#include "cpuminer-config.h" #include "miner.h" #include #include #include "sha3/sph_blake.h" /* Move init out of loop, so init once externally, and then use one single memcpy with that bigger memory block */ typedef struct { sph_blake256_context blake1; } blakehash_context_holder; static blakehash_context_holder base_contexts; void init_blakehash_contexts() { sph_blake256_init(&base_contexts.blake1); } static void blakehash(void *state, const void *input) { blakehash_context_holder ctx; //an array of uint32 uint32_t hashA[8]; //do one memcopy to get fresh contexts, its faster even with a larger block then issuing 9 memcopies memcpy(&ctx, &base_contexts, sizeof(base_contexts)); sph_blake256 (&ctx.blake1, input, 80); sph_blake256_close (&ctx.blake1, hashA); //0 memcpy(state, hashA, 32); } int scanhash_blake(int thr_id, uint32_t *pdata, const uint32_t *ptarget, uint32_t max_nonce, uint64_t *hashes_done) { uint32_t n = pdata[19] - 1; const uint32_t first_nonce = pdata[19]; const uint32_t Htarg = ptarget[7]; uint32_t hash64[8] __attribute__((aligned(32))); uint32_t endiandata[32]; //char testdata[] = {"\x70\x00\x00\x00\x5d\x38\x5b\xa1\x14\xd0\x79\x97\x0b\x29\xa9\x41\x8f\xd0\x54\x9e\x7d\x68\xa9\x5c\x7f\x16\x86\x21\xa3\x14\x20\x10\x00\x00\x00\x00\x57\x85\x86\xd1\x49\xfd\x07\xb2\x2f\x3a\x8a\x34\x7c\x51\x6d\xe7\x05\x2f\x03\x4d\x2b\x76\xff\x68\xe0\xd6\xec\xff\x9b\x77\xa4\x54\x89\xe3\xfd\x51\x17\x32\x01\x1d\xf0\x73\x10\x00"}; //we need bigendian data... //lessons learned: do NOT endianchange directly in pdata, this will all proof-of-works be considered as stale from minerd.... int kk=0; for (; kk < 32; kk++) { be32enc(&endiandata[kk], ((uint32_t*)pdata)[kk]); }; // if (opt_debug) // { // applog(LOG_DEBUG, "Thr: %02d, firstN: %08x, maxN: %08x, ToDo: %d", thr_id, first_nonce, max_nonce, max_nonce-first_nonce); // } /* I'm to lazy to put the loop in an inline function... so dirty copy'n'paste.... */ /* i know that i could set a variable, but i don't know how the compiler will optimize it, not that then the cpu needs to load the value *everytime* in a register */ if (ptarget[7]==0) { do { pdata[19] = ++n; be32enc(&endiandata[19], n); blakehash(hash64, &endiandata); if (((hash64[7]&0xFFFFFFFF)==0) && fulltest(hash64, ptarget)) { *hashes_done = n - first_nonce + 1; return true; } } while (n < max_nonce && !work_restart[thr_id].restart); } else if (ptarget[7]<=0xF) { do { pdata[19] = ++n; be32enc(&endiandata[19], n); blakehash(hash64, &endiandata); if (((hash64[7]&0xFFFFFFF0)==0) && fulltest(hash64, ptarget)) { *hashes_done = n - first_nonce + 1; return true; } } while (n < max_nonce && !work_restart[thr_id].restart); } else if (ptarget[7]<=0xFF) { do { pdata[19] = ++n; be32enc(&endiandata[19], n); blakehash(hash64, &endiandata); if (((hash64[7]&0xFFFFFF00)==0) && fulltest(hash64, ptarget)) { *hashes_done = n - first_nonce + 1; return true; } } while (n < max_nonce && !work_restart[thr_id].restart); } else if (ptarget[7]<=0xFFF) { do { pdata[19] = ++n; be32enc(&endiandata[19], n); blakehash(hash64, &endiandata); if (((hash64[7]&0xFFFFF000)==0) && fulltest(hash64, ptarget)) { *hashes_done = n - first_nonce + 1; return true; } } while (n < max_nonce && !work_restart[thr_id].restart); } else if (ptarget[7]<=0xFFFF) { do { pdata[19] = ++n; be32enc(&endiandata[19], n); blakehash(hash64, &endiandata); if (((hash64[7]&0xFFFF0000)==0) && fulltest(hash64, ptarget)) { *hashes_done = n - first_nonce + 1; return true; } } while (n < max_nonce && !work_restart[thr_id].restart); } else { do { pdata[19] = ++n; be32enc(&endiandata[19], n); blakehash(hash64, &endiandata); if (fulltest(hash64, ptarget)) { *hashes_done = n - first_nonce + 1; return true; } } while (n < max_nonce && !work_restart[thr_id].restart); } *hashes_done = n - first_nonce + 1; pdata[19] = n; return 0; }