- 19 Mar, 2021 1 commit
-
-
Emma Anholt authored
428k trace, nice license, I remember when its shader used to be "hard". First frame of the demo in commit bc2da506d23bdaf3eb0d95c9649a4591b4d912c7 of https://github.com/behdad/glyphy Reviewed-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Part-of: <!15>
-
- 17 Mar, 2021 2 commits
-
-
Gert Wollny authored
The traces that were already in the repository were converted to snappy, then trimmed to their last frame Nrespectively, and repacked back to brotli with quality 11. CS, CS-source, and Portal2, didn't require any specific handling and were trimmed by running apitrace gltrim -f N -o $f-v2.trace $f.trace For Half Life 2 additional setup frames were needed and the trim command command was apitrace gltrim -s 243-476 -f490 $f-v2.trace $f.trace The resulting traces all consist of a setup frame an the final frame from the original trace. Signed-off-by:
Gert Wollny <gert.wollny@collybora.com> Reviewed-by:
Eric Anholt <eric@anholt.net> Part-of: <!30>
-
Version: 2.09/64, linux-x86_64, code revision 9108 The trace is trimmed to contain one setup frame and one rendered frame a few frames after the beginning of a new mission "The Job". The trace was trimmed using apitrace gltrim -f 700 -o thedarkmod.trace thedarkmod-orig.trace and then repacked using brotli with quality 11. Related #2 Signed-off-by:
Gert Wollny <gert.wollny@collabora.com> Reviewed-by:
Eric Anholt <eric@anholt.net> Part-of: <!29>
-
- 18 Feb, 2021 1 commit
-
-
Gert Wollny authored
Version: 20210203 (142693b) on: Linux The trace is trimmed to contain one setup frame and one rendered frame right after the loading screen. The trace was trimmed using apitrace gltrim -s 0-14,483 -f 484 -o pioneer.trace Pioneer-orig.trace and then repacked using brotli with quality 11. Related #2 Signed-off-by:
Gert Wollny <gert.wollny@collabora.com> Reviewed-by:
Eric Anholt <eric@anholt.net>
-
- 20 Jan, 2021 1 commit
-
-
Emma Anholt authored
I got an ack from @Plagman to include Valve titles in traces-db. These come from the traces that Chrome OS uses, trimmed by me to be shorter (just get to an interesting frame in the game), and repacked to brotli level 11 to save disk space. CS: 17MB CS Source: 71MB HL2: 102MB Portal 2: 223MB Acked-by:
Daniel Stone <daniels@collabora.com> Acked-by:
Pierre-Loup Griffais <pgriffais@valvesoftware.com> Part-of: <!27>
-
- 19 Jan, 2021 1 commit
-
- 21 Jul, 2020 2 commits
-
-
Rob Clark authored
From xonotic 0.8.2, with cmdline: xonotic-glx -benchmark demos/the-big-keybench +exec effects-high.cfg Reviewed-by:
Eric Anholt <eric@anholt.net> Part-of: <!21>
-
Rob Clark authored
From neverball 1.6.0 The "stars" above the target use point-sprites, so this is an interesting thing to get CI coverage of. Reviewed-by:
Eric Anholt <eric@anholt.net> Part-of: <!20>
-
- 20 Jul, 2020 4 commits
-
-
Emma Anholt authored
Trace is the first two frames from https://github.com/itoral/gl-terrain-demo/ efbe5e77268f ("terrain: make sure we crate the index buffer before we upload to it") Acked-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Part-of: <!16>
-
Emma Anholt authored
The copyright files from debian are included. I just started a match and took the frame where call count stabilized. 95MB. Acked-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Part-of: <!19>
-
Emma Anholt authored
We don't need 318 frames of gears (which, for some reason, takes about a second per frame on freedreno). This gets us down to a reasonable runtime, while still having a steady-state frame after the setup one. Acked-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Part-of: <!17>
-
Emma Anholt authored
Reviewed-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Acked-by:
Daniel Stone <daniels@collabora.com>
-
- 19 Jul, 2020 2 commits
-
-
Vulkan traces are not device independant. Although gfxreconstruct is working in some portability mode, they will still depend on the exposed extensions at the point of capturing. We now specify that the existing trace was captured with an Intel gen9 device (Kaby Lake) and add an additional trace captured with an AMD POLARIS10 device. Signed-off-by:
Andres Gomez <agomez@igalia.com>
-
Vulkan traces are not device independant. Although gfxreconstruct is working in some portability mode, they will still depend on the exposed extensions at the point of capturing. We now specify that the existing trace was captured with an Intel gen9 device (Kaby Lake) and add an additional trace captured with an AMD POLARIS10 device. Signed-off-by:
Andres Gomez <agomez@igalia.com>
-
- 16 Jul, 2020 2 commits
-
-
Tomeu Vizoso authored
