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

v3.9.0

Browse Source
This commit is contained in:
Jay D Dee
2019-05-19 13:39:45 -04:00
parent bfd1c002f9
commit e1aead3c76
139 changed files with 10907 additions and 4218 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
algo/blake/blake-4way.c
View File

@@ -15,7 +15,7 @@ void blakehash_4way(void *state, const void *input)
memcpy( &ctx, &blake_4w_ctx, sizeof ctx );
blake256r14_4way( &ctx, input + (64<<2), 16 );
blake256r14_4way_close( &ctx, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -37,7 +37,7 @@ int scanhash_blake_4way( int thr_id, struct work *work, uint32_t max_nonce,
// we need big endian data...
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake256r14_4way_init( &blake_4w_ctx );
blake256r14_4way( &blake_4w_ctx, vdata, 64 );

82
algo/blake/blake-hash-4way.c
View File

@@ -363,14 +363,14 @@ static const sph_u64 CB[16] = {
do { \
a = _mm_add_epi32( _mm_add_epi32( _mm_xor_si128( \
_mm_set_epi32( c1, c1, c1, c1 ), m0 ), b ), a ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 16 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 12 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 12 ); \
a = _mm_add_epi32( _mm_add_epi32( _mm_xor_si128( \
_mm_set_epi32( c0, c0, c0, c0 ), m1 ), b ), a ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 8 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 8 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 7 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 7 ); \
} while (0)
#if SPH_COMPACT_BLAKE_32
@@ -562,22 +562,22 @@ do { \
, _mm_set_epi32( CS6, CS6, CS6, CS6 ) ); \
VF = _mm_xor_si128( _mm_set_epi32( T1, T1, T1, T1 ), \
_mm_set_epi32( CS7, CS7, CS7, CS7 ) ); \
M[0x0] = mm_bswap_32( *(buf + 0) ); \
M[0x1] = mm_bswap_32( *(buf + 1) ); \
M[0x2] = mm_bswap_32( *(buf + 2) ); \
M[0x3] = mm_bswap_32( *(buf + 3) ); \
M[0x4] = mm_bswap_32( *(buf + 4) ); \
M[0x5] = mm_bswap_32( *(buf + 5) ); \
M[0x6] = mm_bswap_32( *(buf + 6) ); \
M[0x7] = mm_bswap_32( *(buf + 7) ); \
M[0x8] = mm_bswap_32( *(buf + 8) ); \
M[0x9] = mm_bswap_32( *(buf + 9) ); \
M[0xA] = mm_bswap_32( *(buf + 10) ); \
M[0xB] = mm_bswap_32( *(buf + 11) ); \
M[0xC] = mm_bswap_32( *(buf + 12) ); \
M[0xD] = mm_bswap_32( *(buf + 13) ); \
M[0xE] = mm_bswap_32( *(buf + 14) ); \
M[0xF] = mm_bswap_32( *(buf + 15) ); \
M[0x0] = mm128_bswap_32( *(buf + 0) ); \
M[0x1] = mm128_bswap_32( *(buf + 1) ); \
M[0x2] = mm128_bswap_32( *(buf + 2) ); \
M[0x3] = mm128_bswap_32( *(buf + 3) ); \
M[0x4] = mm128_bswap_32( *(buf + 4) ); \
M[0x5] = mm128_bswap_32( *(buf + 5) ); \
M[0x6] = mm128_bswap_32( *(buf + 6) ); \
M[0x7] = mm128_bswap_32( *(buf + 7) ); \
M[0x8] = mm128_bswap_32( *(buf + 8) ); \
M[0x9] = mm128_bswap_32( *(buf + 9) ); \
M[0xA] = mm128_bswap_32( *(buf + 10) ); \
M[0xB] = mm128_bswap_32( *(buf + 11) ); \
M[0xC] = mm128_bswap_32( *(buf + 12) ); \
M[0xD] = mm128_bswap_32( *(buf + 13) ); \
M[0xE] = mm128_bswap_32( *(buf + 14) ); \
M[0xF] = mm128_bswap_32( *(buf + 15) ); \
for (r = 0; r < rounds; r ++) \
ROUND_S_4WAY(r); \
H0 = _mm_xor_si128( _mm_xor_si128( \
@@ -624,22 +624,22 @@ do { \
VD = _mm_xor_si128( _mm_set1_epi32( T0 ), _mm_set1_epi32( CS5 ) ); \
VE = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS6 ) ); \
VF = _mm_xor_si128( _mm_set1_epi32( T1 ), _mm_set1_epi32( CS7 ) ); \
M0 = mm_bswap_32( * buf ); \
M1 = mm_bswap_32( *(buf+1) ); \
M2 = mm_bswap_32( *(buf+2) ); \
M3 = mm_bswap_32( *(buf+3) ); \
M4 = mm_bswap_32( *(buf+4) ); \
M5 = mm_bswap_32( *(buf+5) ); \
M6 = mm_bswap_32( *(buf+6) ); \
M7 = mm_bswap_32( *(buf+7) ); \
M8 = mm_bswap_32( *(buf+8) ); \
M9 = mm_bswap_32( *(buf+9) ); \
MA = mm_bswap_32( *(buf+10) ); \
MB = mm_bswap_32( *(buf+11) ); \
MC = mm_bswap_32( *(buf+12) ); \
MD = mm_bswap_32( *(buf+13) ); \
ME = mm_bswap_32( *(buf+14) ); \
MF = mm_bswap_32( *(buf+15) ); \
M0 = mm128_bswap_32( * buf ); \
M1 = mm128_bswap_32( *(buf+1) ); \
M2 = mm128_bswap_32( *(buf+2) ); \
M3 = mm128_bswap_32( *(buf+3) ); \
M4 = mm128_bswap_32( *(buf+4) ); \
M5 = mm128_bswap_32( *(buf+5) ); \
M6 = mm128_bswap_32( *(buf+6) ); \
M7 = mm128_bswap_32( *(buf+7) ); \
M8 = mm128_bswap_32( *(buf+8) ); \
M9 = mm128_bswap_32( *(buf+9) ); \
MA = mm128_bswap_32( *(buf+10) ); \
MB = mm128_bswap_32( *(buf+11) ); \
MC = mm128_bswap_32( *(buf+12) ); \
MD = mm128_bswap_32( *(buf+13) ); \
ME = mm128_bswap_32( *(buf+14) ); \
MF = mm128_bswap_32( *(buf+15) ); \
ROUND_S_4WAY(0); \
ROUND_S_4WAY(1); \
ROUND_S_4WAY(2); \
@@ -1073,8 +1073,8 @@ blake32_4way_close( blake_4way_small_context *sc, unsigned ub, unsigned n,
if (out_size_w32 == 8)
buf[52>>2] = _mm_or_si128( buf[52>>2],
_mm_set1_epi32( 0x01000000UL ) );
*(buf+(56>>2)) = mm_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm_bswap_32( _mm_set1_epi32( tl ) );
*(buf+(56>>2)) = mm128_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( sc, buf + (ptr>>2), 64 - ptr );
}
else
@@ -1086,13 +1086,13 @@ blake32_4way_close( blake_4way_small_context *sc, unsigned ub, unsigned n,
memset_zero_128( buf, 56>>2 );
if (out_size_w32 == 8)
buf[52>>2] = _mm_set1_epi32( 0x01000000UL );
*(buf+(56>>2)) = mm_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm_bswap_32( _mm_set1_epi32( tl ) );
*(buf+(56>>2)) = mm128_bswap_32( _mm_set1_epi32( th ) );
*(buf+(60>>2)) = mm128_bswap_32( _mm_set1_epi32( tl ) );
blake32_4way( sc, buf, 64 );
}
out = (__m128i*)dst;
for ( k = 0; k < out_size_w32; k++ )
out[k] = mm_bswap_32( sc->H[k] );
out[k] = mm128_bswap_32( sc->H[k] );
}
#if defined (__AVX2__)

5
algo/blake/blake2s-4way.c
View File

@@ -85,7 +85,8 @@ void blake2s_4way_hash( void *output, const void *input )
blake2s_4way_update( &ctx, input + (64<<2), 16 );
blake2s_4way_final( &ctx, vhash, BLAKE2S_OUTBYTES );
mm_deinterleave_4x32( output, output+32, output+64, output+96, vhash, 256 );
mm128_deinterleave_4x32( output, output+32, output+64, output+96,
vhash, 256 );
}
int scanhash_blake2s_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -104,7 +105,7 @@ int scanhash_blake2s_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint32_t *noncep = vdata + 76; // 19*4
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake2s_4way_init( &blake2s_4w_ctx, BLAKE2S_OUTBYTES );
blake2s_4way_update( &blake2s_4w_ctx, vdata, 64 );

8
algo/blake/blake2s-hash-4way.c
View File

@@ -92,13 +92,13 @@ int blake2s_4way_compress( blake2s_4way_state *S, const __m128i* block )
#define G4W(r,i,a,b,c,d) \
do { \
a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ blake2s_sigma[r][2*i+0] ] ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 16 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 12 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 12 ); \
a = _mm_add_epi32( _mm_add_epi32( a, b ), m[ blake2s_sigma[r][2*i+1] ] ); \
d = mm_ror_32( _mm_xor_si128( d, a ), 8 ); \
d = mm128_ror_32( _mm_xor_si128( d, a ), 8 ); \
c = _mm_add_epi32( c, d ); \
b = mm_ror_32( _mm_xor_si128( b, c ), 7 ); \
b = mm128_ror_32( _mm_xor_si128( b, c ), 7 ); \
} while(0)
#define ROUND4W(r) \

4
algo/blake/blakecoin-4way.c
View File

@@ -17,7 +17,7 @@ void blakecoin_4way_hash(void *state, const void *input)
blake256r8_4way( &ctx, input + (64<<2), 16 );
blake256r8_4way_close( &ctx, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_blakecoin_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -37,7 +37,7 @@ int scanhash_blakecoin_4way( int thr_id, struct work *work, uint32_t max_nonce,
HTarget = 0x7f;
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake256r8_4way_init( &blakecoin_4w_ctx );
blake256r8_4way( &blakecoin_4w_ctx, vdata, 64 );

4
algo/blake/decred-4way.c
View File

@@ -23,7 +23,7 @@ void decred_hash_4way( void *state, const void *input )
memcpy( &ctx, &blake_mid, sizeof(blake_mid) );
blake256_4way( &ctx, tail, tail_len );
blake256_4way_close( &ctx, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_decred_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -44,7 +44,7 @@ int scanhash_decred_4way( int thr_id, struct work *work, uint32_t max_nonce,
memcpy( edata, pdata, 180 );
// use the old way until new way updated for size.
mm_interleave_4x32x( vdata, edata, edata, edata, edata, 180*8 );
mm128_interleave_4x32x( vdata, edata, edata, edata, edata, 180*8 );
blake256_4way_init( &blake_mid );
blake256_4way( &blake_mid, vdata, DECRED_MIDSTATE_LEN );

5
algo/blake/decred-gate.c
View File

@@ -140,6 +140,7 @@ bool decred_ready_to_mine( struct work* work, struct stratum_ctx* stratum,
return true;
}
int decred_get_work_data_size() { return DECRED_DATA_SIZE; }
bool register_decred_algo( algo_gate_t* gate )
{
@@ -154,7 +155,7 @@ bool register_decred_algo( algo_gate_t* gate )
gate->optimizations = AVX2_OPT;
gate->get_nonceptr = (void*)&decred_get_nonceptr;
gate->get_max64 = (void*)&get_max64_0x3fffffLL;
gate->display_extra_data = (void*)&decred_decode_extradata;
gate->decode_extra_data = (void*)&decred_decode_extradata;
gate->build_stratum_request = (void*)&decred_be_build_stratum_request;
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;
@@ -163,7 +164,7 @@ bool register_decred_algo( algo_gate_t* gate )
gate->nbits_index = DECRED_NBITS_INDEX;
gate->ntime_index = DECRED_NTIME_INDEX;
gate->nonce_index = DECRED_NONCE_INDEX;
gate->work_data_size = DECRED_DATA_SIZE;
gate->get_work_data_size = (void*)&decred_get_work_data_size;
gate->work_cmp_size = DECRED_WORK_COMPARE_SIZE;
allow_mininginfo = false;
have_gbt = false;

2
algo/blake/decred.c
View File

@@ -268,7 +268,7 @@ bool register_decred_algo( algo_gate_t* gate )
gate->hash = (void*)&decred_hash;
gate->get_nonceptr = (void*)&decred_get_nonceptr;
gate->get_max64 = (void*)&get_max64_0x3fffffLL;
gate->display_extra_data = (void*)&decred_decode_extradata;
gate->decode_extra_data = (void*)&decred_decode_extradata;
gate->build_stratum_request = (void*)&decred_be_build_stratum_request;
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;

48
algo/bmw/bmw-hash-4way.c
View File

@@ -77,26 +77,26 @@ static const sph_u64 IV512[] = {
#define ss0(x) \
_mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 1), \
_mm_slli_epi32( (x), 3) ), \
_mm_xor_si128( mm_rol_32( (x), 4), \
mm_rol_32( (x), 19) ) )
_mm_xor_si128( mm128_rol_32( (x), 4), \
mm128_rol_32( (x), 19) ) )
#define ss1(x) \
_mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 1), \
_mm_slli_epi32( (x), 2) ), \
_mm_xor_si128( mm_rol_32( (x), 8), \
mm_rol_32( (x), 23) ) )
_mm_xor_si128( mm128_rol_32( (x), 8), \
mm128_rol_32( (x), 23) ) )
#define ss2(x) \
_mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 2), \
_mm_slli_epi32( (x), 1) ), \
_mm_xor_si128( mm_rol_32( (x), 12), \
mm_rol_32( (x), 25) ) )
_mm_xor_si128( mm128_rol_32( (x), 12), \
mm128_rol_32( (x), 25) ) )
#define ss3(x) \
_mm_xor_si128( _mm_xor_si128( _mm_srli_epi32( (x), 2), \
_mm_slli_epi32( (x), 2) ), \
_mm_xor_si128( mm_rol_32( (x), 15), \
mm_rol_32( (x), 29) ) )
_mm_xor_si128( mm128_rol_32( (x), 15), \
mm128_rol_32( (x), 29) ) )
#define ss4(x) \
_mm_xor_si128( (x), _mm_srli_epi32( (x), 1 ) )
@@ -104,16 +104,16 @@ static const sph_u64 IV512[] = {
#define ss5(x) \
_mm_xor_si128( (x), _mm_srli_epi32( (x), 2 ) )
#define rs1(x) mm_rol_32( x, 3 )
#define rs2(x) mm_rol_32( x, 7 )
#define rs3(x) mm_rol_32( x, 13 )
#define rs4(x) mm_rol_32( x, 16 )
#define rs5(x) mm_rol_32( x, 19 )
#define rs6(x) mm_rol_32( x, 23 )
#define rs7(x) mm_rol_32( x, 27 )
#define rs1(x) mm128_rol_32( x, 3 )
#define rs2(x) mm128_rol_32( x, 7 )
#define rs3(x) mm128_rol_32( x, 13 )
#define rs4(x) mm128_rol_32( x, 16 )
#define rs5(x) mm128_rol_32( x, 19 )
#define rs6(x) mm128_rol_32( x, 23 )
#define rs7(x) mm128_rol_32( x, 27 )
#define rol_off_32( M, j, off ) \
mm_rol_32( M[ ( (j) + (off) ) & 0xF ] , \
mm128_rol_32( M[ ( (j) + (off) ) & 0xF ] , \
( ( (j) + (off) ) & 0xF ) + 1 )
#define add_elt_s( M, H, j ) \
@@ -526,42 +526,42 @@ void compress_small( const __m128i *M, const __m128i H[16], __m128i dH[16] )
_mm_slli_epi32( qt[23], 2 ) ) ),
_mm_xor_si128( _mm_xor_si128( xl, qt[31] ), qt[ 7] ));
dH[ 8] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[4], 9 ),
mm128_rol_32( dH[4], 9 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[24] ), M[ 8] )),
_mm_xor_si128( _mm_slli_epi32( xl, 8 ),
_mm_xor_si128( qt[23], qt[ 8] ) ) );
dH[ 9] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[5], 10 ),
mm128_rol_32( dH[5], 10 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[25] ), M[ 9] )),
_mm_xor_si128( _mm_srli_epi32( xl, 6 ),
_mm_xor_si128( qt[16], qt[ 9] ) ) );
dH[10] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[6], 11 ),
mm128_rol_32( dH[6], 11 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[26] ), M[10] )),
_mm_xor_si128( _mm_slli_epi32( xl, 6 ),
_mm_xor_si128( qt[17], qt[10] ) ) );
dH[11] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[7], 12 ),
mm128_rol_32( dH[7], 12 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[27] ), M[11] )),
_mm_xor_si128( _mm_slli_epi32( xl, 4 ),
_mm_xor_si128( qt[18], qt[11] ) ) );
dH[12] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[0], 13 ),
mm128_rol_32( dH[0], 13 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[28] ), M[12] )),
_mm_xor_si128( _mm_srli_epi32( xl, 3 ),
_mm_xor_si128( qt[19], qt[12] ) ) );
dH[13] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[1], 14 ),
mm128_rol_32( dH[1], 14 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[29] ), M[13] )),
_mm_xor_si128( _mm_srli_epi32( xl, 4 ),
_mm_xor_si128( qt[20], qt[13] ) ) );
dH[14] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[2], 15 ),
mm128_rol_32( dH[2], 15 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[30] ), M[14] )),
_mm_xor_si128( _mm_srli_epi32( xl, 7 ),
_mm_xor_si128( qt[21], qt[14] ) ) );
dH[15] = _mm_add_epi32( _mm_add_epi32(
mm_rol_32( dH[3], 16 ),
mm128_rol_32( dH[3], 16 ),
_mm_xor_si128( _mm_xor_si128( xh, qt[31] ), M[15] )),
_mm_xor_si128( _mm_srli_epi32( xl, 2 ),
_mm_xor_si128( qt[22], qt[15] ) ) );

28
algo/cubehash/cube-hash-2way.c
View File

@@ -92,7 +92,6 @@ int cube_2way_reinit( cube_2way_context *sp )
{
memcpy( sp, &cube_2way_ctx_cache, sizeof(cube_2way_context) );
return 0;
}
int cube_2way_init( cube_2way_context *sp, int hashbitlen, int rounds,
@@ -123,7 +122,7 @@ int cube_2way_init( cube_2way_context *sp, int hashbitlen, int rounds,
int cube_2way_update( cube_2way_context *sp, const void *data, size_t size )
{
const int len = size / 16;
const int len = size >> 4;
const __m256i *in = (__m256i*)data;
int i;
@@ -140,7 +139,6 @@ int cube_2way_update( cube_2way_context *sp, const void *data, size_t size )
sp->pos = 0;
}
}
return 0;
}
@@ -151,25 +149,22 @@ int cube_2way_close( cube_2way_context *sp, void *output )
// pos is zero for 64 byte data, 1 for 80 byte data.
sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ],
_mm256_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80,
0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80 ) );
_mm256_set_epi32( 0,0,0,0x80, 0,0,0,0x80 ) );
transform_2way( sp );
sp->h[7] = _mm256_xor_si256( sp->h[7], _mm256_set_epi32( 1,0,0,0,
1,0,0,0 ) );
for ( i = 0; i < 10; ++i )
transform_2way( &cube_2way_ctx_cache );
sp->h[7] = _mm256_xor_si256( sp->h[7],
_mm256_set_epi32( 1,0,0,0, 1,0,0,0 ) );
for ( i = 0; i < sp->hashlen; i++ )
hash[i] = sp->h[i];
for ( i = 0; i < 10; ++i ) transform_2way( sp );
for ( i = 0; i < sp->hashlen; i++ ) hash[i] = sp->h[i];
return 0;
}
int cube_2way_update_close( cube_2way_context *sp, void *output,
const void *data, size_t size )
{
const int len = size / 16;
const int len = size >> 4;
const __m256i *in = (__m256i*)data;
__m256i *hash = (__m256i*)output;
int i;
@@ -187,18 +182,15 @@ int cube_2way_update_close( cube_2way_context *sp, void *output,
// pos is zero for 64 byte data, 1 for 80 byte data.
sp->h[ sp->pos ] = _mm256_xor_si256( sp->h[ sp->pos ],
_mm256_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80,
0,0,0,0, 0,0,0,0, 0,0,0,0, 0,0,0,0x80 ) );
_mm256_set_epi32( 0,0,0,0x80, 0,0,0,0x80 ) );
transform_2way( sp );
sp->h[7] = _mm256_xor_si256( sp->h[7], _mm256_set_epi32( 1,0,0,0,
1,0,0,0 ) );
for ( i = 0; i < 10; ++i )
transform_2way( &cube_2way_ctx_cache );
for ( i = 0; i < sp->hashlen; i++ )
hash[i] = sp->h[i];
for ( i = 0; i < 10; ++i ) transform_2way( sp );
for ( i = 0; i < sp->hashlen; i++ ) hash[i] = sp->h[i];
return 0;
}

