Skip to content

cleanup: no more lock around GPU workloads #890

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
georgeh0 merged 1 commit into from
Aug 20, 2025
Merged

cleanup: no more lock around GPU workloads #890

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 1 addition & 17 deletions python/cocoindex/op.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,10 +11,6 @@
Awaitable,
Callable,
Protocol,
ParamSpec,
TypeVar,
Type,
cast,
dataclass_transform,
Annotated,
get_args,
Expand Down Expand Up @@ -125,9 +121,6 @@ def __call__(
return (result_type, executor)


_gpu_dispatch_lock = asyncio.Lock()


_COCOINDEX_ATTR_PREFIX = "cocoindex.io/"


Expand Down Expand Up @@ -348,16 +341,7 @@ async def __call__(self, *args: Any, **kwargs: Any) -> Any:
decoded_kwargs[kwarg_name] = kwarg_info.decoder(arg)

assert self._acall is not None
if op_args.gpu:
# For GPU executions, data-level parallelism is applied, so we don't want to
# execute different tasks in parallel.
# Besides, multiprocessing is more appropriate for pytorch.
# For now, we use a lock to ensure only one task is executed at a time.
# TODO: Implement multi-processing dispatching.
async with _gpu_dispatch_lock:
output = await self._acall(*decoded_args, **decoded_kwargs)
else:
output = await self._acall(*decoded_args, **decoded_kwargs)
output = await self._acall(*decoded_args, **decoded_kwargs)
return self._result_encoder(output)

def enable_cache(self) -> bool:
Expand Down
Loading