[Flang][OpenMP] readd enable-delayed-privatization-staging for implic… - #204466
Merged
Jason-Van-Beusekom merged 1 commit intoJun 26, 2026
Merged
Conversation
Jason-Van-Beusekom
marked this pull request as ready for review
June 17, 2026 21:37
llvmorg-github-actions
Bot
added
flang
|
@llvm/pr-subscribers-flang-semantics @llvm/pr-subscribers-flang-openmp Author: Jason Van Beusekom (Jason-Van-Beusekom) Changes…t firstprivate In e57ebfd I removed the flag enableDelayedPrivatizationStaging enabling Delayed Privatization for firstprivate on OpenMP target regions, this resulted in regressions for implicit firstprivate tests: module fp_data
integer(8), parameter :: n1=50
real(8), dimension(n1) :: a_8, b_8, c_8
end module
program fp_repro
use fp_data; real(8) :: temp, tempin=100.
do i=1,n1; a_8(i)=i; b_8(i)=i*2-333.; c_8(i)=0; enddo
!$omp target
!$omp teams shared(a_8,b_8,c_8) private(temp)
!$omp distribute private(temp) firstprivate(tempin)
do i=1,n1
temp = a_8(i)/b_8(i)
c_8(i) = temp + 1.0/temp + tempin
end do
!$omp end distribute
!$omp end teams
!$omp end target
print *, c_8(1) ! expected -231.00302
end programThis PR readds enableDelayedPrivatizationStaging flag for the implicit case and also guards the explicit change via Patch is 24.93 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/204466.diff 15 Files Affected:
diff --git a/flang/include/flang/Support/Flags.h b/flang/include/flang/Support/Flags.h
index 979699131ad8f..bcbb72f8e50d0 100644
--- a/flang/include/flang/Support/Flags.h
+++ b/flang/include/flang/Support/Flags.h
@@ -12,5 +12,6 @@
#include "llvm/Support/CommandLine.h"
extern llvm::cl::opt<bool> enableDelayedPrivatization;
+extern llvm::cl::opt<bool> enableDelayedPrivatizationStaging;
#endif // FORTRAN_SUPPORT_FLAGS_H_
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index facca9867e4bb..0ef85dca3234c 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -2010,6 +2010,11 @@ genTargetClauses(lower::AbstractConverter &converter,
cp.processThreadLimit(stmtCtx, clauseOps);
cp.processTODO<clause::Allocate, clause::InReduction, clause::UsesAllocators>(
loc, llvm::omp::Directive::OMPD_target);
+
+ // `target private(..)` is only supported in delayed privatization mode.
+ if (!enableDelayedPrivatization)
+ cp.processTODO<clause::Firstprivate, clause::Private>(
+ loc, llvm::omp::Directive::OMPD_target);
}
static void genTargetDataClauses(
diff --git a/flang/lib/Semantics/resolve-directives.cpp b/flang/lib/Semantics/resolve-directives.cpp
index 0e2dda8398c18..db0c02c9bd5b4 100644
--- a/flang/lib/Semantics/resolve-directives.cpp
+++ b/flang/lib/Semantics/resolve-directives.cpp
@@ -25,6 +25,7 @@
#include "flang/Semantics/openmp-utils.h"
#include "flang/Semantics/symbol.h"
#include "flang/Semantics/tools.h"
+#include "flang/Support/Flags.h"
#include "llvm/ADT/StringMap.h"
#include "llvm/ADT/StringRef.h"
#include "llvm/Frontend/OpenMP/OMP.h.inc"
@@ -2728,7 +2729,8 @@ void OmpAttributeVisitor::CreateImplicitSymbols(
// 4) not mapped target variable -> firstprivate
// - i.e. implicit, but meets OpenMP specification rules for
// firstprivate "promotion"
- if (IsTargetCaptureImplicitlyFirstprivatizeable(*symbol, prevDSA,
+ if (enableDelayedPrivatizationStaging &&
+ IsTargetCaptureImplicitlyFirstprivatizeable(*symbol, prevDSA,
dataSharingAttributeFlags, dataMappingAttributeFlags,
dirContext.defaultMap)) {
prevDSA.set(Symbol::Flag::OmpImplicit);
diff --git a/flang/lib/Support/Flags.cpp b/flang/lib/Support/Flags.cpp
index 3bc8d4685b7dc..02f64981618dd 100644
