[Bug 97035] compilation of WebGL demo iamnop.com/particles shader slowed 5x, run-time perf dropped to 1/10th

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=97035

--- Comment #17 from Tapani Pälli <lemody at gmail.com> ---
(In reply to Tapani Pälli from comment #16)
> (In reply to Eero Tamminen from comment #15)
> > (In reply to Tapani Pälli from comment #14)
> > > Testing with current Mesa (06bfb2d28f) on HSW desktop system,
> > 
> > This bug was originally filed against BSW/SKL.
> > 
> > 
> > > page loads in
> > > ~10 secs and runs 60fps meeting the expected result. I'm proposing to
> > > resolve this as FIXED unless still a problem on chromebooks?
> > 
> > Tested Mesa tip (git-379b24a40d) on SKL GT4e.  Issue is still there, page
> > freezes for ~10 seconds and "galaxy" shape renders at 1/60th speed of the
> > other shapes.
> > 
> > (I verified that Chromium is using right Mesa version from "chrome://gpu".)
> 
> Yep, I tried now on KBL and it also reproduces this. Did not notice this is
> gen specific issue.

Which actually means that this is not just 'generic' spilling issue, this is
something specific to how new gens operate. In the shader the return clause of
Fun_E() is what causes trouble. It does probably cause many variables to be
'live' but even just removing couple of those dot() calls is enough for this to
run 60fps on KBL machine.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20170929/2399328a/attachment.html>