[Mesa-dev] [PATCH 2/2] i965: Fix shared atomic intrinsics to pay attention to base.

Kenneth Graunke kenneth at whitecape.org
Fri Jul 22 04:26:21 UTC 2016


On Tuesday, July 19, 2016 10:23:03 AM PDT Timothy Arceri wrote:
> On Mon, 2016-07-18 at 15:49 -0700, Kenneth Graunke wrote:
> So this fixes a bug with indirects right? Is there a piglit test for
> this?

Not exactly.  Right now, GLSL lowers shared variable access at the
GLSL IR level, and when we translate the GLSL IR intrinsics to NIR,
we always set base to 0 and put everything in offset.  So Piglit
wouldn't have hit this.

Vulkan lowers shared variables in NIR, and was actually using a
non-zero base.  Vulkan tests could have hit this, but I don't think
any actually did.

I recall Jordan saying he ran into some issues when trying to make
GLSL use the NIR-based lowering, so maybe Piglit actually did hit
this base problem.  Not sure.

> With the typo Ilia pointed out fixed, both are:
> 
> Reviewed-by: Timothy Arceri <timothy.arceri at collabora.com>

Thanks!
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: This is a digitally signed message part.
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160721/badb1e54/attachment.sig>


More information about the mesa-dev mailing list