--- a/flang/lib/Support/Flags.cpp
+++ b/flang/lib/Support/Flags.cpp
@@ -12,3 +12,9 @@ llvm::cl::opt<bool> enableDelayedPrivatization("enable-delayed-privatization",
llvm::cl::desc(
"Emit private/local variables as clauses/specifiers on MLIR ops."),
llvm::cl::init(true));
+
+llvm::cl::opt<bool> enableDelayedPrivatizationStaging(
+ "enable-delayed-privatization-staging",
+ llvm::cl::desc("For partially supported constructs, emit private/local "
+ "variables as clauses/specifiers on MLIR ops."),
+ llvm::cl::init(false));
diff --git a/flang/test/Integration/OpenMP/map-types-and-sizes.f90 b/flang/test/Integration/OpenMP/map-types-and-sizes.f90
index 1a266dfff68d0..93f28a7525e19 100644
--- a/flang/test/Integration/OpenMP/map-types-and-sizes.f90
+++ b/flang/test/Integration/OpenMP/map-types-and-sizes.f90
@@ -6,7 +6,8 @@
! added to this directory and sub-directories.
!===----------------------------------------------------------------------===!
-!RUN: %flang_fc1 -emit-llvm -fopenmp -fopenmp-version=51 -fopenmp-targets=amdgcn-amd-amdhsa %s -o - | FileCheck %s
+!RUN: %flang_fc1 -emit-llvm -fopenmp -mmlir --enable-delayed-privatization-staging=false -fopenmp-version=51 -fopenmp-targets=amdgcn-amd-amdhsa %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NO-FPRIV
+!RUN: %flang_fc1 -emit-llvm -fopenmp -mmlir --enable-delayed-privatization-staging=true -fopenmp-version=51 -fopenmp-targets=amdgcn-amd-amdhsa %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-FPRIV
!===============================================================================
@@ -14,7 +15,8 @@
!===============================================================================
!CHECK: @.offload_sizes = private unnamed_addr constant [2 x i64] [i64 4, i64 0]
-!CHECK: @.offload_maptypes = private unnamed_addr constant [2 x i64] [i64 289, i64 288]
+!CHECK-FPRIV: @.offload_maptypes = private unnamed_addr constant [2 x i64] [i64 289, i64 288]
+!CHECK-NO-FPRIV: @.offload_maptypes = private unnamed_addr constant [2 x i64] [i64 800, i64 288]
subroutine mapType_scalar
integer :: a
!$omp target
@@ -382,7 +384,8 @@ subroutine mapType_nested_derived_type_member_idx()
end subroutine
!CHECK: @.offload_sizes{{.*}} = private unnamed_addr constant [3 x i64] [i64 8, i64 4, i64 0]
-!CHECK: @.offload_maptypes{{.*}} = private unnamed_addr constant [3 x i64] [i64 544, i64 289, i64 288]
+!CHECK-FPRIV: @.offload_maptypes{{.*}} = private unnamed_addr constant [3 x i64] [i64 544, i64 289, i64 288]
+!CHECK-NO-FPRIV: @.offload_maptypes{{.*}} = private unnamed_addr constant [3 x i64] [i64 544, i64 800, i64 288]
subroutine mapType_c_ptr
use iso_c_binding, only : c_ptr, c_loc
type(c_ptr) :: a
@@ -393,7 +396,8 @@ subroutine mapType_c_ptr
end subroutine mapType_c_ptr
!CHECK: @.offload_sizes{{.*}} = private unnamed_addr constant [2 x i64] [i64 1, i64 0]
-!CHECK: @.offload_maptypes{{.*}} = private unnamed_addr constant [2 x i64] [i64 289, i64 288]
+!CHECK-FPRIV: @.offload_maptypes{{.*}} = private unnamed_addr constant [2 x i64] [i64 289, i64 288]
+!CHECK-NO-FPRIV: @.offload_maptypes{{.*}} = private unnamed_addr constant [2 x i64] [i64 800, i64 288]
subroutine mapType_char
character :: a
!$omp target
diff --git a/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map-2.f90 b/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map-2.f90
index cc478e8ee6b84..3f1ed7842f560 100644
--- a/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map-2.f90
+++ b/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map-2.f90
@@ -2,8 +2,8 @@
! module boundaries.
