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

v3.8.3

Browse Source
This commit is contained in:
Jay D Dee
2018-02-23 12:39:15 -05:00
parent 502ed0b1fe
commit 3c02653dbe
70 changed files with 3871 additions and 1848 deletions
Download Patch File Download Diff File Expand all files Collapse all files

120
avxdefs.h
View File

@@ -1537,6 +1537,38 @@ static inline void mm256_interleave_8x32( void *dst, const void *src0,
s3[18], s2[18], s1[18], s0[18] );
d[19] = _mm256_set_epi32( s7[19], s6[19], s5[19], s4[19],
s3[19], s2[19], s1[19], s0[19] );
if ( bit_len <= 640 ) return;
d[20] = _mm256_set_epi32( s7[20], s6[20], s5[20], s4[20],
s3[20], s2[20], s1[20], s0[20] );
d[21] = _mm256_set_epi32( s7[21], s6[21], s5[21], s4[21],
s3[21], s2[21], s1[21], s0[21] );
d[22] = _mm256_set_epi32( s7[22], s6[22], s5[22], s4[22],
s3[22], s2[22], s1[22], s0[22] );
d[23] = _mm256_set_epi32( s7[23], s6[23], s5[23], s4[23],
s3[23], s2[23], s1[23], s0[23] );
if ( bit_len <= 768 ) return;
d[24] = _mm256_set_epi32( s7[24], s6[24], s5[24], s4[24],
s3[24], s2[24], s1[24], s0[24] );
d[25] = _mm256_set_epi32( s7[25], s6[25], s5[25], s4[25],
s3[25], s2[25], s1[25], s0[25] );
d[26] = _mm256_set_epi32( s7[26], s6[26], s5[26], s4[26],
s3[26], s2[26], s1[26], s0[26] );
d[27] = _mm256_set_epi32( s7[27], s6[27], s5[27], s4[27],
s3[27], s2[27], s1[27], s0[27] );
d[28] = _mm256_set_epi32( s7[28], s6[28], s5[28], s4[28],
s3[28], s2[28], s1[28], s0[28] );
d[29] = _mm256_set_epi32( s7[29], s6[29], s5[29], s4[29],
s3[29], s2[29], s1[29], s0[29] );
d[30] = _mm256_set_epi32( s7[30], s6[30], s5[30], s4[30],
s3[30], s2[30], s1[30], s0[30] );
d[31] = _mm256_set_epi32( s7[31], s6[31], s5[31], s4[31],
s3[31], s2[31], s1[31], s0[31] );
// bit_len == 1024
}
// probably obsolete with double pack 2x32->64, 4x64->256.
@@ -1615,31 +1647,71 @@ static inline void mm256_deinterleave_8x32( void *dst0, void *dst1, void *dst2,
// null change for overrun space, vector indexing doesn't work for
// 32 bit data
if ( bit_len <= 640 )
{
uint32_t *d = ((uint32_t*)d0) + 8;
d0[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[152], s[144], s[136], s[128] );
d = ((uint32_t*)d1) + 8;
d1[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[153], s[145], s[137], s[129] );
d = ((uint32_t*)d2) + 8;
d2[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[154], s[146], s[138], s[130]);
d = ((uint32_t*)d3) + 8;
d3[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[155], s[147], s[139], s[131] );
d = ((uint32_t*)d4) + 8;
d4[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[156], s[148], s[140], s[132] );
d = ((uint32_t*)d5) + 8;
d5[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[157], s[149], s[141], s[133] );
d = ((uint32_t*)d6) + 8;
d6[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[158], s[150], s[142], s[134] );
d = ((uint32_t*)d7) + 8;
d7[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[159], s[151], s[143], s[135] );
return;
}
uint32_t *d = ((uint32_t*)d0) + 8;
d0[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[152], s[144], s[136], s[128] );
d = ((uint32_t*)d1) + 8;
d1[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[153], s[145], s[137], s[129] );
d = ((uint32_t*)d2) + 8;
d2[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[154], s[146], s[138], s[130]);
d = ((uint32_t*)d3) + 8;
d3[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[155], s[147], s[139], s[131] );
d = ((uint32_t*)d4) + 8;
d4[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[156], s[148], s[140], s[132] );
d = ((uint32_t*)d5) + 8;
d5[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[157], s[149], s[141], s[133] );
d = ((uint32_t*)d6) + 8;
d6[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[158], s[150], s[142], s[134] );
d = ((uint32_t*)d7) + 8;
d7[2] = _mm256_set_epi32( *(d+7), *(d+6), *(d+5), *(d+4),
s[159], s[151], s[143], s[135] );
d0[2] = _mm256_set_epi32( s[184], s[176], s[168], s[160],
s[152], s[144], s[136], s[128] );
d1[2] = _mm256_set_epi32( s[185], s[177], s[169], s[161],
s[153], s[145], s[137], s[129] );
d2[2] = _mm256_set_epi32( s[186], s[178], s[170], s[162],
s[154], s[146], s[138], s[130] );
d3[2] = _mm256_set_epi32( s[187], s[179], s[171], s[163],
s[155], s[147], s[139], s[131] );
d4[2] = _mm256_set_epi32( s[188], s[180], s[172], s[164],
s[156], s[148], s[140], s[132] );
d5[2] = _mm256_set_epi32( s[189], s[181], s[173], s[165],
s[157], s[149], s[141], s[133] );
d6[2] = _mm256_set_epi32( s[190], s[182], s[174], s[166],
s[158], s[150], s[142], s[134] );
d7[2] = _mm256_set_epi32( s[191], s[183], s[175], s[167],
s[159], s[151], s[143], s[135] );
if ( bit_len <= 768 ) return;
d0[3] = _mm256_set_epi32( s[248], s[240], s[232], s[224],
s[216], s[208], s[200], s[192] );
d1[3] = _mm256_set_epi32( s[249], s[241], s[233], s[225],
s[217], s[209], s[201], s[193] );
d2[3] = _mm256_set_epi32( s[250], s[242], s[234], s[226],
s[218], s[210], s[202], s[194] );
d3[3] = _mm256_set_epi32( s[251], s[243], s[235], s[227],
s[219], s[211], s[203], s[195] );
d4[3] = _mm256_set_epi32( s[252], s[244], s[236], s[228],
s[220], s[212], s[204], s[196] );
d5[3] = _mm256_set_epi32( s[253], s[245], s[237], s[229],
s[221], s[213], s[205], s[197] );
d6[3] = _mm256_set_epi32( s[254], s[246], s[238], s[230],
s[222], s[214], s[206], s[198] );
d7[3] = _mm256_set_epi32( s[255], s[247], s[239], s[231],
s[223], s[215], s[207], s[199] );
// bit_len == 1024
}
// Deinterleave 8 arrays into indivdual buffers for scalar processing