989 lines
24 KiB
C
989 lines
24 KiB
C
/*******************************************************************************
|
|
Copyright (c) 2016, The OpenBLAS Project
|
|
All rights reserved.
|
|
Redistribution and use in source and binary forms, with or without
|
|
modification, are permitted provided that the following conditions are
|
|
met:
|
|
1. Redistributions of source code must retain the above copyright
|
|
notice, this list of conditions and the following disclaimer.
|
|
2. Redistributions in binary form must reproduce the above copyright
|
|
notice, this list of conditions and the following disclaimer in
|
|
the documentation and/or other materials provided with the
|
|
distribution.
|
|
3. Neither the name of the OpenBLAS project nor the names of
|
|
its contributors may be used to endorse or promote products
|
|
derived from this software without specific prior written permission.
|
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
|
|
AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
ARE DISCLAIMED. IN NO EVENT SHALL THE OPENBLAS PROJECT OR CONTRIBUTORS BE
|
|
LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
|
|
SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
|
|
CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
|
|
OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE
|
|
USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
|
*******************************************************************************/
|
|
|
|
#include "common.h"
|
|
#include "macros_msa.h"
|
|
|
|
static __attribute__ ((noinline))
|
|
void dsolve_8x4_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
v2f64 src_c0, src_c1, src_c2, src_c3, src_c4, src_c5, src_c6, src_c7;
|
|
v2f64 src_c8, src_c9, src_c10, src_c11, src_c12, src_c13, src_c14, src_c15;
|
|
v2f64 src_b0, src_b1, src_b2, src_b3, src_b5, src_b6, src_b7;
|
|
v2f64 src_b10, src_b11, src_b15;
|
|
FLOAT *c_nxt1line = c + ldc;
|
|
FLOAT *c_nxt2line = c + 2 * ldc;
|
|
FLOAT *c_nxt3line = c + 3 * ldc;
|
|
|
|
LD_DP4(c, 2, src_c0, src_c1, src_c2, src_c3);
|
|
LD_DP4(c_nxt1line, 2, src_c4, src_c5, src_c6, src_c7);
|
|
LD_DP4(c_nxt2line, 2, src_c8, src_c9, src_c10, src_c11);
|
|
LD_DP4(c_nxt3line, 2, src_c12, src_c13, src_c14, src_c15);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i, pref_offset;
|
|
FLOAT *pa0_pref;
|
|
v2f64 src_a0, src_a1, src_a2, src_a3, src_b;
|
|
|
|
pref_offset = (uintptr_t)a & (L1_DATA_LINESIZE - 1);
|
|
|
|
if (pref_offset)
|
|
{
|
|
pref_offset = L1_DATA_LINESIZE - pref_offset;
|
|
pref_offset = pref_offset / sizeof(FLOAT);
|
|
}
|
|
|
|
pa0_pref = a + pref_offset;
|
|
|
|
for (i = (bk >> 1); i--;)
|
|
{
|
|
PREF_OFFSET(pa0_pref, 128);
|
|
PREF_OFFSET(pa0_pref, 160);
|
|
