Bug 84777

*** Bug 84786 has been marked as a duplicate of this bug. ***

*** Bug 85393 has been marked as a duplicate of this bug. ***

*** Bug 84778 has been marked as a duplicate of this bug. ***

*** Bug 84078 has been marked as a duplicate of this bug. ***

*** Bug 84216 has been marked as a duplicate of this bug. ***

*** Bug 84776 has been marked as a duplicate of this bug. ***

*** Bug 84784 has been marked as a duplicate of this bug. ***

*** Bug 84788 has been marked as a duplicate of this bug. ***

*** Bug 84779 has been marked as a duplicate of this bug. ***

*** Bug 84783 has been marked as a duplicate of this bug. ***

*** Bug 84773 has been marked as a duplicate of this bug. ***

*** Bug 84772 has been marked as a duplicate of this bug. ***

*** Bug 84780 has been marked as a duplicate of this bug. ***

*** Bug 84782 has been marked as a duplicate of this bug. ***

*** Bug 84774 has been marked as a duplicate of this bug. ***

*** Bug 86769 has been marked as a duplicate of this bug. ***

*** Bug 84787 has been marked as a duplicate of this bug. ***

*** Bug 84221 has been marked as a duplicate of this bug. ***

*** Bug 84775 has been marked as a duplicate of this bug. ***

Increasing priority, It impacts so many cases.

Currently all tests using the GS fail. I've filed an internal bug, but haven't yet gotten anywhere with it. At least a few hundred piglit failures are caused by this. The GS failures may be masking real failures. Therefore I Agree that this should be high priority (perhaps even higher).

Can you please test my mesa branch?
http://cgit.freedesktop.org/~bwidawsk/mesa/log/?h=qw-mul

Created attachment 110473 [details] [review]
Avoid GS DW mul

You can use this patch instead of my branch.

(In reply to Ben Widawsky from comment #23)
> Created attachment 110473 [details] [review] [review]
> Avoid GS DW mul
> 
> You can use this patch instead of my branch.

Apply this patch, it works well.
root@x-bsw01:/GFX/Test/Piglit/piglit# bin/shader_runner tests/spec/glsl-1.50/execution/geometry-basic.shader_test -auto
libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so
PIGLIT: {"result": "pass" }

Can you please do a full piglit run with this patch? It should fix most, if not all of the BSW bugs that don't exist on BDW.

(In reply to Ben Widawsky from comment #25)
> Can you please do a full piglit run with this patch? It should fix most, if
> not all of the BSW bugs that don't exist on BDW.

OK.

Hi Ben,
I try to test full piglit case on bcc7eb115ed6e with your patch, and test full piglit case on the latest mesa master branch(I notice your patch landed).
Both they have many cases timeout, up to now finished 1000+ cases.
The timeout happens when auto run full piglit case, If run the binary, I am unable the reproduce the timeout(run more than 20 cycles)

Test on bcc7eb115ed6e7, it doesn't have this "timeout" issue.

Run case list:
Piglit/glx_glx-dont-care-mask    PASS
Piglit/glx_glx-fbconfig-compliance    PASS
Piglit/glx_glx-fbconfig-sanity    PASS
Piglit/glx_glx-fbo-binding    TIMEOUT

run log:
@test: Piglit/glx_glx-fbconfig-sanity
errors_ignored!
 libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
 libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so
!
info: @@@Returncode: 0

Errors:
libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so

Dmesg:

Output:
 bin/glx-fbconfig-sanity -auto -fbo
PIGLIT: {"result": "pass" }

Piglit
returncode: 0
result: pass
summary: Piglit/glx_glx-fbconfig-sanity    PASS
@test: Piglit/glx_glx-fbo-binding
errors_ignored!
 libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
 libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so
!
info: @@@Returncode: 1

Errors:
libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so

Dmesg:

Output:
 bin/glx-fbo-binding -auto

Piglit
returncode: 1
result: TIMEOUT
summary: Piglit/glx_glx-fbo-binding    TIMEOUT

I cannot reproduce this locally with mesa master. Can you please try to reproduce this after doing a hard reboot of the target, and running nothing else in between? 

I've seen the behavior you describe on CHV in particular when GPU resets do not actually reset state, and all subsequent tests will hang. If you can still reproduce after the hard reboot, can you please attached the gzipped results.json?

(In reply to Ben Widawsky from comment #28)
> I cannot reproduce this locally with mesa master. Can you please try to
> reproduce this after doing a hard reboot of the target, and running nothing
> else in between? 
> 
> I've seen the behavior you describe on CHV in particular when GPU resets do
> not actually reset state, and all subsequent tests will hang. If you can
> still reproduce after the hard reboot, can you please attached the gzipped
> results.json?

Finished one cycle full piglit testing on latest master branch.
pass count is similar to BDW.
One case has GPU hang issue, the GPU hang issue is unstable. I am unable reproduce it manually, the full log is too large, I will give you the log location via mail.

case                                 BDW    BSW  BUG
spec_ARB_timer_query_timestamp-get   PASS   FAIL bug 84791
spec_arb_shading_language_packing_execution_built-in-functions_vs-packHalf2x16  PASS  FAIL 
spec_arb_shading_language_packing_execution_built-in-functions_vs-unpackHalf2x16  PASS  FAIL  bug 84223
shaders_glsl-routing  PASS   FAIL bug 84826
spec_glsl-1.10_execution_built-in-functions_vs-op-le-int-int-using-if  PASS GPU_hang

log:
@test: Piglit/spec_glsl-1.10_execution_built-in-functions_vs-op-le-int-int-using-if
errors_ignored!
 libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
 libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so
!
info: @@@Returncode: 0


test case start at: Fri Jan  5 16:22:15 2001
test case end at:   Fri Jan  5 16:22:17 2001

Errors:
libGL: OpenDriver: trying /opt/X11R7/lib/dri/tls/i965_dri.so
libGL: OpenDriver: trying /opt/X11R7/lib/dri/i965_dri.so


Dmesg:<3>[20185.674527] [drm:i915_hangcheck_elapsed [i915]] *ERROR* Hangcheck timer elapsed... render ring idle


Output:
 bin/shader_runner /GFX/Test/Piglit/piglit/generated_tests/spec/glsl-1.10/execution/built-in-functions/vs-op-le-int-int-using-if.shader_test -auto
PIGLIT: {"result": "pass" }


Piglit
returncode: 0
result: GPU hang
summary: Piglit/spec_glsl-1.10_execution_built-in-functions_vs-op-le-int-int-using-if    GPU_hang

> pass count is similar to BDW.
> One case has GPU hang issue, the GPU hang issue is unstable. I am unable
> reproduce it manually, 
> 

Due to the "fail" goes away, file Bug 87138 to track GPU hang, Close this bug.

Verified.Fixed.