hash);
if (hash[0] & 0x8)
{
sph_blake512_init(&ctx_blake);
// ZBLAKE;
sph_blake512 (&ctx_blake, (const void*) hash, 64);
sph_blake512_close(&ctx_blake, (void*) hash);
}
else
{
sph_bmw512_init(&ctx_bmw);
// ZBMW;
sph_bmw512 (&ctx_bmw, (const void*) hash, 64);
sph_bmw512_close(&ctx_bmw, (void*) hash);
}
sph_keccak512_init(&ctx_keccak);
// ZKECCAK;
sph_keccak512 (&ctx_keccak, (const void*) hash, 64);
sph_keccak512_close(&ctx_keccak, (void*) hash);
sph_skein512_init(&ctx_skein);
// SKEIN;
sph_skein512 (&ctx_skein, (const void*) hash, 64);
sph_skein512_close(&ctx_skein, (void*) hash);
if (hash[0] & 0x8)
{
sph_keccak512_init(&ctx_keccak);
// ZKECCAK;
sph_keccak512 (&ctx_keccak, (const void*) hash, 64);
sph_keccak512_close(&ctx_keccak, (void*) hash);
}
else
{
sph_jh512_init(&ctx_jh);
// ZJH;
sph_jh512 (&ctx_jh, (const void*) hash, 64);
sph_jh512_close(&ctx_jh, (void*) hash);
}
memcpy(state, hash, 32);
}
/*
struct HashPredicate
{
HashPredicate(uint32_t *hashes, uint32_t startNonce) :
m_hashes(hashes),
m_startNonce(startNonce)
{ }
__device__
bool operator()(const uint32_t x)
{
uint32_t *hash = &m_hashes[(x - m_startNonce)*16];
return hash[0] & 0x8;
}
uint32_t *m_hashes;
uint32_t m_startNonce;
};
*/
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_anime(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
{
const uint32_t first_nonce = pdata[19];
uint32_t throughput = device_intensity(thr_id, __func__, 1 << 19); // 256*256*8
throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x00000f;
if (!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], 16 * sizeof(uint32_t) * throughput));
quark_blake512_cpu_init(thr_id, throughput);
quark_groestl512_cpu_init(thr_id, throughput);
quark_skein512_cpu_init(thr_id, throughput);
quark_bmw512_cpu_init(thr_id, throughput);
quark_keccak512_cpu_init(thr_id, throughput);
quark_jh512_cpu_init(thr_id, throughput);
cuda_check_cpu_init(thr_id, throughput);
quark_compactTest_cpu_init(thr_id, throughput);
CUDA_SAFE_CALL(cudaMalloc(&d_branch1Nonces[thr_id], sizeof(uint32_t)*throughput));
CUDA_SAFE_CALL(cudaMalloc(&d_branch2Nonces[thr_id], sizeof(uint32_t)*throughput));
CUDA_SAFE_CALL(cudaMalloc(&d_branch3Nonces[thr_id], sizeof(uint32_t)*throughput));
init[thr_id] = true;
}
uint32_t endiandata[20];
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
quark_bmw512_cpu_setBlock_80((void*)endiandata);
cuda_check_cpu_setTarget(ptarget);
do {
int order = 0;
uint32_t nrm1=0, nrm2=0, nrm3=0;
// erstes BMW512 Hash mit CUDA
quark_bmw512_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id], order++);
// das ist der unbedingte Branch für Blake512
quark_blake512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_compactTest_single_false_cpu_hash_64(thr_id, throughput, pdata[19], d_hash[thr_id], NULL,
d_branch3Nonces[thr_id], &nrm3,
order++);
// nur den Skein Branch weiterverfolgen
quark_skein512_cpu_hash_64(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
// das ist der unbedingte Branch für Groestl512
quark_groestl512_cpu_hash_64(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
// das ist der unbedingte Branch für JH512
quark_jh512_cpu_hash_64(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
// quarkNonces in branch1 und branch2 aufsplitten gemäss if (hash[0] & 0x8)
quark_compactTest_cpu_hash_64(thr_id, nrm3, pdata[19], d_hash[thr_id], d_branch3Nonces[thr_id],
d_branch1Nonces[thr_id], &nrm1,
d_branch2Nonces[thr_id], &nrm2,
order++);
// das ist der bedingte Branch für Blake512
quark_blake512_cpu_hash_64(thr_id, nrm1, pdata[19], d_branch1Nonces[thr_id], d_hash[thr_id], order++);
// das ist der bedingte Branch für Bmw512
quark_bmw512_cpu_hash_64(thr_id, nrm2, pdata[19], d_branch2Nonces[thr_id], d_hash[thr_id], order++);
// das ist der unbedingte Branch für Keccak512
quark_keccak512_cpu_hash_64(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
// das ist der unbedingte Branch für Skein512
quark_skein512_cpu_hash_64(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
// quarkNonces in branch1 und branch2 aufsplitten gemäss if (hash[0] & 0x8)
quark_compactTest_cpu_hash_64(thr_id, nrm3, pdata[19], d_hash[thr_id], d_branch3Nonces[thr_id],
d_branch1Nonces[thr_id], &nrm1,
d_branch2Nonces[thr_id], &nrm2,
order++);
// das ist der bedingte Branch für Keccak512
quark_keccak512_cpu_hash_64(thr_id, nrm1, pdata[19], d_branch1Nonces[thr_id], d_hash[thr_id], order++);
// das ist der bedingte Branch für JH512
quark_jh512_cpu_hash_64(thr_id, nrm2, pdata[19], d_branch2Nonces[thr_id], d_hash[thr_id], order++);
// Scan nach Gewinner Hashes auf der GPU
uint32_t foundNonce = cuda_check_hash_branch(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
if (foundNonce != UINT32_MAX)
{
const uint32_t Htarg = ptarget[7];
uint32_t vhash64[8];
be32enc(&endiandata[19], foundNonce);
animehash(vhash64, endiandata);
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
if (secNonce != 0) {
pdata[21] = secNonce;
res++;
}
pdata[19] = foundNonce;
return res;
} else {
applog(LOG_WARNING, "GPU #%d: result for nonce %08x does not validate on CPU!", device_map[thr_id], foundNonce);
}
}
if ((uint64_t)pdata[19] + throughput > (uint64_t)max_nonce) {
pdata[19] = max_nonce;
break;
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart);
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
+13 -1
View File
@@ -27,7 +27,7 @@ __device__ cuda_compactTestFunction_t d_QuarkTrueFunction = QuarkTrueTest, d_Qua
cuda_compactTestFunction_t h_QuarkTrueFunction[MAX_GPUS], h_QuarkFalseFunction[MAX_GPUS];
// Setup-Funktionen
// Setup/Alloc Function
__host__ void quark_compactTest_cpu_init(int thr_id, uint32_t threads)
{
cudaMemcpyFromSymbol(&h_QuarkTrueFunction[thr_id], d_QuarkTrueFunction, sizeof(cuda_compactTestFunction_t));
@@ -45,6 +45,18 @@ __host__ void quark_compactTest_cpu_init(int thr_id, uint32_t threads)
cudaMalloc(&d_partSum[1][thr_id], sizeof(uint32_t) * s1); // BLOCKSIZE (Threads/Block)
}
// Because all alloc should have a free...
__host__ void quark_compactTest_cpu_free(int thr_id)
{
cudaFree(d_tempBranch1Nonces[thr_id]);
cudaFree(d_numValid[thr_id]);
cudaFree(d_partSum[0][thr_id]);
cudaFree(d_partSum[1][thr_id]);
cudaFreeHost(h_numValid[thr_id]);
}
#if __CUDA_ARCH__ < 300
/**
* __shfl_up() calculates a source lane ID by subtracting delta from the caller's lane ID, and clamping to the range 0..width-1
+33 -8
View File
@@ -41,6 +41,7 @@ extern void quark_jh512_cpu_init(int thr_id, uint32_t threads);
extern void quark_jh512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_free(int thr_id);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes, uint32_t *d_validNonceTable,
uint32_t *d_nonces1, uint32_t *nrm1,
uint32_t *d_nonces2, uint32_t *nrm2,
@@ -51,6 +52,7 @@ extern void quark_compactTest_single_false_cpu_hash_64(int thr_id, uint32_t thre
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 Quarkhash Funktion aus einem miner Quelltext
extern "C" void quarkhash(void *state, const void *input)
{
@@ -131,10 +133,11 @@ extern "C" void quarkhash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_quark(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_quark(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];
uint32_t throughput = device_intensity(thr_id, __func__, 1 << 20); // 256*4096
@@ -166,7 +169,6 @@ extern "C" int scanhash_quark(int thr_id, uint32_t *pdata,
init[thr_id] = true;
}
uint32_t endiandata[20];
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
@@ -226,19 +228,19 @@ extern "C" int scanhash_quark(int thr_id, uint32_t *pdata,
// das ist der bedingte Branch für JH512
quark_jh512_cpu_hash_64(thr_id, nrm2, pdata[19], d_branch2Nonces[thr_id], d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + 1;
// Scan nach Gewinner Hashes auf der GPU
uint32_t foundNonce = cuda_check_hash_branch(thr_id, nrm3, pdata[19], d_branch3Nonces[thr_id], d_hash[thr_id], order++);
if (foundNonce != 0xffffffff)
{
const uint32_t Htarg = ptarget[7];
uint32_t vhash64[8];
be32enc(&endiandata[19], foundNonce);
quarkhash(vhash64, endiandata);
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[19] = foundNonce;
*hashes_done = foundNonce - first_nonce + 1;
return 1;
} else {
applog(LOG_WARNING, "GPU #%d: result for nonce %08x does not validate on CPU!", device_map[thr_id], foundNonce);
@@ -252,3 +254,26 @@ extern "C" int scanhash_quark(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_quark(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaDeviceSynchronize();
cudaFree(d_hash[thr_id]);
cudaFree(d_branch1Nonces[thr_id]);
cudaFree(d_branch2Nonces[thr_id]);
cudaFree(d_branch3Nonces[thr_id]);
quark_compactTest_cpu_free(thr_id);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+26 -3
View File
@@ -52,10 +52,11 @@ extern "C" void deephash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_deep(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
extern "C" int scanhash_deep(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];
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 19); // 256*256*8
throughput = min(throughput, (max_nonce - first_nonce));
@@ -91,6 +92,8 @@ extern "C" int scanhash_deep(int thr_id, uint32_t *pdata, const uint32_t *ptarge
x11_cubehash512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -101,8 +104,12 @@ extern "C" int scanhash_deep(int thr_id, uint32_t *pdata, const uint32_t *ptarge
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
deephash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -121,3 +128,19 @@ extern "C" int scanhash_deep(int thr_id, uint32_t *pdata, const uint32_t *ptarge
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_deep(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+22 -3
View File
@@ -30,10 +30,11 @@ extern "C" void luffa_hash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_luffa(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
extern "C" int scanhash_luffa(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];
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 22); // 256*256*8*8
throughput = min(throughput, max_nonce - first_nonce);
@@ -44,6 +45,8 @@ extern "C" int scanhash_luffa(int thr_id, uint32_t *pdata, const uint32_t *ptarg
if (!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
if (opt_cudaschedule == -1) // to reduce cpu usage...
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], throughput * 64));
@@ -73,7 +76,7 @@ extern "C" int scanhash_luffa(int thr_id, uint32_t *pdata, const uint32_t *ptarg
luffa_hash(vhash64, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
//*hashes_done = min(max_nonce - first_nonce, (uint64_t) pdata[19] - first_nonce + throughput);
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[19] = foundNonce;
return 1;
} else {
@@ -93,3 +96,19 @@ extern "C" int scanhash_luffa(int thr_id, uint32_t *pdata, const uint32_t *ptarg
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_luffa(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+27 -5
View File
@@ -74,11 +74,11 @@ extern "C" void qubithash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_qubit(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_qubit(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t endiandata[20];
uint32_t _ALIGN(64) endiandata[20];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 19); // 256*256*8
throughput = min(throughput, max_nonce - first_nonce);
@@ -119,6 +119,8 @@ extern "C" int scanhash_qubit(int thr_id, uint32_t *pdata,
x11_simd512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -130,8 +132,12 @@ extern "C" int scanhash_qubit(int thr_id, uint32_t *pdata,
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
qubithash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -150,3 +156,19 @@ extern "C" int scanhash_qubit(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_qubit(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+4 -4
View File
@@ -60,8 +60,8 @@ IDI_ICON1 ICON "ccminer.ico"
//
VS_VERSION_INFO VERSIONINFO
FILEVERSION 1,6,5,0
PRODUCTVERSION 1,6,5,0
FILEVERSION 1,6,6,0
PRODUCTVERSION 1,6,6,0
FILEFLAGSMASK 0x3fL
#ifdef _DEBUG
FILEFLAGS 0x21L
@@ -76,10 +76,10 @@ BEGIN
BEGIN
BLOCK "040904e4"
BEGIN
VALUE "FileVersion", "1.6.5"
VALUE "FileVersion", "1.6.6"
VALUE "LegalCopyright", "Copyright (C) 2015"
VALUE "ProductName", "ccminer"
VALUE "ProductVersion", "1.6.5"
VALUE "ProductVersion", "1.6.6"
END
END
BLOCK "VarFileInfo"
+11 -2
View File
@@ -426,13 +426,21 @@ unsigned char GetNfactor(unsigned int nTimestamp)
return Nfactor;
}
// cleanup
void free_scrypt_jane(int thr_id)
{
// todo ?
}
#define bswap_32x4(x) ((((x) << 24) & 0xff000000u) | (((x) << 8) & 0x00ff0000u) \
| (((x) >> 8) & 0x0000ff00u) | (((x) >> 24) & 0x000000ffu))
static int s_Nfactor = 0;
int scanhash_scrypt_jane(int thr_id, uint32_t *pdata, const uint32_t *ptarget, unsigned char *scratchbuf,
uint32_t max_nonce, unsigned long *hashes_done, struct timeval *tv_start, struct timeval *tv_end)
int scanhash_scrypt_jane(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done,
unsigned char *scratchbuf, struct timeval *tv_start, struct timeval *tv_end)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
uint32_t N;
@@ -594,6 +602,7 @@ int scanhash_scrypt_jane(int thr_id, uint32_t *pdata, const uint32_t *ptarget, u
if (memcmp(thash, &hash[cur][8*i], 32) == 0)
{
bn_store_hash_target_ratio(thash, ptarget, work);
*hashes_done = n - pdata[19];
pdata[19] = tmp_nonce;
scrypt_free(&Vbuf);
+11 -2
View File
@@ -685,14 +685,22 @@ static int lastFactor = 0;
static void computeGold(uint32_t* const input, uint32_t *reference, uchar *scratchpad);
// cleanup
void free_scrypt(int thr_id)
{
// todo ?
