Merge branch 'Foudge-master'

This commit is contained in:
XMRig 2018-02-08 17:34:33 +07:00
commit c50ccd4e06
6 changed files with 58 additions and 50 deletions

View file

@ -100,7 +100,13 @@ void Cpu::initCommon()
m_l2_cache = data.l2_cache * (m_totalCores / 2) * m_sockets;
m_l2_exclusive = true;
}
else {
// Workaround for Intel Core Solo, Core Duo, Core 2 Duo, Core 2 Quad and their Xeon homologue
// These processors have L2 cache shared by 2 cores.
else if (data.vendor == VENDOR_INTEL && data.family == 0x06 && (data.model == 0x0E || data.model == 0x0F || data.model == 0x07)) {
int l2_count_per_socket = m_totalCores > 1 ? m_totalCores / 2 : 1;
m_l2_cache = data.l2_cache > 0 ? data.l2_cache * l2_count_per_socket * m_sockets : 0;
}
else{
m_l2_cache = data.l2_cache > 0 ? data.l2_cache * m_totalCores * m_sockets : 0;
}

View file

@ -46,7 +46,7 @@ NetworkState::NetworkState() :
int NetworkState::connectionTime() const
{
return m_active ? ((uv_now(uv_default_loop()) - m_connectionTime) / 1000) : 0;
return m_active ? (int)((uv_now(uv_default_loop()) - m_connectionTime) / 1000) : 0;
}
@ -56,7 +56,7 @@ uint32_t NetworkState::avgTime() const
return 0;
}
return (uint32_t) connectionTime() / m_latency.size();
return connectionTime() / (uint32_t)m_latency.size();
}

View file

@ -194,14 +194,14 @@ template<bool SOFT_AES>
static inline void aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m128i* x4, __m128i* x5, __m128i* x6, __m128i* x7)
{
if (SOFT_AES) {
*x0 = soft_aesenc(*x0, key);
*x1 = soft_aesenc(*x1, key);
*x2 = soft_aesenc(*x2, key);
*x3 = soft_aesenc(*x3, key);
*x4 = soft_aesenc(*x4, key);
*x5 = soft_aesenc(*x5, key);
*x6 = soft_aesenc(*x6, key);
*x7 = soft_aesenc(*x7, key);
*x0 = soft_aesenc((uint32_t*)x0, key);
*x1 = soft_aesenc((uint32_t*)x1, key);
*x2 = soft_aesenc((uint32_t*)x2, key);
*x3 = soft_aesenc((uint32_t*)x3, key);
*x4 = soft_aesenc((uint32_t*)x4, key);
*x5 = soft_aesenc((uint32_t*)x5, key);
*x6 = soft_aesenc((uint32_t*)x6, key);
*x7 = soft_aesenc((uint32_t*)x7, key);
}
# ifndef XMRIG_ARMv7
else {
@ -361,12 +361,13 @@ inline void cryptonight_hash(const void *__restrict__ input, size_t size, void *
uint64_t idx0 = h0[0] ^ h0[4];
for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
__m128i cx;
if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah0, al0));
cx = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
}
else {
cx = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
# ifndef XMRIG_ARMv7
cx = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0);
# endif
@ -425,14 +426,15 @@ inline void cryptonight_double_hash(const void *__restrict__ input, size_t size,
uint64_t idx1 = h1[0] ^ h1[4];
for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
__m128i cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
__m128i cx0, cx1;
if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1));
cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
}
else {
cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
# ifndef XMRIG_ARMv7
cx0 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx0, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0);
cx1 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx1, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah1, al1);

View file

@ -193,14 +193,14 @@ template<bool SOFT_AES>
static inline void aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m128i* x4, __m128i* x5, __m128i* x6, __m128i* x7)
{
if (SOFT_AES) {
*x0 = soft_aesenc(*x0, key);
*x1 = soft_aesenc(*x1, key);
*x2 = soft_aesenc(*x2, key);
*x3 = soft_aesenc(*x3, key);
*x4 = soft_aesenc(*x4, key);
*x5 = soft_aesenc(*x5, key);
*x6 = soft_aesenc(*x6, key);
*x7 = soft_aesenc(*x7, key);
*x0 = soft_aesenc((uint32_t*)x0, key);
*x1 = soft_aesenc((uint32_t*)x1, key);
*x2 = soft_aesenc((uint32_t*)x2, key);
*x3 = soft_aesenc((uint32_t*)x3, key);
*x4 = soft_aesenc((uint32_t*)x4, key);
*x5 = soft_aesenc((uint32_t*)x5, key);
*x6 = soft_aesenc((uint32_t*)x6, key);
*x7 = soft_aesenc((uint32_t*)x7, key);
}
else {
*x0 = _mm_aesenc_si128(*x0, key);
@ -325,15 +325,14 @@ inline void cryptonight_hash(const void *__restrict__ input, size_t size, void *
for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx;
cx = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah0, al0));
cx = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
}
else {
else {
cx = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
cx = _mm_aesenc_si128(cx, _mm_set_epi64x(ah0, al0));
}
_mm_store_si128((__m128i *) &l0[idx0 & MASK], _mm_xor_si128(bx0, cx));
idx0 = EXTRACT64(cx);
bx0 = cx;
@ -387,14 +386,15 @@ inline void cryptonight_double_hash(const void *__restrict__ input, size_t size,
uint64_t idx1 = h1[0] ^ h1[4];
for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
__m128i cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
__m128i cx0, cx1;
if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1));
cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
}
else {
cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0));
cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1));
}

