Re: Performance questions

New Message Reply Date view Thread view Subject view Author view

Timothy Moore (moore++at++WOLFENET.com)
Tue, 17 Jun 1997 13:30:57 -0700 (PDT)


   Date: Mon, 16 Jun 1997 17:20:03 -0400
   From: Liem Lu <LLU++at++FALCON.AL.WPAFB.AF.MIL>

   To those of you who are doing database paging:

   What kind of performance hit do you get when you
   do database paging compared to when you are not?

   On our machine (Onyx RE, 8 150MHZ R4400 CPUs, 128
   MB memory, 4MB tex mem, 3 pipes), the update rate
   went from 30Hz to 15Hz every time I page in a FLT
   file of approximately 500 polygons. The App
   process was locked on CPU #1, the three Cull
   processes and three Draw processes were locked on
   CPU #2,3,4,5,6, and 7 respectively. I forced the
   dbase process to run on CPU #0. That didn't help
   much.

   Is that the kind of performance I should expect
   when database paging?

Most people doing database paging use the .pfb binary format or a
custom file format. Flight is not very database-paging friendly.

   Also, I've noticed that even though the CPUs were
   locked for my Performer processes as above, the
   update rate would go down when someone run an unix
   app such as the emacs editor or make. This
   happens even when I wasn't doing database paging.
   What's up with that? Isn't the pfuLockDown{App,
   Cull, Draw} command supposed to lock the CPUs away
   from other apps from disturbing the Performer
   processes?

Are you using pfuPrioritizeProcs and running as root?

Tim
=======================================================================
List Archives, FAQ, FTP: http://www.sgi.com/Technology/Performer/
            Submissions: info-performer++at++sgi.com
        Admin. requests: info-performer-request++at++sgi.com


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2.0b2 on Mon Aug 10 1998 - 17:55:28 PDT

This message has been cleansed for anti-spam protection. Replace '++at++' in any mail addresses with the '@' symbol.