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

v3.22.3

Browse Source
This commit is contained in:
Jay D Dee
2023-06-14 11:07:40 -04:00
parent de564ccbde
commit 57a6b7b58b
31 changed files with 3724 additions and 3345 deletions
Download Patch File Download Diff File Expand all files Collapse all files

37
algo/groestl/groestl256-hash-4way.c
View File

@@ -33,8 +33,7 @@ int groestl256_4way_init( groestl256_4way_context* ctx, uint64_t hashlen )
}
// The only non-zero in the IV is len. It can be hard coded.
ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
ctx->chaining[ 3 ] = mm512_bcast128lo_64( 0x0100000000000000 );
ctx->buf_ptr = 0;
ctx->rem_ptr = 0;
@@ -51,9 +50,6 @@ int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
__m512i* in = (__m512i*)input;
int i;
// if (ctx->chaining == NULL || ctx->buffer == NULL)
// return 1;
for ( i = 0; i < SIZE256; i++ )
{
ctx->chaining[i] = m512_zero;
@@ -61,7 +57,7 @@ int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
}
// The only non-zero in the IV is len. It can be hard coded.
ctx->chaining[ 3 ] = m512_const2_64( 0, 0x0100000000000000 );
ctx->chaining[ 3 ] = mm512_bcast128lo_64( 0x0100000000000000 );
ctx->buf_ptr = 0;
// --- update ---
@@ -83,18 +79,18 @@ int groestl256_4way_full( groestl256_4way_context* ctx, void* output,
if ( i == SIZE256 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
}
else
{
// add first padding
ctx->buffer[i] = m512_const2_64( 0, 0x80 );
ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
// add zero padding
for ( i += 1; i < SIZE256 - 1; i++ )
ctx->buffer[i] = m512_zero;
// add length padding, second last byte is zero unless blocks > 255
ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
}
// digest final padding block and do output transform
@@ -140,18 +136,18 @@ int groestl256_4way_update_close( groestl256_4way_context* ctx, void* output,
if ( i == SIZE256 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
}
else
{
// add first padding
ctx->buffer[i] = m512_const2_64( 0, 0x80 );
ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
// add zero padding
for ( i += 1; i < SIZE256 - 1; i++ )
ctx->buffer[i] = m512_zero;
// add length padding, second last byte is zero unless blocks > 255
ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
}
// digest final padding block and do output transform
@@ -186,7 +182,7 @@ int groestl256_2way_init( groestl256_2way_context* ctx, uint64_t hashlen )
}
// The only non-zero in the IV is len. It can be hard coded.
ctx->chaining[ 3 ] = m256_const2_64( 0, 0x0100000000000000 );
ctx->chaining[ 3 ] = mm256_bcast128lo_64( 0x0100000000000000 );
ctx->buf_ptr = 0;
ctx->rem_ptr = 0;
@@ -211,7 +207,7 @@ int groestl256_2way_full( groestl256_2way_context* ctx, void* output,
}
// The only non-zero in the IV is len. It can be hard coded.
