[Beowulf] SPEC MPI2007

Håkon Bugge Hakon.Bugge at scali.com
Thu Apr 26 07:02:59 PDT 2007


Hi Tom,

 From the reference below, 18 benchmarks are 
characterized; viewgraph 15 and 16 show the 
"Message call count". Of the 18 benchmarks, none 
use MPI_Alltoall or MPI_Alltoallv. Now, if the 
suite has been amended with benchmarks using 
MPI_Alltoall, we're only lacking benchmarks using MPI_Alltoallv.

And to be precise, I do not search for said 
benchmarks personally; my intent was to underline 
that this dimension was missing from the 18 
(initial) benchmarks in the MPI2007 suite.



Thanks, Håkon


At 18:44 25.04.2007, Tom Elken wrote:
>Hi Håkon,
>
>you wrote:
>----------------
>I just read http://www.spec.org/workshops/2007/austin/slides/SPEC_MPI2007.pdf
>
>I am lacking applications using MPI_Alltoall and 
>MPI_Alltoallv - these are important dimensions 
>to evaluate. Anyone who knows about any suitable benchmark candidates?
>
>Thanks, Håkon
>----------------
>
>Thanks for the publicity on the forthcoming 
>cluster-relevant benchmark-suite from SPEC.  It 
>is due to be launched at ISC'07 in Dresden in 
>late June.  I am on the SPEC HPG committee that 
>is developing the SPEC MPI2007 benchmark suite, 
>and we would welcome more members to participate 
>in developing these benchmarks, including Scali :)
>
>Are you lacking "applications using MPI_Alltoall 
>and MPI_Alltoallv" or are you pointing out their lack in MPI2007?
>
>Thanks for pointing this out to us.  I did some 
>searching and a molecular dynamics code named 
>CPMD   and a weather code named Hirvda (an 
>"operational weatherforcast used by several 
>weather centers in Europe")  both use MPI_Alltoall extensively.
>
>I hope these applications will suit your 
>needs.  If others on the list can suggest more 
>of these types of applications, that would be great.
>
>It is too late for adding candidate codes for 
>the initial release of MPI2007 which ships with 
>what is called "medium" datasets.  There are 
>plans for more scalable versions of the suite in 
>the future with at least Large and perhaps XL 
>datasets.  With those releases there is the 
>possibility to add more codes, and we will 
>seriously consider codes that use MPI_Alltoall*.
>
>Thanks,
>Tom Elken
>QLogic Corporation
>
>

--
Håkon Bugge
CTO
dir. +47 22 62 89 72
mob. +47 92 48 45 14
fax. +47 22 62 89 51
Hakon.Bugge at scali.com
Skype: hakon_bugge

Scali - http://www.scali.com
Scaling the Linux Datacenter





More information about the Beowulf mailing list