-
Notifications
You must be signed in to change notification settings - Fork 1
/
ResidualCalculation.cu
70 lines (49 loc) · 1.97 KB
/
ResidualCalculation.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
#include <cub/cub.cuh>
#include <cuda.h>
#include "Utilities.cuh"
#include <iostream>
#define BLOCKSIZE 256
#define ITEMS_PER_THREAD 8
const int N = 4096;
/******************************/
/* TRANSFORM REDUCTION KERNEL */
/******************************/
__global__ void TransformSumKernel(const float * __restrict__ indata1, const float * __restrict__ indata2, float * __restrict__ outdata) {
unsigned int tid = threadIdx.x + blockIdx.x * gridDim.x;
// --- Specialize BlockReduce for type float.
typedef cub::BlockReduce<float, BLOCKSIZE * ITEMS_PER_THREAD> BlockReduceT;
__shared__ typename BlockReduceT::TempStorage temp_storage;
float result;
if(tid < N) result = BlockReduceT(temp_storage).Sum((indata1[tid] - indata2[tid]) * (indata1[tid] - indata2[tid]));
if(threadIdx.x == 0) atomicAdd(outdata, result);
return;
}
/********/
/* MAIN */
/********/
int main() {
// --- Allocate host side space for
float *h_data1 = (float *)malloc(N * sizeof(float));
float *h_data2 = (float *)malloc(N * sizeof(float));
float *h_result = (float *)malloc(sizeof(float));
float *d_data1; gpuErrchk(cudaMalloc(&d_data1, N * sizeof(float)));
float *d_data2; gpuErrchk(cudaMalloc(&d_data2, N * sizeof(float)));
float *d_result; gpuErrchk(cudaMalloc(&d_result, sizeof(float)));
for (int i = 0; i < N; i++) {
h_data1[i] = 1.f;
h_data2[i] = 3.f;
}
gpuErrchk(cudaMemcpy(d_data1, h_data1, N * sizeof(float), cudaMemcpyHostToDevice));
gpuErrchk(cudaMemcpy(d_data2, h_data2, N * sizeof(float), cudaMemcpyHostToDevice));
TransformSumKernel<<<iDivUp(N, BLOCKSIZE), BLOCKSIZE>>>(d_data1, d_data2, d_result);
gpuErrchk(cudaPeekAtLastError());
gpuErrchk(cudaDeviceSynchronize());
gpuErrchk(cudaMemcpy(h_result, d_result, sizeof(float), cudaMemcpyDeviceToHost));
std::cout << "output: ";
std::cout << h_result[0];
std::cout << std::endl;
gpuErrchk(cudaFree(d_data1));
gpuErrchk(cudaFree(d_data2));
gpuErrchk(cudaFree(d_result));
return 0;
}