[Bug 111490] [REGRESSION] [BISECTED] Shadow Tactics: Blades of the Shogun - problems rendering water

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Sat Aug 31 19:31:30 UTC 2019


https://bugs.freedesktop.org/show_bug.cgi?id=111490

--- Comment #7 from Ian Romanick <idr at freedesktop.org> ---
I think two things are making it challenging to debug this issue.  First, it
appears that the before output is from before f2dc0f28728 ("nir: Drop imov/fmov
in favor of one mov instruction").  There were 1,777 commits between
f2dc0f28728 and 09705747d72, so there are a lot of differences that are
unrelated to this issue.  Finding a needle in a haystack is hard.

Second, it appears that the use of the shader cache prevents INTEL_DEBUG from
outputing the information that I actually needed. :(  The bisected commit
causes additional ffma instructions to be generated, and there are zero ffma
instructions in any of the shaders that actually came from the game (there are
several from driver-generated shaders and from the Steam runtime).  I guess
when shader cache is used, the NIR output comes from before ffma instructions
are generated... which is before this optimization occurs.  Finding an
invisible needle in an invisible haystack is even harder. :(

Looking for MAD instructions in the GPU assembly has been a little more
fruitful.  For example, the assembly for GLSL246 definitely fits the pattern. 
I also see a couple other shaders, like GLSL15, that fit the pattern, but the
optimization wasn't applied in all cases due to some restrictions in the
optimization.  So far this supports my hypothesis from comment #2 that the app
isn't using 'invariant' when it should.

In a few minutes I'm going to attach a couple modifications to this
optimization that may eliminate the problem observed in this app.  I'd
appreciate it if someone could try these and report back.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20190831/0f945e5d/attachment.html>


More information about the intel-3d-bugs mailing list