| Summary: | dynamic state leak | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Craig Stout <cstout> |
| Component: | Drivers/Vulkan/intel | Assignee: | Intel 3D Bugs Mailing List <intel-3d-bugs> |
| Status: | RESOLVED FIXED | QA Contact: | Intel 3D Bugs Mailing List <intel-3d-bugs> |
| Severity: | normal | ||
| Priority: | medium | CC: | jason |
| Version: | 13.0 | ||
| Hardware: | Other | ||
| OS: | All | ||
| Whiteboard: | |||
| i915 platform: | i915 features: | ||
| Attachments: | workaround or potential fix | ||
Thanks! I never noticed that one. Please send to mesa-dev, drop the Change-id tag, and add Cc: "13.0 17.0" <mesa-stable@lists.freedesktop.org> Will do. What testing should I do before/after posting? (In reply to Craig Stout from comment #2) > Will do. What testing should I do before/after posting? Normally, I'd run the full Vulkan CTS on it. We've got a fast CI system that can do that so just send it to the list and I'll kick off a CI job. Also, please add the following tag as well: Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=100365 Looks like the code in question has changed in branch 17. Is this still worth pushing for 13? That's a good question. 13 is about to go EOL but we might as well fix the bug. Go ahead and send it. Ok will do Jason, cl has been emailed to mesa-dev and mesa-stable and I believe you reviewed it. Let me know if there's any other action for me to take. (In reply to Craig Stout from comment #8) > Let me know if there's any other action for me to take. Just wait for it to show up in the next stable release and close the bug once it does. |
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.
Created attachment 130408 [details] workaround or potential fix We have a vulkan demo that is leaking dynamic state. This demo recently started using a compute shader. I don't have a reduced test case, but check out the attached workaround/fix. Mesa branch 13.0.3 anv_pool_state_emit, which is used only from this one location, requires a matching free. Using the dynamic state stream instead prevents the leak because the stream is cleaned up on finish.