Bug 100061

Summary:	LODQ instruction generated with invalid dst mask
Product:	Mesa	Reporter:	Ilia Mirkin <imirkin>
Component:	Mesa core	Assignee:	mesa-dev
Status:	RESOLVED FIXED	QA Contact:	mesa-dev
Severity:	normal
Priority:	medium	CC:	sephiroth99
Version:	git
Hardware:	Other
OS:	All
Whiteboard:
i915 platform:		i915 features:
Attachments:	shader_test demonstrating the issue

Description Ilia Mirkin 2017-03-04 17:48:51 UTC

Created attachment 130064 [details]
shader_test demonstrating the issue

The attached shader_test produces an instruction like this:

  2: LODQ TEMP[1], TEMP[1], SAMP[0], 2D

Which is clearly bogus (and hits an assert in nouveau code making sure that the destmask is a subset of .xy). I tried to get this to happen with a much simpler shader, but was unsuccessful. Haven't tried whittling this one down yet. It comes from some radeonsi bug with a trace from Hitman.

My guess is that it's an issue in the st_glsl_to_tgsi register renumbering or copy-prop passes... somehow. (Mostly because I hate them, and hope that this hatred is vindicated by them having bugs.)

Comment 1 Ilia Mirkin 2017-03-19 05:14:57 UTC

commit dab88e9af7a35ebcdd0fc87df97f4b13e908552a
Author: Ilia Mirkin <imirkin@alum.mit.edu>
Date:   Sat Mar 4 13:52:48 2017 -0500

    st/mesa: set result writemask based on ir type
    
    This prevents textureQueryLevels, which maps as LODQ, from ending up
    with a xyzw writemask, which is illegal.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100061
    Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
    Cc: mesa-stable@lists.freedesktop.org
    Reviewed-by: Marek Olšák <marek.olsak@amd.com>

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.