On 09/14/2012 11:09 AM, Marek Olšák wrote:
This series reworks blitting and MSAA resolving in Gallium as discussed here:
http://www.mail-archive.com/mesa-dev@lists.freedesktop.org/msg26592.html
I'd like to clarify some more rules for the blit function:
- Z/S blitting must always be supported regardless of shader stencil export and
sampler support (like glBlitFramebuffer). r300_blit is an example of how
stencil can be copied.
- The render condition has no effect.
I have tested r600g, r300g, softpipe, and llvmpipe.
The fbo-blit-stretch test started failing on softpipe. After careful
examination it looks like softpipe is culling one of the two triangles of the
quad u_blitter is rendering. The other tested drivers don't have this issue. I
assume it's a bug in softpipe that was just uncovered.
I haven't implemented the blit in the nv50/nvc0 driver, because Christoph has
already started working on his own implemention.
On the st/mesa side, I couldn't change CopyTexSubImage to use the new blit,
because the blit doesn't allow the alpha channel to be forced to 1. The only
way to achieve that is to set blit.src.format to an RGBX format and we don't
have enough RGBX formats to be able to do that.
That's it. Please review.
I (finally) took a quick look through the patches. The only general
thing is I think some of the new funtions/code are a bit light on
comments.
Looks OK otherwise. I'll test the vmware driver when this is checked in.
Reviewed-by: Brian Paul <bri...@vmware.com>
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
