+++ /dev/null
-// Copyright (c) 2015 The Novacoin developers
-// Distributed under the MIT/X11 software license, see the accompanying
-// file COPYING or http://www.opensource.org/licenses/mit-license.php.
-
-#include <stdint.h>
-
-void copy_swap_hashes(uint32_t *blocks, uint32_t *state)
-{
- blocks[0] = __builtin_bswap32(state[0]);
- blocks[1] = __builtin_bswap32(state[1]);
- blocks[2] = __builtin_bswap32(state[2]);
- blocks[3] = __builtin_bswap32(state[3]);
- blocks[4] = __builtin_bswap32(state[4]);
- blocks[5] = __builtin_bswap32(state[5]);
- blocks[6] = __builtin_bswap32(state[6]);
- blocks[7] = __builtin_bswap32(state[7]);
- blocks[8] = __builtin_bswap32(state[8]);
- blocks[9] = __builtin_bswap32(state[9]);
- blocks[10] = __builtin_bswap32(state[10]);
- blocks[11] = __builtin_bswap32(state[11]);
- blocks[12] = __builtin_bswap32(state[12]);
- blocks[13] = __builtin_bswap32(state[13]);
- blocks[14] = __builtin_bswap32(state[14]);
- blocks[15] = __builtin_bswap32(state[15]);
- blocks[16] = __builtin_bswap32(state[16]);
- blocks[17] = __builtin_bswap32(state[17]);
- blocks[18] = __builtin_bswap32(state[18]);
- blocks[19] = __builtin_bswap32(state[19]);
- blocks[20] = __builtin_bswap32(state[20]);
- blocks[21] = __builtin_bswap32(state[21]);
- blocks[22] = __builtin_bswap32(state[22]);
- blocks[23] = __builtin_bswap32(state[23]);
- blocks[24] = __builtin_bswap32(state[24]);
- blocks[25] = __builtin_bswap32(state[25]);
- blocks[26] = __builtin_bswap32(state[26]);
- blocks[27] = __builtin_bswap32(state[27]);
- blocks[28] = __builtin_bswap32(state[28]);
- blocks[29] = __builtin_bswap32(state[29]);
- blocks[30] = __builtin_bswap32(state[30]);
- blocks[31] = __builtin_bswap32(state[31]);
-}
-
-#ifdef USE_SSSE3
-#include <immintrin.h>
-
-void copy_swap_hashes_ssse3(uint32_t *blocks, uint32_t *state)
-{
- __m128i mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3);
- _mm_storeu_si128((__m128i *)&blocks[0], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[0]), mask));
- _mm_storeu_si128((__m128i *)&blocks[4], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[4]), mask));
- _mm_storeu_si128((__m128i *)&blocks[8], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[8]), mask));
- _mm_storeu_si128((__m128i *)&blocks[12], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[12]), mask));
- _mm_storeu_si128((__m128i *)&blocks[16], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[16]), mask));
- _mm_storeu_si128((__m128i *)&blocks[20], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[20]), mask));
- _mm_storeu_si128((__m128i *)&blocks[24], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[24]), mask));
- _mm_storeu_si128((__m128i *)&blocks[28], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&state[28]), mask));
-}
-#endif
#ifdef USE_ASM
// kernel padding
-static const uint32_t block1_suffix[9] = { 0x80000000, 0, 0, 0, 0, 0, 0, 0, 0xe0000000 };
+static const uint32_t block1_suffix[9] = { 0x80000000, 0, 0, 0, 0, 0, 0, 0, 0x000000e0 };
static const uint32_t block1_suffix_4way[4 * 9] = {
- 0x00000080, 0x00000080, 0x00000080, 0x00000080,
+ 0x80000000, 0x80000000, 0x80000000, 0x80000000,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
- 0xe0000000, 0xe0000000, 0xe0000000, 0xe0000000
+ 0x000000e0, 0x000000e0, 0x000000e0, 0x000000e0
};
// hash padding
-static const uint32_t block2_suffix[8] = { 0x80000000, 0, 0, 0, 0, 0, 0, 0x00010000 };
+static const uint32_t block2_suffix[8] = { 0x80000000, 0, 0, 0, 0, 0, 0, 0x00000100 };
static const uint32_t block2_suffix_4way[4 * 8] = {
- 0x00000080, 0x00000080, 0x00000080, 0x00000080,
+ 0x80000000, 0x80000000, 0x80000000, 0x80000000,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
0, 0, 0, 0,
- 0x00010000, 0x00010000, 0x00010000, 0x00010000
+ 0x00000100, 0x00000100, 0x00000100, 0x00000100
};