PREF_OFFSET(pa0_pref, 192);
|
|
PREF_OFFSET(pa0_pref, 224);
|
|
|
|
LD_DP4_INC(a, 2, src_a0, src_a1, src_a2, src_a3);
|
|
LD_DP2_INC(b, 2, src_b0, src_b1);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
src_c2 -= src_a2 * src_b;
|
|
src_c3 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c4 -= src_a0 * src_b;
|
|
src_c5 -= src_a1 * src_b;
|
|
src_c6 -= src_a2 * src_b;
|
|
src_c7 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c8 -= src_a0 * src_b;
|
|
src_c9 -= src_a1 * src_b;
|
|
src_c10 -= src_a2 * src_b;
|
|
src_c11 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c12 -= src_a0 * src_b;
|
|
src_c13 -= src_a1 * src_b;
|
|
src_c14 -= src_a2 * src_b;
|
|
src_c15 -= src_a3 * src_b;
|
|
|
|
LD_DP4_INC(a, 2, src_a0, src_a1, src_a2, src_a3);
|
|
LD_DP2_INC(b, 2, src_b0, src_b1);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
src_c2 -= src_a2 * src_b;
|
|
src_c3 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c4 -= src_a0 * src_b;
|
|
src_c5 -= src_a1 * src_b;
|
|
src_c6 -= src_a2 * src_b;
|
|
src_c7 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c8 -= src_a0 * src_b;
|
|
src_c9 -= src_a1 * src_b;
|
|
src_c10 -= src_a2 * src_b;
|
|
src_c11 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c12 -= src_a0 * src_b;
|
|
src_c13 -= src_a1 * src_b;
|
|
src_c14 -= src_a2 * src_b;
|
|
src_c15 -= src_a3 * src_b;
|
|
|
|
pa0_pref += 16;
|
|
}
|
|
|
|
if (bk & 1)
|
|
{
|
|
LD_DP4_INC(a, 2, src_a0, src_a1, src_a2, src_a3);
|
|
LD_DP2_INC(b, 2, src_b0, src_b1);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
src_c2 -= src_a2 * src_b;
|
|
src_c3 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c4 -= src_a0 * src_b;
|
|
src_c5 -= src_a1 * src_b;
|
|
src_c6 -= src_a2 * src_b;
|
|
src_c7 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c8 -= src_a0 * src_b;
|
|
src_c9 -= src_a1 * src_b;
|
|
src_c10 -= src_a2 * src_b;
|
|
src_c11 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c12 -= src_a0 * src_b;
|
|
src_c13 -= src_a1 * src_b;
|
|
src_c14 -= src_a2 * src_b;
|
|
src_c15 -= src_a3 * src_b;
|
|
}
|
|
}
|
|
|
|
src_b0 = LD_DP(b + 0);
|
|
src_b1 = (v2f64) __msa_splati_d((v2i64) src_b0, 1);
|
|
src_b0 = (v2f64) __msa_splati_d((v2i64) src_b0, 0);
|
|
src_b2 = LD_DP(b + 2);
|
|
src_b3 = (v2f64) __msa_splati_d((v2i64) src_b2, 1);
|
|
src_b2 = (v2f64) __msa_splati_d((v2i64) src_b2, 0);
|
|
src_b5 = __msa_cast_to_vector_double(*(b + 5));
|
|
src_b5 = (v2f64) __msa_splati_d((v2i64) src_b5, 0);
|
|
src_b6 = LD_DP(b + 6);
|
|
src_b7 = (v2f64) __msa_splati_d((v2i64) src_b6, 1);
|
|
src_b6 = (v2f64) __msa_splati_d((v2i64) src_b6, 0);
|
|
src_b10 = LD_DP(b + 10);
|
|
src_b11 = (v2f64) __msa_splati_d((v2i64) src_b10, 1);
|
|
src_b10 = (v2f64) __msa_splati_d((v2i64) src_b10, 0);
