Add Rust CUDA buffer bindings - #15
Draft
yuanknv wants to merge 12 commits into
Draft
Conversation
Signed-off-by: CY Chen <cyc@nvidia.com>
Signed-off-by: CY Chen <cyc@nvidia.com>
Signed-off-by: CY Chen <cyc@nvidia.com>
Expose typed CPU and CUDA tensor-message conversions, preserve DLPack-backed buffer ownership, and validate zero-copy inter-process rclpy transport.
Add an independent Python Torch vendor and keep CUDA integration optional so CPU and CUDA platforms use the same conversion API.
Keep package documentation centralized and avoid duplicating platform behavior details.
Rename the Python package to pytorch_vendor and align both vendor selectors with official CUDA 13.0 and 13.2 artifacts.
Separate CPU and CUDA conversion behavior behind a private registry so future hardware support does not change the public API.
Make the private extension's scope explicit so future hardware adapters can provide distinct native bridges.
Use a versioned C ABI so CPU users are not forced to install CUDA buffer runtime dependencies.
Keep the core CPU-capable package platform-independent and discover optional conversion adapters through Python entry points.
Expose CUDA buffer allocation through an exception-safe C ABI and a safe Rust RAII package with interprocess zero-copy validation.
yuanknv
force-pushed
the
feature/torch-conversions-py
branch
from
August 27, 2026 23:04
4dddf00 to
f5c5175
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Dependencies
This PR targets
feature/torch-conversions-py, which contains the CUDA/Python backend stack used as its base.Test plan
cudaMemcpy*calls with Nsight Systems