[mlir][flang][openmp] Rework parallel reduction operations #79308
There are no files selected for viewing
This file contains hidden or 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
This file contains hidden or 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
This file contains hidden or 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
This file contains hidden or 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
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,38 @@ | ||
| ! RUN: bbc -emit-hlfir -fopenmp -o - %s 2>&1 | FileCheck %s | ||
| ! RUN: %flang_fc1 -emit-hlfir -fopenmp -o - %s 2>&1 | FileCheck %s | ||
|
|
||
| !CHECK: omp.reduction.declare @[[REDUCTION_DECLARE:[_a-z0-9]+]] : i32 init { | ||
| !CHECK: ^bb0(%{{.*}}: i32): | ||
| !CHECK: %[[I0:[_a-z0-9]+]] = arith.constant 0 : i32 | ||
| !CHECK: omp.yield(%[[I0]] : i32) | ||
| !CHECK: } combiner { | ||
| !CHECK: ^bb0(%[[C0:[_a-z0-9]+]]: i32, %[[C1:[_a-z0-9]+]]: i32): | ||
| !CHECK: %[[CR:[_a-z0-9]+]] = arith.addi %[[C0]], %[[C1]] : i32 | ||
| !CHECK: omp.yield(%[[CR]] : i32) | ||
| !CHECK: } | ||
| !CHECK: func.func @_QQmain() attributes {fir.bindc_name = "mn"} { | ||
| !CHECK: %[[RED_ACCUM_REF:[_a-z0-9]+]] = fir.alloca i32 {bindc_name = "i", uniq_name = "_QFEi"} | ||
| !CHECK: %[[RED_ACCUM_DECL:[_a-z0-9]+]]:2 = hlfir.declare %[[RED_ACCUM_REF]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) | ||
| !CHECK: %[[C0:[_a-z0-9]+]] = arith.constant 0 : i32 | ||
| !CHECK: hlfir.assign %[[C0]] to %[[RED_ACCUM_DECL]]#0 : i32, !fir.ref<i32> | ||
| !CHECK: omp.parallel reduction(@[[REDUCTION_DECLARE]] %[[RED_ACCUM_DECL]]#0 -> %[[PRIVATE_RED:[a-z0-9]+]] : !fir.ref<i32>) { | ||
| !CHECK: %[[PRIVATE_DECL:[_a-z0-9]+]]:2 = hlfir.declare %[[PRIVATE_RED]] {uniq_name = "_QFEi"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>) | ||
| !CHECK: %[[C1:[_a-z0-9]+]] = arith.constant 1 : i32 | ||
| !CHECK: hlfir.assign %[[C1]] to %[[PRIVATE_DECL]]#0 : i32, !fir.ref<i32> | ||
| !CHECK: omp.terminator | ||
| !CHECK: } | ||
| !CHECK: %[[RED_ACCUM_VAL:[_a-z0-9]+]] = fir.load %[[RED_ACCUM_DECL]]#0 : !fir.ref<i32> | ||
| !CHECK: {{.*}} = fir.call @_FortranAioOutputInteger32(%{{.*}}, %[[RED_ACCUM_VAL]]) fastmath<contract> : (!fir.ref<i8>, i32) -> i1 | ||
| !CHECK: return | ||
| !CHECK: } | ||
|
|
||
| program mn | ||
| integer :: i | ||
| i = 0 | ||
|
|
||
| !$omp parallel reduction(+:i) | ||
| i = 1 | ||
| !$omp end parallel | ||
|
|
||
| print *, i | ||
| end program |
This file contains hidden or 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
Oops, something went wrong.
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.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why is
genOpenMPReductionbeing removed here?There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
If genOpenMPReduction is called her it replaces the normal operations (e.g.
arith.addi) with anomp.reductionoperation that is no longer used/needed with the new reduction lowering. At some point when all operations have moved to the new reduction style, those operations should be removed entirely from the OpenMP dialect.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
However we have not changed processing for the
singledirective. So it should continue to use the old flow? Although, the lack of testing suggests it was never used.There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
singleDirective=truedoesn't appear to mean thesingledirective is the one being processed, but rather that a single directive is being processed as opposed to a nested one. The variable name is quite unclear!There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Regardless, of the directives processed by this function, only
parallelandteamscan actually have reduction clauses on them, and I don't think Flang supportsteamsreductions yet?Uh oh!
There was an error while loading. Please reload this page.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, reduction for teams is not supported.