4
algo/cubehash/cube-hash-2way.h
View File

@@ -10,12 +10,12 @@
struct _cube_2way_context
{
__m256i h[8];
int hashlen; // __m128i
int rounds;
int blocksize; // __m128i
int pos; // number of __m128i read into x from current block
__m256i h[8] __attribute__ ((aligned (64)));
};
} __attribute__ ((aligned (64)));
typedef struct _cube_2way_context cube_2way_context;

1
algo/cubehash/sse2/cubehash_sse2.c → algo/cubehash/cubehash_sse2.c
View File

@@ -254,6 +254,7 @@ int cubehashUpdateDigest( cubehashParam *sp, byte *digest,
transform( sp );
sp->x[7] = _mm_xor_si128( sp->x[7], _mm_set_epi32( 1,0,0,0 ) );
transform( sp );
transform( sp );
transform( sp );

0
algo/cubehash/sse2/cubehash_sse2.h → algo/cubehash/cubehash_sse2.h
View File

456
algo/echo/aes_ni/hash.c
View File

@@ -60,336 +60,174 @@ MYALIGN const unsigned int zero[] = {0x00000000, 0x00000000, 0x00000000, 0x000
MYALIGN const unsigned int mul2ipt[] = {0x728efc00, 0x6894e61a, 0x3fc3b14d, 0x25d9ab57, 0xfd5ba600, 0x2a8c71d7, 0x1eb845e3, 0xc96f9234};
//#include "crypto_hash.h"
int crypto_hash(
unsigned char *out,
const unsigned char *in,
unsigned long long inlen
)
{
if(hash_echo(512, in, inlen * 8, out) == SUCCESS)
return 0;
return -1;
}
/*
int main()
{
return 0;
}
*/
#if 0
void DumpState(__m128i *ps)
{
int i, j, k;
unsigned int ucol;
for(j = 0; j < 4; j++)
{
for(i = 0; i < 4; i++)
{
printf("row %d,col %d : ", i, j);
for(k = 0; k < 4; k++)
{
ucol = *((int*)ps + 16 * i + 4 * j + k);
printf("%02x%02x%02x%02x ", (ucol >> 0) & 0xff, (ucol >> 8) & 0xff, (ucol >> 16) & 0xff, (ucol >> 24) & 0xff);
}
printf("\n");
}
}
printf("\n");
}
#endif
#ifndef NO_AES_NI
#define ECHO_SUBBYTES(state, i, j) \
state[i][j] = _mm_aesenc_si128(state[i][j], k1);\
state[i][j] = _mm_aesenc_si128(state[i][j], M128(zero));\
k1 = _mm_add_epi32(k1, M128(const1))
#else
#define ECHO_SUBBYTES(state, i, j) \
AES_ROUND_VPERM(state[i][j], t1, t2, t3, t4, s1, s2, s3);\
state[i][j] = _mm_xor_si128(state[i][j], k1);\
AES_ROUND_VPERM(state[i][j], t1, t2, t3, t4, s1, s2, s3);\
k1 = _mm_add_epi32(k1, M128(const1))
#define ECHO_SUB_AND_MIX(state, i, j, state2, c, r1, r2, r3, r4) \
AES_ROUND_VPERM_CORE(state[i][j], t1, t2, t3, t4, s1, s2, s3);\
ktemp = k1;\
TRANSFORM(ktemp, _k_ipt, t1, t4);\
state[i][j] = _mm_xor_si128(state[i][j], ktemp);\
AES_ROUND_VPERM_CORE(state[i][j], t1, t2, t3, t4, s1, s2, s3);\
k1 = _mm_add_epi32(k1, M128(const1));\
s1 = state[i][j];\
s2 = s1;\
TRANSFORM(s2, mul2ipt, t1, t2);\
s3 = _mm_xor_si128(s1, s2);\
state2[r1][c] = _mm_xor_si128(state2[r1][c], s2);\
state2[r2][c] = _mm_xor_si128(state2[r2][c], s1);\
state2[r3][c] = _mm_xor_si128(state2[r3][c], s1);\
state2[r4][c] = _mm_xor_si128(state2[r4][c], s3)
#endif
state[i][j] = _mm_aesenc_si128(state[i][j], k1);\
state[i][j] = _mm_aesenc_si128(state[i][j], M128(zero));\
k1 = _mm_add_epi32(k1, M128(const1))
#define ECHO_MIXBYTES(state1, state2, j, t1, t2, s2) \
s2 = _mm_add_epi8(state1[0][j], state1[0][j]);\
t1 = _mm_srli_epi16(state1[0][j], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = s2;\
state2[1][j] = state1[0][j];\
state2[2][j] = state1[0][j];\
state2[3][j] = _mm_xor_si128(s2, state1[0][j]);\
s2 = _mm_add_epi8(state1[1][(j + 1) & 3], state1[1][(j + 1) & 3]);\
t1 = _mm_srli_epi16(state1[1][(j + 1) & 3], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = _mm_xor_si128(state2[0][j], _mm_xor_si128(s2, state1[1][(j + 1) & 3]));\
state2[1][j] = _mm_xor_si128(state2[1][j], s2);\
state2[2][j] = _mm_xor_si128(state2[2][j], state1[1][(j + 1) & 3]);\
state2[3][j] = _mm_xor_si128(state2[3][j], state1[1][(j + 1) & 3]);\
s2 = _mm_add_epi8(state1[2][(j + 2) & 3], state1[2][(j + 2) & 3]);\
t1 = _mm_srli_epi16(state1[2][(j + 2) & 3], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = _mm_xor_si128(state2[0][j], state1[2][(j + 2) & 3]);\
state2[1][j] = _mm_xor_si128(state2[1][j], _mm_xor_si128(s2, state1[2][(j + 2) & 3]));\
state2[2][j] = _mm_xor_si128(state2[2][j], s2);\
state2[3][j] = _mm_xor_si128(state2[3][j], state1[2][(j + 2) & 3]);\
s2 = _mm_add_epi8(state1[3][(j + 3) & 3], state1[3][(j + 3) & 3]);\
t1 = _mm_srli_epi16(state1[3][(j + 3) & 3], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = _mm_xor_si128(state2[0][j], state1[3][(j + 3) & 3]);\
state2[1][j] = _mm_xor_si128(state2[1][j], state1[3][(j + 3) & 3]);\
state2[2][j] = _mm_xor_si128(state2[2][j], _mm_xor_si128(s2, state1[3][(j + 3) & 3]));\
state2[3][j] = _mm_xor_si128(state2[3][j], s2)
s2 = _mm_add_epi8(state1[0][j], state1[0][j]);\
t1 = _mm_srli_epi16(state1[0][j], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = s2;\
state2[1][j] = state1[0][j];\
state2[2][j] = state1[0][j];\
state2[3][j] = _mm_xor_si128(s2, state1[0][j]);\
s2 = _mm_add_epi8(state1[1][(j + 1) & 3], state1[1][(j + 1) & 3]);\
t1 = _mm_srli_epi16(state1[1][(j + 1) & 3], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = _mm_xor_si128(state2[0][j], _mm_xor_si128(s2, state1[1][(j + 1) & 3]));\
state2[1][j] = _mm_xor_si128(state2[1][j], s2);\
state2[2][j] = _mm_xor_si128(state2[2][j], state1[1][(j + 1) & 3]);\
state2[3][j] = _mm_xor_si128(state2[3][j], state1[1][(j + 1) & 3]);\
s2 = _mm_add_epi8(state1[2][(j + 2) & 3], state1[2][(j + 2) & 3]);\
t1 = _mm_srli_epi16(state1[2][(j + 2) & 3], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = _mm_xor_si128(state2[0][j], state1[2][(j + 2) & 3]);\
state2[1][j] = _mm_xor_si128(state2[1][j], _mm_xor_si128(s2, state1[2][(j + 2) & 3]));\
state2[2][j] = _mm_xor_si128(state2[2][j], s2);\
state2[3][j] = _mm_xor_si128(state2[3][j], state1[2][(j + 2) & 3]);\
s2 = _mm_add_epi8(state1[3][(j + 3) & 3], state1[3][(j + 3) & 3]);\
t1 = _mm_srli_epi16(state1[3][(j + 3) & 3], 7);\
t1 = _mm_and_si128(t1, M128(lsbmask));\
t2 = _mm_shuffle_epi8(M128(mul2mask), t1);\
s2 = _mm_xor_si128(s2, t2);\
state2[0][j] = _mm_xor_si128(state2[0][j], state1[3][(j + 3) & 3]);\
state2[1][j] = _mm_xor_si128(state2[1][j], state1[3][(j + 3) & 3]);\
state2[2][j] = _mm_xor_si128(state2[2][j], _mm_xor_si128(s2, state1[3][(j + 3) & 3]));\
state2[3][j] = _mm_xor_si128(state2[3][j], s2)
#define ECHO_ROUND_UNROLL2 \
ECHO_SUBBYTES(_state, 0, 0);\
ECHO_SUBBYTES(_state, 1, 0);\
ECHO_SUBBYTES(_state, 2, 0);\
ECHO_SUBBYTES(_state, 3, 0);\
ECHO_SUBBYTES(_state, 0, 1);\
ECHO_SUBBYTES(_state, 1, 1);\
ECHO_SUBBYTES(_state, 2, 1);\
ECHO_SUBBYTES(_state, 3, 1);\
ECHO_SUBBYTES(_state, 0, 2);\
ECHO_SUBBYTES(_state, 1, 2);\
ECHO_SUBBYTES(_state, 2, 2);\
ECHO_SUBBYTES(_state, 3, 2);\
ECHO_SUBBYTES(_state, 0, 3);\
ECHO_SUBBYTES(_state, 1, 3);\
ECHO_SUBBYTES(_state, 2, 3);\
ECHO_SUBBYTES(_state, 3, 3);\
ECHO_MIXBYTES(_state, _state2, 0, t1, t2, s2);\
ECHO_MIXBYTES(_state, _state2, 1, t1, t2, s2);\
ECHO_MIXBYTES(_state, _state2, 2, t1, t2, s2);\
ECHO_MIXBYTES(_state, _state2, 3, t1, t2, s2);\
ECHO_SUBBYTES(_state2, 0, 0);\
ECHO_SUBBYTES(_state2, 1, 0);\
ECHO_SUBBYTES(_state2, 2, 0);\
ECHO_SUBBYTES(_state2, 3, 0);\
ECHO_SUBBYTES(_state2, 0, 1);\
ECHO_SUBBYTES(_state2, 1, 1);\
ECHO_SUBBYTES(_state2, 2, 1);\
ECHO_SUBBYTES(_state2, 3, 1);\
ECHO_SUBBYTES(_state2, 0, 2);\
ECHO_SUBBYTES(_state2, 1, 2);\
ECHO_SUBBYTES(_state2, 2, 2);\
ECHO_SUBBYTES(_state2, 3, 2);\
ECHO_SUBBYTES(_state2, 0, 3);\
ECHO_SUBBYTES(_state2, 1, 3);\
ECHO_SUBBYTES(_state2, 2, 3);\
ECHO_SUBBYTES(_state2, 3, 3);\
ECHO_MIXBYTES(_state2, _state, 0, t1, t2, s2);\
ECHO_MIXBYTES(_state2, _state, 1, t1, t2, s2);\
ECHO_MIXBYTES(_state2, _state, 2, t1, t2, s2);\
ECHO_MIXBYTES(_state2, _state, 3, t1, t2, s2)
ECHO_SUBBYTES(_state, 0, 0);\
ECHO_SUBBYTES(_state, 1, 0);\
ECHO_SUBBYTES(_state, 2, 0);\
ECHO_SUBBYTES(_state, 3, 0);\
ECHO_SUBBYTES(_state, 0, 1);\
ECHO_SUBBYTES(_state, 1, 1);\
ECHO_SUBBYTES(_state, 2, 1);\
ECHO_SUBBYTES(_state, 3, 1);\
ECHO_SUBBYTES(_state, 0, 2);\
ECHO_SUBBYTES(_state, 1, 2);\
ECHO_SUBBYTES(_state, 2, 2);\
ECHO_SUBBYTES(_state, 3, 2);\
ECHO_SUBBYTES(_state, 0, 3);\
ECHO_SUBBYTES(_state, 1, 3);\
ECHO_SUBBYTES(_state, 2, 3);\
ECHO_SUBBYTES(_state, 3, 3);\
ECHO_MIXBYTES(_state, _state2, 0, t1, t2, s2);\
ECHO_MIXBYTES(_state, _state2, 1, t1, t2, s2);\
ECHO_MIXBYTES(_state, _state2, 2, t1, t2, s2);\
ECHO_MIXBYTES(_state, _state2, 3, t1, t2, s2);\
ECHO_SUBBYTES(_state2, 0, 0);\
ECHO_SUBBYTES(_state2, 1, 0);\
ECHO_SUBBYTES(_state2, 2, 0);\
ECHO_SUBBYTES(_state2, 3, 0);\
ECHO_SUBBYTES(_state2, 0, 1);\
ECHO_SUBBYTES(_state2, 1, 1);\
ECHO_SUBBYTES(_state2, 2, 1);\
ECHO_SUBBYTES(_state2, 3, 1);\
ECHO_SUBBYTES(_state2, 0, 2);\
ECHO_SUBBYTES(_state2, 1, 2);\
ECHO_SUBBYTES(_state2, 2, 2);\
ECHO_SUBBYTES(_state2, 3, 2);\
ECHO_SUBBYTES(_state2, 0, 3);\
ECHO_SUBBYTES(_state2, 1, 3);\
ECHO_SUBBYTES(_state2, 2, 3);\
ECHO_SUBBYTES(_state2, 3, 3);\
ECHO_MIXBYTES(_state2, _state, 0, t1, t2, s2);\
ECHO_MIXBYTES(_state2, _state, 1, t1, t2, s2);\
ECHO_MIXBYTES(_state2, _state, 2, t1, t2, s2);\
ECHO_MIXBYTES(_state2, _state, 3, t1, t2, s2)
#define SAVESTATE(dst, src)\
dst[0][0] = src[0][0];\
dst[0][1] = src[0][1];\
dst[0][2] = src[0][2];\
dst[0][3] = src[0][3];\
dst[1][0] = src[1][0];\
dst[1][1] = src[1][1];\
dst[1][2] = src[1][2];\
dst[1][3] = src[1][3];\
dst[2][0] = src[2][0];\
dst[2][1] = src[2][1];\
dst[2][2] = src[2][2];\
dst[2][3] = src[2][3];\
dst[3][0] = src[3][0];\
dst[3][1] = src[3][1];\
dst[3][2] = src[3][2];\
dst[3][3] = src[3][3]
dst[0][0] = src[0][0];\
dst[0][1] = src[0][1];\
dst[0][2] = src[0][2];\
dst[0][3] = src[0][3];\
dst[1][0] = src[1][0];\
dst[1][1] = src[1][1];\
dst[1][2] = src[1][2];\
dst[1][3] = src[1][3];\
dst[2][0] = src[2][0];\
dst[2][1] = src[2][1];\
dst[2][2] = src[2][2];\
dst[2][3] = src[2][3];\
dst[3][0] = src[3][0];\
dst[3][1] = src[3][1];\
dst[3][2] = src[3][2];\
dst[3][3] = src[3][3]
void Compress(hashState_echo *ctx, const unsigned char *pmsg, unsigned int uBlockCount)
{
unsigned int r, b, i, j;
// __m128i t1, t2, t3, t4, s1, s2, s3, k1, ktemp;
__m128i t1, t2, s2, k1;
__m128i _state[4][4], _state2[4][4], _statebackup[4][4];
unsigned int r, b, i, j;
__m128i t1, t2, s2, k1;
__m128i _state[4][4], _state2[4][4], _statebackup[4][4];
for(i = 0; i < 4; i++)
for(j = 0; j < ctx->uHashSize / 256; j++)
_state[i][j] = ctx->state[i][j];
for(i = 0; i < 4; i++)
for(j = 0; j < ctx->uHashSize / 256; j++)
_state[i][j] = ctx->state[i][j];
for(b = 0; b < uBlockCount; b++)
{
ctx->k = _mm_add_epi64(ctx->k, ctx->const1536);
#ifdef NO_AES_NI
// transform cv
for(i = 0; i < 4; i++)
for(j = 0; j < ctx->uHashSize / 256; j++)
{
TRANSFORM(_state[i][j], _k_ipt, t1, t2);
}
#endif
for(b = 0; b < uBlockCount; b++)
// load message
for(j = ctx->uHashSize / 256; j < 4; j++)
{
ctx->k = _mm_add_epi64(ctx->k, ctx->const1536);
// load message
for(j = ctx->uHashSize / 256; j < 4; j++)
{
for(i = 0; i < 4; i++)
{
_state[i][j] = _mm_loadu_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
#ifdef NO_AES_NI
// transform message
TRANSFORM(_state[i][j], _k_ipt, t1, t2);
#endif
}
}
// save state
SAVESTATE(_statebackup, _state);
k1 = ctx->k;
#ifndef NO_AES_NI
for(r = 0; r < ctx->uRounds / 2; r++)
{
ECHO_ROUND_UNROLL2;
}
#else
for(r = 0; r < ctx->uRounds / 2; r++)
{
_state2[0][0] = M128(zero); _state2[1][0] = M128(zero); _state2[2][0] = M128(zero); _state2[3][0] = M128(zero);
_state2[0][1] = M128(zero); _state2[1][1] = M128(zero); _state2[2][1] = M128(zero); _state2[3][1] = M128(zero);
_state2[0][2] = M128(zero); _state2[1][2] = M128(zero); _state2[2][2] = M128(zero); _state2[3][2] = M128(zero);
_state2[0][3] = M128(zero); _state2[1][3] = M128(zero); _state2[2][3] = M128(zero); _state2[3][3] = M128(zero);
ECHO_SUB_AND_MIX(_state, 0, 0, _state2, 0, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state, 1, 0, _state2, 3, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state, 2, 0, _state2, 2, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state, 3, 0, _state2, 1, 3, 0, 1, 2);
ECHO_SUB_AND_MIX(_state, 0, 1, _state2, 1, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state, 1, 1, _state2, 0, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state, 2, 1, _state2, 3, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state, 3, 1, _state2, 2, 3, 0, 1, 2);
ECHO_SUB_AND_MIX(_state, 0, 2, _state2, 2, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state, 1, 2, _state2, 1, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state, 2, 2, _state2, 0, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state, 3, 2, _state2, 3, 3, 0, 1, 2);
ECHO_SUB_AND_MIX(_state, 0, 3, _state2, 3, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state, 1, 3, _state2, 2, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state, 2, 3, _state2, 1, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state, 3, 3, _state2, 0, 3, 0, 1, 2);
_state[0][0] = M128(zero); _state[1][0] = M128(zero); _state[2][0] = M128(zero); _state[3][0] = M128(zero);
_state[0][1] = M128(zero); _state[1][1] = M128(zero); _state[2][1] = M128(zero); _state[3][1] = M128(zero);
_state[0][2] = M128(zero); _state[1][2] = M128(zero); _state[2][2] = M128(zero); _state[3][2] = M128(zero);
_state[0][3] = M128(zero); _state[1][3] = M128(zero); _state[2][3] = M128(zero); _state[3][3] = M128(zero);
ECHO_SUB_AND_MIX(_state2, 0, 0, _state, 0, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state2, 1, 0, _state, 3, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state2, 2, 0, _state, 2, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state2, 3, 0, _state, 1, 3, 0, 1, 2);
ECHO_SUB_AND_MIX(_state2, 0, 1, _state, 1, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state2, 1, 1, _state, 0, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state2, 2, 1, _state, 3, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state2, 3, 1, _state, 2, 3, 0, 1, 2);
ECHO_SUB_AND_MIX(_state2, 0, 2, _state, 2, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state2, 1, 2, _state, 1, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state2, 2, 2, _state, 0, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state2, 3, 2, _state, 3, 3, 0, 1, 2);
ECHO_SUB_AND_MIX(_state2, 0, 3, _state, 3, 0, 1, 2, 3);
ECHO_SUB_AND_MIX(_state2, 1, 3, _state, 2, 1, 2, 3, 0);
ECHO_SUB_AND_MIX(_state2, 2, 3, _state, 1, 2, 3, 0, 1);
ECHO_SUB_AND_MIX(_state2, 3, 3, _state, 0, 3, 0, 1, 2);
}
#endif
if(ctx->uHashSize == 256)
{
for(i = 0; i < 4; i++)
{
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][1]);
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][3]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][1]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][3]);
}
}
else
{
for(i = 0; i < 4; i++)
{
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
_state[i][1] = _mm_xor_si128(_state[i][1], _state[i][3]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][1]);
_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][3]);
}
}
pmsg += ctx->uBlockLength;
for(i = 0; i < 4; i++)
{
_state[i][j] = _mm_loadu_si128((__m128i*)pmsg + 4 * (j - (ctx->uHashSize / 256)) + i);
}
}
#ifdef NO_AES_NI
// transform state
for(i = 0; i < 4; i++)
for(j = 0; j < 4; j++)
{
TRANSFORM(_state[i][j], _k_opt, t1, t2);
}
#endif
// save state
SAVESTATE(_statebackup, _state);
SAVESTATE(ctx->state, _state);
k1 = ctx->k;
for(r = 0; r < ctx->uRounds / 2; r++)
{
ECHO_ROUND_UNROLL2;
}
if(ctx->uHashSize == 256)
{
for(i = 0; i < 4; i++)
{
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][1]);
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][3]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][1]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][3]);
}
}
else
{
for(i = 0; i < 4; i++)
{
_state[i][0] = _mm_xor_si128(_state[i][0], _state[i][2]);
_state[i][1] = _mm_xor_si128(_state[i][1], _state[i][3]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][0]);
_state[i][0] = _mm_xor_si128(_state[i][0], _statebackup[i][2]);
_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][1]);
_state[i][1] = _mm_xor_si128(_state[i][1], _statebackup[i][3]);
}
}
pmsg += ctx->uBlockLength;
}
SAVESTATE(ctx->state, _state);
}

