diff --git a/Makefile.am b/Makefile.am index 4ed9679..0ff48b3 100644 --- a/Makefile.am +++ b/Makefile.am @@ -64,7 +64,7 @@ ccminer_SOURCES = elist.h miner.h compat.h \ qubit/qubit.cu qubit/qubit_luffa512.cu qubit/deep.cu qubit/luffa.cu \ x11/x11.cu x11/fresh.cu x11/cuda_x11_luffa512.cu x11/cuda_x11_cubehash512.cu \ x11/cuda_x11_shavite512.cu x11/cuda_x11_simd512.cu x11/cuda_x11_echo.cu \ - x11/cuda_x11_luffa512_Cubehash.cu x11/x11evo.cu x11/timetravel.cu \ + x11/cuda_x11_luffa512_Cubehash.cu x11/x11evo.cu x11/timetravel.cu x11/bitcore.cu \ x13/x13.cu x13/cuda_x13_hamsi512.cu x13/cuda_x13_fugue512.cu \ x15/x14.cu x15/x15.cu x15/cuda_x14_shabal512.cu x15/cuda_x15_whirlpool.cu \ x15/whirlpool.cu x15/cuda_x15_whirlpool_sm3.cu \ diff --git a/algos.h b/algos.h index ea27f7c..926f2a4 100644 --- a/algos.h +++ b/algos.h @@ -45,6 +45,7 @@ enum sha_algos { ALGO_SKEIN2, ALGO_S3, ALGO_TIMETRAVEL, + ALGO_BITCORE, ALGO_X11EVO, ALGO_X11, ALGO_X13, @@ -105,6 +106,7 @@ static const char *algo_names[] = { "skein2", "s3", "timetravel", + "bitcore", "x11evo", "x11", "x13", @@ -159,6 +161,8 @@ static inline int algo_to_int(char* arg) i = ALGO_SHA256D; else if (!strcasecmp("thorsriddle", arg)) i = ALGO_VELTOR; + else if (!strcasecmp("timetravel10", arg)) + i = ALGO_BITCORE; else if (!strcasecmp("whirl", arg)) i = ALGO_WHIRLPOOL; else if (!strcasecmp("ziftr", arg)) diff --git a/bench.cpp b/bench.cpp index 5fa2791..12854a6 100644 --- a/bench.cpp +++ b/bench.cpp @@ -94,6 +94,7 @@ void algo_free_all(int thr_id) free_scrypt(thr_id); free_scrypt_jane(thr_id); free_timetravel(thr_id); + free_bitcore(thr_id); } // benchmark all algos (called once per mining thread) diff --git a/ccminer.cpp b/ccminer.cpp index 04b166a..1efdfbc 100644 --- a/ccminer.cpp +++ b/ccminer.cpp @@ -224,6 +224,7 @@ Usage: " PROGRAM_NAME " [OPTIONS]\n\ Options:\n\ -a, --algo=ALGO specify the hash algorithm to use\n\ bastion Hefty bastion\n\ + bitcore Timetravel-10\n\ blake Blake 256 (SFR)\n\ blake2s Blake2-S 256 (NEVA)\n\ blakecoin Fast Blake 256 (8 rounds)\n\ @@ -1625,6 +1626,7 @@ static bool stratum_gen_work(struct stratum_ctx *sctx, struct work *work) case ALGO_LYRA2v2: case ALGO_LYRA2Z: case ALGO_TIMETRAVEL: + case ALGO_BITCORE: work_set_target(work, sctx->job.diff / (256.0 * opt_difficulty)); break; case ALGO_KECCAK: @@ -2129,6 +2131,7 @@ static void *miner_thread(void *userdata) case ALGO_LYRA2v2: case ALGO_S3: case ALGO_TIMETRAVEL: + case ALGO_BITCORE: case ALGO_X11EVO: case ALGO_X11: case ALGO_X13: @@ -2347,6 +2350,9 @@ static void *miner_thread(void *userdata) case ALGO_TIMETRAVEL: rc = scanhash_timetravel(thr_id, &work, max_nonce, &hashes_done); break; + case ALGO_BITCORE: + rc = scanhash_bitcore(thr_id, &work, max_nonce, &hashes_done); + break; case ALGO_X11EVO: rc = scanhash_x11evo(thr_id, &work, max_nonce, &hashes_done); break; diff --git a/ccminer.vcxproj b/ccminer.vcxproj index afddcc2..2b26598 100644 --- a/ccminer.vcxproj +++ b/ccminer.vcxproj @@ -538,6 +538,7 @@ + diff --git a/ccminer.vcxproj.filters b/ccminer.vcxproj.filters index 775da32..97302bb 100644 --- a/ccminer.vcxproj.filters +++ b/ccminer.vcxproj.filters @@ -739,6 +739,9 @@ Source Files\CUDA\x11 + + Source Files\CUDA\x11 + Source Files\CUDA\x11 diff --git a/miner.h b/miner.h index c4c739f..221d6ee 100644 --- a/miner.h +++ b/miner.h @@ -308,6 +308,7 @@ extern int scanhash_skeincoin(int thr_id, struct work* work, uint32_t max_nonce, extern int scanhash_skein2(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_timetravel(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done); +extern int scanhash_bitcore(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done); extern