|
|
src_b15 = __msa_cast_to_vector_double(*(b + 15));
|
|
src_b15 = (v2f64) __msa_splati_d((v2i64) src_b15, 0);
|
|
|
|
src_c0 *= src_b0;
|
|
src_c1 *= src_b0;
|
|
src_c2 *= src_b0;
|
|
src_c3 *= src_b0;
|
|
|
|
src_c4 -= src_c0 * src_b1;
|
|
src_c5 -= src_c1 * src_b1;
|
|
src_c6 -= src_c2 * src_b1;
|
|
src_c7 -= src_c3 * src_b1;
|
|
|
|
src_c4 *= src_b5;
|
|
src_c5 *= src_b5;
|
|
src_c6 *= src_b5;
|
|
src_c7 *= src_b5;
|
|
|
|
src_c8 -= src_c0 * src_b2;
|
|
src_c9 -= src_c1 * src_b2;
|
|
src_c10 -= src_c2 * src_b2;
|
|
src_c11 -= src_c3 * src_b2;
|
|
|
|
src_c8 -= src_c4 * src_b6;
|
|
src_c9 -= src_c5 * src_b6;
|
|
src_c10 -= src_c6 * src_b6;
|
|
src_c11 -= src_c7 * src_b6;
|
|
|
|
src_c8 *= src_b10;
|
|
src_c9 *= src_b10;
|
|
src_c10 *= src_b10;
|
|
src_c11 *= src_b10;
|
|
|
|
src_c12 -= src_c0 * src_b3;
|
|
src_c13 -= src_c1 * src_b3;
|
|
src_c14 -= src_c2 * src_b3;
|
|
src_c15 -= src_c3 * src_b3;
|
|
|
|
src_c12 -= src_c4 * src_b7;
|
|
src_c13 -= src_c5 * src_b7;
|
|
src_c14 -= src_c6 * src_b7;
|
|
src_c15 -= src_c7 * src_b7;
|
|
|
|
src_c12 -= src_c8 * src_b11;
|
|
src_c13 -= src_c9 * src_b11;
|
|
src_c14 -= src_c10 * src_b11;
|
|
src_c15 -= src_c11 * src_b11;
|
|
|
|
src_c12 *= src_b15;
|
|
src_c13 *= src_b15;
|
|
src_c14 *= src_b15;
|
|
src_c15 *= src_b15;
|
|
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, c, 2);
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, a, 2);
|
|
ST_DP4(src_c4, src_c5, src_c6, src_c7, c_nxt1line, 2);
|
|
ST_DP4(src_c4, src_c5, src_c6, src_c7, a + 8, 2);
|
|
ST_DP4(src_c8, src_c9, src_c10, src_c11, c_nxt2line, 2);
|
|
ST_DP4(src_c8, src_c9, src_c10, src_c11, a + 16, 2);
|
|
ST_DP4(src_c12, src_c13, src_c14, src_c15, c_nxt3line, 2);
|
|
ST_DP4(src_c12, src_c13, src_c14, src_c15, a + 24, 2);
|
|
}
|
|
|
|
static void dsolve_8x2_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
v2f64 src_c0, src_c1, src_c2, src_c3, src_c4, src_c5, src_c6, src_c7;
|
|
v2f64 src_b0, src_b1, src_b3, src_b;
|
|
|
|
LD_DP4(c, 2, src_c0, src_c1, src_c2, src_c3);
|
|
LD_DP4(c + ldc, 2, src_c4, src_c5, src_c6, src_c7);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
v2f64 src_a0, src_a1, src_a2, src_a3, src_a4, src_a5, src_a6, src_a7;
|
|
|
|
LD_DP4(a, 2, src_a0, src_a1, src_a2, src_a3);
|
|
src_b0 = LD_DP(b);
|
|
|
|
a += 8;
|
|
b += 2;
|
|
|
|
for (i = (bk - 1); i--;)
|
|
{
|
|
LD_DP4(a, 2, src_a4, src_a5, src_a6, src_a7);
|
|
src_b1 = LD_DP(b);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
src_c2 -= src_a2 * src_b;
|
|
src_c3 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c4 -= src_a0 * src_b;
|
|
src_c5 -= src_a1 * src_b;
|
|
src_c6 -= src_a2 * src_b;
|
|
src_c7 -= src_a3 * src_b;
|
|
|
|
src_a0 = src_a4;
|
|
src_a1 = src_a5;
|
