1
0
mirror of https://github.com/GOSTSec/ccminer synced 2025-01-09 22:38:05 +00:00
ccminer/myriadgroestl.cpp
Tanguy Pruvot 61ff92b5b4 never interrupt global benchmark with found nonces
fix some algo weird hashrates (like blake)
and reset device between algos, for better accuracy

but this reset doesnt seems enough to bench all algos correctly...

to test on linux, could be a driver issue...

heavy: fix first alloc and indent with tabs...
2015-11-01 21:12:50 +01:00

112 lines
2.7 KiB
C++

#include <string.h>
#include <stdint.h>
#include <cuda_runtime.h>
#include <openssl/sha.h>
#include "sph/sph_groestl.h"
#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);
void myriadhash(void *state, const void *input)
{
uint32_t _ALIGN(64) hash[16];
sph_groestl512_context ctx_groestl;
SHA256_CTX sha256;
sph_groestl512_init(&ctx_groestl);
sph_groestl512(&ctx_groestl, input, 80);
sph_groestl512_close(&ctx_groestl, hash);
SHA256_Init(&sha256);
SHA256_Update(&sha256,(unsigned char *)hash, 64);
SHA256_Final((unsigned char *)hash, &sha256);
memcpy(state, hash, 32);
}
static bool init[MAX_GPUS] = { 0 };
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 = cuda_default_throughput(thr_id, 1U << 17);
if (init[thr_id]) throughput = min(throughput, max_nonce - start_nonce);
uint32_t *outputHash = (uint32_t*)malloc(throughput * 64);
if (opt_benchmark)
ptarget[7] = 0x0000ff;
// init
if(!init[thr_id])
{
cudaSetDevice(device_map[thr_id]);
myriadgroestl_cpu_init(thr_id, throughput);
init[thr_id] = true;
}
for (int k=0; k < 20; k++)
be32enc(&endiandata[k], pdata[k]);
// Context mit dem Endian gedrehten Blockheader vorbereiten (Nonce wird später ersetzt)
myriadgroestl_cpu_setBlock(thr_id, endiandata, (void*)ptarget);
do {
// GPU
uint32_t foundNounce = UINT32_MAX;
myriadgroestl_cpu_hash(thr_id, throughput, pdata[19], outputHash, &foundNounce);
*hashes_done = pdata[19] - start_nonce + throughput;
if (foundNounce < UINT32_MAX && bench_algo < 0)
{
uint32_t _ALIGN(64) vhash[8];
endiandata[19] = swab32(foundNounce);
myriadhash(vhash, endiandata);
if (vhash[7] <= ptarget[7] && fulltest(vhash, ptarget)) {
work_set_target_ratio(work, vhash);
pdata[19] = foundNounce;
free(outputHash);
return 1;
} else {
gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", foundNounce);
}
}
if ((uint64_t) throughput + pdata[19] >= max_nonce) {
pdata[19] = max_nonce;
break;
}
pdata[19] += throughput;
} while (!work_restart[thr_id].restart);
*hashes_done = max_nonce - start_nonce;
free(outputHash);
return 0;
}
// cleanup
void free_myriad(int thr_id)
{
if (!init[thr_id])
return;
cudaThreadSynchronize();
myriadgroestl_cpu_free(thr_id);
init[thr_id] = false;
cudaDeviceSynchronize();
}