int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done, int8_t blake_rounds); extern int scanhash_veltor(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done); extern int scanhash_whirl(int thr_id, struct work* work, uint32_t max_nonce, unsigned long *hashes_done); @@ -363,6 +364,7 @@ extern void free_skeincoin(int thr_id); extern void free_skein2(int thr_id); extern void free_s3(int thr_id); extern void free_timetravel(int thr_id); +extern void free_bitcore(int thr_id); extern void free_vanilla(int thr_id); extern void free_veltor(int thr_id); extern void free_whirl(int thr_id); @@ -887,6 +889,7 @@ void skeincoinhash(void *output, const void *input); void skein2hash(void *output, const void *input); void s3hash(void *output, const void *input); void timetravel_hash(void *output, const void *input); +void bitcore_hash(void *output, const void *input); void veltorhash(void *output, const void *input); void wcoinhash(void *state, const void *input); void whirlxHash(void *state, const void *input); diff --git a/util.cpp b/util.cpp index a01bf30..1e945c2 100644 --- a/util.cpp +++ b/util.cpp @@ -2252,6 +2252,9 @@ void print_hash_tests(void) timetravel_hash(&hash[0], &buf[0]); printpfx("timetravel", hash); + bitcore_hash(&hash[0], &buf[0]); + printpfx("bitcore", hash); + blake256hash(&hash[0], &buf[0], 8); printpfx("vanilla", hash); diff --git a/x11/bitcore.cu b/x11/bitcore.cu new file mode 100644 index 0000000..7873967 --- /dev/null +++ b/x11/bitcore.cu @@ -0,0 +1,439 @@ +/** + * Timetravel-10 (bitcore) CUDA implementation + * by tpruvot@github - May 2017 + */ + +#include +#include +#include + +#define HASH_FUNC_BASE_TIMESTAMP 1492973331U +#define HASH_FUNC_COUNT 10 +#define HASH_FUNC_COUNT_PERMUTATIONS 40320U + +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" +#if HASH_FUNC_COUNT > 10 +#include "sph/sph_echo.h" +#endif +} + +#include "miner.h" +#include "cuda_helper.h" +#include "cuda_x11.h" + +static uint32_t *d_hash[MAX_GPUS]; + +enum Algo { + BLAKE = 0, + BMW, + GROESTL, + SKEIN, + JH, + KECCAK, + LUFFA, + CUBEHASH, + SHAVITE, + SIMD, +#if HASH_FUNC_COUNT > 10 + ECHO, +#endif + MAX_ALGOS_COUNT +}; + +inline void swap8(uint8_t *a, uint8_t *b) +{ + uint8_t t = *a; + *a = *b; + *b = t; +} + +inline void initPerm(uint8_t n[], int count) +{ + for (int i = 0; i < count; i++) + n[i] = i; +} + +static int nextPerm(uint8_t n[], int count) +{ + int tail, i, j; + + if (count <= 1) + return 0; + + for (i = count - 1; i>0 && n[i - 1] >= n[i]; i--); + tail = i; + + if (tail > 0) { + for (j = count - 1; j>tail && n[j] <= n[tail - 1]; j--); + swap8(&n[tail - 1], &n[j]); + } + + for (i = tail, j = count - 1; i= 10) + sprintf(sptr, "%c", 'A' + (algoList[j] - 10)); + else + sprintf(sptr, "%u", (uint32_t) algoList[j]); + sptr++; + } + *sptr = '\0'; +} + +static __thread uint32_t s_ntime = 0; +static uint32_t s_sequence = UINT32_MAX; +static uint8_t s_firstalgo = 0xFF; +static char hashOrder[HASH_FUNC_COUNT + 1] = { 0 }; + +#define INITIAL_DATE HASH_FUNC_BASE_TIMESTAMP +static inline uint32_t getCurrentAlgoSeq(uint32_t ntime) +{ + // unlike x11evo, the permutation changes often (with ntime) + return (uint32_t) (ntime - INITIAL_DATE) % HASH_FUNC_COUNT_PERMUTATIONS; +} + +// To finish... +static void get_travel_order(uint32_t ntime, char *permstr) +{ + uint32_t seq = getCurrentAlgoSeq(ntime); + if (s_sequence != seq) { + getAlgoString(permstr, seq); + s_sequence = seq; + } +} + +// CPU Hash +extern "C" void bitcore_hash(void *output, const void *input) +{ + uint32_t _ALIGN(64) hash[64/4] = { 0 }; + + sph_blake512_context ctx_blake; + sph_bmw512_context ctx_bmw; + sph_groestl512_context ctx_groestl; + sph_skein512_context