#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
};
+// Sha256 initial state
+static const uint32_t sha256_initial[8] = { 0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a, 0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19 };
+
extern "C" int sha256_use_4way();
-extern "C" void sha256_init(uint32_t *state);
+#ifndef __i386__
extern "C" void sha256_transform(uint32_t *state, const uint32_t *block, int swap);
+#endif
+
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;
bnTargetPerCoinDay.SetCompact(nBits);
uint256 nMaxTarget = (bnTargetPerCoinDay * bnValueIn * nStakeMaxAge / COIN / nOneDay).getuint256();
- uint32_t state1[4 * 8] __attribute__((aligned(16)));
- uint32_t state2[4 * 8] __attribute__((aligned(16)));
uint32_t blocks1[4 * 16] __attribute__((aligned(16)));
uint32_t blocks2[4 * 16] __attribute__((aligned(16)));
+ uint32_t candidates[4 * 8] __attribute__((aligned(16)));
vector<uint32_t> vRow = vector<uint32_t>(4);
uint32_t *pnKernel = (uint32_t *) kernel;
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);
+ sha256_init_4way(blocks2);
+ sha256_init_4way(candidates);
- 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(&blocks2[0], &blocks1[0], 0); // first hashing
+ sha256_transform_4way(&candidates[0], &blocks2[0], 0); // second hashing
+ copyrow_swap32(&nHashes[0], &candidates[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[i] = __builtin_bswap32(candidates[(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]));
}
}
}
}
- void Do_generic()
+ void Do_oneway()
{
SetThreadPriority(THREAD_PRIORITY_LOWEST);
+ // Compute maximum possible target to filter out majority of obviously insufficient hashes
+ CBigNum bnTargetPerCoinDay;
+ bnTargetPerCoinDay.SetCompact(nBits);
+ uint256 nMaxTarget = (bnTargetPerCoinDay * bnValueIn * nStakeMaxAge / COIN / nOneDay).getuint256();
+
+#ifdef __i386__
+ SHA256_CTX ctx, workerCtx;
// Init new sha256 context and update it
// with first 24 bytes of kernel
- SHA256_CTX workerCtx;
- SHA256_Init(&workerCtx);
- SHA256_Update(&workerCtx, kernel, 8 + 16);
- SHA256_CTX ctx = workerCtx;
+ SHA256_Init(&ctx);
+ SHA256_Update(&ctx, kernel, 8 + 16);
+ workerCtx = ctx; // save context
// Sha256 result buffer
uint32_t hashProofOfStake[8];
-
- // Compute maximum possible target to filter out majority of obviously insufficient hashes
- CBigNum bnTargetPerCoinDay;
- bnTargetPerCoinDay.SetCompact(nBits);
-
- uint256 nMaxTarget = (bnTargetPerCoinDay * bnValueIn * nStakeMaxAge / COIN / nOneDay).getuint256(),
- *pnHashProofOfStake = (uint256 *)&hashProofOfStake;
+ uint256 *pnHashProofOfStake = (uint256 *)&hashProofOfStake;
// Search forward in time from the given timestamp
// Stopping search in case of shutting down
if (bnTargetProofOfStake >= CBigNum(*pnHashProofOfStake))
solutions.push_back(std::pair<uint256,uint32_t>(*pnHashProofOfStake, nTimeTx));
}
+#else
+ uint32_t block1[16] __attribute__((aligned(16)));
+ uint32_t block2[16] __attribute__((aligned(16)));
+ uint32_t candidate[8] __attribute__((aligned(16)));
+
+ memcpy(&block1[7], &block1_suffix[0], 36); // sha256 padding
+ memcpy(&block2[8], &block2_suffix[0], 32);
+
+ uint32_t *pnKernel = (uint32_t *) kernel;
+ copyrow_swap32(&block1[0], pnKernel);
+ block1[4] = __builtin_bswap32(pnKernel[4]);
+ block1[5] = __builtin_bswap32(pnKernel[5]);
+
+ // 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++)
+ {
+ memcpy(&block2[0], &sha256_initial[0], 32);
+ memcpy(&candidate[0], &sha256_initial[0], 32);
+
+ block1[6] = __builtin_bswap32(nTimeTx);
+
+ sha256_transform(&block2[0], &block1[0], 0); // first hashing
+ sha256_transform(&candidate[0], &block2[0], 0); // second hashing
+
+ uint32_t nHash7 = __builtin_bswap32(candidate[7]);
+
+ // Skip if hash doesn't satisfy the maximum target
+ if (nHash7 > nMaxTarget32)
+ continue;
+
+ uint256 nHashProofOfStake;
+ uint32_t *pnHashProofOfStake = (uint32_t *) &nHashProofOfStake;
+
+ for (int i = 0; i < 7; i++)
+ pnHashProofOfStake[i] = __builtin_bswap32(candidate[i]);
+ pnHashProofOfStake[7] = nHash7;
+
+ CBigNum bnCoinDayWeight = bnValueIn * GetWeight((int64_t)nInputTxTime, (int64_t)nTimeTx) / COIN / nOneDay;
+ CBigNum bnTargetProofOfStake = bnCoinDayWeight * bnTargetPerCoinDay;
+
+ if (bnTargetProofOfStake >= CBigNum(nHashProofOfStake))
+ solutions.push_back(std::pair<uint256,uint32_t>(nHashProofOfStake, nTimeTx));
+ }
+#endif
}
void Do()
if (fUse4Way)
Do_4way();
else
- Do_generic();
+ Do_oneway();
}
vector<std::pair<uint256,uint32_t> >& GetSolutions()