popov/cpuminer-opt-gpu
1
0
Fork 0
mirror of https://github.com/JayDDee/cpuminer-opt.git synced 2025-09-17 23:44:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

v3.9.1

Browse Source
This commit is contained in:
Jay D Dee
2019-05-30 16:59:49 -04:00
parent eb3f57bfc7
commit 77c5ae80ab
82 changed files with 6906 additions and 3706 deletions
Download Patch File Download Diff File Expand all files Collapse all files

406
algo/shavite/shavite-hash-2way.c Normal file
View File

@@ -0,0 +1,406 @@
#include "shavite-hash-2way.h"
#include "algo/sha/sph_types.h"
#include <stdio.h>
#if defined(__AVX2__)
static const uint32_t IV512[] =
{
0x72FCCDD8, 0x79CA4727, 0x128A077B, 0x40D55AEC,
0xD1901A06, 0x430AE307, 0xB29F5CD1, 0xDF07FBFC,
0x8E45D73D, 0x681AB538, 0xBDE86578, 0xDD577E47,
0xE275EADE, 0x502D9FCD, 0xB9357178, 0x022A4B9A
};
#define mm256_ror2x256hi_1x32( a, b ) \
_mm256_blend_epi32( mm256_ror1x32_128( a ), \
mm256_ror1x32_128( b ), 0x88 )
static void
c512_2way( shavite512_2way_context *ctx, const void *msg )
{
__m256i p0, p1, p2, p3, x;
__m256i k00, k01, k02, k03, k10, k11, k12, k13;
__m256i *m = (__m256i*)msg;
__m256i *h = (__m256i*)ctx->h;
int r;
p0 = h[0];
p1 = h[1];
p2 = h[2];
p3 = h[3];
// round
k00 = m[0];
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k00 ) );
k01 = m[1];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = m[2];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = m[3];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p0 = _mm256_xor_si256( p0, x );
k10 = m[4];
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k10 ) );
k11 = m[5];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = m[6];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = m[7];
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p2 = _mm256_xor_si256( p2, x );
for ( r = 0; r < 3; r ++ )
{
// round 1, 5, 9
k00 = _mm256_xor_si256( k13, mm256_ror1x32_128(
mm256_aesenc_2x128( k00 ) ) );
if ( r == 0 )
k00 = _mm256_xor_si256( k00, _mm256_set_epi32(
~ctx->count3, ctx->count2, ctx->count1, ctx->count0,
~ctx->count3, ctx->count2, ctx->count1, ctx->count0 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p0, k00 ) );
k01 = _mm256_xor_si256( k00,
mm256_ror1x32_128( mm256_aesenc_2x128( k01 ) ) );
if ( r == 1 )
k01 = _mm256_xor_si256( k01, _mm256_set_epi32(
~ctx->count0, ctx->count1, ctx->count2, ctx->count3,
~ctx->count0, ctx->count1, ctx->count2, ctx->count3 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( k01,
mm256_ror1x32_128( mm256_aesenc_2x128( k02 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( k02,
mm256_ror1x32_128( mm256_aesenc_2x128( k03 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p3 = _mm256_xor_si256( p3, x );
k10 = _mm256_xor_si256( k03,
mm256_ror1x32_128( mm256_aesenc_2x128( k10 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p2, k10 ) );
k11 = _mm256_xor_si256( k10,
mm256_ror1x32_128( mm256_aesenc_2x128( k11 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( k11,
mm256_ror1x32_128( mm256_aesenc_2x128( k12 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( k12,
mm256_ror1x32_128( mm256_aesenc_2x128( k13 ) ) );
if ( r == 2 )
k13 = _mm256_xor_si256( k13, _mm256_set_epi32(
~ctx->count1, ctx->count0, ctx->count3, ctx->count2,
~ctx->count1, ctx->count0, ctx->count3, ctx->count2 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p1 = _mm256_xor_si256( p1, x );
// round 2, 6, 10
k00 = _mm256_xor_si256( k00, mm256_ror2x256hi_1x32( k12, k13 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k00 ) );
k01 = _mm256_xor_si256( k01, mm256_ror2x256hi_1x32( k13, k00 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( k02, mm256_ror2x256hi_1x32( k00, k01 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( k03, mm256_ror2x256hi_1x32( k01, k02 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p2 = _mm256_xor_si256( p2, x );
k10 = _mm256_xor_si256( k10, mm256_ror2x256hi_1x32( k02, k03 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k10 ) );