}
// Scrypt proof of work algorithm
// using SSE2 vectorized HMAC SHA256 on CPU and
// a salsa core implementation on GPU with CUDA
//
int scanhash_scrypt(int thr_id, uint32_t *pdata, const uint32_t *ptarget, unsigned char *scratchbuf,
uint32_t max_nonce, unsigned long *hashes_done, struct timeval *tv_start, struct timeval *tv_end)
int scanhash_scrypt(int thr_id, struct work *work, uint32_t max_nonce, unsigned long *hashes_done,
unsigned char *scratchbuf, struct timeval *tv_start, struct timeval *tv_end)
{
int result = 0;
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
int throughput = cuda_throughput(thr_id);
if(throughput == 0)
@@ -904,6 +912,7 @@ int scanhash_scrypt(int thr_id, uint32_t *pdata, const uint32_t *ptarget, unsign
device_map[thr_id], device_name[thr_id], i, cur);
} else {
*hashes_done = n - pdata[19];
bn_store_hash_target_ratio(refhash, ptarget, work);
pdata[19] = nonce[cur] + i;
result = 1;
goto byebye;
+29 -7
View File
@@ -346,10 +346,11 @@ static __inline uint32_t swab32_if(uint32_t val, bool iftrue) {
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_skeincoin(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
extern "C" int scanhash_skeincoin(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];
const int swap = 1;
@@ -409,23 +410,28 @@ extern "C" int scanhash_skeincoin(int thr_id, uint32_t *pdata, const uint32_t *p
if (foundNonce != UINT32_MAX)
{
uint32_t _ALIGN(64) vhash64[8];
uint32_t _ALIGN(64) vhash[8];
endiandata[19] = swab32_if(foundNonce, swap);
skeincoinhash(vhash64, endiandata);
skeincoinhash(vhash, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
int res = 1;
uint8_t num = res;
bn_store_hash_target_ratio(vhash, ptarget, work);
if (checkSecnonce) {
secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], num);
}
while (secNonce != 0 && res < 2) /* todo: up to 6 */
{
endiandata[19] = swab32_if(secNonce, swap);
skeincoinhash(vhash64, endiandata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
skeincoinhash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
// todo: use 19 20 21... zr5 pok to adapt...
endiandata[19] = swab32_if(secNonce, swap);
skeincoinhash(vhash, endiandata);
if (bn_hash_target_ratio(vhash, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash, ptarget, work);
pdata[19+res*2] = swab32_if(secNonce, !swap);
res++;
}
@@ -458,3 +464,19 @@ extern "C" int scanhash_skeincoin(int thr_id, uint32_t *pdata, const uint32_t *p
return 0;
}
// cleanup
extern "C" void free_skeincoin(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+25 -2
View File
@@ -36,9 +36,10 @@ void skein2hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
int scanhash_skein2(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
int scanhash_skein2(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t throughput = device_intensity(thr_id, __func__, 1 << 19); // 256*256*8
@@ -88,9 +89,15 @@ int scanhash_skein2(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
if (!opt_quiet)
applog(LOG_BLUE, "GPU #%d: found second nonce %08x !", device_map[thr_id], swab32(secNonce));
endiandata[19] = secNonce;
skein2hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = swab32(secNonce);
res++;
}
@@ -113,3 +120,19 @@ int scanhash_skein2(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
return 0;
}
// cleanup
void free_skein2(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+10 -8
View File
@@ -468,8 +468,8 @@ static inline void sha256d_ms(uint32_t *hash, uint32_t *W,
void sha256d_ms_4way(uint32_t *hash, uint32_t *data,
const uint32_t *midstate, const uint32_t *prehash);
static inline int scanhash_sha256d_4way(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done)
static inline int scanhash_sha256d_4way(int thr_id, uint32_t *pdata,
const uint32_t *ptarget uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t data[4 * 64] __attribute__((aligned(128)));
uint32_t hash[4 * 8] __attribute__((aligned(32)));
@@ -508,6 +508,7 @@ static inline int scanhash_sha256d_4way(int thr_id, uint32_t *pdata,
pdata[19] = data[4 * 3 + i];
sha256d_80_swap(hash, pdata);
if (fulltest(hash, ptarget)) {
bn_store_hash_target_ratio(hash, ptarget, work);
*hashes_done = n - first_nonce + 1;
return 1;
}
@@ -581,13 +582,14 @@ static inline int scanhash_sha256d_8way(int thr_id, uint32_t *pdata,
#endif /* HAVE_SHA256_8WAY */
int scanhash_sha256d(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
int scanhash_sha256d(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t data[64] /* __attribute__((aligned(128))) */;
uint32_t hash[8] /* __attribute__((aligned(32))) */;
uint32_t midstate[8] /* __attribute__((aligned(32))) */;
uint32_t prehash[8] /* __attribute__((aligned(32))) */;
uint32_t _ALIGN(128) data[64];
uint32_t hash[8];
uint32_t midstate[8];
uint32_t prehash[8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19] - 1;
const uint32_t first_nonce = pdata[19];
const uint32_t Htarg = ptarget[7];
+695
View File
@@ -0,0 +1,695 @@
/*
Based on libtommath and tomsfastmath by Tom St Denis
This is basically Barret reduction and it's setup from libtommath and
basic bignum types and functions from tomsfastmath with some application
specific optimizations like dropping sign handling and branches that are
never reached, hard coding for a specific digit size (replacing slow ops like
x % DIGIT_BIT with x & 31 etc).
Could probably do a lot better with CUDA, this code spills to local memory
like a mofo due to dynamic array indexes.
*/
#include <stdint.h>
#ifndef CHAR_BIT
#define CHAR_BIT 8
#endif
#ifndef MIN
#define MIN(x,y) ((x)<(y)?(x):(y))
#endif
#ifndef MAX
#define MAX(x,y) ((x)>(y)?(x):(y))
#endif
#define DIGIT_BIT 32
#define FP_SIZE 20
typedef uint32_t fp_digit;
typedef uint64_t fp_word;
typedef fp_digit fp_min_u32;
/* return codes */
#define FP_OKAY 0
#define FP_VAL 1
#define FP_MEM 2
/* equalities */
#define FP_LT -1 /* less than */
#define FP_EQ 0 /* equal to */
#define FP_GT 1 /* greater than */
/* replies */
#define FP_YES 1 /* yes response */
#define FP_NO 0 /* no response */
typedef int fp_err;
typedef struct {
int used;
fp_digit dp[FP_SIZE];
} fp_int;
#define USED(m) ((m)->used)
#define DIGIT(m,k) ((m)->dp[(k)])
#define fp_iszero(a) (((a)->used == 0) ? FP_YES : FP_NO)
#define fp_iseven(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 0)) ? FP_YES : FP_NO)
#define fp_isodd(a) (((a)->used > 0 && (((a)->dp[0] & 1) == 1)) ? FP_YES : FP_NO)
#define fp_init(a) fp_zero((a))
#define fp_mul(a,b,c) fp_mul_comba((a), (b), (c))
/* clamp digits */
#define fp_clamp(a) { while ((a)->used && (a)->dp[(a)->used-1] == 0) --((a)->used); }
__device__ __host__ void fp_zero( fp_int *a )
{
for( int i = 0; i < FP_SIZE; i++ ) a->dp[i] = 0;
a->used = 0;
}
__device__ __host__ void fp_set( fp_int *a, fp_digit b )
{
fp_zero(a);
a->dp[0] = b;
a->used = a->dp[0] ? 1 : 0;
}
__device__ __host__ int fp_cmp_mag( fp_int *a, fp_int *b )
{
if( a->used > b->used ) return FP_GT;
else if( a->used < b->used ) return FP_LT;
for( int x = a->used - 1; x >= 0; x-- ) {
if( a->dp[x] > b->dp[x] ) return FP_GT;
else if( a->dp[x] < b->dp[x] ) return FP_LT;
}
return FP_EQ;
}
/* unsigned addition */
__device__ __host__ void s_fp_add(fp_int *a, fp_int *b, fp_int *c)
{
int x, y, oldused;
fp_word t;
y = MAX(a->used, b->used);
oldused = c->used;
c->used = y;
t = 0;
for (x = 0; x < y; x++) {
t += ((fp_word)a->dp[x]) + ((fp_word)b->dp[x]);
c->dp[x] = (fp_digit)t;
t >>= DIGIT_BIT;
}
if (t != 0 && x < FP_SIZE) {
c->dp[c->used++] = (fp_digit)t;
++x;
}
c->used = x;
for (; x < oldused; x++) {
c->dp[x] = 0;
}
fp_clamp(c);
}
/* unsigned subtraction ||a|| >= ||b|| ALWAYS! */
__device__ __host__ void s_fp_sub(fp_int *a, fp_int *b, fp_int *c)
{
int x, oldbused, oldused;
fp_word t;
oldused = c->used;
oldbused = b->used;
c->used = a->used;
t = 0;
for (x = 0; x < oldbused; x++) {
t = ((fp_word)a->dp[x]) - (((fp_word)b->dp[x]) + t);
c->dp[x] = (fp_digit)t;
t = (t >> DIGIT_BIT)&1;
}
for (; x < a->used; x++) {
t = ((fp_word)a->dp[x]) - t;
c->dp[x] = (fp_digit)t;
t = (t >> DIGIT_BIT)&1;
}
for (; x < oldused; x++) {
c->dp[x] = 0;
}
fp_clamp(c);
}
__device__ __host__ void fp_add( fp_int *a, fp_int *b, fp_int *c )
{
s_fp_add (a, b, c);
}
/* c = a - b */
__device__ __host__ void fp_sub(fp_int *a, fp_int *b, fp_int *c)
{
// int sa, sb;
/* The first has a larger or equal magnitude */
s_fp_sub (a, b, c);
}
__device__ __host__ int fp_cmp(fp_int *a, fp_int *b)
{
return fp_cmp_mag(a, b);
}
__device__ __host__ void fp_copy( fp_int *a, fp_int *b )
{
for( int i = 0; i < FP_SIZE; i++ ) b->dp[i] = a->dp[i];
b->used = a->used;
}
// TODO: use clz, return a->used - clz(a->dp[a->used-1])
__device__ __host__ int fp_count_bits( fp_int *a )
{
int r;
fp_digit q;
/* shortcut */
if (a->used == 0) return 0;
/* get number of digits and add that */
r = (a->used - 1) << 5;
/* take the last digit and count the bits in it */
q = a->dp[a->used - 1];
while (q > ((fp_digit) 0)) {
++r;
q >>= ((fp_digit) 1);
}
return r;
}
__device__ __host__ void fp_rshd( fp_int *a, int x )
{
int y;
/* too many digits just zero and return */
if (x >= a->used) {
fp_zero(a);
return;
}
/* shift */
for (y = 0; y < a->used - x; y++) a->dp[y] = a->dp[y+x];
/* zero rest */
for (; y < a->used; y++) a->dp[y] = 0;
/* decrement count */
a->used -= x;
fp_clamp(a);
}
__device__ __host__ void fp_lshd(fp_int *a, int x)
{
int y;
/* move up and truncate as required */
y = MIN(a->used + x - 1, (int)(FP_SIZE-1));
/* store new size */
a->used = y + 1;
/* move digits */
for (; y >= x; y--) a->dp[y] = a->dp[y-x];
/* zero lower digits */
for (; y >= 0; y--) a->dp[y] = 0;
/* clamp digits */
fp_clamp(a);
}
/* c = a * 2**d */
__device__ __host__ void fp_mul_2d( fp_int *a, int b, fp_int *c )
{
fp_digit carry, carrytmp, shift;
int x;
/* copy it */
fp_copy(a, c);
/* handle whole digits */
if (b >= DIGIT_BIT) {
fp_lshd(c, b >> 5);
}
b &= 31;
/* shift the digits */
if (b != 0) {
carry = 0;
shift = DIGIT_BIT - b;
for (x = 0; x < c->used; x++) {
carrytmp = c->dp[x] >> shift;
c->dp[x] = (c->dp[x] << b) + carry;
carry = carrytmp;
}
/* store last carry if room */
if (carry && x < FP_SIZE) c->dp[c->used++] = carry;
}
fp_clamp(c);
}
/* c = a * b */
__device__ __host__ void fp_mul_d( fp_int *a, fp_digit b, fp_int *c )
{
fp_word w;
int x, oldused;
oldused = c->used;
c->used = a->used;
w = 0;
for (x = 0; x < a->used; x++) {
w = ((fp_word)a->dp[x]) * ((fp_word)b) + w;
c->dp[x] = (fp_digit)w;
w = w >> DIGIT_BIT;
}
if (w != 0 && (a->used != FP_SIZE)) {
c->dp[c->used++] = w;
++x;
}
for (; x < oldused; x++) c->dp[x] = 0;
fp_clamp(c);
}
/* c = a mod 2**d */
__device__ __host__ void fp_mod_2d( fp_int *a, int b, fp_int *c )
{
int x;
/* zero if count less than or equal to zero */
if (b <= 0) {
fp_zero(c);
return;
}
/* get copy of input */
fp_copy(a, c);
/* if 2**d is larger than we just return */
if (b >= (a->used << 5)) return;
/* zero digits above the last digit of the modulus */
for (x = (b >> 5) + ((b & 31) == 0 ? 0 : 1); x < c->used; x++) {
c->dp[x] = 0;
}
/* clear the digit that is not completely outside/inside the modulus */
c->dp[b >> 5] &= ~((fp_digit)0) >> (DIGIT_BIT - b);
