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

Merge pull request #572 from wernsaar/develop 

added optimized cscal and zscal functions for steamroller
tags/v0.2.15^2
wernsaar 10 years ago
parent
685be40339 24f58c8bb1
commit
7fb718a7d8
9 changed files with 1064 additions and 1 deletions
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +1
    -0
      kernel/x86_64/KERNEL.BULLDOZER
  2. +3
    -0
      kernel/x86_64/KERNEL.PILEDRIVER
  3. +1
    -0
      kernel/x86_64/KERNEL.SANDYBRIDGE
  4. +4
    -0
      kernel/x86_64/KERNEL.STEAMROLLER
  5. +6
    -0
      kernel/x86_64/cscal.c
  6. +348
    -0
      kernel/x86_64/cscal_microk_bulldozer-2.c
  7. +349
    -0
      kernel/x86_64/cscal_microk_steamroller-2.c
  8. +3
    -1
      kernel/x86_64/zscal.c
  9. +349
    -0
      kernel/x86_64/zscal_microk_steamroller-2.c

+ 1
- 0
kernel/x86_64/KERNEL.BULLDOZER View File

@@ -1,4 +1,5 @@
DSCALKERNEL = dscal.c
CSCALKERNEL = cscal.c
ZSCALKERNEL = zscal.c

DAXPYKERNEL = daxpy.c


+ 3
- 0
kernel/x86_64/KERNEL.PILEDRIVER View File

@@ -1,4 +1,7 @@
DSCALKERNEL = dscal.c
CSCALKERNEL = cscal.c
ZSCALKERNEL = zscal.c


SAXPYKERNEL = saxpy.c
DAXPYKERNEL = daxpy.c


+ 1
- 0
kernel/x86_64/KERNEL.SANDYBRIDGE View File

@@ -1,4 +1,5 @@
DSCALKERNEL = dscal.c
CSCALKERNEL = cscal.c

SGERKERNEL = sger.c
DGERKERNEL = dger.c


+ 4
- 0
kernel/x86_64/KERNEL.STEAMROLLER View File

@@ -1,3 +1,7 @@
DSCALKERNEL = dscal.c
CSCALKERNEL = cscal.c
ZSCALKERNEL = zscal.c

SAXPYKERNEL = saxpy.c
DAXPYKERNEL = daxpy.c
CAXPYKERNEL = caxpy.c


+ 6
- 0
kernel/x86_64/cscal.c View File

@@ -30,6 +30,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

#if defined(HASWELL)
#include "cscal_microk_haswell-2.c"
#elif defined(BULLDOZER) || defined(PILEDRIVER)
#include "cscal_microk_bulldozer-2.c"
#elif defined(STEAMROLLER)
#include "cscal_microk_steamroller-2.c"
#elif defined(SANDYBRIDGE)
#include "cscal_microk_bulldozer-2.c"
#endif




+ 348
- 0
kernel/x86_64/cscal_microk_bulldozer-2.c View File

@@ -0,0 +1,348 @@
/***************************************************************************
Copyright (c) 2014-2015, 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.
*****************************************************************************/


#define HAVE_KERNEL_16 1

static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vbroadcastss (%2), %%xmm0 \n\t" // da_r
"vbroadcastss 4(%2), %%xmm1 \n\t" // da_i

"addq $128, %1 \n\t"

"vmovups -128(%1), %%xmm4 \n\t"
"vmovups -112(%1), %%xmm5 \n\t"
"vmovups -96(%1), %%xmm6 \n\t"
"vmovups -80(%1), %%xmm7 \n\t"

"vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t"
"vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t"
"vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t"
"vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t"

"subq $8 , %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

"prefetcht0 320(%1) \n\t"
// ".align 2 \n\t"

"vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1
"vmovups -64(%1), %%xmm4 \n\t"
"vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t"
"vmovups -48(%1), %%xmm5 \n\t"
"vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t"
"vmovups -32(%1), %%xmm6 \n\t"
"vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t"
"vmovups -16(%1), %%xmm7 \n\t"

"vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%xmm12 , %%xmm8 , %%xmm8 \n\t"
"vmulps %%xmm1, %%xmm13, %%xmm13 \n\t"
"vaddsubps %%xmm13 , %%xmm9 , %%xmm9 \n\t"
"vmulps %%xmm1, %%xmm14, %%xmm14 \n\t"
"vaddsubps %%xmm14 , %%xmm10, %%xmm10 \n\t"
"vmulps %%xmm1, %%xmm15, %%xmm15 \n\t"
"vaddsubps %%xmm15 , %%xmm11, %%xmm11 \n\t"

"vmovups %%xmm8 , -128(%1) \n\t"
"vmovups %%xmm9 , -112(%1) \n\t"
"vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t"
"vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t"
"vmovups %%xmm10, -96(%1) \n\t"
"vmovups %%xmm11, -80(%1) \n\t"
"vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t"
"vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t"

"addq $64 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"


"vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1
"vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t"
"vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t"
"vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t"

"vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%xmm12 , %%xmm8 , %%xmm8 \n\t"
"vmulps %%xmm1, %%xmm13, %%xmm13 \n\t"
"vaddsubps %%xmm13 , %%xmm9 , %%xmm9 \n\t"
"vmulps %%xmm1, %%xmm14, %%xmm14 \n\t"
"vaddsubps %%xmm14 , %%xmm10, %%xmm10 \n\t"
"vmulps %%xmm1, %%xmm15, %%xmm15 \n\t"
"vaddsubps %%xmm15 , %%xmm11, %%xmm11 \n\t"

"vmovups %%xmm8 , -128(%1) \n\t"
"vmovups %%xmm9 , -112(%1) \n\t"
"vmovups %%xmm10, -96(%1) \n\t"
"vmovups %%xmm11, -80(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}

static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vxorps %%xmm0, %%xmm0, %%xmm0 \n\t"
"vbroadcastss 4(%2), %%xmm1 \n\t" // da_i

"addq $128, %1 \n\t"

"vmovups -128(%1), %%xmm4 \n\t"
"vmovups -112(%1), %%xmm5 \n\t"
"vmovups -96(%1), %%xmm6 \n\t"
"vmovups -80(%1), %%xmm7 \n\t"

"vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t"
"vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t"
"vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t"
"vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t"

"subq $8 , %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmovups -64(%1), %%xmm4 \n\t"
"vmovups -48(%1), %%xmm5 \n\t"
"vmovups -32(%1), %%xmm6 \n\t"
"vmovups -16(%1), %%xmm7 \n\t"

"vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%xmm12 , %%xmm0 , %%xmm8 \n\t"
"vmulps %%xmm1, %%xmm13, %%xmm13 \n\t"
"vaddsubps %%xmm13 , %%xmm0 , %%xmm9 \n\t"
"vmulps %%xmm1, %%xmm14, %%xmm14 \n\t"
"vaddsubps %%xmm14 , %%xmm0 , %%xmm10 \n\t"
"vmulps %%xmm1, %%xmm15, %%xmm15 \n\t"
"vaddsubps %%xmm15 , %%xmm0 , %%xmm11 \n\t"

"vmovups %%xmm8 , -128(%1) \n\t"
"vpermilps $0xb1 , %%xmm4, %%xmm12 \n\t"
"vmovups %%xmm9 , -112(%1) \n\t"
"vpermilps $0xb1 , %%xmm5, %%xmm13 \n\t"
"vmovups %%xmm10, -96(%1) \n\t"
"vpermilps $0xb1 , %%xmm6, %%xmm14 \n\t"
"vmovups %%xmm11, -80(%1) \n\t"
"vpermilps $0xb1 , %%xmm7, %%xmm15 \n\t"

"addq $64 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"

"vmulps %%xmm1, %%xmm12, %%xmm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%xmm12 , %%xmm0 , %%xmm8 \n\t"
"vmulps %%xmm1, %%xmm13, %%xmm13 \n\t"
"vaddsubps %%xmm13 , %%xmm0 , %%xmm9 \n\t"
"vmulps %%xmm1, %%xmm14, %%xmm14 \n\t"
"vaddsubps %%xmm14 , %%xmm0 , %%xmm10 \n\t"
"vmulps %%xmm1, %%xmm15, %%xmm15 \n\t"
"vaddsubps %%xmm15 , %%xmm0 , %%xmm11 \n\t"

"vmovups %%xmm8 , -128(%1) \n\t"
"vmovups %%xmm9 , -112(%1) \n\t"
"vmovups %%xmm10, -96(%1) \n\t"
"vmovups %%xmm11, -80(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}



static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vbroadcastss (%2), %%xmm0 \n\t" // da_r

