Marc Erich Latoschik (marcl++at++TechFak.Uni-Bielefeld.DE)
Mon, 01 Feb 1999 13:26:36 +0100
> In effect you were porting a single threaded application which had been
> written over a period of 3 years to a multiprocessor system.
The application itself was multithreaded from the beginning. I had to do
lots of
serial and multicast input and wanted to avoid blocking. That is why i
started
the Performer and pthreads thread a few weeks back.
>
> The realloc(0) problem was probably related to making draw calls in the
> application.
*SIGH* Thats what i thought too, after me (and some collegues) looked
over the code
at least 10 times, i am quite sure there were no draw calls in the app.
E.g. running
the ported one in APPCULLDRAW should have worked then but didn't.
Anyway, lets not cry over history :),
>
> It's fair to compare your porting effort to the difficulty you'd have
> had efficiently exploiting your OCTANE multi-processing without
> performer. Despite your experience I expect that Performer was in fact a
> big help to you in this respect.
Well, to be honest, i can't really admit that. It wasn't the first
multi-processing
step i did. And like you can see, there was no second node in my machine
anyway.
>
> I think your suggestions are great, I expect everyone who has
> implemented a sizeable performer application has encountered similar
> problems and solved them independently.
Yes, i think i will start soon and give a few explanations about what
kinds
of classes i used in which context.
>
> Cheers,Angus.
>
Admin. requests: info-performer-request++at++sgi.com
--bye Marc
************************************************************************** Marc Erich Latoschik, AG-WBS, Technische Fakultaet, Universitaet Bielefeld Universitaetsstrasse 25, PO 100 131, 33501 Bielefeld Room M4-122 Tel: (0521) 106 2919, Fax: (0521) 106 2962, marcl++at++techfak.uni-bielefeld.de http://www.TechFak.Uni-Bielefeld.DE/techfak/persons/marcl/
This archive was generated by hypermail 2.0b2 on Mon Feb 01 1999 - 04:26:44 PST