fp_clamp (c);
}
/* c = a / 2**b */
__device__ __host__ void fp_div_2d( fp_int *a, int b, fp_int *c, fp_int *d )
{
fp_digit D, r, rr;
int x;
fp_int t;
/* if the shift count is <= 0 then we do no work */
if (b <= 0) {
fp_copy (a, c);
if (d != NULL) fp_zero (d);
return;
}
fp_zero(&t);
/* get the remainder */
if (d != NULL) fp_mod_2d (a, b, &t);
/* copy */
fp_copy(a, c);
/* shift by as many digits in the bit count */
if (b >= (int)DIGIT_BIT) fp_rshd (c, b >> 5);
/* shift any bit count < DIGIT_BIT */
D = (fp_digit) (b & 31);
if (D != 0) {
fp_digit *tmpc, mask, shift;
/* mask */
mask = (((fp_digit)1) << D) - 1;
/* shift for lsb */
shift = DIGIT_BIT - D;
/* alias */
tmpc = c->dp + (c->used - 1);
/* carry */
r = 0;
for (x = c->used - 1; x >= 0; x--) {
/* get the lower bits of this word in a temp */
rr = *tmpc & mask;
/* shift the current word and mix in the carry bits from the previous word */
*tmpc = (*tmpc >> D) | (r << shift);
--tmpc;
/* set the carry to the carry bits of the current word found above */
r = rr;
}
}
fp_clamp (c);
if (d != NULL) fp_copy (&t, d);
}
/* a/b => cb + d == a */
__device__ __host__ void fp_div( fp_int *a, fp_int *b, fp_int *c, fp_int *d )
{
fp_int q, x, y, t1, t2;
int n, t, i, norm;
/* if a < b then q=0, r = a */
if( fp_cmp_mag(a, b) == FP_LT ) {
if (d != NULL) fp_copy(a, d);
if (c != NULL) fp_zero(c);
return;
}
fp_init(&q);
q.used = a->used + 2;
fp_init(&t1);
fp_init(&t2);
fp_copy(a, &x);
fp_copy(b, &y);
/* normalize both x and y, ensure that y >= b/2, [b == 2**DIGIT_BIT] */
norm = fp_count_bits(&y) & 31;
if (norm < (int)(DIGIT_BIT-1)) {
norm = (DIGIT_BIT-1) - norm;
fp_mul_2d (&x, norm, &x);
fp_mul_2d (&y, norm, &y);
}
else {
norm = 0;
}
/* note hac does 0 based, so if used==5 then its 0,1,2,3,4, e.g. use 4 */
n = x.used - 1;
t = y.used - 1;
/* while (x >= y*b**n-t) do { q[n-t] += 1; x -= y*b**{n-t} } */
fp_lshd (&y, n - t); /* y = y*b**{n-t} */
while (fp_cmp (&x, &y) != FP_LT) {
++(q.dp[n - t]);
fp_sub (&x, &y, &x);
}
/* reset y by shifting it back down */
fp_rshd (&y, n - t);
/* step 3. for i from n down to (t + 1) */
for (i = n; i >= (t + 1); i--) {
if (i > x.used) continue;
/* step 3.1 if xi == yt then set q{i-t-1} to b-1,
* otherwise set q{i-t-1} to (xi*b + x{i-1})/yt */
if (x.dp[i] == y.dp[t]) {
q.dp[i - t - 1] = ((((fp_word)1) << DIGIT_BIT) - 1);
}
else {
fp_word tmp;
tmp = ((fp_word) x.dp[i]) << ((fp_word) DIGIT_BIT);
tmp |= ((fp_word) x.dp[i - 1]);
tmp /= ((fp_word) y.dp[t]);
q.dp[i - t - 1] = (fp_digit) (tmp);
}
/* while (q{i-t-1} * (yt * b + y{t-1})) >
xi * b**2 + xi-1 * b + xi-2
do q{i-t-1} -= 1;
*/
q.dp[i - t - 1] = (q.dp[i - t - 1] + 1);
do {
q.dp[i - t - 1] = (q.dp[i - t - 1] - 1);
/* find left hand */
fp_zero (&t1);
t1.dp[0] = (t - 1 < 0) ? 0 : y.dp[t - 1];
t1.dp[1] = y.dp[t];
t1.used = 2;
fp_mul_d (&t1, q.dp[i - t - 1], &t1);
/* find right hand */
t2.dp[0] = (i - 2 < 0) ? 0 : x.dp[i - 2];
t2.dp[1] = (i - 1 < 0) ? 0 : x.dp[i - 1];
t2.dp[2] = x.dp[i];
t2.used = 3;
} while (fp_cmp_mag(&t1, &t2) == FP_GT);
/* step 3.3 x = x - q{i-t-1} * y * b**{i-t-1} */
fp_mul_d (&y, q.dp[i - t - 1], &t1);
fp_lshd (&t1, i - t - 1);
fp_sub (&x, &t1, &x);
}
/* now q is the quotient and x is the remainder
* [which we have to normalize]
*/
if (c != NULL) {
fp_clamp (&q);
fp_copy (&q, c);
}
if (d != NULL) {
fp_div_2d (&x, norm, &x, NULL);
/* the following is a kludge, essentially we were seeing the right remainder but
with excess digits that should have been zero
*/
for (i = b->used; i < x.used; i++) x.dp[i] = 0;
fp_clamp(&x);
fp_copy (&x, d);
}
}
__device__ __host__ void fp_2expt( fp_int *a, int b )
{
int z;
fp_zero (a);
z = b >> 5;
/* set the used count of where the bit will go */
a->used = z + 1;
/* put the single bit in its place */
a->dp[z] = ((fp_digit)1) << (b & 31);
}
/* generic PxQ multiplier */
__device__ __host__ void fp_mul_comba(fp_int *A, fp_int *B, fp_int *C)
{
int ix, iy, iz, tx, ty, pa;
fp_digit c0, c1, c2, *tmpx, *tmpy;
fp_int tmp, *dst;
c0 = c1 = c2 = 0;
/* get size of output and trim */
pa = A->used + B->used;
if (pa >= FP_SIZE) pa = FP_SIZE-1;
if (A == C || B == C) {
fp_zero(&tmp);
dst = &tmp;
}
else {
fp_zero(C);
dst = C;
}
for (ix = 0; ix < pa; ix++) {
/* get offsets into the two bignums */
ty = MIN(ix, B->used-1);
tx = ix - ty;
/* setup temp aliases */
tmpx = A->dp + tx;
tmpy = B->dp + ty;
/* this is the number of times the loop will iterrate, essentially its
while (tx++ < a->used && ty-- >= 0) { ... }
*/
iy = MIN(A->used-tx, ty+1);
/* execute loop */
c0 = c1; c1 = c2; c2 = 0;
for (iz = 0; iz < iy; ++iz) {
fp_word t;
t = (fp_word)c0 + ((fp_word)*tmpx++) * ((fp_word)*tmpy--);
c0 = t;
t = (fp_word)c1 + (t >> DIGIT_BIT);
c1 = t;
c2 += t >> DIGIT_BIT;
}
/* store term */
dst->dp[ix] = c0;
}
dst->used = pa;
fp_clamp(dst);
fp_copy(dst, C);
}
__device__ __host__ void fp_read_unsigned_bin(fp_int *a, unsigned char *b, int c)
{
/* zero the int */
fp_zero (a);
unsigned char *pd = (unsigned char *)a->dp;
if ((unsigned)c > (FP_SIZE * sizeof(fp_digit))) {
int excess = c - (FP_SIZE * sizeof(fp_digit));
c -= excess;
b += excess;
}
a->used = (c + sizeof(fp_digit) - 1)/sizeof(fp_digit);
/* read the bytes in */
for (c -= 1; c >= 0; c -= 1) {
pd[c] = *b++;
}
fp_clamp (a);
}
__device__ __host__ void fp_reverse( unsigned char *s, int len )
{
int ix, iy;
unsigned char t;
ix = 0;
iy = len - 1;
while (ix < iy) {
t = s[ix];
s[ix] = s[iy];
s[iy] = t;
++ix;
--iy;
}
}
__device__ __host__ void fp_to_unsigned_bin( fp_int *a, unsigned char *b )
{
int x;
fp_int t;
fp_copy(a, &t);
x = 0;
while (fp_iszero (&t) == FP_NO) {
b[x++] = (unsigned char) (t.dp[0] & 255);
fp_div_2d (&t, 8, &t, NULL);
}
fp_reverse (b, x);
}
__device__ __host__ void fp_reduce_setup( fp_int *a, fp_int *b )
{
fp_2expt (a, b->used << 6);
fp_div (a, b, a, NULL);
}
/* compare against a single digit */
__device__ __host__ int fp_cmp_d( fp_int *a, fp_digit b )
{
/* compare based on magnitude */
if (a->used > 1) return FP_GT;
/* compare the only digit of a to b */
if (a->dp[0] > b) return FP_GT;
else if (a->dp[0] < b) return FP_LT;
return FP_EQ;
}
__device__ __host__ void fp_reduce( fp_int *x, fp_int *m, fp_int *mu )
{
fp_int q;
int um = m->used;
if( fp_cmp (x, m) == FP_LT ) return;
/* q = x */
fp_copy (x, &q);
/* q1 = x / b**(k-1) */
fp_rshd (&q, um - 1);
fp_mul (&q, mu, &q);
/* q3 = q2 / b**(k+1) */
fp_rshd (&q, um + 1);
/* x = x mod b**(k+1), quick (no division) */
fp_mod_2d (x, ((um + 1) << 5), x);
/* q = q * m mod b**(k+1), quick (no division) */
//s_fp_mul_digs (&q, m, &q, um + 1);
// FIXME: actually multiply up to required digits instead of full multiply + truncation
fp_mul(&q, m, &q);
q.used = um + 1;
for( int i = q.used; i < FP_SIZE; i++ ) q.dp[i] = 0;
/* x = x - q */
fp_sub (x, &q, x);
/* If x < 0, add b**(k+1) to it */
// FIXME: does this ever happen..?
if (fp_cmp_d (x, 0) == FP_LT) {
fp_set (&q, 1);
fp_lshd (&q, um + 1);
fp_add (x, &q, x);
}
int count = 0;
/* Back off if it's too big */
while( fp_cmp (x, m) != FP_LT ) {
s_fp_sub (x, m, x);
count++;
if( count > 100 ) {
//printf("FUCK ME I GOT STUCK IN THE REDUCE WHILE LOOP, AGAIN!!!!\n");
break;
}
}
}
+291
View File
@@ -0,0 +1,291 @@
#include <stdint.h>
#include <stdio.h>
#include <memory.h>
#include "../cuda_helper.h"
__constant__ uint64_t c_PaddedMessage[32];
__constant__ uint32_t c_sigma[16][16];
const uint32_t host_sigma[16][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 },
{ 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 }
};
static __device__ uint32_t HIWORD(const uint64_t x)
{
uint32_t result;
asm(
"{\n\t"
".reg .u32 xl; \n\t"
"mov.b64 {xl,%0},%1; \n\t"
"}" : "=r"(result) : "l"(x)
);
return result;
}
static __device__ uint32_t LOWORD(const uint64_t x)
{
uint32_t result;
asm(
"{\n\t"
".reg .u32 xh; \n\t"
"mov.b64 {%0,xh},%1; \n\t"
"}" : "=r"(result) : "l"(x)
);
return result;
}
__device__ __forceinline__ uint64_t SWAP64( const uint64_t x)
{
uint64_t temp[2];
temp[0] = __byte_perm(HIWORD(x), 0, 0x0123);
temp[1] = __byte_perm(LOWORD(x), 0, 0x0123);
return temp[0] | (temp[1]<<32);
}
__host__ uint64_t host_SWAP64( uint64_t x )
{
return
((x & 0xff00000000000000) >> 56) |
((x & 0x00ff000000000000) >> 40) |
((x & 0x0000ff0000000000) >> 24) |
((x & 0x000000ff00000000) >> 8) |
((x & 0x00000000ff000000) << 8) |
((x & 0x0000000000ff0000) << 24) |
((x & 0x000000000000ff00) << 40) |
((x & 0x00000000000000ff) << 56);
}
__constant__ uint64_t c_u512[16];
const uint64_t host_u512[16] =
{
0x243f6a8885a308d3ULL, 0x13198a2e03707344ULL,
0xa4093822299f31d0ULL, 0x082efa98ec4e6c89ULL,
0x452821e638d01377ULL, 0xbe5466cf34e90c6cULL,
0xc0ac29b7c97c50ddULL, 0x3f84d5b5b5470917ULL,
0x9216d5d98979fb1bULL, 0xd1310ba698dfb5acULL,
0x2ffd72dbd01adfb7ULL, 0xb8e1afed6a267e96ULL,
0xba7c9045f12c7f99ULL, 0x24a19947b3916cf7ULL,
0x0801f2e2858efc16ULL, 0x636920d871574e69ULL
};
__forceinline__ __device__ uint64_t ROTR(const uint64_t value, const int offset) {
uint64_t result;
#if __CUDA_ARCH__ >= 320
if(offset < 32) {
asm("{\n\t"
" .reg .u32 tl,th,vl,vh; \n\t"
"mov.b64 {tl,th},%1; \n\t"
"shf.r.wrap.b32 vl,tl,th,%2; \n\t"
"shf.r.wrap.b32 vh,th,tl,%2; \n\t"
"mov.b64 %0,{vl,vh}; \n\t"
"}"
: "=l"(result) : "l"(value) , "r"(offset));
} else {
asm("{\n\t"
" .reg .u32 tl,th,vl,vh; \n\t"
"mov.b64 {tl,th},%1; \n\t"
"shf.r.wrap.b32 vh,tl,th,%2; \n\t"
"shf.r.wrap.b32 vl,th,tl,%2; \n\t"
"mov.b64 %0,{vl,vh}; \n\t"
"}"
: "=l"(result) : "l"(value) , "r"(offset));
}
#else
result = (value >> offset) | (value << (64-offset));
#endif
return result;
}
__device__ __forceinline__ uint64_t ROTR64_32( uint64_t val )
{
asm("{\n\t"
".reg .u32 tl,th; \n\t"
"mov.b64 {tl,th},%1; \n\t"
"mov.b64 %0,{th,tl}; \n\t"
"}"
: "=l"(val) : "l"(val));
return val;
}
__device__ __forceinline__ uint64_t ROTR64_16( uint64_t val )
{
asm("{\n\t"
".reg .u16 b0, b1, b2, b3; \n\t"
"mov.b64 { b0, b1, b2, b3 }, %1; \n\t"
"mov.b64 %0, {b1, b2, b3, b0}; \n\t"
"}"
: "=l"(val) : "l"(val));
return val;
}
#define G(a,b,c,d,e) \
v[a] += (m[c_sigma[i][e]] ^ c_u512[c_sigma[i][e+1]]) + v[b]; \
v[d] = ROTR64_32(v[d] ^ v[a]); \
v[c] += v[d]; \
v[b] = ROTR( v[b] ^ v[c],25); \
v[a] += (m[c_sigma[i][e+1]] ^ c_u512[c_sigma[i][e]])+v[b]; \
v[d] = ROTR64_16(v[d] ^ v[a]); \
v[c] += v[d]; \
v[b] = ROTR( v[b] ^ v[c],11); \
__device__ void blake_compress( uint64_t *h, const uint64_t *m, const int bits )
{
uint64_t v[16];
#pragma unroll 8
for( int i = 0; i < 8; ++i ) v[i] = h[i];
v[ 8] = c_u512[0];
v[ 9] = c_u512[1];
v[10] = c_u512[2];
v[11] = c_u512[3];
v[12] = c_u512[4] ^ bits;
v[13] = c_u512[5] ^ bits;
v[14] = c_u512[6];
v[15] = c_u512[7];
#pragma unroll 2
for( int i = 0; i < 16; ++i )
{
/* column step */
G( 0, 4, 8, 12, 0 );
G( 1, 5, 9, 13, 2 );
G( 2, 6, 10, 14, 4 );
G( 3, 7, 11, 15, 6 );
/* diagonal step */
G( 0, 5, 10, 15, 8 );
G( 1, 6, 11, 12, 10 );
G( 2, 7, 8, 13, 12 );
G( 3, 4, 9, 14, 14 );
}
#pragma unroll 8
for( int i = 0; i < 8; ++i ) h[i] ^= v[i] ^ v[i+8];
}
static __device__ uint32_t cuda_swap32(uint32_t x)
{
return __byte_perm(x, 0, 0x0123);
}
static __constant__ uint64_t d_IV[8];
static const uint64_t h_IV[8] = {
0x6a09e667f3bcc908ULL,
