fzorb/REDACTED-rig
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Merge branch 'master'

Browse source 
Conflicts:
	src/net/Client.cpp
This commit is contained in:
BenDroid 2018-03-02 23:22:06 +01:00
commit 563fdf6485
5 changed files with 140 additions and 99 deletions
Download patch file Download diff file Expand all files Collapse all files

4
src/Options.cpp
View file

@ -130,6 +130,9 @@ static char const short_options[] = "a:c:khBp:Px:r:R:s:t:T:o:u:O:v:Vl:S";
static struct option const options[] = { static struct option const options[] = {
{ "algo", 1, nullptr, 'a' }, { "algo", 1, nullptr, 'a' },
{ "api-access-token", 1, nullptr, 4001 },
{ "api-port", 1, nullptr, 4000 },
{ "api-worker-id", 1, nullptr, 4002 },
{ "av", 1, nullptr, 'v' }, { "av", 1, nullptr, 'v' },
{ "aesni", 1, nullptr, 'A' }, { "aesni", 1, nullptr, 'A' },
{ "multihash-factor", 1, nullptr, 'm' }, { "multihash-factor", 1, nullptr, 'm' },
@ -532,7 +535,6 @@ bool Options::parseArg(int key, const char *arg)
case 'S': /* --syslog */ case 'S': /* --syslog */
case 1005: /* --safe */ case 1005: /* --safe */
case 1006: /* --nicehash */ case 1006: /* --nicehash */
return parseBoolean(key, true);
case 1002: /* --no-color */ case 1002: /* --no-color */
case 1009: /* --no-huge-pages */ case 1009: /* --no-huge-pages */

4
src/api/NetworkState.cpp
View file

@ -46,7 +46,7 @@ NetworkState::NetworkState() :
int NetworkState::connectionTime() const 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 0;
} }
return (uint32_t) connectionTime() / m_latency.size(); return connectionTime() / (uint32_t)m_latency.size();
} }