|
src_a2 = src_a6;
|
|
src_a3 = src_a7;
|
|
src_b0 = src_b1;
|
|
|
|
a += 8;
|
|
b += 2;
|
|
}
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
src_c2 -= src_a2 * src_b;
|
|
src_c3 -= src_a3 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c4 -= src_a0 * src_b;
|
|
src_c5 -= src_a1 * src_b;
|
|
src_c6 -= src_a2 * src_b;
|
|
src_c7 -= src_a3 * src_b;
|
|
}
|
|
|
|
src_b0 = LD_DP(b + 0);
|
|
src_b1 = (v2f64) __msa_splati_d((v2i64) src_b0, 1);
|
|
src_b0 = (v2f64) __msa_splati_d((v2i64) src_b0, 0);
|
|
src_b3 = __msa_cast_to_vector_double(*(b + 3));
|
|
src_b3 = (v2f64) __msa_splati_d((v2i64) src_b3, 0);
|
|
|
|
src_c0 *= src_b0;
|
|
src_c1 *= src_b0;
|
|
src_c2 *= src_b0;
|
|
src_c3 *= src_b0;
|
|
|
|
src_c4 -= src_c0 * src_b1;
|
|
src_c5 -= src_c1 * src_b1;
|
|
src_c6 -= src_c2 * src_b1;
|
|
src_c7 -= src_c3 * src_b1;
|
|
|
|
src_c4 *= src_b3;
|
|
src_c5 *= src_b3;
|
|
src_c6 *= src_b3;
|
|
src_c7 *= src_b3;
|
|
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, c, 2);
|
|
ST_DP4(src_c4, src_c5, src_c6, src_c7, c + ldc, 2);
|
|
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, a, 2);
|
|
ST_DP4(src_c4, src_c5, src_c6, src_c7, a + 8, 2);
|
|
}
|
|
|
|
static void dsolve_8x1_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG bk)
|
|
{
|
|
v2f64 src_c0, src_c1, src_c2, src_c3;
|
|
v2f64 src_b0;
|
|
|
|
LD_DP4(c, 2, src_c0, src_c1, src_c2, src_c3);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
v2f64 src_a0, src_a1, src_a2, src_a3, src_b;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
LD_DP4(a, 2, src_a0, src_a1, src_a2, src_a3);
|
|
src_b = LD_DP(b);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b, (v2i64) src_b);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
src_c2 -= src_a2 * src_b;
|
|
src_c3 -= src_a3 * src_b;
|
|
|
|
a += 8;
|
|
b += 1;
|
|
}
|
|
}
|
|
|
|
src_b0 = __msa_cast_to_vector_double(*b);
|
|
src_b0 = (v2f64) __msa_splati_d((v2i64) src_b0, 0);
|
|
|
|
src_c0 *= src_b0;
|
|
src_c1 *= src_b0;
|
|
src_c2 *= src_b0;
|
|
src_c3 *= src_b0;
|
|
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, c, 2);
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, a, 2);
|
|
}
|
|
|
|
static void dsolve_4x4_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
v2f64 src_c0, src_c1, src_c2, src_c3, src_c4, src_c5, src_c6, src_c7;
|
|
v2f64 src_b0, src_b1, src_b2, src_b3, src_b5, src_b6, src_b7;
|
|
v2f64 src_b10, src_b11, src_b15;
|
|
|
|
LD_DP2(c, 2, src_c0, src_c1);
|
|
LD_DP2(c + ldc, 2, src_c2, src_c3);
|
|
LD_DP2(c + 2 * ldc, 2, src_c4, src_c5);
|
|
LD_DP2(c + 3 * ldc, 2, src_c6, src_c7);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
v2f64 src_a0, src_a1, src_b, src_b0, src_b1;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
