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

v3.10.5

Browse Source
This commit is contained in:
Jay D Dee
2019-12-21 13:19:29 -05:00
parent a17ff6f189
commit c65b0ff7a6
72 changed files with 9090 additions and 1336 deletions
Download Patch File Download Diff File Expand all files Collapse all files

178
algo/lyra2/sponge-2way.c
View File

@@ -19,7 +19,7 @@
* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*/
#include "algo-gate.h"
//#include "algo-gate.h"
#include <string.h>
#include <stdio.h>
#include <time.h>
@@ -40,19 +40,26 @@ inline void squeeze_2way( uint64_t *State, byte *Out, unsigned int len )
//Squeezes full blocks
for ( i = 0; i < fullBlocks; i++ )
{
memcpy_512( out, state, BLOCK_LEN_M256I*2 );
LYRA_ROUND_2WAY_AVX2( state[0], state[1], state[2], state[3] );
out += BLOCK_LEN_M256I*2;
memcpy_512( out, state, BLOCK_LEN_M256I );
LYRA_ROUND_2WAY_AVX512( state[0], state[1], state[2], state[3] );
out += BLOCK_LEN_M256I;
}
//Squeezes remaining bytes
memcpy_512( out, state, ( (len_m256i % BLOCK_LEN_M256I) * 2 ) );
memcpy_512( out, state, len_m256i % BLOCK_LEN_M256I );
}
inline void absorbBlock_2way( uint64_t *State, const uint64_t *In )
inline void absorbBlock_2way( uint64_t *State, const uint64_t *In0,
const uint64_t *In1 )
{
register __m512i state0, state1, state2, state3;
__m512i *in = (__m512i*)In;
__m512i in[3];
casti_m256i( in, 0 ) = casti_m256i( In0, 0 );
casti_m256i( in, 1 ) = casti_m256i( In1, 1 );
casti_m256i( in, 2 ) = casti_m256i( In0, 2 );
casti_m256i( in, 3 ) = casti_m256i( In1, 3 );
casti_m256i( in, 4 ) = casti_m256i( In0, 4 );
casti_m256i( in, 5 ) = casti_m256i( In1, 5 );
state0 = _mm512_load_si512( (__m512i*)State );
state1 = _mm512_load_si512( (__m512i*)State + 1 );
state2 = _mm512_load_si512( (__m512i*)State + 2 );
@@ -90,7 +97,7 @@ inline void absorbBlockBlake2Safe_2way( uint64_t *State, const uint64_t *In,
state1 = _mm512_xor_si512( state1, in[1] );
LYRA_12_ROUNDS_2WAY_AVX512( state0, state1, state2, state3 );
In += block_len * 2;
In += block_len*2;
}
_mm512_store_si512( (__m512i*)State, state0 );
@@ -109,7 +116,7 @@ inline void reducedSqueezeRow0_2way( uint64_t* State, uint64_t* rowOut,
register __m512i state0, state1, state2, state3;
__m512i* out = (__m512i*)rowOut + ( (nCols-1) * BLOCK_LEN_M256I * 2 );
__m512i* out = (__m512i*)rowOut + ( (nCols-1) * BLOCK_LEN_M256I );
state0 = _mm512_load_si512( (__m512i*)State );
state1 = _mm512_load_si512( (__m512i*)State + 1 );
@@ -126,13 +133,13 @@ inline void reducedSqueezeRow0_2way( uint64_t* State, uint64_t* rowOut,
{
_mm_prefetch( out - 9, _MM_HINT_T0 );
_mm_prefetch( out - 11, _MM_HINT_T0 );
out[0] = state0;
out[1] = state1;
out[2] = state2;
//Goes to next block (column) that will receive the squeezed data
out -= BLOCK_LEN_M256I * 2;
out -= BLOCK_LEN_M256I;
LYRA_ROUND_2WAY_AVX512( state0, state1, state2, state3 );
}
@@ -143,15 +150,14 @@ inline void reducedSqueezeRow0_2way( uint64_t* State, uint64_t* rowOut,
_mm512_store_si512( (__m512i*)State + 3, state3 );
}
// This function has to deal with gathering 2 256 bit rowin vectors from
// non-contiguous memory. Extra work and performance penalty.
