OSchip
/
OpenBLAS
Not watched
1
0
Fork 0
Code
Releases 66 Wiki evaluate Activity
Issues 0 Pull Requests 0 Datasets Model Cloudbrain HPC
Browse Source

DGEMM optimization for MIPS P5600 and I6400 using MSA 

Signed-off-by: Shivraj Patil <shivraj.patil@imgtec.com>
tags/v0.2.19^2
Shivraj Patil 9 years ago
parent
2c3dfe2bf3
commit
b7b3d8ec8e
9 changed files with 1655 additions and 7 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +3
    -0
      CONTRIBUTORS.md
  2. +7
    -3
      kernel/mips/KERNEL.P5600
  3. +720
    -0
      kernel/mips/dgemm_kernel_8x4_msa.c
  4. +135
    -0
      kernel/mips/dgemm_ncopy_4_msa.c
  5. +228
    -0
      kernel/mips/dgemm_ncopy_8_msa.c
  6. +162
    -0
      kernel/mips/dgemm_tcopy_4_msa.c
  7. +317
    -0
      kernel/mips/dgemm_tcopy_8_msa.c
  8. +79
    -0
      kernel/mips/macros_msa.h
  9. +4
    -4
      param.h

+ 3
- 0
CONTRIBUTORS.md View File

@@ -150,3 +150,6 @@ In chronological order:
* theoractice <https://github.com/theoractice/>
* [2016-03-20] Fix compiler error in VisualStudio with CMake
* [2016-03-22] Fix access violation on Windows while static linking

* Shivraj Patil <https://github.com/sva-img/>
* [2016-05-03] DGEMM optimization for MIPS P5600 and I6400 using MSA

+ 7
- 3
kernel/mips/KERNEL.P5600 View File

@@ -91,9 +91,13 @@ SGEMMOTCOPY = ../generic/gemm_tcopy_2.c
SGEMMONCOPYOBJ = sgemm_oncopy.o
SGEMMOTCOPYOBJ = sgemm_otcopy.o

DGEMMKERNEL = ../generic/gemmkernel_2x2.c
DGEMMONCOPY = ../generic/gemm_ncopy_2.c
DGEMMOTCOPY = ../generic/gemm_tcopy_2.c
DGEMMKERNEL = ../mips/dgemm_kernel_8x4_msa.c
DGEMMINCOPY = ../mips/dgemm_ncopy_8_msa.c
DGEMMITCOPY = ../mips/dgemm_tcopy_8_msa.c
DGEMMONCOPY = ../mips/dgemm_ncopy_4_msa.c
DGEMMOTCOPY = ../mips/dgemm_tcopy_4_msa.c
DGEMMINCOPYOBJ = dgemm_incopy.o
DGEMMITCOPYOBJ = dgemm_itcopy.o
DGEMMONCOPYOBJ = dgemm_oncopy.o
DGEMMOTCOPYOBJ = dgemm_otcopy.o



+ 720
- 0
kernel/mips/dgemm_kernel_8x4_msa.c View File

@@ -0,0 +1,720 @@
/*******************************************************************************
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"

int CNAME(BLASLONG m, BLASLONG n, BLASLONG k, FLOAT alpha, FLOAT *A, FLOAT *B,
FLOAT *C, BLASLONG ldc
#ifdef TRMMKERNEL
, BLASLONG offset
#endif
)
{
BLASLONG i, j, l;
FLOAT *pc0, *pc1, *pc2, *pc3;
FLOAT *pa0, *pb0;
FLOAT tmp0, tmp1, tmp2, tmp3;
FLOAT a0;
FLOAT b0, b1, b2, b3;
v2f64 v_alpha = {alpha, alpha};
v2f64 src_a0, src_a1, src_a2, src_a3, src_b, src_b0, src_b1;
v2f64 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;
v2f64 res0, res1, res2, res3, res4, res5, res6, res7;
v2f64 res8, res9, res10, res11, res12, res13, res14, res15;

for (j = (n / 4); j--;)
{
pc0 = C;
pc1 = pc0 + ldc;
pc2 = pc1 + ldc;
pc3 = pc2 + ldc;

pa0 = A;

for (i = (m / 8); i--;)
{
pb0 = B;

LD_DP4(pa0, 2, src_a0, src_a1, src_a2, src_a3);
LD_DP2(pb0, 2, src_b0, src_b1);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 = src_a0 * src_b;
res1 = src_a1 * src_b;
res2 = src_a2 * src_b;
res3 = src_a3 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res4 = src_a0 * src_b;
res5 = src_a1 * src_b;
res6 = src_a2 * src_b;
res7 = src_a3 * src_b;

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
res8 = src_a0 * src_b;
res9 = src_a1 * src_b;
res10 = src_a2 * src_b;
res11 = src_a3 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
res12 = src_a0 * src_b;
res13 = src_a1 * src_b;
res14 = src_a2 * src_b;
res15 = src_a3 * src_b;

pa0 += 8;
pb0 += 4;

for (l = (k - 1); l--;)
{
LD_DP4(pa0, 2, src_a0, src_a1, src_a2, src_a3);
LD_DP2(pb0, 2, src_b0, src_b1);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 += src_a0 * src_b;
res1 += src_a1 * src_b;
res2 += src_a2 * src_b;
res3 += src_a3 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res4 += src_a0 * src_b;
res5 += src_a1 * src_b;
res6 += src_a2 * src_b;
res7 += src_a3 * src_b;