k11 = _mm256_xor_si256( k11, mm256_ror2x256hi_1x32( k03, k10 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( k12, mm256_ror2x256hi_1x32( k10, k11 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( k13, mm256_ror2x256hi_1x32( k11, k12 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p0 = _mm256_xor_si256( p0, x );
// round 3, 7, 11
k00 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k00 ) ), k13 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p2, k00 ) );
k01 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k01 ) ), k00 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k02 ) ), k01 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k03 ) ), k02 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p1 = _mm256_xor_si256( p1, x );
k10 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k10 ) ), k03 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p0, k10 ) );
k11 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k11 ) ), k10 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k12 ) ), k11 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k13 ) ), k12 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p3 = _mm256_xor_si256( p3, x );
// round 4, 8, 12
k00 = _mm256_xor_si256( k00, mm256_ror2x256hi_1x32( k12, k13 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p1, k00 ) );
k01 = _mm256_xor_si256( k01, mm256_ror2x256hi_1x32( k13, k00 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( k02, mm256_ror2x256hi_1x32( k00, k01 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( k03, mm256_ror2x256hi_1x32( k01, k02 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p0 = _mm256_xor_si256( p0, x );
k10 = _mm256_xor_si256( k10, mm256_ror2x256hi_1x32( k02, k03 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( p3, k10 ) );
k11 = _mm256_xor_si256( k11, mm256_ror2x256hi_1x32( k03, k10 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = _mm256_xor_si256( k12, mm256_ror2x256hi_1x32( k10, k11 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( k13, mm256_ror2x256hi_1x32( k11, k12 ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p2 = _mm256_xor_si256( p2, x );
}
// round 13
k00 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k00 ) ), k13 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p0, k00 ) );
k01 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k01 ) ), k00 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k01 ) );
k02 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k02 ) ), k01 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k02 ) );
k03 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k03 ) ), k02 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k03 ) );
p3 = _mm256_xor_si256( p3, x );
k10 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k10 ) ), k03 );
x = mm256_aesenc_2x128( _mm256_xor_si256( p2, k10 ) );
k11 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k11 ) ), k10 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k11 ) );
k12 = mm256_ror1x32_128( mm256_aesenc_2x128( k12 ) );
k12 = _mm256_xor_si256( k12, _mm256_xor_si256( k11, _mm256_set_epi32(
~ctx->count2, ctx->count3, ctx->count0, ctx->count1,
~ctx->count2, ctx->count3, ctx->count0, ctx->count1 ) ) );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k12 ) );
k13 = _mm256_xor_si256( mm256_ror1x32_128(
mm256_aesenc_2x128( k13 ) ), k12 );
x = mm256_aesenc_2x128( _mm256_xor_si256( x, k13 ) );
p1 = _mm256_xor_si256( p1, x );
h[0] = _mm256_xor_si256( h[0], p2 );
h[1] = _mm256_xor_si256( h[1], p3 );
h[2] = _mm256_xor_si256( h[2], p0 );
h[3] = _mm256_xor_si256( h[3], p1 );
}
void shavite512_2way_init( shavite512_2way_context *ctx )
{
casti_m256i( ctx->h, 0 ) =
_mm256_set_epi32( IV512[ 3], IV512[ 2], IV512[ 1], IV512[ 0],
IV512[ 3], IV512[ 2], IV512[ 1], IV512[ 0] );
casti_m256i( ctx->h, 1 ) =
_mm256_set_epi32( IV512[ 7], IV512[ 6], IV512[ 5], IV512[ 4],
IV512[ 7], IV512[ 6], IV512[ 5], IV512[ 4] );
casti_m256i( ctx->h, 2 ) =