0xbb67ae8584caa73bULL,
0x3c6ef372fe94f82bULL,
0xa54ff53a5f1d36f1ULL,
0x510e527fade682d1ULL,
0x9b05688c2b3e6c1fULL,
0x1f83d9abfb41bd6bULL,
0x5be0cd19137e2179ULL
};
__global__ void blake_gpu_hash_185(uint32_t threads, uint32_t startNonce, uint32_t *outputHash, uint32_t *g_signature, uint32_t *g_hashwholeblock)
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
// bestimme den aktuellen Zähler
uint32_t nonce = startNonce + thread;
uint32_t idx64 = thread >> 6;
uint64_t h[8];
uint64_t buf[16];
uint64_t *hashwholeblock = (uint64_t *)&g_hashwholeblock[idx64 << 3];
uint64_t *signature = (uint64_t *)&g_signature[idx64 << 3];
#pragma unroll 8
for(int i=0;i<8;i++)
h[i] = d_IV[i];
#pragma unroll 11
for (int i=0; i < 11; ++i) buf[i] = SWAP64(c_PaddedMessage[i]);
((uint32_t *)buf)[20] = cuda_swap32(nonce);
#pragma unroll 4
for (int i=0; i < 4; ++i) buf[i+11] = SWAP64(hashwholeblock[i]);
buf[15] = SWAP64(c_PaddedMessage[15]);
blake_compress( h, buf, 1024 );
#pragma unroll 16
for (int i=0; i < 16; ++i) buf[i] = c_PaddedMessage[i+16];
#pragma unroll 32
for (int i=0; i < 32; ++i) ((unsigned char *)buf)[25+i] = ((unsigned char *)signature)[i];//SWAP64(signature[i]);
for (int i=0; i < 16; ++i) buf[i] = SWAP64(buf[i]);
blake_compress( h, buf, 1480 );
uint32_t *outHash = (uint32_t *)outputHash + 16 * thread;
#pragma unroll 8
for (int i=0; i < 8; ++i) {
outHash[2*i+0] = cuda_swap32( HIWORD(h[i]) );
outHash[2*i+1] = cuda_swap32( LOWORD(h[i]) );
}
}
}
__host__
void blake_cpu_init(int thr_id, uint32_t threads)
{
cudaMemcpyToSymbol(c_sigma, host_sigma, sizeof(host_sigma), 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(c_u512, host_u512, sizeof(host_u512), 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(d_IV, h_IV, sizeof(h_IV), 0, cudaMemcpyHostToDevice);
}
__host__
void blake_cpu_setBlock_185(void *pdata)
{
unsigned char PaddedMessage[256];
memcpy(PaddedMessage, pdata, 185);
memset(PaddedMessage+185, 0, 71);
PaddedMessage[185] = 0x80;
PaddedMessage[239] = 1;
PaddedMessage[254] = 0x05;
PaddedMessage[255] = 0xC8;
//for( int i = 0; i < 32; i++ ) ((uint64_t *)PaddedMessage)[i] = host_SWAP64(((uint64_t *)PaddedMessage)[i]);
cudaMemcpyToSymbol(c_PaddedMessage, PaddedMessage, 32*sizeof(uint64_t), 0, cudaMemcpyHostToDevice);
}
__host__
void blake_cpu_hash_185(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_outputHash, uint32_t *d_signature, uint32_t *d_hashwholeblock)
{
const int threadsperblock = 64;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
size_t shared_size = 0;
blake_gpu_hash_185<<<grid, block, shared_size>>>(threads, startNonce, d_outputHash, d_signature, d_hashwholeblock);
cudaDeviceSynchronize();
}
+501
View File
@@ -0,0 +1,501 @@
#include <stdio.h>
#include <stdint.h>
#include <memory.h>
#include "miner.h"
#include "spreadx11.h"
__constant__ uint32_t c_PaddedMessage88[32]; // padded message, 88 bytes + padding to 128 bytes
// precalculated state for sha256double_88
static __constant__ uint32_t d_state[8];
// init vector
static __constant__ uint32_t d_IV[8];
static const uint32_t h_IV[8] = {
0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a,
0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19
};
// round constants
static __constant__ uint32_t d_K[64];
static const uint32_t h_K[64] = {
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
};
// padding for 400000 bytes of input data
static __constant__ uint32_t padding400000[16] = {
0x80000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x0030d400
};
// padding for 32 bytes of input data
static __constant__ uint32_t padding32[16] = {
0x80000000, 0x00000000, 0x00000000, 0x00000000,
0x00000000, 0x00000000, 0x00000000, 0x00000100
};
__device__ __host__ __forceinline__ uint32_t ROTR32( const uint32_t val, const size_t offset )
{
#if __CUDA_ARCH__ >= 320
return __funnelshift_r(val, val, offset);
#else
return (val >> offset) | (val << (32-offset));
#endif
}
static __device__ uint32_t cuda_swab32(uint32_t x)
{
return __byte_perm(x, x, 0x0123);
}
__host__ uint32_t bswap32( const uint32_t x )
{
return
((x & 0xff000000) >> 24 ) |
((x & 0x00ff0000) >> 8 ) |
((x & 0x0000ff00) << 8 ) |
((x & 0x000000ff) << 24 );
}
__device__ __host__ __forceinline__ uint32_t BSG2_0( const uint32_t x )
{
return
ROTR32(x, 2) ^
ROTR32(x, 13) ^
ROTR32(x, 22);
}
__device__ __host__ __forceinline__ uint32_t BSG2_1( const uint32_t x )
{
return
ROTR32(x, 6) ^
ROTR32(x, 11) ^
ROTR32(x, 25);
}
__device__ __host__ __forceinline__ uint32_t SSG2_0( const uint32_t x )
{
return
ROTR32(x, 7) ^
ROTR32(x, 18) ^
(x >> 3);
}
__device__ __host__ __forceinline__ uint32_t SSG2_1( const uint32_t x )
{
return
ROTR32(x, 17) ^
ROTR32(x, 19) ^
(x >> 10);
}
__device__ __host__ __forceinline__ uint32_t CH( const uint32_t x, const uint32_t y, const uint32_t z )
{
return ((y ^ z) & x) ^ z;
}
__device__ __host__ __forceinline__ uint32_t MAJ( const uint32_t x, const uint32_t y, const uint32_t z )
{
return (x & y) | ((x | y) & z);
}
__device__ __host__ __forceinline__ void sha2_step1( uint32_t *W, const uint32_t *K, const uint32_t a, const uint32_t b, const uint32_t c, uint32_t &d,
const uint32_t e, const uint32_t f, const uint32_t g, uint32_t &h, const uint32_t *in, const uint32_t pc, const uint32_t pcount )
{
W[pc] = in[pc];
uint32_t t1 =
h + BSG2_1(e) + CH(e, f, g) +
K[pcount + pc] + W[pc];
d += t1;
h = t1 + BSG2_0(a) + MAJ(a, b, c);
}
__device__ __host__ __forceinline__ void sha2_step2( uint32_t *W, const uint32_t *K, const uint32_t a, const uint32_t b, const uint32_t c, uint32_t &d,
const uint32_t e, const uint32_t f, const uint32_t g, uint32_t &h, const uint32_t *in, const uint32_t pc, const uint32_t pcount )
{
uint32_t t1;
W[pc] = SSG2_1(W[(pc - 2) & 0x0F])
+ W[(pc - 7) & 0x0F]
+ SSG2_0(W[(pc - 15) & 0x0F]) + W[pc];
t1 = h + BSG2_1(e) + CH(e, f, g)
+ K[pcount + (pc)] + W[(pc)];
d += t1;
h = t1 + BSG2_0(a) + MAJ(a, b, c);
}
__device__
void sha2_transform(const uint32_t *in, uint32_t *state)
{
uint32_t a, b, c, d, e, f, g, h;
uint32_t w[16];
uint32_t pcount;
a = state[0];
b = state[1];
c = state[2];
d = state[3];
e = state[4];
f = state[5];
g = state[6];
h = state[7];
pcount = 0;
sha2_step1(w, d_K, a, b, c, d, e, f, g, h, in, 0, pcount);
sha2_step1(w, d_K, h, a, b, c, d, e, f, g, in, 1, pcount);
sha2_step1(w, d_K, g, h, a, b, c, d, e, f, in, 2, pcount);
sha2_step1(w, d_K, f, g, h, a, b, c, d, e, in, 3, pcount);
sha2_step1(w, d_K, e, f, g, h, a, b, c, d, in, 4, pcount);
sha2_step1(w, d_K, d, e, f, g, h, a, b, c, in, 5, pcount);
sha2_step1(w, d_K, c, d, e, f, g, h, a, b, in, 6, pcount);
sha2_step1(w, d_K, b, c, d, e, f, g, h, a, in, 7, pcount);
sha2_step1(w, d_K, a, b, c, d, e, f, g, h, in, 8, pcount);
sha2_step1(w, d_K, h, a, b, c, d, e, f, g, in, 9, pcount);
sha2_step1(w, d_K, g, h, a, b, c, d, e, f, in, 10, pcount);
sha2_step1(w, d_K, f, g, h, a, b, c, d, e, in, 11, pcount);
sha2_step1(w, d_K, e, f, g, h, a, b, c, d, in, 12, pcount);
sha2_step1(w, d_K, d, e, f, g, h, a, b, c, in, 13, pcount);
sha2_step1(w, d_K, c, d, e, f, g, h, a, b, in, 14, pcount);
sha2_step1(w, d_K, b, c, d, e, f, g, h, a, in, 15, pcount);
for (pcount = 16; pcount < 64; pcount += 16) {
sha2_step2(w, d_K, a, b, c, d, e, f, g, h, in, 0, pcount);
sha2_step2(w, d_K, h, a, b, c, d, e, f, g, in, 1, pcount);
sha2_step2(w, d_K, g, h, a, b, c, d, e, f, in, 2, pcount);
sha2_step2(w, d_K, f, g, h, a, b, c, d, e, in, 3, pcount);
sha2_step2(w, d_K, e, f, g, h, a, b, c, d, in, 4, pcount);
sha2_step2(w, d_K, d, e, f, g, h, a, b, c, in, 5, pcount);
sha2_step2(w, d_K, c, d, e, f, g, h, a, b, in, 6, pcount);
sha2_step2(w, d_K, b, c, d, e, f, g, h, a, in, 7, pcount);
sha2_step2(w, d_K, a, b, c, d, e, f, g, h, in, 8, pcount);
sha2_step2(w, d_K, h, a, b, c, d, e, f, g, in, 9, pcount);
sha2_step2(w, d_K, g, h, a, b, c, d, e, f, in, 10, pcount);
sha2_step2(w, d_K, f, g, h, a, b, c, d, e, in, 11, pcount);
sha2_step2(w, d_K, e, f, g, h, a, b, c, d, in, 12, pcount);
sha2_step2(w, d_K, d, e, f, g, h, a, b, c, in, 13, pcount);
sha2_step2(w, d_K, c, d, e, f, g, h, a, b, in, 14, pcount);
sha2_step2(w, d_K, b, c, d, e, f, g, h, a, in, 15, pcount);
}
state[0] += a;
state[1] += b;
state[2] += c;
state[3] += d;
state[4] += e;
state[5] += f;
state[6] += g;
state[7] += h;
}
__host__
void host_sha2_transform(const uint32_t *in, uint32_t *state)
{
uint32_t a, b, c, d, e, f, g, h;
uint32_t w[16];
uint32_t pcount;
a = state[0];
b = state[1];
c = state[2];
d = state[3];
e = state[4];
f = state[5];
g = state[6];
h = state[7];
pcount = 0;
sha2_step1(w, h_K, a, b, c, d, e, f, g, h, in, 0, pcount);
sha2_step1(w, h_K, h, a, b, c, d, e, f, g, in, 1, pcount);
sha2_step1(w, h_K, g, h, a, b, c, d, e, f, in, 2, pcount);
sha2_step1(w, h_K, f, g, h, a, b, c, d, e, in, 3, pcount);
sha2_step1(w, h_K, e, f, g, h, a, b, c, d, in, 4, pcount);
sha2_step1(w, h_K, d, e, f, g, h, a, b, c, in, 5, pcount);
sha2_step1(w, h_K, c, d, e, f, g, h, a, b, in, 6, pcount);
sha2_step1(w, h_K, b, c, d, e, f, g, h, a, in, 7, pcount);
sha2_step1(w, h_K, a, b, c, d, e, f, g, h, in, 8, pcount);
sha2_step1(w, h_K, h, a, b, c, d, e, f, g, in, 9, pcount);
sha2_step1(w, h_K, g, h, a, b, c, d, e, f, in, 10, pcount);
sha2_step1(w, h_K, f, g, h, a, b, c, d, e, in, 11, pcount);
sha2_step1(w, h_K, e, f, g, h, a, b, c, d, in, 12, pcount);
sha2_step1(w, h_K, d, e, f, g, h, a, b, c, in, 13, pcount);
sha2_step1(w, h_K, c, d, e, f, g, h, a, b, in, 14, pcount);
sha2_step1(w, h_K, b, c, d, e, f, g, h, a, in, 15, pcount);
for (pcount = 16; pcount < 64; pcount += 16) {
sha2_step2(w, h_K, a, b, c, d, e, f, g, h, in, 0, pcount);
sha2_step2(w, h_K, h, a, b, c, d, e, f, g, in, 1, pcount);
sha2_step2(w, h_K, g, h, a, b, c, d, e, f, in, 2, pcount);
sha2_step2(w, h_K, f, g, h, a, b, c, d, e, in, 3, pcount);
sha2_step2(w, h_K, e, f, g, h, a, b, c, d, in, 4, pcount);
sha2_step2(w, h_K, d, e, f, g, h, a, b, c, in, 5, pcount);
sha2_step2(w, h_K, c, d, e, f, g, h, a, b, in, 6, pcount);
sha2_step2(w, h_K, b, c, d, e, f, g, h, a, in, 7, pcount);
sha2_step2(w, h_K, a, b, c, d, e, f, g, h, in, 8, pcount);
sha2_step2(w, h_K, h, a, b, c, d, e, f, g, in, 9, pcount);
sha2_step2(w, h_K, g, h, a, b, c, d, e, f, in, 10, pcount);
sha2_step2(w, h_K, f, g, h, a, b, c, d, e, in, 11, pcount);
sha2_step2(w, h_K, e, f, g, h, a, b, c, d, in, 12, pcount);
sha2_step2(w, h_K, d, e, f, g, h, a, b, c, in, 13, pcount);
sha2_step2(w, h_K, c, d, e, f, g, h, a, b, in, 14, pcount);
sha2_step2(w, h_K, b, c, d, e, f, g, h, a, in, 15, pcount);
}
state[0] += a;
state[1] += b;
state[2] += c;
state[3] += d;
state[4] += e;
state[5] += f;
state[6] += g;
state[7] += h;
}
__global__
void spreadx11_sha256_gpu_hash_wholeblock(uint32_t threads, uint32_t startNonce, uint32_t *g_input, uint32_t *g_hash, uint32_t *g_signature)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t state[8];
uint32_t input[16];
uint32_t *outputHash = (uint32_t *)&g_hash[thread<<3];
uint8_t *signature = (uint8_t *)&g_signature[thread<<3];
#pragma unroll 8
for( int i=0; i < 8; i++ ) state[i] = d_IV[i];
/*
Input is as follows, we need to replace nonce and MinerSignature for each hash.
