Skip to content

[SYCL] Optimize use of shared_ptr on kernel enqueue fast path - #17569

Merged
uditagarwal97 merged 9 commits into
intel:syclfrom
slawekptak:shared_ptrs_refs_local_vars_2
Mar 26, 2025
Merged

uditagarwal97 merged 9 commits into
intel:syclfrom
slawekptak:shared_ptrs_refs_local_vars_2

Conversation

@slawekptak

Copy link
Copy Markdown
Contributor

Introduce a series of optimizations to how the shared_ptr type variables are passed, to minimize the number of reference counter increments and decrements.

Comment thread sycl/source/accessor.cpp Outdated
Use references to shared_ptr as local variables.
Pass the ContextImplPtr by reference to selected
functions (on the kernel enqueue fast path).
@slawekptak slawekptak changed the title Optimize use of shared_ptr on kernel enqueue fast path [SYCL] Optimize use of shared_ptr on kernel enqueue fast path Mar 24, 2025
Comment thread sycl/source/handler.cpp Outdated
@slawekptak
slawekptak requested a review from a team as a code owner March 25, 2025 09:31
@slawekptak
slawekptak requested a review from fabiomestre March 25, 2025 09:31
@slawekptak
slawekptak force-pushed the shared_ptrs_refs_local_vars_2 branch from 12430be to bdcd30e Compare March 25, 2025 09:56
Comment thread sycl/source/handler.cpp Outdated
directly, instead of creating the device object instance.
@uditagarwal97

Copy link
Copy Markdown
Contributor

Jenkins pre-commit failure is unrelated and an infrastructure issue. Merging

@uditagarwal97
uditagarwal97 merged commit 0f42c75 into intel:sycl Mar 26, 2025

This branch was previously deployed

1 inactive deployment
WindowsCILock — 4612758a Deployed Mar 25, 2025 by slawekptak
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

5 participants