Newsgroups: comp.parallel.mpi,comp.parallel.pvm
From: tagata@kcc.co.jp
Subject: Re: Memory Channel: Message Passing Latency 5.8 usecs, ~60 MB/sec @ 4KB messages
Organization: The Kaihatsu Computing Service Center Ltd.
Date: 20 Aug 1996 04:01:21 GMT
Message-ID: <4vbdah$66m@s4leia1.kcc.co.jp>


In <4urfml$cnr@bach.convex.com>, romero@convex.com (Paco Romero) writes:
Paco> In article <01bb8966$3b025a20$c4400410@richard.pa.dec.com>,
Paco> Nice job, guys!  I am afraid that you didn't get there first, though :-)
Paco> The HP PVM and MPI implementations on the HP/Convex Exemplar SPP1200 achieved
Paco> a 5 usec latency back in 1995.  With additional tuning we  achieve an 
Paco> impressive 3.99 usec latency and 135 MB/s bandwidth on the SPP1600, and we 
Paco> expect the bandwidth to increase significantly with the PA8000 family of HP 
Paco> Exemplar servers.  We used Dongarra's ping-pong code available from netlib 
Paco> for these measurements.
Paco>
Paco> Best regards
Paco>
Paco> Paco Romero (paco@rsn.hp.com)
Paco> HP Explicit Parallel Models, Project Manager.
Paco> HP Convex Division, Richardson, TX. 
Paco>

I am interesting in a Scalable Parallel Computer. It's a nice speed. I would like to know 
where I can find the Dongarra's ping-pong code.

Thanks.

-----------------
TAGATA, Kazuo
E-mail : tagata@kcc.co.jp