4
algo/echo/aes_ni/hash_api.h
View File

@@ -30,6 +30,7 @@
typedef struct
{
__m128i state[4][4];
BitSequence buffer[192];
__m128i k;
__m128i hashsize;
__m128i const1536;
@@ -39,9 +40,8 @@ typedef struct
unsigned int uBlockLength;
unsigned int uBufferBytes;
DataLength processed_bits;
BitSequence buffer[192];
} hashState_echo;
} hashState_echo __attribute__ ((aligned (64)));
HashReturn init_echo(hashState_echo *state, int hashbitlen);

1031
algo/echo/sse2/echo.c
View File

File diff suppressed because it is too large Load Diff

320
algo/echo/sse2/sph_echo.h
View File

@@ -1,320 +0,0 @@
/* $Id: sph_echo.h 216 2010-06-08 09:46:57Z tp $ */
/**
* ECHO interface. ECHO is a family of functions which differ by
* their output size; this implementation defines ECHO for output
* sizes 224, 256, 384 and 512 bits.
*
* ==========================(LICENSE BEGIN)============================
*
* Copyright (c) 2007-2010 Projet RNRT SAPHIR
*
* Permission is hereby granted, free of charge, to any person obtaining
* a copy of this software and associated documentation files (the
* "Software"), to deal in the Software without restriction, including
* without limitation the rights to use, copy, modify, merge, publish,
* distribute, sublicense, and/or sell copies of the Software, and to
* permit persons to whom the Software is furnished to do so, subject to
* the following conditions:
*
* The above copyright notice and this permission notice shall be
* included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*
* ===========================(LICENSE END)=============================
*
* @file sph_echo.h
* @author Thomas Pornin <thomas.pornin@cryptolog.com>
*/
#ifndef SPH_ECHO_H__
#define SPH_ECHO_H__
#ifdef __cplusplus
extern "C"{
#endif
#include <stddef.h>
#include "algo/sha/sph_types.h"
/**
* Output size (in bits) for ECHO-224.
*/
#define SPH_SIZE_echo224 224
/**
* Output size (in bits) for ECHO-256.
*/
#define SPH_SIZE_echo256 256
/**
* Output size (in bits) for ECHO-384.
*/
#define SPH_SIZE_echo384 384
/**
* Output size (in bits) for ECHO-512.
*/
#define SPH_SIZE_echo512 512
/**
* This structure is a context for ECHO computations: it contains the
* intermediate values and some data from the last entered block. Once
* an ECHO computation has been performed, the context can be reused for
* another computation. This specific structure is used for ECHO-224
* and ECHO-256.
*
* The contents of this structure are private. A running ECHO computation
* can be cloned by copying the context (e.g. with a simple
* <code>memcpy()</code>).
*/
typedef struct {
#ifndef DOXYGEN_IGNORE
unsigned char buf[192]; /* first field, for alignment */
size_t ptr;
union {
sph_u32 Vs[4][4];
#if SPH_64
sph_u64 Vb[4][2];
#endif
} u;
sph_u32 C0, C1, C2, C3;
#endif
} sph_echo_small_context;
/**
* This structure is a context for ECHO computations: it contains the
* intermediate values and some data from the last entered block. Once
* an ECHO computation has been performed, the context can be reused for
* another computation. This specific structure is used for ECHO-384
* and ECHO-512.
*
* The contents of this structure are private. A running ECHO computation
* can be cloned by copying the context (e.g. with a simple
* <code>memcpy()</code>).
*/
typedef struct {
#ifndef DOXYGEN_IGNORE
unsigned char buf[128]; /* first field, for alignment */
size_t ptr;
union {
sph_u32 Vs[8][4];
#if SPH_64
sph_u64 Vb[8][2];
#endif
} u;
sph_u32 C0, C1, C2, C3;
#endif
} sph_echo_big_context;
/**
* Type for a ECHO-224 context (identical to the common "small" context).
*/
typedef sph_echo_small_context sph_echo224_context;
/**
* Type for a ECHO-256 context (identical to the common "small" context).
*/
typedef sph_echo_small_context sph_echo256_context;
/**
* Type for a ECHO-384 context (identical to the common "big" context).
*/
typedef sph_echo_big_context sph_echo384_context;
/**
* Type for a ECHO-512 context (identical to the common "big" context).
*/
typedef sph_echo_big_context sph_echo512_context;
/**
* Initialize an ECHO-224 context. This process performs no memory allocation.
*
* @param cc the ECHO-224 context (pointer to a
* <code>sph_echo224_context</code>)
*/
void sph_echo224_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the ECHO-224 context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_echo224(void *cc, const void *data, size_t len);
/**
* Terminate the current ECHO-224 computation and output the result into
* the provided buffer. The destination buffer must be wide enough to
* accomodate the result (28 bytes). The context is automatically
* reinitialized.
*
* @param cc the ECHO-224 context
* @param dst the destination buffer
*/
void sph_echo224_close(void *cc, void *dst);
/**
* Add a few additional bits (0 to 7) to the current computation, then
* terminate it and output the result in the provided buffer, which must
* be wide enough to accomodate the result (28 bytes). If bit number i
* in <code>ub</code> has value 2^i, then the extra bits are those
* numbered 7 downto 8-n (this is the big-endian convention at the byte
* level). The context is automatically reinitialized.
*
* @param cc the ECHO-224 context
* @param ub the extra bits
* @param n the number of extra bits (0 to 7)
* @param dst the destination buffer
*/
void sph_echo224_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
/**
* Initialize an ECHO-256 context. This process performs no memory allocation.
*
* @param cc the ECHO-256 context (pointer to a
* <code>sph_echo256_context</code>)
*/
void sph_echo256_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the ECHO-256 context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_echo256(void *cc, const void *data, size_t len);
/**
* Terminate the current ECHO-256 computation and output the result into
* the provided buffer. The destination buffer must be wide enough to
* accomodate the result (32 bytes). The context is automatically
* reinitialized.
*
* @param cc the ECHO-256 context
* @param dst the destination buffer
*/
void sph_echo256_close(void *cc, void *dst);
/**
* Add a few additional bits (0 to 7) to the current computation, then
* terminate it and output the result in the provided buffer, which must
* be wide enough to accomodate the result (32 bytes). If bit number i
* in <code>ub</code> has value 2^i, then the extra bits are those
* numbered 7 downto 8-n (this is the big-endian convention at the byte
* level). The context is automatically reinitialized.
*
* @param cc the ECHO-256 context
* @param ub the extra bits
* @param n the number of extra bits (0 to 7)
* @param dst the destination buffer
*/
void sph_echo256_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
/**
* Initialize an ECHO-384 context. This process performs no memory allocation.
*
* @param cc the ECHO-384 context (pointer to a
* <code>sph_echo384_context</code>)
*/
void sph_echo384_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the ECHO-384 context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_echo384(void *cc, const void *data, size_t len);
/**
* Terminate the current ECHO-384 computation and output the result into
* the provided buffer. The destination buffer must be wide enough to
* accomodate the result (48 bytes). The context is automatically
* reinitialized.
*
* @param cc the ECHO-384 context
* @param dst the destination buffer
*/
void sph_echo384_close(void *cc, void *dst);
/**
* Add a few additional bits (0 to 7) to the current computation, then
* terminate it and output the result in the provided buffer, which must
* be wide enough to accomodate the result (48 bytes). If bit number i
* in <code>ub</code> has value 2^i, then the extra bits are those
* numbered 7 downto 8-n (this is the big-endian convention at the byte
* level). The context is automatically reinitialized.
*
* @param cc the ECHO-384 context
* @param ub the extra bits
* @param n the number of extra bits (0 to 7)
* @param dst the destination buffer
*/
void sph_echo384_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
/**
* Initialize an ECHO-512 context. This process performs no memory allocation.
*
* @param cc the ECHO-512 context (pointer to a
* <code>sph_echo512_context</code>)
*/
void sph_echo512_init(void *cc);
/**
* Process some data bytes. It is acceptable that <code>len</code> is zero
* (in which case this function does nothing).
*
* @param cc the ECHO-512 context
* @param data the input data
* @param len the input data length (in bytes)
*/
void sph_echo512(void *cc, const void *data, size_t len);
/**
* Terminate the current ECHO-512 computation and output the result into
* the provided buffer. The destination buffer must be wide enough to
* accomodate the result (64 bytes). The context is automatically
* reinitialized.
*
* @param cc the ECHO-512 context
* @param dst the destination buffer
*/
void sph_echo512_close(void *cc, void *dst);
/**
* Add a few additional bits (0 to 7) to the current computation, then
* terminate it and output the result in the provided buffer, which must
* be wide enough to accomodate the result (64 bytes). If bit number i
* in <code>ub</code> has value 2^i, then the extra bits are those
* numbered 7 downto 8-n (this is the big-endian convention at the byte
* level). The context is automatically reinitialized.
*
* @param cc the ECHO-512 context
* @param ub the extra bits
* @param n the number of extra bits (0 to 7)
* @param dst the destination buffer
*/
void sph_echo512_addbits_and_close(
void *cc, unsigned ub, unsigned n, void *dst);
#ifdef __cplusplus
}
#endif
#endif

8
algo/groestl/myrgr-4way.c
View File

@@ -33,7 +33,7 @@ void myriad_4way_hash( void *output, const void *input )
myrgr_4way_ctx_holder ctx;
memcpy( &ctx, &myrgr_4way_ctx, sizeof(myrgr_4way_ctx) );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, input, 640 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, input, 640 );
update_and_final_groestl( &ctx.groestl, (char*)hash0, (char*)hash0, 640 );
memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
@@ -43,12 +43,12 @@ void myriad_4way_hash( void *output, const void *input )
memcpy( &ctx.groestl, &myrgr_4way_ctx.groestl, sizeof(hashState_groestl) );
update_and_final_groestl( &ctx.groestl, (char*)hash3, (char*)hash3, 640 );
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
sha256_4way( &ctx.sha, vhash, 64 );
sha256_4way_close( &ctx.sha, vhash );
mm_deinterleave_4x32( output, output+32, output+64, output+96,
mm128_deinterleave_4x32( output, output+32, output+64, output+96,
vhash, 256 );
}
@@ -79,7 +79,7 @@ int scanhash_myriad_4way( int thr_id, struct work *work, uint32_t max_nonce,
( (uint32_t*)ptarget )[7] = 0x0000ff;
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
do {
be32enc( noncep, n );

8
algo/haval/haval-hash-4way.c
View File

@@ -83,7 +83,7 @@ extern "C"{
_mm_xor_si128( _mm_xor_si128( _mm_and_si128( x1, x2 ), \
_mm_or_si128( x4, x6 ) ), x5 ) ), \
_mm_and_si128( x4, \
_mm_xor_si128( _mm_xor_si128( _mm_and_si128( mm_not(x2), x5 ), \
_mm_xor_si128( _mm_xor_si128( _mm_and_si128( mm128_not(x2), x5 ), \
_mm_xor_si128( x1, x6 ) ), x0 ) ) ), \
_mm_xor_si128( _mm_and_si128( x2, x6 ), x0 ) )
@@ -91,7 +91,7 @@ extern "C"{
#define F5(x6, x5, x4, x3, x2, x1, x0) \
_mm_xor_si128( \
_mm_and_si128( x0, \
mm_not( _mm_xor_si128( \
mm128_not( _mm_xor_si128( \
_mm_and_si128( _mm_and_si128( x1, x2 ), x3 ), x5 ) ) ), \
_mm_xor_si128( _mm_xor_si128( _mm_and_si128( x1, x4 ), \
_mm_and_si128( x2, x5 ) ), \
@@ -136,8 +136,8 @@ extern "C"{
#define STEP(n, p, x7, x6, x5, x4, x3, x2, x1, x0, w, c) \
do { \
__m128i t = FP ## n ## _ ## p(x6, x5, x4, x3, x2, x1, x0); \
x7 = _mm_add_epi32( _mm_add_epi32( mm_ror_32( t, 7 ), \
mm_ror_32( x7, 11 ) ), \
x7 = _mm_add_epi32( _mm_add_epi32( mm128_ror_32( t, 7 ), \
mm128_ror_32( x7, 11 ) ), \
_mm_add_epi32( w, _mm_set1_epi32( c ) ) ); \
} while (0)

23
algo/keccak/keccak-4way.c
View File

@@ -10,14 +10,10 @@
void keccakhash_4way(void *state, const void *input)
{
uint64_t vhash[4*4] __attribute__ ((aligned (64)));
keccak256_4way_context ctx;
keccak256_4way_init( &ctx );
keccak256_4way( &ctx, input, 80 );
keccak256_4way_close( &ctx, vhash );
mm256_deinterleave_4x64( state, state+32, state+64, state+96, vhash, 256 );
keccak256_4way_close( &ctx, state );
}
int scanhash_keccak_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -25,6 +21,8 @@ int scanhash_keccak_4way( int thr_id, struct work *work, uint32_t max_nonce,
{
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[25]); // 3*8+1
uint32_t lane_hash[8];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t n = pdata[19];
@@ -49,13 +47,16 @@ int scanhash_keccak_4way( int thr_id, struct work *work, uint32_t max_nonce,
keccakhash_4way( hash, vdata );
for ( int i = 0; i < 4; i++ )
if ( ( ( (hash+(i<<3))[7] & 0xFFFFFF00 ) == 0 )
&& fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( ( ( hash7[ lane<<1 ] & 0xFFFFFF00 ) == 0 ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm256_extract_lane_4x64( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;

20
algo/luffa/luffa_for_sse2.c
View File

@@ -272,8 +272,8 @@ HashReturn update_luffa( hashState_luffa *state, const BitSequence *data,
// full blocks
for ( i = 0; i < blocks; i++ )
{
rnd512( state, mm_bswap_32( casti_m128i( data, 1 ) ),
mm_bswap_32( casti_m128i( data, 0 ) ) );
rnd512( state, mm128_bswap_32( casti_m128i( data, 1 ) ),
mm128_bswap_32( casti_m128i( data, 0 ) ) );
data += MSG_BLOCK_BYTE_LEN;
}
@@ -282,7 +282,7 @@ HashReturn update_luffa( hashState_luffa *state, const BitSequence *data,
if ( state->rembytes )
{
// remaining data bytes
casti_m128i( state->buffer, 0 ) = mm_bswap_32( cast_m128i( data ) );
casti_m128i( state->buffer, 0 ) = mm128_bswap_32( cast_m128i( data ) );
// padding of partial block
casti_m128i( state->buffer, 1 ) =
_mm_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0x80,0,0,0 );
@@ -324,8 +324,8 @@ HashReturn update_and_final_luffa( hashState_luffa *state, BitSequence* output,
// full blocks
for ( i = 0; i < blocks; i++ )
{
rnd512( state, mm_bswap_32( casti_m128i( data, 1 ) ),
mm_bswap_32( casti_m128i( data, 0 ) ) );
rnd512( state, mm128_bswap_32( casti_m128i( data, 1 ) ),
mm128_bswap_32( casti_m128i( data, 0 ) ) );
data += MSG_BLOCK_BYTE_LEN;
}
@@ -334,7 +334,7 @@ HashReturn update_and_final_luffa( hashState_luffa *state, BitSequence* output,
{
// padding of partial block
rnd512( state, _mm_set_epi8( 0,0,0,0, 0,0,0,0, 0,0,0,0, 0x80,0,0,0 ),
mm_bswap_32( cast_m128i( data ) ) );
mm128_bswap_32( cast_m128i( data ) ) );
}
else
{
@@ -587,8 +587,8 @@ static void finalization512( hashState_luffa *state, uint32 *b )
_mm_store_si128((__m128i*)&hash[0], t[0]);
_mm_store_si128((__m128i*)&hash[4], t[1]);
casti_m128i( b, 0 ) = mm_bswap_32( casti_m128i( hash, 0 ) );
casti_m128i( b, 1 ) = mm_bswap_32( casti_m128i( hash, 1 ) );
casti_m128i( b, 0 ) = mm128_bswap_32( casti_m128i( hash, 0 ) );
casti_m128i( b, 1 ) = mm128_bswap_32( casti_m128i( hash, 1 ) );
rnd512( state, zero, zero );
@@ -609,8 +609,8 @@ static void finalization512( hashState_luffa *state, uint32 *b )
_mm_store_si128((__m128i*)&hash[0], t[0]);
_mm_store_si128((__m128i*)&hash[4], t[1]);
casti_m128i( b, 2 ) = mm_bswap_32( casti_m128i( hash, 0 ) );
casti_m128i( b, 3 ) = mm_bswap_32( casti_m128i( hash, 1 ) );
casti_m128i( b, 2 ) = mm128_bswap_32( casti_m128i( hash, 0 ) );
casti_m128i( b, 3 ) = mm128_bswap_32( casti_m128i( hash, 1 ) );
}
#endif

6
algo/lyra2/allium-4way.c
View File

@@ -1,4 +1,4 @@
#include "allium-gate.h"
#include "lyra2-gate.h"
#include <memory.h>
#include <mm_malloc.h>
@@ -7,7 +7,7 @@
#include "algo/blake/blake-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/skein/skein-hash-4way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/groestl/aes_ni/hash-groestl256.h"
typedef struct {
@@ -108,7 +108,7 @@ int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
( (uint32_t*)ptarget )[7] = 0x0000ff;
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake256_4way_init( &allium_4way_ctx.blake );
blake256_4way( &allium_4way_ctx.blake, vdata, 64 );

22
algo/lyra2/allium-gate.c
View File

@@ -1,22 +0,0 @@
#include "allium-gate.h"
int64_t get_max64_0xFFFFLL() { return 0xFFFFLL; }
bool register_allium_algo( algo_gate_t* gate )
{
#if defined (ALLIUM_4WAY)
gate->miner_thread_init = (void*)&init_allium_4way_ctx;
gate->scanhash = (void*)&scanhash_allium_4way;
gate->hash = (void*)&allium_4way_hash;
#else
gate->miner_thread_init = (void*)&init_allium_ctx;
gate->scanhash = (void*)&scanhash_allium;
gate->hash = (void*)&allium_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->set_target = (void*)&alt_set_target;
gate->get_max64 = (void*)&get_max64_0xFFFFLL;
return true;
};