ctx->chaining[ 3 ] = m256_const2_64( 0, 0x0100000000000000 );
ctx->chaining[ 3 ] = mm256_bcast128lo_64( 0x0100000000000000 );
ctx->buf_ptr = 0;
// --- update ---
@@ -233,18 +229,18 @@ int groestl256_2way_full( groestl256_2way_context* ctx, void* output,
if ( i == SIZE256 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
}
else
{
// add first padding
ctx->buffer[i] = m256_const2_64( 0, 0x80 );
ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
// add zero padding
for ( i += 1; i < SIZE256 - 1; i++ )
ctx->buffer[i] = m256_zero;
// add length padding, second last byte is zero unless blocks > 255
ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
}
// digest final padding block and do output transform
@@ -289,23 +285,22 @@ int groestl256_2way_update_close( groestl256_2way_context* ctx, void* output,
if ( i == SIZE256 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
}
else
{
// add first padding
ctx->buffer[i] = m256_const2_64( 0, 0x80 );
ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
// add zero padding
for ( i += 1; i < SIZE256 - 1; i++ )
ctx->buffer[i] = m256_zero;
// add length padding, second last byte is zero unless blocks > 255
ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
}
// digest final padding block and do output transform
TF512_2way( ctx->chaining, ctx->buffer );
OF512_2way( ctx->chaining );
// store hash result in output

170
algo/groestl/groestl256-intr-4way.h
View File

@@ -165,7 +165,7 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
\
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
/* compute w_i : add y_{i+4} */\
b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b ); \
b1 = _mm512_set1_epi64( 0x1b1b1b1b1b1b1b1b ); \
MUL2( a0, b0, b1 ); \
a0 = _mm512_xor_si512( a0, TEMP0 ); \
MUL2( a1, b0, b1 ); \
@@ -205,116 +205,18 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
b1 = _mm512_xor_si512( b1, a4 ); \
}/*MixBytes*/
#if 0
#define MixBytes(a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
/* t_i = a_i + a_{i+1} */\
b6 = a0;\
b7 = a1;\
a0 = _mm512_xor_si512(a0, a1);\
b0 = a2;\
a1 = _mm512_xor_si512(a1, a2);\
b1 = a3;\
a2 = _mm512_xor_si512(a2, a3);\
b2 = a4;\
a3 = _mm512_xor_si512(a3, a4);\
b3 = a5;\
a4 = _mm512_xor_si512(a4, a5);\
b4 = a6;\
a5 = _mm512_xor_si512(a5, a6);\
b5 = a7;\
a6 = _mm512_xor_si512(a6, a7);\
a7 = _mm512_xor_si512(a7, b6);\
\
/* build y4 y5 y6 ... in regs xmm8, xmm9, xmm10 by adding t_i*/\
b0 = _mm512_xor_si512(b0, a4);\
b6 = _mm512_xor_si512(b6, a4);\
b1 = _mm512_xor_si512(b1, a5);\
b7 = _mm512_xor_si512(b7, a5);\
b2 = _mm512_xor_si512(b2, a6);\
b0 = _mm512_xor_si512(b0, a6);\
/* spill values y_4, y_5 to memory */\
TEMP0 = b0;\
b3 = _mm512_xor_si512(b3, a7);\
b1 = _mm512_xor_si512(b1, a7);\
TEMP1 = b1;\
b4 = _mm512_xor_si512(b4, a0);\
b2 = _mm512_xor_si512(b2, a0);\
/* save values t0, t1, t2 to xmm8, xmm9 and memory */\
b0 = a0;\
b5 = _mm512_xor_si512(b5, a1);\
b3 = _mm512_xor_si512(b3, a1);\
b1 = a1;\
b6 = _mm512_xor_si512(b6, a2);\