LD_DP2(a, 2, src_a0, src_a1);
|
|
LD_DP2(b, 2, src_b0, src_b1);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c2 -= src_a0 * src_b;
|
|
src_c3 -= src_a1 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c4 -= src_a0 * src_b;
|
|
src_c5 -= src_a1 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
|
|
src_c6 -= src_a0 * src_b;
|
|
src_c7 -= src_a1 * src_b;
|
|
|
|
a += 4;
|
|
b += 4;
|
|
}
|
|
}
|
|
|
|
src_b0 = LD_DP(b + 0);
|
|
src_b1 = (v2f64) __msa_splati_d((v2i64) src_b0, 1);
|
|
src_b0 = (v2f64) __msa_splati_d((v2i64) src_b0, 0);
|
|
src_b2 = LD_DP(b + 2);
|
|
src_b3 = (v2f64) __msa_splati_d((v2i64) src_b2, 1);
|
|
src_b2 = (v2f64) __msa_splati_d((v2i64) src_b2, 0);
|
|
src_b5 = __msa_cast_to_vector_double(*(b + 5));
|
|
src_b5 = (v2f64) __msa_splati_d((v2i64) src_b5, 0);
|
|
src_b6 = LD_DP(b + 6);
|
|
src_b7 = (v2f64) __msa_splati_d((v2i64) src_b6, 1);
|
|
src_b6 = (v2f64) __msa_splati_d((v2i64) src_b6, 0);
|
|
src_b10 = LD_DP(b + 10);
|
|
src_b11 = (v2f64) __msa_splati_d((v2i64) src_b10, 1);
|
|
src_b10 = (v2f64) __msa_splati_d((v2i64) src_b10, 0);
|
|
src_b15 = __msa_cast_to_vector_double(*(b + 15));
|
|
src_b15 = (v2f64) __msa_splati_d((v2i64) src_b15, 0);
|
|
|
|
src_c0 *= src_b0;
|
|
src_c1 *= src_b0;
|
|
|
|
src_c2 -= src_c0 * src_b1;
|
|
src_c3 -= src_c1 * src_b1;
|
|
|
|
src_c2 *= src_b5;
|
|
src_c3 *= src_b5;
|
|
|
|
src_c4 -= src_c0 * src_b2;
|
|
src_c5 -= src_c1 * src_b2;
|
|
|
|
src_c4 -= src_c2 * src_b6;
|
|
src_c5 -= src_c3 * src_b6;
|
|
|
|
src_c4 *= src_b10;
|
|
src_c5 *= src_b10;
|
|
|
|
src_c6 -= src_c0 * src_b3;
|
|
src_c7 -= src_c1 * src_b3;
|
|
|
|
src_c6 -= src_c2 * src_b7;
|
|
src_c7 -= src_c3 * src_b7;
|
|
|
|
src_c6 -= src_c4 * src_b11;
|
|
src_c7 -= src_c5 * src_b11;
|
|
|
|
src_c6 *= src_b15;
|
|
src_c7 *= src_b15;
|
|
|
|
ST_DP2(src_c0, src_c1, c, 2);
|
|
ST_DP2(src_c2, src_c3, c + ldc, 2);
|
|
ST_DP2(src_c4, src_c5, c + 2 * ldc, 2);
|
|
ST_DP2(src_c6, src_c7, c + 3 * ldc, 2);
|
|
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, a, 2);
|
|
ST_DP4(src_c4, src_c5, src_c6, src_c7, a + 8, 2);
|
|
}
|
|
|
|
static void dsolve_4x2_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
v2f64 src_c0, src_c1, src_c2, src_c3, src_b0, src_b1, src_b3;
|
|
|
|
LD_DP2(c, 2, src_c0, src_c1);
|
|
LD_DP2(c + ldc, 2, src_c2, src_c3);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
v2f64 src_a0, src_a1, src_b, src_b0;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
LD_DP2(a, 2, src_a0, src_a1);
|
|
src_b0 = LD_DP(b);
|
|
|
|
src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c0 -= src_a0 * src_b;
|
|
src_c1 -= src_a1 * src_b;
|
|
|
|
src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
|
|
src_c2 -= src_a0 * src_b;
|
|
