5c079506
 Laila Saadatifard
upload the ivote ...
|
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
|
#include <stim/cuda/cudatools/error.h>
template<typename T>
__global__ void mag3(T* out, T* in, int x, int y, int z){
//calculate x,y,z coordinates for this thread
int xi = blockIdx.x * blockDim.x + threadIdx.x;
//find the grid size along y
int grid_y = y / blockDim.y;
int blockidx_y = blockIdx.y % grid_y;
int yi = blockidx_y * blockDim.y + threadIdx.y;
int zi = blockIdx.y / grid_y;
int i = zi * x * y + yi * x + xi;
if(xi >= x|| yi >= y || zi>= z) return;
float xl = in[i * 3 + 0];
float yl = in[i * 3 + 1];
float zl = in[i * 3 + 2];
out[i] = sqrt(xl * xl + yl * yl + zl * zl);
}
template<typename T>
void gpu_mag3(T* gpu_mag, T* gpu_cart, unsigned int x, unsigned int y, unsigned int z){
int max_threads = stim::maxThreadsPerBlock();
dim3 threads(sqrt (max_threads),sqrt (max_threads));
dim3 blocks(x / threads.x + 1, (y / threads.y + 1) * z);
//call the GPU kernel to determine the gradient
mag3<T> <<< blocks, threads >>>(gpu_mag, gpu_cart, x, y, z);
}
template<typename T>
void cpu_mag3(T* out, T* in, unsigned int x, unsigned int y, unsigned int z){
//calculate the number of bytes in the array
unsigned int bytes = x* y* z * sizeof(T);
//allocate memory on the GPU for the input data.
T* gpu_cart;
cudaMalloc(&gpu_cart, bytes*3);
cudaMemcpy(gpu_cart, in, bytes*3, cudaMemcpyHostToDevice);
//allocate memory on the GPU for the magnitude
T* gpu_mag;
cudaMalloc(&gpu_mag, bytes);
//call the GPU version of this function
gpu_mag3<T>(gpu_mag, gpu_cart, x, y, z);
//copy the array back to the CPU
cudaMemcpy(out, gpu_mag, bytes, cudaMemcpyDeviceToHost);
//free allocated memory
cudaFree(gpu_cart);
cudaFree(gpu_mag);
}
#endif
|
