Re: PETSc mpich and lam "benchmarks"

[Junichi Uekawa <dancer@netfort.gr.jp>]

In Thu, 15 Feb 2001 10:14:59 -0500 Adam cum veritate scripsit :

Hi,

> >                               # Procs  mpich  lam
> >                               "0"      88.26  88.53
> >                               1        89.26  90.31
> >                               2        53.89  56.15
> >                               4        22.65  22.71

Hmmm.. interesting result. It seems to show how this benchmark scales pretty 
well. ;)

I would love to see results for larger clusters (with more nodes). Maybe that
 I have to go and hunt around for the systems myself.
It is rumoured that MPICH does not scale up like lam. I have not 
quite tested it.


# Maybe doing a tcpdump and looking at all the packets might enlighten me.



regards,
	junichi

--
Netfort: dancer@netfort.gr.jp           Debian: dancer@debian.org 
dancer, a.k.a. Junichi Uekawa   http://www.netfort.gr.jp/~dancer
 Dept. of Knowledge Engineering and Computer Science, Doshisha University.