chapter5_6_bitmap.cu

#include <cuda_runtime.h>
#include <device_launch_parameters.h>
#include <math.h>
#include "cpu_bitmap.h"
#define DIM 512
#define PI 3.14159
__global__ void kernel(unsigned char* ptr) {
	int x = threadIdx.x + blockIdx.x * blockDim.x;
	int y = threadIdx.y + blockIdx.y * blockDim.y;
	int offset = x + y * blockDim.x * gridDim.x;
	__shared__ float shared[16][16];
	const float period = 128.0f;
	shared[threadIdx.x][threadIdx.y] = 255 * (sinf(x * 2.0f * PI / period) + 1.0f) *
		(sinf(y * 2.0f * PI / period) + 1.0f) / 4.0f;
	__syncthreads();
	ptr[offset * 4 + 0] = 0;
	ptr[offset * 4 + 1] = shared[15- threadIdx.x][15 - threadIdx.y];
	ptr[offset * 4 + 2] = 0;
	ptr[offset * 4 + 3] = 255;
}
int main(void) {
	CPUBitmap bitmap(DIM, DIM);
	unsigned char* dev_bitmap;
	cudaMalloc((void**)&dev_bitmap, bitmap.image_size());
	dim3 grids(DIM / 16, DIM / 16);
	dim3 threads(16, 16);
	kernel << <grids, threads >> > (dev_bitmap);
	cudaMemcpy(bitmap.get_ptr(), dev_bitmap, bitmap.image_size(),cudaMemcpyDeviceToHost);
	bitmap.display_and_exit();
}