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
res8 += src_a0 * src_b;
res9 += src_a1 * src_b;
res10 += src_a2 * src_b;
res11 += src_a3 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
res12 += src_a0 * src_b;
res13 += src_a1 * src_b;
res14 += src_a2 * src_b;
res15 += src_a3 * src_b;

pa0 += 8;
pb0 += 4;
}

LD_DP4(pc0, 2, dst0, dst1, dst2, dst3);
LD_DP4(pc1, 2, dst4, dst5, dst6, dst7);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;
dst2 += res2 * v_alpha;
dst3 += res3 * v_alpha;
dst4 += res4 * v_alpha;
dst5 += res5 * v_alpha;
dst6 += res6 * v_alpha;
dst7 += res7 * v_alpha;

ST_DP4(dst0, dst1, dst2, dst3, pc0, 2);
ST_DP4(dst4, dst5, dst6, dst7, pc1, 2);

LD_DP4(pc2, 2, dst0, dst1, dst2, dst3);
LD_DP4(pc3, 2, dst4, dst5, dst6, dst7);

dst0 += res8 * v_alpha;
dst1 += res9 * v_alpha;
dst2 += res10 * v_alpha;
dst3 += res11 * v_alpha;
dst4 += res12 * v_alpha;
dst5 += res13 * v_alpha;
dst6 += res14 * v_alpha;
dst7 += res15 * v_alpha;

ST_DP4(dst0, dst1, dst2, dst3, pc2, 2);
ST_DP4(dst4, dst5, dst6, dst7, pc3, 2);

pc0 += 8;
pc1 += 8;
pc2 += 8;
pc3 += 8;
}

for (i = ((m & 4) / 4); i--;)
{
pb0 = B;

LD_DP2(pa0, 2, src_a0, src_a1);
LD_DP2(pb0, 2, src_b0, src_b1);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 = src_a0 * src_b;
res1 = src_a1 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res2 = src_a0 * src_b;
res3 = src_a1 * src_b;

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
res4 = src_a0 * src_b;
res5 = src_a1 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
res6 = src_a0 * src_b;
res7 = src_a1 * src_b;

pa0 += 4;
pb0 += 4;

for (l = (k - 1); l--;)
{
LD_DP2(pa0, 2, src_a0, src_a1);
LD_DP2(pb0, 2, src_b0, src_b1);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 += src_a0 * src_b;
res1 += src_a1 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res2 += src_a0 * src_b;
res3 += src_a1 * src_b;

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
res4 += src_a0 * src_b;
res5 += src_a1 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
res6 += src_a0 * src_b;
res7 += src_a1 * src_b;

pa0 += 4;
pb0 += 4;
}

LD_DP2(pc0, 2, dst0, dst1);
LD_DP2(pc1, 2, dst2, dst3);
LD_DP2(pc2, 2, dst4, dst5);
LD_DP2(pc3, 2, dst6, dst7);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;
dst2 += res2 * v_alpha;
dst3 += res3 * v_alpha;
dst4 += res4 * v_alpha;
dst5 += res5 * v_alpha;
dst6 += res6 * v_alpha;
dst7 += res7 * v_alpha;

ST_DP2(dst0, dst1, pc0, 2);
ST_DP2(dst2, dst3, pc1, 2);
ST_DP2(dst4, dst5, pc2, 2);
ST_DP2(dst6, dst7, pc3, 2);

pc0 += 4;
pc1 += 4;
pc2 += 4;
pc3 += 4;
}

for (i = ((m & 2) / 2); i--;)
{
pb0 = B;

src_a0 = LD_DP(pa0);
LD_DP2(pb0, 2, src_b0, src_b1);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 = src_a0 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res1 = src_a0 * src_b;

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
res2 = src_a0 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
res3 = src_a0 * src_b;

pa0 += 2;
pb0 += 4;

for (l = (k - 1); l--;)
{
src_a0 = LD_DP(pa0);
LD_DP2(pb0, 2, src_b0, src_b1);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 += src_a0 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res1 += src_a0 * src_b;

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b1, (v2i64) src_b1);
res2 += src_a0 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b1, (v2i64) src_b1);
res3 += src_a0 * src_b;

pa0 += 2;
pb0 += 4;
}

dst0 = LD_DP(pc0);
dst1 = LD_DP(pc1);
dst2 = LD_DP(pc2);
dst3 = LD_DP(pc3);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;
dst2 += res2 * v_alpha;
dst3 += res3 * v_alpha;

ST_DP(dst0, pc0);
ST_DP(dst1, pc1);
ST_DP(dst2, pc2);
ST_DP(dst3, pc3);

pc0 += 2;
pc1 += 2;
pc2 += 2;
pc3 += 2;
}

for (i = (m & 1); i--;)
{
pb0 = B;

a0 = pa0[0];
b0 = pb0[0];
tmp0 = a0 * b0;

b1 = pb0[1];
tmp1 = a0 * b1;

b2 = pb0[2];
tmp2 = a0 * b2;

b3 = pb0[3];
tmp3 = a0 * b3;

pa0 += 1;
pb0 += 4;

for (l = (k - 1); l--;)
{
a0 = pa0[0];
b0 = pb0[0];
tmp0 += a0 * b0;

b1 = pb0[1];
tmp1 += a0 * b1;

b2 = pb0[2];
tmp2 += a0 * b2;

b3 = pb0[3];
tmp3 += a0 * b3;

pa0 += 1;
pb0 += 4;
}

tmp0 = alpha * tmp0;
tmp1 = alpha * tmp1;
tmp2 = alpha * tmp2;
tmp3 = alpha * tmp3;

pc0[0] += tmp0;
pc1[0] += tmp1;
pc2[0] += tmp2;
pc3[0] += tmp3;

pc0 += 1;
pc1 += 1;
pc2 += 1;
pc3 += 1;
}

l = (k << 2);
B = B + l;
i = (ldc << 2);
C = C + i;
}

for (j = ((n & 2) / 2); j--;)
{
pc0 = C;
pc1 = pc0 + ldc;

pa0 = A;

for (i = (m / 8); i--;)
{
pb0 = B;

