[Bug 92196] [BDW] ES3-CTS.shaders.uniform_block.random.nested_structs_arrays.9 Fails

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=92196

--- Comment #3 from Jason Ekstrand <jason at jlekstrand.net> ---
So, I did a little more investigation last night and Matt and I looked at it a
bit this morning.  Here's what we found:

1) If you step through it in the simulator, the simulator produces "correct"
(would pass the test) results.  I double and tripple-checked this and I can't
get the simulator to give me yellow.

2) If you hack up could_coissue() in brw_fs_combine_constants.cpp, it generates
the exact same code for HSW as it does for BDW but HSW passes.

3) We generate the exact same code for BSW as we do for BDW and BSW passes.

The above seems to rule out any sort of register allocation or scheduling
errors.  Instead, it's almost certain that we're hitting some undocumented and
unsimulated hardware bug.  Unfortunately, the shader (and hitting the bug) is
very sensitive to register allocation and anything you do that changes register
allocation is liable to make the shader not hit the bug.

This means that there are a pile of different things that one could do to make
the shader go from pass to fail but none of them actually fix anything!  We
need to be very careful when we claim that a patch fixes this bug to make sure
that the patch actually does fix something and doesn't just hide it again.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20151008/52e985f6/attachment.html>