GOSTcoin support for ccminer CUDA miner project, compatible with most nvidia cards
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

252 lines
7.7 KiB

/*
* X13 algorithm
*/
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 "sph/sph_hamsi.h"
#include "sph/sph_fugue.h"
}
#include "miner.h"
#include "cuda_helper.h"
#include "x11/cuda_x11.h"
static uint32_t *d_hash[MAX_GPUS];
extern void x13_hamsi512_cpu_init(int thr_id, uint32_t threads);
extern void x13_hamsi512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
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 x13_fugue512_cpu_free(int thr_id);
// X13 CPU Hash
extern "C" void x13hash(void *output, const void *input)
{
// blake1-bmw2-grs3-skein4-jh5-keccak6-luffa7-cubehash8-shavite9-simd10-echo11-hamsi12-fugue13
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;
sph_hamsi512_context ctx_hamsi;
sph_fugue512_context ctx_fugue;
uint32_t hash[32];
memset(hash, 0, sizeof hash);
sph_blake512_init(&ctx_blake);
sph_blake512 (&ctx_blake, input, 80);
sph_blake512_close(&ctx_blake, (void*) hash);
sph_bmw512_init(&ctx_bmw);
sph_bmw512 (&ctx_bmw, (const void*) hash, 64);
sph_bmw512_close(&ctx_bmw, (void*) hash);
sph_groestl512_init(&ctx_groestl);
sph_groestl512 (&ctx_groestl, (const void*) hash, 64);
sph_groestl512_close(&ctx_groestl, (void*) hash);
sph_skein512_init(&ctx_skein);
sph_skein512 (&ctx_skein, (const void*) hash, 64);
sph_skein512_close(&ctx_skein, (void*) hash);
sph_jh512_init(&ctx_jh);
sph_jh512 (&ctx_jh, (const void*) hash, 64);
sph_jh512_close(&ctx_jh, (void*) hash);
sph_keccak512_init(&ctx_keccak);
sph_keccak512 (&ctx_keccak, (const void*) hash, 64);
sph_keccak512_close(&ctx_keccak, (void*) hash);
sph_luffa512_init(&ctx_luffa);
sph_luffa512 (&ctx_luffa, (const void*) hash, 64);
sph_luffa512_close (&ctx_luffa, (void*) hash);
sph_cubehash512_init(&ctx_cubehash);
sph_cubehash512 (&ctx_cubehash, (const void*) hash, 64);
sph_cubehash512_close(&ctx_cubehash, (void*) hash);
sph_shavite512_init(&ctx_shavite);
sph_shavite512 (&ctx_shavite, (const void*) hash, 64);
sph_shavite512_close(&ctx_shavite, (void*) hash);
sph_simd512_init(&ctx_simd);
sph_simd512 (&ctx_simd, (const void*) hash, 64);
sph_simd512_close(&ctx_simd, (void*) hash);
sph_echo512_init(&ctx_echo);
sph_echo512 (&ctx_echo, (const void*) hash, 64);
sph_echo512_close(&ctx_echo, (void*) hash);
sph_hamsi512_init(&ctx_hamsi);
sph_hamsi512 (&ctx_hamsi, (const void*) hash, 64);
sph_hamsi512_close(&ctx_hamsi, (void*) hash);
sph_fugue512_init(&ctx_fugue);
sph_fugue512 (&ctx_fugue, (const void*) hash, 64);
sph_fugue512_close(&ctx_fugue, (void*) hash);
memcpy(output, hash, 32);
}
static bool init[MAX_GPUS] = { 0 };
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 = cuda_default_throughput(thr_id, 1 << intensity); // 19=256*256*8;
//if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x000f;
if (!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset();
// reduce cpu usage
cudaSetDeviceFlags(cudaDeviceScheduleBlockingSync);
CUDA_LOG_ERROR();
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
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);
x11_luffaCubehash512_cpu_init(thr_id, throughput);
x11_shavite512_cpu_init(thr_id, throughput);
if (x11_simd512_cpu_init(thr_id, throughput) != 0) {
return 0;
}
x11_echo512_cpu_init(thr_id, throughput);
x13_hamsi512_cpu_init(thr_id, throughput);
x13_fugue512_cpu_init(thr_id, throughput);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_hash[thr_id], 16 * sizeof(uint32_t) * throughput), 0);
cuda_check_cpu_init(thr_id, throughput);
init[thr_id] = true;
}
uint32_t endiandata[20];
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
quark_blake512_cpu_setBlock_80(thr_id, endiandata);
cuda_check_cpu_setTarget(ptarget);
do {
uint32_t foundNonce;
int order = 0;
quark_blake512_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id]); order++;
quark_bmw512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_groestl512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_skein512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_jh512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
quark_keccak512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
x11_luffaCubehash512_cpu_hash_64(thr_id, throughput, d_hash[thr_id], order++);
x11_shavite512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
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++);
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;
CUDA_LOG_ERROR();
foundNonce = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]);
if (foundNonce != UINT32_MAX)
{
uint32_t vhash[8];
be32enc(&endiandata[19], foundNonce);
x13hash(vhash, endiandata);
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);
work_set_target_ratio(work, vhash);
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) {
work_set_target_ratio(work, vhash);
xchg(pdata[19], pdata[21]);
}
res++;
}
return res;
} else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", 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;
CUDA_LOG_ERROR();
return 0;
}
// cleanup
extern "C" void free_x13(int thr_id)
{
if (!init[thr_id])
return;
cudaThreadSynchronize();
cudaFree(d_hash[thr_id]);
quark_blake512_cpu_free(thr_id);
quark_groestl512_cpu_free(thr_id);
x11_simd512_cpu_free(thr_id);
x13_fugue512_cpu_free(thr_id);
cuda_check_cpu_free(thr_id);
CUDA_LOG_ERROR();
cudaDeviceSynchronize();
init[thr_id] = false;
}