"addq $128, %1 \n\t"

"vmovups -128(%1), %%xmm4 \n\t"
"vmovups -112(%1), %%xmm5 \n\t"
"vmovups -96(%1), %%xmm6 \n\t"
"vmovups -80(%1), %%xmm7 \n\t"


"subq $8 , %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1
"vmovups -64(%1), %%xmm4 \n\t"
"vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t"
"vmovups -48(%1), %%xmm5 \n\t"
"vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t"
"vmovups -32(%1), %%xmm6 \n\t"
"vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t"
"vmovups -16(%1), %%xmm7 \n\t"

"vmovups %%xmm8 , -128(%1) \n\t"
"vmovups %%xmm9 , -112(%1) \n\t"
"vmovups %%xmm10, -96(%1) \n\t"
"vmovups %%xmm11, -80(%1) \n\t"

"addq $64 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"


"vmulps %%xmm0, %%xmm4 , %%xmm8 \n\t" // da_r*x0 , da_r *x1
"vmulps %%xmm0, %%xmm5 , %%xmm9 \n\t"
"vmulps %%xmm0, %%xmm6 , %%xmm10 \n\t"
"vmulps %%xmm0, %%xmm7 , %%xmm11 \n\t"

"vmovups %%xmm8 , -128(%1) \n\t"
"vmovups %%xmm9 , -112(%1) \n\t"
"vmovups %%xmm10, -96(%1) \n\t"
"vmovups %%xmm11, -80(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}


static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vxorps %%xmm0, %%xmm0, %%xmm0 \n\t"

"addq $128, %1 \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmovups %%xmm0 , -128(%1) \n\t"
"vmovups %%xmm0 , -112(%1) \n\t"
"vmovups %%xmm0 , -96(%1) \n\t"
"vmovups %%xmm0 , -80(%1) \n\t"

"addq $64 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}




+ 349
- 0
kernel/x86_64/cscal_microk_steamroller-2.c View File

@@ -0,0 +1,349 @@
/***************************************************************************
Copyright (c) 2014-2015, 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.
*****************************************************************************/


#define HAVE_KERNEL_16 1

static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vbroadcastss (%2), %%ymm0 \n\t" // da_r
"vbroadcastss 4(%2), %%ymm1 \n\t" // da_i

"addq $128, %1 \n\t"

"vmovups -128(%1), %%ymm4 \n\t"
"vmovups -96(%1), %%ymm5 \n\t"
"vmovups -64(%1), %%ymm6 \n\t"
"vmovups -32(%1), %%ymm7 \n\t"

"vpermilps $0xb1 , %%ymm4, %%ymm12 \n\t"
"vpermilps $0xb1 , %%ymm5, %%ymm13 \n\t"
"vpermilps $0xb1 , %%ymm6, %%ymm14 \n\t"
"vpermilps $0xb1 , %%ymm7, %%ymm15 \n\t"

"subq $16, %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

"prefetcht0 512(%1) \n\t"
// ".align 2 \n\t"

"vmulps %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmovups 0(%1), %%ymm4 \n\t"
"vmulps %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmovups 32(%1), %%ymm5 \n\t"
"prefetcht0 768(%1) \n\t"
"vmulps %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmovups 64(%1), %%ymm6 \n\t"
"vmulps %%ymm0, %%ymm7 , %%ymm11 \n\t"
"vmovups 96(%1), %%ymm7 \n\t"

"vmulps %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vmulps %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubps %%ymm12 , %%ymm8 , %%ymm8 \n\t"
"vaddsubps %%ymm13 , %%ymm9 , %%ymm9 \n\t"
"vmulps %%ymm1, %%ymm14, %%ymm14 \n\t"
"vmulps %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubps %%ymm14 , %%ymm10, %%ymm10 \n\t"
"vaddsubps %%ymm15 , %%ymm11, %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"
"vpermilps $0xb1 , %%ymm4, %%ymm12 \n\t"
"vpermilps $0xb1 , %%ymm5, %%ymm13 \n\t"
"vpermilps $0xb1 , %%ymm6, %%ymm14 \n\t"
"vpermilps $0xb1 , %%ymm7, %%ymm15 \n\t"