29
algo/lyra2/allium-gate.h
View File

@@ -1,29 +0,0 @@
#ifndef ALLIUM_GATE_H__
#define ALLIUM_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#include "lyra2.h"
#if defined(__AVX2__) && defined(__AES__)
#define ALLIUM_4WAY
#endif
bool register_allium_algo( algo_gate_t* gate );
#if defined(ALLIUM_4WAY)
void allium_4way_hash( void *state, const void *input );
int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_allium_4way_ctx();
#endif
void allium_hash( void *state, const void *input );
int scanhash_allium( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_allium_ctx();
#endif

4
algo/lyra2/allium.c
View File

@@ -1,9 +1,9 @@
#include "allium-gate.h"
#include "lyra2-gate.h"
#include <memory.h>
#include "algo/blake/sph_blake.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#if defined(__AES__)
#include "algo/groestl/aes_ni/hash-groestl256.h"
#else

178
algo/lyra2/lyra2-gate.c Normal file
View File

@@ -0,0 +1,178 @@
#include "lyra2-gate.h"
__thread uint64_t* l2v3_wholeMatrix;
bool lyra2rev3_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
int size = (int64_t)ROW_LEN_BYTES * 4; // nRows;
l2v3_wholeMatrix = _mm_malloc( size, 64 );
#if defined (LYRA2REV3_4WAY)
init_lyra2rev3_4way_ctx();;
#else
init_lyra2rev3_ctx();
#endif
return l2v3_wholeMatrix;
}
bool register_lyra2rev3_algo( algo_gate_t* gate )
{
#if defined (LYRA2REV3_4WAY)
gate->scanhash = (void*)&scanhash_lyra2rev3_4way;
gate->hash = (void*)&lyra2rev3_4way_hash;
#else
gate->scanhash = (void*)&scanhash_lyra2rev3;
gate->hash = (void*)&lyra2rev3_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2rev3_thread_init;
gate->set_target = (void*)&alt_set_target;
return true;
};
//////////////////////////////////
__thread uint64_t* l2v2_wholeMatrix;
bool lyra2rev2_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
int size = (int64_t)ROW_LEN_BYTES * 4; // nRows;
l2v2_wholeMatrix = _mm_malloc( size, 64 );
#if defined (LYRA2REV2_4WAY)
init_lyra2rev2_4way_ctx();;
#else
init_lyra2rev2_ctx();
#endif
return l2v2_wholeMatrix;
}
bool register_lyra2rev2_algo( algo_gate_t* gate )
{
#if defined (LYRA2REV2_4WAY)
gate->scanhash = (void*)&scanhash_lyra2rev2_4way;
gate->hash = (void*)&lyra2rev2_4way_hash;
#else
gate->scanhash = (void*)&scanhash_lyra2rev2;
gate->hash = (void*)&lyra2rev2_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2rev2_thread_init;
gate->set_target = (void*)&alt_set_target;
return true;
};
/////////////////////////////
bool register_lyra2z_algo( algo_gate_t* gate )
{
#if defined(LYRA2Z_8WAY)
gate->miner_thread_init = (void*)&lyra2z_8way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z_8way;
gate->hash = (void*)&lyra2z_8way_hash;
#elif defined(LYRA2Z_4WAY)
gate->miner_thread_init = (void*)&lyra2z_4way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z_4way;
gate->hash = (void*)&lyra2z_4way_hash;
#else
gate->miner_thread_init = (void*)&lyra2z_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z;
gate->hash = (void*)&lyra2z_hash;
#endif
gate->optimizations = SSE42_OPT | AVX2_OPT;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&alt_set_target;
return true;
};
////////////////////////
bool register_lyra2h_algo( algo_gate_t* gate )
{
#ifdef LYRA2H_4WAY
gate->miner_thread_init = (void*)&lyra2h_4way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h_4way;
gate->hash = (void*)&lyra2h_4way_hash;
#else
gate->miner_thread_init = (void*)&lyra2h_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h;
gate->hash = (void*)&lyra2h_hash;
#endif
gate->optimizations = SSE42_OPT | AVX2_OPT;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&alt_set_target;
return true;
};
/////////////////////////////////
int64_t allium_get_max64_0xFFFFLL() { return 0xFFFFLL; }
bool register_allium_algo( algo_gate_t* gate )
{
#if defined (ALLIUM_4WAY)
gate->miner_thread_init = (void*)&init_allium_4way_ctx;
gate->scanhash = (void*)&scanhash_allium_4way;
gate->hash = (void*)&allium_4way_hash;
#else
gate->miner_thread_init = (void*)&init_allium_ctx;
gate->scanhash = (void*)&scanhash_allium;
gate->hash = (void*)&allium_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->set_target = (void*)&alt_set_target;
gate->get_max64 = (void*)&allium_get_max64_0xFFFFLL;
return true;
};
/////////////////////////////////////////
bool phi2_has_roots;
bool phi2_use_roots = false;
int phi2_get_work_data_size() { return phi2_use_roots ? 144 : 128; }
void phi2_decode_extra_data( struct work *work )
{
if ( work->data[0] & ( 1<<30 ) ) phi2_use_roots = true;
else for ( int i = 20; i < 36; i++ )
{
if (work->data[i]) { phi2_use_roots = true; break; }
}
}
void phi2_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
{
uchar merkle_tree[64] = { 0 };
size_t t;
algo_gate.gen_merkle_root( merkle_tree, sctx );
// Increment extranonce2
for ( t = 0; t < sctx->xnonce2_size && !( ++sctx->job.xnonce2[t] ); t++ );
// Assemble block header
algo_gate.build_block_header( g_work, le32dec( sctx->job.version ),
(uint32_t*) sctx->job.prevhash, (uint32_t*) merkle_tree,
le32dec( sctx->job.ntime ), le32dec(sctx->job.nbits) );
for ( t = 0; t < 16; t++ )
g_work->data[ 20+t ] = ((uint32_t*)sctx->job.extra)[t];
}
bool register_phi2_algo( algo_gate_t* gate )
{
init_phi2_ctx();
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->get_work_data_size = (void*)&phi2_get_work_data_size;
gate->decode_extra_data = (void*)&phi2_decode_extra_data;
gate->build_extraheader = (void*)&phi2_build_extraheader;
gate->set_target = (void*)&alt_set_target;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->scanhash = (void*)&scanhash_phi2;
return true;
}

154
algo/lyra2/lyra2-gate.h Normal file
View File

@@ -0,0 +1,154 @@
#ifndef LYRA2_GATE_H__
#define LYRA2_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#include "lyra2.h"
#if defined(__AVX2__)
#define LYRA2REV3_4WAY
#endif
extern __thread uint64_t* l2v3_wholeMatrix;
bool register_lyra2rev3_algo( algo_gate_t* gate );
#if defined(LYRA2REV3_4WAY)
void lyra2rev3_4way_hash( void *state, const void *input );
int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_lyra2rev3_4way_ctx();
#else
void lyra2rev3_hash( void *state, const void *input );
int scanhash_lyra2rev3( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_lyra2rev3_ctx();
#endif
//////////////////////////////////
#if defined(__AVX2__)
#define LYRA2REV2_4WAY
#endif
extern __thread uint64_t* l2v2_wholeMatrix;
bool register_lyra2rev2_algo( algo_gate_t* gate );
#if defined(LYRA2REV2_4WAY)
void lyra2rev2_4way_hash( void *state, const void *input );
int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_lyra2rev2_4way_ctx();
#else
void lyra2rev2_hash( void *state, const void *input );
int scanhash_lyra2rev2( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_lyra2rev2_ctx();
#endif
/////////////////////////
#if defined(__SSE4_2__)
#define LYRA2Z_4WAY
#endif
#if defined(__AVX2__)
// #define LYRA2Z_8WAY
#endif
#define LYRA2Z_MATRIX_SIZE BLOCK_LEN_INT64 * 8 * 8 * 8
#if defined(LYRA2Z_8WAY)
void lyra2z_8way_hash( void *state, const void *input );
int scanhash_lyra2z_8way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2z_8way_thread_init();
#elif defined(LYRA2Z_4WAY)
void lyra2z_4way_hash( void *state, const void *input );
int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2z_4way_thread_init();
#else
void lyra2z_hash( void *state, const void *input );
int scanhash_lyra2z( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2z_thread_init();
#endif
////////////////////
#if defined(__AVX2__)
#define LYRA2H_4WAY
#endif
#define LYRA2H_MATRIX_SIZE BLOCK_LEN_INT64 * 16 * 16 * 8
#if defined(LYRA2H_4WAY)
void lyra2h_4way_hash( void *state, const void *input );
int scanhash_lyra2h_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2h_4way_thread_init();
#else
void lyra2h_hash( void *state, const void *input );
int scanhash_lyra2h( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2h_thread_init();
#endif
//////////////////////////////////
#if defined(__AVX2__) && defined(__AES__)
#define ALLIUM_4WAY
#endif
bool register_allium_algo( algo_gate_t* gate );
#if defined(ALLIUM_4WAY)
void allium_4way_hash( void *state, const void *input );
int scanhash_allium_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_allium_4way_ctx();
#else
void allium_hash( void *state, const void *input );
int scanhash_allium( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_allium_ctx();
#endif
/////////////////////////////////////////
bool phi2_has_roots;
bool register_phi2_algo( algo_gate_t* gate );
void phi2_hash( void *state, const void *input );
int scanhash_phi2( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
void init_phi2_ctx();
#endif // LYRA2_GATE_H__

180
algo/lyra2/lyra2.c
View File

@@ -211,6 +211,186 @@ int LYRA2REV2( uint64_t* wholeMatrix, void *K, uint64_t kLen, const void *pwd,
return 0;
}
/////////////////////////////////////////////////
int LYRA2REV3( uint64_t* wholeMatrix, void *K, uint64_t kLen, const void *pwd,
const uint64_t pwdlen, const void *salt, const uint64_t saltlen,
const uint64_t timeCost, const uint64_t nRows,
const uint64_t nCols )
{
//====================== Basic variables ============================//
uint64_t _ALIGN(256) state[16];
int64_t row = 2; //index of row to be processed
int64_t prev = 1; //index of prev (last row ever computed/modified)
int64_t rowa = 0; //index of row* (a previous row, deterministically picked during Setup and randomly picked while Wandering)
int64_t tau; //Time Loop iterator
int64_t step = 1; //Visitation step (used during Setup and Wandering phases)
int64_t window = 2; //Visitation window (used to define which rows can be revisited during Setup)
int64_t gap = 1; //Modifier to the step, assuming the values 1 or -1
int64_t i; //auxiliary iteration counter
int64_t v64; // 64bit var for memcpy
uint64_t instance = 0;
//====================================================================/
//=== Initializing the Memory Matrix and pointers to it =============//
//Tries to allocate enough space for the whole memory matrix
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * nCols;
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
const int64_t BLOCK_LEN = BLOCK_LEN_BLAKE2_SAFE_INT64;
/*
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * nCols;
// const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
// for Lyra2REv2, nCols = 4, v1 was using 8
const int64_t BLOCK_LEN = (nCols == 4) ? BLOCK_LEN_BLAKE2_SAFE_INT64
: BLOCK_LEN_BLAKE2_SAFE_BYTES;
*/
uint64_t *ptrWord = wholeMatrix;
// memset( wholeMatrix, 0, ROW_LEN_BYTES * nRows );
//=== Getting the password + salt + basil padded with 10*1 ==========//
//OBS.:The memory matrix will temporarily hold the password: not for saving memory,
//but this ensures that the password copied locally will be overwritten as soon as possible
//First, we clean enough blocks for the password, salt, basil and padding
int64_t nBlocksInput = ( ( saltlen + pwdlen + 6 * sizeof(uint64_t) )
/ BLOCK_LEN_BLAKE2_SAFE_BYTES ) + 1;
byte *ptrByte = (byte*) wholeMatrix;
//Prepends the password
memcpy(ptrByte, pwd, pwdlen);
ptrByte += pwdlen;
//Concatenates the salt
memcpy(ptrByte, salt, saltlen);
ptrByte += saltlen;
memset( ptrByte, 0, nBlocksInput * BLOCK_LEN_BLAKE2_SAFE_BYTES
- (saltlen + pwdlen) );
//Concatenates the basil: every integer passed as parameter, in the order they are provided by the interface
memcpy(ptrByte, &kLen, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = pwdlen;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = saltlen;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = timeCost;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = nRows;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
v64 = nCols;
memcpy(ptrByte, &v64, sizeof(int64_t));
ptrByte += sizeof(uint64_t);
//Now comes the padding
*ptrByte = 0x80; //first byte of padding: right after the password
ptrByte = (byte*) wholeMatrix; //resets the pointer to the start of the memory matrix
ptrByte += nBlocksInput * BLOCK_LEN_BLAKE2_SAFE_BYTES - 1; //sets the pointer to the correct position: end of incomplete block
*ptrByte ^= 0x01; //last byte of padding: at the end of the last incomplete block
// from here on it's all simd acces to state and matrix
// define vector pointers and adjust sizes and pointer offsets
//================= Initializing the Sponge State ====================//
//Sponge state: 16 uint64_t, BLOCK_LEN_INT64 words of them for the bitrate (b) and the remainder for the capacity (c)
initState( state );
//========================= Setup Phase =============================//
//Absorbing salt, password and basil: this is the only place in which the block length is hard-coded to 512 bits
ptrWord = wholeMatrix;
for (i = 0; i < nBlocksInput; i++)
{
absorbBlockBlake2Safe( state, ptrWord ); //absorbs each block of pad(pwd || salt || basil)
ptrWord += BLOCK_LEN; //goes to next block of pad(pwd || salt || basil)
}
//Initializes M[0] and M[1]
reducedSqueezeRow0( state, &wholeMatrix[0], nCols ); //The locally copied password is most likely overwritten here
reducedDuplexRow1( state, &wholeMatrix[0], &wholeMatrix[ROW_LEN_INT64],
nCols);
do
{
//M[row] = rand; //M[row*] = M[row*] XOR rotW(rand)
reducedDuplexRowSetup( state, &wholeMatrix[prev*ROW_LEN_INT64],
&wholeMatrix[rowa*ROW_LEN_INT64],
&wholeMatrix[row*ROW_LEN_INT64], nCols );
//updates the value of row* (deterministically picked during Setup))
rowa = (rowa + step) & (window - 1);
//update prev: it now points to the last row ever computed
prev = row;
//updates row: goes to the next row to be computed
row++;
//Checks if all rows in the window where visited.
if (rowa == 0)
{
step = window + gap; //changes the step: approximately doubles its value
window *= 2; //doubles the size of the re-visitation window
gap = -gap; //inverts the modifier to the step
}
} while (row < nRows);
//===================== Wandering Phase =============================//
row = 0; //Resets the visitation to the first row of the memory matrix
for (tau = 1; tau <= timeCost; tau++)
{
//Step is approximately half the number of all rows of the memory matrix for an odd tau; otherwise, it is -1
step = ((tau & 1) == 0) ? -1 : (nRows >> 1) - 1;
// step = (tau % 2 == 0) ? -1 : nRows / 2 - 1;
do
{
//Selects a pseudorandom index row*
//-----------------------------------------------
instance = state[instance & 0xF];
rowa = state[instance & 0xF] & (unsigned int)(nRows-1);
// rowa = state[0] & (unsigned int)(nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
//rowa = state[0] % nRows; //(USE THIS FOR THE "GENERIC" CASE)
//-------------------------------------------
//Performs a reduced-round duplexing operation over M[row*] XOR M[prev], updating both M[row*] and M[row]
reducedDuplexRow( state, &wholeMatrix[prev*ROW_LEN_INT64],
&wholeMatrix[rowa*ROW_LEN_INT64],
&wholeMatrix[row*ROW_LEN_INT64], nCols );
//update prev: it now points to the last row ever computed
prev = row;
//updates row: goes to the next row to be computed
//----------------------------------------------------
row = (row + step) & (unsigned int)(nRows-1); //(USE THIS IF nRows IS A POWER OF 2)
//row = (row + step) % nRows; //(USE THIS FOR THE "GENERIC" CASE)
//----------------------------------------------------
} while (row != 0);
}
//===================== Wrap-up Phase ===============================//
//Absorbs the last block of the memory matrix
absorbBlock(state, &wholeMatrix[rowa*ROW_LEN_INT64]);
//Squeezes the key
squeeze(state, K, (unsigned int) kLen);
return 0;
}
//////////////////////////////////////////////////
int LYRA2Z( uint64_t* wholeMatrix, void *K, uint64_t kLen, const void *pwd,
const uint64_t pwdlen, const void *salt, const uint64_t saltlen,
const uint64_t timeCost, const uint64_t nRows,

4
algo/lyra2/lyra2.h
View File

@@ -50,6 +50,10 @@ int LYRA2REV2( uint64_t*, void *K, uint64_t kLen, const void *pwd,
uint64_t pwdlen, const void *salt, uint64_t saltlen,
uint64_t timeCost, uint64_t nRows, uint64_t nCols );
int LYRA2REV3( uint64_t*, void *K, uint64_t kLen, const void *pwd,
uint64_t pwdlen, const void *salt, uint64_t saltlen,
uint64_t timeCost, uint64_t nRows, uint64_t nCols );
int LYRA2Z( uint64_t*, void *K, uint64_t kLen, const void *pwd,
uint64_t pwdlen, const void *salt, uint64_t saltlen,
uint64_t timeCost, uint64_t nRows, uint64_t nCols );

6
algo/lyra2/lyra2h-4way.c
View File

@@ -1,4 +1,4 @@
#include "lyra2h-gate.h"
#include "lyra2-gate.h"
#ifdef LYRA2H_4WAY
@@ -36,7 +36,7 @@ void lyra2h_4way_hash( void *state, const void *input )
blake256_4way( &ctx_blake, input + (64*4), 16 );
blake256_4way_close( &ctx_blake, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
LYRA2Z( lyra2h_4way_matrix, hash0, 32, hash0, 32, hash0, 32, 16, 16, 16 );
LYRA2Z( lyra2h_4way_matrix, hash1, 32, hash1, 32, hash1, 32, 16, 16, 16 );
@@ -70,7 +70,7 @@ int scanhash_lyra2h_4way( int thr_id, struct work *work, uint32_t max_nonce,
for ( int i=0; i < 20; i++ )
be32enc( &edata[i], pdata[i] );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
lyra2h_4way_midstate( vdata );

25
algo/lyra2/lyra2h-gate.c
View File

@@ -1,25 +0,0 @@
#include "lyra2h-gate.h"
#include "lyra2.h"
void lyra2h_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool register_lyra2h_algo( algo_gate_t* gate )
{
#ifdef LYRA2H_4WAY
gate->miner_thread_init = (void*)&lyra2h_4way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h_4way;
gate->hash = (void*)&lyra2h_4way_hash;
#else
gate->miner_thread_init = (void*)&lyra2h_thread_init;
gate->scanhash = (void*)&scanhash_lyra2h;
gate->hash = (void*)&lyra2h_hash;
#endif
gate->optimizations = SSE42_OPT | AVX2_OPT;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&lyra2h_set_target;
return true;
};

32
algo/lyra2/lyra2h-gate.h
View File

@@ -1,32 +0,0 @@
#ifndef LYRA2H_GATE_H__
#define LYRA2H_GATE_H__
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__AVX2__)
#define LYRA2H_4WAY
#endif
#define LYRA2H_MATRIX_SIZE BLOCK_LEN_INT64 * 16 * 16 * 8
#if defined(LYRA2H_4WAY)
void lyra2h_4way_hash( void *state, const void *input );
int scanhash_lyra2h_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2h_4way_thread_init();
#endif
void lyra2h_hash( void *state, const void *input );
int scanhash_lyra2h( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2h_thread_init();
#endif

2
algo/lyra2/lyra2h.c
View File

@@ -1,4 +1,4 @@
#include "lyra2h-gate.h"
#include "lyra2-gate.h"
#include <memory.h>
#include <mm_malloc.h>
#include "lyra2.h"

12
algo/lyra2/lyra2rev2-4way.c
View File

@@ -1,13 +1,13 @@
#include "lyra2rev2-gate.h"
#include "lyra2-gate.h"
#include <memory.h>
#if defined (__AVX2__)
#if defined (LYRA2REV2_4WAY)
#include "algo/blake/blake-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/skein/skein-hash-4way.h"
#include "algo/bmw/bmw-hash-4way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
typedef struct {
blake256_4way_context blake;
@@ -74,11 +74,11 @@ void lyra2rev2_4way_hash( void *state, const void *input )
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
bmw256_4way( &ctx.bmw, vhash, 32 );
bmw256_4way_close( &ctx.bmw, vhash );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -101,7 +101,7 @@ int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
swab32_array( edata, pdata, 20 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
blake256_4way_init( &l2v2_4way_ctx.blake );
blake256_4way( &l2v2_4way_ctx.blake, vdata, 64 );

40
algo/lyra2/lyra2rev2-gate.c
View File

@@ -1,40 +0,0 @@
#include "lyra2rev2-gate.h"
__thread uint64_t* l2v2_wholeMatrix;
void lyra2rev2_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool lyra2rev2_thread_init()
{
const int64_t ROW_LEN_INT64 = BLOCK_LEN_INT64 * 4; // nCols
const int64_t ROW_LEN_BYTES = ROW_LEN_INT64 * 8;
int i = (int64_t)ROW_LEN_BYTES * 4; // nRows;
l2v2_wholeMatrix = _mm_malloc( i, 64 );
#if defined (LYRA2REV2_4WAY)
init_lyra2rev2_4way_ctx();;
#else
init_lyra2rev2_ctx();
#endif
return l2v2_wholeMatrix;
}
bool register_lyra2rev2_algo( algo_gate_t* gate )
{
#if defined (LYRA2REV2_4WAY)
gate->scanhash = (void*)&scanhash_lyra2rev2_4way;
gate->hash = (void*)&lyra2rev2_4way_hash;
#else
gate->scanhash = (void*)&scanhash_lyra2rev2;
gate->hash = (void*)&lyra2rev2_hash;
#endif
gate->optimizations = SSE2_OPT | AES_OPT | SSE42_OPT | AVX2_OPT;
gate->miner_thread_init = (void*)&lyra2rev2_thread_init;
gate->set_target = (void*)&lyra2rev2_set_target;
return true;
};

35
algo/lyra2/lyra2rev2-gate.h
View File

@@ -1,35 +0,0 @@
#ifndef LYRA2REV2_GATE_H__
#define LYRA2REV2_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#include "lyra2.h"
#if defined(__AVX2__)
#define LYRA2REV2_4WAY
#endif
extern __thread uint64_t* l2v2_wholeMatrix;
bool register_lyra2rev2_algo( algo_gate_t* gate );
#if defined(LYRA2REV2_4WAY)
void lyra2rev2_4way_hash( void *state, const void *input );
int scanhash_lyra2rev2_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_lyra2rev2_4way_ctx();
#endif
void lyra2rev2_hash( void *state, const void *input );
int scanhash_lyra2rev2( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool init_lyra2rev2_ctx();
#endif

4
algo/lyra2/lyra2rev2.c
View File

@@ -1,11 +1,11 @@
#include "lyra2rev2-gate.h"
#include "lyra2-gate.h"
#include <memory.h>
#include "algo/blake/sph_blake.h"
#include "algo/cubehash/sph_cubehash.h"
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
//#include "lyra2.h"
typedef struct {

110
algo/lyra2/lyra2rev3-4way.c Normal file
View File

@@ -0,0 +1,110 @@
#include "lyra2-gate.h"
#include <memory.h>
#if defined (LYRA2REV3_4WAY)
#include "algo/blake/blake-hash-4way.h"
#include "algo/bmw/bmw-hash-4way.h"
#include "algo/cubehash/cubehash_sse2.h"
typedef struct {
blake256_4way_context blake;
cubehashParam cube;
bmw256_4way_context bmw;
} lyra2v3_4way_ctx_holder;
static lyra2v3_4way_ctx_holder l2v3_4way_ctx;
bool init_lyra2rev3_4way_ctx()
{
blake256_4way_init( &l2v3_4way_ctx.blake );
cubehashInit( &l2v3_4way_ctx.cube, 256, 16, 32 );
bmw256_4way_init( &l2v3_4way_ctx.bmw );
return true;
}
void lyra2rev3_4way_hash( void *state, const void *input )
{
uint32_t vhash[8*4] __attribute__ ((aligned (64)));
uint32_t hash0[8] __attribute__ ((aligned (64)));
uint32_t hash1[8] __attribute__ ((aligned (32)));
uint32_t hash2[8] __attribute__ ((aligned (32)));
uint32_t hash3[8] __attribute__ ((aligned (32)));
lyra2v3_4way_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &l2v3_4way_ctx, sizeof(l2v3_4way_ctx) );
blake256_4way( &ctx.blake, input, 80 );
blake256_4way_close( &ctx.blake, vhash );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
LYRA2REV3( l2v3_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
LYRA2REV3( l2v3_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
cubehashUpdateDigest( &ctx.cube, (byte*) hash0, (const byte*) hash0, 32 );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash1, (const byte*) hash1, 32 );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash2, (const byte*) hash2, 32 );
cubehashReinit( &ctx.cube );
cubehashUpdateDigest( &ctx.cube, (byte*) hash3, (const byte*) hash3, 32 );
LYRA2REV3( l2v3_wholeMatrix, hash0, 32, hash0, 32, hash0, 32, 1, 4, 4 );
LYRA2REV3( l2v3_wholeMatrix, hash1, 32, hash1, 32, hash1, 32, 1, 4, 4 );
LYRA2REV3( l2v3_wholeMatrix, hash2, 32, hash2, 32, hash2, 32, 1, 4, 4 );
LYRA2REV3( l2v3_wholeMatrix, hash3, 32, hash3, 32, hash3, 32, 1, 4, 4 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 256 );
bmw256_4way( &ctx.bmw, vhash, 32 );
bmw256_4way_close( &ctx.bmw, vhash );
mm128_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_lyra2rev3_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
const uint32_t Htarg = ptarget[7];
uint32_t *nonces = work->nonces;
int num_found = 0;
uint32_t *noncep = vdata + 76; // 19*4
if ( opt_benchmark )
( (uint32_t*)ptarget )[7] = 0x0000ff;
swab32_array( edata, pdata, 20 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
do {
be32enc( noncep, n );
be32enc( noncep+1, n+1 );
be32enc( noncep+2, n+2 );
be32enc( noncep+3, n+3 );
lyra2rev3_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
}
n += 4;
} while ( (num_found == 0) && (n < max_nonce-4)
&& !work_restart[thr_id].restart);
*hashes_done = n - first_nonce + 1;
return num_found;
}
#endif

102
algo/lyra2/lyra2rev3.c Normal file
View File

@@ -0,0 +1,102 @@
#include "lyra2-gate.h"
#include <memory.h>
#include "algo/blake/sph_blake.h"
#include "algo/cubehash/sph_cubehash.h"
#include "algo/bmw/sph_bmw.h"
#include "algo/cubehash/cubehash_sse2.h"
//#include "lyra2.h"
typedef struct {
cubehashParam cube;
// cubehashParam cube2;
sph_blake256_context blake;
sph_bmw256_context bmw;
} lyra2v3_ctx_holder;
static lyra2v3_ctx_holder lyra2v3_ctx;
static __thread sph_blake256_context l2v3_blake_mid;
bool init_lyra2rev3_ctx()
{
cubehashInit( &lyra2v3_ctx.cube, 256, 16, 32 );
// cubehashInit( &lyra2v3_ctx.cube2, 256, 16, 32 );
sph_blake256_init( &lyra2v3_ctx.blake );
sph_bmw256_init( &lyra2v3_ctx.bmw );
return true;
}
void l2v3_blake256_midstate( const void* input )
{
memcpy( &l2v3_blake_mid, &lyra2v3_ctx.blake, sizeof l2v3_blake_mid );
sph_blake256( &l2v3_blake_mid, input, 64 );
}
void lyra2rev3_hash( void *state, const void *input )
{
lyra2v3_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &lyra2v3_ctx, sizeof(lyra2v3_ctx) );
uint8_t hash[128] __attribute__ ((aligned (64)));
#define hashA hash
#define hashB hash+64
const int midlen = 64; // bytes
const int tail = 80 - midlen; // 16
memcpy( &ctx.blake, &l2v3_blake_mid, sizeof l2v3_blake_mid );
sph_blake256( &ctx.blake, (uint8_t*)input + midlen, tail );
sph_blake256_close( &ctx.blake, hash );
LYRA2REV3( l2v3_wholeMatrix, hash, 32, hash, 32, hash, 32, 1, 4, 4 );
cubehashUpdateDigest( &ctx.cube, (byte*) hashA,
(const byte*) hash, 32 );
LYRA2REV3( l2v3_wholeMatrix, hash, 32, hash, 32, hash, 32, 1, 4, 4 );
sph_bmw256( &ctx.bmw, hash, 32 );
sph_bmw256_close( &ctx.bmw, hash );
memcpy( state, hash, 32 );
}
int scanhash_lyra2rev3(int thr_id, struct work *work,
uint32_t max_nonce, uint64_t *hashes_done)
{
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
uint32_t endiandata[20] __attribute__ ((aligned (64)));
uint32_t hash[8] __attribute__((aligned(64)));
const uint32_t first_nonce = pdata[19];
uint32_t nonce = first_nonce;
const uint32_t Htarg = ptarget[7];
if (opt_benchmark)
((uint32_t*)ptarget)[7] = 0x0000ff;
swab32_array( endiandata, pdata, 20 );
l2v3_blake256_midstate( endiandata );
do {
be32enc(&endiandata[19], nonce);
lyra2rev3_hash(hash, endiandata);
if (hash[7] <= Htarg )
{
if( fulltest(hash, ptarget) )
{
pdata[19] = nonce;
work_set_target_ratio( work, hash );
*hashes_done = pdata[19] - first_nonce;
return 1;
}
}
nonce++;
} while (nonce < max_nonce && !work_restart[thr_id].restart);
pdata[19] = nonce;
*hashes_done = pdata[19] - first_nonce + 1;
return 0;
}

6
algo/lyra2/lyra2z-4way.c
View File

@@ -1,4 +1,4 @@
#include "lyra2z-gate.h"
#include "lyra2-gate.h"
#ifdef LYRA2Z_4WAY
@@ -36,7 +36,7 @@ void lyra2z_4way_hash( void *state, const void *input )
blake256_4way( &ctx_blake, input + (64*4), 16 );
blake256_4way_close( &ctx_blake, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 256 );
LYRA2Z( lyra2z_4way_matrix, hash0, 32, hash0, 32, hash0, 32, 8, 8, 8 );
LYRA2Z( lyra2z_4way_matrix, hash1, 32, hash1, 32, hash1, 32, 8, 8, 8 );
@@ -70,7 +70,7 @@ int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
for ( int i=0; i < 20; i++ )
be32enc( &edata[i], pdata[i] );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
lyra2z_4way_midstate( vdata );

29
algo/lyra2/lyra2z-gate.c
View File

@@ -1,29 +0,0 @@
#include "lyra2z-gate.h"
#include "lyra2.h"
void lyra2z_set_target( struct work* work, double job_diff )
{
work_set_target( work, job_diff / (256.0 * opt_diff_factor) );
}
bool register_lyra2z_algo( algo_gate_t* gate )
{
#if defined(LYRA2Z_8WAY)
gate->miner_thread_init = (void*)&lyra2z_8way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z_8way;
gate->hash = (void*)&lyra2z_8way_hash;
#elif defined(LYRA2Z_4WAY)
gate->miner_thread_init = (void*)&lyra2z_4way_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z_4way;
gate->hash = (void*)&lyra2z_4way_hash;
#else
gate->miner_thread_init = (void*)&lyra2z_thread_init;
gate->scanhash = (void*)&scanhash_lyra2z;
gate->hash = (void*)&lyra2z_hash;
#endif
gate->optimizations = SSE42_OPT | AVX2_OPT;
gate->get_max64 = (void*)&get_max64_0xffffLL;
gate->set_target = (void*)&lyra2z_set_target;
return true;
};

46
algo/lyra2/lyra2z-gate.h
View File

@@ -1,46 +0,0 @@
#ifndef LYRA2Z_GATE_H__
#define LYRA2Z_GATE_H__ 1
#include "algo-gate-api.h"
#include <stdint.h>
#if defined(__SSE4_2__)
#define LYRA2Z_4WAY
#endif
#if defined(__AVX2__)
// #define LYRA2Z_8WAY
#endif
#define LYRA2Z_MATRIX_SIZE BLOCK_LEN_INT64 * 8 * 8 * 8
#if defined(LYRA2Z_8WAY)
void lyra2z_8way_hash( void *state, const void *input );
int scanhash_lyra2z_8way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2z_8way_thread_init();
#elif defined(LYRA2Z_4WAY)
void lyra2z_4way_hash( void *state, const void *input );
int scanhash_lyra2z_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2z_4way_thread_init();
#else
void lyra2z_hash( void *state, const void *input );
int scanhash_lyra2z( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done );
bool lyra2z_thread_init();
#endif
#endif

2
algo/lyra2/lyra2z.c
View File

@@ -1,6 +1,6 @@
#include <memory.h>
#include <mm_malloc.h>
#include "lyra2z-gate.h"
#include "lyra2-gate.h"
#include "lyra2.h"
#include "algo/blake/sph_blake.h"
#include "avxdefs.h"

133
algo/lyra2/phi2.c Normal file
View File

@@ -0,0 +1,133 @@
/**
* Phi-2 algo Implementation
*/
#include "lyra2-gate.h"
#include "algo/skein/sph_skein.h"
#include "algo/jh/sph_jh.h"
#include "algo/gost/sph_gost.h"
#include "algo/cubehash/cubehash_sse2.h"
#ifdef __AES__
#include "algo/echo/aes_ni/hash_api.h"
#else
#include "algo/echo/sph_echo.h"
#endif
typedef struct {
cubehashParam cube;
sph_jh512_context jh;
#if defined(__AES__)
hashState_echo echo1;
hashState_echo echo2;
#else
sph_echo512_context echo1;
sph_echo512_context echo2;
#endif
sph_gost512_context gost;
sph_skein512_context skein;
} phi2_ctx_holder;
phi2_ctx_holder phi2_ctx;
void init_phi2_ctx()
{
cubehashInit( &phi2_ctx.cube, 512, 16, 32 );
sph_jh512_init(&phi2_ctx.jh);
#if defined(__AES__)
init_echo( &phi2_ctx.echo1, 512 );
init_echo( &phi2_ctx.echo2, 512 );
#else
sph_echo512_init(&phi2_ctx.echo1);
sph_echo512_init(&phi2_ctx.echo2);
#endif
sph_gost512_init(&phi2_ctx.gost);
sph_skein512_init(&phi2_ctx.skein);
};
void phi2_hash(void *state, const void *input)
{
unsigned char _ALIGN(128) hash[64];
unsigned char _ALIGN(128) hashA[64];
unsigned char _ALIGN(128) hashB[64];
phi2_ctx_holder ctx __attribute__ ((aligned (64)));
memcpy( &ctx, &phi2_ctx, sizeof(phi2_ctx) );
cubehashUpdateDigest( &ctx.cube, (byte*)hashB, (const byte*)input,
phi2_has_roots ? 144 : 80 );
LYRA2RE( &hashA[ 0], 32, &hashB[ 0], 32, &hashB[ 0], 32, 1, 8, 8 );
LYRA2RE( &hashA[32], 32, &hashB[32], 32, &hashB[32], 32, 1, 8, 8 );
sph_jh512( &ctx.jh, (const void*)hashA, 64 );
sph_jh512_close( &ctx.jh, (void*)hash );
if ( hash[0] & 1 )
{
sph_gost512( &ctx.gost, (const void*)hash, 64 );
sph_gost512_close( &ctx.gost, (void*)hash );
}
else
{
#if defined(__AES__)
update_final_echo ( &ctx.echo1, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
update_final_echo ( &ctx.echo2, (BitSequence *)hash,
(const BitSequence *)hash, 512 );
#else
sph_echo512( &ctx.echo1, (const void*)hash, 64 );
sph_echo512_close( &ctx.echo1, (void*)hash );
sph_echo512( &ctx.echo2, (const void*)hash, 64 );
sph_echo512_close( &ctx.echo2, (void*)hash );
#endif
}
sph_skein512( &ctx.skein, (const void*)hash, 64 );
sph_skein512_close( &ctx.skein, (void*)hash );
for (int i=0; i<4; i++)
((uint64_t*)hash)[i] ^= ((uint64_t*)hash)[i+4];
memcpy(state, hash, 32);
}
int scanhash_phi2(int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done)
{
uint32_t _ALIGN(128) hash[8];
uint32_t _ALIGN(128) endiandata[36];
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
if(opt_benchmark){
ptarget[7] = 0x00ff;
}
phi2_has_roots = false;
for (int i=0; i < 36; i++) {
be32enc(&endiandata[i], pdata[i]);
if (i >= 20 && pdata[i]) phi2_has_roots = true;
}
do {
be32enc(&endiandata[19], n);
phi2_hash(hash, endiandata);
if (hash[7] < Htarg && fulltest(hash, ptarget)) {
work_set_target_ratio(work, hash);
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 1;
}
n++;
} while (n < max_nonce && !work_restart[thr_id].restart);
*hashes_done = n - first_nonce + 1;
pdata[19] = n;
return 0;
}

32
algo/lyra2/sponge.h
View File

@@ -48,6 +48,10 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
return ( w >> c ) | ( w << ( 64 - c ) );
}
// serial data is only 32 bytes so AVX2 is the limit for that dimension.
// However, 2 way parallel looks trivial to code for AVX512 except for
// a data dependency with rowa.
#if defined __AVX2__
// only available with avx2
@@ -65,13 +69,13 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
#define LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
G_4X64( s0, s1, s2, s3 ); \
s1 = mm256_ror256_1x64( s1); \
s1 = mm256_ror_1x64( s1); \
s2 = mm256_swap_128( s2 ); \
s3 = mm256_rol256_1x64( s3 ); \
s3 = mm256_rol_1x64( s3 ); \
G_4X64( s0, s1, s2, s3 ); \
s1 = mm256_rol256_1x64( s1 ); \
s1 = mm256_rol_1x64( s1 ); \
s2 = mm256_swap_128( s2 ); \
s3 = mm256_ror256_1x64( s3 );
s3 = mm256_ror_1x64( s3 );
#define LYRA_12_ROUNDS_AVX2( s0, s1, s2, s3 ) \
LYRA_ROUND_AVX2( s0, s1, s2, s3 ) \
@@ -93,25 +97,25 @@ static inline uint64_t rotr64( const uint64_t w, const unsigned c ){
// returns void, all args updated
#define G_2X64(a,b,c,d) \
a = _mm_add_epi64( a, b ); \
d = mm_ror_64( _mm_xor_si128( d, a), 32 ); \
d = mm128_ror_64( _mm_xor_si128( d, a), 32 ); \
c = _mm_add_epi64( c, d ); \
b = mm_ror_64( _mm_xor_si128( b, c ), 24 ); \
b = mm128_ror_64( _mm_xor_si128( b, c ), 24 ); \
a = _mm_add_epi64( a, b ); \
d = mm_ror_64( _mm_xor_si128( d, a ), 16 ); \
d = mm128_ror_64( _mm_xor_si128( d, a ), 16 ); \
c = _mm_add_epi64( c, d ); \
b = mm_ror_64( _mm_xor_si128( b, c ), 63 );
b = mm128_ror_64( _mm_xor_si128( b, c ), 63 );
#define LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \
G_2X64( s0, s2, s4, s6 ); \
G_2X64( s1, s3, s5, s7 ); \
mm_ror256_1x64( s2, s3 ); \
mm_swap_128( s4, s5 ); \
mm_rol256_1x64( s6, s7 ); \
mm128_ror256_1x64( s2, s3 ); \
mm128_swap256_128( s4, s5 ); \
mm128_rol256_1x64( s6, s7 ); \
G_2X64( s0, s2, s4, s6 ); \
G_2X64( s1, s3, s5, s7 ); \
mm_rol256_1x64( s2, s3 ); \
mm_swap_128( s4, s5 ); \
mm_ror256_1x64( s6, s7 );
mm128_rol256_1x64( s2, s3 ); \
mm128_swap256_128( s4, s5 ); \
mm128_ror256_1x64( s6, s7 );
#define LYRA_12_ROUNDS_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \
LYRA_ROUND_AVX(s0,s1,s2,s3,s4,s5,s6,s7) \

