[MLIR][OpenMP] Update translation of DISTRIBUTE PARALLEL DO to LLVM IR #137
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
skatrak
requested review from
ergawy,
jsjodin,
abidh,
mjklemm,
raghavendhra,
agozillon,
DominikAdamski,
TIFitis,
kparzysz and
bhandarkar-pranav
This was referenced Aug 7, 2024
|
Here's a simple application and the MLIR and LLVM IR produced for the host by the current version of ATD (pre.*) and by this PR stack (post.*) for reference: test.zip |
This patch removes the `LoopWrapperInterface` from `omp.parallel` and updates the semantics of the interface to make loop wrapper restrictions mandatory to operations that have it, rather than a role they might optionally take. MLIR operation verifiers are updated to expect the "hoisted omp.parallel" representation for `distribute parallel do`, to be later implemented in place of a loop wrapper `omp.parallel`.
This patch updates the Flang lowering process for `distribute parallel do` to follow the "hoisted `omp.parallel`" representation. Now temporary allocations produced while lowering the loop body reside inside of that operation instead of the loop wrappers' parent region. Special handling of `omp.parallel` with regards to alloca creation is removed, as it's no longer necessary to make this distinction. Impacted Lit tests are updated according to the new representation.
This patch makes changes to the `DoConcurrentConversion` pass to follow the "hoisted omp.parallel" representation when converting `do concurrent` constructs into `target teams distribute parallel do`.
This patch updates the LLVM IR translation pass to expect the new "hoisted omp.parallel" representation for `distribute parallel do`.
skatrak
force-pushed
the
skatrak/dpd-03-doconcurrent
branch
from
97b9697
to
9a95922
Compare
skatrak
force-pushed
the
skatrak/dpd-04-llvm
branch
from
354d4bc
to
a67931a
Compare
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.
This patch updates the LLVM IR translation pass to expect the new "hoisted omp.parallel" representation for
distribute parallel do.Complete PR stack: