[L0] move platform cache into the adapter structure #1252
Merged
Conversation
This file contains 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
Codecov ReportAll modified and coverable lines are covered by tests ✅
❗ Your organization needs to install the Codecov GitHub app to enable full functionality. Additional details and impacted files@@ Coverage Diff @@
## main #1252 +/- ##
=======================================
Coverage 15.80% 15.80%
=======================================
Files 223 223
Lines 31483 31483
Branches 3558 3558
=======================================
Hits 4975 4975
Misses 26457 26457
Partials 51 51 ☔ View full report in Codecov by Sentry. |
pbalcer
force-pushed
the
l0-platform-hang
branch
from
6f4c425
to
d35ea94
Compare
nrspruit
approved these changes
Jan 16, 2024
steffenlarsen
pushed a commit
to intel/llvm
that referenced
this pull request
Jan 17, 2024
The ur.hpp and ur.cpp files were moved to the unified runtime repo. This is currently blocking #12403, oneapi-src/unified-runtime#1252.
The platform cache is a global variable used exclusively by the L0 adapter, and it's protected by a loosely-associated spin lock. However, its destruction is associated with the lifetime of the adapter structure and is deleted the first time adapter refcount reaches 0. This was causing issues whenever the adapter was initialized and destroyed multiple time inside of a single process, which, for example, happens during tests. This patch fixes the above problem by moving the platform cache from the global state into the adapter structure. This allows for a simpler implementation that no longer requires an explicit lock and instead uses lazy loading (std::call_once). With this patch, all platform tests are now passing for L0. Closes oneapi-src#824
pbalcer
force-pushed
the
l0-platform-hang
branch
from
d35ea94
to
c56b44e
Compare
nrspruit
pushed a commit
to nrspruit/llvm
that referenced
this pull request
Feb 6, 2024
This solves the same issues as intel#12432, but for urDeviceCreateWithNativeHandle. Includes required oneapi-src/unified-runtime#1252
aarongreig
pushed a commit
to aarongreig/unified-runtime
that referenced
this pull request
Feb 9, 2024
[L0] move platform cache into the adapter structure
4 tasks
aarongreig
pushed a commit
to aarongreig/unified-runtime
that referenced
this pull request
Feb 9, 2024
[L0] move platform cache into the adapter structure
againull
pushed a commit
to againull/unified-runtime
that referenced
this pull request
Feb 9, 2024
[L0] move platform cache into the adapter structure
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.
The platform cache is a global variable used exclusively by the L0 adapter, and it's protected by a loosely-associated spin lock. However, its destruction is associated with the lifetime of the adapter structure and is deleted the first time adapter refcount reaches 0. This was causing issues whenever the adapter was initialized and destroyed multiple time inside of a single process, which, for example, happens during tests.
This patch fixes the above problem by moving the platform cache from the global state into the adapter structure. This allows for a simpler implementation that no longer requires an explicit lock and instead uses lazy loading (std::call_once).
With this patch, all platform tests are now passing for L0. Closes #824