simultaneous run of CPU and CPU

Accelerated Computing CUDA CUDA Programming and Performance
1

(typo: CPU and GPU should be the proper subject title, my mistake.)

can i simultaneously run an extern “c” function and cuda from the main? in a extremely simple example,

main()
{

…

an_extern_c(A,B);

cuda_kernel<<<dimGrid,dimBlock>>>(C,D);

…

}

in the above, how do i arrange those two calls (an_extern and cuda_kernel) so that they run simultaneously?

can i do that?

any comments are welcome and thanks in advance.

2

Just reverse the order of the calls so the GPU is launched first (which is asynchronous) and the CPU function happens afterwards (which is blocking.)

You’ll probably want a synchronization afterwards so you can continue after you know both CPU and GPU have finished.

cuda_kernel<<<dimGrid,dimBlock>>>(C,D);

an_extern_c(A,B);

cudaThreadSynchronize();
3

do not forget that gpu call are batched, so to run simultaneously you need to release batch to gpu.