98
src/crypto/CryptoNight_arm.h
View file

@ -208,14 +208,14 @@ aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m12
__m128i* x7) __m128i* x7)
{ {
if (SOFT_AES) { if (SOFT_AES) {
*x0 = soft_aesenc(*x0, key); *x0 = soft_aesenc((uint32_t*)x0, key);
*x1 = soft_aesenc(*x1, key); *x1 = soft_aesenc((uint32_t*)x1, key);
*x2 = soft_aesenc(*x2, key); *x2 = soft_aesenc((uint32_t*)x2, key);
*x3 = soft_aesenc(*x3, key); *x3 = soft_aesenc((uint32_t*)x3, key);
*x4 = soft_aesenc(*x4, key); *x4 = soft_aesenc((uint32_t*)x4, key);
*x5 = soft_aesenc(*x5, key); *x5 = soft_aesenc((uint32_t*)x5, key);
*x6 = soft_aesenc(*x6, key); *x6 = soft_aesenc((uint32_t*)x6, key);
*x7 = soft_aesenc(*x7, key); *x7 = soft_aesenc((uint32_t*)x7, key);
} }
# ifndef XMRIG_ARMv7 # ifndef XMRIG_ARMv7
else { else {
@ -392,11 +392,11 @@ public:
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
for (size_t hashBlock = 0; hashBlock < NUM_HASH_BLOCKS; ++hashBlock) { for (size_t hashBlock = 0; hashBlock < NUM_HASH_BLOCKS; ++hashBlock) {
__m128i cx; __m128i cx;
cx = _mm_load_si128((__m128i*) &l[hashBlock][idx[hashBlock] & MASK]);
if (SOFT_AES) { if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah[hashBlock], al[hashBlock])); cx = soft_aesenc(cx, _mm_set_epi64x(ah[hashBlock], al[hashBlock]));
} else { } else {
cx = _mm_load_si128((__m128i*) &l[hashBlock][idx[hashBlock] & MASK]);
cx = _mm_aesenc_si128(cx, _mm_set_epi64x(ah[hashBlock], al[hashBlock])); cx = _mm_aesenc_si128(cx, _mm_set_epi64x(ah[hashBlock], al[hashBlock]));
} }
@ -460,15 +460,17 @@ public:
idx = h[0] ^ h[4]; idx = h[0] ^ h[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx = _mm_load_si128((__m128i*) &l[idx & MASK]); __m128i cx;
if (SOFT_AES) { if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah, al)); cx = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
} else {
# ifndef XMRIG_ARMv7
cx = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah, al);
# endif
} }
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
}
_mm_store_si128((__m128i*) &l[idx & MASK], _mm_xor_si128(bx, cx)); _mm_store_si128((__m128i*) &l[idx & MASK], _mm_xor_si128(bx, cx));
idx = EXTRACT64(cx); idx = EXTRACT64(cx);
@ -528,13 +530,16 @@ public:
uint64_t idx1 = h1[0] ^h1[4]; uint64_t idx1 = h1[0] ^h1[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1)); cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1));
} else { } else {
cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]);
# ifndef XMRIG_ARMv7 # ifndef XMRIG_ARMv7
cx0 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx0, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0); 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); cx1 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx1, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah1, al1);
@ -630,20 +635,24 @@ public:
uint64_t idx1 = h1[0] ^h1[4]; uint64_t idx1 = h1[0] ^h1[4];
uint64_t idx2 = h2[0] ^h2[4]; uint64_t idx2 = h2[0] ^h2[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
__m128i cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]); __m128i cx2;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1)); cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
cx2 = soft_aesenc(cx2, _mm_set_epi64x(ah2, al2)); cx2 = soft_aesenc((uint32_t*)&l2[idx2 & MASK], _mm_set_epi64x(ah2, al2));
} else { }
# ifndef XMRIG_ARMv7 else {
cx0 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx0, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0); cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
cx1 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx1, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah1, al1); cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
cx2 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx2, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah2, al2); cx2 = _mm_load_si128((__m128i *) &l2[idx2 & 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);
cx2 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx2, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah2, al2);
# endif # endif
} }
@ -768,10 +777,10 @@ public:
uint64_t idx3 = h3[0] ^h3[4]; uint64_t idx3 = h3[0] ^h3[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
__m128i cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]); __m128i cx2;
__m128i cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]); __m128i cx3;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
@ -780,6 +789,11 @@ public:
cx3 = soft_aesenc(cx3, _mm_set_epi64x(ah3, al3)); cx3 = soft_aesenc(cx3, _mm_set_epi64x(ah3, al3));
} else { } else {
# ifndef XMRIG_ARMv7 # ifndef XMRIG_ARMv7
cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]);
cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]);
cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]);
cx0 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx0, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0); 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); cx1 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx1, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah1, al1);
cx2 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx2, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah2, al2); cx2 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx2, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah2, al2);
@ -937,11 +951,11 @@ public:
uint64_t idx4 = h4[0] ^h4[4]; uint64_t idx4 = h4[0] ^h4[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
__m128i cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]); __m128i cx2;
__m128i cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]); __m128i cx3;
__m128i cx4 = _mm_load_si128((__m128i*) &l4[idx4 & MASK]); __m128i cx4;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
@ -951,6 +965,12 @@ public:
cx4 = soft_aesenc(cx4, _mm_set_epi64x(ah4, al4)); cx4 = soft_aesenc(cx4, _mm_set_epi64x(ah4, al4));
} else { } else {
# ifndef XMRIG_ARMv7 # ifndef XMRIG_ARMv7
cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]);
cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]);
cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]);
cx4 = _mm_load_si128((__m128i*) &l4[idx4 & MASK]);
cx0 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx0, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah0, al0); 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); cx1 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx1, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah1, al1);
cx2 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx2, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah2, al2); cx2 = vreinterpretq_m128i_u8(vaesmcq_u8(vaeseq_u8(cx2, vdupq_n_u8(0)))) ^ _mm_set_epi64x(ah2, al2);