! RUN: split-file %s %t
-! RUN: bbc -emit-hlfir -fopenmp -fopenmp-version=50 %t/imp_scalar_map_module.f90 -o - | FileCheck %s --check-prefix=CHECK-MOD
-! RUN: bbc -emit-hlfir -fopenmp -fopenmp-version=50 %t/imp_scalar_map_target.f90 -o - | FileCheck %s --check-prefix=CHECK-PROG
+! RUN: bbc -emit-hlfir -fopenmp --enable-delayed-privatization-staging -fopenmp-version=50 %t/imp_scalar_map_module.f90 -o - | FileCheck %s --check-prefix=CHECK-MOD
+! RUN: bbc -emit-hlfir -fopenmp --enable-delayed-privatization-staging -fopenmp-version=50 %t/imp_scalar_map_target.f90 -o - | FileCheck %s --check-prefix=CHECK-PROG
!--- imp_scalar_map_module.f90
module test_data
diff --git a/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map.f90 b/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map.f90
index 7246136f0a2f8..52a1f31cc3ce3 100644
--- a/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map.f90
+++ b/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-implicit-scalar-map.f90
@@ -1,7 +1,7 @@
! Tests delayed privatization works for implicit capture of scalars similarly to
! the way it works for explicitly firstprivitized scalars.
-! RUN: %flang_fc1 -emit-mlir -fopenmp \
+! RUN: %flang_fc1 -emit-mlir -fopenmp -mmlir --enable-delayed-privatization-staging \
! RUN: -o - %s 2>&1 | FileCheck %s
! CHECK-LABEL: omp.private {type = firstprivate} @_QFExdgfx_firstprivate_i32 : i32 copy {
diff --git a/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-multiple-variables.f90 b/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-multiple-variables.f90
index b6fc246799d32..3cc50129dfab6 100644
--- a/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-multiple-variables.f90
+++ b/flang/test/Lower/OpenMP/DelayedPrivatization/target-private-multiple-variables.f90
@@ -1,7 +1,9 @@
! Tests delayed privatization for `targets ... private(..)` for allocatables.
-! RUN: %flang_fc1 -emit-hlfir -fopenmp -o - %s 2>&1 | FileCheck %s
-! RUN: bbc -emit-hlfir -fopenmp -o - %s 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --enable-delayed-privatization-staging \
+! RUN: -o - %s 2>&1 | FileCheck %s
+! RUN: bbc -emit-hlfir -fopenmp --enable-delayed-privatization-staging -o - %s 2>&1 \
+! RUN: | FileCheck %s
subroutine target_allocatable(lb, ub, l)
implicit none
diff --git a/flang/test/Lower/OpenMP/DelayedPrivatization/target-teams-private-implicit-scalar-map.f90 b/flang/test/Lower/OpenMP/DelayedPrivatization/target-teams-private-implicit-scalar-map.f90
index f288c4c4b1b65..7661ced82bf69 100644
--- a/flang/test/Lower/OpenMP/DelayedPrivatization/target-teams-private-implicit-scalar-map.f90
+++ b/flang/test/Lower/OpenMP/DelayedPrivatization/target-teams-private-implicit-scalar-map.f90
@@ -1,7 +1,7 @@
! Tests delayed privatization works for implicit capture of scalars similarly to
! the way it works for explicitly firstprivitized scalars.
