[slimtensor] Add non-owning storage mode for wrapping external memory #16801
Conversation
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
Pull Request resolved: #16769 This diff adds CUDA storage infrastructure to SlimTensor, enabling GPU memory allocation and management. **Key changes:** 1. **`cuda/Guard.h`** - CUDAGuard RAII class: - Saves current CUDA device on construction, restores on destruction - Exception-safe device context switching - Constructors accept device index or Device object 2. **`core/Storage.h`** - Extended for CUDA support: - Added `DeviceTraits<DeviceType::CUDA>` specialization with: - `allocate()` - Uses cudaMalloc with CUDAGuard for device selection - `free()` - Uses cudaFree with warning on error - `memcpy()` - Supports Host↔Device and Device↔Device copies - Added `DEFAULT_CUDA_DEVICE` constant - Updated `MaybeOwningStorage` constructor to handle CUDA devices - Stub implementation when `CUDA_AVAILABLE` is not defined (throws error) ghstack-source-id: 335102161 @exported-using-ghexport Differential Revision: [D91202899](https://our.internmc.facebook.com/intern/diff/D91202899/)
Pull Request resolved: #16770 This diff enables CUDA tensor creation with basic tensor functionality and factory function support **Key changes:* 1. **`core/SlimTensor.h`** - Extended for CUDA support: - Added `is_cuda()` method to check if tensor is on CUDA device 2. **`factory/Empty.h`** - Supports CUDA: - `empty_strided()` and `empty()` work with CUDA device via `new_storage()` - Device routing is handled by `MaybeOwningStorage` constructor ghstack-source-id: 335102160 @exported-using-ghexport Differential Revision: [D91202897](https://our.internmc.facebook.com/intern/diff/D91202897/)
Pull Request resolved: #16771 This diff enables CUDA tensor copy operations in SlimTensor. **Key changes:** **`core/SlimTensor.h`** - Extended for CUDA support: - Updated `copy_()` to handle cross-device copies: - CPU→CUDA (cudaMemcpyHostToDevice) - CUDA→CPU (cudaMemcpyDeviceToHost) - CUDA→CUDA (cudaMemcpyDeviceToDevice, same device) - Cross-device copies require contiguous tensors - CPU-to-CPU copies continue to support non-contiguous (strided) tensors ghstack-source-id: 335102159 @exported-using-ghexport Differential Revision: [D91202900](https://our.internmc.facebook.com/intern/diff/D91202900/)
Pull Request resolved: #16772 Add non-owning constructor to MaybeOwningStorage that allows wrapping external memory without taking ownership. This is needed for from_blob() which creates tensors from user-provided data pointers. Changes: - Add MaybeOwningStorage(device, data, nbytes) non-owning constructor - Add is_owning() method to query ownership mode - Non-owning storage uses no-op deleter so external memory is not freed ghstack-source-id: 335102163 @exported-using-ghexport Differential Revision: [D91202898](https://our.internmc.facebook.com/intern/diff/D91202898/)
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/16801
Note: Links to docs will display an error until the docs builds have been completed. ❌ 1 New Failure, 1 Cancelled Job, 34 PendingAs of commit 4a28e6e with merge base 37aa87e ( NEW FAILURE - The following job has failed:
CANCELLED JOB - The following job was cancelled. Please retry:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
meta-cla
bot
added
the
CLA Signed
Gasoonjia
approved these changes
Jan 22, 2026
This PR needs a
|
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
This PR was created by the merge bot to help merge the original PR into the main branch.
ghstack PR number: #16772 by @Gasoonjia
^ Please use this as the source of truth for the PR details, comments, and reviews
ghstack PR base: https://github.com/pytorch/executorch/tree/gh/gasoonjia/112/base
ghstack PR head: https://github.com/pytorch/executorch/tree/gh/gasoonjia/112/head
Merge bot PR base: https://github.com/pytorch/executorch/tree/gh/gasoonjia/111/orig
Merge bot PR head: https://github.com/pytorch/executorch/tree/gh/gasoonjia/112/orig
Differential Revision: D91202898
@diff-train-skip-merge