Bug 99343

Summary:	dEQP-GLES31.functional.shaders.builtin_functions.precision.{min,max}.highp_compute.scalar failures
Product:	Mesa	Reporter:	Samuel Pitoiset <samuel.pitoiset>
Component:	Drivers/Gallium/radeonsi	Assignee:	Default DRI bug account <dri-devel>
Status:	RESOLVED MOVED	QA Contact:	Default DRI bug account <dri-devel>
Severity:	normal
Priority:	medium
Version:	git
Hardware:	Other
OS:	All
Whiteboard:
i915 platform:		i915 features:

Description Samuel Pitoiset 2017-01-10 12:50:30 UTC

Hi,

I have noticed that the following dEQP GLES31 tests fail with latest mesa/llvm:

dEQP-GLES31.functional.shaders.builtin_functions.precision.min.highp_compute.scalar
dEQP-GLES31.functional.shaders.builtin_functions.precision.max.highp_compute.scalar

The same tests with lowp instead of highp work fine. Note that it's unrelated to compute because they also fail with vertex shaders.

After glancing at the code, it seems like enabling denorms for 32-bit floats fix them. But it's currently only enabled for 64-bit and 16-bit floats for some reasons.

Marek, Nicolai, any thoughts about that?

I didn't try if these tests pass with Catalyst though, but they do at least on NVIDIA hw.

Comment 1 Marek Olšák 2017-01-11 14:58:18 UTC

Only VI can do FP32 denormals at full rate, which means only VI can enable them and not lose performance.

Comment 2 GitLab Migration User 2019-09-25 17:56:42 UTC

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1247.

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.