Call a series kernels in a program

Accelerated Computing CUDA CUDA Programming and Performance
1

Hi everyone,
Can I call a variety of different kernels with different block sizes and thread sizes ?
Ex:
main()
{
…
alpha04 << < (index/16+1, index/16+1), (16,16) >> > (iw, ih, index, cu_x, cu_y, cu_label, cu_strength, cu_rin, cu_gin, cu_bin);
cudaThreadSynchronize();
alpha02 << < block, thread >> > (iw, ih,cu_x, cu_y, cu_label1, cu_label, cu_strength);
cudaThreadSynchronize();
alpha03 << <block1, thread1 >> > (iw, ih, cu_strength, cu_label);
cudaThreadSynchronize();
…
}
Note: block != block1, thread != thread1
when I use block and thread for all kernel. that run right. else they run fail.

2

yes, you can use different grid (blocks and threads) sizes.

Use normal processes to debug failures, such as proper CUDA error checking (google that) and run your code with cuda-memcheck