"addq $128 ,%1 \n\t"
"subq $16, %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"


"vmulps %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmulps %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmulps %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmulps %%ymm0, %%ymm7 , %%ymm11 \n\t"

"vmulps %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%ymm12 , %%ymm8 , %%ymm8 \n\t"
"vmulps %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubps %%ymm13 , %%ymm9 , %%ymm9 \n\t"
"vmulps %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubps %%ymm14 , %%ymm10, %%ymm10 \n\t"
"vmulps %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubps %%ymm15 , %%ymm11, %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "0", "1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}

static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vxorps %%ymm0, %%ymm0, %%ymm0 \n\t"
"vbroadcastss 4(%2), %%ymm1 \n\t" // da_i

"addq $128, %1 \n\t"

"vmovups -128(%1), %%ymm4 \n\t"
"vmovups -96(%1), %%ymm5 \n\t"
"vmovups -64(%1), %%ymm6 \n\t"
"vmovups -32(%1), %%ymm7 \n\t"

"vpermilps $0xb1 , %%ymm4, %%ymm12 \n\t"
"vpermilps $0xb1 , %%ymm5, %%ymm13 \n\t"
"vpermilps $0xb1 , %%ymm6, %%ymm14 \n\t"
"vpermilps $0xb1 , %%ymm7, %%ymm15 \n\t"

"subq $16, %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmovups 0(%1), %%ymm4 \n\t"
"vmovups 32(%1), %%ymm5 \n\t"
"vmovups 64(%1), %%ymm6 \n\t"
"vmovups 96(%1), %%ymm7 \n\t"

"vmulps %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%ymm12 , %%ymm0 , %%ymm8 \n\t"
"vmulps %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubps %%ymm13 , %%ymm0 , %%ymm9 \n\t"
"vmulps %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubps %%ymm14 , %%ymm0 , %%ymm10 \n\t"
"vmulps %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubps %%ymm15 , %%ymm0 , %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vpermilps $0xb1 , %%ymm4, %%ymm12 \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vpermilps $0xb1 , %%ymm5, %%ymm13 \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vpermilps $0xb1 , %%ymm6, %%ymm14 \n\t"
"vmovups %%ymm11, -32(%1) \n\t"
"vpermilps $0xb1 , %%ymm7, %%ymm15 \n\t"

"addq $128 ,%1 \n\t"
"subq $16, %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"

"vmulps %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubps %%ymm12 , %%ymm0 , %%ymm8 \n\t"
"vmulps %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubps %%ymm13 , %%ymm0 , %%ymm9 \n\t"
"vmulps %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubps %%ymm14 , %%ymm0 , %%ymm10 \n\t"
"vmulps %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubps %%ymm15 , %%ymm0 , %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "0", "1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}



static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vbroadcastss (%2), %%ymm0 \n\t" // da_r

"addq $128, %1 \n\t"

"vmovups -128(%1), %%ymm4 \n\t"
"vmovups -96(%1), %%ymm5 \n\t"
"vmovups -64(%1), %%ymm6 \n\t"
"vmovups -32(%1), %%ymm7 \n\t"


"subq $16, %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmulps %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmovups 0(%1), %%ymm4 \n\t"
"vmulps %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmovups 32(%1), %%ymm5 \n\t"
"vmulps %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmovups 64(%1), %%ymm6 \n\t"
"vmulps %%ymm0, %%ymm7 , %%ymm11 \n\t"
"vmovups 96(%1), %%ymm7 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"addq $128 ,%1 \n\t"
"subq $16, %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"


"vmulps %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmulps %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmulps %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmulps %%ymm0, %%ymm7 , %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}


static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void cscal_kernel_16_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vxorps %%ymm0, %%ymm0, %%ymm0 \n\t"

"addq $128, %1 \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmovups %%ymm0 , -128(%1) \n\t"
"vmovups %%ymm0 , -96(%1) \n\t"
"vmovups %%ymm0 , -64(%1) \n\t"
"vmovups %%ymm0 , -32(%1) \n\t"

"addq $128 ,%1 \n\t"
"subq $16, %0 \n\t"
"jnz 1b \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "0", "1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}




