-
Notifications
You must be signed in to change notification settings - Fork 0
/
syncthreads.cu
73 lines (54 loc) · 1.7 KB
/
syncthreads.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
#include <stdio.h>
#define ARRAYDIM 64
#define THREADS_PER_BLOCK 64
__global__ void globFunction(int *arr) {
__shared__ int local_array[THREADS_PER_BLOCK]; //local block memory cache
int idx = blockIdx.x * blockDim.x + threadIdx.x;
if(idx >= ARRAYDIM) {
return;
}
// read values and increase if not already 2
if(arr[idx] != 2) {
local_array[threadIdx.x] = arr[idx] + 1;
} else {
local_array[threadIdx.x] = arr[idx];
}
__syncthreads(); // <- wait for all threads to finish outside the if statement
// read the results of neighbor element
int val = local_array[(threadIdx.x + 1) % THREADS_PER_BLOCK];
// write back the value to global memory
arr[idx] = val;
return;
}
int main() {
int i, cudaRtrn;
dim3 thrds_per_block, blcks_per_grid;
int *a;
if (cudaRtrn = cudaMallocManaged(&a, ARRAYDIM * sizeof(int)) != 0) {
printf("*** allocation failed for array a[], %d ***\n", cudaRtrn);
}
// first half is 1
for(i = 0; i < ARRAYDIM/2; ++i) {
a[i] = 1;
}
// second half is 2
for(i = ARRAYDIM/2; i < ARRAYDIM; ++i) {
a[i] = 2;
}
thrds_per_block.x = THREADS_PER_BLOCK;
blcks_per_grid.x = 1;
printf("Running computation...\n");
globFunction<<<blcks_per_grid, thrds_per_block>>>(a);
if(cudaRtrn = cudaDeviceSynchronize() != 0) {
printf("*** error synchronizing device, %d ***\n", cudaRtrn);
}
printf("Computation finished!\n");
// sanity check on the result
for (i=0; i<ARRAYDIM-1; i++) {
if(a[i] != 2) {
printf("ERROR, expected a[%d]=2 but found %d\n", i, a[i]);
}
}
cudaFree(a);
return(0);
}