[Bug 106774] GLSL IR copy propagates loads of SSBOs

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Wed Jun 6 21:21:01 UTC 2018


https://bugs.freedesktop.org/show_bug.cgi?id=106774

--- Comment #11 from Caio Marcelo de Oliveira Filho <caio.oliveira at intel.com> ---
Even with the fixes to the GLSL copy propagation (and other passes) logic, the
shader was still hanging. After a debugging session with Jason, he figured that
the helper invocations were not making progress, getting into an infinite loop.

While the reads of value give correct values, the result from atomicCompareSwap
is undefined for helper invocations, and comparing two was always failing.

The alternative version of the test in the report (the one that works), avoids
this issue since the return value for atomicCompareSwap seems to be always the
same (although not something we should rely on, I guess), allowing the loop to
end.

Early returning based on gl_HelperInvocation makes things work with the
original loop, but depends on newer GL version.

-- 
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/20180606/b56a2e42/attachment.html>


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