b4 = _mm512_xor_si512(b4, a2);\
TEMP2 = a2;\
b7 = _mm512_xor_si512(b7, a3);\
b5 = _mm512_xor_si512(b5, a3);\
\
/* compute x_i = t_i + t_{i+3} */\
a0 = _mm512_xor_si512(a0, a3);\
a1 = _mm512_xor_si512(a1, a4);\
a2 = _mm512_xor_si512(a2, a5);\
a3 = _mm512_xor_si512(a3, a6);\
a4 = _mm512_xor_si512(a4, a7);\
a5 = _mm512_xor_si512(a5, b0);\
a6 = _mm512_xor_si512(a6, b1);\
a7 = _mm512_xor_si512(a7, TEMP2);\
\
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
/* compute w_i : add y_{i+4} */\
b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b );\
MUL2(a0, b0, b1);\
a0 = _mm512_xor_si512(a0, TEMP0);\
MUL2(a1, b0, b1);\
a1 = _mm512_xor_si512(a1, TEMP1);\
MUL2(a2, b0, b1);\
a2 = _mm512_xor_si512(a2, b2);\
MUL2(a3, b0, b1);\
a3 = _mm512_xor_si512(a3, b3);\
MUL2(a4, b0, b1);\
a4 = _mm512_xor_si512(a4, b4);\
MUL2(a5, b0, b1);\
a5 = _mm512_xor_si512(a5, b5);\
MUL2(a6, b0, b1);\
a6 = _mm512_xor_si512(a6, b6);\
MUL2(a7, b0, b1);\
a7 = _mm512_xor_si512(a7, b7);\
\
/* compute v_i : double w_i */\
/* add to y_4 y_5 .. v3, v4, ... */\
MUL2(a0, b0, b1);\
b5 = _mm512_xor_si512(b5, a0);\
MUL2(a1, b0, b1);\
b6 = _mm512_xor_si512(b6, a1);\
MUL2(a2, b0, b1);\
b7 = _mm512_xor_si512(b7, a2);\
MUL2(a5, b0, b1);\
b2 = _mm512_xor_si512(b2, a5);\
MUL2(a6, b0, b1);\
b3 = _mm512_xor_si512(b3, a6);\
MUL2(a7, b0, b1);\
b4 = _mm512_xor_si512(b4, a7);\
MUL2(a3, b0, b1);\
MUL2(a4, b0, b1);\
b0 = TEMP0;\
b1 = TEMP1;\
b0 = _mm512_xor_si512(b0, a3);\
b1 = _mm512_xor_si512(b1, a4);\
}/*MixBytes*/
#endif
#define MASK_NOT( a ) _mm512_mask_ternarylogic_epi64( a, 0xaa, a, a, 1 )
#define ROUND(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
/* AddRoundConstant */\
b1 = m512_const2_64( 0xffffffffffffffff, 0 ); \
a0 = _mm512_xor_si512( a0, m512_const1_128( round_const_l0[i] ) );\
a1 = _mm512_xor_si512( a1, b1 );\
a2 = _mm512_xor_si512( a2, b1 );\
a3 = _mm512_xor_si512( a3, b1 );\
a4 = _mm512_xor_si512( a4, b1 );\
a5 = _mm512_xor_si512( a5, b1 );\
a6 = _mm512_xor_si512( a6, b1 );\
a7 = _mm512_xor_si512( a7, m512_const1_128( round_const_l7[i] ) );\
a0 = _mm512_xor_si512( a0, mm512_bcast_m128( round_const_l0[i] ) );\
a1 = MASK_NOT( a1 ); \
a2 = MASK_NOT( a2 ); \
a3 = MASK_NOT( a3 ); \
a4 = MASK_NOT( a4 ); \
a5 = MASK_NOT( a5 ); \
a6 = MASK_NOT( a6 ); \
a7 = _mm512_xor_si512( a7, mm512_bcast_m128( round_const_l7[i] ) );\
\
/* ShiftBytes + SubBytes (interleaved) */\
b0 = _mm512_xor_si512( b0, b0 );\
@@ -450,7 +352,7 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
* outputs: (i0-7) = (0|S)
*/
#define Matrix_Transpose_O_B(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
t0 = _mm512_xor_si512( t0, t0 );\
t0 = m512_zero;\
i1 = i0;\
i3 = i2;\
i5 = i4;\
@@ -481,11 +383,11 @@ static const __m512i SUBSH_MASK7 = { 0x090c000306080b07, 0x02050f0a0d01040e,
void TF512_4way( __m512i* chaining, __m512i* message )
{
static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m512i TEMP0;
static __m512i TEMP1;
static __m512i TEMP2;
