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Outdated suggestions cannot be applied.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.
Overall, SimSYCL has very little shared mutable state:
Fiber scheduling always happens inside a
parallel_for
, so it will never cross thread boundaries. The correspondingcooperative_schedule
and "return fiber" can be made thread-local.For where there is globally visible shared state (system config, USM allocations, buffer access validation), I've introduced a global lock (named
system_lock
, which is backed by a singleton mutex). All shared mutable state is now explicitly wrapped in amutable shared_state<>
, which only grants access to its interior when asystem_lock
is in scope.To guarantee command-group ordering across threads without tracking dependencies explicitly,
submit
(and all similar functions) hold thesystem_lock
while they're running.With this patch, Celerity works fine on my machine ™️ with workarounds removed.