[flang-commits] [flang] [mlir] [flang][openacc][openmp] Support implicit casting on the atomic interface (PR #114390)

via flang-commits flang-commits at lists.llvm.org
Mon Nov 4 11:30:20 PST 2024


https://github.com/khaki3 updated https://github.com/llvm/llvm-project/pull/114390

>From b62b2d2b0c6e27975cc24d40b71b870b320df638 Mon Sep 17 00:00:00 2001
From: Kazuaki Matsumura <kmatsumura at nvidia.com>
Date: Thu, 31 Oct 2024 03:54:22 -0700
Subject: [PATCH 1/5] [flang][openacc][openmp] Support implicit casting on the
 atomic interface

---
 flang/lib/Lower/DirectivesCommon.h            |  55 ++++------
 .../Fir/convert-to-llvm-openmp-and-fir.fir    |   4 +-
 .../test/Lower/OpenACC/acc-atomic-capture.f90 | 103 ++++++++++++++++--
 flang/test/Lower/OpenACC/acc-atomic-read.f90  |  19 ++--
 .../Lower/OpenACC/acc-atomic-update-array.f90 |   4 +-
 flang/test/Lower/OpenMP/atomic-capture.f90    |   6 +-
 flang/test/Lower/OpenMP/atomic-read.f90       |  14 +--
 .../mlir/Dialect/OpenACC/OpenACCOps.td        |   5 +-
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td |   5 +-
 9 files changed, 145 insertions(+), 70 deletions(-)

diff --git a/flang/lib/Lower/DirectivesCommon.h b/flang/lib/Lower/DirectivesCommon.h
index 421a44b128c017..88514b16743278 100644
--- a/flang/lib/Lower/DirectivesCommon.h
+++ b/flang/lib/Lower/DirectivesCommon.h
@@ -179,7 +179,11 @@ static inline void genOmpAccAtomicWriteStatement(
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
 
   mlir::Type varType = fir::unwrapRefType(lhsAddr.getType());
+  // Create a conversion outside the capture block.
+  auto insertionPoint = firOpBuilder.saveInsertionPoint();
+  firOpBuilder.setInsertionPointAfter(rhsExpr.getDefiningOp());
   rhsExpr = firOpBuilder.createConvert(loc, varType, rhsExpr);
+  firOpBuilder.restoreInsertionPoint(insertionPoint);
 
   processOmpAtomicTODO<AtomicListT>(varType, loc);
 
@@ -410,10 +414,6 @@ void genOmpAccAtomicRead(Fortran::lower::AbstractConverter &converter,
       fir::getBase(converter.genExprAddr(fromExpr, stmtCtx));
   mlir::Value toAddress = fir::getBase(converter.genExprAddr(
       *Fortran::semantics::GetExpr(assignmentStmtVariable), stmtCtx));
-  fir::FirOpBuilder &builder = converter.getFirOpBuilder();
-  if (fromAddress.getType() != toAddress.getType())
-    fromAddress =
-        builder.create<fir::ConvertOp>(loc, toAddress.getType(), fromAddress);
   genOmpAccAtomicCaptureStatement(converter, fromAddress, toAddress,
                                   leftHandClauseList, rightHandClauseList,
                                   elementType, loc);
@@ -497,23 +497,12 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
   // a `atomic.read`, `atomic.write`, or `atomic.update` operation
   // inside `atomic.capture`
   Fortran::lower::StatementContext stmtCtx;
-  mlir::Value stmt1LHSArg, stmt1RHSArg, stmt2LHSArg, stmt2RHSArg;
-  mlir::Type elementType;
   // LHS evaluations are common to all combinations of `atomic.capture`
-  stmt1LHSArg = fir::getBase(converter.genExprAddr(assign1.lhs, stmtCtx));
-  stmt2LHSArg = fir::getBase(converter.genExprAddr(assign2.lhs, stmtCtx));
+  mlir::Value stmt1LHSArg =
+      fir::getBase(converter.genExprAddr(assign1.lhs, stmtCtx));
+  mlir::Value stmt2LHSArg =
+      fir::getBase(converter.genExprAddr(assign2.lhs, stmtCtx));
 
-  // Operation specific RHS evaluations
-  if (Fortran::semantics::checkForSingleVariableOnRHS(stmt1)) {
-    // Atomic capture construct is of the form [capture-stmt, update-stmt] or
-    // of the form [capture-stmt, write-stmt]
-    stmt1RHSArg = fir::getBase(converter.genExprAddr(assign1.rhs, stmtCtx));
-    stmt2RHSArg = fir::getBase(converter.genExprValue(assign2.rhs, stmtCtx));
-  } else {
-    // Atomic capture construct is of the form [update-stmt, capture-stmt]
-    stmt1RHSArg = fir::getBase(converter.genExprValue(assign1.rhs, stmtCtx));
-    stmt2RHSArg = fir::getBase(converter.genExprAddr(assign2.lhs, stmtCtx));
-  }
   // Type information used in generation of `atomic.update` operation
   mlir::Type stmt1VarType =
       fir::getBase(converter.genExprValue(assign1.lhs, stmtCtx)).getType();
@@ -545,44 +534,46 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
       // Atomic capture construct is of the form [capture-stmt, update-stmt]
       const Fortran::semantics::SomeExpr &fromExpr =
           *Fortran::semantics::GetExpr(stmt1Expr);
-      elementType = converter.genType(fromExpr);
+      mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
-          converter, stmt1RHSArg, stmt1LHSArg,
+          converter, stmt2LHSArg, stmt1LHSArg,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, elementType, loc);
       genOmpAccAtomicUpdateStatement<AtomicListT>(
-          converter, stmt1RHSArg, stmt2VarType, stmt2Var, stmt2Expr,
+          converter, stmt2LHSArg, stmt2VarType, stmt2Var, stmt2Expr,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, loc, atomicCaptureOp);
     } else {
       // Atomic capture construct is of the form [capture-stmt, write-stmt]
+      firOpBuilder.setInsertionPoint(atomicCaptureOp);
+      mlir::Value stmt2RHSArg =
+          fir::getBase(converter.genExprValue(assign2.rhs, stmtCtx));
+      firOpBuilder.setInsertionPointToStart(&block);
       const Fortran::semantics::SomeExpr &fromExpr =
           *Fortran::semantics::GetExpr(stmt1Expr);
-      elementType = converter.genType(fromExpr);
+      mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
-          converter, stmt1RHSArg, stmt1LHSArg,
+          converter, stmt2LHSArg, stmt1LHSArg,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, elementType, loc);
       genOmpAccAtomicWriteStatement<AtomicListT>(
-          converter, stmt1RHSArg, stmt2RHSArg,
+          converter, stmt2LHSArg, stmt2RHSArg,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, loc);
     }
   } else {
     // Atomic capture construct is of the form [update-stmt, capture-stmt]
-    firOpBuilder.setInsertionPointToEnd(&block);
     const Fortran::semantics::SomeExpr &fromExpr =
         *Fortran::semantics::GetExpr(stmt2Expr);
-    elementType = converter.genType(fromExpr);
-    genOmpAccAtomicCaptureStatement<AtomicListT>(
-        converter, stmt1LHSArg, stmt2LHSArg,
-        /*leftHandClauseList=*/nullptr,
-        /*rightHandClauseList=*/nullptr, elementType, loc);
-    firOpBuilder.setInsertionPointToStart(&block);
+    mlir::Type elementType = converter.genType(fromExpr);
     genOmpAccAtomicUpdateStatement<AtomicListT>(
         converter, stmt1LHSArg, stmt1VarType, stmt1Var, stmt1Expr,
         /*leftHandClauseList=*/nullptr,
         /*rightHandClauseList=*/nullptr, loc, atomicCaptureOp);
+    genOmpAccAtomicCaptureStatement<AtomicListT>(
+        converter, stmt1LHSArg, stmt2LHSArg,
+        /*leftHandClauseList=*/nullptr,
+        /*rightHandClauseList=*/nullptr, elementType, loc);
   }
   firOpBuilder.setInsertionPointToEnd(&block);
   if constexpr (std::is_same<AtomicListT,
diff --git a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
index 168526518865b4..184abe24fe967d 100644
--- a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
+++ b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
@@ -781,11 +781,11 @@ func.func @_QPsimple_reduction(%arg0: !fir.ref<!fir.array<100x!fir.logical<4>>>
 // -----
 