ctx_skein; + sph_jh512_context ctx_jh; + sph_keccak512_context ctx_keccak; + sph_luffa512_context ctx_luffa1; + sph_cubehash512_context ctx_cubehash1; + sph_shavite512_context ctx_shavite1; + sph_simd512_context ctx_simd1; +#if HASH_FUNC_COUNT > 10 + sph_echo512_context ctx_echo1; +#endif + + if (s_sequence == UINT32_MAX) { + uint32_t *data = (uint32_t*) input; + const uint32_t ntime = (opt_benchmark || !data[17]) ? (uint32_t) time(NULL) : data[17]; + get_travel_order(ntime, hashOrder); + } + + void *in = (void*) input; + int size = 80; + + const int hashes = (int) strlen(hashOrder); + + for (int i = 0; i < hashes; i++) + { + const char elem = hashOrder[i]; + uint8_t algo = elem >= 'A' ? elem - 'A' + 10 : elem - '0'; + + if (i > 0) { + in = (void*) hash; + size = 64; + } + + switch (algo) { + case BLAKE: + sph_blake512_init(&ctx_blake); + sph_blake512(&ctx_blake, in, size); + sph_blake512_close(&ctx_blake, hash); + break; + case BMW: + sph_bmw512_init(&ctx_bmw); + sph_bmw512(&ctx_bmw, in, size); + sph_bmw512_close(&ctx_bmw, hash); + break; + case GROESTL: + sph_groestl512_init(&ctx_groestl); + sph_groestl512(&ctx_groestl, in, size); + sph_groestl512_close(&ctx_groestl, hash); + break; + case SKEIN: + sph_skein512_init(&ctx_skein); + sph_skein512(&ctx_skein, in, size); + sph_skein512_close(&ctx_skein, hash); + break; + case JH: + sph_jh512_init(&ctx_jh); + sph_jh512(&ctx_jh, in, size); + sph_jh512_close(&ctx_jh, hash); + break; + case KECCAK: + sph_keccak512_init(&ctx_keccak); + sph_keccak512(&ctx_keccak, in, size); + sph_keccak512_close(&ctx_keccak, hash); + break; + case LUFFA: + sph_luffa512_init(&ctx_luffa1); + sph_luffa512(&ctx_luffa1, in, size); + sph_luffa512_close(&ctx_luffa1, hash); + break; + case CUBEHASH: + sph_cubehash512_init(&ctx_cubehash1); + sph_cubehash512(&ctx_cubehash1, in, size); + sph_cubehash512_close(&ctx_cubehash1, hash); + break; + case SHAVITE: + sph_shavite512_init(&ctx_shavite1); + sph_shavite512(&ctx_shavite1, in, size); + sph_shavite512_close(&ctx_shavite1, hash); + break; + case SIMD: + sph_simd512_init(&ctx_simd1); + sph_simd512(&ctx_simd1, in, size); + sph_simd512_close(&ctx_simd1, hash); + break; +#if HASH_FUNC_COUNT > 10 + case ECHO: + sph_echo512_init(&ctx_echo1); + sph_echo512(&ctx_echo1, in, size); + sph_echo512_close(&ctx_echo1, hash); + break; +#endif + } + } + + memcpy(output, hash, 32); +} + +//#define _DEBUG +#define _DEBUG_PREFIX "tt-" +#include "cuda_debug.cuh" + +void quark_blake512_cpu_hash_64(int thr_id, uint32_t threads, uint32_t startNounce, uint32_t *d_nonceVector, uint32_t *d_outputHash, int order); + +static bool init[MAX_GPUS] = { 0 }; + +extern "C" int scanhash_bitcore(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 = cuda_default_throughput(thr_id, 1U << intensity); // 19=256*256*8; + //if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce); + + if (opt_benchmark) pdata[17] = swab32(0x59090909); + + if (opt_debug || s_ntime != pdata[17] || s_sequence == UINT32_MAX) { + uint32_t ntime = swab32(work->data[17]); + get_travel_order(ntime, hashOrder); + s_ntime = pdata[17]; + if (opt_debug && !thr_id) { + applog(LOG_DEBUG, "timetravel10 hash order %s (%08x)", hashOrder, ntime); + } + } + + if (opt_benchmark) + ptarget[7] = 0x5; + + 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_bmw512_cpu_init(thr_id, throughput); + quark_groestl512_cpu_init(thr_id, throughput); + quark_skein512_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); + if (x11_simd512_cpu_init(thr_id, throughput) != 0) { + return 0; + } +#if HASH_FUNC_COUNT > 10 + x11_echo512_cpu_init(thr_id, throughput); +#endif + CUDA_CALL_OR_RET_X(cudaMalloc(&d_hash[thr_id], (size_t) 