This means the first two blocks need to be handled separately.
4 bytes (nNonce & ~NONCE_MASK)
8 bytes nTime;
65 bytes MinerSignature;
4 bytes nVersion;
and so on....
*/
// we process the same 200000 bytes of data twice
for( int round = 0; round < 2; round++ ) {
// insert nonce
input[0] = startNonce + (thread << 6);
// read the rest of the block
#pragma unroll 15
for( int i=1; i < 16; i++ ) input[i] = g_input[i];
// first 19 bytes of signature go into the first block
#pragma unroll 19
for( int i = 0; i < 19; i++ ) ((uint8_t *)input)[45+i] = signature[i];
// endian swap the block and transform
#pragma unroll 16
for( int i=0; i < 16; i++ ) input[i] = cuda_swab32(input[i]);
sha2_transform(input, state);
// read the second block
#pragma unroll 16
for( int i=0; i < 16; i++ ) input[i] = g_input[i+16];
// replace in the remaining bytes of the signature
#pragma unroll 13
for( int i = 0; i < 13; i++ ) ((uint8_t *)input)[i] = signature[i+19];
// endian swap the second block and transform
#pragma unroll 16
for( int i=0; i < 16; i++ ) input[i] = cuda_swab32(input[i]);
sha2_transform(input, state);
// transform the remaining blocks as they are
for( int block = 2; block < 3125; block++ ) {
#pragma unroll 16
for( int i=0; i < 16; i++ ) input[i] = cuda_swab32(g_input[i+(block << 4)]);
sha2_transform(input, state);
}
}
// load and transform the padding block
#pragma unroll 16
for( int i = 0; i < 16; i++ ) input[i] = padding400000[i];
sha2_transform(input, state);
#pragma unroll 8
for( int i=0; i < 8; i++ ) outputHash[i] = cuda_swab32(state[i]);
}
}
__global__
void spreadx11_sha256double_gpu_hash_88(uint32_t threads, uint32_t startNonce, uint32_t *g_hash)
{
uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
uint32_t state[8];
uint32_t input[16];
uint32_t nonce = startNonce + (thread << 6);
uint32_t *outputHash = (uint32_t *)&g_hash[thread<<3];
// load precalculated state from the first block
#pragma unroll 8
for( int i = 0; i < 8; i++ ) state[i] = d_state[i];
#pragma unroll 16
for( int i = 0; i < 16; i++ ) input[i] = cuda_swab32(c_PaddedMessage88[i+16]);
input[5] = cuda_swab32(nonce);
sha2_transform(input, state);
// finished hashing the input data, next step: hash the first result hash
#pragma unroll 8
for( int i = 0; i < 8; i++ ) {
input[i] = state[i];
input[i+8] = padding32[i];
state[i] = d_IV[i];
}
sha2_transform(input, state);
#pragma unroll 8
for( int i = 0; i < 8; i++ ) outputHash[i] = cuda_swab32(state[i]);
}
}
__host__
void spreadx11_sha256_cpu_init(int thr_id, uint32_t threads)
{
cudaMemcpyToSymbol(d_K, h_K, sizeof(h_K), 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(d_IV, h_IV, sizeof(h_IV), 0, cudaMemcpyHostToDevice);
}
__host__
void spreadx11_sha256double_cpu_hash_88(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_hash)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
spreadx11_sha256double_gpu_hash_88<<<grid, block>>>(threads, startNonce, d_hash);
cudaDeviceSynchronize();
}
__host__
void spreadx11_sha256double_setBlock_88(void *data)
{
unsigned char PaddedMessage[128];
uint32_t tmpstate[8], tmpinput[16];
memcpy(PaddedMessage, data, 88);
memset(PaddedMessage+88, 0, 40);
PaddedMessage[ 88] = 0x80;
PaddedMessage[126] = 0x02;
PaddedMessage[127] = 0xC0;
// we can transform the first block here and start the CUDA kernel with the precalculated state
for( int i = 0; i < 8; i++ ) tmpstate[i] = h_IV[i];
for( int i = 0; i < 16; i++ ) tmpinput[i] = bswap32(((uint32_t *)PaddedMessage)[i]);
host_sha2_transform(tmpinput, tmpstate);
cudaMemcpyToSymbol(d_state, tmpstate, 8*sizeof(uint32_t), 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(c_PaddedMessage88, PaddedMessage, 128, 0, cudaMemcpyHostToDevice);
}
__host__
void spreadx11_sha256_setBlock_wholeblock(struct work *work, uint32_t *d_wholeblock)
{
unsigned char paddedmessage[200000];
struct spreadwork *swork = (struct spreadwork *) work->extradata;
/*
We need a hash of:
offset size field
0 4 nNonce & 0xffffffc0
4 8 nTime
12 65 MinerSignature
77 4 nVersion
81 32 hashPrevBlock
113 32 hashMerkleRoot
145 4 nBits
149 4 nHeight
153 x transactions, variable length
padding calculated backwards from the copy of hashPrevBlock at the end
199968 32 hashPrevBlock
We have in work->data:
offset size field data
0 4 nVersion 02000000
4 32 hashPrevBlock 83ac9d7e96c4bb15d1a9a2fe3dbc81747768125009efc14e3824fbacaec65fad
36 32 HashMerkleRoot 4e06a8366a959df705ea34dfe7eafea0d8a9e6fe3ba6b83afddb145b27da2362
68 8 nTime ed71585400000000
76 4 nBits 19850d1d
80 4 nHeight 33d20100
84 4 nNonce 00000000
88 32 hashWholeBlock 0000000000000000000000000000000000000000000000000000000000000000
120 33 MinerSignature1 1cad95a3e3e3598b74a1b8e2e3c013ababf6c45c83534029dbe8b4698916f78255
153 32 MinerSignature2 0000000000000000000000000000000000000000000000000000000000000000
*/
memcpy(paddedmessage+0, swork->data+84, 4); // nNonce
memcpy(paddedmessage+4, swork->data+68, 8); // nTime
memcpy(paddedmessage+12, swork->data+120, 65); // MinerSignature
memcpy(paddedmessage+77, swork->data+0, 4); // nVersion
memcpy(paddedmessage+81, swork->data+4, 32); // hashPrevBlock
memcpy(paddedmessage+113, swork->data+36, 32); // HashMerkleRoot
memcpy(paddedmessage+145, swork->data+76, 4); // nBits
memcpy(paddedmessage+149, swork->data+80, 4); // nHeight
memcpy(paddedmessage+153, swork->tx, swork->txsize); // tx
// the total amount of bytes in our data
int blocksize = 153 + swork->txsize;
// pad the block with 0x07 bytes to make it a multiple of uint32_t
while( blocksize & 3 ) paddedmessage[blocksize++] = 0x07;
// figure out the offsets for the padding
uint32_t *pFillBegin = (uint32_t*)&paddedmessage[blocksize];
uint32_t *pFillEnd = (uint32_t*)&paddedmessage[MAX_BLOCK_SIZE]; // FIXME: isn't this out of bounds by one... but it seems to work out...
uint32_t *pFillFooter = pFillBegin > pFillEnd - 8 ? pFillBegin : pFillEnd - 8;
memcpy(pFillFooter, swork->data+4, (pFillEnd - pFillFooter)*4);
for (uint32_t *pI = pFillFooter; pI < pFillEnd; pI++)
*pI |= 1;
for (uint32_t *pI = pFillFooter - 1; pI >= pFillBegin; pI--)
pI[0] = pI[3]*pI[7];
// copy the whole monstrosity into device memory for processing
cudaMemcpy(d_wholeblock, paddedmessage, 200000, cudaMemcpyHostToDevice);
}
__host__
void spreadx11_sha256_cpu_hash_wholeblock(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_hash, uint32_t *d_signature, uint32_t *d_wholeblock)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
spreadx11_sha256_gpu_hash_wholeblock<<<grid, block>>>(threads, startNonce, d_wholeblock, d_hash, d_signature);
cudaDeviceSynchronize();
}
+97
View File
@@ -0,0 +1,97 @@
#include <stdint.h>
#include <stdio.h>
#include <memory.h>
#include "miner.h"
#include "bignum_fp.h"
#include "spreadx11.h"
__constant__ uint32_t d_privkey_data[8]; // 32 bytes
__constant__ uint32_t d_kinv_data[8]; // 32 bytes
__constant__ uint32_t d_m_data[8] = {
0xd0364141, 0xbfd25e8c, 0xaf48a03b, 0xbaaedce6,
0xfffffffe, 0xffffffff, 0xffffffff, 0xffffffff
};
__constant__ uint32_t d_mu_data[9] = {
0x2fc9bec0, 0x402da173, 0x50b75fc4, 0x45512319,
0x00000001, 0x00000000, 0x00000000, 0x00000000,
0x00000001
};
__global__
void spreadx11_sign_gpu(uint32_t threads, uint32_t startNonce, uint32_t *g_sha256hash, uint32_t *g_signature )
{
const uint32_t thread = (blockDim.x * blockIdx.x + threadIdx.x);
if (thread < threads)
{
fp_int hash, result, m, mu, privkey, kinv;
unsigned char *sha256hash = (unsigned char *)&g_sha256hash[thread<<3];
uint32_t *signature = (uint32_t *)&g_signature[thread<<3];
fp_zero(&result);
fp_read_unsigned_bin(&hash, sha256hash, 32);
privkey.used = 8;
kinv.used = 8;
m.used = 8;
mu.used = 9;
#pragma unroll 8
for( int i = 0; i < 8; i++ ) {
privkey.dp[i] = d_privkey_data[i];
kinv.dp[i] = d_kinv_data[i];
m.dp[i] = d_m_data[i];
mu.dp[i] = d_mu_data[i];
signature[i] = 0;
}
mu.dp[8] = d_mu_data[8];
m.dp[8] = 0;
#pragma unroll
for( int i = 9; i < FP_SIZE; i++ ) {
m.dp[i] = 0;
mu.dp[i] = 0;
}
fp_add(&hash, &privkey, &result);
fp_mul(&result, &kinv, &result);
fp_reduce(&result, &m, &mu);
fp_to_unsigned_bin(&result, &((unsigned char *)signature)[32-(fp_count_bits(&result)+7)/8]);
}
}
__host__
void spreadx11_sign_cpu_init(int thr_id, uint32_t throughput)
{
}
__host__
void spreadx11_sign_cpu_setInput(struct work *work)
{
fp_int fp_privkey, fp_kinv;
struct spreadwork* swork = (struct spreadwork*) work->extradata;
fp_read_unsigned_bin(&fp_privkey, swork->privkey, 32);
fp_read_unsigned_bin(&fp_kinv, swork->kinv, 32);
cudaMemcpyToSymbol(d_privkey_data, fp_privkey.dp, 32, 0, cudaMemcpyHostToDevice);
cudaMemcpyToSymbol(d_kinv_data, fp_kinv.dp, 32, 0, cudaMemcpyHostToDevice);
}
__host__
void spreadx11_sign_cpu(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_hash, uint32_t *d_signature)
{
const uint32_t threadsperblock = 256;
dim3 grid((threads + threadsperblock-1)/threadsperblock);
dim3 block(threadsperblock);
spreadx11_sign_gpu<<<grid, block>>>(threads, startNonce, d_hash, d_signature);
cudaDeviceSynchronize();
}
+540
View File
@@ -0,0 +1,540 @@
/**
* From the SpreadX11 spreadminer Cuda Implementation by tsiv
*
* Adapted to common ccminer by tpruvot@github - 2015
*/
#include <string.h>
#ifndef _WIN32
#include <unistd.h>
#endif
extern "C"
{
#include "sph/sph_blake.h"
#include "sph/sph_bmw.h"
#include "sph/sph_groestl.h"
#include "sph/sph_skein.h"
#include "sph/sph_jh.h"
#include "sph/sph_keccak.h"
#include "sph/sph_luffa.h"
#include "sph/sph_cubehash.h"
#include "sph/sph_shavite.h"
#include "sph/sph_simd.h"
#include "sph/sph_echo.h"
#include <openssl/bn.h>
#include <openssl/sha.h>
}
#include "miner.h"
#include "spreadx11.h"
#include "cuda_helper.h"
static uint32_t *d_sha256hash[MAX_GPUS];
static uint32_t *d_signature[MAX_GPUS];
static uint32_t *d_hashwholeblock[MAX_GPUS];
static uint32_t *d_hash[MAX_GPUS];
static uint32_t *d_wholeblockdata[MAX_GPUS];
extern void spreadx11_sha256double_cpu_hash_88(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_hash);
extern void spreadx11_sha256double_setBlock_88(void *data);
extern void spreadx11_sha256_cpu_init( int thr_id, uint32_t threads);
extern void spreadx11_sha256_cpu_hash_wholeblock(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_hash, uint32_t *d_signature, uint32_t *d_wholeblock);
extern void spreadx11_sha256_setBlock_wholeblock(struct work *work, uint32_t *d_wholeblock);
extern void spreadx11_sign_cpu_init(int thr_id, uint32_t threads);
extern void spreadx11_sign_cpu_setInput(struct work *work);
extern void spreadx11_sign_cpu(int thr_id, uint32_t threads, uint32_t startNonce, uint32_t *d_hash, uint32_t *d_signature);
extern void blake_cpu_init(int thr_id, uint32_t threads);
extern void blake_cpu_setBlock_185(void *pdata);
extern void blake_cpu_hash_185(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_hash, uint32_t *d_signature, uint32_t *d_hashwholeblock);
extern void quark_bmw512_cpu_init(int thr_id, uint32_t threads);
extern void quark_bmw512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_groestl512_cpu_init(int thr_id, uint32_t threads);
extern void quark_groestl512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_skein512_cpu_init(int thr_id, uint32_t threads);
extern void quark_skein512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_keccak512_cpu_init(int thr_id, uint32_t threads);
extern void quark_keccak512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_jh512_cpu_init(int thr_id, uint32_t threads);
extern void quark_jh512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void x11_luffa512_cpu_init(int thr_id, uint32_t threads);
extern void x11_luffa512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void x11_cubehash512_cpu_init(int thr_id, uint32_t threads);