 // CHECK: llvm.func @_QPs
-// CHECK: omp.atomic.read %{{.*}} = %{{.*}}   : !llvm.ptr, !llvm.struct<(f32, f32)>
+// CHECK: omp.atomic.read %{{.*}} = %{{.*}}   : !llvm.ptr, !llvm.ptr, !llvm.struct<(f32, f32)>
 
 func.func @_QPs(%arg0: !fir.ref<complex<f32>> {fir.bindc_name = "x"}) {
   %0 = fir.alloca complex<f32> {bindc_name = "v", uniq_name = "_QFsEv"}
-  omp.atomic.read %0 = %arg0   : !fir.ref<complex<f32>>, complex<f32>
+  omp.atomic.read %0 = %arg0   : !fir.ref<complex<f32>>, !fir.ref<complex<f32>>, complex<f32>
   return
 }
 
diff --git a/flang/test/Lower/OpenACC/acc-atomic-capture.f90 b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
index 373683386fda90..66b8e8c5843a81 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-capture.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
@@ -11,7 +11,7 @@ program acc_atomic_capture_test
 !CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %2 {uniq_name = "_QFEy"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK: %[[temp:.*]] = fir.load %[[X_DECL]]#0 : !fir.ref<i32>
 !CHECK: acc.atomic.capture {
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: acc.atomic.update %[[Y_DECL]]#1 : !fir.ref<i32> {
 !CHECK: ^bb0(%[[ARG:.*]]: i32):
 !CHECK: %[[result:.*]] = arith.addi %[[temp]], %[[ARG]] : i32
@@ -32,7 +32,7 @@ program acc_atomic_capture_test
 !CHECK: %[[result:.*]] = arith.muli %[[temp]], %[[ARG]] : i32
 !CHECK: acc.yield %[[result]] : i32
 !CHECK: }
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: }
 
     !$acc atomic capture
@@ -47,7 +47,7 @@ program acc_atomic_capture_test
 !CHECK: %[[result_noreassoc:.*]] = hlfir.no_reassoc %[[result]] : i32
 !CHECK: %[[result:.*]] = arith.addi %[[constant_20]], %[[result_noreassoc]] : i32
 !CHECK: acc.atomic.capture {
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: acc.atomic.write %[[Y_DECL]]#1 = %[[result]] : !fir.ref<i32>, i32
 !CHECK: }
 
@@ -82,7 +82,7 @@ subroutine pointers_in_atomic_capture()
 !CHECK: %[[result:.*]] = arith.addi %[[ARG]], %[[loaded_value]] : i32
 !CHECK: acc.yield %[[result]] : i32
 !CHECK: }
-!CHECK: acc.atomic.read %[[loaded_B_addr]] = %[[loaded_A_addr]] : !fir.ptr<i32>, i32
+!CHECK: acc.atomic.read %[[loaded_B_addr]] = %[[loaded_A_addr]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
 !CHECK: }
     integer, pointer :: a, b
     integer, target :: c, d
@@ -118,10 +118,95 @@ subroutine capture_with_convert_f32_to_i32()
 ! CHECK: %[[MUL:.*]] = arith.mulf %{{.*}}, %[[CST]] fastmath<contract> : f32
 ! CHECK: %[[CONV:.*]] = fir.convert %[[MUL]] : (f32) -> i32
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[K_DECL]]#1 : !fir.ref<i32>, i32
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[K_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:   acc.atomic.write %[[K_DECL]]#1 = %[[CONV]] : !fir.ref<i32>, i32
 ! CHECK: }
 