-! RUN: %flang_fc1 -emit-mlir -fopenmp \
+! RUN: %flang_fc1 -emit-mlir -fopenmp -mmlir --enable-delayed-privatization-staging \
! RUN: -o - %s 2>&1 | FileCheck %s
!CHECK: omp.private {type = private} @[[SYM_K:.*]] : i32
diff --git a/flang/test/Lower/OpenMP/defaultmap.f90 b/flang/test/Lower/OpenMP/defaultmap.f90
index 42bc47a719df7..093f8fb79c6df 100644
--- a/flang/test/Lower/OpenMP/defaultmap.f90
+++ b/flang/test/Lower/OpenMP/defaultmap.f90
@@ -1,4 +1,5 @@
-!RUN: %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=52 %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-FPRIV
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=52 -mmlir --enable-delayed-privatization-staging=false %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NO-FPRIV
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=52 -mmlir --enable-delayed-privatization-staging=true %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-FPRIV
subroutine defaultmap_allocatable_present()
implicit none
@@ -49,9 +50,11 @@ subroutine defaultmap_pointer_to()
integer, dimension(:), pointer :: arr_ptr(:)
integer :: scalar_int
+! CHECK-NO-FPRIV: %[[MAP_1:.*]] = omp.map.info var_ptr({{.*}} : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.box<!fir.ptr<!fir.array<?xi32>>>) map_clauses(implicit, to) capture(ByRef) var_ptr_ptr({{.*}}) bounds({{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xi32>>> {name = ""}
! CHECK-FPRIV: %[[MAP_1:.*]] = omp.map.info var_ptr({{.*}} : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.box<!fir.ptr<!fir.array<?xi32>>>) map_clauses(implicit, to) capture(ByRef) var_ptr_ptr({{.*}}) bounds({{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xi32>>> {name = ""}
! CHECK: %[[MAP_2:.*]] = omp.map.info var_ptr({{.*}} : !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>>, !fir.box<!fir.ptr<!fir.array<?xi32>>>) map_clauses(always, implicit, to) capture(ByRef) members({{.*}}) -> !fir.ref<!fir.box<!fir.ptr<!fir.array<?xi32>>>> {name = "arr_ptr"}
! CHECK-FPRIV: %[[MAP_3:.*]] = omp.map.info var_ptr({{.*}} : !fir.ref<i32>, i32) map_clauses(to) capture(ByCopy) -> !fir.ref<i32>
+! CHECK-NO-FPRIV: %[[MAP_3:.*]] = omp.map.info var_ptr({{.*}} : !fir.ref<i32>, i32) map_clauses(implicit) capture(ByCopy) -> !fir.ref<i32> {name = "scalar_int"}
!$omp target defaultmap(to: pointer)
arr_ptr(1) = scalar_int + 20
!$omp end target
diff --git a/flang/test/Lower/OpenMP/optional-argument-map-2.f90 b/flang/test/Lower/OpenMP/optional-argument-map-2.f90
index 78122547bf828..b6366e2d83f2b 100644
--- a/flang/test/Lower/OpenMP/optional-argument-map-2.f90
+++ b/flang/test/Lower/OpenMP/optional-argument-map-2.f90
@@ -1,4 +1,5 @@
-!RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-FPRIV