_mm256_set_epi32( IV512[11], IV512[10], IV512[ 9], IV512[ 8],
IV512[11], IV512[10], IV512[ 9], IV512[ 8] );
casti_m256i( ctx->h, 3 ) =
_mm256_set_epi32( IV512[15], IV512[14], IV512[13], IV512[12],
IV512[15], IV512[14], IV512[13], IV512[12] );
ctx->ptr = 0;
ctx->count0 = 0;
ctx->count1 = 0;
ctx->count2 = 0;
ctx->count3 = 0;
}
void shavite512_2way_update( shavite512_2way_context *ctx, const void *data,
size_t len )
{
unsigned char *buf = ctx->buf;
size_t ptr = ctx->ptr;
while ( len > 0 )
{
size_t clen;
clen = (sizeof ctx->buf) - ptr;
if ( clen > len << 1 )
clen = len << 1;
memcpy( buf + ptr, data, clen );
data = (const unsigned char *)data + clen;
ptr += clen;
len -= clen >> 1;
if ( ptr == sizeof ctx->buf )
{
if ( ( ctx->count0 = ctx->count0 + 1024 ) == 0 )
{
ctx->count1 = ctx->count1 + 1;
if ( ctx->count1 == 0 )
{
ctx->count2 = ctx->count2 + 1;
if ( ctx->count2 == 0 )
ctx->count3 = ctx->count3 + 1;
}
}
c512_2way( ctx, buf );
ptr = 0;
}
}
ctx->ptr = ptr;
}
void shavite512_2way_close( shavite512_2way_context *ctx, void *dst )
{
unsigned char *buf;
union
{
uint32_t u32[4];
uint16_t u16[8];
} count;
buf = ctx->buf;
uint32_t vp = ctx->ptr>>5;
// Terminating byte then zero pad
casti_m256i( buf, vp++ ) = _mm256_set_epi32( 0,0,0,0x80, 0,0,0,0x80 );
// Zero pad full vectors up to count
for ( ; vp < 6; vp++ )
casti_m256i( buf, vp ) = m256_zero;
// Count = { 0, 16, 64, 80 }. Outsize = 16 u32 = 512 bits = 0x0200
// Count is misaligned to 16 bits and straddles a vector.
// Use u32 overlay to stage then u16 to load buf.
count.u32[0] = ctx->count0 += (ctx->ptr << 2); // ptr/2 * 8
count.u32[1] = ctx->count1;
count.u32[2] = ctx->count2;
count.u32[3] = ctx->count3;
casti_m256i( buf, 6 ) = _mm256_set_epi16( count.u16[0], 0,0,0,0,0,0,0,
count.u16[0], 0,0,0,0,0,0,0 );
casti_m256i( buf, 7 ) = _mm256_set_epi16(
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1],
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] );
c512_2way( ctx, buf);
casti_m256i( dst, 0 ) = casti_m256i( ctx->h, 0 );
casti_m256i( dst, 1 ) = casti_m256i( ctx->h, 1 );
casti_m256i( dst, 2 ) = casti_m256i( ctx->h, 2 );
casti_m256i( dst, 3 ) = casti_m256i( ctx->h, 3 );
}
void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
const void *data, size_t len )
{
unsigned char *buf = ctx->buf;
size_t ptr = ctx->ptr;
// process full blocks and load buf with remainder.
while ( len > 0 )
{
size_t clen;
clen = (sizeof ctx->buf) - ptr;
if ( clen > len << 1 )
clen = len << 1;
memcpy( buf + ptr, data, clen );
data = (const unsigned char *)data + clen;
ptr += clen;
len -= clen >> 1;
if ( ptr == sizeof ctx->buf )
{
if ( ( ctx->count0 = ctx->count0 + 1024 ) == 0 )
{
ctx->count1 = ctx->count1 + 1;
if ( ctx->count1 == 0 )
{
ctx->count2 = ctx->count2 + 1;
if ( ctx->count2 == 0 )
ctx->count3 = ctx->count3 + 1;
}
}
c512_2way( ctx, buf );
ptr = 0;
}
}
uint32_t vp = ptr>>5;
// Terminating byte then zero pad
casti_m256i( buf, vp++ ) = _mm256_set_epi32( 0,0,0,0x80, 0,0,0,0x80 );
// Zero pad full vectors up to count
for ( ; vp < 6; vp++ )
casti_m256i( buf, vp ) = m256_zero;
// Count = { 0, 16, 64, 80 }. Outsize = 16 u32 = 512 bits = 0x0200
// Count is misaligned to 16 bits and straddles a vector.
// Use u32 overlay to stage then u16 to load buf.
union
{
uint32_t u32[4];
uint16_t u16[8];
} count;
count.u32[0] = ctx->count0 += (ptr << 2); // ptr/2 * 8
count.u32[1] = ctx->count1;
count.u32[2] = ctx->count2;
count.u32[3] = ctx->count3;
casti_m256i( buf, 6 ) = _mm256_set_epi16( count.u16[0], 0,0,0,0,0,0,0,
count.u16[0], 0,0,0,0,0,0,0 );
casti_m256i( buf, 7 ) = _mm256_set_epi16(
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1],
0x0200 , count.u16[7], count.u16[6], count.u16[5],
count.u16[4], count.u16[3], count.u16[2], count.u16[1] );
c512_2way( ctx, buf);
casti_m256i( dst, 0 ) = casti_m256i( ctx->h, 0 );
casti_m256i( dst, 1 ) = casti_m256i( ctx->h, 1 );
casti_m256i( dst, 2 ) = casti_m256i( ctx->h, 2 );
casti_m256i( dst, 3 ) = casti_m256i( ctx->h, 3 );
}
#endif // AVX2