__m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m512i TEMP0;
__m512i TEMP1;
__m512i TEMP2;
/* load message into registers xmm12 - xmm15 */
xmm12 = message[0];
@@ -547,11 +449,11 @@ void TF512_4way( __m512i* chaining, __m512i* message )
void OF512_4way( __m512i* chaining )
{
static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m512i TEMP0;
static __m512i TEMP1;
static __m512i TEMP2;
__m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m512i TEMP0;
__m512i TEMP1;
__m512i TEMP2;
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
xmm8 = chaining[0];
@@ -696,7 +598,7 @@ static const __m256i SUBSH_MASK7_2WAY =
\
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
/* compute w_i : add y_{i+4} */\
b1 = m256_const1_64( 0x1b1b1b1b1b1b1b1b );\
b1 = _mm256_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
MUL2_2WAY(a0, b0, b1);\
a0 = _mm256_xor_si256(a0, TEMP0);\
MUL2_2WAY(a1, b0, b1);\
@@ -738,15 +640,15 @@ static const __m256i SUBSH_MASK7_2WAY =
#define ROUND_2WAY(i, a0, a1, a2, a3, a4, a5, a6, a7, b0, b1, b2, b3, b4, b5, b6, b7){\
/* AddRoundConstant */\
b1 = m256_const2_64( 0xffffffffffffffff, 0 ); \
a0 = _mm256_xor_si256( a0, m256_const1_128( round_const_l0[i] ) );\
b1 = mm256_bcast_m128( mm128_mask_32( m128_neg1, 0x3 ) ); \
a0 = _mm256_xor_si256( a0, mm256_bcast_m128( round_const_l0[i] ) );\
a1 = _mm256_xor_si256( a1, b1 );\
a2 = _mm256_xor_si256( a2, b1 );\
a3 = _mm256_xor_si256( a3, b1 );\
a4 = _mm256_xor_si256( a4, b1 );\
a5 = _mm256_xor_si256( a5, b1 );\
a6 = _mm256_xor_si256( a6, b1 );\
a7 = _mm256_xor_si256( a7, m256_const1_128( round_const_l7[i] ) );\
a7 = _mm256_xor_si256( a7, mm256_bcast_m128( round_const_l7[i] ) );\
\
/* ShiftBytes + SubBytes (interleaved) */\
b0 = _mm256_xor_si256( b0, b0 );\
@@ -850,7 +752,7 @@ static const __m256i SUBSH_MASK7_2WAY =
}/**/
#define Matrix_Transpose_O_B_2way(i0, i1, i2, i3, i4, i5, i6, i7, t0){\
t0 = _mm256_xor_si256( t0, t0 );\
t0 = m256_zero;\
i1 = i0;\
i3 = i2;\
i5 = i4;\
@@ -874,11 +776,11 @@ static const __m256i SUBSH_MASK7_2WAY =
void TF512_2way( __m256i* chaining, __m256i* message )
{
static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m256i TEMP0;
static __m256i TEMP1;
static __m256i TEMP2;
__m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m256i TEMP0;
__m256i TEMP1;
__m256i TEMP2;
/* load message into registers xmm12 - xmm15 */
xmm12 = message[0];
@@ -940,11 +842,11 @@ void TF512_2way( __m256i* chaining, __m256i* message )
void OF512_2way( __m256i* chaining )
{
static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m256i TEMP0;
static __m256i TEMP1;
static __m256i TEMP2;
__m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m256i TEMP0;
__m256i TEMP1;
__m256i TEMP2;
/* load CV into registers xmm8, xmm10, xmm12, xmm14 */
xmm8 = chaining[0];

33
algo/groestl/groestl512-hash-4way.c
View File

@@ -25,8 +25,7 @@ int groestl512_4way_init( groestl512_4way_context* ctx, uint64_t hashlen )
memset_zero_512( ctx->buffer, SIZE512 );
// The only non-zero in the IV is len. It can be hard coded.
ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );
ctx->chaining[ 6 ] = mm512_bcast128hi_64( 0x0200000000000000 );
ctx->buf_ptr = 0;
ctx->rem_ptr = 0;
@@ -61,14 +60,14 @@ int groestl512_4way_update_close( groestl512_4way_context* ctx, void* output,
if ( i == SIZE512 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
}
else
{
ctx->buffer[i] = m512_const2_64( 0, 0x80 );
ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
for ( i += 1; i < SIZE512 - 1; i++ )
ctx->buffer[i] = m512_zero;
ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
}
TF1024_4way( ctx->chaining, ctx->buffer );
@@ -94,7 +93,7 @@ int groestl512_4way_full( groestl512_4way_context* ctx, void* output,
memset_zero_512( ctx->chaining, SIZE512 );
memset_zero_512( ctx->buffer, SIZE512 );
ctx->chaining[ 6 ] = m512_const2_64( 0x0200000000000000, 0 );
ctx->chaining[ 6 ] = mm512_bcast128hi_64( 0x0200000000000000 );
ctx->buf_ptr = 0;
// --- update ---
@@ -113,14 +112,14 @@ int groestl512_4way_full( groestl512_4way_context* ctx, void* output,
if ( i == SIZE512 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m512_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm512_set2_64( blocks << 56, 0x80 );
}
else
{
ctx->buffer[i] = m512_const2_64( 0, 0x80 );
ctx->buffer[i] = mm512_bcast128lo_64( 0x80 );
for ( i += 1; i < SIZE512 - 1; i++ )
ctx->buffer[i] = m512_zero;
ctx->buffer[i] = m512_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm512_bcast128hi_64( blocks << 56 );
}
TF1024_4way( ctx->chaining, ctx->buffer );
@@ -143,7 +142,7 @@ int groestl512_2way_init( groestl512_2way_context* ctx, uint64_t hashlen )
memset_zero_256( ctx->buffer, SIZE512 );
// The only non-zero in the IV is len. It can be hard coded.
ctx->chaining[ 6 ] = m256_const2_64( 0x0200000000000000, 0 );
ctx->chaining[ 6 ] = mm256_bcast128hi_64( 0x0200000000000000 );
ctx->buf_ptr = 0;
ctx->rem_ptr = 0;
@@ -179,14 +178,14 @@ int groestl512_2way_update_close( groestl512_2way_context* ctx, void* output,
if ( i == SIZE512 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
}
else
{
ctx->buffer[i] = m256_const2_64( 0, 0x80 );
ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
for ( i += 1; i < SIZE512 - 1; i++ )
ctx->buffer[i] = m256_zero;
ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
}
TF1024_2way( ctx->chaining, ctx->buffer );
@@ -212,7 +211,7 @@ int groestl512_2way_full( groestl512_2way_context* ctx, void* output,
memset_zero_256( ctx->chaining, SIZE512 );
memset_zero_256( ctx->buffer, SIZE512 );
ctx->chaining[ 6 ] = m256_const2_64( 0x0200000000000000, 0 );
ctx->chaining[ 6 ] = mm256_bcast128hi_64( 0x0200000000000000 );
ctx->buf_ptr = 0;
// --- update ---
@@ -231,14 +230,14 @@ int groestl512_2way_full( groestl512_2way_context* ctx, void* output,
if ( i == SIZE512 - 1 )
{
// only 1 vector left in buffer, all padding at once
ctx->buffer[i] = m256_const2_64( blocks << 56, 0x80 );
ctx->buffer[i] = mm256_set2_64( blocks << 56, 0x80 );
}
else
{
ctx->buffer[i] = m256_const2_64( 0, 0x80 );
ctx->buffer[i] = mm256_bcast128lo_64( 0x80 );
for ( i += 1; i < SIZE512 - 1; i++ )
ctx->buffer[i] = m256_zero;
ctx->buffer[i] = m256_const2_64( blocks << 56, 0 );
ctx->buffer[i] = mm256_bcast128hi_64( blocks << 56 );
}
TF1024_2way( ctx->chaining, ctx->buffer );

72
algo/groestl/groestl512-intr-4way.h
