[Starpu-devel] Combined workers and OpenMP code

[Keisuke Fukuda <fukuda@matsulab.is.titech.ac.jp>]

Hi all,

I'm working on StarPU's "combined workers" feature and have got some 
questions.
I've wrote a very simple code, which is similar to
"example/openmp/vector_scal.c".
I put it at https://gist.github.com/2488259 .

When I run the code on my environment,
starpu_combined_worker_get_size() always returns 1
and only 1 thread runs in #pragma omp parallel for loop.
OTOH, vector_scal example run 2-4 threads (although I expect 6 workers since
the CPU is 6-core sandy bridge).

So, here are my questions:
(1) How does StarPU determine the number of combined workers ?
 (in other words, why vector_scal runs 2-4 workers and my code runs only 1?
  What causes the difference?)

(2) Is there any way to let (or even force) StarPU to use a certain
number of workers, i.e.
just as the same number as of physical cores.

I know that splitting the code into multiple tasks are preferable option,
but we have existing OpenMP code and we don't want to do that.

StarPU version : starpu-1.0 branch, rev. 5648
hwloc : available.

Regards,
Keisuke

-- 
------------------------------------------
FUKUDA, Keisuke<fukuda@matsulab.is.titech.ac.jp>
Dept. of Math. & Comp. Sciences
Satoshi Matsuoka Lab.,
Tokyo Institute of Technology