25
algo/shavite/shavite-hash-2way.h Normal file
View File

@@ -0,0 +1,25 @@
#ifndef SHAVITE_HASH_2WAY_H__
#define SHAVITE_HASH_2WAY_H__
#if defined(__AVX2__)
#include "avxdefs.h"
typedef struct {
unsigned char buf[128<<1];
uint32_t h[16<<1];
size_t ptr;
uint32_t count0, count1, count2, count3;
} shavite512_2way_context __attribute__ ((aligned (64)));
void shavite512_2way_init( shavite512_2way_context *ctx );
void shavite512_2way_update( shavite512_2way_context *ctx, const void *data,
size_t len );
void shavite512_2way_close( shavite512_2way_context *ctx, void *dst );
void shavite512_2way_update_close( shavite512_2way_context *ctx, void *dst,
const void *data, size_t len );
#endif // AVX2
#endif // SHAVITE_HASH_2WAY_H__

62
algo/shavite/sph-shavite-aesni.c
View File

@@ -102,35 +102,31 @@ c512( sph_shavite_big_context *sc, const void *msg )
k00 = m[0];
x = _mm_xor_si128( p1, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = m[1];
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = m[2];
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = m[3];
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p0 = _mm_xor_si128( p0, x );
k10 = m[4];
x = _mm_xor_si128( p3, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = m[5];
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = m[6];
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = m[7];
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p2 = _mm_xor_si128( p2, x );
for ( r = 0; r < 3; r ++ )
@@ -156,15 +152,15 @@ c512( sph_shavite_big_context *sc, const void *msg )
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm128_ror_1x32( _mm_aesenc_si128( k02, m128_zero ) );
k02 = _mm_xor_si128( k02, k01 );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = mm128_ror_1x32( _mm_aesenc_si128( k03, m128_zero ) );
k03 = _mm_xor_si128( k03, k02 );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p3 = _mm_xor_si128( p3, x );
k10 = mm128_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
k10 = _mm_xor_si128( k10, k03 );
@@ -172,12 +168,10 @@ c512( sph_shavite_big_context *sc, const void *msg )
x = _mm_aesenc_si128( x, m128_zero );
k11 = mm128_ror_1x32( _mm_aesenc_si128( k11, m128_zero ) );
k11 = _mm_xor_si128( k11, k10 );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = mm128_ror_1x32( _mm_aesenc_si128( k12, m128_zero ) );
k12 = _mm_xor_si128( k12, k11 );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
@@ -196,118 +190,103 @@ c512( sph_shavite_big_context *sc, const void *msg )
k00 = _mm_xor_si128( k00, mm128_ror256hi_1x32( k12, k13 ) );
x = _mm_xor_si128( p3, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = _mm_xor_si128( k01, mm128_ror256hi_1x32( k13, k00 ) );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = _mm_xor_si128( k02, mm128_ror256hi_1x32( k00, k01 ) );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = _mm_xor_si128( k03, mm128_ror256hi_1x32( k01, k02 ) );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p2 = _mm_xor_si128( p2, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
x = _mm_xor_si128( p1, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = _mm_xor_si128( k11, mm128_ror256hi_1x32( k03, k10 ) );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = _mm_xor_si128( k12, mm128_ror256hi_1x32( k10, k11 ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p0 = _mm_xor_si128( p0, x );
// round 3, 7, 11
k00 = mm128_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
k00 = _mm_xor_si128( k00, k13 );
x = _mm_xor_si128( p2, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = mm128_ror_1x32( _mm_aesenc_si128( k01, m128_zero ) );
k01 = _mm_xor_si128( k01, k00 );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm128_ror_1x32( _mm_aesenc_si128( k02, m128_zero ) );
k02 = _mm_xor_si128( k02, k01 );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = mm128_ror_1x32( _mm_aesenc_si128( k03, m128_zero ) );
k03 = _mm_xor_si128( k03, k02 );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p1 = _mm_xor_si128( p1, x );
k10 = mm128_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
k10 = _mm_xor_si128( k10, k03 );
x = _mm_xor_si128( p0, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = mm128_ror_1x32( _mm_aesenc_si128( k11, m128_zero ) );
k11 = _mm_xor_si128( k11, k10 );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = mm128_ror_1x32( _mm_aesenc_si128( k12, m128_zero ) );
k12 = _mm_xor_si128( k12, k11 );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p3 = _mm_xor_si128( p3, x );
// round 4, 8, 12
k00 = _mm_xor_si128( k00, mm128_ror256hi_1x32( k12, k13 ) );
x = _mm_xor_si128( p1, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = _mm_xor_si128( k01, mm128_ror256hi_1x32( k13, k00 ) );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = _mm_xor_si128( k02, mm128_ror256hi_1x32( k00, k01 ) );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = _mm_xor_si128( k03, mm128_ror256hi_1x32( k01, k02 ) );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p0 = _mm_xor_si128( p0, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
p0 = _mm_xor_si128( p0, x );
k10 = _mm_xor_si128( k10, mm128_ror256hi_1x32( k02, k03 ) );
x = _mm_xor_si128( p3, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = _mm_xor_si128( k11, mm128_ror256hi_1x32( k03, k10 ) );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = _mm_xor_si128( k12, mm128_ror256hi_1x32( k10, k11 ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p2 = _mm_xor_si128( p2, x );
}
@@ -315,46 +294,41 @@ c512( sph_shavite_big_context *sc, const void *msg )
k00 = mm128_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
k00 = _mm_xor_si128( k00, k13 );
x = _mm_xor_si128( p0, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = mm128_ror_1x32( _mm_aesenc_si128( k01, m128_zero ) );
k01 = _mm_xor_si128( k01, k00 );
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm128_ror_1x32( _mm_aesenc_si128( k02, m128_zero ) );
k02 = _mm_xor_si128( k02, k01 );
x = _mm_xor_si128( x, k02 );
x = _mm_aesenc_si128( x, m128_zero );
k03 = mm128_ror_1x32( _mm_aesenc_si128( k03, m128_zero ) );
k03 = _mm_xor_si128( k03, k02 );
x = _mm_xor_si128( x, k03 );
x = _mm_aesenc_si128( x, m128_zero );
p3 = _mm_xor_si128( p3, x );
k10 = mm128_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
k10 = _mm_xor_si128( k10, k03 );
x = _mm_xor_si128( p2, k10 );
x = _mm_aesenc_si128( x, m128_zero );
k11 = mm128_ror_1x32( _mm_aesenc_si128( k11, m128_zero ) );
k11 = _mm_xor_si128( k11, k10 );
x = _mm_xor_si128( x, k11 );
x = _mm_aesenc_si128( x, m128_zero );
k12 = mm128_ror_1x32( _mm_aesenc_si128( k12, m128_zero ) );
k12 = _mm_xor_si128( k12, _mm_xor_si128( k11, _mm_set_epi32(
~sc->count2, sc->count3, sc->count0, sc->count1 ) ) );
x = _mm_xor_si128( x, k12 );
x = _mm_aesenc_si128( x, m128_zero );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
p1 = _mm_xor_si128( p1, x );
h[0] = _mm_xor_si128( h[0], p2 );
@@ -427,6 +401,9 @@ shavite_big_aesni_close( sph_shavite_big_context *sc, unsigned ub, unsigned n,
count1 = sc->count1;
count2 = sc->count2;
count3 = sc->count3;
z = 0x80 >> n;
z = ((ub & -z) | z) & 0xFF;
if (ptr == 0 && n == 0) {
@@ -443,6 +420,7 @@ shavite_big_aesni_close( sph_shavite_big_context *sc, unsigned ub, unsigned n,
memset(buf, 0, 110);
sc->count0 = sc->count1 = sc->count2 = sc->count3 = 0;
}
sph_enc32le(buf + 110, count0);
sph_enc32le(buf + 114, count1);
sph_enc32le(buf + 118, count2);

1764
algo/shavite/sse2/shavite.c
View File

File diff suppressed because it is too large Load Diff