KhronosGroup / KhronosGroup/OpenCL-Docs
Implicit command queue flushes triggered by `clReleaseCommandQueue`
- Dominant language
- Python
- Stars
- 420
- Forks
- 131
- Avg merge
- 5d 13h
- Merged PRs (30d)
- 11
Description
The specification currently states
>`clReleaseCommandQueue` performs an implicit flush to issue any previously queued OpenCL
commands in command_queue.
A few implementers observed in the 2024/09/03 teleconference that this leads to unwanted flushes when RAII objects that hold command queues frequently call `clRetainCommandQueue` and `clReleaseCommandQueue`. The intent may have been to solve the issue described in https://github.com/KhronosGroup/OpenCL-Docs/issues/1238.
Do people rely on this behaviour? Is this something we want to relax? Would we want explicit opt-in/opt-out control for this behaviour?
Contributor guide
Research direction
Start by reading the cited issue #1238 and the current specification wording quoted here, then review the 2024/09/03 teleconference context. Determine whether the implicit flush behavior should remain, be relaxed, or gain explicit control; the issue is done only when that design question has consensus and the specification change is defined.
Written by the indexing model from the issue text.
Assessment
- Domain
- documentation
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 20/100