| Ilja H. Friedel | 8faad30 | 2011-04-26 14:40:49 -0700 | [diff] [blame] | 1 | GLBench runs a bunch of OpenGL or OpenGL ES performance tests and writes |
| 2 | performance numbers to stdout and resulting images to a directory for |
| 3 | verification. |
| 4 | |
| 5 | For the test to pass the performance numbers have to be better than a predefined |
| 6 | threshold, while the resulting images have to be found in a repository of |
| 7 | reference images. As the image name encodes the raw pixel MD5 this can be |
| 8 | done as a simple file existence check. If we ever get too much pixel |
| 9 | variation using a tool like perceptualdiff to waive small differences |
| 10 | should be acceptable. |
| 11 | |
| 12 | |
| 13 | Executable options |
| 14 | ================== |
| 15 | |
| 16 | ./glbench [-save [-outdir=<directory>]] |
| 17 | |
| 18 | |
| 19 | Example |
| 20 | ======= |
| 21 | |
| 22 | ./glbench -save -outdir=img |
| 23 | # board_id: NVIDIA Corporation - Quadro FX 380/PCI/SSE2 |
| Daniel Kurtz | be25242 | 2014-06-18 12:44:26 +0800 | [diff] [blame] | 24 | swap_swap = 214.77 us [swap_swap.pixmd5-20dbc406b95e214a799a6a7f9c700d2f.png] |
| 25 | clear_color = 4448.28 mpixels_sec [clear_color.pixmd5-e3609de1022a164fe240a562c69367de.png] |
| 26 | clear_depth = 10199.76 mpixels_sec [clear_depth.pixmd5-e3609de1022a164fe240a562c69367de.png] |
| 27 | clear_colordepth = 3250.57 mpixels_sec [clear_colordepth.pixmd5-e3609de1022a164fe240a562c69367de.png] |
| 28 | clear_depthstencil = 26447.22 mpixels_sec [clear_depthstencil.pixmd5-e3609de1022a164fe240a562c69367de.png] |
| Ilja H. Friedel | 8faad30 | 2011-04-26 14:40:49 -0700 | [diff] [blame] | 29 | [...] |
| 30 | |
| 31 | ls img |
| Daniel Kurtz | be25242 | 2014-06-18 12:44:26 +0800 | [diff] [blame] | 32 | clear_color.pixmd5-e3609de122a164fe240a562c69367de.png |
| 33 | clear_colordepth.pixmd5-e3609de122a164fe240a562c69367de.png |
| 34 | clear_colordepthstencil.pixmd5-e3609de122a164fe240a562c69367de.png |
| 35 | compositing.pixmd5-7d02a16a7ac15cd6cbbc5c786f1.png |
| Ilja H. Friedel | 8faad30 | 2011-04-26 14:40:49 -0700 | [diff] [blame] | 36 | [...] |
| 37 | |
| 38 | |
| 39 | Running from the autotest harness |
| 40 | ================================= |
| 41 | |
| 42 | The autotest script graphics_GLBench.py will |
| 43 | 0) run glbench -save |
| 44 | 1) first try to identify known buggy images by searching in |
| 45 | deps/glbench/glbench_knownbad_images.txt |
| 46 | 2) then identify good images by searching in |
| 47 | deps/glbench/glbench_reference_images.txt |
| 48 | 3) if it is unable to find the test image it will do a second lookup at |
| 49 | http://commondatastorage.googleapis.com/chromeos-localmirror/distfiles/glbench_* |
| 50 | 4) TODO(ihf) use perceptualdiff to do a fuzzy compare |
| 51 | 5) raise an error if the image is completely unknown |
| 52 | 6) report performance numbers back to the harness |
| 53 | |
| 54 | |
| 55 | Handling of reference images |
| 56 | ============================ |
| 57 | |
| 58 | Good reference images themselves are located at ../glbench-images/glbench_reference_images/ |
| Ilja H. Friedel | 6eec492 | 2015-04-17 18:44:03 -0700 | [diff] [blame] | 59 | Images that have outstanding defects and an open bug filed are at ../glbench-images/glbench_knownbad_images/chromium-bug-NNNNN/ |
| 60 | When that bug is closed the directory should be moved to ../glbench-images/glbench_fixedbad_images/chromium-bug-NNNNN/ |
| Ilja H. Friedel | 8faad30 | 2011-04-26 14:40:49 -0700 | [diff] [blame] | 61 | |
| 62 | To push out new reference images place them in the appropriate |
| 63 | directories (create a new bug if needed) and run |
| 64 | > update_glbench_image_filelists.sh |
| 65 | |
| Ilja H. Friedel | 6eec492 | 2015-04-17 18:44:03 -0700 | [diff] [blame] | 66 | to update the image filelists in deps/glbench/glbench_reference_images.txt etc. |