一个很简单的CUDA程序,适合刚刚接触CUDA的人了解CUDA的工作原理,以及与OpenCV结合的基本用法。
#include <stdlib.h>
#include <stdio.h>
#include <cv.h>
#include <highgui.h>
#include "cutil_inline.h"
#define GAMMA 0.4
void runTest(int argc, char** argv);
__global__ void testKernel(float* d_idata, float* d_odata, int width, int height, float gamma)
{
unsigned int tid_in_grid_x = blockDim.x*blockIdx.x + threadIdx.x;
unsigned int tid_in_grid_y = blockDim.y*blockIdx.y + threadIdx.y;
unsigned int tid_in_grid = tid_in_grid_y*width + tid_in_grid_x;
d_odata[tid_in_grid] = powf(d_idata[tid_in_grid], gamma);
}
int main(int argc, char** argv)
{
runTest(argc, argv);
CUT_EXIT(argc, argv);
}
void runTest(int argc, char** argv)
{
if( cutCheckCmdLineFlag(argc, (const char**)argv, "device") ) //设备初始化并设定计时器。
cutilDeviceInit(argc, argv);
else
cudaSetDevice(cutGetMaxGflopsDeviceId() );
unsigned int timer = 0;
cutilCheckError(cutCreateTimer(&timer));
cutilCheckError(cutStartTimer(timer));
IplImage* pImg;
if((pImg = cvLoadImage("lena_gray.jpg", CV_LOAD_IMAGE_GRAYSCALE)) != 0 )
{
cvNamedWindow("Image", 1);
cvShowImage("Image", pImg);
unsigned int num_blocks_x = pImg->width/16;
unsigned int num_blocks_y = pImg->height/16;
unsigned int mem_size = sizeof(float)*pImg->widthStep*pImg->height;
float* h_idata = (float*)malloc(mem_size);
float* h_odata = (float*)malloc(mem_size);
float* d_idata;
CUDA_SAFE_CALL(cudaMalloc((void**)&d_idata, mem_size));
float* d_odata;
CUDA_SAFE_CALL(cudaMalloc((void**)&d_odata, mem_size));
for(int i = 0; i < pImg->widthStep*pImg->height; i++) //从图像中复制数据时要特别注意变量类型,char类型的数据计算精度
h_idata[i] = ((uchar)pImg->imageData[i])/255.0; //太低,但是图像中只能存储char类型,所以要进行转换
CUDA_SAFE_CALL(cudaMemcpy(d_idata, h_idata, mem_size, cudaMemcpyHostToDevice));
dim3 grid(num_blocks_x, num_blocks_y, 1);
dim3 threads(16, 16, 1);
testKernel<<<grid, threads>>>(d_idata, d_odata, pImg->width, pImg->height, GAMMA);
CUT_CHECK_ERROR("Kernel execution failed");
CUDA_SAFE_CALL(cudaMemcpy(h_odata, d_odata, mem_size, cudaMemcpyDeviceToHost));
IplImage* oImg = cvCreateImage(cvSize(pImg->width,pImg->height), IPL_DEPTH_8U, 1);
for(int i = 0; i < pImg->widthStep*pImg->height; i++)
oImg->imageData[i] = (uchar)(int)(h_odata[i]*255);
cvNamedWindow("Result", CV_WINDOW_AUTOSIZE);
cvShowImage("Result", oImg);
// cvSaveImage("result.jpg", oImg);
cutilCheckError( cutStopTimer( timer));
printf( "Processing time: %f (ms)/n", cutGetTimerValue( timer));
cutilCheckError( cutDeleteTimer( timer));
cvWaitKey(0);
cvDestroyWindow("Image");
cvReleaseImage(&pImg);
cvDestroyWindow("Result");
cvReleaseImage(&oImg);
free(h_idata);
free(h_odata);
CUDA_SAFE_CALL(cudaFree(d_idata));
CUDA_SAFE_CALL(cudaFree(d_odata));
}
}
http://blog.csdn.net/mmjwung/article/details/6273653