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --enable-delayed-privatization-staging=false %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NO-FPRIV
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --enable-delayed-privatization-staging=true %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-FPRIV
module mod
implicit none
@@ -82,3 +83,36 @@ end module mod
! CHECK-FPRIV: hlfir.assign %[[VAL_28]]#0 to %[[VAL_26]]#0 : !fir.boxchar<1>, !fir.ref<!fir.char<1,4>>
! CHECK-FPRIV: return
! CHECK-FPRIV: }
+! CHECK-NO-FPRIV: %[[VAL_8:.*]] = fir.is_present %[[VAL_3]]#1 : (!fir.ref<!fir.char<1,?>>) -> i1
+! CHECK-NO-FPRIV: %[[VAL_9:.*]] = arith.constant 0 : index
+! CHECK-NO-FPRIV: %[[VAL_10:.*]] = arith.constant 1 : index
+! CHECK-NO-FPRIV: %[[VAL_11:.*]]:2 = fir.if %[[VAL_8]] -> (index, index) {
+! CHECK-NO-FPRIV: %[[VAL_12:.*]]:2 = fir.unboxchar %[[VAL_3]]#0 : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)
+! CHECK-NO-FPRIV: fir.result %[[VAL_12]]#1, %[[VAL_10]] : index, index
+! CHECK-NO-FPRIV: } else {
+! CHECK-NO-FPRIV: fir.result %[[VAL_9]], %[[VAL_9]] : index, index
+! CHECK-NO-FPRIV: }
+! CHECK-NO-FPRIV: %[[VAL_13:.*]] = arith.subi %[[VAL_14:.*]]#0, %[[VAL_10]] : index
+! CHECK-NO-FPRIV: %[[VAL_15:.*]] = omp.map.bounds lower_bound(%[[VAL_9]] : index) upper_bound(%[[VAL_13]] : index) extent(%[[VAL_14]]#0 : index) stride(%[[VAL_14]]#1 : index) start_idx(%[[VAL_9]] : index) {stride_in_bytes = true}
+! CHECK-NO-FPRIV: %[[VAL_16:.*]] = omp.map.info var_ptr(%[[VAL_3]]#1 : !fir.ref<!fir.char<1,?>>, !fir.char<1,?>) map_clauses(implicit) capture(ByCopy) bounds(%[[VAL_15]]) -> !fir.ref<!fir.char<1,?>> {name = "a"}
+! CHECK-NO-FPRIV: fir.store %[[ARG0]] to %[[VAL_0]] : !fir.ref<!fir.boxchar<1>>
+! CHECK-NO-FPRIV: %[[VAL_17:.*]] = arith.constant 0 : index
+! CHECK-NO-FPRIV: %[[VAL_18:.*]] = arith.constant 1 : index
+! CHECK-NO-FPRIV: %[[VAL_19:.*]]:2 = fir.unboxchar %[[ARG0]] : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)
+! CHECK-NO-FPRIV: %[[VAL_20:.*]] = arith.subi %[[VAL_19]]#1, %[[VAL_18]] : index
+! CHECK-NO-FPRIV: %[[VAL_21:.*]] = omp.map.bounds lower_bound(%[[VAL_17]] : index) upper_bound(%[[VAL_20]] : index) extent(%[[VAL_19]]#1 : index) stride(%[[VAL_18]] : index) start_idx(%[[VAL_17]] : index) {stride_in_bytes = true}
+! CHECK-NO-FPRIV: %[[VAL_22:.*]] = fir.box_offset %[[VAL_0]] base_addr : (!fir.ref<!fir.boxchar<1>>) -> !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>
+! CHECK-NO-FPRIV: %[[VAL_23:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.boxchar<1>>, !fir.boxchar<1>) map_clauses(implicit, to) capture(ByRef) var_ptr_ptr(%[[VAL_22]] : !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>, !fir.char<1,?>) bounds(%14) -> !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>> {name = ""}
