-
Notifications
You must be signed in to change notification settings - Fork 0
/
vector_add.cu
82 lines (63 loc) · 2.46 KB
/
vector_add.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
// MP 1
#include <wb.h>
__global__ void vecAdd(float *in1, float *in2, float *out, int len) {
//@@ Insert code to implement vector addition here
int outIdx = blockDim.x * blockIdx.x + threadIdx.x;
if(outIdx<len){
out[outIdx] = in1[outIdx] + in2[outIdx];
}
}
int main(int argc, char **argv) {
wbArg_t args;
int inputLength;
float *hostInput1;
float *hostInput2;
float *hostOutput;
float *deviceInput1;
float *deviceInput2;
float *deviceOutput;
args = wbArg_read(argc, argv);
wbTime_start(Generic, "Importing data and creating memory on host");
hostInput1 =
(float *)wbImport(wbArg_getInputFile(args, 0), &inputLength);
hostInput2 =
(float *)wbImport(wbArg_getInputFile(args, 1), &inputLength);
hostOutput = (float *)malloc(inputLength * sizeof(float));
wbTime_stop(Generic, "Importing data and creating memory on host");
wbLog(TRACE, "The input length is ", inputLength);
wbTime_start(GPU, "Allocating GPU memory.");
//@@Allocating memory
cudaMalloc(&deviceInput1,inputLength*sizeof(float));
cudaMalloc(&deviceInput2,inputLength*sizeof(float));
cudaMalloc(&deviceOutput,inputLength*sizeof(float));
wbTime_stop(GPU, "Allocating GPU memory.");
wbTime_start(GPU, "Copying input memory to the GPU.");
//@@transfer data to device
cudaMemcpy(deviceInput1,hostInput1,inputLength*sizeof(float),cudaMemcpyHostToDevice);
cudaMemcpy(deviceInput2,hostInput2,inputLength*sizeof(float),cudaMemcpyHostToDevice);
cudaMemset(deviceOutput,0,inputLength*sizeof(float));
wbTime_stop(GPU, "Copying input memory to the GPU.");
//@@Dimensions
dim3 dimGrid((inputLength-1)/256+1,1,1);
dim3 dimBlock(256,1,1);
wbTime_start(Compute, "Performing CUDA computation");
//@@ Launch the GPU Kernel here
vecAdd<<<dimGrid,dimBlock>>>(deviceInput1, deviceInput2, deviceOutput, inputLength);
cudaDeviceSynchronize();
wbTime_stop(Compute, "Performing CUDA computation");
wbTime_start(Copy, "Copying output memory to the CPU");
//@@ Copy the GPU memory back to the CPU here
cudaMemcpy(hostOutput,deviceOutput,inputLength*sizeof(float),cudaMemcpyDeviceToHost);
wbTime_stop(Copy, "Copying output memory to the CPU");
wbTime_start(GPU, "Freeing GPU Memory");
//@@ Free the GPU memory here
cudaFree(deviceInput1);
cudaFree(deviceInput2);
cudaFree(deviceOutput);
wbTime_stop(GPU, "Freeing GPU Memory");
wbSolution(args, hostOutput, inputLength);
free(hostInput1);
free(hostInput2);
free(hostOutput);
return 0;
}