Summary: | [ALL bisected] Mesa change cause performance case manhattan fail. | ||
---|---|---|---|
Product: | Mesa | Reporter: | Ding Heng <hengx.ding> |
Component: | Mesa core | Assignee: | Tapani Pälli <lemody> |
Status: | VERIFIED FIXED | QA Contact: | mesa-dev |
Severity: | normal | ||
Priority: | medium | CC: | christophe.prigent, eero.t.tamminen, valtteri.rantala |
Version: | git | ||
Hardware: | Other | ||
OS: | All | ||
Whiteboard: | |||
i915 platform: | i915 features: | ||
Attachments: |
xorg log
patch to fix the issue fix using static values fix using static values |
Description
Ding Heng
2015-06-01 03:33:18 UTC
Heng, Manhattan and Unigine Valley don't render correctly anymore in our tests (e.g. on BSW). As both require overrides to render correctly, could you check whether that issue is also caused by the Mesa change in this bug? PS. Please use --format=fuller when pasting commit info (this seems to be added to Mesa on 28th of April). Created attachment 116229 [details] [review] patch to fix the issue this fixes the issue, will have to test with different override combinations (In reply to Tapani Pälli from comment #2) > Created attachment 116229 [details] [review] [review] > patch to fix the issue > > this fixes the issue, will have to test with different override combinations This patch has fix this issue. Change state to resolved. (In reply to Ding Heng from comment #3) > (In reply to Tapani Pälli from comment #2) > > Created attachment 116229 [details] [review] [review] [review] > > patch to fix the issue > > > > this fixes the issue, will have to test with different override combinations > > This patch has fix this issue. Change state to resolved. The patch is not upstream. The bug is not fixed until the patch is upstream. (In reply to Tapani Pälli from comment #2) > Created attachment 116229 [details] [review] [review] > patch to fix the issue > > this fixes the issue, will have to test with different override combinations Ugh... can we make it not do all the work every time a context is created? That was the purpose of the static variables, etc. (In reply to Ian Romanick from comment #5) > (In reply to Tapani Pälli from comment #2) > > Created attachment 116229 [details] [review] [review] [review] > > patch to fix the issue > > > > this fixes the issue, will have to test with different override combinations > > Ugh... can we make it not do all the work every time a context is created? > That was the purpose of the static variables, etc. It should be doable by having static array of versions for each gl_api, then it runs once for each. I can try if this works ok. Created attachment 116276 [details] [review] fix using static values here's alternative fix with using static so that we initialize only once per api Created attachment 116277 [details] [review] fix using static values sorry, there were some issues, here is a fixed version (In reply to Tapani Pälli from comment #8) > Created attachment 116277 [details] [review] [review] > fix using static values > > sorry, there were some issues, here is a fixed version This patch has been test on latest mesa. The result is pass. Fixed in master: commit 7d88ab42b9dda825feddbae774a2a48ddf3cbec2 Author: Tapani Pälli <tapani.palli@intel.com> Date: Tue Jun 16 13:46:47 2015 +0300 mesa: set override_version per api version override Before 9b5e92f get_gl_override was called only once, but now it is called for multiple APIs (GLES2, GL), version needs to be set always. Bug has been verified as fixed with latest Mea master branch: Libdrm: (master)libdrm-2.4.61-13-g97be70b45eccc37e98a1cecf360593f36956ea42 Mesa: (master)028590cbc758e877b963ba430f0a0cb49e882a6b Xserver: (master)xorg-server-1.17.0-158-gfa12f2c150b2f50de9dac4a2b09265f13af353af Xf86_video_intel: (master)2.99.917-366-g87b5ca2c3ed45a46ac037e71e96922b1a7f8c10f Cairo: (master)2cf2d8e340a325adb205baf8e4bd64e1d1858008 Libva: (master)5d07b29687db6d17811b7ecf9b779377e9851a27 Libva_intel_driver: (master)0f6143e2e4bd70217887e1ad86a76200ad8b38a8 Kernel: (drm-intel-nightly)be39a7ed43b45f63aff1ad3b61008f67692db202 |
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.