kenel2 has repeated calculations

dim3 gridDim(CEIL_DIV(M, 32), CEIL_DIV(N, 32)); dim3 blockDim(32 * 32); // BLOCKSIZE=32 const int cRow = blockIdx.x * BLOCKSIZE + (threadIdx.x / BLOCKSIZE); const int cCol = blockIdx.y * BLOCKSIZE + (threadIdx.x % BLOCKSIZE); I think there is a problem, threadIdx.y is not used for cRow and cCol, which cause different threads(has diff value of threadIdx.y) to do the same work.