+subroutine capture_with_convert_i32_to_f64()
+  real(8) :: x
+  integer :: v
+  x = 1.0
+  v = 0
+  !$acc atomic capture
+  v = x
+  x = v
+  !$acc end atomic
+end subroutine capture_with_convert_i32_to_f64
+
+! CHECK-LABEL: func.func @_QPcapture_with_convert_i32_to_f64()
+! CHECK: %[[V:.*]] = fir.alloca i32 {bindc_name = "v", uniq_name = "_QFcapture_with_convert_i32_to_f64Ev"}
+! CHECK: %[[V_DECL:.*]]:2 = hlfir.declare %[[V]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ev"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[X:.*]] = fir.alloca f64 {bindc_name = "x", uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"}
+! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"} : (!fir.ref<f64>) -> (!fir.ref<f64>, !fir.ref<f64>)
+! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
+! CHECK: %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f64
+! CHECK: acc.atomic.capture {
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i32>, !fir.ref<f64>, f64
+! CHECK:   acc.atomic.write %[[X_DECL]]#1 = %[[CONV]] : !fir.ref<f64>, f64
+! CHECK: }
+
+subroutine capture_with_convert_f64_to_i32()
+  integer :: x
+  real(8) :: v
+  x = 1
+  v = 0
+  !$acc atomic capture
+  x = v * v
+  v = x
+  !$acc end atomic
+end subroutine capture_with_convert_f64_to_i32
+
+! CHECK-LABEL: func.func @_QPcapture_with_convert_f64_to_i32()
+! CHECK: %[[V:.*]] = fir.alloca f64 {bindc_name = "v", uniq_name = "_QFcapture_with_convert_f64_to_i32Ev"}
+! CHECK: %[[V_DECL:.*]]:2 = hlfir.declare %[[V]] {uniq_name = "_QFcapture_with_convert_f64_to_i32Ev"} : (!fir.ref<f64>) -> (!fir.ref<f64>, !fir.ref<f64>)
+! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFcapture_with_convert_f64_to_i32Ex"}
+! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcapture_with_convert_f64_to_i32Ex"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %c1_i32 = arith.constant 1 : i32
+! CHECK: hlfir.assign %c1_i32 to %[[X_DECL]]#0 : i32, !fir.ref<i32>
+! CHECK: %[[CST:.*]] = arith.constant 0.000000e+00 : f64
+! CHECK: hlfir.assign %[[CST]] to %[[V_DECL]]#0 : f64, !fir.ref<f64>
+! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<f64>
+! CHECK: acc.atomic.capture {
+! CHECK:   acc.atomic.update %[[X_DECL]]#1 : !fir.ref<i32> {
+! CHECK:   ^bb0(%arg0: i32):
+! CHECK:     %[[MUL:.*]] = arith.mulf %[[LOAD]], %[[LOAD]] fastmath<contract> : f64
+! CHECK:     %[[CONV:.*]] = fir.convert %[[MUL]] : (f64) -> i32
+! CHECK:     acc.yield %[[CONV]] : i32
+! CHECK:   }
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<f64>, !fir.ref<i32>, i32
+! CHECK: }
+
+subroutine capture_with_convert_i32_to_f32()
+  real(4) :: x
+  integer :: v
+  x = 1.0
+  v = 0
+  !$acc atomic capture
+  v = x
+  x = x + v
+  !$acc end atomic
+end subroutine capture_with_convert_i32_to_f32
+
+! CHECK-LABEL: func.func @_QPcapture_with_convert_i32_to_f32()
+! CHECK: %[[V:.*]] = fir.alloca i32 {bindc_name = "v", uniq_name = "_QFcapture_with_convert_i32_to_f32Ev"}
+! CHECK: %[[V_DECL:.*]]:2 = hlfir.declare %[[V]] {uniq_name = "_QFcapture_with_convert_i32_to_f32Ev"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[X:.*]] = fir.alloca f32 {bindc_name = "x", uniq_name = "_QFcapture_with_convert_i32_to_f32Ex"}
+! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcapture_with_convert_i32_to_f32Ex"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
+! CHECK: %[[CST:.*]] = arith.constant 1.000000e+00 : f32
+! CHECK: hlfir.assign %[[CST]] to %[[X_DECL]]#0 : f32, !fir.ref<f32>
+! CHECK: %c0_i32 = arith.constant 0 : i32
+! CHECK: hlfir.assign %c0_i32 to %[[V_DECL]]#0 : i32, !fir.ref<i32>
+! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
+! CHECK: acc.atomic.capture {
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i32>, !fir.ref<f32>, f32
+! CHECK:   acc.atomic.update %[[X_DECL]]#1 : !fir.ref<f32> {
+! CHECK:   ^bb0(%arg0: f32):
+! CHECK:     %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f32
+! CHECK:     %[[ADD:.*]] = arith.addf %arg0, %[[CONV]] fastmath<contract> : f32
+! CHECK:     acc.yield %[[ADD]] : f32
+! CHECK:   }
+! CHECK: }
+
 subroutine array_ref_in_atomic_capture1
   integer :: x(10), v
   !$acc atomic capture
@@ -136,7 +221,7 @@ end subroutine array_ref_in_atomic_capture1
 ! CHECK:           %[[X_DECL:.*]]:2 = hlfir.declare %[[X]](%{{.*}}) {uniq_name = "_QFarray_ref_in_atomic_capture1Ex"} : (!fir.ref<!fir.array<10xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<10xi32>>, !fir.ref<!fir.array<10xi32>>)
 ! CHECK:           %[[X_REF:.*]] = hlfir.designate %[[X_DECL]]#0 (%{{.*}})  : (!fir.ref<!fir.array<10xi32>>, index) -> !fir.ref<i32>
 ! CHECK:           acc.atomic.capture {
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:             acc.atomic.update %[[X_REF]] : !fir.ref<i32> {
 ! CHECK:             ^bb0(%[[VAL_7:.*]]: i32):
 ! CHECK:               %[[VAL_8:.*]] = arith.addi %[[VAL_7]], %{{.*}} : i32
@@ -163,7 +248,7 @@ end subroutine array_ref_in_atomic_capture2
 ! CHECK:               %[[VAL_8:.*]] = arith.addi %[[VAL_7]], %{{.*}} : i32
 ! CHECK:               acc.yield %[[VAL_8]] : i32
 ! CHECK:             }
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:           }
 