+! CHECK-NO-FPRIV: %[[VAL_24:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.boxchar<1>>, !fir.boxchar<1>) map_clauses(always, implicit, to) capture(ByRef) members(%[[VAL_23]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>) -> !fir.ref<!fir.boxchar<1>> {name = ""}
+! CHECK-NO-FPRIV: %[[VAL_26:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.boxchar<1>>, !fir.boxchar<1>) map_clauses(attach, ref_ptr, ref_ptee) capture(ByRef) var_ptr_ptr(%[[VAL_22]] : !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>, !fir.char<1,?>) bounds(%[[VAL_21]]) -> !fir.ref<!fir.boxchar<1>>
+! CHECK-NO-FPRIV: omp.target map_entries(%[[VAL_7]] -> %[[VAL_27:.*]], %[[VAL_16]] -> %[[VAL_28:.*]], %[[VAL_24]] -> %[[VAL_29:.*]], %[[VAL_26]] -> %[[VAL_30:.*]], %[[VAL_23]] -> %[[VAL_30:.*]] : !fir.ref<!fir.char<1,4>>, !fir.ref<!fir.char<1,?>>, !fir.ref<!fir.boxchar<1>>, !fir.ref<!fir.boxchar<1>>, !fir.llvm_ptr<!fir.ref<!fir.char<1,?>>>) {
+! CHECK-NO-FPRIV: %[[VAL_31:.*]] = fir.load %[[VAL_29]] : !fir.ref<!fir.boxchar<1>>
+! CHECK-NO-FPRIV: %[[VAL_32:.*]]:2 = fir.unboxchar %[[VAL_31]] : (!fir.boxchar<1>) -> (!fir.ref<!fir.char<1,?>>, index)
+! CHECK-NO-FPRIV: %[[VAL_33:.*]] = arith.constant 4 : index
+! CHECK-NO-FPRIV: %[[VAL_34:.*]]:2 = hlfir.declare %[[VAL_27]] typeparams %[[VAL_33]] {uniq_name = "_QMmodFroutine_boxcharEb"} : (!fir.ref<!fir.char<1,4>>, index) -> (!fir.ref<!fir.char<1,4>>, !fir.ref<!fir.char<1,4>>)
+! CHECK-NO-FPRIV: %[[VAL_35:.*]]:2 = hlfir.declare %[[VAL_28]] typeparams %[[VAL_32]]#1 {fortran_attrs = #fir.var_attrs<intent_in, optional>, uniq_name = "_QMmodFroutine_boxcharEa"} : (!fir.ref<!fir.char<1,?>>, index) -> (!fir.boxchar<1>, !fir.ref<!fir.char<1,?>>)
+! CHECK-NO-FPRIV: hlfir.assign %[[VAL_35]]#0 to %[[VAL_34]]#0 : !fir.boxchar<1>, !fir.ref<!fir.char<1,4>>
+! CHECK-NO-FPRIV: omp.terminator
+! CHECK-NO-FPRIV: }
+! CHECK-NO-FPRIV: return
+! CHECK-NO-FPRIV:}
diff --git a/flang/test/Lower/OpenMP/target-map-complex.f90 b/flang/test/Lower/OpenMP/target-map-complex.f90
index a0a3abf6ef3d2..2325aec79e65b 100644
--- a/flang/test/Lower/OpenMP/target-map-complex.f90
+++ b/flang/test/Lower/OpenMP/target-map-complex.f90
@@ -1,17 +1,21 @@
-!RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --enable-delayed-privatization-staging=false %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-NO-FPRIV
+!RUN: %flang_fc1 -emit-hlfir -fopenmp -mmlir --enable-delayed-privatization-staging=true %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-FPRIV
! Check that the complex*4 is passed by value. but complex*8 is passed by
! reference
-!CHECK: omp.private {type = firstprivate} @[[PRIV_64:.*]] : complex<f64> copy {
-!CHECK: omp.private {type = firstprivate} @[[PRIV_32:.*]] : complex<f32> copy {
+!CHECK-FPRIV: omp.private {type = firstprivate} @[[PRIV_64:.*]] : complex<f64> copy {