LD_DP4(pa0, 2, src_a0, src_a1, src_a2, src_a3);
src_b0 = LD_DP(pb0);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 = src_a0 * src_b;
res1 = src_a1 * src_b;
res2 = src_a2 * src_b;
res3 = src_a3 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res4 = src_a0 * src_b;
res5 = src_a1 * src_b;
res6 = src_a2 * src_b;
res7 = src_a3 * src_b;

pa0 += 8;
pb0 += 2;

for (l = (k - 1); l--;)
{
LD_DP4(pa0, 2, src_a0, src_a1, src_a2, src_a3);
src_b0 = LD_DP(pb0);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 += src_a0 * src_b;
res1 += src_a1 * src_b;
res2 += src_a2 * src_b;
res3 += src_a3 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res4 += src_a0 * src_b;
res5 += src_a1 * src_b;
res6 += src_a2 * src_b;
res7 += src_a3 * src_b;

pa0 += 8;
pb0 += 2;
}

LD_DP4(pc0, 2, dst0, dst1, dst2, dst3);
LD_DP4(pc1, 2, dst4, dst5, dst6, dst7);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;
dst2 += res2 * v_alpha;
dst3 += res3 * v_alpha;
dst4 += res4 * v_alpha;
dst5 += res5 * v_alpha;
dst6 += res6 * v_alpha;
dst7 += res7 * v_alpha;

ST_DP4(dst0, dst1, dst2, dst3, pc0, 2);
ST_DP4(dst4, dst5, dst6, dst7, pc1, 2);

pc0 += 8;
pc1 += 8;
}

for (i = ((m & 4) / 4); i--;)
{
pb0 = B;

LD_DP2(pa0, 2, src_a0, src_a1);
src_b0 = LD_DP(pb0);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 = src_a0 * src_b;
res1 = src_a1 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res2 = src_a0 * src_b;
res3 = src_a1 * src_b;

pa0 += 4;
pb0 += 2;

for (l = (k - 1); l--;)
{
LD_DP2(pa0, 2, src_a0, src_a1);
src_b0 = LD_DP(pb0);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 += src_a0 * src_b;
res1 += src_a1 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res2 += src_a0 * src_b;
res3 += src_a1 * src_b;

pa0 += 4;
pb0 += 2;
}

LD_DP2(pc0, 2, dst0, dst1);
LD_DP2(pc1, 2, dst2, dst3);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;
dst2 += res2 * v_alpha;
dst3 += res3 * v_alpha;

ST_DP2(dst0, dst1, pc0, 2);
ST_DP2(dst2, dst3, pc1, 2);

pc0 += 4;
pc1 += 4;
}

for (i = ((m & 2) / 2); i--;)
{
pb0 = B;

src_a0 = LD_DP(pa0);
src_b0 = LD_DP(pb0);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 = src_a0 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res1 = src_a0 * src_b;

pa0 += 2;
pb0 += 2;

for (l = (k - 1); l--;)
{
src_a0 = LD_DP(pa0);
src_b0 = LD_DP(pb0);

src_b = (v2f64) __msa_ilvr_d((v2i64) src_b0, (v2i64) src_b0);
res0 += src_a0 * src_b;

src_b = (v2f64) __msa_ilvl_d((v2i64) src_b0, (v2i64) src_b0);
res1 += src_a0 * src_b;

pa0 += 2;
pb0 += 2;
}

dst0 = LD_DP(pc0);
dst1 = LD_DP(pc1);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;

ST_DP(dst0, pc0);
ST_DP(dst1, pc1);

pc0 += 2;
pc1 += 2;
}

for (i = (m & 1); i--;)
{
pb0 = B;

a0 = pa0[0];
b0 = pb0[0];
tmp0 = a0 * b0;

b1 = pb0[1];
tmp1 = a0 * b1;

pa0 += 1;
pb0 += 2;

for (l = (k - 1); l--;)
{
a0 = pa0[0];
b0 = pb0[0];
tmp0 += a0 * b0;

b1 = pb0[1];
tmp1 += a0 * b1;

pa0 += 1;
pb0 += 2;
}

tmp0 = alpha * tmp0;
tmp1 = alpha * tmp1;

pc0[0] += tmp0;
pc1[0] += tmp1;

pc0 += 1;
pc1 += 1;
}

l = (k << 1);
B = B + l;
i = (ldc << 1);
C = C + i;
}

for (j = (n & 1); j--;)
{
pc0 = C;
pa0 = A;

for (i = (m / 8); i--;)
{
pb0 = B;

