Re: [Starpu-devel] StarPU OpenMP+CUDA

[alankrutha reddy <alankruthareddy46@gmail.com>]

My scenario is as below:

#pragma omp parallel for

for(...)
{

        kernel<<<,>>>(...);

}

In this scenario how to use StarPU.?

On Wed, Jul 3, 2013 at 1:04 PM, Samuel Thibault <samuel.thibault@ens-lyon.org> wrote:

Hello,

alankrutha reddy, le Wed 03 Jul 2013 06:54:21 +0200, a écrit :

> I want to run a program with both OpenMP and CUDA. From cpu_func I would like
> to call gpu_func. Is it possible.?

We don't have plans to have tasks running on both OpenMP and CUDA.  But
is it really what you want?  Usually applications have both an OpenMP
and a CUDA implementation of a given task, or some tasks implemented in
OpenMP and some tasks implemented on CUDA, and there is no real need to
run the execution at the same time, but instead simply express data
dependencies.

Samuel

--
Regards
Alankrutha