101
src/crypto/CryptoNight_x86.h
View file

@ -204,15 +204,16 @@ aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m12
__m128i* x7) __m128i* x7)
{ {
if (SOFT_AES) { if (SOFT_AES) {
*x0 = soft_aesenc(*x0, key); *x0 = soft_aesenc((uint32_t*)x0, key);
*x1 = soft_aesenc(*x1, key); *x1 = soft_aesenc((uint32_t*)x1, key);
*x2 = soft_aesenc(*x2, key); *x2 = soft_aesenc((uint32_t*)x2, key);
*x3 = soft_aesenc(*x3, key); *x3 = soft_aesenc((uint32_t*)x3, key);
*x4 = soft_aesenc(*x4, key); *x4 = soft_aesenc((uint32_t*)x4, key);
*x5 = soft_aesenc(*x5, key); *x5 = soft_aesenc((uint32_t*)x5, key);
*x6 = soft_aesenc(*x6, key); *x6 = soft_aesenc((uint32_t*)x6, key);
*x7 = soft_aesenc(*x7, key); *x7 = soft_aesenc((uint32_t*)x7, key);
} else { }
else {
*x0 = _mm_aesenc_si128(*x0, key); *x0 = _mm_aesenc_si128(*x0, key);
*x1 = _mm_aesenc_si128(*x1, key); *x1 = _mm_aesenc_si128(*x1, key);
*x2 = _mm_aesenc_si128(*x2, key); *x2 = _mm_aesenc_si128(*x2, key);
@ -352,8 +353,7 @@ public:
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
for (size_t hashBlock = 0; hashBlock < NUM_HASH_BLOCKS; ++hashBlock) { for (size_t hashBlock = 0; hashBlock < NUM_HASH_BLOCKS; ++hashBlock) {
__m128i cx; __m128i cx = _mm_load_si128((__m128i*) &l[hashBlock][idx[hashBlock] & MASK]);
cx = _mm_load_si128((__m128i*) &l[hashBlock][idx[hashBlock] & MASK]);
if (SOFT_AES) { if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah[hashBlock], al[hashBlock])); cx = soft_aesenc(cx, _mm_set_epi64x(ah[hashBlock], al[hashBlock]));
@ -421,11 +421,12 @@ public:
idx = h[0] ^ h[4]; idx = h[0] ^ h[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx = _mm_load_si128((__m128i*) &l[idx & MASK]); __m128i cx;
if (SOFT_AES) { if (SOFT_AES) {
cx = soft_aesenc(cx, _mm_set_epi64x(ah, al)); cx = soft_aesenc((uint32_t*)&l[idx & MASK], _mm_set_epi64x(ah, al));
} else { } else {
cx = _mm_load_si128((__m128i*) &l[idx & MASK]);
cx = _mm_aesenc_si128(cx, _mm_set_epi64x(ah, al)); cx = _mm_aesenc_si128(cx, _mm_set_epi64x(ah, al));
} }
@ -487,13 +488,16 @@ public:
uint64_t idx1 = h1[0] ^h1[4]; uint64_t idx1 = h1[0] ^h1[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1)); cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
} else { } 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)); cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0));
cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1)); cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1));
} }
@ -588,15 +592,19 @@ public:
uint64_t idx2 = h2[0] ^h2[4]; uint64_t idx2 = h2[0] ^h2[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
__m128i cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]); __m128i cx2;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1)); cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
cx2 = soft_aesenc(cx2, _mm_set_epi64x(ah2, al2)); cx2 = soft_aesenc((uint32_t*)&l2[idx2 & MASK], _mm_set_epi64x(ah2, al2));
} else { } else {
cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]);
cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]);
cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0)); cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0));
cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1)); cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1));
cx2 = _mm_aesenc_si128(cx2, _mm_set_epi64x(ah2, al2)); cx2 = _mm_aesenc_si128(cx2, _mm_set_epi64x(ah2, al2));
@ -723,17 +731,22 @@ public:
uint64_t idx3 = h3[0] ^h3[4]; uint64_t idx3 = h3[0] ^h3[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
__m128i cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]); __m128i cx2;
__m128i cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]); __m128i cx3;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1)); cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
cx2 = soft_aesenc(cx2, _mm_set_epi64x(ah2, al2)); cx2 = soft_aesenc((uint32_t*)&l2[idx2 & MASK], _mm_set_epi64x(ah2, al2));
cx3 = soft_aesenc(cx3, _mm_set_epi64x(ah3, al3)); cx3 = soft_aesenc((uint32_t*)&l3[idx3 & MASK], _mm_set_epi64x(ah3, al3));
} else { } else {
cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]);
cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]);
cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]);
cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0)); cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0));
cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1)); cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1));
cx2 = _mm_aesenc_si128(cx2, _mm_set_epi64x(ah2, al2)); cx2 = _mm_aesenc_si128(cx2, _mm_set_epi64x(ah2, al2));
@ -890,19 +903,25 @@ public:
uint64_t idx4 = h4[0] ^h4[4]; uint64_t idx4 = h4[0] ^h4[4];
for (size_t i = 0; i < ITERATIONS; i++) { for (size_t i = 0; i < ITERATIONS; i++) {
__m128i cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]); __m128i cx0;
__m128i cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]); __m128i cx1;
__m128i cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]); __m128i cx2;
__m128i cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]); __m128i cx3;
__m128i cx4 = _mm_load_si128((__m128i*) &l4[idx4 & MASK]); __m128i cx4;
if (SOFT_AES) { if (SOFT_AES) {
cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0)); cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1)); cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
cx2 = soft_aesenc(cx2, _mm_set_epi64x(ah2, al2)); cx2 = soft_aesenc((uint32_t*)&l2[idx2 & MASK], _mm_set_epi64x(ah2, al2));
cx3 = soft_aesenc(cx3, _mm_set_epi64x(ah3, al3)); cx3 = soft_aesenc((uint32_t*)&l3[idx3 & MASK], _mm_set_epi64x(ah3, al3));
cx4 = soft_aesenc(cx4, _mm_set_epi64x(ah4, al4)); cx4 = soft_aesenc((uint32_t*)&l4[idx4 & MASK], _mm_set_epi64x(ah4, al4));
} else { } else {
cx0 = _mm_load_si128((__m128i*) &l0[idx0 & MASK]);
cx1 = _mm_load_si128((__m128i*) &l1[idx1 & MASK]);
cx2 = _mm_load_si128((__m128i*) &l2[idx2 & MASK]);
cx3 = _mm_load_si128((__m128i*) &l3[idx3 & MASK]);
cx4 = _mm_load_si128((__m128i*) &l4[idx4 & MASK]);
cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0)); cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0));
cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1)); cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1));
cx2 = _mm_aesenc_si128(cx2, _mm_set_epi64x(ah2, al2)); cx2 = _mm_aesenc_si128(cx2, _mm_set_epi64x(ah2, al2));

