Skip to content

Commit

Permalink
Final
Browse files Browse the repository at this point in the history
  • Loading branch information
Rio Yokota committed May 28, 2017
1 parent a1f47a5 commit a9be582
Show file tree
Hide file tree
Showing 2 changed files with 29 additions and 78 deletions.
73 changes: 12 additions & 61 deletions blislab/bl_sgemm_asm_24x4.c → blislab/micro_kernel.h
View file
Original file line number Diff line number Diff line change
@@ -1,53 +1,19 @@
/*
BLIS
An object-based framework for developing high-performance BLAS-like
libraries.
Copyright (C) 2014, The University of Texas at Austin
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are
met:
- Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
- 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.
- Neither the name of The University of Texas at Austin 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 COPYRIGHT
HOLDER 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 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 <stdio.h>
#include <immintrin.h> // AVX

typedef union {
__m256d v;
__m256i u;
double d[ 4 ];
} v4df_t;

typedef unsigned long long dim_t;

typedef union {
__m128i v;
int d[ 4 ];
} v4li_t;

#define inc_t unsigned long long
struct aux_t {
float *b_next;
float *b_next_s;
char *flag;
int pc;
int m;
int n;
};

#define SGEMM_INPUT_GS_BETA_NZ \
"vmovlps (%%rcx ), %%xmm0, %%xmm0 \n\t" \
Expand Down Expand Up @@ -79,14 +45,9 @@ typedef union {
"vpermilps $0x39, %%xmm2, %%xmm1 \n\t" \
"vmovss %%xmm1, (%%rcx,%%r10 ) \n\t"

void micro_kernel(int k,
float* a,
float* b,
float* c,
dim_t ldc,
aux_t* data) {
const inc_t cs_c = ldc;
const inc_t rs_c = 1;
void micro_kernel(int k, float* a, float* b, float* c, dim_t ldc, aux_t* data) {
const dim_t cs_c = ldc;
const dim_t rs_c = 1;
float alpha_val = 1.0, beta_val = 1.0;
float *alpha = &alpha_val;
float *beta = &beta_val;
Expand All @@ -100,7 +61,6 @@ void micro_kernel(int k,
" \n\t"
"movq %2, %%rax \n\t" // load address of a.
"movq %3, %%rbx \n\t" // load address of b.
//"movq %9, %%r15 \n\t" // load address of b_next.
" \n\t"
" \n\t" // initialize loop by pre-loading
"vmovaps 0 * 32(%%rax), %%ymm0 \n\t"
Expand Down Expand Up @@ -389,7 +349,6 @@ void micro_kernel(int k,
SGEMM_INPUT_GS_BETA_NZ
"vfmadd213ps %%ymm13, %%ymm3, %%ymm0 \n\t"
SGEMM_OUTPUT_GS_BETA_NZ
//"addq %%rdi, %%rcx \n\t" // c += cs_c;
" \n\t"
" \n\t"
"movq %%rdx, %%rcx \n\t" // rcx = c + 8*rs_c
Expand All @@ -416,7 +375,6 @@ void micro_kernel(int k,
SGEMM_INPUT_GS_BETA_NZ
"vfmadd213ps %%ymm14, %%ymm3, %%ymm0 \n\t"
SGEMM_OUTPUT_GS_BETA_NZ
//"addq %%rdi, %%rcx \n\t" // c += cs_c;
" \n\t"
" \n\t"
"movq %%r12, %%rcx \n\t" // rcx = c + 16*rs_c
Expand All @@ -443,7 +401,6 @@ void micro_kernel(int k,
SGEMM_INPUT_GS_BETA_NZ
"vfmadd213ps %%ymm15, %%ymm3, %%ymm0 \n\t"
SGEMM_OUTPUT_GS_BETA_NZ
//"addq %%rdi, %%rcx \n\t" // c += cs_c;
" \n\t"
" \n\t"
" \n\t"
Expand Down Expand Up @@ -547,7 +504,6 @@ void micro_kernel(int k,
" \n\t"
"vmovaps %%ymm13, %%ymm0 \n\t"
SGEMM_OUTPUT_GS_BETA_NZ
//"addq %%rdi, %%rcx \n\t" // c += cs_c;
" \n\t"
" \n\t"
"movq %%rdx, %%rcx \n\t" // rcx = c + 8*rs_c
Expand All @@ -570,7 +526,6 @@ void micro_kernel(int k,
" \n\t"
"vmovaps %%ymm14, %%ymm0 \n\t"
SGEMM_OUTPUT_GS_BETA_NZ
//"addq %%rdi, %%rcx \n\t" // c += cs_c;
" \n\t"
" \n\t"
"movq %%r12, %%rcx \n\t" // rcx = c + 16*rs_c
Expand All @@ -593,7 +548,6 @@ void micro_kernel(int k,
" \n\t"
"vmovaps %%ymm15, %%ymm0 \n\t"
SGEMM_OUTPUT_GS_BETA_NZ
//"addq %%rdi, %%rcx \n\t" // c += cs_c;
" \n\t"
" \n\t"
" \n\t"
Expand Down Expand Up @@ -630,11 +584,8 @@ void micro_kernel(int k,
"addq %%rdi, %%r12 \n\t"
" \n\t"
"vmovaps %%ymm13, (%%rcx) \n\t"
//"addq %%rdi, %%rcx \n\t"
"vmovaps %%ymm14, (%%rdx) \n\t"
//"addq %%rdi, %%rdx \n\t"
"vmovaps %%ymm15, (%%r12) \n\t"
//"addq %%rdi, %%r12 \n\t"
" \n\t"
" \n\t"
" \n\t"
Expand Down
34 changes: 17 additions & 17 deletions blislab/test.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,32 +2,21 @@
#include <cmath>
#include <cstdio>
#include <cstdlib>
#include <omp.h>
#include <limits.h>
#include <immintrin.h>
#include <omp.h>
#include <time.h>

typedef unsigned long long dim_t;

struct aux_t {
float *b_next;
float *b_next_s;
char *flag;
int pc;
int m;
int n;
};

extern "C" void sgemm_(char*, char*, int*, int*, int*, float*, float*,
int*, float*, int*, float*, float*, int*);

#define GEMM_SIMD_ALIGN_SIZE 32
#define SGEMM_MC 264
#define SGEMM_NC 128
#define SGEMM_KC 256
#define SGEMM_MR 24
#define SGEMM_NR 4

#include "bl_sgemm_asm_24x4.c"
extern "C" void sgemm_(char*, char*, int*, int*, int*, float*, float*,
int*, float*, int*, float*, float*, int*);


float *bl_malloc_aligned(int m, int n, int size) {
float *ptr;
Expand Down Expand Up @@ -72,10 +61,21 @@ inline void packB_kcxnc_d(int n, int k, float *XB, int ldXB, int offsetb, float
}
}

#if 1
#include "micro_kernel.h"
#else
void micro_kernel(int k, float* A, float* B, float* C, dim_t ldc, aux_t* aux) {
for (int i=0; i<aux.m; i++) {
for (int j=0; j<aux.n; j++) {
}
}
}
#endif

void macro_kernel(int m, int n, int k, float *packA, float *packB, float *C, int ldc) {
aux_t aux;
aux.b_next = packB;
for (int j=0; j<n; j+=SGEMM_NR ) {
for (int j=0; j<n; j+=SGEMM_NR) {
aux.n = std::min(n-j, SGEMM_NR);
for (int i=0; i<m; i+=SGEMM_MR) {
aux.m = std::min(m-i, SGEMM_MR);
Expand Down

0 comments on commit a9be582

Please sign in to comment.