extern void x11_cubehash512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void x11_shavite512_cpu_init(int thr_id, uint32_t threads);
extern void x11_shavite512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void x11_simd512_cpu_init(int thr_id, uint32_t threads);
extern void x11_simd512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
//extern void spreadx11_echo512_cpu_init(int thr_id, uint32_t threads);
//extern void spreadx11_echo512_cpu_setTarget(const void *ptarget);
//extern uint32_t spreadx11_echo512_cpu_hash_64_final(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, uint32_t *hashidx);
extern void x11_echo512_cpu_init(int thr_id, uint32_t threads);
extern void x11_echo512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
#define PROFILE 0
#if PROFILE == 1
#define PRINTTIME(s) do { \
double duration; \
gettimeofday(&tv_end, NULL); \
duration = 1e-6 * (tv_end.tv_usec-tv_start.tv_usec) + (tv_end.tv_sec-tv_start.tv_sec); \
printf("%s: %.2f ms, %.2f MH/s\n", s, duration*1000.0, (double)throughput / 1000000.0 / duration); \
} while(0)
#define TRACE(algo) { \
if (nonce == first_nonce) { \
uint32_t* debugbuf = NULL; \
cudaMallocHost(&debugbuf, 8*sizeof(uint32_t)); \
cudaMemcpy(debugbuf, d_hash[thr_id], 32, cudaMemcpyDeviceToHost); \
printf(" %s %08x %08x %08x %08x...\n", algo, swab32(debugbuf[0]), swab32(debugbuf[1]), \
swab32(debugbuf[2]), swab32(debugbuf[3])); \
cudaFreeHost(debugbuf); \
} \
}
#else
#define PRINTTIME(s)
#define TRACE(s)
#endif
void hextobin(unsigned char *p, const char *hexstr, size_t len)
{
char hex_byte[3];
char *ep;
hex_byte[2] = '\0';
while (*hexstr && len) {
if (!hexstr[1]) {
applog(LOG_ERR, "hex2bin str truncated");
return;
}
hex_byte[0] = hexstr[0];
hex_byte[1] = hexstr[1];
*p = (unsigned char) strtol(hex_byte, &ep, 16);
if (*ep) {
applog(LOG_ERR, "hex2bin failed on '%s'", hex_byte);
return;
}
p++;
hexstr += 2;
len--;
}
}
static void spreadx11_hash(void *output, struct work *work, uint32_t nonce)
{
SHA256_CTX ctx_sha;
sph_blake512_context ctx_blake;
sph_bmw512_context ctx_bmw;
sph_groestl512_context ctx_groestl;
sph_jh512_context ctx_jh;
sph_keccak512_context ctx_keccak;
sph_skein512_context ctx_skein;
sph_luffa512_context ctx_luffa;
sph_cubehash512_context ctx_cubehash;
sph_shavite512_context ctx_shavite;
sph_simd512_context ctx_simd;
sph_echo512_context ctx_echo;
unsigned char mod[32] = {
0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xff, 0xfe,
0xba, 0xae, 0xdc, 0xe6, 0xaf, 0x48, 0xa0, 0x3b, 0xbf, 0xd2, 0x5e, 0x8c, 0xd0, 0x36, 0x41, 0x41
};
unsigned char tmp[185];
unsigned char hashforsig[32];
unsigned char wholeblock[MAX_BLOCK_SIZE];
unsigned char finalhash[64];
BIGNUM *bn_hash, *bn_privkey, *bn_kinv, *bn_mod, *bn_res;
BN_CTX *bn_ctx;
struct spreadwork* swork = (struct spreadwork*) work->extradata;
uint32_t *nonceptr = (uint32_t *)&tmp[84];
unsigned char *hashwholeblockptr = &tmp[88];
unsigned char *signatureptr = &tmp[153];
memcpy(tmp, swork->data, 185);
nonceptr[0] = nonce & 0xffffffc0;
SHA256_Init(&ctx_sha);
SHA256_Update(&ctx_sha, tmp, 88);
SHA256_Final(hashforsig, &ctx_sha);
SHA256_Init(&ctx_sha);
SHA256_Update(&ctx_sha, hashforsig, 32);
SHA256_Final(hashforsig, &ctx_sha);
bn_ctx = BN_CTX_new();
bn_hash = BN_new();
bn_privkey = BN_new();
bn_kinv = BN_new();
bn_mod = BN_new();
bn_res = BN_new();
BN_bin2bn(hashforsig, 32, bn_hash);
BN_bin2bn(swork->privkey, 32, bn_privkey);
BN_bin2bn(swork->kinv, 32, bn_kinv);
BN_bin2bn(mod, 32, bn_mod);
BN_mod_add_quick(bn_privkey, bn_privkey, bn_hash, bn_mod);
BN_mod_mul(bn_res, bn_privkey, bn_kinv, bn_mod, bn_ctx);
int nBitsS = BN_num_bits(bn_res);
memset(signatureptr, 0, 32);
BN_bn2bin(bn_res, &signatureptr[32-(nBitsS+7)/8]);
BN_CTX_free(bn_ctx);
BN_clear_free(bn_hash);
BN_clear_free(bn_privkey);
BN_clear_free(bn_kinv);
BN_clear_free(bn_mod);
BN_clear_free(bn_res);
memcpy(wholeblock+0, tmp+84, 4); // nNonce
memcpy(wholeblock+4, tmp+68, 8); // nTime
memcpy(wholeblock+12, tmp+120, 65); // MinerSignature
memcpy(wholeblock+77, tmp+0, 4); // nVersion
memcpy(wholeblock+81, tmp+4, 32); // hashPrevBlock
memcpy(wholeblock+113, tmp+36, 32); // HashMerkleRoot
memcpy(wholeblock+145, tmp+76, 4); // nBits
memcpy(wholeblock+149, tmp+80, 4); // nHeight
memcpy(wholeblock+153, swork->tx, swork->txsize); // tx
// the total amount of bytes in our data
int blocksize = 153 + swork->txsize;
// pad the block with 0x07 bytes to make it a multiple of uint32_t
while( blocksize & 3 ) wholeblock[blocksize++] = 0x07;
// figure out the offsets for the padding
uint32_t *pFillBegin = (uint32_t*)&wholeblock[blocksize];
uint32_t *pFillEnd = (uint32_t*)&wholeblock[MAX_BLOCK_SIZE]; // FIXME: isn't this out of bounds by one... but it seems to work out...
uint32_t *pFillFooter = pFillBegin > pFillEnd - 8 ? pFillBegin : pFillEnd - 8;
memcpy(pFillFooter, tmp+4, (pFillEnd - pFillFooter)*4);
for (uint32_t *pI = pFillFooter; pI < pFillEnd; pI++)
*pI |= 1;
for (uint32_t *pI = pFillFooter - 1; pI >= pFillBegin; pI--)
pI[0] = pI[3]*pI[7];
SHA256_Init(&ctx_sha);
SHA256_Update(&ctx_sha, wholeblock, MAX_BLOCK_SIZE);
SHA256_Update(&ctx_sha, wholeblock, MAX_BLOCK_SIZE);
SHA256_Final(hashwholeblockptr, &ctx_sha);
//applog_hash(hashwholeblockptr);
nonceptr[0] = nonce;
sph_blake512_init(&ctx_blake);
sph_blake512 (&ctx_blake, tmp, 185);
sph_blake512_close(&ctx_blake, (void*) finalhash);
//applog_hash(finalhash);
sph_bmw512_init(&ctx_bmw);
sph_bmw512 (&ctx_bmw, (const void*) finalhash, 64);
sph_bmw512_close(&ctx_bmw, (void*) finalhash);
sph_groestl512_init(&ctx_groestl);
sph_groestl512 (&ctx_groestl, (const void*) finalhash, 64);
sph_groestl512_close(&ctx_groestl, (void*) finalhash);
sph_skein512_init(&ctx_skein);
sph_skein512 (&ctx_skein, (const void*) finalhash, 64);
sph_skein512_close(&ctx_skein, (void*) finalhash);
sph_jh512_init(&ctx_jh);
sph_jh512 (&ctx_jh, (const void*) finalhash, 64);
sph_jh512_close(&ctx_jh, (void*) finalhash);
sph_keccak512_init(&ctx_keccak);
sph_keccak512 (&ctx_keccak, (const void*) finalhash, 64);
sph_keccak512_close(&ctx_keccak, (void*) finalhash);
sph_luffa512_init(&ctx_luffa);
sph_luffa512 (&ctx_luffa, (const void*) finalhash, 64);
sph_luffa512_close (&ctx_luffa, (void*) finalhash);
sph_cubehash512_init(&ctx_cubehash);
sph_cubehash512 (&ctx_cubehash, (const void*) finalhash, 64);
sph_cubehash512_close(&ctx_cubehash, (void*) finalhash);
sph_shavite512_init(&ctx_shavite);
sph_shavite512 (&ctx_shavite, (const void*) finalhash, 64);
sph_shavite512_close(&ctx_shavite, (void*) finalhash);
sph_simd512_init(&ctx_simd);
sph_simd512 (&ctx_simd, (const void*) finalhash, 64);
sph_simd512_close(&ctx_simd, (void*) finalhash);
sph_echo512_init(&ctx_echo);
sph_echo512 (&ctx_echo, (const void*) finalhash, 64);
sph_echo512_close(&ctx_echo, (void*) finalhash);
memcpy(output, finalhash, 32);
}
// cpu hash only used as dev test (--cputest)
extern "C" void spreadhash(void *output, const void *input)
{
struct work work;
struct spreadwork swork;
memset(&work, 0, sizeof(struct work));
memset(&swork, 0, sizeof(struct spreadwork));
memcpy(swork.data, input, 128); // 185 but made for a simple empty
const char hextest[371] = { 0 };
hex2bin(swork.data, &hextest[0], 185);
work.extradata = (void*) &swork;
uint32_t *nonceptr = (uint32_t*) (&swork.data[84]);
spreadx11_hash(output, &work, nonceptr[0]);
}
static bool init[MAX_GPUS] = { 0 };
// GPU Search
extern "C" int scanhash_spreadx11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
int dev_id = device_map[thr_id % MAX_GPUS];
// multiple of 64 to keep things simple with signatures
//const uint32_t throughput = opt_throughput * 1024 * 64;
uint32_t *ptarget = work->target;
struct spreadwork* swork = (struct spreadwork*) work->extradata;
if (!swork) {
work->extradata = swork = (struct spreadwork*) calloc(1, sizeof(struct spreadwork));
applog(LOG_ERR, "GPU #%d: missing block data!", dev_id);
sleep(1);
//return -1;
}
unsigned char *blocktemplate = swork->data;
uint32_t *pnonce = (uint32_t*) (&blocktemplate[84]);
uint32_t nonce = (*pnonce);
uint32_t first_nonce = nonce;
int intensity = (device_sm[dev_id] >= 500 && !is_windows()) ? 19 : 18;
uint32_t throughput = device_intensity(thr_id, __func__, 1U << intensity); // 19=256*256*8;
throughput = min(throughput, max_nonce - first_nonce) & 0xFFFFFFC0UL;
if (opt_benchmark)
ptarget[7] = 0x00ff;
if (!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
//cudaDeviceReset();
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
cudaDeviceSetCacheConfig(cudaFuncCachePreferL1);
// a 512-bit buffer for every nonce to hold the x11 intermediate hashes
CUDA_SAFE_CALL(cudaMalloc(&d_hash[thr_id], (size_t)64 * throughput));
// sha256 hashes used for signing, 32 bytes for every 64 nonces
cudaMalloc(&d_sha256hash[thr_id], (size_t)32*(throughput>>6));
// changing part of MinerSignature, 32 bytes for every 64 nonces
cudaMalloc(&d_signature[thr_id], (size_t)32*(throughput>>6));
// sha256 hashes for the whole block, 32 bytes for every 64 nonces
cudaMalloc(&d_hashwholeblock[thr_id], (size_t)32*(throughput>>6));
spreadx11_sha256_cpu_init(thr_id, throughput);
spreadx11_sign_cpu_init(thr_id, throughput);
blake_cpu_init(thr_id, throughput);
quark_groestl512_cpu_init(thr_id, throughput);
quark_skein512_cpu_init(thr_id, throughput);
quark_bmw512_cpu_init(thr_id, throughput);
quark_keccak512_cpu_init(thr_id, throughput);
quark_jh512_cpu_init(thr_id, throughput);
x11_luffa512_cpu_init(thr_id, throughput);
x11_cubehash512_cpu_init(thr_id, throughput);
x11_shavite512_cpu_init(thr_id, throughput);
x11_simd512_cpu_init(thr_id, throughput);
//spreadx11_echo512_cpu_init(thr_id, throughput);
x11_echo512_cpu_init(thr_id, throughput);
cuda_check_cpu_init(thr_id, throughput);
// single buffer to hold the padded whole block data
CUDA_SAFE_CALL(cudaMalloc(&d_wholeblockdata[thr_id], 200000));
init[thr_id] = true;
}
struct timeval tv_start;
#if PROFILE == 1
struct timeval tv_end;
#endif
spreadx11_sign_cpu_setInput(work);
spreadx11_sha256_setBlock_wholeblock(work, d_wholeblockdata[thr_id]);
spreadx11_sha256double_setBlock_88((void *)blocktemplate);
blake_cpu_setBlock_185((void *)blocktemplate);
//spreadx11_echo512_cpu_setTarget(ptarget);
cuda_check_cpu_setTarget(ptarget);
do {
int order = 0;
gettimeofday(&tv_start, NULL);
spreadx11_sha256double_cpu_hash_88(thr_id, throughput>>6, nonce, d_sha256hash[thr_id]);
//PRINTTIME("sha256 sig");
gettimeofday(&tv_start, NULL);
spreadx11_sign_cpu(thr_id, throughput>>6, nonce, d_sha256hash[thr_id], d_signature[thr_id]);
//PRINTTIME("signing ");
gettimeofday(&tv_start, NULL);
spreadx11_sha256_cpu_hash_wholeblock(thr_id, throughput>>6, nonce, d_hashwholeblock[thr_id], d_signature[thr_id], d_wholeblockdata[thr_id]);
//PRINTTIME("hashwhole ");
gettimeofday(&tv_start, NULL);
blake_cpu_hash_185(thr_id, throughput, nonce, d_hash[thr_id], d_signature[thr_id], d_hashwholeblock[thr_id]);
TRACE("blake ");
gettimeofday(&tv_start, NULL);
quark_bmw512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("bmw ");
gettimeofday(&tv_start, NULL);
quark_groestl512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("groestl ");
gettimeofday(&tv_start, NULL);
quark_skein512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("skein ");
gettimeofday(&tv_start, NULL);
quark_jh512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("jh ");
gettimeofday(&tv_start, NULL);
quark_keccak512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("keccak ");
gettimeofday(&tv_start, NULL);
x11_luffa512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("luffa ");
gettimeofday(&tv_start, NULL);