LD_DP4(pa0, 2, src_a0, src_a1, src_a2, src_a3);
src_b[0] = pb0[0];
src_b[1] = pb0[0];

res0 = src_a0 * src_b;
res1 = src_a1 * src_b;
res2 = src_a2 * src_b;
res3 = src_a3 * src_b;

pa0 += 8;
pb0 += 1;

for (l = (k - 1); l--;)
{
LD_DP4(pa0, 2, src_a0, src_a1, src_a2, src_a3);
src_b[0] = pb0[0];
src_b[1] = pb0[0];

res0 += src_a0 * src_b;
res1 += src_a1 * src_b;
res2 += src_a2 * src_b;
res3 += src_a3 * src_b;

pa0 += 8;
pb0 += 1;
}

LD_DP4(pc0, 2, dst0, dst1, dst2, dst3);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;
dst2 += res2 * v_alpha;
dst3 += res3 * v_alpha;

ST_DP4(dst0, dst1, dst2, dst3, pc0, 2);

pc0 += 8;
}

for (i = ((m & 4) / 4); i--;)
{
pb0 = B;

LD_DP2(pa0, 2, src_a0, src_a1);
src_b[0] = pb0[0];
src_b[1] = pb0[0];

res0 = src_a0 * src_b;
res1 = src_a1 * src_b;

pa0 += 4;
pb0 += 1;

for (l = (k - 1); l--;)
{
LD_DP2(pa0, 2, src_a0, src_a1);
src_b[0] = pb0[0];
src_b[1] = pb0[0];

res0 += src_a0 * src_b;
res1 += src_a1 * src_b;

pa0 += 4;
pb0 += 1;
}

LD_DP2(pc0, 2, dst0, dst1);

dst0 += res0 * v_alpha;
dst1 += res1 * v_alpha;

ST_DP2(dst0, dst1, pc0, 2);

pc0 += 4;
}

for (i = ((m & 2) / 2); i--;)
{
pb0 = B;

src_a0 = LD_DP(pa0);
src_b[0] = pb0[0];
src_b[1] = pb0[0];

res0 = src_a0 * src_b;

pa0 += 2;
pb0 += 1;

for (l = (k - 1); l--;)
{
src_a0 = LD_DP(pa0);
src_b[0] = pb0[0];
src_b[1] = pb0[0];

res0 += src_a0 * src_b;

pa0 += 2;
pb0 += 1;
}

dst0 = LD_DP(pc0);

dst0 += res0 * v_alpha;

ST_DP(dst0, pc0);

pc0 += 2;
}

for (i = (m & 1); i--;)
{
pb0 = B;

a0 = pa0[0];
b0 = pb0[0];
tmp0 = a0 * b0;

pa0 += 1;
pb0 += 1;

for (l = (k - 1); l--;)
{
a0 = pa0[0];
b0 = pb0[0];
tmp0 += a0 * b0;

pa0 += 1;
pb0 += 1;
}

pc0[0] += alpha * tmp0;

pc0 += 1;
}

l = (k << 0);
B = B + l;
i = (ldc << 0);
C = C + i;
}
return 0;
}

+ 135
- 0
kernel/mips/dgemm_ncopy_4_msa.c View File

@@ -0,0 +1,135 @@
/*******************************************************************************
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"

int CNAME(BLASLONG m, BLASLONG n, FLOAT * __restrict src, BLASLONG lda,
FLOAT * __restrict dst)
{
BLASLONG i, j;
FLOAT *psrc0, *psrc1, *psrc2, *psrc3, *psrc4;
FLOAT *pdst;
v2f64 src0, src1, src2, src3, src4, src5, src6, src7;
v2f64 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;

psrc0 = src;
pdst = dst;

for (j = (n >> 2); j--;)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc3 = psrc2 + lda;
psrc4 = psrc3 + lda;
psrc0 += 4 * lda;

for (i = (m >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
LD_DP2(psrc3, 2, src4, src5);
LD_DP2(psrc4, 2, src6, src7);

psrc1 += 4;
psrc2 += 4;
psrc3 += 4;
psrc4 += 4;

dst0 = (v2f64) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src6, (v2i64) src4);
dst2 = (v2f64) __msa_ilvr_d((v2i64) src3, (v2i64) src1);
dst3 = (v2f64) __msa_ilvr_d((v2i64) src7, (v2i64) src5);

dst4 = (v2f64) __msa_ilvl_d((v2i64) src2, (v2i64) src0);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src6, (v2i64) src4);
dst6 = (v2f64) __msa_ilvl_d((v2i64) src3, (v2i64) src1);
dst7 = (v2f64) __msa_ilvl_d((v2i64) src7, (v2i64) src5);

ST_DP8(dst0, dst1, dst4, dst5, dst2, dst3, dst6, dst7, pdst, 2);
pdst += 16;
}

for (i = (m & 3); i--;)
{
*pdst++ = *psrc1++;
*pdst++ = *psrc2++;
*pdst++ = *psrc3++;
*pdst++ = *psrc4++;
}
}

if (n & 2)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc0 += 2 * lda;

for (i = (m >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
psrc1 += 4;
psrc2 += 4;

dst0 = (v2f64) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src3, (v2i64) src1);
dst4 = (v2f64) __msa_ilvl_d((v2i64) src2, (v2i64) src0);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src3, (v2i64) src1);

ST_DP4(dst0, dst4, dst1, dst5, pdst, 2);
pdst += 8;
}

for (i = (m & 3); i--;)
{
*pdst++ = *psrc1++;
*pdst++ = *psrc2++;
}
}

if (n & 1)
{
psrc1 = psrc0;

for (i = (m >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
psrc1 += 4;

