Matrix Copy

1
#include <cuda_runtime.h>
2

3
__global__ void copy_matrix_kernel(const float* A, float* B, int total) {
4
    int idx = blockDim.x * blockIdx.x + threadIdx.x;
5
    if (idx > total) { return; }
6
    if (idx < total) {
7
        B[idx] = A[idx];
8
    }
9
}
10

11
// A, B are device pointers (i.e. pointers to memory on the GPU)
12
extern "C" void solve(const float* A, float* B, int N) {
13
    int total = N * N;
14
    int threadsPerBlock = 256;
15
    int blocksPerGrid = (total + threadsPerBlock - 1) / threadsPerBlock;
16
    copy_matrix_kernel<<<blocksPerGrid, threadsPerBlock>>>(A, B, total);
17
    cudaDeviceSynchronize();
18
}