4
algo/neoscrypt/neoscrypt.c
View File

@@ -1080,6 +1080,8 @@ void neoscrypt_wait_for_diff( struct stratum_ctx *stratum )
}
}
int neoscrypt_get_work_data_size () { return 80; }
bool register_neoscrypt_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT;
@@ -1092,7 +1094,7 @@ bool register_neoscrypt_algo( algo_gate_t* gate )
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;
gate->set_work_data_endian = (void*)&set_work_data_big_endian;
gate->work_data_size = 80;
gate->get_work_data_size = (void*)&neoscrypt_get_work_data_size;
return true;
};

25
algo/nist5/nist5-4way.c
View File

@@ -62,15 +62,15 @@ void nist5hash_4way( void *out, const void *input )
skein512_4way_init( &ctx_skein );
skein512_4way( &ctx_skein, vhash, 64 );
skein512_4way_close( &ctx_skein, vhash );
mm256_deinterleave_4x64( out, out+32, out+64, out+96, vhash, 256 );
skein512_4way_close( &ctx_skein, out );
}
int scanhash_nist5_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done)
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]);
uint32_t lane_hash[8];
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
@@ -120,15 +120,16 @@ int scanhash_nist5_4way( int thr_id, struct work *work, uint32_t max_nonce,
nist5hash_4way( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( ( !( (hash+(i<<3))[7] & mask ) == 0 )
&& fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( ( hash7[ lane ] & mask ) == 0 )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm256_extract_lane_4x64( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )

6
algo/nist5/zr5.c
View File

@@ -219,6 +219,8 @@ void zr5_display_pok( struct work* work )
applog(LOG_BLUE, "POK received: %08xx", work->data[0] );
}
int zr5_get_work_data_size() { return 80; }
bool register_zr5_algo( algo_gate_t* gate )
{
gate->optimizations = SSE2_OPT | AES_OPT;
@@ -227,12 +229,12 @@ bool register_zr5_algo( algo_gate_t* gate )
gate->scanhash = (void*)&scanhash_zr5;
gate->hash = (void*)&zr5hash;
gate->get_max64 = (void*)&zr5_get_max64;
gate->display_extra_data = (void*)&zr5_display_pok;
gate->decode_extra_data = (void*)&zr5_display_pok;
gate->build_stratum_request = (void*)&std_be_build_stratum_request;
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;
gate->set_work_data_endian = (void*)&set_work_data_big_endian;
gate->work_data_size = 80;
gate->get_work_data_size = (void*)&zr5_get_work_data_size;
gate->work_cmp_size = 72;
return true;
};

2
algo/qubit/deep-2way.c
View File

@@ -7,7 +7,7 @@
#include <string.h>
#include <stdio.h>
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/qubit/deep.c
View File

@@ -4,7 +4,7 @@
#include <string.h>
#include <stdio.h>
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#ifndef NO_AES_NI
#include "algo/echo/aes_ni/hash_api.h"
#else

2
algo/qubit/qubit-2way.c
View File

@@ -7,7 +7,7 @@
#include <string.h>
#include <stdio.h>
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/qubit/qubit.c
View File

@@ -4,7 +4,7 @@
#include <string.h>
#include <stdio.h>
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/shavite/sph_shavite.h"
#ifndef NO_AES_NI

5
algo/ripemd/lbry-4way.c
View File

@@ -184,7 +184,8 @@ void lbry_4way_hash( void* output, const void* input )
sha256_4way( &ctx_sha256, vhashA, 32 );
sha256_4way_close( &ctx_sha256, vhashA );
mm_deinterleave_4x32( output, output+32, output+64, output+96, vhashA, 256 );
mm128_deinterleave_4x32( output, output+32, output+64, output+96,
vhashA, 256 );
}
int scanhash_lbry_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -209,7 +210,7 @@ int scanhash_lbry_4way( int thr_id, struct work *work, uint32_t max_nonce,
// we need bigendian data...
swab32_array( edata, pdata, 32 );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 1024 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 1024 );
sha256_4way_init( &sha256_mid );
sha256_4way( &sha256_mid, vdata, LBRY_MIDSTATE );

15
algo/ripemd/lbry-gate.c
View File

@@ -41,8 +41,6 @@ void lbry_le_build_stratum_request( char *req, struct work *work,
free(xnonce2str);
}
// don't use lbry_build_block_header, it can't handle clasim, do it inline
// in lbry_build_extraheader. The side effect is no gbt support for lbry.
void lbry_build_block_header( struct work* g_work, uint32_t version,
uint32_t *prevhash, uint32_t *merkle_root,
uint32_t ntime, uint32_t nbits )
@@ -61,9 +59,6 @@ void lbry_build_block_header( struct work* g_work, uint32_t version,
for ( i = 0; i < 8; i++ )
g_work->data[9 + i] = be32dec( merkle_root + i );
// for ( int i = 0; i < 8; i++ )
// g_work->data[17 + i] = claim[i];
g_work->data[ LBRY_NTIME_INDEX ] = ntime;
g_work->data[ LBRY_NBITS_INDEX ] = nbits;
g_work->data[28] = 0x80000000;
@@ -80,10 +75,6 @@ void lbry_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
for ( t = 0; t < sctx->xnonce2_size && !( ++sctx->job.xnonce2[t] ); t++ );
// Assemble block header
// algo_gate.build_block_header( g_work, le32dec( sctx->job.version ),
// (uint32_t*) sctx->job.prevhash, (uint32_t*) merkle_root,
// le32dec( sctx->job.ntime ), le32dec( sctx->job.nbits ) );
memset( g_work->data, 0, sizeof(g_work->data) );
g_work->data[0] = le32dec( sctx->job.version );
@@ -94,7 +85,7 @@ void lbry_build_extraheader( struct work* g_work, struct stratum_ctx* sctx )
g_work->data[9 + i] = be32dec( (uint32_t *) merkle_root + i );
for ( int i = 0; i < 8; i++ )
g_work->data[17 + i] = ((uint32_t*)sctx->job.claim)[i];
g_work->data[17 + i] = ((uint32_t*)sctx->job.extra)[i];
g_work->data[ LBRY_NTIME_INDEX ] = le32dec(sctx->job.ntime);
g_work->data[ LBRY_NBITS_INDEX ] = le32dec(sctx->job.nbits);
@@ -108,6 +99,8 @@ void lbry_set_target( struct work* work, double job_diff )
int64_t lbry_get_max64() { return 0x1ffffLL; }
int lbry_get_work_data_size() { return LBRY_WORK_DATA_SIZE; }
bool register_lbry_algo( algo_gate_t* gate )
{
gate->optimizations = AVX2_OPT | SHA_OPT;
@@ -130,7 +123,7 @@ bool register_lbry_algo( algo_gate_t* gate )
gate->ntime_index = LBRY_NTIME_INDEX;
gate->nbits_index = LBRY_NBITS_INDEX;
gate->nonce_index = LBRY_NONCE_INDEX;
gate->work_data_size = LBRY_WORK_DATA_SIZE;
gate->get_work_data_size = (void*)&lbry_get_work_data_size;
return true;
}

8
algo/ripemd/ripemd-hash-4way.c
View File

@@ -32,20 +32,20 @@ static const uint32_t IV[5] =
_mm_xor_si128( _mm_and_si128( _mm_xor_si128( y, z ), x ), z )
#define F3(x, y, z) \
_mm_xor_si128( _mm_or_si128( x, mm_not( y ) ), z )
_mm_xor_si128( _mm_or_si128( x, mm128_not( y ) ), z )
#define F4(x, y, z) \
_mm_xor_si128( _mm_and_si128( _mm_xor_si128( x, y ), z ), y )
#define F5(x, y, z) \
_mm_xor_si128( x, _mm_or_si128( y, mm_not( z ) ) )
_mm_xor_si128( x, _mm_or_si128( y, mm128_not( z ) ) )
#define RR(a, b, c, d, e, f, s, r, k) \
do{ \
a = _mm_add_epi32( mm_rol_32( _mm_add_epi32( _mm_add_epi32( \
a = _mm_add_epi32( mm128_rol_32( _mm_add_epi32( _mm_add_epi32( \
_mm_add_epi32( a, f( b ,c, d ) ), r ), \
_mm_set1_epi32( k ) ), s ), e ); \
c = mm_rol_32( c, 10 );\
c = mm128_rol_32( c, 10 );\
} while (0)
#define ROUND1(a, b, c, d, e, f, s, r, k) \

46
algo/sha/sha2-hash-4way.c
View File

@@ -98,19 +98,19 @@ static const sph_u32 K256[64] = {
#define BSG2_0(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_ror_32(x, 2), mm_ror_32(x, 13) ), mm_ror_32( x, 22) )
mm128_ror_32(x, 2), mm128_ror_32(x, 13) ), mm128_ror_32( x, 22) )
#define BSG2_1(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_ror_32(x, 6), mm_ror_32(x, 11) ), mm_ror_32( x, 25) )
mm128_ror_32(x, 6), mm128_ror_32(x, 11) ), mm128_ror_32( x, 25) )
#define SSG2_0(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_ror_32(x, 7), mm_ror_32(x, 18) ), _mm_srli_epi32(x, 3) )
mm128_ror_32(x, 7), mm128_ror_32(x, 18) ), _mm_srli_epi32(x, 3) )
#define SSG2_1(x) \
_mm_xor_si128( _mm_xor_si128( \
mm_ror_32(x, 17), mm_ror_32(x, 19) ), _mm_srli_epi32(x, 10) )
mm128_ror_32(x, 17), mm128_ror_32(x, 19) ), _mm_srli_epi32(x, 10) )
#define SHA2s_4WAY_STEP(A, B, C, D, E, F, G, H, i, j) \
do { \
@@ -129,22 +129,22 @@ sha256_4way_round( __m128i *in, __m128i r[8] )
register __m128i A, B, C, D, E, F, G, H;
__m128i W[16];
W[ 0] = mm_bswap_32( in[ 0] );
W[ 1] = mm_bswap_32( in[ 1] );
W[ 2] = mm_bswap_32( in[ 2] );
W[ 3] = mm_bswap_32( in[ 3] );
W[ 4] = mm_bswap_32( in[ 4] );
W[ 5] = mm_bswap_32( in[ 5] );
W[ 6] = mm_bswap_32( in[ 6] );
W[ 7] = mm_bswap_32( in[ 7] );
W[ 8] = mm_bswap_32( in[ 8] );
W[ 9] = mm_bswap_32( in[ 9] );
W[10] = mm_bswap_32( in[10] );
W[11] = mm_bswap_32( in[11] );
W[12] = mm_bswap_32( in[12] );
W[13] = mm_bswap_32( in[13] );
W[14] = mm_bswap_32( in[14] );
W[15] = mm_bswap_32( in[15] );
W[ 0] = mm128_bswap_32( in[ 0] );
W[ 1] = mm128_bswap_32( in[ 1] );
W[ 2] = mm128_bswap_32( in[ 2] );
W[ 3] = mm128_bswap_32( in[ 3] );
W[ 4] = mm128_bswap_32( in[ 4] );
W[ 5] = mm128_bswap_32( in[ 5] );
W[ 6] = mm128_bswap_32( in[ 6] );
W[ 7] = mm128_bswap_32( in[ 7] );
W[ 8] = mm128_bswap_32( in[ 8] );
W[ 9] = mm128_bswap_32( in[ 9] );
W[10] = mm128_bswap_32( in[10] );
W[11] = mm128_bswap_32( in[11] );
W[12] = mm128_bswap_32( in[12] );
W[13] = mm128_bswap_32( in[13] );
W[14] = mm128_bswap_32( in[14] );
W[15] = mm128_bswap_32( in[15] );
A = r[0];
B = r[1];
@@ -289,13 +289,13 @@ void sha256_4way_close( sha256_4way_context *sc, void *dst )
low = low << 3;
sc->buf[ pad >> 2 ] =
mm_bswap_32( _mm_set1_epi32( high ) );
mm128_bswap_32( _mm_set1_epi32( high ) );
sc->buf[ ( pad+4 ) >> 2 ] =
mm_bswap_32( _mm_set1_epi32( low ) );
mm128_bswap_32( _mm_set1_epi32( low ) );
sha256_4way_round( sc->buf, sc->val );
for ( u = 0; u < 8; u ++ )
((__m128i*)dst)[u] = mm_bswap_32( sc->val[u] );
((__m128i*)dst)[u] = mm128_bswap_32( sc->val[u] );
}
#if defined(__AVX2__)

57
algo/sha/sha256t-4way.c
View File

@@ -4,7 +4,6 @@
#include <string.h>
#include <stdio.h>
#include "sha2-hash-4way.h"
//#include <openssl/sha.h>
#if defined(SHA256T_8WAY)
@@ -25,11 +24,8 @@ void sha256t_8way_hash( void* output, const void* input )
sha256_8way_init( &ctx );
sha256_8way( &ctx, vhash, 32 );
sha256_8way_close( &ctx, vhash );
sha256_8way_close( &ctx, output );
mm256_deinterleave_8x32( output, output+ 32, output+ 64, output+ 96,
output+128, output+160, output+192, output+224,
vhash, 256 );
}
int scanhash_sha256t_8way( int thr_id, struct work *work,
@@ -84,14 +80,22 @@ int scanhash_sha256t_8way( int thr_id, struct work *work,
sha256t_8way_hash( hash, vdata );
for ( int i = 0; i < 8; i++ )
if ( ( !( ( hash+(i<<3) )[7] & mask ) )
&& fulltest( hash+(i<<3), ptarget ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
}
uint32_t *hash7 = &(hash[7<<3]);
for ( int lane = 0; lane < 8; lane++ )
if ( !( hash7[ lane ] & mask ) )
{
// deinterleave hash for lane
uint32_t lane_hash[8];
mm256_extract_lane_8x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 8;
} while ( (num_found == 0) && (n < max_nonce)
@@ -122,10 +126,8 @@ void sha256t_4way_hash( void* output, const void* input )
sha256_4way_init( &ctx );
sha256_4way( &ctx, vhash, 32 );
sha256_4way_close( &ctx, vhash );
sha256_4way_close( &ctx, output );
mm_deinterleave_4x32( output, output+ 32, output+ 64, output+ 96,
vhash, 256 );
}
int scanhash_sha256t_4way( int thr_id, struct work *work,
@@ -133,6 +135,8 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (32)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t lane_hash[8];
uint32_t edata[20] __attribute__ ((aligned (32)));;
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
@@ -159,7 +163,7 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
for ( int k = 0; k < 19; k++ )
be32enc( &edata[k], pdata[k] );
mm_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
mm128_interleave_4x32( vdata, edata, edata, edata, edata, 640 );
sha256_4way_init( &sha256_ctx4 );
sha256_4way( &sha256_ctx4, vdata, 64 );
@@ -175,15 +179,20 @@ int scanhash_sha256t_4way( int thr_id, struct work *work,
sha256t_4way_hash( hash, vdata );
for ( int i = 0; i < 4; i++ )
if ( ( !( ( hash+(i<<3) )[7] & mask ) )
&& fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( !( hash7[ lane ] & mask ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
n += 4;
} while ( (num_found == 0) && (n < max_nonce)
&& !work_restart[thr_id].restart );

4
algo/sha/sha256t-gate.c
View File

@@ -3,16 +3,18 @@
bool register_sha256t_algo( algo_gate_t* gate )
{
#if defined(SHA256T_8WAY)
gate->optimizations = SSE42_OPT | AVX2_OPT;
gate->scanhash = (void*)&scanhash_sha256t_8way;
gate->hash = (void*)&sha256t_8way_hash;
#elif defined(SHA256T_4WAY)
gate->optimizations = SSE42_OPT | AVX2_OPT;
gate->scanhash = (void*)&scanhash_sha256t_4way;
gate->hash = (void*)&sha256t_4way_hash;
#else
gate->optimizations = SSE42_OPT | AVX2_OPT | SHA_OPT;
gate->scanhash = (void*)&scanhash_sha256t;
gate->hash = (void*)&sha256t_hash;
#endif
gate->optimizations = SSE42_OPT | AVX2_OPT | SHA_OPT;
gate->get_max64 = (void*)&get_max64_0x3ffff;
return true;
}

68
algo/shabal/shabal-hash-4way.c
View File

@@ -248,22 +248,22 @@ do { \
*/
#define SWAP_BC \
do { \
mm_swap_128( B0, C0 ); \
mm_swap_128( B1, C1 ); \
mm_swap_128( B2, C2 ); \
mm_swap_128( B3, C3 ); \
mm_swap_128( B4, C4 ); \
mm_swap_128( B5, C5 ); \
mm_swap_128( B6, C6 ); \
mm_swap_128( B7, C7 ); \
mm_swap_128( B8, C8 ); \
mm_swap_128( B9, C9 ); \
mm_swap_128( BA, CA ); \
mm_swap_128( BB, CB ); \
mm_swap_128( BC, CC ); \
mm_swap_128( BD, CD ); \
mm_swap_128( BE, CE ); \
mm_swap_128( BF, CF ); \
mm128_swap256_128( B0, C0 ); \
mm128_swap256_128( B1, C1 ); \
mm128_swap256_128( B2, C2 ); \
mm128_swap256_128( B3, C3 ); \
mm128_swap256_128( B4, C4 ); \
mm128_swap256_128( B5, C5 ); \
mm128_swap256_128( B6, C6 ); \
mm128_swap256_128( B7, C7 ); \
mm128_swap256_128( B8, C8 ); \
mm128_swap256_128( B9, C9 ); \
mm128_swap256_128( BA, CA ); \
mm128_swap256_128( BB, CB ); \
mm128_swap256_128( BC, CC ); \
mm128_swap256_128( BD, CD ); \
mm128_swap256_128( BE, CE ); \
mm128_swap256_128( BF, CF ); \
} while (0)
#define PERM_ELT(xa0, xa1, xb0, xb1, xb2, xb3, xc, xm) \
@@ -271,9 +271,9 @@ do { \
xa0 = _mm_xor_si128( xm, _mm_xor_si128( xb1, _mm_xor_si128( \
_mm_andnot_si128( xb3, xb2 ), \
_mm_mullo_epi32( _mm_xor_si128( xa0, _mm_xor_si128( xc, \
_mm_mullo_epi32( mm_rol_32( xa1, 15 ), _mm_set1_epi32(5UL) ) \
_mm_mullo_epi32( mm128_rol_32( xa1, 15 ), _mm_set1_epi32(5UL) ) \
) ), _mm_set1_epi32(3UL) ) ) ) ); \
xb0 = mm_not( _mm_xor_si128( xa0, mm_rol_32( xb0, 1 ) ) ); \
xb0 = mm128_not( _mm_xor_si128( xa0, mm128_rol_32( xb0, 1 ) ) ); \
} while (0)
#define PERM_STEP_0 do { \
@@ -335,22 +335,22 @@ do { \
#define APPLY_P \
do { \
B0 = mm_ror_32( B0, 15 ); \
B1 = mm_ror_32( B1, 15 ); \
B2 = mm_ror_32( B2, 15 ); \
B3 = mm_ror_32( B3, 15 ); \
B4 = mm_ror_32( B4, 15 ); \
B5 = mm_ror_32( B5, 15 ); \
B6 = mm_ror_32( B6, 15 ); \
B7 = mm_ror_32( B7, 15 ); \
B8 = mm_ror_32( B8, 15 ); \
B9 = mm_ror_32( B9, 15 ); \
BA = mm_ror_32( BA, 15 ); \
BB = mm_ror_32( BB, 15 ); \
BC = mm_ror_32( BC, 15 ); \
BD = mm_ror_32( BD, 15 ); \
BE = mm_ror_32( BE, 15 ); \
BF = mm_ror_32( BF, 15 ); \
B0 = mm128_ror_32( B0, 15 ); \
B1 = mm128_ror_32( B1, 15 ); \
B2 = mm128_ror_32( B2, 15 ); \
B3 = mm128_ror_32( B3, 15 ); \
B4 = mm128_ror_32( B4, 15 ); \
B5 = mm128_ror_32( B5, 15 ); \
B6 = mm128_ror_32( B6, 15 ); \
B7 = mm128_ror_32( B7, 15 ); \
B8 = mm128_ror_32( B8, 15 ); \
B9 = mm128_ror_32( B9, 15 ); \
BA = mm128_ror_32( BA, 15 ); \
BB = mm128_ror_32( BB, 15 ); \
BC = mm128_ror_32( BC, 15 ); \
BD = mm128_ror_32( BD, 15 ); \
BE = mm128_ror_32( BE, 15 ); \
BF = mm128_ror_32( BF, 15 ); \
PERM_STEP_0; \
PERM_STEP_1; \
PERM_STEP_2; \