ST_DP2(src0, src1, pdst, 2);
pdst += 4;
}

for (i = (m & 3); i--;)
{
*pdst++ = *psrc1++;
}
}

return 0;
}

+ 228
- 0
kernel/mips/dgemm_ncopy_8_msa.c View File

@@ -0,0 +1,228 @@
/*******************************************************************************
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"

int CNAME(BLASLONG m, BLASLONG n, FLOAT * __restrict src, BLASLONG lda,
FLOAT * __restrict dst)
{
BLASLONG i, j;
FLOAT *psrc0, *psrc1, *psrc2, *psrc3, *psrc4;
FLOAT *psrc5, *psrc6, *psrc7, *psrc8;
FLOAT *pdst;
v2f64 src0, src1, src2, src3, src4, src5, src6, src7;
v2f64 src8, src9, src10, src11, src12, src13, src14, src15;
v2f64 dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7;

psrc0 = src;
pdst = dst;

for (j = (n >> 3); j--;)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc3 = psrc2 + lda;
psrc4 = psrc3 + lda;
psrc5 = psrc4 + lda;
psrc6 = psrc5 + lda;
psrc7 = psrc6 + lda;
psrc8 = psrc7 + lda;
psrc0 += 8 * lda;

for (i = (m >> 3); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
LD_DP2(psrc3, 2, src4, src5);
LD_DP2(psrc4, 2, src6, src7);
LD_DP2(psrc5, 2, src8, src9);
LD_DP2(psrc6, 2, src10, src11);
LD_DP2(psrc7, 2, src12, src13);
LD_DP2(psrc8, 2, src14, src15);

dst0 = (v2f64) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src6, (v2i64) src4);
dst2 = (v2f64) __msa_ilvr_d((v2i64) src10, (v2i64) src8);
dst3 = (v2f64) __msa_ilvr_d((v2i64) src14, (v2i64) src12);
dst4 = (v2f64) __msa_ilvl_d((v2i64) src2, (v2i64) src0);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src6, (v2i64) src4);
dst6 = (v2f64) __msa_ilvl_d((v2i64) src10, (v2i64) src8);
dst7 = (v2f64) __msa_ilvl_d((v2i64) src14, (v2i64) src12);

ST_DP8(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, pdst, 2);

dst0 = (v2f64) __msa_ilvr_d((v2i64) src3, (v2i64) src1);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src7, (v2i64) src5);
dst2 = (v2f64) __msa_ilvr_d((v2i64) src11, (v2i64) src9);
dst3 = (v2f64) __msa_ilvr_d((v2i64) src15, (v2i64) src13);
dst4 = (v2f64) __msa_ilvl_d((v2i64) src3, (v2i64) src1);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src7, (v2i64) src5);
dst6 = (v2f64) __msa_ilvl_d((v2i64) src11, (v2i64) src9);
dst7 = (v2f64) __msa_ilvl_d((v2i64) src15, (v2i64) src13);

ST_DP8(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, pdst + 16,
2);

LD_DP2(psrc1 + 4, 2, src0, src1);
LD_DP2(psrc2 + 4, 2, src2, src3);
LD_DP2(psrc3 + 4, 2, src4, src5);
LD_DP2(psrc4 + 4, 2, src6, src7);
LD_DP2(psrc5 + 4, 2, src8, src9);
LD_DP2(psrc6 + 4, 2, src10, src11);
LD_DP2(psrc7 + 4, 2, src12, src13);
LD_DP2(psrc8 + 4, 2, src14, src15);

dst0 = (v2f64) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src6, (v2i64) src4);
dst2 = (v2f64) __msa_ilvr_d((v2i64) src10, (v2i64) src8);
dst3 = (v2f64) __msa_ilvr_d((v2i64) src14, (v2i64) src12);
dst4 = (v2f64) __msa_ilvl_d((v2i64) src2, (v2i64) src0);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src6, (v2i64) src4);
dst6 = (v2f64) __msa_ilvl_d((v2i64) src10, (v2i64) src8);
dst7 = (v2f64) __msa_ilvl_d((v2i64) src14, (v2i64) src12);

ST_DP8(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, pdst + 32,
2);

dst0 = (v2f64) __msa_ilvr_d((v2i64) src3, (v2i64) src1);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src7, (v2i64) src5);
dst2 = (v2f64) __msa_ilvr_d((v2i64) src11, (v2i64) src9);
dst3 = (v2f64) __msa_ilvr_d((v2i64) src15, (v2i64) src13);
dst4 = (v2f64) __msa_ilvl_d((v2i64) src3, (v2i64) src1);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src7, (v2i64) src5);
dst6 = (v2f64) __msa_ilvl_d((v2i64) src11, (v2i64) src9);
dst7 = (v2f64) __msa_ilvl_d((v2i64) src15, (v2i64) src13);

