Last active
February 9, 2018 19:31
-
-
Save thirdwing/5b64aa80afa37b1c4cf27cefbd49306c to your computer and use it in GitHub Desktop.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
// gcc -msse3 -O3 sse_4x4.c -l cblas | |
#include <stdio.h> | |
#include <stdlib.h> | |
#include <time.h> | |
#include <math.h> | |
#include <mmintrin.h> | |
#include <xmmintrin.h> | |
#include <pmmintrin.h> | |
#include <emmintrin.h> | |
// use cblas to verify the results | |
#include <cblas.h> | |
int main(int argc, char **argv) { | |
double A[16]; | |
double B[16]; | |
double C_sse[16]; | |
double C_blas[16]; | |
srand(time(NULL)); | |
for (int i = 0; i < 16; ++i) { | |
A[i] = (double) rand() / RAND_MAX; | |
B[i] = (double) rand() / RAND_MAX; | |
C_sse[i] = C_blas[i] = (double) rand() / RAND_MAX; | |
} | |
for (int j = 0; j < 4; j++) { | |
__m128d c1 = _mm_loadu_pd(C_sse + j * 4); | |
__m128d c2 = _mm_loadu_pd(C_sse + j * 4 + 2); | |
for (int i = 0; i < 4; i++) { | |
__m128d b1 = _mm_loadu_pd(B + i * 4); | |
__m128d b2 = _mm_loadu_pd(B + i * 4 + 2); | |
__m128d a = _mm_load1_pd(A + j * 4 + i); | |
c1 = _mm_add_pd(c1, _mm_mul_pd(a, b1)); // multiply and add | |
c2 = _mm_add_pd(c2, _mm_mul_pd(a, b2)); | |
} | |
// store the result back into the array | |
_mm_storeu_pd(C_sse + j * 4, c1); | |
_mm_storeu_pd(C_sse + j * 4 + 2, c2); | |
} | |
double alpha = 1, beta = 1; | |
cblas_dgemm(CblasRowMajor, CblasNoTrans, CblasNoTrans, 4, 4, 4, | |
alpha, A, 4, B, 4, beta, C_blas, 4); | |
for (int i = 0; i < 16; ++i) { | |
printf("C_sse[%d] = %f, C_blas[%d] = %f\n", i, C_sse[i], i, C_blas[i]); | |
} | |
return 0; | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment