/
APSP.cu
211 lines (179 loc) · 4.91 KB
/
APSP.cu
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
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <sys/time.h>
#include <cuda_runtime.h>
#include <cuda_profiler_api.h>
#include "APSP.cuh"
#define BLKSZ_1 32
#define BLKSZ_2 32
#define BLKSZ_3 32
#define ROUND 6
#define gpuErrchk(ans) { gpuAssert((ans), __FILE__, __LINE__); }
inline void gpuAssert(cudaError_t code, const char *file, int line, bool abort=true)
{
if (code != cudaSuccess)
{
fprintf(stderr,"GPUassert: %s %s %d\n", cudaGetErrorString(code), file, line);
if (abort) exit(code);
}
}
__global__ void floyd_kernel_1(int k, int *devMat, int N)
{
__shared__ int d_d[BLKSZ_1][BLKSZ_1];
int base = k * BLKSZ_1;
int i = threadIdx.y;
int j = threadIdx.x;
int d_i = i + base;
int d_j = j + base;
base = d_i * N + d_j;
d_d[i][j] = devMat[base];
__syncthreads();
int newD;
for (int t = 0; t < BLKSZ_1; t++) {
newD = d_d[i][t] + d_d[t][j];
if (newD < d_d[i][j])
d_d[i][j] = newD;
}
devMat[base] = d_d[i][j];
}
__global__ void floyd_kernel_2(int k, int *devMat, int N)
{
if (blockIdx.x == k) return;
__shared__ int d_d[BLKSZ_2][BLKSZ_2], d_c[BLKSZ_2][BLKSZ_2];
int base = k * BLKSZ_2;
int i = threadIdx.y;
int j = threadIdx.x;
int d_i = i + base;
int d_j = j + base;
base = d_i * N + d_j;
d_d[i][j] = devMat[base]; // (k,k) matrix
if (blockIdx.y == 0) {
d_j = BLKSZ_2 * blockIdx.x + threadIdx.x;
} else {
d_i = BLKSZ_2 * blockIdx.x + threadIdx.y;
}
int current_base = d_i * N + d_j;
d_c[i][j] = devMat[current_base]; // updating matrix
__syncthreads();
int newD;
if (blockIdx.y == 0) {
for (int t = 0; t < BLKSZ_2; t++) {
newD = d_d[i][t] + d_c[t][j];
if (newD < d_c[i][j])
d_c[i][j] = newD;
}
} else {
for (int t = 0; t < BLKSZ_2; t++) {
newD = d_c[i][t] + d_d[t][j];
if (newD < d_c[i][j])
d_c[i][j] = newD;
}
}
devMat[current_base] = d_c[i][j];
}
__global__ void floyd_kernel_3(int k, int *devMat, int N)
{
if (blockIdx.x == k || blockIdx.y == k) return;
__shared__ int d_c[BLKSZ_3][BLKSZ_3], d_r[BLKSZ_3][BLKSZ_3];
int base = k * BLKSZ_3;
int d_i = blockDim.y * blockIdx.y + threadIdx.y;
int d_j = blockDim.x * blockIdx.x + threadIdx.x;
int i = threadIdx.y;
int j = threadIdx.x;
int col_base = (base + i) * N + d_j;
int row_base = d_i * N + base + j;
base = d_i * N + d_j;
d_r[i][j] = devMat[col_base];
d_c[i][j] = devMat[row_base];
int oldD = devMat[base];
__syncthreads();
int newD;
for (int t = 0; t < BLKSZ_3; t++) {
newD = d_c[i][t] + d_r[t][j];
if (newD < oldD)
oldD = newD;
}
devMat[base] = oldD;
}
int floyd(int* devMat, int N)
{
int N_blk = N / BLKSZ_1;
dim3 blockSize1(BLKSZ_1, BLKSZ_1);
dim3 gridSizeP2(N / BLKSZ_2, 2);
dim3 blockSize2(BLKSZ_2, BLKSZ_2);
dim3 gridSizeP3(N / BLKSZ_3, N / BLKSZ_3);
dim3 blockSize3(BLKSZ_3, BLKSZ_3);
for (int k = 0; k < N_blk; k++) {
cudaProfilerStart();
floyd_kernel_1 <<<1, blockSize1>>> (k, devMat, N);
floyd_kernel_2 <<<gridSizeP2, blockSize2>>> (k, devMat, N);
cudaProfilerStop();
floyd_kernel_3 <<<gridSizeP3, blockSize3>>> (k, devMat, N);
}
gpuErrchk(cudaPeekAtLastError());
return 0;
}
int main(int argc, char* argv[])
{
int N;
int flag = 0;
int *mat, *ref, *result;
struct timespec start, end;
double diff1, diff2;
if (argc == 1) {
printf("USAGE N [1]; N graph size, N = 2^k; 1 means doing validation\n");
exit(-1);
}
if (argc == 2) {
if (!sscanf(argv[1], "%d", &N)) {
printf("USAGE N [1]; N graph size, N = 2^k; 1 means doing validation\n");
exit(-1);
}
}
if (argc == 3) {
if (!sscanf(argv[1], "%d", &N)) {
printf("USAGE N [1]; N graph size, N = 2^k; 1 means doing validation\n");
exit(-1);
}
if (!sscanf(argv[2], "%d", &flag)) {
printf("USAGE N [1]; N graph size, N = 2^k; 1 means doing validation\n");
exit(-1);
}
}
int ttt;
cudaGetDeviceCount(&ttt);
printf("device number = %d\n", ttt);
mat = (int *)malloc(N * N * sizeof(int));
ref = (int *)malloc(sizeof(int) * N * N);
GenMatrix(mat, N);
if (flag) {
memcpy(ref, mat, sizeof(int) * N * N);
clock_gettime(CLOCK_MONOTONIC, &start);
ST_APSP(ref, N);
clock_gettime(CLOCK_MONOTONIC, &end);
diff1 = 1000000 * (end.tv_sec-start.tv_sec) + (end.tv_nsec-start.tv_nsec)/1000;
}
cudaMallocManaged(&result, N * N * sizeof(int));
cudaMemcpy(result, mat, sizeof(int) * N * N, cudaMemcpyHostToDevice);
diff2 = 0;
floyd(result, N);
floyd(result, N);
for (int i = 0; i < ROUND; i++) {
clock_gettime(CLOCK_MONOTONIC, &start);
floyd(result, N);
cudaDeviceSynchronize();
clock_gettime(CLOCK_MONOTONIC, &end);
diff2 = 1000000 * (end.tv_sec-start.tv_sec) + (end.tv_nsec-start.tv_nsec)/1000;
printf("%d Problem SIZE ------ CUDA_APSP elasped time =%.2lfus\n", N, diff2);
}
if (flag) {
if (CmpArray(result, ref, N * N))
printf("Your result is correct.\n");
else
printf("Your result is wrong.\n");
printf("ST_APSP elasped time = %.2lf us\n", diff1);
printf("============================================\n");
printf("Speedup = %.2lf\n", diff1 / diff2);
}
}