ST_DP8(dst0, dst1, dst2, dst3, dst4, dst5, dst6, dst7, pdst + 48,
2);

psrc1 += 8;
psrc2 += 8;
psrc3 += 8;
psrc4 += 8;
psrc5 += 8;
psrc6 += 8;
psrc7 += 8;
psrc8 += 8;
pdst += 64;
}

for (i = (m & 7); i--;)
{
*pdst++ = *psrc1++;
*pdst++ = *psrc2++;
*pdst++ = *psrc3++;
*pdst++ = *psrc4++;
*pdst++ = *psrc5++;
*pdst++ = *psrc6++;
*pdst++ = *psrc7++;
*pdst++ = *psrc8++;
}
}

if (n & 4)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc3 = psrc2 + lda;
psrc4 = psrc3 + lda;
psrc0 += 4 * lda;

for (i = (m >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
LD_DP2(psrc3, 2, src4, src5);
LD_DP2(psrc4, 2, src6, src7);
psrc1 += 4;
psrc2 += 4;
psrc3 += 4;
psrc4 += 4;

dst0 = (v2f64) __msa_ilvr_d((v2i64) src2, (v2i64) src0);
dst1 = (v2f64) __msa_ilvr_d((v2i64) src6, (v2i64) src4);
dst2 = (v2f64) __msa_ilvr_d((v2i64) src3, (v2i64) src1);
dst3 = (v2f64) __msa_ilvr_d((v2i64) src7, (v2i64) src5);

dst4 = (v2f64) __msa_ilvl_d((v2i64) src2, (v2i64) src0);
dst5 = (v2f64) __msa_ilvl_d((v2i64) src6, (v2i64) src4);
dst6 = (v2f64) __msa_ilvl_d((v2i64) src3, (v2i64) src1);
dst7 = (v2f64) __msa_ilvl_d((v2i64) src7, (v2i64) src5);

ST_DP8(dst0, dst1, dst4, dst5, dst2, dst3, dst6, dst7, pdst, 2);
pdst += 16;
}

for (i = (m & 3); i--;)
{
*pdst++ = *psrc1++;
*pdst++ = *psrc2++;
*pdst++ = *psrc3++;
*pdst++ = *psrc4++;
}
}

if (n & 2)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc0 += 2 * lda;

for (i = (m >> 1); i--;)
{
src0 = LD_DP(psrc1);
src1 = LD_DP(psrc2);
psrc1 += 2;
psrc2 += 2;

dst0 = (v2f64) __msa_ilvr_d((v2i64) src1, (v2i64) src0);
dst1 = (v2f64) __msa_ilvl_d((v2i64) src1, (v2i64) src0);

ST_DP2(dst0, dst1, pdst, 2);
pdst += 4;
}

if (m & 1)
{
*pdst++ = *psrc1++;
*pdst++ = *psrc2++;
}
}

if (n & 1)
{
psrc1 = psrc0;

for (i = m; i--;)
{
*pdst++ = *psrc1++;
}
}

return 0;
}

+ 162
- 0
kernel/mips/dgemm_tcopy_4_msa.c View File

@@ -0,0 +1,162 @@
/*******************************************************************************
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"

int CNAME(BLASLONG m, BLASLONG n, FLOAT * __restrict src, BLASLONG lda,
FLOAT * __restrict dst)
{
BLASLONG i, j;
FLOAT *psrc0, *psrc1, *psrc2, *psrc3, *psrc4;
FLOAT *pdst0, *pdst1, *pdst2, *pdst3;
v2f64 src0, src1, src2, src3, src4, src5, src6, src7;

psrc0 = src;
pdst0 = dst;

pdst2 = dst + m * (n & ~3);
pdst3 = dst + m * (n & ~1);

for (j = (m >> 2); j--;)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc3 = psrc2 + lda;
psrc4 = psrc3 + lda;
psrc0 += 4 * lda;

pdst1 = pdst0;
pdst0 += 16;

for (i = (n >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
LD_DP2(psrc3, 2, src4, src5);
LD_DP2(psrc4, 2, src6, src7);
psrc1 += 4;
psrc2 += 4;
psrc3 += 4;
psrc4 += 4;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst1, 2);
pdst1 += m * 4;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
src1 = LD_DP(psrc2);
src2 = LD_DP(psrc3);
src3 = LD_DP(psrc4);
psrc1 += 2;
psrc2 += 2;
psrc3 += 2;
psrc4 += 2;

ST_DP4(src0, src1, src2, src3, pdst2, 2);
pdst2 += 8;
}

if (n & 1)
{
*pdst3++ = *psrc1++;
*pdst3++ = *psrc2++;
*pdst3++ = *psrc3++;
*pdst3++ = *psrc4++;
}
}

if (m & 2)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc0 += 2 * lda;

pdst1 = pdst0;
pdst0 += 8;

for (i = (n >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
psrc1 += 4;
psrc2 += 4;

ST_DP4(src0, src1, src2, src3, pdst1, 2);
pdst1 += m * 4;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
src1 = LD_DP(psrc2);
psrc1 += 2;
psrc2 += 2;

ST_DP2(src0, src1, pdst2, 2);
pdst2 += 4;
}

if (n & 1)
{
*pdst3++ = *psrc1++;
*pdst3++ = *psrc2++;
}
}

if (m & 1)
{
psrc1 = psrc0;
pdst1 = pdst0;

for (i = (n >> 2); i--;)
{
LD_DP2(psrc1, 2, src0, src1);
psrc1 += 4;

ST_DP2(src0, src1, pdst1, 2);
pdst1 += 4 * m;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
psrc1 += 2;