84
algo/shavite/sph-shavite-aesni.c
View File

@@ -64,11 +64,11 @@ static const sph_u32 IV512[] = {
// a[3:0] = { b[0], a[3], a[2], a[1] }
#if defined(__SSSE3__)
#define mm_ror256hi_1x32( a, b ) _mm_alignr_epi8( b, a, 4 )
#define mm128_ror256hi_1x32( a, b ) _mm_alignr_epi8( b, a, 4 )
#else // SSE2
#define mm_ror256hi_1x32( a, b ) \
#define mm128_ror256hi_1x32( a, b ) \
_mm_or_si128( _mm_srli_si128( a, 4 ), \
_mm_slli_si128( b, 12 ) )
@@ -136,7 +136,7 @@ c512( sph_shavite_big_context *sc, const void *msg )
for ( r = 0; r < 3; r ++ )
{
// round 1, 5, 9
k00 = mm_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
k00 = mm128_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
k00 = _mm_xor_si128( k00, k13 );
if ( r == 0 )
@@ -145,7 +145,7 @@ c512( sph_shavite_big_context *sc, const void *msg )
x = _mm_xor_si128( p0, k00 );
x = _mm_aesenc_si128( x, m128_zero );
k01 = mm_ror_1x32( _mm_aesenc_si128( k01, m128_zero ) );
k01 = mm128_ror_1x32( _mm_aesenc_si128( k01, m128_zero ) );
k01 = _mm_xor_si128( k01, k00 );
if ( r == 1 )
@@ -154,33 +154,33 @@ c512( sph_shavite_big_context *sc, const void *msg )
x = _mm_xor_si128( x, k01 );
x = _mm_aesenc_si128( x, m128_zero );
k02 = mm_ror_1x32( _mm_aesenc_si128( k02, 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 = mm_ror_1x32( _mm_aesenc_si128( k03, 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 = mm_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
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 = mm_ror_1x32( _mm_aesenc_si128( k11, 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 = mm_ror_1x32( _mm_aesenc_si128( k12, 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 = mm_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
if ( r == 2 )
@@ -193,80 +193,80 @@ c512( sph_shavite_big_context *sc, const void *msg )
// round 2, 6, 10
k00 = _mm_xor_si128( k00, mm_ror256hi_1x32( k12, k13 ) );
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, mm_ror256hi_1x32( k13, k00 ) );
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, mm_ror256hi_1x32( k00, k01 ) );
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, mm_ror256hi_1x32( k01, k02 ) );
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, mm_ror256hi_1x32( k02, k03 ) );
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, mm_ror256hi_1x32( k03, k10 ) );
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, mm_ror256hi_1x32( k10, k11 ) );
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, mm_ror256hi_1x32( k11, k12 ) );
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 = mm_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
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 = mm_ror_1x32( _mm_aesenc_si128( k01, 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 = mm_ror_1x32( _mm_aesenc_si128( k02, 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 = mm_ror_1x32( _mm_aesenc_si128( k03, 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 = mm_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
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 = mm_ror_1x32( _mm_aesenc_si128( k11, 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 = mm_ror_1x32( _mm_aesenc_si128( k12, 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 = mm_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
x = _mm_xor_si128( x, k13 );
@@ -275,36 +275,36 @@ c512( sph_shavite_big_context *sc, const void *msg )
// round 4, 8, 12
k00 = _mm_xor_si128( k00, mm_ror256hi_1x32( k12, k13 ) );
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, mm_ror256hi_1x32( k13, k00 ) );
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, mm_ror256hi_1x32( k00, k01 ) );
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, mm_ror256hi_1x32( k01, k02 ) );
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, mm_ror256hi_1x32( k02, k03 ) );
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, mm_ror256hi_1x32( k03, k10 ) );
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, mm_ror256hi_1x32( k10, k11 ) );
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, mm_ror256hi_1x32( k11, k12 ) );
k13 = _mm_xor_si128( k13, mm128_ror256hi_1x32( k11, k12 ) );
x = _mm_xor_si128( x, k13 );
x = _mm_aesenc_si128( x, m128_zero );
@@ -313,44 +313,44 @@ c512( sph_shavite_big_context *sc, const void *msg )
// round 13
k00 = mm_ror_1x32( _mm_aesenc_si128( k00, m128_zero ) );
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 = mm_ror_1x32( _mm_aesenc_si128( k01, 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 = mm_ror_1x32( _mm_aesenc_si128( k02, 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 = mm_ror_1x32( _mm_aesenc_si128( k03, 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 = mm_ror_1x32( _mm_aesenc_si128( k10, m128_zero ) );
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 = mm_ror_1x32( _mm_aesenc_si128( k11, 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 = mm_ror_1x32( _mm_aesenc_si128( k12, 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 = mm_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = mm128_ror_1x32( _mm_aesenc_si128( k13, m128_zero ) );
k13 = _mm_xor_si128( k13, k12 );
x = _mm_xor_si128( x, k13 );

46
algo/simd/simd-hash-2way.c
View File

@@ -198,13 +198,13 @@ do { \
#undef BUTTERFLY_N
// Multiply by twiddle factors
X(6) = _mm256_mullo_epi16( X(6), FFT64_Twiddle[0].m256i );
X(5) = _mm256_mullo_epi16( X(5), FFT64_Twiddle[1].m256i );
X(4) = _mm256_mullo_epi16( X(4), FFT64_Twiddle[2].m256i );
X(3) = _mm256_mullo_epi16( X(3), FFT64_Twiddle[3].m256i );
X(2) = _mm256_mullo_epi16( X(2), FFT64_Twiddle[4].m256i );
X(1) = _mm256_mullo_epi16( X(1), FFT64_Twiddle[5].m256i );
X(0) = _mm256_mullo_epi16( X(0), FFT64_Twiddle[6].m256i );
X(6) = _mm256_mullo_epi16( X(6), FFT64_Twiddle[0].v256 );
X(5) = _mm256_mullo_epi16( X(5), FFT64_Twiddle[1].v256 );
X(4) = _mm256_mullo_epi16( X(4), FFT64_Twiddle[2].v256 );
X(3) = _mm256_mullo_epi16( X(3), FFT64_Twiddle[3].v256 );
X(2) = _mm256_mullo_epi16( X(2), FFT64_Twiddle[4].v256 );
X(1) = _mm256_mullo_epi16( X(1), FFT64_Twiddle[5].v256 );
X(0) = _mm256_mullo_epi16( X(0), FFT64_Twiddle[6].v256 );
// Transpose the FFT state with a revbin order permutation
// on the rows and the column.
@@ -319,7 +319,7 @@ void fft128_2way( void *a )
B[ i ] = REDUCE_FULL_S( B[ i ] );
A[ i+8 ] = _mm256_sub_epi16( A[ i ], A[ i+8 ] );
A[ i+8 ] = REDUCE_FULL_S( A[ i+8 ] );
A[ i+8 ] = _mm256_mullo_epi16( A[ i+8 ], FFT128_Twiddle[i].m256i );
A[ i+8 ] = _mm256_mullo_epi16( A[ i+8 ], FFT128_Twiddle[i].v256 );
A[ i+8 ] = REDUCE_FULL_S( A[ i+8 ] );
}
@@ -347,10 +347,10 @@ void fft128_2way_msg( uint16_t *a, const uint8_t *x, int final )
do { \
__m256i t = X[i]; \
A[2*i] = _mm256_unpacklo_epi8( t, m256_zero ); \
A[2*i+8] = _mm256_mullo_epi16( A[2*i], FFT128_Twiddle[2*i].m256i ); \
A[2*i+8] = _mm256_mullo_epi16( A[2*i], FFT128_Twiddle[2*i].v256 ); \
A[2*i+8] = REDUCE(A[2*i+8]); \
A[2*i+1] = _mm256_unpackhi_epi8( t, m256_zero ); \
A[2*i+9] = _mm256_mullo_epi16(A[2*i+1], FFT128_Twiddle[2*i+1].m256i ); \
A[2*i+9] = _mm256_mullo_epi16(A[2*i+1], FFT128_Twiddle[2*i+1].v256 ); \
A[2*i+9] = REDUCE(A[2*i+9]); \
} while(0)
@@ -360,12 +360,12 @@ do { \
__m256i t = X[i]; \
__m256i tmp; \
A[2*i] = _mm256_unpacklo_epi8( t, m256_zero ); \
A[2*i+8] = _mm256_mullo_epi16( A[ 2*i ], FFT128_Twiddle[ 2*i ].m256i ); \
A[2*i+8] = _mm256_mullo_epi16( A[ 2*i ], FFT128_Twiddle[ 2*i ].v256 ); \
A[2*i+8] = REDUCE( A[ 2*i+8 ] ); \
tmp = _mm256_unpackhi_epi8( t, m256_zero ); \
A[2*i+1] = _mm256_add_epi16( tmp, tw ); \
A[2*i+9] = _mm256_mullo_epi16( _mm256_sub_epi16( tmp, tw ), \
FFT128_Twiddle[ 2*i+1 ].m256i );\
FFT128_Twiddle[ 2*i+1 ].v256 );\
A[2*i+9] = REDUCE( A[ 2*i+9 ] ); \
} while(0)
@@ -373,9 +373,9 @@ do { \
UNPACK( 1 );
UNPACK( 2 );
if ( final )
UNPACK_TWEAK( 3, FinalTweak.m256i );
UNPACK_TWEAK( 3, FinalTweak.v256 );
else
UNPACK_TWEAK( 3, Tweak.m256i );
UNPACK_TWEAK( 3, Tweak.v256 );
#undef UNPACK
#undef UNPACK_TWEAK
@@ -398,11 +398,11 @@ do { \
__m256i t = X[i]; \
A[ 2*i ] = _mm256_unpacklo_epi8( t, m256_zero ); \
A[ 2*i + 16 ] = _mm256_mullo_epi16( A[ 2*i ], \
FFT256_Twiddle[ 2*i ].m256i ); \
FFT256_Twiddle[ 2*i ].v256 ); \
A[ 2*i + 16 ] = REDUCE( A[ 2*i + 16 ] ); \
A[ 2*i + 1 ] = _mm256_unpackhi_epi8( t, m256_zero ); \
A[ 2*i + 17 ] = _mm256_mullo_epi16( A[ 2*i + 1 ], \
FFT256_Twiddle[ 2*i + 1 ].m256i ); \
FFT256_Twiddle[ 2*i + 1 ].v256 ); \
A[ 2*i + 17 ] = REDUCE( A[ 2*i + 17 ] ); \
} while(0)
@@ -413,12 +413,12 @@ do { \
__m256i tmp; \
A[ 2*i ] = _mm256_unpacklo_epi8( t, m256_zero ); \
A[ 2*i + 16 ] = _mm256_mullo_epi16( A[ 2*i ], \
FFT256_Twiddle[ 2*i ].m256i ); \
FFT256_Twiddle[ 2*i ].v256 ); \
A[ 2*i + 16 ] = REDUCE( A[ 2*i + 16 ] ); \
tmp = _mm256_unpackhi_epi8( t, m256_zero ); \
A[ 2*i + 1 ] = _mm256_add_epi16( tmp, tw ); \
A[ 2*i + 17 ] = _mm256_mullo_epi16( _mm256_sub_epi16( tmp, tw ), \
FFT256_Twiddle[ 2*i + 1 ].m256i ); \
FFT256_Twiddle[ 2*i + 1 ].v256 ); \
} while(0)
UNPACK( 0 );
@@ -429,9 +429,9 @@ do { \
UNPACK( 5 );
UNPACK( 6 );
if ( final )
UNPACK_TWEAK( 7, FinalTweak.m256i );
UNPACK_TWEAK( 7, FinalTweak.v256 );
else
UNPACK_TWEAK( 7, Tweak.m256i );
UNPACK_TWEAK( 7, Tweak.v256 );
#undef UNPACK
#undef UNPACK_TWEAK
@@ -447,7 +447,7 @@ void rounds512_2way( uint32_t *state, const uint8_t *msg, uint16_t *fft )
__m256i *S = (__m256i*) state;
__m256i *M = (__m256i*) msg;
__m256i *W = (__m256i*) fft;
static const m256_v16 code[] = { mm256_setc1_16(185), mm256_setc1_16(233) };
static const m256_v16 code[] = { mm256_const1_16(185), mm256_const1_16(233) };
S0l = _mm256_xor_si256( S[0], M[0] );
S0h = _mm256_xor_si256( S[1], M[1] );
@@ -612,9 +612,9 @@ do { \
int a = MSG_##u(hh); \
int b = MSG_##u(ll); \
w##l = _mm256_unpacklo_epi16( W[a], W[b] ); \
w##l = _mm256_mullo_epi16( w##l, code[z].m256i ); \
w##l = _mm256_mullo_epi16( w##l, code[z].v256 ); \
w##h = _mm256_unpackhi_epi16( W[a], W[b]) ; \
w##h = _mm256_mullo_epi16( w##h, code[z].m256i ); \
w##h = _mm256_mullo_epi16( w##h, code[z].v256 ); \
} while(0)
#define ROUND( h0,l0,u0,h1,l1,u1,h2,l2,u2,h3,l3,u3,fun,r,s,t,u,z ) \

21
algo/skein/skein-4way.c
View File

@@ -21,9 +21,10 @@ void skeinhash_4way( void *state, const void *input )
sha256_4way_init( &ctx_sha256 );
sha256_4way( &ctx_sha256, vhash32, 64 );
sha256_4way_close( &ctx_sha256, vhash32 );
sha256_4way_close( &ctx_sha256, state );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash32, 256 );
mm128_deinterleave_4x32( state, state+32, state+64, state+96,
vhash32, 256 );
}
int scanhash_skein_4way( int thr_id, struct work *work, uint32_t max_nonce,
@@ -31,6 +32,8 @@ int scanhash_skein_4way( int thr_id, struct work *work, uint32_t max_nonce,
{
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t lane_hash[8];
uint32_t *hash7 = &(hash[7<<2]);
uint32_t edata[20] __attribute__ ((aligned (64)));
uint32_t *pdata = work->data;
uint32_t *ptarget = work->target;
@@ -58,12 +61,16 @@ int scanhash_skein_4way( int thr_id, struct work *work, uint32_t max_nonce,
skeinhash_4way( hash, vdata );
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( hash7[ lane ] <= Htarg )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( (num_found == 0) && (n < max_nonce)

23
algo/skein/skein2-4way.c
View File

@@ -9,7 +9,6 @@ void skein2hash_4way( void *output, const void *input )
{
skein512_4way_context ctx;
uint64_t hash[8*4] __attribute__ ((aligned (64)));
uint64_t *out64 = (uint64_t*)output;
skein512_4way_init( &ctx );
skein512_4way( &ctx, input, 80 );
@@ -17,15 +16,14 @@ void skein2hash_4way( void *output, const void *input )
skein512_4way_init( &ctx );
skein512_4way( &ctx, hash, 64 );
skein512_4way_close( &ctx, hash );
mm256_deinterleave_4x64( out64, out64+4, out64+8, out64+12, hash, 256 );
skein512_4way_close( &ctx, output );
}
int scanhash_skein2_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[8*4] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[25]);
uint32_t vdata[20*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__ ((aligned (64)));
uint64_t *edata = (uint64_t*)endiandata;
@@ -34,7 +32,6 @@ int scanhash_skein2_4way( int thr_id, struct work *work, uint32_t max_nonce,
const uint32_t Htarg = ptarget[7];
const uint32_t first_nonce = pdata[19];
uint32_t n = first_nonce;
// hash is returned deinterleaved
uint32_t *nonces = work->nonces;
int num_found = 0;
@@ -53,12 +50,18 @@ int scanhash_skein2_4way( int thr_id, struct work *work, uint32_t max_nonce,
skein2hash( hash, vdata );
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( hash7[ lane ] <= Htarg )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
// deinterleave hash for lane
uint32_t lane_hash[8];
mm256_extract_lane_4x64( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( (num_found == 0) && (n < max_nonce)

40
algo/sm3/sm3-hash-4way.c
View File

@@ -125,20 +125,20 @@ void sm3_4way_close( void *cc, void *dst )
memset_zero_128( block, ( SM3_BLOCK_SIZE - 8 ) >> 2 );
}
count[0] = mm_bswap_32(
count[0] = mm128_bswap_32(
_mm_set1_epi32( ctx->nblocks >> 23 ) );
count[1] = mm_bswap_32( _mm_set1_epi32( ( ctx->nblocks << 9 ) +
count[1] = mm128_bswap_32( _mm_set1_epi32( ( ctx->nblocks << 9 ) +
( ctx->num << 3 ) ) );
sm3_4way_compress( ctx->digest, block );
for ( i = 0; i < 8 ; i++ )
hash[i] = mm_bswap_32( ctx->digest[i] );
hash[i] = mm128_bswap_32( ctx->digest[i] );
}
#define P0(x) _mm_xor_si128( x, _mm_xor_si128( mm_rol_32( x, 9 ), \
mm_rol_32( x, 17 ) ) )
#define P1(x) _mm_xor_si128( x, _mm_xor_si128( mm_rol_32( x, 15 ), \
mm_rol_32( x, 23 ) ) )
#define P0(x) _mm_xor_si128( x, _mm_xor_si128( mm128_rol_32( x, 9 ), \
mm128_rol_32( x, 17 ) ) )
#define P1(x) _mm_xor_si128( x, _mm_xor_si128( mm128_rol_32( x, 15 ), \
mm128_rol_32( x, 23 ) ) )
#define FF0(x,y,z) _mm_xor_si128( x, _mm_xor_si128( y, z ) )
#define FF1(x,y,z) _mm_or_si128( _mm_or_si128( _mm_and_si128( x, y ), \
@@ -165,13 +165,13 @@ void sm3_4way_compress( __m128i *digest, __m128i *block )
int j;
for ( j = 0; j < 16; j++ )
W[j] = mm_bswap_32( block[j] );
W[j] = mm128_bswap_32( block[j] );
for ( j = 16; j < 68; j++ )
W[j] = _mm_xor_si128( P1( _mm_xor_si128( _mm_xor_si128( W[ j-16 ],
W[ j-9 ] ),
mm_rol_32( W[ j-3 ], 15 ) ) ),
_mm_xor_si128( mm_rol_32( W[ j-13 ], 7 ),
mm128_rol_32( W[ j-3 ], 15 ) ) ),
_mm_xor_si128( mm128_rol_32( W[ j-13 ], 7 ),
W[ j-6 ] ) );
for( j = 0; j < 64; j++ )
@@ -180,19 +180,19 @@ void sm3_4way_compress( __m128i *digest, __m128i *block )
T = _mm_set1_epi32( 0x79CC4519UL );
for( j =0; j < 16; j++ )
{
SS1 = mm_rol_32( _mm_add_epi32( _mm_add_epi32( mm_rol_32( A, 12 ), E ),
mm_rol_32( T, j ) ), 7 );
SS2 = _mm_xor_si128( SS1, mm_rol_32( A, 12 ) );
SS1 = mm128_rol_32( _mm_add_epi32( _mm_add_epi32( mm128_rol_32(A,12), E ),
mm128_rol_32( T, j ) ), 7 );
SS2 = _mm_xor_si128( SS1, mm128_rol_32( A, 12 ) );
TT1 = _mm_add_epi32( _mm_add_epi32( _mm_add_epi32( FF0( A, B, C ), D ),
SS2 ), W1[j] );
TT2 = _mm_add_epi32( _mm_add_epi32( _mm_add_epi32( GG0( E, F, G ), H ),
SS1 ), W[j] );
D = C;
C = mm_rol_32( B, 9 );
C = mm128_rol_32( B, 9 );
B = A;
A = TT1;
H = G;
G = mm_rol_32( F, 19 );
G = mm128_rol_32( F, 19 );
F = E;
E = P0( TT2 );
}
@@ -200,19 +200,19 @@ void sm3_4way_compress( __m128i *digest, __m128i *block )
T = _mm_set1_epi32( 0x7A879D8AUL );
for( j =16; j < 64; j++ )
{
SS1 = mm_rol_32( _mm_add_epi32( _mm_add_epi32( mm_rol_32( A, 12 ), E ),
mm_rol_32( T, j&31 ) ), 7 );
SS2 = _mm_xor_si128( SS1, mm_rol_32( A, 12 ) );
SS1 = mm128_rol_32( _mm_add_epi32( _mm_add_epi32( mm128_rol_32(A,12), E ),
mm128_rol_32( T, j&31 ) ), 7 );
SS2 = _mm_xor_si128( SS1, mm128_rol_32( A, 12 ) );
TT1 = _mm_add_epi32( _mm_add_epi32( _mm_add_epi32( FF1( A, B, C ), D ),
SS2 ), W1[j] );
TT2 = _mm_add_epi32( _mm_add_epi32( _mm_add_epi32( GG1( E, F, G ), H ),
SS1 ), W[j] );
D = C;
C = mm_rol_32( B, 9 );
C = mm128_rol_32( B, 9 );
B = A;
A = TT1;
H = G;
G = mm_rol_32( F, 19 );
G = mm128_rol_32( F, 19 );
F = E;
E = P0( TT2 );
}

2
algo/x11/c11-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/x11/c11.c
View File

@@ -23,7 +23,7 @@
#endif
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/keccak/sse2/keccak.c"

2
algo/x11/timetravel-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
static __thread uint32_t s_ntime = UINT32_MAX;
static __thread int permutation[TT8_FUNC_COUNT] = { 0 };

2
algo/x11/timetravel.c
View File

@@ -10,7 +10,7 @@
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#ifdef NO_AES_NI
#include "algo/groestl/sph_groestl.h"
#else

2
algo/x11/timetravel10-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"

2
algo/x11/timetravel10.c
View File

@@ -9,7 +9,7 @@
#include "algo/keccak/sph_keccak.h"
#include "algo/skein/sph_skein.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/nist.h"

2
algo/x11/x11-4way.c
View File

@@ -12,7 +12,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/x11/x11.c
View File

@@ -20,7 +20,7 @@
#endif
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/keccak/sse2/keccak.c"

2
algo/x11/x11evo-4way.c
View File

@@ -15,7 +15,7 @@
#include "algo/groestl/aes_ni/hash-groestl.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/simd-hash-2way.h"
typedef struct {

2
algo/x11/x11evo.c
View File

@@ -23,7 +23,7 @@
#endif
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
typedef struct {

2
algo/x11/x11gost-4way.c
View File

@@ -14,7 +14,7 @@
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/gost/sph_gost.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/x11/x11gost.c
View File

@@ -11,7 +11,7 @@
#include "algo/echo/sph_echo.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/keccak/sse2/keccak.c"

2
algo/x12/x12-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/x12/x12.c
View File

@@ -20,7 +20,7 @@
//#include "algo/fugue/sph_fugue.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"

6
algo/x13/drop.c
View File

@@ -238,6 +238,8 @@ void drop_display_pok( struct work* work )
applog(LOG_BLUE, "POK received: %08xx", work->data[0] );
}
int drop_get_work_data_size() { return 80; }
// Need to fix POK offset problems like zr5
bool register_drop_algo( algo_gate_t* gate )
{
@@ -250,8 +252,8 @@ bool register_drop_algo( algo_gate_t* gate )
gate->work_decode = (void*)&std_be_work_decode;
gate->submit_getwork_result = (void*)&std_be_submit_getwork_result;
gate->set_work_data_endian = (void*)&set_work_data_big_endian;
gate->display_extra_data = (void*)&drop_display_pok;
gate->work_data_size = 80;
gate->decode_extra_data = (void*)&drop_display_pok;
gate->get_work_data_size = (void*)&drop_get_work_data_size;
gate->work_cmp_size = 72;
return true;
};

2
algo/x13/phi1612-4way.c
View File

@@ -8,7 +8,7 @@
#include <stdio.h>
#include "algo/skein/skein-hash-4way.h"
#include "algo/jh/jh-hash-4way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/gost/sph_gost.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/x13/phi1612.c
View File

@@ -8,7 +8,7 @@
#include "algo/gost/sph_gost.h"
#include "algo/echo/sph_echo.h"
#include "algo/fugue//sph_fugue.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/skein/sse2/skein.c"
#include "algo/jh/sph_jh.h"

2
algo/x13/skunk-4way.c
View File

@@ -9,7 +9,7 @@
#include "algo/skein/skein-hash-4way.h"
#include "algo/gost/sph_gost.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
typedef struct {
skein512_4way_context skein;

2
algo/x13/skunk.c
View File

@@ -6,7 +6,7 @@
#include "algo/gost/sph_gost.h"
#include "algo/skein/sph_skein.h"
#include "algo/fugue/sph_fugue.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
typedef struct {
sph_skein512_context skein;

2
algo/x13/x13-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"

2
algo/x13/x13.c
View File

@@ -20,7 +20,7 @@
#include "algo/fugue/sph_fugue.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"

6
algo/x13/x13sm3-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
@@ -166,7 +166,7 @@ void x13sm3_4way_hash( void *state, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
// SM3 parallel 32 bit
uint32_t sm3_vhash[32*4] __attribute__ ((aligned (64)));
@@ -182,7 +182,7 @@ void x13sm3_4way_hash( void *state, const void *input )
sm3_4way( &ctx.sm3, vhash, 64 );
sm3_4way_close( &ctx.sm3, sm3_vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, sm3_vhash, 512 );
// Hamsi parallel 4x32x2
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );

3
algo/x13/x13sm3.c
View File

@@ -16,9 +16,8 @@
#include "algo/sm3/sph_sm3.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/echo/sse2/sph_echo.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"
#include "algo/keccak/sse2/keccak.c"

2
algo/x14/polytimos-4way.c
View File

@@ -52,7 +52,7 @@ void polytimos_4way_hash( void *output, const void *input )
mm256_reinterleave_4x32( vhash32, vhash, 512 );
shabal512_4way( &ctx.shabal, vhash32, 64 );
shabal512_4way_close( &ctx.shabal, vhash32 );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash32, 512 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash32, 512 );
update_final_echo ( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *)hash0, 512 );

4
algo/x14/veltor-4way.c
View File

@@ -54,10 +54,10 @@ void veltor_4way_hash( void *output, const void *input )
sph_shavite512( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
sph_gost512( &ctx.gost, hash0, 64 );
sph_gost512_close( &ctx.gost, hash0 );

35
algo/x14/x14-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
@@ -183,21 +183,16 @@ void x14_4way_hash( void *state, const void *input )
sph_fugue512_close( &ctx.fugue, hash3 );
// 14 Shabal, parallel 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
memcpy( state, hash0, 32 );
memcpy( state+32, hash1, 32 );
memcpy( state+64, hash2, 32 );
memcpy( state+96, hash3, 32 );
shabal512_4way_close( &ctx.shabal, state );
}
int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t hash[4*16] __attribute__ ((aligned (64)));
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
@@ -233,13 +228,21 @@ int scanhash_x14_4way( int thr_id, struct work *work, uint32_t max_nonce,
x14_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( ( ( (hash+(i<<3))[7] & mask ) == 0 )
&& fulltest( hash+(i<<3), ptarget ) )
uint32_t *hash7 = &(hash[7<<2]);
for ( int lane = 0; lane < 4; lane++ )
if ( ( hash7[ lane ] & mask ) == 0 )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
// deinterleave hash for lane
uint32_t lane_hash[8];
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )

3
algo/x14/x14.c
View File

@@ -21,9 +21,8 @@
#include "algo/shabal/sph_shabal.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/echo/sse2/sph_echo.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"
#include "algo/keccak/sse2/keccak.c"

6
algo/x15/x15-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
@@ -186,10 +186,10 @@ void x15_4way_hash( void *state, const void *input )
sph_fugue512_close( &ctx.fugue, hash3 );
// 14 Shabal, parallel 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
// 15 Whirlpool
sph_whirlpool( &ctx.whirlpool, hash0, 64 );

2
algo/x15/x15.c
View File

@@ -22,7 +22,7 @@
#include "algo/whirlpool/sph_whirlpool.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"

2
algo/x17/hmq1725.c
View File

@@ -24,7 +24,7 @@
#include "algo/echo/aes_ni/hash_api.h"
#endif
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/jh/sse2/jh_sse2_opt64.h"

6
algo/x17/x16r-4way.c
View File

@@ -20,7 +20,7 @@
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
@@ -256,11 +256,11 @@ void x16r_4way_hash( void* output, const void* input )
sph_fugue512_close( &ctx.fugue, hash3 );
break;
case SHABAL:
mm_interleave_4x32( vhash, in0, in1, in2, in3, size<<3 );
mm128_interleave_4x32( vhash, in0, in1, in2, in3, size<<3 );
shabal512_4way_init( &ctx.shabal );
shabal512_4way( &ctx.shabal, vhash, size );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
break;
case WHIRLPOOL:
sph_whirlpool_init( &ctx.whirlpool );

2
algo/x17/x16r.c
View File

@@ -17,7 +17,7 @@
#include "algo/skein/sph_skein.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/echo/sph_echo.h"
#include "algo/hamsi/sph_hamsi.h"

79
algo/x17/x17-4way.c
View File

@@ -13,7 +13,7 @@
#include "algo/jh/jh-hash-4way.h"
#include "algo/keccak/keccak-hash-4way.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cube-hash-2way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
@@ -32,7 +32,7 @@ typedef struct {
jh512_4way_context jh;
keccak512_4way_context keccak;
luffa_2way_context luffa;
cubehashParam cube;
cube_2way_context cube;
sph_shavite512_context shavite;
simd_2way_context simd;
hashState_echo echo;
@@ -55,7 +55,7 @@ void init_x17_4way_ctx()
jh512_4way_init( &x17_4way_ctx.jh );
keccak512_4way_init( &x17_4way_ctx.keccak );
luffa_2way_init( &x17_4way_ctx.luffa, 512 );
cubehashInit( &x17_4way_ctx.cube, 512, 16, 32 );
cube_2way_init( &x17_4way_ctx.cube, 512, 16, 32 );
sph_shavite512_init( &x17_4way_ctx.shavite );
simd_2way_init( &x17_4way_ctx.simd, 512 );
init_echo( &x17_4way_ctx.echo, 512 );
@@ -73,11 +73,11 @@ void x17_4way_hash( void *state, const void *input )
uint64_t hash2[8] __attribute__ ((aligned (64)));
uint64_t hash3[8] __attribute__ ((aligned (64)));
uint64_t vhash[8*4] __attribute__ ((aligned (64)));
uint64_t vhash32[8*4] __attribute__ ((aligned (64)));
uint64_t vhashB[8*4] __attribute__ ((aligned (64)));
x17_4way_ctx_holder ctx;
memcpy( &ctx, &x17_4way_ctx, sizeof(x17_4way_ctx) );
// 1 Blake
// 1 Blake 4 way 64 bit
blake512_4way( &ctx.blake, input, 80 );
blake512_4way_close( &ctx.blake, vhash );
@@ -114,25 +114,22 @@ void x17_4way_hash( void *state, const void *input )
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 7 Luffa
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
// 7 Luffa parallel 2 way
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
mm256_interleave_2x128( vhashB, hash2, hash3, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_interleave_2x128( vhash, hash2, hash3, 512 );
luffa_2way_init( &ctx.luffa, 512 );
luffa_2way_update_close( &ctx.luffa, vhash, vhash, 64 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
luffa_2way_update_close( &ctx.luffa, vhashB, vhashB, 64 );
// 8 Cubehash
cubehashUpdateDigest( &ctx.cube, (byte*)hash0, (const byte*) hash0, 64 );
memcpy( &ctx.cube, &x17_4way_ctx.cube, sizeof(cubehashParam) );
cubehashUpdateDigest( &ctx.cube, (byte*)hash1, (const byte*) hash1, 64 );
memcpy( &ctx.cube, &x17_4way_ctx.cube, sizeof(cubehashParam) );
cubehashUpdateDigest( &ctx.cube, (byte*)hash2, (const byte*) hash2, 64 );
memcpy( &ctx.cube, &x17_4way_ctx.cube, sizeof(cubehashParam) );
cubehashUpdateDigest( &ctx.cube, (byte*)hash3, (const byte*) hash3, 64 );
// 8 Cubehash parallel 2 way
cube_2way_update_close( &ctx.cube, vhash, vhash, 64 );
cube_2way_reinit( &ctx.cube );
cube_2way_update_close( &ctx.cube, vhashB, vhashB, 64 );
// 9 Shavite
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhashB, 512 );
// 9 Shavite serial
sph_shavite512( &ctx.shavite, hash0, 64 );
sph_shavite512_close( &ctx.shavite, hash0 );
memcpy( &ctx.shavite, &x17_4way_ctx.shavite,
@@ -148,7 +145,7 @@ void x17_4way_hash( void *state, const void *input )
sph_shavite512( &ctx.shavite, hash3, 64 );
sph_shavite512_close( &ctx.shavite, hash3 );
// 10 Simd
// 10 Simd parallel 2 way 128 bit
mm256_interleave_2x128( vhash, hash0, hash1, 512 );
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash0, hash1, vhash, 512 );
@@ -157,7 +154,7 @@ void x17_4way_hash( void *state, const void *input )
simd_2way_update_close( &ctx.simd, vhash, vhash, 512 );
mm256_deinterleave_2x128( hash2, hash3, vhash, 512 );
// 11 Echo
// 11 Echo serial
update_final_echo( &ctx.echo, (BitSequence *)hash0,
(const BitSequence *) hash0, 512 );
memcpy( &ctx.echo, &x17_4way_ctx.echo, sizeof(hashState_echo) );
@@ -170,13 +167,13 @@ void x17_4way_hash( void *state, const void *input )
update_final_echo( &ctx.echo, (BitSequence *)hash3,
(const BitSequence *) hash3, 512 );
// 12 Hamsi
// 12 Hamsi parallel 4 way 64 bit
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, 512 );
hamsi512_4way( &ctx.hamsi, vhash, 64 );
hamsi512_4way_close( &ctx.hamsi, vhash );
mm256_deinterleave_4x64( hash0, hash1, hash2, hash3, vhash, 512 );
// 13 Fugue
// 13 Fugue serial
sph_fugue512( &ctx.fugue, hash0, 64 );
sph_fugue512_close( &ctx.fugue, hash0 );
memcpy( &ctx.fugue, &x17_4way_ctx.fugue, sizeof(sph_fugue512_context) );
@@ -189,11 +186,11 @@ void x17_4way_hash( void *state, const void *input )
sph_fugue512( &ctx.fugue, hash3, 64 );
sph_fugue512_close( &ctx.fugue, hash3 );
// 14 Shabal, parallel 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
// 14 Shabal, parallel 4 way 32 bit SSE
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, 512 );
shabal512_4way( &ctx.shabal, vhash, 64 );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, 512 );
// 15 Whirlpool
sph_whirlpool( &ctx.whirlpool, hash0, 64 );
@@ -217,17 +214,18 @@ void x17_4way_hash( void *state, const void *input )
sha512_4way_close( &ctx.sha512, vhash );
// 17 Haval parallel 32 bit
mm256_reinterleave_4x32( vhash32, vhash, 512 );
haval256_5_4way( &ctx.haval, vhash32, 64 );
haval256_5_4way_close( &ctx.haval, vhash );
mm256_reinterleave_4x32( vhashB, vhash, 512 );
haval256_5_4way( &ctx.haval, vhashB, 64 );
haval256_5_4way_close( &ctx.haval, state );
mm_deinterleave_4x32( state, state+32, state+64, state+96, vhash, 256 );
}
int scanhash_x17_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t lane_hash[8];
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t endiandata[20] __attribute__((aligned(64)));
uint32_t *pdata = work->data;
@@ -261,17 +259,20 @@ int scanhash_x17_4way( int thr_id, struct work *work, uint32_t max_nonce,
be32enc( noncep+6, n+3 );
x17_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( ( ( (hash+(i<<3))[7] & mask ) == 0 )
&& fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( ( ( hash7[ lane ] & mask ) == 0 ) )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )
&& !work_restart[thr_id].restart );
break;

2
algo/x17/x17.c
View File

@@ -22,7 +22,7 @@
#include "algo/haval/sph-haval.h"
#include "algo/luffa/luffa_for_sse2.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/nist.h"
#include "algo/blake/sse2/blake.c"
#include "algo/bmw/sse2/bmw.c"

36
algo/x17/xevan-4way.c
View File

@@ -14,7 +14,7 @@
#include "algo/skein/skein-hash-4way.h"
#include "algo/shavite/sph_shavite.h"
#include "algo/luffa/luffa-hash-2way.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include "algo/simd/simd-hash-2way.h"
#include "algo/echo/aes_ni/hash_api.h"
#include "algo/hamsi/hamsi-hash-4way.h"
@@ -201,10 +201,10 @@ void xevan_4way_hash( void *output, const void *input )
sph_fugue512_close( &ctx.fugue, hash3 );
// Parallel 4way 32 bit
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
shabal512_4way( &ctx.shabal, vhash, dataLen );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
// Serial
sph_whirlpool( &ctx.whirlpool, hash0, dataLen );
@@ -229,7 +229,7 @@ void xevan_4way_hash( void *output, const void *input )
mm256_reinterleave_4x32( vhash32, vhash, dataLen<<3 );
haval256_5_4way( &ctx.haval, vhash32, dataLen );
haval256_5_4way_close( &ctx.haval, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
mm256_interleave_4x64( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
memset( &vhash[ 4<<2 ], 0, (dataLen-32) << 2 );
@@ -339,10 +339,10 @@ void xevan_4way_hash( void *output, const void *input )
sph_fugue512( &ctx.fugue, hash3, dataLen );
sph_fugue512_close( &ctx.fugue, hash3 );
mm_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
mm128_interleave_4x32( vhash, hash0, hash1, hash2, hash3, dataLen<<3 );
shabal512_4way( &ctx.shabal, vhash, dataLen );
shabal512_4way_close( &ctx.shabal, vhash );
mm_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
mm128_deinterleave_4x32( hash0, hash1, hash2, hash3, vhash, dataLen<<3 );
sph_whirlpool( &ctx.whirlpool, hash0, dataLen );
sph_whirlpool_close( &ctx.whirlpool, hash0 );
@@ -365,16 +365,15 @@ void xevan_4way_hash( void *output, const void *input )
mm256_reinterleave_4x32( vhash32, vhash, dataLen<<3 );
haval256_5_4way( &ctx.haval, vhash32, dataLen );
haval256_5_4way_close( &ctx.haval, vhash32 );
mm_deinterleave_4x32( output, output+32, output+64, output+96,
vhash32, 256 );
haval256_5_4way_close( &ctx.haval, output );
}
int scanhash_xevan_4way( int thr_id, struct work *work, uint32_t max_nonce,
uint64_t *hashes_done )
{
uint32_t hash[4*8] __attribute__ ((aligned (64)));
uint32_t *hash7 = &(hash[7<<2]);
uint32_t lane_hash[8];
uint32_t vdata[24*4] __attribute__ ((aligned (64)));
uint32_t _ALIGN(64) endiandata[20];
uint32_t *pdata = work->data;
@@ -405,15 +404,16 @@ int scanhash_xevan_4way( int thr_id, struct work *work, uint32_t max_nonce,
be32enc( noncep+6, n+3 );
xevan_4way_hash( hash, vdata );
pdata[19] = n;
for ( int i = 0; i < 4; i++ )
if ( (hash+(i<<3))[7] <= Htarg && fulltest( hash+(i<<3), ptarget ) )
for ( int lane = 0; lane < 4; lane++ )
if ( hash7[ lane ] <= Htarg )
{
pdata[19] = n+i;
nonces[ num_found++ ] = n+i;
work_set_target_ratio( work, hash+(i<<3) );
mm128_extract_lane_4x32( lane_hash, hash, lane, 256 );
if ( fulltest( lane_hash, ptarget ) )
{
pdata[19] = n + lane;
nonces[ num_found++ ] = n + lane;
work_set_target_ratio( work, lane_hash );
}
}
n += 4;
} while ( ( num_found == 0 ) && ( n < max_nonce )

2
algo/x17/xevan.c
View File

@@ -19,7 +19,7 @@
#include "algo/sha/sph_sha2.h"
#include "algo/haval/sph-haval.h"
#include "algo/simd/nist.h"
#include "algo/cubehash/sse2/cubehash_sse2.h"
#include "algo/cubehash/cubehash_sse2.h"
#include <openssl/sha.h>
#ifdef NO_AES_NI
#include "algo/groestl/sph_groestl.h"

Some files were not shown because too many files have changed in this diff Show More