cuda卷积

一D卷积

#include<cuda_runtime.h>
#include<device_launch_parameters.h>
#include<iostream>
using namespace std;
__global__ void juan(int* a, int* b, int* c, int m, int n) {
	int id = blockIdx.x * blockDim.x + threadIdx.x;
	int sum = 0;
	int k = id - (m / 2);
	if (id < n) {
		for (int i = 0; i < m; i++) {
			if (i + k >= 0 && i + k < n) {
				sum += a[i + k] * b[i];
			}
		}
		c[id] = sum;
	}
}
int main() {
	int ha[7], hb[7], hc[7];
	int* da, * db, * dc;
	for (int i = 0; i < 7; i++) {
		cin >> ha[i];
	}
	for (int i = 0; i < 5; i++) {
		cin >> hb[i];
	}
	cudaMalloc((void**)&da, sizeof(int) * 7);
	cudaMalloc((void**)&db, sizeof(int) * 7);
	cudaMalloc((void**)&dc, sizeof(int) * 7);
	cudaMemcpy(da, ha, sizeof(int) * 7, cudaMemcpyHostToDevice);
	cudaMemcpy(db, hb, sizeof(int) * 7, cudaMemcpyHostToDevice);
	dim3 bl(7);
	juan << <1, bl >> > (da, db, dc, 5, 7);
	cudaMemcpy(hc, dc, sizeof(int) * 7, cudaMemcpyDeviceToHost);
	for (int i = 0; i < 7; i++) {
		cout << hc[i] << "  ";
	}
	cudaFree(da);
	cudaFree(db);
	cudaFree(dc);
}

二D卷积

#include<cuda_runtime.h>
#include<device_launch_parameters.h>
#include<iostream>
using namespace std;
__global__ void juan(int* a, int* b, int* c, int m, int n) {
	int idx = blockIdx.x * blockDim.x + threadIdx.x;
	int idy = blockIdx.y * blockDim.y + threadIdx.y;
	int sum = 0;
	int l = idx - (m / 2);
	int r = idy - (m / 2);
	if (idx < n && idy < n) {
		for (int i = 0; i < m; i++) {
			for (int j = 0; j < m; j++) {
				int nc = l + j;
				int nr = r + i;
				if (nc >= 0 && nc < n && nr >= 0 && nr < n) {
					sum += a[nr * n + nc] * b[i * m + j];
				}
			}
		}
		c[idy * n + idx] = sum;
	}
}
int main() {
	int ha[7][7], hb[5][5], hc[49];
	int* da, * db, * dc;
	for (int i = 0; i < 7; i++) {
		for (int j = 0; j < 7; j++)
			cin >> ha[i][j];
	}
	for (int i = 0; i < 5; i++) {
		for (int j = 0; j < 5; j++)
			cin >> hb[i][j];
	}
	cudaMalloc((void**)&da, sizeof(int) * 49);
	cudaMalloc((void**)&db, sizeof(int) * 49);
	cudaMalloc((void**)&dc, sizeof(int) * 49);
	cudaMemcpy(da, ha, sizeof(int) * 49, cudaMemcpyHostToDevice);
	cudaMemcpy(db, hb, sizeof(int) * 49, cudaMemcpyHostToDevice);
	dim3 bl(7, 7);
	juan << <1, bl >> > (da, db, dc, 5, 7);
	cudaMemcpy(hc, dc, sizeof(int) * 49, cudaMemcpyDeviceToHost);
	for (int i = 0; i < 7; i++) {
		for (int j = 0; j < 7; j++)
			cout << hc[i * 7 + j] << "  ";
		cout << endl;
	}
	cudaFree(da);
	cudaFree(db);
	cudaFree(dc);
}