Re: Cull and Draw Process Overlap and MultiGen Models with External R

New Message Reply Date view Thread view Subject view Author view

Marcin Romaszewicz (marcin++at++asmodean.engr.sgi.com)
Wed, 8 Sep 1999 18:40:24 -0700 (PDT)

Hi Stuart,

The particular method of communication between CULL and DRAW which the
PFMP_CULLoDRAW flag selects is broken in Performer. This problem has
nothing to do with the models you are loading, nor with the hardware
that you are running on. For now, the only workaround is to use
PFMP_APP_CULL_DRAW mode. I've run perfly with some very simple models
and they all manage to lock it up, so removing the external references
probably won't help.

-- Marcin

>
> Hi Angus,
>
> I'm developing a Performer application (based on Ian Bickerstaff's Landscape
> from BAe Flight Sim at Warton which you might know of) to run on a
> multiprocessor Onyx2 IR with a MultiGen database. In single procesor mode
> everything's fine, but in multiprocessor mode the application hangs if I set
> PFMP_CULLoDRAW in the call to pfMultiprocess when some of the models in the
> database come into the field of view (the models all load ok). If
> PFMP_CULLoDRAW isn't set everything's fine but obviously the latency is
> higher. I think the common factor between models causing the trouble is
> that they're all MultiGen models invoked as external references from other
> models. Any suggestions as to the cause and a fix other than getting rid of
> all the external references or not using PFMP_CULLoDRAW and putting up with
> the latency?
>
> System and software details as follows:
>
> Onyx 2 IR with 4xR1000 CPU 2xRM
> IRIX 6.5.4 and Performer v2.2
> MultiGen II v1.5
>
> pfMultiprocess(PFMP_APP_CULL_DRAW | PFMP_FORK_ISECT | PFMP_FORK_DBASE) is ok
> with 4 cpus
> pfMultiprocess(PFMP_APP_CULL_DRAW | PFMP_CULLoDRAW | PFMP_FORK_ISECT |
> PFMP_FORK_DBASE) hangs with 2 cpus at 100% 2 cpus at 0%
>
> Regards,
>
> Stuart Beaton
>
> c/o
> Andrew Ryder
> Software Engineer
> BAeA MS & S
> Salisbury, Adelaide, SA
> Ph: +61 8 8290 7891 Fx: +61 8 8290 8881
>
> -----------------------------------------------------------------------
> List Archives, FAQ, FTP: http://www.sgi.com/software/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 Wed Sep 08 1999 - 18:40:40 PDT

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