ST_DP(src0, pdst2);
}

if (n & 1)
{
*pdst3 = *psrc1;
}
}

return 0;
}

+ 317
- 0
kernel/mips/dgemm_tcopy_8_msa.c View File

@@ -0,0 +1,317 @@
/*******************************************************************************
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"

int CNAME(BLASLONG m, BLASLONG n, FLOAT * __restrict src, BLASLONG lda,
FLOAT * __restrict dst)
{
BLASLONG i, j;
FLOAT *psrc0, *psrc1, *psrc2, *psrc3, *psrc4;
FLOAT *psrc5, *psrc6, *psrc7, *psrc8;
FLOAT *pdst0, *pdst1, *pdst2, *pdst3, *pdst4;
v2f64 src0, src1, src2, src3, src4, src5, src6, src7;
v2f64 src8, src9, src10, src11, src12, src13, src14, src15;

psrc0 = src;
pdst0 = dst;

pdst2 = dst + m * (n & ~7);
pdst3 = dst + m * (n & ~3);
pdst4 = dst + m * (n & ~1);

for (j = (m >> 3); j--;)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc3 = psrc2 + lda;
psrc4 = psrc3 + lda;
psrc5 = psrc4 + lda;
psrc6 = psrc5 + lda;
psrc7 = psrc6 + lda;
psrc8 = psrc7 + lda;
psrc0 += 8 * lda;

pdst1 = pdst0;
pdst0 += 64;

for (i = (n >> 3); i--;)
{
LD_DP4(psrc1, 2, src0, src1, src2, src3);
LD_DP4(psrc2, 2, src4, src5, src6, src7);
LD_DP4(psrc3, 2, src8, src9, src10, src11);
LD_DP4(psrc4, 2, src12, src13, src14, src15);
psrc1 += 8;
psrc2 += 8;
psrc3 += 8;
psrc4 += 8;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst1, 2);
ST_DP8(src8, src9, src10, src11, src12, src13, src14, src15,
pdst1 + 16, 2);

LD_DP4(psrc5, 2, src0, src1, src2, src3);
LD_DP4(psrc6, 2, src4, src5, src6, src7);
LD_DP4(psrc7, 2, src8, src9, src10, src11);
LD_DP4(psrc8, 2, src12, src13, src14, src15);
psrc5 += 8;
psrc6 += 8;
psrc7 += 8;
psrc8 += 8;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst1 + 32,
2);
ST_DP8(src8, src9, src10, src11, src12, src13, src14, src15,
pdst1 + 48, 2);
pdst1 += m * 8;
}

if (n & 4)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
LD_DP2(psrc3, 2, src4, src5);
LD_DP2(psrc4, 2, src6, src7);
LD_DP2(psrc5, 2, src8, src9);
LD_DP2(psrc6, 2, src10, src11);
LD_DP2(psrc7, 2, src12, src13);
LD_DP2(psrc8, 2, src14, src15);
psrc1 += 4;
psrc2 += 4;
psrc3 += 4;
psrc4 += 4;
psrc5 += 4;
psrc6 += 4;
psrc7 += 4;
psrc8 += 4;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst2, 2);
ST_DP8(src8, src9, src10, src11, src12, src13, src14, src15,
pdst2 + 16, 2);
pdst2 += 32;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
src1 = LD_DP(psrc2);
src2 = LD_DP(psrc3);
src3 = LD_DP(psrc4);
src4 = LD_DP(psrc5);
src5 = LD_DP(psrc6);
src6 = LD_DP(psrc7);
src7 = LD_DP(psrc8);
psrc1 += 2;
psrc2 += 2;
psrc3 += 2;
psrc4 += 2;
psrc5 += 2;
psrc6 += 2;
psrc7 += 2;
psrc8 += 2;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst3, 2);
pdst3 += 16;
}

if (n & 1)
{
*pdst4++ = *psrc1++;
*pdst4++ = *psrc2++;
*pdst4++ = *psrc3++;
*pdst4++ = *psrc4++;
*pdst4++ = *psrc5++;
*pdst4++ = *psrc6++;
*pdst4++ = *psrc7++;
*pdst4++ = *psrc8++;
}
}

if (m & 4)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc3 = psrc2 + lda;
psrc4 = psrc3 + lda;
psrc0 += 4 * lda;

pdst1 = pdst0;
pdst0 += 32;

for (i = (n >> 3); i--;)
{
LD_DP4(psrc1, 2, src0, src1, src2, src3);
LD_DP4(psrc2, 2, src4, src5, src6, src7);
LD_DP4(psrc3, 2, src8, src9, src10, src11);
LD_DP4(psrc4, 2, src12, src13, src14, src15);
psrc1 += 8;
psrc2 += 8;
psrc3 += 8;
psrc4 += 8;
psrc5 += 8;
psrc6 += 8;
psrc7 += 8;
psrc8 += 8;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst1, 2);
ST_DP8(src8, src9, src10, src11, src12, src13, src14, src15,
pdst1 + 16, 2);
pdst1 += 8 * m;
}

if (n & 4)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
LD_DP2(psrc3, 2, src4, src5);
LD_DP2(psrc4, 2, src6, src7);
psrc1 += 4;
psrc2 += 4;
psrc3 += 4;
psrc4 += 4;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst2, 2);
pdst2 += 16;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
src1 = LD_DP(psrc2);
src2 = LD_DP(psrc3);
src3 = LD_DP(psrc4);
psrc1 += 2;
psrc2 += 2;
psrc3 += 2;
psrc4 += 2;

