SUMMARY: Performance of a single job in a multi-CPU environment

From: Zina Yung (zina@cs.ust.hk)
Date: Thu Mar 13 1997 - 01:39:05 CST


Hello SUN managers:

******************************************************************
My original posting:

  Can one get any speed gain from the dual-CPU config on a single
  *ordinary* job? I.e., assuming no attempt has been made by a
  programmer to write a parallel program, and you run just one single
  job, does the system automatically do any low-level parallelizing
  to speed up your single job? If so, what kind of speed gain could
  one reasonably hope to get? (I realize that any speed gain would
  be highly dependent on the nature of the job, but just want a rough
  idea.)

*****************************************************************
My summary:
Thanks for all those replied! The majority said there is NO gain
unless:
        . you use parallelizing compiler such as SUN's Impact.
        . use some parallel virtual machine (e.g. PVM). This will
          allow your source to be modified to treat each processor as
          a distributed node.
        . the program does some forking, or spawning of other jobs.

Many thanks to:
       Bert N. Shure
       Michael R. Zika
       Matthew Stier
       Rich Kulawiec
       Kai O'Yang
       James H. McG. Sibl
       Greg Price
       Peter Bestel
       Kevin Sheehan
       Alex Finkel
       Marc S. Gibian
       Jay Lessert

*****************************************************************
Details of the responses:

From: "Bert Shure" <bert@virtual.com>

no luck unless the application is written to take advantage of multiple
cpu's.

the only winning aspect of the two processor system is that one
processor could use all the memory if nothing else is happening on the
system.

-------------------------------------------------------------------
From: "Michael R. Zika" <zika@oconto.tamu.edu>

No.
None.
If you're looking at a parallel applications, I would recommend one of
the following:

   o Purchase the iMPact toolkit with the SunSoft compilers. This will
     let you code shared-memory parallel directives into your code.
     I've used these and had good success with some small real world
     applications. It also provides and auto parallelization option.

   o Install PVM (Parallel Virtual Machine) or MPI (Message Passing
     Interface). This will allow your source to be modified to treat
     each processor as a distributed node. This allows you to treat
     your shared memory machine as a distributed memory machine (if
     that's what you want)

In both cases, source code modifications will be required.
-------------------------------------------------------------------
From: "Matthew Stier" <mstier@hotmail.com>

1) There never is just 'one' job running on a computer.

2) If the program does any forking, or spawning of other jobs, you can expect
some improvement.

3) You don't necessarily need multi-threading, to get a performance boost from
a multi-processor computer.

-----------------------From sun-managers-relay@ra.mcs.anl.gov Thu Mar 13 06:28:41 1997
Received: from mail.itw.com (localhost [127.0.0.1])
        by itw.com (8.8.5/8.8.5) with ESMTP id GAA27575
        for <rsk>; Thu, 13 Mar 1997 06:28:36 -0500 (EST)
X-POP3-Rcpt: rsk@mail
Received: from ra.mcs.anl.gov (ra.mcs.anl.gov [140.221.9.21]) by mail.itw.net (8.8.0/8.7.3) with ESMTP id FAA16086 for <rsk@itw.com>; Thu, 13 Mar 1997 05:49:25 -0500
Received: from localhost (daemon@localhost) by ra.mcs.anl.gov (8.8.3/8.8.3) with SMTP id EAA20670; Thu, 13 Mar 1997 04:16:38 -0600 (CST)
Received: by ra.mcs.anl.gov (bulk_mailer v1.5); Thu, 13 Mar 1997 04:14:54 -0600
Received: (from daemon@localhost) by ra.mcs.anl.gov (8.8.3/8.8.3) id DAA20561 for sun-managers-outbound; Thu, 13 Mar 1997 03:23:41 -0600 (CST)
Sender: sun-managers-relay@ra.mcs.anl.gov
Received: from morse.co.uk (gw.morse.co.uk [194.73.188.200]) by ra.mcs.anl.gov (8.8.3/8.8.3) with SMTP id DAA20554 for <sun-managers@ra.mcs.anl.gov>; Thu, 13 Mar 1997 03:23:33 -0600 (CST)
Received: by morse.co.uk with SMTP
        id JAA12917; Thu, 13 Mar 1997 09:37:17 GMT
Received: from scottw by scotserv. (SMI-8.6/SMI-SVR4)
        id JAA20995; Thu, 13 Mar 1997 09:26:19 GMT
Message-Id: <199703130926.JAA20995@scotserv.>
Comments: Authenticated sender is <scottw@scotserv>
From: "Scott Williamson" <scottw@morse.co.uk>
Reply-to: "Scott Williamson" <scottw@morse.co.uk>
Followup-to: "Scott Williamson" <scottw@morse.co.uk>
Organization: Morse Computers
To: sun-managers@ra.mcs.anl.gov
Date: Thu, 13 Mar 1997 09:23:50 +0000
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: SUMMARY: Boot Prom version
Priority: normal
X-mailer: Pegasus Mail for Win32 (v2.52)
Precedence: bulk

Thanks to all who wrote.

"prtconf -V" shows Open Boot Prom revision, which is not quite what
I wanted. Got a value of "OBP 3.1.1 1996/03/08 14:20" on my system.

Stop-A on the console, followed by "banner" then "go" to resume gives
the ROM revision, 2.28, which is what I was really after. Maybe I
didn't ask the correct question.

This is the value which, for example, has to be at least 2.25 to be
able to use Ross HyperSPARC modules in a SPARC 20.

Other people mentioned the "sysinfo" package as something which might
give this information. It would be nice to think that there's some
way of getting it without hanging the system.

-- 

Scott Williamson Morse Computers Tel.: 0131 226 3300 Senior Consultant 60 Melville Street Direct: 0131 260 3314 mailto:scottw@morse.co.uk Edinburgh Fax: 0131 226 3535 http://www.morse.co.uk EH3 7HF Mobile: 0370 865370



This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:11:48 CDT