+ 3
- 1
kernel/x86_64/zscal.c View File

@@ -30,8 +30,10 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

#if defined(HASWELL)
#include "zscal_microk_haswell-2.c"
#elif defined(BULLDOZER)
#elif defined(BULLDOZER) || defined(PILEDRIVER)
#include "zscal_microk_bulldozer-2.c"
#elif defined(STEAMROLLER)
#include "zscal_microk_steamroller-2.c"
#endif




+ 349
- 0
kernel/x86_64/zscal_microk_steamroller-2.c View File

@@ -0,0 +1,349 @@
/***************************************************************************
Copyright (c) 2014-2015, 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.
*****************************************************************************/


#define HAVE_KERNEL_8 1

static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void zscal_kernel_8( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vbroadcastsd (%2), %%ymm0 \n\t" // da_r
"vbroadcastsd 8(%2), %%ymm1 \n\t" // da_i

"addq $128, %1 \n\t"

"vmovups -128(%1), %%ymm4 \n\t"
"vmovups -96(%1), %%ymm5 \n\t"
"vmovups -64(%1), %%ymm6 \n\t"
"vmovups -32(%1), %%ymm7 \n\t"

"vpermilpd $0x05 , %%ymm4, %%ymm12 \n\t"
"vpermilpd $0x05 , %%ymm5, %%ymm13 \n\t"
"vpermilpd $0x05 , %%ymm6, %%ymm14 \n\t"
"vpermilpd $0x05 , %%ymm7, %%ymm15 \n\t"

"subq $8 , %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

"prefetcht0 320(%1) \n\t"
"prefetcht0 384(%1) \n\t"
// ".align 2 \n\t"

"vmulpd %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmovups 0(%1), %%ymm4 \n\t"
"vmulpd %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmovups 32(%1), %%ymm5 \n\t"
"vmulpd %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmovups 64(%1), %%ymm6 \n\t"
"vmulpd %%ymm0, %%ymm7 , %%ymm11 \n\t"
"vmovups 96(%1), %%ymm7 \n\t"

"vmulpd %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubpd %%ymm12 , %%ymm8 , %%ymm8 \n\t"
"vmulpd %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubpd %%ymm13 , %%ymm9 , %%ymm9 \n\t"
"vmulpd %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubpd %%ymm14 , %%ymm10, %%ymm10 \n\t"
"vmulpd %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubpd %%ymm15 , %%ymm11, %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vpermilpd $0x05 , %%ymm4, %%ymm12 \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vpermilpd $0x05 , %%ymm5, %%ymm13 \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vpermilpd $0x05 , %%ymm6, %%ymm14 \n\t"
"vmovups %%ymm11, -32(%1) \n\t"
"vpermilpd $0x05 , %%ymm7, %%ymm15 \n\t"

"addq $128 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"


"vmulpd %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmulpd %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmulpd %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmulpd %%ymm0, %%ymm7 , %%ymm11 \n\t"

"vmulpd %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubpd %%ymm12 , %%ymm8 , %%ymm8 \n\t"
"vmulpd %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubpd %%ymm13 , %%ymm9 , %%ymm9 \n\t"
"vmulpd %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubpd %%ymm14 , %%ymm10, %%ymm10 \n\t"
"vmulpd %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubpd %%ymm15 , %%ymm11, %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}

static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void zscal_kernel_8_zero_r( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vxorpd %%ymm0, %%ymm0, %%ymm0 \n\t"
"vbroadcastsd 8(%2), %%ymm1 \n\t" // da_i

"addq $128, %1 \n\t"

"vmovups -128(%1), %%ymm4 \n\t"
"vmovups -96(%1), %%ymm5 \n\t"
"vmovups -64(%1), %%ymm6 \n\t"
"vmovups -32(%1), %%ymm7 \n\t"

"vpermilpd $0x05 , %%ymm4, %%ymm12 \n\t"
"vpermilpd $0x05 , %%ymm5, %%ymm13 \n\t"
"vpermilpd $0x05 , %%ymm6, %%ymm14 \n\t"
"vpermilpd $0x05 , %%ymm7, %%ymm15 \n\t"

"subq $8 , %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmovups 0(%1), %%ymm4 \n\t"
"vmovups 32(%1), %%ymm5 \n\t"
"vmovups 64(%1), %%ymm6 \n\t"
"vmovups 96(%1), %%ymm7 \n\t"

