Bug 90350

Trace available at:

http://people.freedesktop.org/~imirkin/traces/portal_bug_portalrendering_G96.trace.xz

Verified to render correctly on i965 haswell. Would be nice if someone else with a G96 or older card can verify it also renders wrong. Otherwise it's an issue on the macbook gpu setup.

Bisected to

commit 44673512a84c0897c8eddabf4a56e79b7d5b3395
Author: Roy Spliet <rspliet@eclipso.eu>
Date:   Mon Jan 5 00:22:17 2015 +0100

    nv50/ir: Add sat modifier for mul
    
    Signed-off-by: Roy Spliet <rspliet@eclipso.eu>
    Reviewed-by: Ilia Mirkin <imirkin@alum.mit.edu>

Created attachment 115652 [details]
saturate shader test

On each of your NV96 and NVAC, please run the attached test through shader_runner. Also remove the "+ arg2" from the result computation and run it again. I suspect that only NV96 + arg2 removed will fail.

[The results that *actually* test clamping are 16, 20, 28, and 36.]

Looks like I was right in my guess. Since NVAC tends to have the same things as NVA0, I'm assuming that FMUL.SAT becomes a thing on G200+. Fix pushed:

commit da136dc07ddb6147d181c96f475b94f6281efd73
Author: Ilia Mirkin <imirkin@alum.mit.edu>
Date:   Sat May 9 03:26:07 2015 -0400

    nv50/ir: only enable mul saturate on G200+
    
    Commit 44673512a84 enabled support for saturating fmul. However
    experimentally this does not seem to work on the older chips. Restrict
    the feature to G200 (NVA0) and later.
    
    Reported-by: Pierre Moreau <pierre.morrow@free.fr>
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=90350
    Signed-off-by: Ilia Mirkin <imirkin@alum.mit.edu>
    Tested-by: Pierre Moreau <pierre.morrow@free.fr>
    Reviewed-by: Tobias Klausmann <tobias.johannes.klausmann@mni.thm.de>
    Cc: mesa-stable@lists.freedesktop.org