On Tue, 2009-12-15 at 10:27 +0000, Keith Whitwell wrote:
> On Tue, 2009-12-15 at 01:49 -0800, Marek Olšák wrote:
> > On Tue, Dec 15, 2009 at 10:28 AM, Keith Whitwell <[email protected]> wrote:
> > > On Mon, 2009-12-14 at 16:31 -0800, Marek Olšák wrote:
> > >> On Mon, Dec 14, 2009 at 5:42 PM, Corbin Simpson
> > >> <[email protected]> wrote:
> > >> > As far as immediate verts, why don't we just add support to r300g to 
> > >> > switch
> > >> > to immediate mode for small VBOs?
> > >> >
> > >> > Posting from a mobile, pardon my terseness. ~ C.
> > >> >
> > >>
> > >> Corbin,
> > >>
> > >> that seems reasonable, and it's the reason I killed the draw_quad
> > >> function. BTW immediate mode doubles the performance in glxgears.
> > >
> > > Shouldn't gears upload its vertices one time only and then leave them
> > > resident in VRAM?
> > >
> > > Keith
> > >
> > >
> > 
> > I meant the clear path with a quad using immediate mode, that made the
> > difference in comparison with a straightforward VBO path.
> 
> OK, understood.
> 
> I'm interested to understand why the VBO path is slow.  Is it the
> allocation of the VBO itself?  Mapping & uploading the vertices?
> Command submission (the extra reloc)?
> 
> It seems to me that none of these should be taking half your time.  In
> fact, it makes me wonder if there wasn't something else going on -- eg.
> was the same buffer getting reused to hold those 4 vertices frame after
> frame, resulting in a sync when the driver tried to map the buffer?
> 
> That could be happening in the state tracker or in the driver, I guess.
> 
> Keith



------------------------------------------------------------------------------
Return on Information:
Google Enterprise Search pays you back
Get the facts.
http://p.sf.net/sfu/google-dev2dev
_______________________________________________
Mesa3d-dev mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/mesa3d-dev

Reply via email to