inline void reducedDuplexRow1_2way( uint64_t *State, uint64_t *rowIn,
uint64_t *rowOut, uint64_t nCols )
{
int i;
register __m512i state0, state1, state2, state3;
__m512i *in = (__m256i*)rowIn;
__m512i *in = (__m512i*)rowIn;
__m512i *out = (__m512i*)rowOut + ( (nCols-1) * BLOCK_LEN_M256I );
state0 = _mm512_load_si512( (__m512i*)State );
state1 = _mm512_load_si512( (__m512i*)State + 1 );
@@ -171,28 +177,25 @@ inline void reducedDuplexRow1_2way( uint64_t *State, uint64_t *rowIn,
out[2] = _mm512_xor_si512( state2, in[2] );
//Input: next column (i.e., next block in sequence)
in0 += BLOCK_LEN_M256I;
in1 += BLOCK_LEN_M256I;
in += BLOCK_LEN_M256I;
//Output: goes to previous column
out -= BLOCK_LEN_M256I * 2;
out -= BLOCK_LEN_M256I;
}
_mm512_store_si256( (__m512i*)State, state0 );
_mm512_store_si256( (__m512i*)State + 1, state1 );
_mm512_store_si256( (__m512i*)State + 2, state2 );
_mm512_store_si256( (__m512i*)State + 3, state3 );
}
_mm512_store_si512( (__m512i*)State, state0 );
_mm512_store_si512( (__m512i*)State + 1, state1 );
_mm512_store_si512( (__m512i*)State + 2, state2 );
_mm512_store_si512( (__m512i*)State + 3, state3 );
}
inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
uint64_t *rowInOut, uint64_t *rowOut, uint64_t nCols )
{
int i;
register __m512i state0, state1, state2, state3;
__m512i* in = (__m512i*)rowIn;
__m512i* inout = (__m512i*)rowInOut;
__m512i* out = (__m512i*)rowOut + ( (nCols-1) * BLOCK_LEN_M256I * 2 );
__m512i* out = (__m512i*)rowOut + ( (nCols-1) * BLOCK_LEN_M256I );
__m512i t0, t1, t2;
state0 = _mm512_load_si512( (__m512i*)State );
@@ -209,7 +212,7 @@ inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
state2 = _mm512_xor_si512( state2,
_mm512_add_epi64( in[2], inout[2] ) );
LYRA_ROUND_2WAY AVX512( state0, state1, state2, state3 );
LYRA_ROUND_2WAY_AVX512( state0, state1, state2, state3 );
out[0] = _mm512_xor_si512( state0, in[0] );
out[1] = _mm512_xor_si512( state1, in[1] );
@@ -221,17 +224,18 @@ inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
t2 = _mm512_permutex_epi64( state2, 0x93 );
inout[0] = _mm512_xor_si512( inout[0],
_mm512_mask_blend_epi32( t0, t2, 0x03 ) );
_mm512_mask_blend_epi32( 0x0303, t0, t2 ) );
inout[1] = _mm512_xor_si512( inout[1],
_mm512_mask_blend_epi32( t1, t0, 0x03 ) );
_mm512_mask_blend_epi32( 0x0303, t1, t0 ) );
inout[2] = _mm512_xor_si512( inout[2],
_mm512_mask_blend_epi32( t2, t1, 0x03 ) );
_mm512_mask_blend_epi32( 0x0303, t2, t1 ) );
//Inputs: next column (i.e., next block in sequence)
in += BLOCK_LEN_M256I * 2;
inout += BLOCK_LEN_M256I * 2;
in += BLOCK_LEN_M256I;
inout += BLOCK_LEN_M256I;
//Output: goes to previous column
out -= BLOCK_LEN_M256I * 2;
out -= BLOCK_LEN_M256I;
}
_mm512_store_si512( (__m512i*)State, state0 );
@@ -240,49 +244,61 @@ inline void reducedDuplexRowSetup_2way( uint64_t *State, uint64_t *rowIn,
_mm512_store_si512( (__m512i*)State + 3, state3 );
}
inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn1,
uint64_t *rowIn0, uint64_t *rowInOut, uint64_t *rowOut,
uint64_t nCols )
// big ugly workaound for pointer aliasing, use a union of pointers.
// Access matrix using m512i for in and out, m256i for inout
inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn,
uint64_t *rowInOut0, uint64_t *rowInOut1,
uint64_t *rowOut, uint64_t nCols)
{
int i;
register __m512i state0, state1, state2, state3;
__m256i *in0 = (__m256i*)rowIn0;
__m256i *in0 = (__m256i*)rowIn0;
__m2512* in = (__m512i*)rowIn;
__m2512* inout = (__m512i*)rowInOut;
__m512i* out = (__m512i*)rowOut;
__m512i t0, t1, t2;
__m512i *in = (__m512i*)rowIn;
__m256i *inout0 = (__m256i*)rowInOut0;
__m256i *inout1 = (__m256i*)rowInOut1;
__m512i *out = (__m512i*)rowOut;
__m512i io[3];
povly inout;