32
src/crypto/soft_aes.h
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 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); 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 x0 = in[0];
const uint32_t x1 = _mm_cvtsi128_si32(_mm_shuffle_epi32(in, 0x55)); const uint32_t x1 = in[1];
const uint32_t x2 = _mm_cvtsi128_si32(_mm_shuffle_epi32(in, 0xAA)); const uint32_t x2 = in[2];
const uint32_t x3 = _mm_cvtsi128_si32(_mm_shuffle_epi32(in, 0xFF)); const uint32_t x3 = in[3];
__m128i out = _mm_set_epi32( __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][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][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][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])); (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) static inline uint32_t sub_word(uint32_t key)
{ {
return (saes_sbox[key >> 24 ] << 24) | return (saes_sbox[key >> 24 ] << 24) |
(saes_sbox[(key >> 16) & 0xff] << 16 ) | (saes_sbox[(key >> 16) & 0xff] << 16 ) |
(saes_sbox[(key >> 8) & 0xff] << 8 ) | (saes_sbox[(key >> 8) & 0xff] << 8 ) |
saes_sbox[key & 0xff]; saes_sbox[key & 0xff];
} }
#if defined(__clang__) || defined(XMRIG_ARM) #if defined(__clang__) || defined(XMRIG_ARM)
static inline uint32_t _rotr(uint32_t value, uint32_t amount) 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 #endif