 subroutine comp_ref_in_atomic_capture1
@@ -184,7 +269,7 @@ end subroutine comp_ref_in_atomic_capture1
 ! CHECK:           %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcomp_ref_in_atomic_capture1Ex"} : (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>) -> (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>, !fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>)
 ! CHECK:           %[[C:.*]] = hlfir.designate %[[X_DECL]]#0{"c"}   : (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>) -> !fir.ref<i32>
 ! CHECK:           acc.atomic.capture {
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:             acc.atomic.update %[[C]] : !fir.ref<i32> {
 ! CHECK:             ^bb0(%[[VAL_5:.*]]: i32):
 ! CHECK:               %[[VAL_6:.*]] = arith.addi %[[VAL_5]], %{{.*}} : i32
@@ -215,5 +300,5 @@ end subroutine comp_ref_in_atomic_capture2
 ! CHECK:               %[[VAL_6:.*]] = arith.addi %[[VAL_5]], %{{.*}} : i32
 ! CHECK:               acc.yield %[[VAL_6]] : i32
 ! CHECK:             }
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:           }
diff --git a/flang/test/Lower/OpenACC/acc-atomic-read.f90 b/flang/test/Lower/OpenACC/acc-atomic-read.f90
index c1a97a9e5f74f3..f2cbe6e45596a4 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-read.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-read.f90
@@ -13,7 +13,7 @@ end program acc_atomic_test
 ! CHECK: %[[G_DECL:.*]]:2 = hlfir.declare %[[VAR_G]] {uniq_name = "_QFEg"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
 ! CHECK: %[[VAR_H:.*]] = fir.alloca f32 {bindc_name = "h", uniq_name = "_QFEh"}
 ! CHECK: %[[H_DECL:.*]]:2 = hlfir.declare %[[VAR_H]] {uniq_name = "_QFEh"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
-! CHECK: acc.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1 : !fir.ref<f32>, f32
+! CHECK: acc.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1 : !fir.ref<f32>, !fir.ref<f32>, f32
 ! CHECK: return
 ! CHECK: }
 