ST_DP4(src0, src1, src2, src3, pdst3, 2);
pdst3 += 8;
}

if (n & 1)
{
*pdst4++ = *psrc1++;
*pdst4++ = *psrc2++;
*pdst4++ = *psrc3++;
*pdst4++ = *psrc4++;
}
}

if (m & 2)
{
psrc1 = psrc0;
psrc2 = psrc1 + lda;
psrc0 += 2 * lda;

pdst1 = pdst0;
pdst0 += 16;

for (i = (n >> 3); i--;)
{
LD_DP4(psrc1, 2, src0, src1, src2, src3);
LD_DP4(psrc2, 2, src4, src5, src6, src7);
psrc1 += 8;
psrc2 += 8;

ST_DP8(src0, src1, src2, src3, src4, src5, src6, src7, pdst1, 2);
pdst1 += 8 * m;
}

if (n & 4)
{
LD_DP2(psrc1, 2, src0, src1);
LD_DP2(psrc2, 2, src2, src3);
psrc1 += 4;
psrc2 += 4;

ST_DP4(src0, src1, src2, src3, pdst2, 2);
pdst2 += 8;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
src1 = LD_DP(psrc2);
psrc1 += 2;
psrc2 += 2;

ST_DP2(src0, src1, pdst3, 2);
pdst3 += 4;
}

if (n & 1)
{
*pdst4++ = *psrc1++;
*pdst4++ = *psrc2++;
}
}

if (m & 1)
{
psrc1 = psrc0;
psrc0 += lda;

pdst1 = pdst0;
pdst0 += 8;

for (i = (n >> 3); i--;)
{
LD_DP4(psrc1, 2, src0, src1, src2, src3);
psrc1 += 8;

ST_DP4(src0, src1, src2, src3, pdst1, 2);
pdst1 += 8 * m;
}

if (n & 4)
{
LD_DP2(psrc1, 2, src0, src1);
psrc1 += 4;

ST_DP2(src0, src1, pdst2, 2);
pdst2 += 4;
}

if (n & 2)
{
src0 = LD_DP(psrc1);
psrc1 += 2;

ST_DP(src0, pdst3);
pdst3 += 2;
}

if (n & 1)
{
*pdst4++ = *psrc1++;
}
}

return 0;
}

+ 79
- 0
kernel/mips/macros_msa.h View File

@@ -0,0 +1,79 @@
/*******************************************************************************
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.
*******************************************************************************/

#ifndef __MACROS_MSA_H__
#define __MACROS_MSA_H__

#include <msa.h>

#define LD_D(RTYPE, psrc) *((RTYPE *)(psrc))
#define LD_DP(...) LD_D(v2f64, __VA_ARGS__)

#define ST_D(RTYPE, in, pdst) *((RTYPE *)(pdst)) = (in)
#define ST_DP(...) ST_D(v2f64, __VA_ARGS__)

/* Description : Load 2 vectors of double precision floating point elements with stride
Arguments : Inputs - psrc, stride
Outputs - out0, out1
Return Type - double precision floating point
*/
#define LD_DP2(psrc, stride, out0, out1) \
{ \
out0 = LD_DP((psrc)); \
out1 = LD_DP((psrc) + stride); \
}

#define LD_DP4(psrc, stride, out0, out1, out2, out3) \
{ \
LD_DP2(psrc, stride, out0, out1) \
LD_DP2(psrc + 2 * stride, stride, out2, out3) \
}

/* Description : Store vectors of double precision floating point elements with stride
Arguments : Inputs - in0, in1, pdst, stride
Details : Store 2 double precision floating point elements from 'in0' to (pdst)
Store 2 double precision floating point elements from 'in1' to (pdst + stride)
*/
#define ST_DP2(in0, in1, pdst, stride) \
{ \
ST_DP(in0, (pdst)); \
ST_DP(in1, (pdst) + stride); \
}

#define ST_DP4(in0, in1, in2, in3, pdst, stride) \
{ \
ST_DP2(in0, in1, (pdst), stride); \
ST_DP2(in2, in3, (pdst) + 2 * stride, stride); \
}

#define ST_DP8(in0, in1, in2, in3, in4, in5, in6, in7, pdst, stride) \
{ \
ST_DP4(in0, in1, in2, in3, (pdst), stride); \
ST_DP4(in4, in5, in6, in7, (pdst) + 4 * stride, stride); \
}

#endif /* __MACROS_MSA_H__ */

+ 4
- 4
param.h View File

@@ -2185,8 +2185,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define SGEMM_DEFAULT_UNROLL_M 2
#define SGEMM_DEFAULT_UNROLL_N 2

#define DGEMM_DEFAULT_UNROLL_M 2
#define DGEMM_DEFAULT_UNROLL_N 2
#define DGEMM_DEFAULT_UNROLL_M 8
#define DGEMM_DEFAULT_UNROLL_N 4

#define CGEMM_DEFAULT_UNROLL_M 2
#define CGEMM_DEFAULT_UNROLL_N 2
@@ -2224,8 +2224,8 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define SGEMM_DEFAULT_UNROLL_M 2
#define SGEMM_DEFAULT_UNROLL_N 2

#define DGEMM_DEFAULT_UNROLL_M 2
#define DGEMM_DEFAULT_UNROLL_N 2
#define DGEMM_DEFAULT_UNROLL_M 8
#define DGEMM_DEFAULT_UNROLL_N 4

#define CGEMM_DEFAULT_UNROLL_M 2
#define CGEMM_DEFAULT_UNROLL_N 2


Write Preview
Loading…
Cancel
Save