Skip to content

Conversation

@hdelan
Copy link
Contributor

@hdelan hdelan commented Jul 23, 2024

Wrong command type was being used for enqueue native command, causing a race condition as the host task codepath was being taken, where another path should have been taken. This fixes #14715

@sarnex

Copy link
Contributor

@sarnex sarnex left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm in terms of i think it should fix the nightly failure. we should have a seperate tracker to fix the root cause and reenable this case

@hdelan
Copy link
Contributor Author

hdelan commented Jul 23, 2024

lgtm in terms of i think it should fix the nightly failure. we should have a seperate tracker to fix the root cause and reenable this case

Yep I'm investigating as we speak

@sarnex
Copy link
Contributor

sarnex commented Jul 23, 2024

thx

Wrong command type was being used, causing code to follow host task
codepath. This fixes that.
@hdelan hdelan force-pushed the early-exit-test branch from fa35acb to 3738af7 Compare July 23, 2024 15:19
@hdelan
Copy link
Contributor Author

hdelan commented Jul 23, 2024

@sarnex bug found. See update

@sarnex
Copy link
Contributor

sarnex commented Jul 23, 2024

thanks! im not qualified to review this fix but i really appreciate the quick investigation and root cause fix!

@hdelan hdelan changed the title [SYCL] Add early exit for single device context [SYCL] Fix race condition in EnqueueNativeCommand Jul 23, 2024
@hdelan hdelan temporarily deployed to WindowsCILock July 23, 2024 15:29 — with GitHub Actions Inactive
@hdelan hdelan temporarily deployed to WindowsCILock July 23, 2024 16:21 — with GitHub Actions Inactive
Copy the CodeplayHostTask command in how to get the context.
@hdelan hdelan temporarily deployed to WindowsCILock July 24, 2024 09:26 — with GitHub Actions Inactive
@hdelan hdelan temporarily deployed to WindowsCILock July 24, 2024 10:01 — with GitHub Actions Inactive
@cperkinsintel cperkinsintel requested review from sergey-semenov and removed request for sergey-semenov July 25, 2024 16:16
@hdelan hdelan requested review from a team as code owners July 26, 2024 10:20
@hdelan hdelan requested review from a team as code owners July 26, 2024 10:20
@hdelan hdelan requested a review from npmiller July 26, 2024 10:20
Not to be merged. Get the SYCL PI TRACE.
@hdelan hdelan force-pushed the early-exit-test branch from 58f48a5 to 3ad165f Compare July 26, 2024 11:08
@hdelan hdelan removed request for a team and npmiller July 26, 2024 11:08
@hdelan hdelan temporarily deployed to WindowsCILock July 26, 2024 11:12 — with GitHub Actions Inactive
@hdelan hdelan temporarily deployed to WindowsCILock July 26, 2024 11:54 — with GitHub Actions Inactive
@hdelan hdelan temporarily deployed to WindowsCILock July 29, 2024 10:01 — with GitHub Actions Inactive
@hdelan hdelan temporarily deployed to WindowsCILock July 29, 2024 10:30 — with GitHub Actions Inactive
@sarnex sarnex merged commit 619185f into intel:sycl Jul 29, 2024
AlexeySachkov pushed a commit to AlexeySachkov/llvm that referenced this pull request Nov 26, 2024
Wrong command type was being used for enqueue native command, causing a
race condition as the host task codepath was being taken, where another
path should have been taken. This fixes
intel#14715

@sarnex
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.

SYCL :: EnqueueNativeCommand/custom-command-multiple-dev-cuda.cpp failing in nightly on CUDA

4 participants