Add a job that will make sure that all traces are stored in MinIO, for cached download of traces. Traces added to the upstream repository will be uploaded to: https://minio-packet.freedesktop.org/mesa-tracie-public/ Traces added to a fork will be uploaded to: https://minio-packet.freedesktop.org/artifacts/$CI_PROJECT_PATH/$CI_PIPELINE_ID/ Remember to update the traces-db:download-url property in the traces.yml file when testing with trace-db forks. Signed-off-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Reviewed-by:
Andres Gomez <agomez@igalia.com> Reviewed-By:
Rohan Garrg <rohan.garg@collabora.com>
-
Tomeu Vizoso authored
The caching system that we are going to implement so devices don't have to pull from git-lfs requires that files' contents are immutable. So add a CI job that checks that there's no modifications to existing traces, and recommend in that case to append a version suffix to the new file. Signed-off-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com> Reviewed-by:
Andres Gomez <agomez@igalia.com> Reviewed-By:
Rohan Garrg <rohan.garg@collabora.com>
-
- 20 Apr, 2020 1 commit
-
-
Rohan Garg authored
This reverts commit e119f323. Fixes: 1
-
- 16 Apr, 2020 4 commits
-
-
Based off of https://bugs.freedesktop.org/show_bug.cgi?id=27343 I've added a capture for Celestia. This was tricky due to the fact that Celestia has no OpenGL 3.2+ support for renderdoc and the traces from apitrace were too big ( ~700MB ). Instead, we've captured a renderdoc trace of the apitrace replay. Note that we cycle through the rendering modes (Ctrl+V) when capturing the trace with apitrace as mentioned in the bug report in order to faithfully reproduce the bug. The trace was captured with Celestia version 1.6.1 Signed-off-by:
Rohan Garg <rohan.garg@collabora.com> Reviewed-by:
Eric Anholt <eric@anholt.net>
-
This trace captures the 'Material Testers' example from the Godot demo projects (https://github.com/godotengine/godot-demo-projects/ ) with Godot engine 3.2.1. Signed-off-by:
Rohan Garg <rohan.garg@collabora.com> Reviewed-by:
Eric Anholt <eric@anholt.net>
-
From commit ce3b40cd5c9e73dcfa71e38702e516caf479451d on master on Intel iris. Requires GL 3.3. <2MB of traces.
-
12MB for 2-frame captures (1 setup frame and 1 actual render). MESA_GL_VERSION_OVERRIDE was used to ensure that the benchmark didn't use a higher context version than it needed (otherwise, all of them required GL 4.6, even triangle which just had a basic color shader in GLSL #version 150 mode. Requirements: GL 4.0: tessmark, julia fp32 GL 3.3: gimark GL 2.1: the rest
-
- 24 Mar, 2020 1 commit
-
-
Emma Anholt authored
-
- 23 Mar, 2020 1 commit
-
-
HelloWorld and Cloth Physics Test. Signed-off-by:
Andres Gomez <agomez@igalia.com>
-
- 19 Mar, 2020 3 commits
-
-
Rohan Garg authored
Signed-off-by:
Rohan Garg <rohan.garg@collabora.com>
-
HelloWorld and Lightmap Bake Test. Signed-off-by:
Andres Gomez <agomez@igalia.com>
-
-
- 12 Mar, 2020 1 commit
-
-
Rohan Garg authored
Signed-off-by:
Rohan Garg <rohan.garg@collabora.com>
-
- 04 Mar, 2020 4 commits
-
-
Alexandros Frantzis authored
Since we are now using checksums to track image changes in Mesa master tracie jobs, the reference images here are unused and are at risk of getting out of sync with their respective checksums. Remove them to avoid bit rot. An up to date set of reference images can be created by using the TRACIE_STORE_IMAGES=1 CI/Pipeline environment variable in Mesa CI runs. Signed-off-by:
Alexandros Frantzis <alexandros.frantzis@collabora.com>
-
Signed-off-by:
Andres Gomez <agomez@igalia.com>
-
Signed-off-by:
Andres Gomez <agomez@igalia.com>
-
-
- 22 Oct, 2019 1 commit
-
-
Tomeu Vizoso authored
Signed-off-by:
Tomeu Vizoso <tomeu.vizoso@collabora.com>
-
- 26 Sep, 2019 1 commit
-
-
Alexandros Frantzis authored
Signed-off-by:
Alexandros Frantzis <alexandros.frantzis@collabora.com>
-
- 25 Sep, 2019 1 commit
-
-
Alexandros Frantzis authored
Mesa commit [1] results in slight changes to the produced supertuxkart images. [1] mesa/mesa@92f70df8 Signed-off-by:
Alexandros Frantzis <alexandros.frantzis@collabora.com>
-
- 24 Sep, 2019 1 commit
-
-
Alexandros Frantzis authored
Only run them on intel for now since it's not compatible with softpipe. Signed-off-by:
Alexandros Frantzis <alexandros.frantzis@collabora.com>
-
- 28 Aug, 2019 1 commit
-
-
Gert Wollny authored
-
- 27 Aug, 2019 1 commit
-
-
Gert Wollny authored
-
- 26 Aug, 2019 1 commit
-
-
Gert Wollny authored
Maybeit fails, then it would be a nice test for now Signed-off-by:
Gert Wollny <gw.fossdev@gmail.com>
-
- 26 Jul, 2019 1 commit
-
-
Elie Tournier authored
Signed-off-by:
Elie Tournier <elie.tournier@collabora.com>
-
- 10 Jul, 2019 1 commit
-
-
Elie Tournier authored
Signed-off-by:
Elie Tournier <elie.tournier@collabora.com>
-