extern "C" int sha256_use_4way();
extern "C" void sha256_transform(uint32_t *state, const uint32_t *block, int swap);
extern "C" void sha256_init_4way(uint32_t *state);
extern "C" void sha256_transform_4way(uint32_t *state, const uint32_t *block, int swap);
-extern "C" void copy_swap_hashes(uint32_t *blocks, uint32_t *state); // Generic block copy function
#ifdef USE_SSSE3
+#include <immintrin.h>
+
extern "C" int sha256_use_ssse3();
-extern "C" void copy_swap_hashes_ssse3(uint32_t *blocks, uint32_t *state); // SSSE3 optimized block copy function
+bool fUseSSSE3 = sha256_use_ssse3() != 0;
-void (*copy_swap)(uint32_t *, uint32_t *) = (sha256_use_ssse3() != 0) ? ©_swap_hashes_ssse3 : copy_swap_hashes;
+inline void copyrow_swap32(uint32_t *to, uint32_t *from)
+{
+ if (!fUseSSSE3)
+ {
+ for (int i = 0; i < 4; i++)
+ to[i] = __builtin_bswap32(from[i]);
+ }
+ else
+ {
+ __m128i mask = _mm_set_epi8(12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3);
+ _mm_storeu_si128((__m128i *)&to[0], _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)&from[0]), mask));
+ }
+}
#else
-void (*copy_swap)(uint32_t *, uint32_t *) = ©_swap_hashes;
+inline void copyrow_swap32(uint32_t *to, uint32_t *from)
+{
+ for (int i = 0; i < 4; i++)
+ to[i] = __builtin_bswap32(from[i]);
+}
#endif
bool fUse4Way = sha256_use_4way() != 0;
for(int i = 0; i < 7; i++)
{
- uint32_t nVal = pnKernel[i];
- fill(vRow.begin(), vRow.end(), nVal);
-
- for (int j = 0; j < 4; j++)
- {
- memcpy(&blocks1[i*4], &vRow[0], 16);
- }
+ fill(vRow.begin(), vRow.end(), pnKernel[i]);
+ copyrow_swap32(&blocks1[i*4], &vRow[0]);
}
memcpy(&blocks1[28], &block1_suffix_4way[0], 36*4); // sha256 padding
memcpy(&blocks2[32], &block2_suffix_4way[0], 32*4);
+ uint32_t nTimeStamps[4] = {0, 0, 0, 0};
+ uint32_t nHashes[4] = {0, 0, 0, 0};
+
// Search forward in time from the given timestamp
// Stopping search in case of shutting down
for (uint32_t nTimeTx=nIntervalBegin, nMaxTarget32 = nMaxTarget.Get32(7); nTimeTx<nIntervalEnd && !fShutdown; nTimeTx +=4)
sha256_init_4way(state1);
sha256_init_4way(state2);
- blocks1[24] = nTimeTx;
- blocks1[25] = nTimeTx+1;
- blocks1[26] = nTimeTx+2;
- blocks1[27] = nTimeTx+3;
+ nTimeStamps[0] = nTimeTx;
+ nTimeStamps[1] = nTimeTx+1;
+ nTimeStamps[2] = nTimeTx+2;
+ nTimeStamps[3] = nTimeTx+3;
- sha256_transform_4way(&state1[0], &blocks1[0], 1); // first hashing
- copy_swap(&blocks2[0], &state1[0]);
- sha256_transform_4way(&state2[0], &blocks2[0], 1); // second hashing
+ copyrow_swap32(&blocks1[24], &nTimeStamps[0]); // Kernel timestamps
+ sha256_transform_4way(&state1[0], &blocks1[0], 0); // first hashing
+ memcpy(&blocks2[0], &state1[0], 128);
+ sha256_transform_4way(&state2[0], &blocks2[0], 0); // second hashing
+ copyrow_swap32(&nHashes[0], &state2[28]);
for(int nResult = 0; nResult < 4; nResult++)
{
- uint32_t nHash = __builtin_bswap32(state2[28+nResult]);
-
- if (nHash <= nMaxTarget32) // Possible hit
+ if (nHashes[nResult] <= nMaxTarget32) // Possible hit
{
- uint32_t nTime = blocks1[24+nResult];
uint256 nHashProofOfStake = 0;
uint32_t *pnHashProofOfStake = (uint32_t *) &nHashProofOfStake;
- pnHashProofOfStake[7] = nHash;
for (int i = 0; i < 7; i++)
pnHashProofOfStake[i] = __builtin_bswap32(state2[(i*4) + nResult]);
+ pnHashProofOfStake[7] = nHashes[nResult];
- CBigNum bnCoinDayWeight = bnValueIn * GetWeight((int64_t)nInputTxTime, (int64_t)nTimeTx) / COIN / nOneDay;
+ CBigNum bnCoinDayWeight = bnValueIn * GetWeight((int64_t)nInputTxTime, (int64_t)nTimeStamps[nResult]) / COIN / nOneDay;
CBigNum bnTargetProofOfStake = bnCoinDayWeight * bnTargetPerCoinDay;
if (bnTargetProofOfStake >= CBigNum(nHashProofOfStake))
- solutions.push_back(std::pair<uint256,uint32_t>(nHashProofOfStake, nTime));
+ solutions.push_back(std::pair<uint256,uint32_t>(nHashProofOfStake, nTimeStamps[nResult]));
}
}
}