[DON'T BE REVIEWED]ClearBuffer in advance to Support concurrent use with explicit Session streams - #32669
Draft
Xiaofei Han (xiaofeihan1) wants to merge 7 commits into
Draft
[DON'T BE REVIEWED]ClearBuffer in advance to Support concurrent use with explicit Session streams#32669Xiaofei Han (xiaofeihan1) wants to merge 7 commits into
Xiaofei Han (xiaofeihan1) wants to merge 7 commits into
Conversation
Partition command recording, graph capture, and buffer cache state per session while retaining the context-level buffer manager for session-less plugin EP allocation and data transfer. Guard shared pipeline cache access and keep deferred-dispatch resources alive until submission.
Keep command recording state per session while protecting context-level buffer cache containers with an internal mutex. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Preserve explicit streams for single plugin tensor copies so I/O binding uploads follow stream-ordered buffer clears. Restore deferred dispatch encoding before WebGPU uploads and device copies, add regression coverage, and fix formatting. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Allocate plugin kernel scratch tensors on their owning Session streams while preserving immediate cached-buffer clears for ordinary allocations. Require implicit device synchronization for native devices and document allocator ownership and usage. Add regressions for mixed copies, scratch reuse, allocator concurrency, and native device feature validation. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Preserve the complete explicit Session stream implementation with unified recycle-time clearing. Remove the separately tracked core forwarding change and five explicitly excluded tests; document the remaining mixed-output correctness limitation. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
Allow caller-supplied native devices without ImplicitDeviceSynchronization with an initialization warning requiring caller-serialized access. Preserve the internal device invariant and automatic feature request. Replace the external-device rejection test with a focused acceptance/warning regression and document the concurrency boundary. Co-authored-by: Copilot App <223556219+Copilot@users.noreply.github.com>
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
1 participant
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.
Description
Motivation and Context