View file

@ -89,34 +89,34 @@
alignas(16) const uint32_t saes_table[4][256] = { saes_data(saes_u0), saes_data(saes_u1), saes_data(saes_u2), saes_data(saes_u3) };
alignas(16) const uint8_t saes_sbox[256] = saes_data(saes_h0);
static inline __m128i soft_aesenc(__m128i in, __m128i key)
static inline __m128i soft_aesenc(const uint32_t* in, __m128i key)
{
const uint32_t x0 = _mm_cvtsi128_si32(in);
const uint32_t x1 = _mm_cvtsi128_si32(_mm_shuffle_epi32(in, 0x55));
const uint32_t x2 = _mm_cvtsi128_si32(_mm_shuffle_epi32(in, 0xAA));
const uint32_t x3 = _mm_cvtsi128_si32(_mm_shuffle_epi32(in, 0xFF));
const uint32_t x0 = in[0];
const uint32_t x1 = in[1];
const uint32_t x2 = in[2];
const uint32_t x3 = in[3];
__m128i out = _mm_set_epi32(
(saes_table[0][x3 & 0xff] ^ saes_table[1][(x0 >> 8) & 0xff] ^ saes_table[2][(x1 >> 16) & 0xff] ^ saes_table[3][x2 >> 24]),
(saes_table[0][x2 & 0xff] ^ saes_table[1][(x3 >> 8) & 0xff] ^ saes_table[2][(x0 >> 16) & 0xff] ^ saes_table[3][x1 >> 24]),
(saes_table[0][x1 & 0xff] ^ saes_table[1][(x2 >> 8) & 0xff] ^ saes_table[2][(x3 >> 16) & 0xff] ^ saes_table[3][x0 >> 24]),
(saes_table[0][x0 & 0xff] ^ saes_table[1][(x1 >> 8) & 0xff] ^ saes_table[2][(x2 >> 16) & 0xff] ^ saes_table[3][x3 >> 24]));
__m128i out = _mm_set_epi32(
(saes_table[0][x3 & 0xff] ^ saes_table[1][(x0 >> 8) & 0xff] ^ saes_table[2][(x1 >> 16) & 0xff] ^ saes_table[3][x2 >> 24]),
(saes_table[0][x2 & 0xff] ^ saes_table[1][(x3 >> 8) & 0xff] ^ saes_table[2][(x0 >> 16) & 0xff] ^ saes_table[3][x1 >> 24]),
(saes_table[0][x1 & 0xff] ^ saes_table[1][(x2 >> 8) & 0xff] ^ saes_table[2][(x3 >> 16) & 0xff] ^ saes_table[3][x0 >> 24]),
(saes_table[0][x0 & 0xff] ^ saes_table[1][(x1 >> 8) & 0xff] ^ saes_table[2][(x2 >> 16) & 0xff] ^ saes_table[3][x3 >> 24]));
return _mm_xor_si128(out, key);
return _mm_xor_si128(out, key);
}
static inline uint32_t sub_word(uint32_t key)
{
return (saes_sbox[key >> 24 ] << 24) |
(saes_sbox[(key >> 16) & 0xff] << 16 ) |
(saes_sbox[(key >> 8) & 0xff] << 8 ) |
saes_sbox[key & 0xff];
return (saes_sbox[key >> 24 ] << 24) |
(saes_sbox[(key >> 16) & 0xff] << 16 ) |
(saes_sbox[(key >> 8) & 0xff] << 8 ) |
saes_sbox[key & 0xff];
}
#if defined(__clang__) || defined(XMRIG_ARM)
static inline uint32_t _rotr(uint32_t value, uint32_t amount)
{
return (value >> amount) | (value << ((32 - amount) & 31));
return (value >> amount) | (value << ((32 - amount) & 31));
}
#endif

View file

@ -532,7 +532,7 @@ void Client::onAllocBuffer(uv_handle_t *handle, size_t suggested_size, uv_buf_t
auto client = getClient(handle->data);
buf->base = &client->m_recvBuf.base[client->m_recvBufPos];
buf->len = client->m_recvBuf.len - client->m_recvBufPos;
buf->len = client->m_recvBuf.len - (unsigned long)client->m_recvBufPos;
}