[Openchrome-users] new dri options

[Boris Dušek]

Hi Benno,

On Saturday 07 October 2006 15:26, you wrote:
> Benno Schulenberg wrote:
> > Boris Du?ek wrote:
> > > The question is whether SSE can make the performance difference
> > > in the rest of the code, i.e. where vmesa->memcpy is invoked.
> > > If the difference is negligible, then best would be to simply
> > > forget about SSE at all and make things simpler.
> >
> > With glxgears it makes zero difference, no difference at all.
> > In slune and bzflag the fps are quite variable, so it is hard to
> > compare them.  But on a quiet and straight stretch in slune the
> > difference is also nil.
>
> Okay, I've dropped the entire sse_memcpy part from the patch.
> Could someone please test the attached patch and report either
> success or failure?  Or just inspect it and report any problems
> they see?

Just by going through the patch, I discovered one thing: in via_tex.c, it adds 
the "driOptionCache * optionCache" parameter to the "viaInitTextureFuncs". 
But the patch does not add any use of this new parameter.

It was added there by me because this intended to be used to query DRI 
configuration whether to allow SSE memcpy, because the memcpy routine is 
setup in this function. but since we drop the SSE, we do not need that 
parameter any more.

So I attach a version of your patch which corrects just this. The patch is 
against the default branch (6.5) as of today. Otherwise, all seems OK.

Boris

>
> Best try it with with the 6.4 branch of Mesa; it applies there with
> some fuzz but without error.  After applying it, don't forget to
> first run 'make' in src/mesa/drivers/dri/common/xmlpool, otherwise
> making Mesa will fail.
>
> Benno
-------------- next part --------------
A non-text attachment was scrubbed...
Name: mesa-unichrome--new-options-10.patch
Type: text/x-diff
Size: 6827 bytes
Desc: not available
Url : http://wiki.openchrome.org/pipermail/openchrome-users/attachments/20061009/fab0944b/mesa-unichrome--new-options-10.bin