x11_cubehash512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("cubehash ");
gettimeofday(&tv_start, NULL);
x11_shavite512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("shavite ");
gettimeofday(&tv_start, NULL);
x11_simd512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("simd ");
uint32_t winnerthread = 0;
uint32_t foundNonce;
gettimeofday(&tv_start, NULL);
//foundNonce = spreadx11_echo512_cpu_hash_64_final(thr_id, throughput, nonce, NULL, d_hash[thr_id], &winnerthread);
x11_echo512_cpu_hash_64(thr_id, throughput, nonce, NULL, d_hash[thr_id], order++);
TRACE("echo ");
*hashes_done = nonce - first_nonce + throughput;
foundNonce = cuda_check_hash(thr_id, throughput, nonce, d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
uint32_t cpuhash[8], hash[8];
char hexbuffer[MAX_BLOCK_SIZE*2] = { 0 };
uint32_t *resnonce = (uint32_t*) (&swork->data[84]);
uint32_t *reshashwholeblock = (uint32_t*) (&swork->data[88]);
uint32_t *ressignature = (uint32_t*) (&swork->data[153]);
for (uint32_t u = 0; u < swork->txsize && u < MAX_BLOCK_SIZE; u++)
sprintf(&hexbuffer[u*2], "%02x", swork->tx[u]);
winnerthread = (foundNonce - nonce);
uint32_t idx64 = winnerthread >> 6;
if (opt_debug)
applog(LOG_DEBUG,
"Thread %d found a solution\n"
"First nonce : %08x\n"
"Found nonce : %08x\n"
"Threadidx : %d\n"
"Threadidx64 : %d\n"
"VTX (%d) : %s\n",
dev_id, first_nonce, foundNonce, winnerthread, idx64, swork->txsize, hexbuffer);
else applog(LOG_INFO, "GPU #%d: found a solution, nonce %08x", dev_id, foundNonce);
*resnonce = foundNonce;
cudaMemcpy(reshashwholeblock, d_hashwholeblock[thr_id] + idx64 * 8, 32, cudaMemcpyDeviceToHost);
cudaMemcpy(ressignature, d_signature[thr_id] + idx64 * 8, 32, cudaMemcpyDeviceToHost);
cudaMemcpy(hash, d_hash[thr_id] + winnerthread * 16, 32, cudaMemcpyDeviceToHost);
if (opt_debug) {
memset(hexbuffer, 0, sizeof(hexbuffer));
for( int i = 0; i < 32; i++ ) sprintf(&hexbuffer[i*2], "%02x", ((uint8_t *)hash)[i]);
applog(LOG_DEBUG, "Final hash 256 : %s", hexbuffer);
memset(hexbuffer, 0, sizeof(hexbuffer));
for( int i = 0; i < 185; i++ ) sprintf(&hexbuffer[i*2], "%02x", ((uint8_t *)swork->data)[i]);
applog(LOG_DEBUG, "Submit data : %s", hexbuffer);
memset(hexbuffer, 0, sizeof(hexbuffer));
for( int i = 0; i < 32; i++ ) sprintf(&hexbuffer[i*2], "%02x", ((uint8_t *)reshashwholeblock)[i]);
applog(LOG_DEBUG, "HashWholeBlock : %s", hexbuffer);
memset(hexbuffer, 0, sizeof(hexbuffer));
for( int i = 0; i < 32; i++ ) sprintf(&hexbuffer[i*2], "%02x", ((uint8_t *)ressignature)[i]);
applog(LOG_DEBUG, "MinerSignature : %s", hexbuffer);
}
spreadx11_hash((void *)cpuhash, work, foundNonce);
if(cpuhash[7] == hash[7] && fulltest(hash, ptarget) ) {
bn_store_hash_target_ratio(cpuhash, ptarget, work);
return 1;
} else if(cpuhash[7] != hash[7]) {
applog(LOG_WARNING, "GPU #%d: result for nonce %08x does not validate on CPU %08x<>%08x!", dev_id, foundNonce, cpuhash[7], hash[7]);
applog_hash((uchar*)cpuhash);
} else {
applog(LOG_WARNING, "GPU #%d: result for nonce %08x does not pass the full test!", dev_id, foundNonce);
}
}
nonce += throughput;
} while (nonce < max_nonce && !work_restart[thr_id].restart);
return 0;
}
// cleanup
extern "C" void free_spreadx11(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+9
View File
@@ -0,0 +1,9 @@
#define MAX_BLOCK_SIZE 200000
struct spreadwork {
unsigned char data[185];
unsigned char privkey[32];
unsigned char kinv[32];
unsigned char tx[MAX_BLOCK_SIZE];
size_t txsize;
};
+97 -37
View File
@@ -38,6 +38,7 @@
extern pthread_mutex_t stratum_sock_lock;
extern pthread_mutex_t stratum_work_lock;
extern bool opt_debug_diff;
bool opt_tracegpu = false;
@@ -105,13 +106,10 @@ void applog(int prio, const char *fmt, ...)
const char* color = "";
char *f;
int len;
struct tm tm, *tm_p;
struct tm tm;
time_t now = time(NULL);
pthread_mutex_lock(&applog_lock);
tm_p = localtime(&now);
memcpy(&tm, tm_p, sizeof(tm));
pthread_mutex_unlock(&applog_lock);
localtime_r(&now, &tm);
switch (prio) {
case LOG_ERR: color = CL_RED; break;
@@ -404,7 +402,7 @@ static json_t *json_rpc_call(CURL *curl, const char *url,
char *httpdata;
char len_hdr[64], hashrate_hdr[64];
char curl_err_str[CURL_ERROR_SIZE] = { 0 };
long timeout = longpoll ? opt_timeout*2 : opt_timeout;
long timeout = longpoll ? opt_timeout : opt_timeout/2;
struct header_info hi = { 0 };
bool lp_scanning = longpoll_scan && !have_longpoll;
@@ -598,11 +596,56 @@ json_t *json_rpc_longpoll(CURL *curl, char *lp_url, struct pool_infos *pool, con
strlen(pool->pass)?':':'\0', pool->pass);
// on pool rotate by time-limit, this keepalive can be a problem
bool keepalive = pool->time_limit == 0 || pool->time_limit > opt_timeout*4;
bool keepalive = pool->time_limit == 0 || pool->time_limit > opt_timeout;
return json_rpc_call(curl, lp_url, userpass, req, false, true, keepalive, curl_err);
}
json_t *json_load_url(char* cfg_url, json_error_t *err)
{
char err_str[CURL_ERROR_SIZE] = { 0 };
struct data_buffer all_data = { 0 };
int rc = 0; json_t *cfg = NULL;
CURL *curl = curl_easy_init();
if (unlikely(!curl)) {
applog(LOG_ERR, "Remote config init failed!");
return NULL;
}
curl_easy_setopt(curl, CURLOPT_URL, cfg_url);
curl_easy_setopt(curl, CURLOPT_FRESH_CONNECT, 1);
curl_easy_setopt(curl, CURLOPT_CONNECTTIMEOUT, 15);
curl_easy_setopt(curl, CURLOPT_ERRORBUFFER, err_str);
curl_easy_setopt(curl, CURLOPT_NOSIGNAL, 1);
curl_easy_setopt(curl, CURLOPT_TCP_NODELAY, 1);
curl_easy_setopt(curl, CURLOPT_WRITEFUNCTION, all_data_cb);
curl_easy_setopt(curl, CURLOPT_WRITEDATA, &all_data);
if (opt_proxy) {
curl_easy_setopt(curl, CURLOPT_PROXY, opt_proxy);
curl_easy_setopt(curl, CURLOPT_PROXYTYPE, opt_proxy_type);
} else if (getenv("http_proxy")) {
if (getenv("all_proxy"))
curl_easy_setopt(curl, CURLOPT_PROXY, getenv("all_proxy"));
else if (getenv("ALL_PROXY"))
curl_easy_setopt(curl, CURLOPT_PROXY, getenv("ALL_PROXY"));
else
curl_easy_setopt(curl, CURLOPT_PROXY, "");
}
rc = curl_easy_perform(curl);
if (rc) {
applog(LOG_ERR, "Remote config read failed: %s", err_str);
goto err_out;
}
if (!all_data.buf || !all_data.len) {
applog(LOG_ERR, "Empty data received for config");
goto err_out;
}
cfg = JSON_LOADS((char*)all_data.buf, err);
err_out:
curl_easy_cleanup(curl);
return cfg;
}
/**
* Unlike malloc, calloc set the memory to zero
*/
@@ -723,7 +766,7 @@ bool fulltest(const uint32_t *hash, const uint32_t *target)
}
}
if (!rc && opt_debug) {
if ((!rc && opt_debug) || opt_debug_diff) {
uint32_t hash_be[8], target_be[8];
char *hash_str, *target_str;
@@ -747,11 +790,15 @@ bool fulltest(const uint32_t *hash, const uint32_t *target)
return rc;
}
void diff_to_target(uint32_t *target, double diff)
// Only used by stratum pools
void diff_to_target(struct work* work, double diff)
{
uint32_t *target = work->target;
uint64_t m;
int k;
work->targetdiff = diff;
for (k = 6; k > 0 && diff > 1.0; k--)
diff /= 4294967296.0;
m = (uint64_t)(4294901760.0 / diff);
@@ -764,6 +811,26 @@ void diff_to_target(uint32_t *target, double diff)
}
}
// Only used by longpoll pools
double target_to_diff(uint32_t* target)
{
uchar* tgt = (uchar*) target;
uint64_t m =
(uint64_t)tgt[29] << 56 |
(uint64_t)tgt[28] << 48 |
(uint64_t)tgt[27] << 40 |
(uint64_t)tgt[26] << 32 |
(uint64_t)tgt[25] << 24 |
(uint64_t)tgt[24] << 16 |
(uint64_t)tgt[23] << 8 |
(uint64_t)tgt[22] << 0;
if (!m)
return 0.;
else
return (double)0x0000ffff00000000/m;
}
#ifdef WIN32
#define socket_blocks() (WSAGetLastError() == WSAEWOULDBLOCK)
#else
@@ -853,6 +920,7 @@ char *stratum_recv_line(struct stratum_ctx *sctx)
{
ssize_t len, buflen;
char *tok, *sret = NULL;
int timeout = opt_timeout;
if (!sctx->sockbuf)
return NULL;
@@ -860,7 +928,7 @@ char *stratum_recv_line(struct stratum_ctx *sctx)
if (!strstr(sctx->sockbuf, "\n")) {
bool ret = true;
time_t rstart = time(NULL);
if (!socket_full(sctx->sock, opt_timeout)) {
if (!socket_full(sctx->sock, timeout)) {
applog(LOG_ERR, "stratum_recv_line timed out");
goto out;
}
@@ -881,7 +949,7 @@ char *stratum_recv_line(struct stratum_ctx *sctx)
}
} else
stratum_buffer_append(sctx, s);
} while (time(NULL) - rstart < opt_timeout && !strstr(sctx->sockbuf, "\n"));
} while (time(NULL) - rstart < timeout && !strstr(sctx->sockbuf, "\n"));
if (!ret) {
applog(LOG_ERR, "stratum_recv_line failed");
@@ -1756,33 +1824,19 @@ void do_gpu_tests(void)
unsigned long done;
char s[128] = { '\0' };
uchar buf[160];
uint32_t tgt[8] = { 0 };
struct work work;
memset(&work, 0, sizeof(work));
opt_tracegpu = true;
work_restart = (struct work_restart*) malloc(sizeof(struct work_restart));
work_restart[0].restart = 1;
tgt[7] = 0xffff;
work.target[7] = 0xffff;
//memset(buf, 0, sizeof buf);
//scanhash_skeincoin(0, (uint32_t*)buf, tgt, 1, &done);
//memset(work.data, 0, sizeof(work.data));
//scanhash_spreadx11(0, &work, 1, &done);
//memset(buf, 0, sizeof buf);
//memcpy(buf, zrtest, 80);
//scanhash_zr5(0, (uint32_t*)buf, tgt, zrtest[19]+1, &done);
//struct timeval tv;
//memset(buf, 0, sizeof buf);
//scanhash_scrypt_jane(0, (uint32_t*)buf, tgt, NULL, 1, &done, &tv, &tv);
memset(buf, 0, sizeof buf);
scanhash_x11(0, (uint32_t*)buf, tgt, 1, &done);
//memset(buf, 0, sizeof buf);
// buf[0] = 1; buf[64] = 2; // for endian tests
//scanhash_blake256(0, (uint32_t*)buf, tgt, 1, &done, 14);
//memset(buf, 0, sizeof buf);
//scanhash_heavy(0, (uint32_t*)buf, tgt, 1, &done, 1, 84); // HEAVYCOIN_BLKHDR_SZ=84
memset(work.data, 0, sizeof(work.data));
scanhash_lyra2(0, &work, 1, &done);
free(work_restart);
work_restart = NULL;
@@ -1805,15 +1859,15 @@ void print_hash_tests(void)
printf(CL_WHT "CPU HASH ON EMPTY BUFFER RESULTS:" CL_N "\n");
animehash(&hash[0], &buf[0]);
printpfx("anime", hash);
blake256hash(&hash[0], &buf[0], 8);
printpfx("blakecoin", hash);
blake256hash(&hash[0], &buf[0], 14);
printpfx("blake", hash);
bmw_hash(&hash[0], &buf[0]);
printpfx("bmw", hash);
c11hash(&hash[0], &buf[0]);
printpfx("c11", hash);
@@ -1841,9 +1895,12 @@ void print_hash_tests(void)
luffa_hash(&hash[0], &buf[0]);
printpfx("luffa", hash);
lyra2_hash(&hash[0], &buf[0]);
lyra2re_hash(&hash[0], &buf[0]);
printpfx("lyra2", hash);
lyra2v2_hash(&hash[0], &buf[0]);
printpfx("lyra2v2", hash);
myriadhash(&hash[0], &buf[0]);
printpfx("myriad", hash);
@@ -1874,6 +1931,9 @@ void print_hash_tests(void)
skein2hash(&hash[0], &buf[0]);
printpfx("skein2", hash);
spreadhash(&hash[0], &buf[0]);
printpfx("spread", hash);
s3hash(&hash[0], &buf[0]);
printpfx("S3", hash);
+26 -5
View File
@@ -54,9 +54,6 @@ extern void x11_simd512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t start
extern void x11_echo512_cpu_init(int thr_id, uint32_t threads);
extern void x11_echo512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse, int order);
// Flax/C11 CPU Hash
extern "C" void c11hash(void *output, const void *input)
@@ -142,8 +139,10 @@ extern "C" void c11hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_c11(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uint32_t max_nonce, unsigned long *hashes_done)
extern "C" int scanhash_c11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int intensity = (device_sm[device_map[thr_id]] >= 500 && !is_windows()) ? 20 : 19;
uint32_t throughput = device_intensity(thr_id, __func__, 1U << intensity); // 19=256*256*8;
@@ -208,6 +207,8 @@ extern "C" int scanhash_c11(int thr_id, uint32_t *pdata, const uint32_t *ptarget
x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
TRACE("echo => ");
*hashes_done = pdata[19] - first_nonce + throughput;
foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -220,8 +221,12 @@ extern "C" int scanhash_c11(int thr_id, uint32_t *pdata, const uint32_t *ptarget
int res = 1;
// check if there was some other ones...
