Re: Database Paging

New Message Reply Date view Thread view Subject view Author view

From: Angus Dorbie (dorbie++at++sgi.com)
Date: 04/11/2000 07:56:20


Can you use amallinfo to get information on the arena?

This will help determine the rate of growth and whether it's a leak or
fragmentation.

Also use amallopt on the arena to try and improve the memory reuse in
the case of fragmentation.

Do you have any idea where the leak (if that's what it is) is being
caused? Is it the loaders the database utilities or the scene graph not
being cleaned up when you delete it...e.t.c.?

CHeers,ANgus.

"Heibel, R.J." wrote:
>
> Howdy!
>
> I am currently debugging a Performer 2.2 database paging routine for a radar
> simulation running on an octane with 256MB of memory. I must be able to page
> because already 20 geocells appear to be too much for loading/keeping into
> memory. Eventually, the database will be > 200 geocells.
>
> Currently, I am attempting load 5 geocells into memory and swap geocells as
> the aircraft position changes. Even though the Performer remove and add
> child routines are being applied reasonably, it doesn't appear to deallocate
> memory correctly. I have added the calls to pfTmpDir and pfSharedArenaSize
> after the first failures, and it holds up for a little while longer, but I
> need a permanent solution and lasting fix.
>
> I have noticed a simular problem while creating database loaders for
> preprocessing other database formats, but made work arounds since these
> utilities were not realtime anyway.
>
>
> Please see attached routine.
>
>
>
> RJ Heibel
>
> ------------------------------------------------------------------------
>
> database_pager.CName: database_pager.C
> Type: unspecified type (application/octet-stream)

-- 
For Performer+OpenGL tutorials http://www.dorbie.com/

"In the middle of difficulty lies opportunity." --Albert Einstein


New Message Reply Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue Apr 11 2000 - 07:56:26 PDT

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