+!CHECK-FPRIV: omp.private {type = firstprivate} @[[PRIV_32:.*]] : complex<f32> copy {
!CHECK-LABEL: func.func @_QMmPbar()
!CHECK: %[[V1:[0-9]+]]:2 = hlfir.declare {{.*}} (!fir.ref<complex<f32>>) -> (!fir.ref<complex<f32>>, !fir.ref<complex<f32>>)
!CHECK: %[[V0:[0-9]+]]:2 = hlfir.declare {{.*}} (!fir.ref<complex<f64>>) -> (!fir.ref<complex<f64>>, !fir.ref<complex<f64>>)
-!CHECK: %[[V2:[0-9]+]] = omp.map.info var_ptr(%[[V1]]#0 : !fir.ref<complex<f32>>, complex<f32>) {{.*}} capture(ByCopy)
-!CHECK: %[[V3:[0-9]+]] = omp.map.info var_ptr(%[[V0]]#0 : !fir.ref<complex<f64>>, complex<f64>) {{.*}} capture(ByRef)
-!CHECK: omp.target map_entries(%[[V2]] -> {{.*}}, %[[V3]] -> {{.*}} : !fir.ref<complex<f32>>, !fir.ref<complex<f64>>) private(@[[PRIV_32]] %[[V1]]#0 -> %{{.*}} [map_idx=0], @[[PRIV_64]] %[[V0]]#0 -> %{{.*}} [map_idx=1] : !fir.ref<complex<f32>>, !fir.ref<complex<f64>>) {
+!CHECK-FPRIV: %[[V2:[0-9]+]] = omp.map.info var_ptr(%[[V1]]#0 : !fir.ref<complex<f32>>, complex<f32>) {{.*}} capture(ByCopy)
+!CHECK-FPRIV: %[[V3:[0-9]+]] = omp.map.info var_ptr(%[[V0]]#0 : !fir.ref<complex<f64>>, complex<f64>) {{.*}} capture(ByRef)
+!CHECK-FPRIV: omp.target map_entries(%[[V2]] -> {{.*}}, %[[V3]] -> {{.*}} : !fir.ref<complex<f32>>, !fir.ref<complex<f64>>) private(@[[PRIV_32]] %[[V1]]#0 -> %{{.*}} [map_idx=0], @[[PRIV_64]] %[[V0]]#0 -> %{{.*}} [map_idx=1] : !fir.ref<complex<f32>>, !fir.ref<complex<f64>>) {
+!CHECK-NO-FPRIV: %[[V2:[0-9]+]] = omp.map.info var_ptr(%[[V1]]#1 : !fir.ref<complex<f32>>, complex<f32>) {{.*}} capture(ByCopy)
+!CHECK-NO-FPRIV: %[[V3:[0-9]+]] = omp.map.info var_ptr(%[[V0]]#1 : !fir.ref<complex<f64>>, complex<f64>) {{.*}} capture(ByRef)
+!CHECK-NO-PRIV: omp.target map_entries(%[[V2]] -> {{.*}}, %[[V3]] -> {{.*}} : !fir.ref<complex<f32>>, !fir.ref<complex<f64>>)
module m
implicit none
diff --git a/flang/test/Lower/OpenMP/target-scope.f90 b/flang/test/Lower/OpenMP/target-scope.f90
index fcc476595a4cd..d0900bd20e81f 100644
--- a/flang/test/Lower/OpenMP/target-scope.f90
+++ b/flang/test/Lower/OpenMP/target-scope.f90
@@ -28,7 +28,7 @@ subroutine target_scope_basic()
!$omp target
! CHECK: omp.map.info var_ptr(%{{.*}} : !fir.ref<i32>, i32)
- ! CHECK: omp.target map_entries(%{{.*}} -> %{{.*}} : !fir.ref<i32>) private({{.*}} -> %[[XARG:.*]] [map_idx=0] : !fir.ref<i32>) {
+ ! CHECK: omp.target map_entries(%{{.*}} -> %[[XARG:.*]] : !fir.ref<i32>) {
! CHECK: hlfir.declare %[[XARG]] {uniq_name = "_QFtarget_scope_basicEx"}
! CHECK: omp.scope {
!$omp scope
@@ -45,7 +45,7 @@ subroutine target_scope_nowait()
x = 10
!$omp target
- ! CHECK: omp.target map_entries(%{{.*}} -> %{{.*}} : !fir.ref<i32>) private({{.*}} -> %[[XARG:.*]] [map_idx=0] : !fir.ref<i32>) {
+ ! CHECK: omp.target map_entries(%{{.*}} -> %[[XARG:.*]] : !fir.ref<i32>) {
! CHECK: hlfir.declare %[[XARG]] {uniq_...
