Bug 102995

Summary:	[BAT] igt@gem_mmap_gtt@basic-wc - Failed assertion: gtt_writes > cpu_writes/2
Product:	DRI	Reporter:	Marta Löfstedt <marta.lofstedt>
Component:	DRM/Intel	Assignee:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Status:	CLOSED FIXED	QA Contact:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity:	critical
Priority:	high	CC:	intel-gfx-bugs
Version:	DRI git
Hardware:	Other
OS:	All
Whiteboard:	ReadyForDev
i915 platform:	BYT	i915 features:	GEM/Other

Description Marta Löfstedt 2017-09-26 09:08:16 UTC

CI_DRM_3133 fi-byt-n2820: 

(gem_mmap_gtt:3108) CRITICAL: Test assertion failure function test_wc, file gem_mmap_gtt.c:286:
(gem_mmap_gtt:3108) CRITICAL: Failed assertion: gtt_writes > cpu_writes/2
(gem_mmap_gtt:3108) CRITICAL: Write-Combined writes are expected to be roughly equivalent to WB writes: WC (gtt)=1686.82MiB/s, WB (cpu)=3437.54MiB/s
Subtest basic-wc failed.


https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_3133/fi-byt-n2820/igt@gem_mmap_gtt@basic-wc.html

Comment 1 Chris Wilson 2017-09-26 09:18:16 UTC

Hence why using GTT is such a big performance hit on byt. This test is to enumerate that hw regression.

Comment 2 Elizabeth 2017-09-29 20:20:03 UTC

Raising priority since it's BAT test.

Comment 3 Elizabeth 2017-10-10 20:53:58 UTC

Please note this is a dup of bug 97942 and was moved to IGT component. Also note: 
https://bugs.freedesktop.org/show_bug.cgi?id=97942#c1
(In reply to Chris Wilson from comment #1)
> Yes, expected failure for this *hardware*.

Comment 4 Elizabeth 2017-10-30 18:36:17 UTC

We are still hitting the same issue on BYT with: IGT-Version: 1.20-g1fc4de1 (x86_64) (Linux: 4.14.0-rc6-drm-intel-qa-ww44-commit-2badde8+ x86_64)

Comment 5 Marta Löfstedt 2017-11-01 13:44:03 UTC

Never seen again. Closing

Comment 6 Elizabeth 2017-11-07 16:07:00 UTC

On QA BYT the issue still exists, not sure what's different.
$ : time sudo -E ./gem_mmap_gtt --r basic-wc
IGT-Version: 1.20-g9fe5a9a (x86_64) (Linux: 4.14.0-rc8-drm-intel-qa-ww45-commit-b911f67+ x86_64)
(gem_mmap_gtt:6261) CRITICAL: Test assertion failure function test_wc, file gem_mmap_gtt.c:286:
(gem_mmap_gtt:6261) CRITICAL: Failed assertion: gtt_writes > cpu_writes/2
(gem_mmap_gtt:6261) CRITICAL: Write-Combined writes are expected to be roughly equivalent to WB writes: WC (gtt)=2623.16MiB/s, WB (cpu)=5517.15MiB/s

Comment 7 Elizabeth 2017-11-14 16:32:11 UTC

Still present on QA with IGT-Version: 1.20-gf370d59 (x86_64) (Linux: 4.14.0-drm-intel-qa-ww46-commit-712af85+ x86_64)

Comment 8 Elizabeth 2017-11-21 19:49:04 UTC

Still there with IGT-Version: 1.20-gf8f6db9 (x86_64) (Linux: 4.14.0-drm-intel-qa-ww47-commit-f710441+ x86_64)

(gem_mmap_gtt:14136) CRITICAL: Test assertion failure function test_wc, file gem_mmap_gtt.c:286:
(gem_mmap_gtt:14136) CRITICAL: Failed assertion: gtt_writes > cpu_writes/2
(gem_mmap_gtt:14136) CRITICAL: Write-Combined writes are expected to be roughly equivalent to WB writes: WC (gtt)=2553.05MiB/s, WB (cpu)=5613.75MiB/s
Subtest basic-wc failed.

Comment 9 Elizabeth 2017-11-29 23:38:57 UTC

IGT-Version: 1.20-g380cc81 (x86_64) (Linux: 4.15.0-rc1-drm-intel-qa-ww48-commit-0645c6d+ x86_64), same Failed assertion: gtt_writes > cpu_writes/2

Comment 10 Elizabeth 2017-12-06 15:51:57 UTC

Still failing, IGT-Version: 1.20-g1db1246 (x86_64) (Linux: 4.15.0-rc2-drm-intel-qa-ww49-commit-1a0d67e+ x86_64)
Will try to bisect.

Comment 11 Chris Wilson 2017-12-06 18:08:59 UTC

(In reply to Elizabeth from comment #10)
> Still failing, IGT-Version: 1.20-g1db1246 (x86_64) (Linux:
> 4.15.0-rc2-drm-intel-qa-ww49-commit-1a0d67e+ x86_64)
> Will try to bisect.

It's meant to fail. The HW didn't live up to expectations, the test is trying to document that and warn others not to use GTT.

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.