[SCore-users-jp] Re: [SCore-users] GM availability on SCore, when ???

[Shinji Sumimoto]

Hi.

Could you let me know your measurement envinronment?

  1) Which Myrinet Cards are you using, PCI64B or PCI64C?
  2) How about your host environment, CPU, chipset and memory?

Shinji.

From: Shinji Sumimoto <s-sumi at bd6.so-net.ne.jp>
Subject: Re: [SCore-users-jp] Re: [SCore-users] GM availability on SCore, when ???
Date: Fri, 20 Dec 2002 00:21:57 +0900 (JST)
Message-ID: <20021220.002157.74750815.s-sumi at bd6.so-net.ne.jp>

s-sumi> Hi.
s-sumi> 
s-sumi> Thank you for the information.
s-sumi> 
s-sumi> I will invesitigate the performance.
s-sumi> 
s-sumi> Shinji.
s-sumi> 
s-sumi> From: Emile CARCAMO <emile.carcamo at nec.fr>
s-sumi> Subject: [SCore-users-jp] Re: [SCore-users] GM availability on SCore, when ???
s-sumi> Date: Thu, 19 Dec 2002 15:53:46 +0100
s-sumi> Message-ID: <200212191453.gBJErkY07276 at emilepc.ess.nec.fr>
s-sumi> 
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> s-sumi at flab.fujitsu.co.jp said:
s-sumi> emile.carcamo> > Have you used mpi_zerocopy=on on your benchmark? The default option
s-sumi> emile.carcamo> > MPICH/SCore is mpi_zerocopy=off. 
s-sumi> emile.carcamo> >
s-sumi> emile.carcamo> > Which benchmark are you using? Could you show the performance results?
s-sumi> emile.carcamo> > 
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> Dear Shinji-san,
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 	Sorry for not replying earlier ... Yes we have used the
s-sumi> emile.carcamo> 	option "mpi_zerocopy=on" when issuing the scrun command
s-sumi> emile.carcamo> 	within a SCore scout session :
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 	     scrun -nodes=2x1,mpi_zerocopy=on ./sendrecv
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 	With this sample benchmark, we noticed a peak transfer
s-sumi> emile.carcamo> 	rate looking like this :
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 			405 Mbytes/sec with GM/Myrinet
s-sumi> emile.carcamo> 			325 Mbytes/sec with PM/Myrinet
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 	Here is the URL of another test we've also ran on our
s-sumi> emile.carcamo> 	Myrinet-2k cluster :
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 		   http://www.hlrs.de/mpi/b_eff
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 	and we noticed in that case about 100 Mbytes/sec with
s-sumi> emile.carcamo> 	PM/Myrinet and two times faster using GM/Myrinet ...
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 	Please let us know if you need more inputs. I'll be glad
s-sumi> emile.carcamo> 	to hear from anybody who already performed such kind of
s-sumi> emile.carcamo> 	tests ... Thanks for your help, and best regards.
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> -- 
s-sumi> emile.carcamo> Emile_CARCAMO           NEC European            http://www.ess.nec.de
s-sumi> emile.carcamo> System Engineer         Supercomputer           mailto:emile.carcamo at nec.fr
s-sumi> emile.carcamo>                         Systems France      GSM mailto:emile.carcamo at orange.fr
s-sumi> emile.carcamo> (+33)1-3930-6601   FAX
s-sumi> emile.carcamo> (+33)1-3930-6613  PHONE  / Your mouse has moved. Windows NT must be restarted \
s-sumi> emile.carcamo> (+33)6-8063-7003   GSM   \ for the change to take effect. Reboot now?  [ OK ] /
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> 
s-sumi> emile.carcamo> _______________________________________________
s-sumi> emile.carcamo> SCore-users mailing list
s-sumi> emile.carcamo> SCore-users at pccluster.org
s-sumi> emile.carcamo> http://www.pccluster.org/mailman/listinfo/score-users
s-sumi> emile.carcamo> _______________________________________________
s-sumi> emile.carcamo> SCore-users-jp mailing list
s-sumi> emile.carcamo> SCore-users-jp at pccluster.org
s-sumi> emile.carcamo> http://www.pccluster.org/mailman/listinfo/score-users-jp
s-sumi> emile.carcamo> 
s-sumi> -----
s-sumi> Shinji Sumimoto    E-Mail: s-sumi at bd6.so-net.ne.jp
s-sumi> _______________________________________________
s-sumi> SCore-users mailing list
s-sumi> SCore-users at pccluster.org
s-sumi> http://www.pccluster.org/mailman/listinfo/score-users
s-sumi> 
s-sumi> 
------
Shinji Sumimoto, Fujitsu Labs