src_c3 -= src_a1 * src_b;
|
|
|
|
a += 4;
|
|
b += 2;
|
|
}
|
|
}
|
|
|
|
src_b0 = LD_DP(b + 0);
|
|
src_b1 = (v2f64) __msa_splati_d((v2i64) src_b0, 1);
|
|
src_b0 = (v2f64) __msa_splati_d((v2i64) src_b0, 0);
|
|
src_b3 = __msa_cast_to_vector_double(*(b + 3));
|
|
src_b3 = (v2f64) __msa_splati_d((v2i64) src_b3, 0);
|
|
|
|
src_c0 *= src_b0;
|
|
src_c1 *= src_b0;
|
|
|
|
src_c2 -= src_c0 * src_b1;
|
|
src_c3 -= src_c1 * src_b1;
|
|
|
|
src_c2 *= src_b3;
|
|
src_c3 *= src_b3;
|
|
|
|
ST_DP2(src_c0, src_c1, c, 2);
|
|
ST_DP2(src_c2, src_c3, c + ldc, 2);
|
|
|
|
ST_DP4(src_c0, src_c1, src_c2, src_c3, a, 2);
|
|
}
|
|
|
|
static void dsolve_4x1_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG bk)
|
|
{
|
|
FLOAT c0, c1, c2, c3;
|
|
|
|
c0 = *(c + 0);
|
|
c1 = *(c + 1);
|
|
c2 = *(c + 2);
|
|
c3 = *(c + 3);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
c0 -= a[0] * b[0];
|
|
c1 -= a[1] * b[0];
|
|
c2 -= a[2] * b[0];
|
|
c3 -= a[3] * b[0];
|
|
|
|
a += 4;
|
|
b += 1;
|
|
}
|
|
}
|
|
|
|
c0 *= *b;
|
|
c1 *= *b;
|
|
c2 *= *b;
|
|
c3 *= *b;
|
|
|
|
*(a + 0) = c0;
|
|
*(a + 1) = c1;
|
|
*(a + 2) = c2;
|
|
*(a + 3) = c3;
|
|
|
|
*(c + 0) = c0;
|
|
*(c + 1) = c1;
|
|
*(c + 2) = c2;
|
|
*(c + 3) = c3;
|
|
}
|
|
|
|
static void dsolve_2x4_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
FLOAT b0, b1, b2, b3, b5, b6, b7, b10, b11, b15;
|
|
FLOAT c0, c0_nxt1, c0_nxt2, c0_nxt3;
|
|
FLOAT c1, c1_nxt1, c1_nxt2, c1_nxt3;
|
|
|
|
c0 = *(c + 0);
|
|
c1 = *(c + 1);
|
|
c0_nxt1 = *(c + 0 + 1 * ldc);
|
|
c1_nxt1 = *(c + 1 + 1 * ldc);
|
|
c0_nxt2 = *(c + 0 + 2 * ldc);
|
|
c1_nxt2 = *(c + 1 + 2 * ldc);
|
|
c0_nxt3 = *(c + 0 + 3 * ldc);
|
|
c1_nxt3 = *(c + 1 + 3 * ldc);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
c0 -= a[0] * b[0];
|
|
c1 -= a[1] * b[0];
|
|
c0_nxt1 -= a[0] * b[1];
|
|
c1_nxt1 -= a[1] * b[1];
|
|
c0_nxt2 -= a[0] * b[2];
|
|
c1_nxt2 -= a[1] * b[2];
|
|
c0_nxt3 -= a[0] * b[3];
|
|
c1_nxt3 -= a[1] * b[3];
|
|
|
|
a += 2;
|
|
b += 4;
|
|
}
|
|
}
|
|
|
|
b0 = *(b + 0);
|
|
b1 = *(b + 1);
|
|
b2 = *(b + 2);
|
|
b3 = *(b + 3);
|
|
b5 = *(b + 5);
|
|
b6 = *(b + 6);
|
|
b7 = *(b + 7);
|
|
b10 = *(b + 10);
|
|
b11 = *(b + 11);
|
|
b15 = *(b + 15);
|
|
|
|
c0 *= b0;
|
|
c1 *= b0;
|
|
|
|
c0_nxt1 -= c0 * b1;
|
|
c1_nxt1 -= c1 * b1;
|
|
c0_nxt1 *= b5;
|
|
c1_nxt1 *= b5;
|
|
|
|
c0_nxt2 -= c0 * b2;
|
|
c1_nxt2 -= c1 * b2;
|
|
c0_nxt2 -= c0_nxt1 * b6;
|
|
c1_nxt2 -= c1_nxt1 * b6;
|
|
c0_nxt2 *= b10;
|
|
c1_nxt2 *= b10;
|
|
|
|
c0_nxt3 -= c0 * b3;
|
|
c1_nxt3 -= c1 * b3;
|
|
c0_nxt3 -= c0_nxt1 * b7;
|
|
c1_nxt3 -= c1_nxt1 * b7;
|
|
c0_nxt3 -= c0_nxt2 * b11;
|
|
c1_nxt3 -= c1_nxt2 * b11;
|
|
c0_nxt3 *= b15;
|
|
c1_nxt3 *= b15;
|
|
|
|
*(a + 0) = c0;
|
|
*(a + 1) = c1;
|
|
*(a + 2) = c0_nxt1;
