Repository navigation

cuda.core: add a bounded asyncio wait for NVML device and system events - #2924

Open

0z5a wants to merge 14 commits into

NVIDIA:mainfrom

0z5a:cp1458-async-nvml-events

Open

cuda.core: add a bounded asyncio wait for NVML device and system events#2924
0z5a wants to merge 14 commits into
NVIDIA:mainfrom
0z5a:cp1458-async-nvml-events

Commits on Sep 24, 2026
- Add wait_async to NVML device and system event sets
  0z5a
  committed
  2896838
- Add contract tests for bounded async NVML event waits
  0z5a
  committed
  d54d4b3
- Treat a slice cancelled before it ran as no result
  0z5a
  committed
  a11b05b
- Run wait slices on reused workers instead of per-request pool tasks
  0z5a
  committed
  934e26c
- Complete a drained slice from the worker that owns it
  0z5a
  committed
  79c7f1c
- Return the wait coroutine directly from wait_async
  0z5a
  committed
  5ec11f7
- Restore async def methods and satisfy pre-commit
  0z5a
  committed
  28dbed4
- Type the dispatcher's queue, thread list and request slot
  0z5a
  committed
  3c7dfff
- Cover resource accounting, saturated scheduling and cross-loop reuse
  0z5a
  committed
  ab5b1a7
Commits on Oct 5, 2026
- Fix NVML event waiting after review
  0z5a
  committed
  30feadc
- Merge current main into async NVML events
  0z5a
  committed
  2252549
- Merge latest main and reconcile NVML event release notes
  0z5a
  committed
  5199f5e
Commits on Oct 6, 2026
- Merge branch 'main' into cp1458-async-nvml-events
  mdboom
  authored
  a0faf4a
Commits on Oct 16, 2026
- Merge branch 'main' into cp1458-async-nvml-events
  mdboom
  authored
  ecd2f08