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

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

Mario Tacconi, le Fri 25 May 2012 18:43:00 +0200, a écrit :
> > Mmm, the ws scheduling policy actually does not make any use of
> > performance models.  So dmda performs worse than ws?  That's bad news :)
> No, wait... I used the ws policy only to harvest data for the
> performance model. The idea was that by using ws (or eager) policy all the
> workers would have got an adequate amount of tasks and speed up the
> statistics "convergence" of the history based model. 

Ah, OK!

> During the benchmarks I found that the dmdas policy actually gives the
> best performance.

Good, that's what is expected :) though heft is supposed to achieve the
same. Is the difference very big?

Samuel