Re: [Starpu-devel] bad performance scaling behavior of cholesky decomposition

[Samuel Thibault <samuel.thibault@ens-lyon.org>]

Mario Tacconi, le Fri 25 May 2012 17:08:07 +0200, a écrit :
> As you suggest below the cause of the problematic scaling behavior was a
> bogus performance model.
> By resetting the performance model I was able to obtain a pretty decent
> scaling over a wide range of GPU+CPU configurations, as the graphs in
> attach show :)
> I was able to obtain a good performance model by using 2 GPU, 2 CPU
> worker, STARPU_CALIBRATE=1 and the
> "ws" scheduling policy.

Mmm, the ws scheduling policy actually does not make any use of
performance models.  So dmda performs worse than ws?  That's bad news :)

> > That could be odd, or maybe not. Our cholesky factorization is really
> > not advanced, and suffers from lack of parallelism, so it is not
> > necessarily surprising that it does not scale ; maybe you could rather
> > give a try at Magma 1.1 or later revision, which is the state of the
> > art for cholesky, and supports using StarPU for exploiting hybrid
> > machines.
> I know the magma library. Actually I discovered the StarPU project trying
> to compile their dynamically-scheduled version of the LU and Cholesky
> factorization.
> I have to admit that I could not succeed in obtaining working code froma
> that compilation: I think the Magma are using
> an old version of StarPU along with the StarPU's deprecated interface.

Mmm, it's odd, we tested that compilation before releasing StarPU 1.0,
there should only be warnings, and no errors.
Which error are you getting?

> Actually, I'm using the cholesky benchmarks I attached to this mail in a
> internal seminar on StarPU:
> I hope I can convince my collegues (and my boss...) to use StarPU in our
> future projects.

Great :)

Samuel