"vmulpd %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubpd %%ymm12 , %%ymm0 , %%ymm8 \n\t"
"vmulpd %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubpd %%ymm13 , %%ymm0 , %%ymm9 \n\t"
"vmulpd %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubpd %%ymm14 , %%ymm0 , %%ymm10 \n\t"
"vmulpd %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubpd %%ymm15 , %%ymm0 , %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vpermilpd $0x05 , %%ymm4, %%ymm12 \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vpermilpd $0x05 , %%ymm5, %%ymm13 \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vpermilpd $0x05 , %%ymm6, %%ymm14 \n\t"
"vmovups %%ymm11, -32(%1) \n\t"
"vpermilpd $0x05 , %%ymm7, %%ymm15 \n\t"

"addq $128 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"

"vmulpd %%ymm1, %%ymm12, %%ymm12 \n\t" // da_i*x1 , da_i *x0
"vaddsubpd %%ymm12 , %%ymm0 , %%ymm8 \n\t"
"vmulpd %%ymm1, %%ymm13, %%ymm13 \n\t"
"vaddsubpd %%ymm13 , %%ymm0 , %%ymm9 \n\t"
"vmulpd %%ymm1, %%ymm14, %%ymm14 \n\t"
"vaddsubpd %%ymm14 , %%ymm0 , %%ymm10 \n\t"
"vmulpd %%ymm1, %%ymm15, %%ymm15 \n\t"
"vaddsubpd %%ymm15 , %%ymm0 , %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}



static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void zscal_kernel_8_zero_i( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vbroadcastsd (%2), %%ymm0 \n\t" // da_r

"addq $128, %1 \n\t"

"vmovups -128(%1), %%ymm4 \n\t"
"vmovups -96(%1), %%ymm5 \n\t"
"vmovups -64(%1), %%ymm6 \n\t"
"vmovups -32(%1), %%ymm7 \n\t"


"subq $8 , %0 \n\t"
"jz 2f \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmulpd %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmovups 0(%1), %%ymm4 \n\t"
"vmulpd %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmovups 32(%1), %%ymm5 \n\t"
"vmulpd %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmovups 64(%1), %%ymm6 \n\t"
"vmulpd %%ymm0, %%ymm7 , %%ymm11 \n\t"
"vmovups 96(%1), %%ymm7 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"addq $128 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"2: \n\t"


"vmulpd %%ymm0, %%ymm4 , %%ymm8 \n\t" // da_r*x0 , da_r *x1
"vmulpd %%ymm0, %%ymm5 , %%ymm9 \n\t"
"vmulpd %%ymm0, %%ymm6 , %%ymm10 \n\t"
"vmulpd %%ymm0, %%ymm7 , %%ymm11 \n\t"

"vmovups %%ymm8 , -128(%1) \n\t"
"vmovups %%ymm9 , -96(%1) \n\t"
"vmovups %%ymm10, -64(%1) \n\t"
"vmovups %%ymm11, -32(%1) \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}


static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x) __attribute__ ((noinline));

static void zscal_kernel_8_zero( BLASLONG n, FLOAT *alpha, FLOAT *x)
{


__asm__ __volatile__
(
"vxorpd %%ymm0, %%ymm0, %%ymm0 \n\t"

"addq $128, %1 \n\t"

".align 16 \n\t"
"1: \n\t"

//"prefetcht0 128(%1) \n\t"
// ".align 2 \n\t"

"vmovups %%ymm0 , -128(%1) \n\t"
"vmovups %%ymm0 , -96(%1) \n\t"
"vmovups %%ymm0 , -64(%1) \n\t"
"vmovups %%ymm0 , -32(%1) \n\t"

"addq $128 ,%1 \n\t"
"subq $8 , %0 \n\t"
"jnz 1b \n\t"

"vzeroupper \n\t"

:
:
"r" (n), // 0
"r" (x), // 1
"r" (alpha) // 2
: "cc", "%0", "%1",
"%xmm0", "%xmm1", "%xmm2", "%xmm3",
"%xmm4", "%xmm5", "%xmm6", "%xmm7",
"%xmm8", "%xmm9", "%xmm10", "%xmm11",
"%xmm12", "%xmm13", "%xmm14", "%xmm15",
"memory"
);

}




Write Preview
Loading…
Cancel
Save