[Bug 91857] Mesa 10.6.3 linker is slow

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Fri Sep 4 07:34:15 PDT 2015


--- Comment #9 from Clive McCarthy <CliveMcCarthy at gmail.com> ---
(In reply to Tapani Pälli from comment #8)
> Thanks for attaching the shader, will do some profiling on this later.
> In case this is a 'real life' case and you are having problem with this:
> while this all is 'legal', IMO it would make more sense to write multiple
> shaders for different purposes and then minimize state changes by grouping
> the objects for rendering pass (a lot of apps and games do this). This will
> likely result in better performance on any given platform compared to a
> 'über-shader' approach.
> What comes to 'apples and oranges' comparison, on Nvidia you will need to
> measure without shader cache. According to documentation below you should
> either remove '$HOME/.nv/GLCache' or export __GL_SHADER_DISK_CACHE=false
> before measuring (don't know if these work because docs seem a bit old):
> http://us.download.nvidia.com/XFree86/Linux-x86/319.32/README/
> openglenvvariables.html
> Then the comparison should be more fair.

The shader itself is fast enough and works just fine (weird as it might be).
The speed issue is with the Mesa linker, running x86 code. My über-shader
allows me to try various approaches and experiment very quickly.

The timing of the linker was done with gettimeofday() and represents how long
the linker took before any rendering even began. This is not a rendering

You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20150904/8598e8dc/attachment.html>

More information about the intel-3d-bugs mailing list