GOSTSec/ccminer
1
0
Fork 0
mirror of https://github.com/GOSTSec/ccminer synced 2025-01-24 21:45:04 +00:00
Code Issues Releases Wiki Activity

x12: apply echo512 optimised kernel on recent cards

Browse Source
This commit is contained in:
Tanguy Pruvot 2018-04-09 17:45:20 +02:00
parent 5d0c0a665d
commit 16ebe53b72
3 changed files with 18 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
x12/x12.cu
View File

@ -22,6 +22,8 @@ extern "C" {
static uint32_t *d_hash[MAX_GPUS]; static uint32_t *d_hash[MAX_GPUS];
extern void x16_echo512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t *d_hash);
extern void x13_hamsi512_cpu_init(int thr_id, uint32_t threads); 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_hamsi512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_hash, int order);
@ -96,13 +98,15 @@ extern "C" void x12hash(void *output, const void *input)
} }
static bool init[MAX_GPUS] = { 0 }; static bool init[MAX_GPUS] = { 0 };
static bool use_compat_kernels[MAX_GPUS] = { 0 };
extern "C" int scanhash_x12(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done) extern "C" int scanhash_x12(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done)
{ {
uint32_t *pdata = work->data; uint32_t *pdata = work->data;
uint32_t *ptarget = work->target; uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19]; const uint32_t first_nonce = pdata[19];
int intensity = (device_sm[device_map[thr_id]] > 500 && !is_windows()) ? 20 : 19; const int dev_id = device_map[thr_id];
int intensity = (device_sm[dev_id] > 500 && !is_windows()) ? 20 : 19;
uint32_t throughput = cuda_default_throughput(thr_id, 1 << intensity); // 19=256*256*8; 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 (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
@ -111,7 +115,7 @@ extern "C" int scanhash_x12(int thr_id, struct work* work, uint32_t max_nonce, u
if (!init[thr_id]) if (!init[thr_id])
{ {
cudaSetDevice(device_map[thr_id]); cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) { if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset(); cudaDeviceReset();
// reduce cpu usage // reduce cpu usage
@ -120,13 +124,17 @@ extern "C" int scanhash_x12(int thr_id, struct work* work, uint32_t max_nonce, u
} }
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput); gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
use_compat_kernels[thr_id] = (cuda_arch[dev_id] < 500);
if (use_compat_kernels[thr_id])
x11_echo512_cpu_init(thr_id, throughput);
quark_blake512_cpu_init(thr_id, throughput); quark_blake512_cpu_init(thr_id, throughput);
x11_luffaCubehash512_cpu_init(thr_id, throughput); x11_luffaCubehash512_cpu_init(thr_id, throughput);
x11_shavite512_cpu_init(thr_id, throughput); x11_shavite512_cpu_init(thr_id, throughput);
if (x11_simd512_cpu_init(thr_id, throughput) != 0) { if (x11_simd512_cpu_init(thr_id, throughput) != 0) {
return 0; return 0;
} }
x11_echo512_cpu_init(thr_id, throughput);
quark_groestl512_cpu_init(thr_id, throughput); quark_groestl512_cpu_init(thr_id, throughput);
quark_skein512_cpu_init(thr_id, throughput); quark_skein512_cpu_init(thr_id, throughput);
quark_bmw512_cpu_init(thr_id, throughput); quark_bmw512_cpu_init(thr_id, throughput);
@ -156,7 +164,11 @@ extern "C" int scanhash_x12(int thr_id, struct work* work, uint32_t max_nonce, u
x11_luffaCubehash512_cpu_hash_64(thr_id, throughput, 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_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_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++); if (use_compat_kernels[thr_id])
x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);
else {
x16_echo512_cpu_hash_64(thr_id, throughput, d_hash[thr_id]); order++;
}
quark_groestl512_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_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_jh512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], order++);

2
x16/x16r.cu
View File

@ -250,7 +250,7 @@ extern "C" int scanhash_x16r(int thr_id, struct work* work, uint32_t max_nonce,
if (!init[thr_id]) if (!init[thr_id])
{ {
cudaSetDevice(device_map[thr_id]); cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) { if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset(); cudaDeviceReset();
// reduce cpu usage // reduce cpu usage

2
x16/x16s.cu
View File

@ -248,7 +248,7 @@ extern "C" int scanhash_x16s(int thr_id, struct work* work, uint32_t max_nonce,
if (!init[thr_id]) if (!init[thr_id])
{ {
cudaSetDevice(device_map[thr_id]); cudaSetDevice(dev_id);
if (opt_cudaschedule == -1 && gpu_threads == 1) { if (opt_cudaschedule == -1 && gpu_threads == 1) {
cudaDeviceReset(); cudaDeviceReset();
// reduce cpu usage // reduce cpu usage