Bug 96298

Summary: 32-bit failures for spec.arb_compute_shader.execution.atomic-counter
Product: Mesa Reporter: Mark Janes <mark.a.janes>
Component: Drivers/DRI/i965Assignee: Jordan Justen <jljusten>
Status: RESOLVED DUPLICATE QA Contact: Intel 3D Bugs Mailing List <intel-3d-bugs>
Severity: normal    
Priority: medium CC: currojerez
Version: git   
Hardware: Other   
OS: All   
i915 platform: i915 features:

Description Mark Janes 2016-05-31 21:39:36 UTC
with mesa 357495b94dad4101a5491ed30782574162de58db, this test was enabled, and found to be failing *only* on 32 bit platforms.

This could indicate a bug in mesa or a bug in the test.

/tmp/build_root/m32/lib/piglit/bin/shader_runner /tmp/build_root/m32/lib/piglit/tests/spec/arb_compute_shader/execution/atomic-counter.shader_test -auto
piglit: debug: Requested an OpenGL 3.2 Core Context, and received a matching 4.2 context

Atomic counter 1 test failed: Reference == Observed
  Reference: 2147483647
  Observed:  4294966784

It is notable that the reference value is 7FFFFFFF, which may indicate an 32-bit overflow error.

Jordan, can you verify that the test is correct for 32 bit platforms?
Comment 1 Jordan Justen 2016-05-31 22:13:59 UTC

uses atomic_uint, which should be a 32-bit value on 64-bit processes
as well. The test has a uint atomic counter that starts at 0, and gets
decremented 512 times.

Does this test have a similar failure?

says "Increments and decrements at the limit of the range will wrap
to [0, 2^32-1]", so I think the test is valid. The shader_runner
code looks valid for treating the buffer as a uint32_t pointer.
Comment 2 Mark Janes 2016-05-31 22:41:33 UTC
I don't see any failures from that shader test.
Comment 3 Jordan Justen 2016-05-31 23:12:36 UTC
Sorry. I guess that was a test that I made when adding shader_runner
atomic counter support, but I didn't send it to piglit.

I added it in my freedesktop.org piglit repo atomic-fs branch.
Comment 4 Mark Janes 2016-06-03 22:36:47 UTC

*** This bug has been marked as a duplicate of bug 93333 ***

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.