New issue
Advanced search Search tips
Note: Color blocks (like or ) mean that a user may not be available. Tooltip shows the reason.

Issue 922205 link

Starred by 2 users

Issue metadata

Status: Untriaged
Owner:
Cc:
EstimatedDays: ----
NextAction: ----
OS: ----
Pri: 1
Type: Bug



Sign in to add a comment

Measure crostini GPU performance.

Project Member Reported by cylee@chromium.org, Jan 15

Issue description

I've tried to install glmark2 and run it on both host and containers on eve.

I first tried to configure glmark2 to use OpenGL (not ES) in the container and it throws  errors like
** GLX does not support GLX_EXT_swap_control or GLX_MESA_swap_control!
** Failed to set swap interval. Results may be bounded above by refresh rate.

If I configure it to use OpenGL ES, it still gives me an error like
** Failed to set swap interval. Results may be bounded above by refresh rate.

But it still reports some numbers with those errors:

* Host
=======================================================
glmark2 2012.12
=======================================================
OpenGL Information
GL_VENDOR:     Intel Open Source Technology Center
GL_RENDERER:   Mesa DRI Intel(R) HD Graphics 615 (Kaby Lake GT2)
GL_VERSION:    OpenGL ES 3.2 Mesa 18.2.0-devel (git-0f959215c3)
...
glmark2 Score: 316


* Container using openGL
=======================================================
    glmark2 2017.07
=======================================================
    OpenGL Information
    GL_VENDOR:     VMware, Inc.
    GL_RENDERER:   Gallium 0.4 on llvmpipe (LLVM 3.9, 256 bits)
    GL_VERSION:    3.0 Mesa 13.0.6
=======================================================
...
glmark2 Score: 193

* Container using OpenGL ES
=======================================================
    glmark2 2017.07
=======================================================
    OpenGL Information
    GL_VENDOR:     VMware, Inc.
    GL_RENDERER:   Gallium 0.4 on llvmpipe (LLVM 3.9, 256 bits)
    GL_VERSION:    OpenGL ES 3.0 Mesa 13.0.6
=======================================================
...
glmark2 Score: 153


I'm studying why it failed to set the swap rate. With current settings the performance drop is significant. Full logs are attached.

 
glmark2_container
7.8 KB View Download
glmark2_es_container
5.3 KB View Download
glmark2_host
3.0 KB View Download
Cc: davidri...@chromium.org
Cc: selcott@chromium.org
I don't recall if I tried running with desktop GL or just ES.  crbug.com/916721 had my findings on glmark2 before, which I built with:
git clone https://github.com/glmark2/glmark2
sudo apt-get install libjpeg-dev libgles2-mesa-dev
./waf configure --with-flavors=drm-gl,drm-glesv2,wayland-gl,wayland-glesv2,x11-gl,x11-glesv2
./waf
sudo ./waf install

Comment 4 by cylee@chromium.org, Jan 16 (6 days ago)

Hi David,

  Are you using your own container instead of the default "penguin" container? I see your environment like:
=======================================================
    OpenGL Information
    GL_VENDOR:     Red Hat
    GL_RENDERER:   virgl
    GL_VERSION:    OpenGL ES 3.2 Mesa 18.2.6 (git-3e741344d7)
=======================================================

For the build command, I'm using roughly the same as yours:

git clone https://github.com/glmark2/glmark2
sudo apt-get install libx11-dev libgl1-mesa-dev libgles2-mesa-dev
sudo apt-get install libjpeg-dev libpng-dev 
./waf configure --with-flavors=x11-gl,x11-glesv2
./waf
sudo ./waf install
(Your configuration enabled more flavors and a I missed some dependency like "gbm".  I think we only need x11-gl or x11-glesv2 here).

But I still get the error for glmark2-es2
** Failed to set swap interval. Results may be bounded above by refresh rate.
So the score remains low
=======================================================
                                  glmark2 Score: 154
=======================================================

I think our running environment (container settings) are different. Could you provide some info about it?

Comment 5 by davidri...@chromium.org, Jan 16 (6 days ago)

I think for the bug where I tried to narrow it down I was running sid on crosvm on either eve or nami.  The instructions at go/crosvm-gpu-care should get you to what I was running.

Sign in to add a comment