|
|
*(a + 3) = c1_nxt1;
|
|
*(a + 4) = c0_nxt2;
|
|
*(a + 5) = c1_nxt2;
|
|
*(a + 6) = c0_nxt3;
|
|
*(a + 7) = c1_nxt3;
|
|
|
|
*(c + 0) = c0;
|
|
*(c + 1) = c1;
|
|
*(c + 1 * ldc) = c0_nxt1;
|
|
*(c + 1 + 1 * ldc) = c1_nxt1;
|
|
*(c + 2 * ldc) = c0_nxt2;
|
|
*(c + 1 + 2 * ldc) = c1_nxt2;
|
|
*(c + 3 * ldc) = c0_nxt3;
|
|
*(c + 1 + 3 * ldc) = c1_nxt3;
|
|
}
|
|
|
|
static void dsolve_2x2_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
FLOAT b0, b1, b3, c0, c0_nxt, c1, c1_nxt;
|
|
|
|
c0 = *(c + 0);
|
|
c1 = *(c + 1);
|
|
c0_nxt = *(c + 0 + ldc);
|
|
c1_nxt = *(c + 1 + ldc);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
c0 -= a[0] * b[0];
|
|
c1 -= a[1] * b[0];
|
|
|
|
c0_nxt -= a[0] * b[1];
|
|
c1_nxt -= a[1] * b[1];
|
|
|
|
a += 2;
|
|
b += 2;
|
|
}
|
|
}
|
|
|
|
b0 = *(b + 0);
|
|
b1 = *(b + 1);
|
|
b3 = *(b + 3);
|
|
|
|
c0 *= b0;
|
|
c1 *= b0;
|
|
|
|
c0_nxt -= c0 * b1;
|
|
c1_nxt -= c1 * b1;
|
|
|
|
c0_nxt *= b3;
|
|
c1_nxt *= b3;
|
|
|
|
*(a + 0) = c0;
|
|
*(a + 1) = c1;
|
|
*(a + 2) = c0_nxt;
|
|
*(a + 3) = c1_nxt;
|
|
|
|
*(c + 0) = c0;
|
|
*(c + 1) = c1;
|
|
*(c + 0 + ldc) = c0_nxt;
|
|
*(c + 1 + ldc) = c1_nxt;
|
|
}
|
|
|
|
static void dsolve_2x1_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG bk)
|
|
{
|
|
FLOAT b0, c0, c1;
|
|
|
|
c0 = *(c + 0);
|
|
c1 = *(c + 1);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
c0 -= a[0] * b[0];
|
|
c1 -= a[1] * b[0];
|
|
|
|
a += 2;
|
|
b += 1;
|
|
}
|
|
}
|
|
|
|
b0 = *b;
|
|
|
|
c0 *= b0;
|
|
c1 *= b0;
|
|
|
|
*(a + 0) = c0;
|
|
*(a + 1) = c1;
|
|
|
|
*(c + 0) = c0;
|
|
*(c + 1) = c1;
|
|
}
|
|
|
|
static void dsolve_1x4_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
FLOAT b0, b1, b2, b3, b5, b6, b7, b10, b11, b15, c0, c1, c2, c3;
|
|
|
|
c0 = *(c + 0);
|
|
c1 = *(c + 1 * ldc);
|
|
c2 = *(c + 2 * ldc);
|
|
c3 = *(c + 3 * ldc);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
c0 -= a[0] * b[0];
|
|
c1 -= a[0] * b[1];
|
|
c2 -= a[0] * b[2];
|
|
c3 -= a[0] * b[3];
|
|
|
|
a += 1;
|
|
b += 4;
|
|
}
|
|
}
|
|
|
|
b0 = *(b + 0);
|
|
b1 = *(b + 1);
|
|
b2 = *(b + 2);
|
|
b3 = *(b + 3);
|
|
b5 = *(b + 5);
|
|
b6 = *(b + 6);
|
|
b7 = *(b + 7);
|
|
b10 = *(b + 10);
|
|
b11 = *(b + 11);
|
|
b15 = *(b + 15);
|
|
|
|
c0 *= b0;
|
|
|
|
c1 -= c0 * b1;
|
|
c1 *= b5;
|
|
|
|
c2 -= c0 * b2;
|
|
c2 -= c1 * b6;
|
|
c2 *= b10;
|
|
|
|
c3 -= c0 * b3;
|
|
c3 -= c1 * b7;
|
|
c3 -= c2 * b11;
|
|
c3 *= b15;
|
|
|
|
*(a + 0) = c0;
|
|
*(a + 1) = c1;
|
|
*(a + 2) = c2;
|
|
*(a + 3) = c3;
|
|
|
|
*(c + 0) = c0;
|
|
*(c + 1 * ldc) = c1;
|
|
*(c + 2 * ldc) = c2;
|
|
*(c + 3 * ldc) = c3;
|
|
}
|
|
|
|
static void dsolve_1x2_rn_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG ldc, BLASLONG bk)
|
|
{
|
|
FLOAT b0, b1, b3, c0, c1;