64 * throughput), -1); + CUDA_CALL_OR_RET_X(cudaMemset(d_hash[thr_id], 0, (size_t) 64 * throughput), -1); + + cuda_check_cpu_init(thr_id, throughput); + + init[thr_id] = true; + } + + uint32_t endiandata[20]; + for (int k=0; k < 19; k++) + be32enc(&endiandata[k], pdata[k]); + + cuda_check_cpu_setTarget(ptarget); + + const int hashes = (int) strlen(hashOrder); + const char first = hashOrder[0]; + const uint8_t algo80 = first >= 'A' ? first - 'A' + 10 : first - '0'; + if (algo80 != s_firstalgo) { + s_firstalgo = algo80; + } + + // first algo seems locked to blake in bitcore, fine! + quark_blake512_cpu_setBlock_80(thr_id, endiandata); + + do { + // Hash with CUDA + + quark_blake512_cpu_hash_80(thr_id, throughput, pdata[19], d_hash[thr_id]); + TRACE("blake80:"); + + for (int i = 1; i < hashes; i++) + { + const char elem = hashOrder[i]; + const uint8_t algo64 = elem >= 'A' ? elem - 'A' + 10 : elem - '0'; + + switch (algo64) { + case BLAKE: + quark_blake512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("blake :"); + break; + case BMW: + quark_bmw512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("bmw :"); + break; + case GROESTL: + quark_groestl512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("groestl:"); + break; + case SKEIN: + quark_skein512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("skein :"); + break; + case JH: + quark_jh512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("jh512 :"); + break; + case KECCAK: + quark_keccak512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("keccak :"); + break; + case LUFFA: + x11_luffa512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("luffa :"); + break; + case CUBEHASH: + x11_cubehash512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("cube :"); + break; + case SHAVITE: + x11_shavite512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("shavite:"); + break; + case SIMD: + x11_simd512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("simd :"); + break; +#if HASH_FUNC_COUNT > 10 + case ECHO: + x11_echo512_cpu_hash_64(thr_id, throughput, pdata[19], NULL, d_hash[thr_id], i); + TRACE("echo :"); + break; +#endif + } + } + + *hashes_done = pdata[19] - first_nonce + throughput; + + work->nonces[0] = cuda_check_hash(thr_id, throughput, pdata[19], d_hash[thr_id]); + if (work->nonces[0] != UINT32_MAX) + { + uint32_t _ALIGN(64) vhash[8]; + const uint32_t Htarg = ptarget[7]; + be32enc(&endiandata[19], work->nonces[0]); + bitcore_hash(vhash, endiandata); + + if (vhash[7] <= Htarg && fulltest(vhash, ptarget)) { + work->valid_nonces = 1; + work_set_target_ratio(work, vhash); + work->nonces[1] = cuda_check_hash_suppl(thr_id, throughput, pdata[19], d_hash[thr_id], 1); + pdata[19] = work->nonces[0]; + if (work->nonces[1] != 0) { + be32enc(&endiandata[19], work->nonces[1]); + bitcore_hash(vhash, endiandata); + if (vhash[7] <= Htarg && fulltest(vhash, ptarget)) { + bn_set_target_ratio(work, vhash, 1); + work->valid_nonces++; + } + pdata[19] = max(pdata[19], work->nonces[1]) + 1; + } + return work->valid_nonces; + } else if (vhash[7] > Htarg) { + gpulog(LOG_WARNING, thr_id, "result for %08x does not validate on CPU!", work->nonces[0]); + pdata[19] = work->nonces[0] + 1; + continue; + } + } + + if ((uint64_t) throughput + pdata[19] >= max_nonce) { + pdata[19] = max_nonce; + break; + } + pdata[19] += throughput; + + } while (!work_restart[thr_id].restart); + + *hashes_done = pdata[19] - first_nonce; + return 0; +} + +// cleanup +extern "C" void free_bitcore(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); + + cuda_check_cpu_free(thr_id); + init[thr_id] = false; + + cudaDeviceSynchronize(); +}