GLBench runs a bunch of OpenGL or OpenGL ES performance tests and writes 
performance numbers to stdout and resulting images to a directory for 
verification.

For the test to pass the performance numbers have to be better than a predefined 
threshold, while the resulting images have to be found in a repository of 
reference images. As the image name encodes the raw pixel MD5 this can be 
done as a simple file existence check. If we ever get too much pixel 
variation using a tool like perceptualdiff to waive small differences 
should be acceptable.


Executable options
==================

./glbench [-save [-outdir=<directory>]]


Example
=======

./glbench -save -outdir=img
# board_id: NVIDIA Corporation - Quadro FX 380/PCI/SSE2
us_swap_swap                    =   214.77   [us_swap_swap.pixmd5-20dbc406b95e214a799a6a7f9c700d2f.png]
mpixels_sec_clear_color         =  4448.28   [mpixels_sec_clear_color.pixmd5-e3609de1022a164fe240a562c69367de.png]
mpixels_sec_clear_depth         = 10199.76   [mpixels_sec_clear_depth.pixmd5-e3609de1022a164fe240a562c69367de.png]
mpixels_sec_clear_colordepth    =  3250.57   [mpixels_sec_clear_colordepth.pixmd5-e3609de1022a164fe240a562c69367de.png]
mpixels_sec_clear_depthstencil  = 26447.22   [mpixels_sec_clear_depthstencil.pixmd5-e3609de1022a164fe240a562c69367de.png]
[...]

ls img
1280x768_fps_compositing.pixmd5-7d02a16a7ac15cd6cbbc5c786f1.png
1280x768_fps_no_fill_compositing.pixmd5-642537874c2920ca7ddf48d08c17f74a.png
mpixels_sec_clear_color.pixmd5-e3609de122a164fe240a562c69367de.png
mpixels_sec_clear_colordepth.pixmd5-e3609de122a164fe240a562c69367de.png
mpixels_sec_clear_colordepthstencil.pixmd5-e3609de122a164fe240a562c69367de.png
[...]


Running from the autotest harness
=================================

The autotest script graphics_GLBench.py will 
0) run glbench -save
1) first try to identify known buggy images by searching in 
       deps/glbench/glbench_knownbad_images.txt
2) then identify good images by searching in 
       deps/glbench/glbench_reference_images.txt
3) if it is unable to find the test image it will do a second lookup at
       http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/glbench_*
4) TODO(ihf) use perceptualdiff to do a fuzzy compare
5) raise an error if the image is completely unknown
6) report performance numbers back to the harness


Handling of reference images
============================

Good reference images themselves are located at       ../glbench-images/glbench_reference_images/
while images that have defects and a bug filed are at ../glbench-images/glbench_knownbad_images/cros-bug-NNNNN/

To push out new reference images place them in the appropriate 
directories (create a new bug if needed) and run 
    > update_glbench_image_filelists.sh 

to update the image filelists in deps/glbench/glbench_reference_images.txt
and deps/glbench/glbench_knownbad_images.txt 

Unfortunately one has to manually keep the web locations in sync. 
But they are not critical for now.
http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/glbench_reference_images/
http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/glbench_knownbad_images/