//This is my kernel function
__global__ void createSCM(Pixel*pixelMat, //image
int imgRows, //image dimensions
int imgCols,
int*matrizSCM, //Coocurrence matrix
int numNiveles, //coocurrence matrix levels = 256
int delta_R, //value = {-1,0 or 1}
int delta_C) //value = {-1,0 or 1}
{
int i = blockIdx.y*blockDim.y+threadIdx.y;
int j = blockIdx.x*blockDim.x+threadIdx.x;
int cols = numNiveles;
int posx,posy;
if ( (j + delta_C) < imgCols && (i + delta_R) < imgRows &&
((j + delta_C) >= 0) && ((i + delta_R) >= 0) )
{
posx = pixelMat[i*imgCols+j].channel_0;
posy = pixelMat[(i + delta_R)*imgCols+(j + delta_C)].channel_0;
matrizSCM[posx*cols+posy]++;
matrizSCM[posy*cols+posx]++;
}
}
struct Pixel {
int channel_0;
};
我在共现矩阵中有计数错误,因为
pixelMat[i*imgCols+j]
和pixelMat[(i + delta_R)*imgCols+(j + delta_C)]
正在使用同一个线程访问不同的位置。
这是我的内核调用
int Grid_Dim_x=imagenTest.rows, Grid_Dim_y=imagenTest.cols;
int Block_Dim_x=1, Block_Dim_y=1;
dim3 Grid(Grid_Dim_x, Grid_Dim_y);
dim3 Block(Block_Dim_x,Block_Dim_x);
createSCM<<<Grid,Block>>>(...)
每个块上只有一个线程,每个块代表一个像素
这个问题有很好的解决方案吗?
谢谢 :)