1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
|
// Copyright (c) 2017-2023, University of Tennessee. All rights reserved.
// SPDX-License-Identifier: BSD-3-Clause
// This program is free software: you can redistribute it and/or modify it under
// the terms of the BSD 3-Clause license. See the accompanying LICENSE file.
#include <cuda_runtime.h>
#include <cublas_v2.h>
#include <stdexcept>
#include <cassert>
#include <cstdio>
//------------------------------------------------------------------------------
void error_check_( cudaError_t err, const char* file, int line )
{
if (err != cudaSuccess) {
printf( "CUDA error %d: %s at %s:%d\n",
err, cudaGetErrorString(err), file, line );
exit(1);
}
}
//------------------------------------------------------------------------------
void error_check_( cublasStatus_t err, const char* file, int line )
{
if (err != CUBLAS_STATUS_SUCCESS) {
printf( "cuBLAS error %d at %s:%d\n",
err, file, line );
exit(1);
}
}
#define error_check( err ) \
error_check_( (err), __FILE__, __LINE__ )
//------------------------------------------------------------------------------
int main()
{
double alpha = 2, beta = 3;
int n = 2;
double A[] = { 1, 2, 3, 4 };
double B[] = { 5, 4, 3, 2 };
double C[] = { 2, 3, 1, 0 };
double D[] = { 40, 61, 21, 28 };
cudaError_t err = cudaSetDevice( 0 );
if (err != cudaSuccess) {
printf( "cudaSetDevice failed: %s (%d).\n"
"Cannot run on GPU; skipping test.\n",
cudaGetErrorString(err), err );
return 0;
}
double *dA, *dB, *dC;
error_check(
cudaMalloc( &dA, n*n*sizeof(double) ) );
error_check(
cudaMalloc( &dB, n*n*sizeof(double) ) );
error_check(
cudaMalloc( &dC, n*n*sizeof(double) ) );
assert( dA != nullptr );
assert( dB != nullptr );
assert( dC != nullptr );
// dA = A, dB = B, dC = c
error_check(
cudaMemcpy( dA, A, n*n*sizeof(double), cudaMemcpyDefault ) );
error_check(
cudaMemcpy( dB, B, n*n*sizeof(double), cudaMemcpyDefault ) );
error_check(
cudaMemcpy( dC, C, n*n*sizeof(double), cudaMemcpyDefault ) );
// C = alpha A B + beta C
cublasHandle_t handle;
error_check(
cublasCreate( &handle ) );
error_check(
cublasDgemm( handle, CUBLAS_OP_N, CUBLAS_OP_N, n, n, n,
&alpha, dA, n, dB, n, &beta, dC, n ) );
error_check(
cublasDestroy( handle ) );
// C = dC
error_check(
cudaMemcpy( C, dC, n*n*sizeof(double), cudaMemcpyDefault ) );
error_check(
cudaFree( dA ) );
error_check(
cudaFree( dB ) );
error_check(
cudaFree( dC ) );
// verify C == D
double result = 0;
for (int i = 0; i < n*n; ++i) {
printf( "C[%d] = %.2f, D = %.2f\n", i, C[i], D[i] );
result += std::abs( D[i] - C[i] );
}
bool okay = (result == 0);
printf( "%s\n", okay ? "ok" : "failed" );
return ! okay;
}
|
