Bug 38942

Summary: Mathematica Plot3D gpu hang
Product: Mesa Reporter: Ivan Iakoupov <voxiac>
Component: Drivers/DRI/i965Assignee: Ian Romanick <idr>
Severity: normal    
Priority: medium    
Version: git   
Hardware: Other   
OS: All   
i915 platform: i915 features:
Attachments: trace of opening a file with Plot3D in it
A Mathematica 8 notebook with a Plot3D command

Description Ivan Iakoupov 2011-07-04 02:58:42 UTC
Created attachment 48738 [details]
trace of opening a file with Plot3D in it

This is qualitatively different from bug 37177, as steps to reproduce are simply:


Plot3D[Sin[x*y], {x, 0, 1}, {y, 0, 1}]

in Mathematica Notebook and Shift+Enter to run. Before the graph appears the gpu hangs.
Alternatively if one has any Mathematica notebooks with a Plot3D commands, attempting to open them results in a gpu hang.

I did a bisect and it says that it's caused by

commit c173541d9769d41a85cc899bc49699a3587df4bf
Author: Eric Anholt
Date:   Wed Apr 27 13:33:10 2011 -0700

    i965: Use state streaming on programs, and state base address on gen5+.
    There will be a little bit of thrashing of the program cache BO as the
    cache warms up, but once the application is in steady state, this
    reduces relocations on gen5 and later.
    On my T420 laptop, cairogl firefox-talos-gfx performance improves 2.6%
    +/- 1.3% (n=6).  No statistically significant performance difference
    on nexuiz (n=5).

My system is kernel 3.0-rc5, libdrm-2.5.26, xf86-video-intel-2.15,
xserver-1.10.2 and

OpenGL renderer string: Mesa DRI Mobile IntelĀ® GM45 Express Chipset 
OpenGL version string: 2.1 Mesa 7.12-devel (git-424b121)

Now I'm not sure how to help you reproduce this hang. There's
which is supposed to be able to open Mathematica worksheets and hopefully renders then in the same way but it's only 32 bit build which I cannot test as my system is pure 64 bit. If you make it work I can attach a sample Mathematica file to test with.

I also tried making an apitrace file for it but since apitrace doesn't support replaying glXCreateGLXPixmap calls may be not reproducing what Mathematica does. Replaying it does lead to some gpu hang (but this one is partially recoverable compared to non-recoverable one from Mathematica), so I'll attach it anyway.
Comment 1 Ivan Iakoupov 2011-07-04 03:01:41 UTC
Created attachment 48739 [details]
A Mathematica 8 notebook with a Plot3D command
Comment 2 Ivan Iakoupov 2011-07-09 10:35:20 UTC
I cannot reproduce it with master anymore. I guess it's the same as bug 38771.

*** This bug has been marked as a duplicate of bug 38771 ***

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.