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
c11hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -240,3 +245,19 @@ extern "C" int scanhash_c11(int thr_id, uint32_t *pdata, const uint32_t *ptarget
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_c11(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+25 -8
View File
@@ -25,10 +25,6 @@ extern void x11_simd512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t start
extern void x11_echo512_cpu_init(int thr_id, uint32_t threads);
extern void x11_echo512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse,
int order);
// CPU Hash
extern "C" void fresh_hash(void *state, const void *input)
@@ -70,10 +66,10 @@ extern "C" void fresh_hash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_fresh(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_fresh(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t endiandata[20];
@@ -122,6 +118,7 @@ extern "C" int scanhash_fresh(int thr_id, uint32_t *pdata,
CUDA_SAFE_CALL(cudaThreadSynchronize());
print_hash((unsigned char*)buf); printf("\n");
#endif
*hashes_done = pdata[19] - first_nonce + throughput;
foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
@@ -133,8 +130,12 @@ extern "C" int scanhash_fresh(int thr_id, uint32_t *pdata,
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
fresh_hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -152,3 +153,19 @@ extern "C" int scanhash_fresh(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_fresh(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+26 -4
View File
@@ -52,10 +52,10 @@ extern "C" void s3hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
/* Main S3 entry point */
extern "C" int scanhash_s3(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_s3(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int intensity = 20; // 256*256*8*2;
#ifdef WIN32
@@ -99,6 +99,8 @@ extern "C" int scanhash_s3(int thr_id, uint32_t *pdata,
x11_simd512_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[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
@@ -110,8 +112,12 @@ extern "C" int scanhash_s3(int thr_id, uint32_t *pdata,
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
s3hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -130,3 +136,19 @@ extern "C" int scanhash_s3(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_s3(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+24 -7
View File
@@ -54,10 +54,6 @@ extern void x11_simd512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t start
extern void x11_echo512_cpu_init(int thr_id, uint32_t threads);
extern void x11_echo512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse, int order);
// X11 CPU Hash
extern "C" void x11hash(void *output, const void *input)
{
@@ -142,10 +138,10 @@ extern "C" void x11hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_x11(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_x11(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int intensity = (device_sm[device_map[thr_id]] >= 500 && !is_windows()) ? 20 : 19;
uint32_t throughput = device_intensity(thr_id, __func__, 1U << intensity); // 19=256*256*8;
@@ -222,8 +218,13 @@ extern "C" int scanhash_x11(int thr_id, uint32_t *pdata,
int res = 1;
// check if there was some other ones...
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
bn_store_hash_target_ratio(vhash64, ptarget, work);
*hashes_done = pdata[19] - first_nonce + throughput;
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
x11hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -242,3 +243,19 @@ extern "C" int scanhash_x11(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_x11(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+32 -12
View File
@@ -62,9 +62,6 @@ extern void x13_hamsi512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t star
extern void x13_fugue512_cpu_init(int thr_id, uint32_t threads);
extern void x13_fugue512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse, int order);
// X13 CPU Hash
extern "C" void x13hash(void *output, const void *input)
@@ -145,10 +142,10 @@ extern "C" void x13hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_x13(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_x13(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
int intensity = 19; // (device_sm[device_map[thr_id]] > 500 && !is_windows()) ? 20 : 19;
uint32_t throughput = device_intensity(thr_id, __func__, 1 << intensity); // 19=256*256*8;
@@ -207,23 +204,30 @@ extern "C" int scanhash_x13(int thr_id, uint32_t *pdata,
x13_hamsi512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x13_fugue512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
const uint32_t Htarg = ptarget[7];
uint32_t vhash64[8];
uint32_t vhash[8];
be32enc(&endiandata[19], foundNonce);
x13hash(vhash64, endiandata);
x13hash(vhash, endiandata);
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash, ptarget, work);
pdata[19] = foundNonce;
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
x13hash(vhash, endiandata);
pdata[21] = secNonce;
if (bn_hash_target_ratio(vhash, ptarget) > work->shareratio) {
bn_store_hash_target_ratio(vhash, ptarget, work);
xchg(pdata[19], pdata[21]);
}
res++;
}
pdata[19] = foundNonce;
return res;
} else {
applog(LOG_WARNING, "GPU #%d: result for %08x does not validate on CPU!", device_map[thr_id], foundNonce);
@@ -241,3 +245,19 @@ extern "C" int scanhash_x13(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_x13(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+9
View File
@@ -544,6 +544,15 @@ extern void whirlpoolx_cpu_init(int thr_id, uint32_t threads)
CUDA_SAFE_CALL(cudaMalloc(&d_tmp[thr_id], 8 * 9 * sizeof(uint64_t))); // d_tmp[threadIdx.x+64] (7+64)
}
__host__
extern void whirlpoolx_cpu_free(int thr_id)
{
cudaFree(d_WXNonce[thr_id]);
cudaFreeHost(h_wxnounce[thr_id]);
cudaFree(d_xtra[thr_id]);
cudaFree(d_tmp[thr_id]);
}
__host__
void whirlpoolx_setBlock_80(void *pdata, const void *ptarget)
{
+24 -3
View File
@@ -12,6 +12,7 @@ extern "C" {
static uint32_t *d_hash[MAX_GPUS];
extern void whirlpoolx_cpu_init(int thr_id, uint32_t threads);
extern void whirlpoolx_cpu_free(int thr_id);
extern void whirlpoolx_setBlock_80(void *pdata, const void *ptarget);
extern uint32_t whirlpoolx_cpu_hash(int thr_id, uint32_t threads, uint32_t startNounce);
extern void whirlpoolx_precompute(int thr_id);
@@ -37,9 +38,10 @@ extern "C" void whirlxHash(void *state, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_whirlpoolx(int thr_id, uint32_t *pdata, const uint32_t *ptarget,
uint32_t max_nonce, unsigned long *hashes_done)
extern "C" int scanhash_whirlx(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t endiandata[20];
int intensity = is_windows() ? 20 : 22;
@@ -65,6 +67,7 @@ extern "C" int scanhash_whirlpoolx(int thr_id, uint32_t *pdata, const uint32_t *
whirlpoolx_setBlock_80((void*)endiandata, ptarget);
whirlpoolx_precompute(thr_id);
do {
uint32_t foundNonce = whirlpoolx_cpu_hash(thr_id, throughput, pdata[19]);
if (foundNonce != UINT32_MAX)
@@ -74,8 +77,10 @@ extern "C" int scanhash_whirlpoolx(int thr_id, uint32_t *pdata, const uint32_t *
be32enc(&endiandata[19], foundNonce);
whirlxHash(vhash64, endiandata);
*hashes_done = pdata[19] - first_nonce + throughput;
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[19] = foundNonce;
return 1;
} else {
@@ -95,3 +100,19 @@ extern "C" int scanhash_whirlpoolx(int thr_id, uint32_t *pdata, const uint32_t *
return 0;
}
// cleanup
extern "C" void free_whirlx(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
whirlpoolx_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+26 -7
View File
@@ -69,9 +69,6 @@ extern void x13_fugue512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t star
extern void x14_shabal512_cpu_init(int thr_id, uint32_t threads);
extern void x14_shabal512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse, int order);
// X14 CPU Hash function
extern "C" void x14hash(void *output, const void *input)
@@ -157,10 +154,10 @@ extern "C" void x14hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_x14(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_x14(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t endiandata[20];
@@ -216,6 +213,8 @@ extern "C" int scanhash_x14(int thr_id, uint32_t *pdata,
x13_fugue512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x14_shabal512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -228,8 +227,12 @@ extern "C" int scanhash_x14(int thr_id, uint32_t *pdata,
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
x14hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -246,3 +249,19 @@ extern "C" int scanhash_x14(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_x14(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+26 -7
View File
@@ -74,9 +74,6 @@ extern void x15_whirlpool_cpu_init(int thr_id, uint32_t threads, int mode);
extern void x15_whirlpool_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void x15_whirlpool_cpu_free(int thr_id);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse, int order);
// X15 CPU Hash function
extern "C" void x15hash(void *output, const void *input)
@@ -167,10 +164,10 @@ extern "C" void x15hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_x15(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_x15(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t endiandata[20];
@@ -228,6 +225,8 @@ extern "C" int scanhash_x15(int thr_id, uint32_t *pdata,
x14_shabal512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x15_whirlpool_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -240,8 +239,12 @@ extern "C" int scanhash_x15(int thr_id, uint32_t *pdata,
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
x15hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -261,3 +264,19 @@ extern "C" int scanhash_x15(int thr_id, uint32_t *pdata,
x15_whirlpool_cpu_free(thr_id);
return 0;
}
// cleanup
extern "C" void free_x15(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+26 -8
View File
@@ -82,10 +82,6 @@ extern void x17_sha512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startN
extern void x17_haval256_cpu_init(int thr_id, uint32_t threads);
extern void x17_haval256_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
extern void quark_compactTest_cpu_init(int thr_id, uint32_t threads);
extern void quark_compactTest_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *inpHashes,
uint32_t *d_noncesTrue, size_t *nrmTrue, uint32_t *d_noncesFalse, size_t *nrmFalse,
int order);
// X17 Hashfunktion
extern "C" void x17hash(void *output, const void *input)
@@ -186,10 +182,10 @@ extern "C" void x17hash(void *output, const void *input)
static bool init[MAX_GPUS] = { 0 };
extern "C" int scanhash_x17(int thr_id, uint32_t *pdata,
const uint32_t *ptarget, uint32_t max_nonce,
unsigned long *hashes_done)
extern "C" int scanhash_x17(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t throughput = device_intensity(thr_id, __func__, 1U << 19); // 19=256*256*8;
@@ -254,6 +250,8 @@ extern "C" int scanhash_x17(int thr_id, uint32_t *pdata,
x17_sha512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x17_haval256_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
*hashes_done = pdata[19] - first_nonce + throughput;
uint32_t foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
@@ -265,8 +263,12 @@ extern "C" int scanhash_x17(int thr_id, uint32_t *pdata,
if (vhash64[7] <= Htarg && fulltest(vhash64, ptarget)) {
int res = 1;
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1);
*hashes_done = pdata[19] - first_nonce + throughput;
bn_store_hash_target_ratio(vhash64, ptarget, work);
if (secNonce != 0) {
be32enc(&endiandata[19], secNonce);
x17hash(vhash64, endiandata);
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
res++;
}
@@ -284,3 +286,19 @@ extern "C" int scanhash_x17(int thr_id, uint32_t *pdata,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_x17(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}
+30
View File
@@ -435,6 +435,7 @@ extern "C" int scanhash_zr5(int thr_id, struct work *work,
zr5hash(vhash64, pdata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
int res = 1;
bn_store_hash_target_ratio(vhash64, ptarget, work);
uint32_t secNonce = cuda_check_hash_suppl(thr_id, throughput, oldp19, d_hash[thr_id], 1);
if (secNonce != 0) {
offset = secNonce - oldp19;
@@ -444,6 +445,8 @@ extern "C" int scanhash_zr5(int thr_id, struct work *work,
tmpdata[0] = pok; tmpdata[19] = secNonce;
zr5hash(vhash64, tmpdata);
if (vhash64[7] <= ptarget[7] && fulltest(vhash64, ptarget)) {
if (bn_hash_target_ratio(vhash64, ptarget) > work->shareratio)
bn_store_hash_target_ratio(vhash64, ptarget, work);
pdata[21] = secNonce;
pdata[22] = pok;
res++;
@@ -466,3 +469,30 @@ extern "C" int scanhash_zr5(int thr_id, struct work *work,
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}
// cleanup
extern "C" void free_zr5(int thr_id)
{
if (!init[thr_id])
return;
cudaSetDevice(device_map[thr_id]);
cudaFree(d_hash[thr_id]);
cudaFree(d_poks[thr_id]);
cudaFree(d_permut[thr_id]);
cudaFree(d_buffers[thr_id]);
cudaFree(d_blake[thr_id]);
cudaFree(d_groes[thr_id]);
cudaFree(d_jh512[thr_id]);
cudaFree(d_skein[thr_id]);
cudaFree(d_txs[thr_id]);
cuda_check_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}