View File

@@ -174,7 +174,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
\
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
/* compute w_i : add y_{i+4} */\
b1 = m512_const1_64( 0x1b1b1b1b1b1b1b1b ); \
b1 = _mm512_set1_epi64( 0x1b1b1b1b1b1b1b1b ); \
MUL2( a0, b0, b1 ); \
a0 = _mm512_xor_si512( a0, TEMP0 ); \
MUL2( a1, b0, b1 ); \
@@ -238,7 +238,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
for ( round_counter = 0; round_counter < 14; round_counter += 2 ) \
{ \
/* AddRoundConstant P1024 */\
xmm8 = _mm512_xor_si512( xmm8, m512_const1_128( \
xmm8 = _mm512_xor_si512( xmm8, mm512_bcast_m128( \
casti_m128i( round_const_p, round_counter ) ) ); \
/* ShiftBytes P1024 + pre-AESENCLAST */\
xmm8 = _mm512_shuffle_epi8( xmm8, SUBSH_MASK0 ); \
@@ -253,7 +253,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
SUBMIX(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
\
/* AddRoundConstant P1024 */\
xmm0 = _mm512_xor_si512( xmm0, m512_const1_128( \
xmm0 = _mm512_xor_si512( xmm0, mm512_bcast_m128( \
casti_m128i( round_const_p, round_counter+1 ) ) ); \
/* ShiftBytes P1024 + pre-AESENCLAST */\
xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK0 );\
@@ -282,7 +282,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
xmm12 = _mm512_xor_si512( xmm12, xmm1 );\
xmm13 = _mm512_xor_si512( xmm13, xmm1 );\
xmm14 = _mm512_xor_si512( xmm14, xmm1 );\
xmm15 = _mm512_xor_si512( xmm15, m512_const1_128( \
xmm15 = _mm512_xor_si512( xmm15, mm512_bcast_m128( \
casti_m128i( round_const_q, round_counter ) ) ); \
/* ShiftBytes Q1024 + pre-AESENCLAST */\
xmm8 = _mm512_shuffle_epi8( xmm8, SUBSH_MASK1 );\
@@ -305,7 +305,7 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
xmm4 = _mm512_xor_si512( xmm4, xmm9 );\
xmm5 = _mm512_xor_si512( xmm5, xmm9 );\
xmm6 = _mm512_xor_si512( xmm6, xmm9 );\
xmm7 = _mm512_xor_si512( xmm7, m512_const1_128( \
xmm7 = _mm512_xor_si512( xmm7, mm512_bcast_m128( \
casti_m128i( round_const_q, round_counter+1 ) ) ); \
/* ShiftBytes Q1024 + pre-AESENCLAST */\
xmm0 = _mm512_shuffle_epi8( xmm0, SUBSH_MASK1 );\
@@ -471,8 +471,8 @@ static const __m512i SUBSH_MASK7 = { 0x06090c0f0205080b, 0x0e0104070a0d0003,
void INIT_4way( __m512i* chaining )
{
static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
/* load IV into registers xmm8 - xmm15 */
xmm8 = chaining[0];
@@ -500,12 +500,12 @@ void INIT_4way( __m512i* chaining )
void TF1024_4way( __m512i* chaining, const __m512i* message )
{
static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m512i QTEMP[8];
static __m512i TEMP0;
static __m512i TEMP1;
static __m512i TEMP2;
__m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m512i QTEMP[8];
__m512i TEMP0;
__m512i TEMP1;
__m512i TEMP2;
/* load message into registers xmm8 - xmm15 (Q = message) */
xmm8 = message[0];
@@ -606,11 +606,11 @@ void TF1024_4way( __m512i* chaining, const __m512i* message )
void OF1024_4way( __m512i* chaining )
{
static __m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m512i TEMP0;
static __m512i TEMP1;
static __m512i TEMP2;
__m512i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m512i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m512i TEMP0;
__m512i TEMP1;
__m512i TEMP2;
/* load CV into registers xmm8 - xmm15 */
xmm8 = chaining[0];
@@ -758,7 +758,7 @@ static const __m256i SUBSH_MASK7_2WAY =
\
/* compute z_i : double x_i using temp xmm8 and 1B xmm9 */\
/* compute w_i : add y_{i+4} */\
b1 = m256_const1_64( 0x1b1b1b1b1b1b1b1b );\
b1 = _mm256_set1_epi64x( 0x1b1b1b1b1b1b1b1b );\
MUL2_2WAY(a0, b0, b1);\
a0 = _mm256_xor_si256(a0, TEMP0);\
MUL2_2WAY(a1, b0, b1);\
@@ -822,7 +822,7 @@ static const __m256i SUBSH_MASK7_2WAY =
for ( round_counter = 0; round_counter < 14; round_counter += 2 ) \
{ \
/* AddRoundConstant P1024 */\
xmm8 = _mm256_xor_si256( xmm8, m256_const1_128( \
xmm8 = _mm256_xor_si256( xmm8, mm256_bcast_m128( \
casti_m128i( round_const_p, round_counter ) ) ); \
/* ShiftBytes P1024 + pre-AESENCLAST */\
xmm8 = _mm256_shuffle_epi8( xmm8, SUBSH_MASK0_2WAY ); \
@@ -837,7 +837,7 @@ static const __m256i SUBSH_MASK7_2WAY =
SUBMIX_2WAY(xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15, xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7);\
\
/* AddRoundConstant P1024 */\
xmm0 = _mm256_xor_si256( xmm0, m256_const1_128( \
xmm0 = _mm256_xor_si256( xmm0, mm256_bcast_m128( \
casti_m128i( round_const_p, round_counter+1 ) ) ); \
/* ShiftBytes P1024 + pre-AESENCLAST */\
xmm0 = _mm256_shuffle_epi8( xmm0, SUBSH_MASK0_2WAY );\
@@ -866,7 +866,7 @@ static const __m256i SUBSH_MASK7_2WAY =
xmm12 = _mm256_xor_si256( xmm12, xmm1 );\
xmm13 = _mm256_xor_si256( xmm13, xmm1 );\
xmm14 = _mm256_xor_si256( xmm14, xmm1 );\
xmm15 = _mm256_xor_si256( xmm15, m256_const1_128( \
xmm15 = _mm256_xor_si256( xmm15, mm256_bcast_m128( \
casti_m128i( round_const_q, round_counter ) ) ); \
/* ShiftBytes Q1024 + pre-AESENCLAST */\
xmm8 = _mm256_shuffle_epi8( xmm8, SUBSH_MASK1_2WAY );\
@@ -889,7 +889,7 @@ static const __m256i SUBSH_MASK7_2WAY =
xmm4 = _mm256_xor_si256( xmm4, xmm9 );\
xmm5 = _mm256_xor_si256( xmm5, xmm9 );\
xmm6 = _mm256_xor_si256( xmm6, xmm9 );\
xmm7 = _mm256_xor_si256( xmm7, m256_const1_128( \
xmm7 = _mm256_xor_si256( xmm7, mm256_bcast_m128( \
casti_m128i( round_const_q, round_counter+1 ) ) ); \
/* ShiftBytes Q1024 + pre-AESENCLAST */\
xmm0 = _mm256_shuffle_epi8( xmm0, SUBSH_MASK1_2WAY );\
@@ -1040,8 +1040,8 @@ static const __m256i SUBSH_MASK7_2WAY =
void INIT_2way( __m256i *chaining )
{
static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
/* load IV into registers xmm8 - xmm15 */
xmm8 = chaining[0];
@@ -1069,12 +1069,12 @@ void INIT_2way( __m256i *chaining )
void TF1024_2way( __m256i *chaining, const __m256i *message )
{
static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m256i QTEMP[8];
static __m256i TEMP0;
static __m256i TEMP1;
static __m256i TEMP2;
__m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m256i QTEMP[8];
__m256i TEMP0;
__m256i TEMP1;
__m256i TEMP2;
/* load message into registers xmm8 - xmm15 (Q = message) */
xmm8 = message[0];
@@ -1175,11 +1175,11 @@ void TF1024_2way( __m256i *chaining, const __m256i *message )
void OF1024_2way( __m256i* chaining )
{
static __m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
static __m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
static __m256i TEMP0;
static __m256i TEMP1;
static __m256i TEMP2;
__m256i xmm0, xmm1, xmm2, xmm3, xmm4, xmm5, xmm6, xmm7;
__m256i xmm8, xmm9, xmm10, xmm11, xmm12, xmm13, xmm14, xmm15;
__m256i TEMP0;
__m256i TEMP1;
__m256i TEMP2;
/* load CV into registers xmm8 - xmm15 */
xmm8 = chaining[0];