[truncated]
|
Contributor
Author
|
Issue #203890 appears to be related to this aswell |
🐧 Linux x64 Test Results
Failed Tests(click on a test name to see its output) FlangFlang.Semantics/cuf05.cufIf these failures are unrelated to your changes (for example tests are broken or flaky at HEAD), please open an issue at https://github.com/llvm/llvm-project/issues and add the |
Jason-Van-Beusekom
force-pushed
the
implicit-firstprivate-revert
branch
from
June 24, 2026 16:17
d5ff8bb to
5e2178c
Compare
Contributor
Author
|
Previous tests were failing due to unrelated issue, rebased with upstream main and tests are now passing |
bhandarkar-pranav
approved these changes
Jun 25, 2026
bhandarkar-pranav
left a comment
bhandarkar-pranav
left a comment
Contributor
There was a problem hiding this comment.
Thanks @Jason-Van-Beusekom for this PR. LGTM aside from a single question for context.
| loc, llvm::omp::Directive::OMPD_target); | ||
|
|
||
| // `target private(..)` is only supported in delayed privatization mode. | ||
| if (!enableDelayedPrivatization) |
Contributor
There was a problem hiding this comment.
Just curious, why you have chosen to put this back under this flag?
Contributor
Author
There was a problem hiding this comment.
That is the opt out flag, we have:
extern llvm::cl::opt<bool> enableDelayedPrivatization;
extern llvm::cl::opt<bool> enableDelayedPrivatizationStaging;explicit firstprivate was originally under enableDelayedPrivatizationStaging which is default false, while enableDelayedPrivatization is default true
I know it is a bit out of scope, I can move it to a different PR of desired, but since it is a small change I wanted to avoid another pr for just that.
Without this there would be no way to disable delayed privatization for explicit fristprivate.
Jason-Van-Beusekom
force-pushed
the
implicit-firstprivate-revert
branch
from
June 25, 2026 20:53
5e2178c to
16dab28
Compare
Contributor
Author
|
Rebased to address merge conflicts. |
🪟 Windows x64 Test Results
Failed Tests(click on a test name to see its output) FlangFlang.Semantics/cuf05.cufIf these failures are unrelated to your changes (for example tests are broken or flaky at HEAD), please open an issue at https://github.com/llvm/llvm-project/issues and add the |
Contributor
Author
|
test failures are unrelated to this PR |
Contributor
|
@Jason-Van-Beusekom |
Contributor
Author
Yes, however this fix does not address the core problem, it just guards it behind the delayed privatization staging flag. |
Jason-Van-Beusekom
enabled auto-merge (squash)
June 26, 2026 20:02
Jason-Van-Beusekom
disabled auto-merge
June 26, 2026 20:02
LouisLu060211
pushed a commit
to LouisLu060211/llvm-project
that referenced
this pull request
Jun 30, 2026
llvm#204466) …t firstprivate In e57ebfd the flag enableDelayedPrivatizationStaging was removed, enabling Delayed Privatization for firstprivate on OpenMP target regions, this resulted in regressions for implicit firstprivate tests: This commit readds enableDelayedPrivatizationStaging flag for the implicit case and also guards the explicit change via `enableDelayedPrivatization` instead of enableDelayedPrivatizationStaging.
maarcosrmz
pushed a commit
to maarcosrmz/llvm-project
that referenced
this pull request
Jul 1, 2026
llvm#204466) …t firstprivate In e57ebfd the flag enableDelayedPrivatizationStaging was removed, enabling Delayed Privatization for firstprivate on OpenMP target regions, this resulted in regressions for implicit firstprivate tests: This commit readds enableDelayedPrivatizationStaging flag for the implicit case and also guards the explicit change via `enableDelayedPrivatization` instead of enableDelayedPrivatizationStaging.
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
…t firstprivate
In e57ebfd I removed the flag enableDelayedPrivatizationStaging enabling Delayed Privatization for firstprivate on OpenMP target regions, this resulted in regressions for implicit firstprivate tests:
This PR readds enableDelayedPrivatizationStaging flag for the implicit case and also guards the explicit change via
enableDelayedPrivatizationinstead of enableDelayedPrivatizationStaging.