http://bugs.freedesktop.org/show_bug.cgi?id=12240





------- Comment #7 from [EMAIL PROTECTED]  2007-12-28 16:51 PST -------
(In reply to comment #3)
> I have a patch which fixes the t_vp_build.c code to not require more temp
> registers then it actually needs, by allocating them only when needed and
> freeing them asap. It takes 2 of the number of temp registers needed by the
> generated code in the case of maniadrive, making maniadrive work even with the
> off by one bug still present in the r300 code. I'll attach this patch right
> away.

Looks good to me. Ultimately, a better solution would be to have a driver
backend which optimizes temp usage, but avoiding unnecessary temps in the first
place can't hurt (though might make the code slightly more complicated to
read). I think there are more temps which could be avoided in some places. For
instance those ambient/diffuse/specular possible temps are only used one after
another, though changing that will make the code a bit harder to read and
scheduling of the instructions less optimal (if using a backend which doesn't
reschedule things). build_sphere_texgen seems to use a unnecessary inv_m temp
(could just use tmp instead).
Oh, and if you have a r4xx card you should be able to use more temps, since for
PS2.0 the limit was 12, but for PS2.0b it was 32, and I don't think it was
upped without having ati hw support for it. Dunno if those are addressed the
same or need some special treatment (like requiring different reg class)
however.


-- 
Configure bugmail: http://bugs.freedesktop.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.

-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2005.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
--
_______________________________________________
Dri-devel mailing list
Dri-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/dri-devel

Reply via email to