|
|
|
|
c0 = *c;
|
|
c1 = *(c + ldc);
|
|
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
c0 -= *a * b[0];
|
|
c1 -= *a * b[1];
|
|
|
|
a += 1;
|
|
b += 2;
|
|
}
|
|
}
|
|
|
|
b0 = *(b + 0);
|
|
b1 = *(b + 1);
|
|
b3 = *(b + 3);
|
|
|
|
c0 *= b0;
|
|
|
|
c1 -= c0 * b1;
|
|
c1 *= b3;
|
|
|
|
*(a + 0) = c0;
|
|
*(a + 1) = c1;
|
|
|
|
*(c + 0) = c0;
|
|
*(c + ldc) = c1;
|
|
}
|
|
|
|
static void dgmm_dsolve_1x1_msa(FLOAT *a, FLOAT *b, FLOAT *c, BLASLONG bk)
|
|
{
|
|
if (bk)
|
|
{
|
|
BLASLONG i;
|
|
|
|
for (i = bk; i--;)
|
|
{
|
|
*c -= *a * *b;
|
|
|
|
a += 1;
|
|
b += 1;
|
|
}
|
|
}
|
|
|
|
*c *= *a;
|
|
*b = *c;
|
|
}
|
|
|
|
int CNAME(BLASLONG m, BLASLONG n, BLASLONG k, FLOAT dummy1, FLOAT *a, FLOAT *b,
|
|
FLOAT *c, BLASLONG ldc, BLASLONG offset)
|
|
{
|
|
BLASLONG i, j, kk;
|
|
FLOAT *aa, *cc;
|
|
|
|
kk = -offset;
|
|
|
|
for (j = (n >> 2); j--;)
|
|
{
|
|
aa = a;
|
|
cc = c;
|
|
|
|
for (i = (m >> 3); i--;)
|
|
{
|
|
dsolve_8x4_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += 8 * k;
|
|
cc += 8;
|
|
}
|
|
|
|
if (m & 7)
|
|
{
|
|
if (m & 4)
|
|
{
|
|
dsolve_4x4_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += 4 * k;
|
|
cc += 4;
|
|
}
|
|
|
|
if (m & 2)
|
|
{
|
|
dsolve_2x4_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += 2 * k;
|
|
cc += 2;
|
|
}
|
|
|
|
if (m & 1)
|
|
{
|
|
dsolve_1x4_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += k;
|
|
cc += 1;
|
|
}
|
|
}
|
|
|
|
kk += 4;
|
|
b += 4 * k;
|
|
c += 4 * ldc;
|
|
}
|
|
|
|
if (n & 3)
|
|
{
|
|
if (n & 2)
|
|
{
|
|
aa = a;
|
|
cc = c;
|
|
|
|
for (i = (m >> 3); i--;)
|
|
{
|
|
dsolve_8x2_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += 8 * k;
|
|
cc += 8;
|
|
}
|
|
|
|
if (m & 7)
|
|
{
|
|
if (m & 4)
|
|
{
|
|
dsolve_4x2_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += 4 * k;
|
|
cc += 4;
|
|
}
|
|
|
|
if (m & 2)
|
|
{
|
|
dsolve_2x2_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += 2 * k;
|
|
cc += 2;
|
|
}
|
|
|
|
if (m & 1)
|
|
{
|
|
dsolve_1x2_rn_msa(aa, b, cc, ldc, kk);
|
|
|
|
aa += k;
|
|
cc += 1;
|
|
}
|
|
}
|
|
|
|
b += 2 * k;
|
|
c += 2 * ldc;
|
|
kk += 2;
|
|
}
|
|
|
|
if (n & 1)
|
|
{
|
|
aa = a;
|
|
cc = c;
|
|
|
|
for (i = (m >> 3); i--;)
|
|
{
|
|
dsolve_8x1_rn_msa(aa, b, cc, kk);
|
|
|
|
aa += 8 * k;
|
|
cc += 8;
|
|
}
|
|
|
|
if (m & 7)
|
|
{
|
|
if (m & 4)
|
|
{
|
|
dsolve_4x1_rn_msa(aa, b, cc, kk);
|
|
|
|
aa += 4 * k;
|
|
cc += 4;
|
|
}
|
|
|
|
if (m & 2)
|
|
{
|
|
dsolve_2x1_rn_msa(aa, b, cc, kk);
|
|
|
|
aa += 2 * k;
|
|
cc += 2;
|
|
}
|
|
|
|
if (m & 1)
|
|
{
|
|
dgmm_dsolve_1x1_msa(b, aa, cc, kk);
|
|
|
|
aa += k;
|
|
cc += 1;
|
|
}
|
|
}
|
|
|
|
b += k;
|
|
c += ldc;
|
|
kk += 1;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|