38 lines
1 KiB
C
38 lines
1 KiB
C
#ifndef _MATRIXMUL_H_
|
|
#define _MATRIXMUL_H_
|
|
#include <stdlib.h>
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
#include <math.h>
|
|
|
|
// includes, project
|
|
#include <cuda_runtime.h>
|
|
//#include <helper_functions.h>
|
|
//#include <helper_cuda.h>
|
|
#include "cublas.h"
|
|
// Thread block size
|
|
#define BLOCK_SIZE 16
|
|
|
|
typedef enum type_version_kernel {v0,v0_bis,v1} Type_version_kernel;
|
|
|
|
void computeGold( float*, const float*, const float*, unsigned int, unsigned int, unsigned int);
|
|
// Matrix dimensions
|
|
// (chosen as multiples of the thread block size for simplicity)
|
|
/*
|
|
#define WA (64 * BLOCK_SIZE) // Matrix A width
|
|
#define HA (64 * BLOCK_SIZE) // Matrix A height
|
|
#define WB (64 * BLOCK_SIZE) // Matrix B width
|
|
#define HB WA // Matrix B height
|
|
#define WC WB // Matrix C width
|
|
#define HC HA // Matrix C height
|
|
*/
|
|
|
|
#
|
|
void compute_matrixMul_C(int W);
|
|
void compute_matrixMul_cublas(int N);
|
|
void compute_matrixMul_cuda(int N,Type_version_kernel v);
|
|
void randomInit(float*, int);
|
|
void printDiff(float*, float*, int, int);
|
|
|
|
|
|
#endif // _MATRIXMUL_H_
|