inout.v512 = &io[0];
__m512i t0, t1, t2;
_mm_prefetch( in0, _MM_HINT_T0 );
_mm_prefetch( in1, _MM_HINT_T0 );
_mm_prefetch( in0 + 2, _MM_HINT_T0 );
_mm_prefetch( in1 + 2, _MM_HINT_T0 );
_mm_prefetch( in0 + 4, _MM_HINT_T0 );
_mm_prefetch( in1 + 4, _MM_HINT_T0 );
_mm_prefetch( in0 + 6, _MM_HINT_T0 );
_mm_prefetch( in1 + 6, _MM_HINT_T0 );
state0 = _mm512_load_si512( (__m512i*)State );
state1 = _mm512_load_si512( (__m512i*)State + 1 );
state2 = _mm512_load_si512( (__m512i*)State + 2 );
state3 = _mm512_load_si512( (__m512i*)State + 3 );
_mm_prefetch( in, _MM_HINT_T0 );
_mm_prefetch( inout0, _MM_HINT_T0 );
_mm_prefetch( inout1, _MM_HINT_T0 );
_mm_prefetch( in + 2, _MM_HINT_T0 );
_mm_prefetch( inout0 + 2, _MM_HINT_T0 );
_mm_prefetch( inout1 + 2, _MM_HINT_T0 );
_mm_prefetch( in + 4, _MM_HINT_T0 );
_mm_prefetch( inout0 + 4, _MM_HINT_T0 );
_mm_prefetch( inout1 + 4, _MM_HINT_T0 );
_mm_prefetch( in + 6, _MM_HINT_T0 );
_mm_prefetch( inout0 + 6, _MM_HINT_T0 );
_mm_prefetch( inout1 + 6, _MM_HINT_T0 );
for ( i = 0; i < nCols; i++ )
{
//Absorbing "M[prev] [+] M[row*]"
inout.v256[0] = inout0[0];
inout.v256[1] = inout1[1];
inout.v256[2] = inout0[2];
inout.v256[3] = inout1[3];
inout.v256[4] = inout0[4];
inout.v256[5] = inout1[5];
// state0 = _mm512_xor_si512( state0, mm512_concat_256( in1[0], in0[0] );
// state1 = _mm512_xor_si512( state1, mm512_concat_256( in1[1], in0[1] );
// state2 = _mm512_xor_si512( state2, mm512_concat_256( in1[2], in0[2] );
t0 = mm512_concat_256( in1[0], in0[0] );
t1 = mm512_concat_256( in1[1], in0[1] );
t2 = mm512_concat_256( in1[2], in0[2] );
state0 = _mm512_xor_si512( state0,
_mm512_add_epi64( t0, inout[0] ) );
_mm512_add_epi64( in[0], inout.v512[0] ) );
state1 = _mm512_xor_si512( state1,
_mm512_add_epi64( t1, inout[1] ) );
_mm512_add_epi64( in[1], inout.v512[1] ) );
state2 = _mm512_xor_si512( state2,
_mm512_add_epi64( t2, inout[2] ) );
_mm512_add_epi64( in[2], inout.v512[2] ) );
//Applies the reduced-round transformation f to the sponge's state
LYRA_ROUND_2WAY_AVX512( state0, state1, state2, state3 );
@@ -292,22 +308,44 @@ inline void reducedDuplexRow_2way( uint64_t *State, uint64_t *rowIn1,
out[1] = _mm512_xor_si512( out[1], state1 );
out[2] = _mm512_xor_si512( out[2], state2 );
// if inout is the same row as out it was just overwritten, reload.
if ( rowOut == rowInOut0 )
{
inout.v256[0] = inout0[0];
inout.v256[2] = inout0[2];
inout.v256[4] = inout0[4];
}
if ( rowOut == rowInOut1 )
{
inout.v256[1] = inout1[1];
inout.v256[3] = inout1[3];
inout.v256[5] = inout1[5];
}
//M[rowInOut][col] = M[rowInOut][col] XOR rotW(rand)
t0 = _mm512_permutex_epi64( state0, 0x93 );
t1 = _mm512_permutex_epi64( state1, 0x93 );
t2 = _mm512_permutex_epi64( state2, 0x93 );
inout[0] = _mm512_xor_si512( inout[0],
_mm512_mask_blend_epi32( t0, t2, 0x03 ) );
inout[1] = _mm512_xor_si512( inout[1],
_mm512_mask_blend_epi32( t1, t0, 0x03 ) );
inout[2] = _mm512_xor_si512( inout[2],
_mm512_mask_blend_epi32( t2, t1, 0x03 ) );
inout.v512[0] = _mm512_xor_si512( inout.v512[0],
_mm512_mask_blend_epi32( 0x0303, t0, t2 ) );
inout.v512[1] = _mm512_xor_si512( inout.v512[1],
_mm512_mask_blend_epi32( 0x0303, t1, t0 ) );
inout.v512[2] = _mm512_xor_si512( inout.v512[2],
_mm512_mask_blend_epi32( 0x0303, t2, t1 ) );
inout0[0] = inout.v256[0];
inout1[1] = inout.v256[1];
inout0[2] = inout.v256[2];
inout1[3] = inout.v256[3];
inout0[4] = inout.v256[4];
inout1[5] = inout.v256[5];
//Goes to next block
in += BLOCK_LEN_M256I * 2;
out += BLOCK_LEN_M256I * 2;
inout += BLOCK_LEN_M256I * 2;
in += BLOCK_LEN_M256I;
inout0 += BLOCK_LEN_M256I * 2;
inout1 += BLOCK_LEN_M256I * 2;
out += BLOCK_LEN_M256I;
}
_mm512_store_si512( (__m512i*)State, state0 );