@@ -39,10 +39,10 @@ subroutine atomic_read_pointer()
 ! CHECK:   %[[BOX_ADDR_X:.*]] = fir.box_addr %[[LOAD_X]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 ! CHECK:   %[[LOAD_Y:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 ! CHECK:   %[[BOX_ADDR_Y:.*]] = fir.box_addr %[[LOAD_Y]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
-! CHECK:   acc.atomic.read %[[BOX_ADDR_Y]] = %[[BOX_ADDR_X]] : !fir.ptr<i32>, i32
+! CHECK:   acc.atomic.read %[[BOX_ADDR_Y]] = %[[BOX_ADDR_X]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
 ! CHECK: }
 
-subroutine atomic_read_with_convert()
+subroutine atomic_read_with_cast()
   integer(4) :: x
   integer(8) :: y
 
@@ -50,10 +50,9 @@ subroutine atomic_read_with_convert()
   y = x
 end
 
-! CHECK-LABEL: func.func @_QPatomic_read_with_convert() {
-! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_convertEx"}
-! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_convertEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_convertEy"}
-! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_convertEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
-! CHECK: %[[CONV:.*]] = fir.convert %[[X_DECL]]#1 : (!fir.ref<i32>) -> !fir.ref<i64>
-! CHECK: acc.atomic.read %[[Y_DECL]]#1 = %[[CONV]] : !fir.ref<i64>, i32
+! CHECK-LABEL: func.func @_QPatomic_read_with_cast() {
+! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_castEx"}
+! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_castEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_castEy"}
+! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_castEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
+! CHECK: acc.atomic.read %[[Y_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i64>, !fir.ref<i32>, i32
diff --git a/flang/test/Lower/OpenACC/acc-atomic-update-array.f90 b/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
index eeb7ea29940862..f89a9ab457d499 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
@@ -45,7 +45,7 @@ subroutine atomic_read_array1(r, n, x)
 ! CHECK: %[[DECL_X:.*]]:2 = hlfir.declare %[[ARG2]] dummy_scope %{{[0-9]+}} {uniq_name = "_QFatomic_read_array1Ex"} : (!fir.ref<f32>, !fir.dscope) -> (!fir.ref<f32>, !fir.ref<f32>)
 ! CHECK: %[[DECL_R:.*]]:2 = hlfir.declare %[[ARG0]](%{{.*}}) dummy_scope %{{[0-9]+}} {uniq_name = "_QFatomic_read_array1Er"} : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.ref<!fir.array<?xf32>>)
 ! CHECK: %[[DES:.*]] = hlfir.designate %[[DECL_R]]#0 (%{{.*}})  : (!fir.box<!fir.array<?xf32>>, i64) -> !fir.ref<f32>
-! CHECK: acc.atomic.read %[[DECL_X]]#1 = %[[DES]] : !fir.ref<f32>, f32
+! CHECK: acc.atomic.read %[[DECL_X]]#1 = %[[DES]] : !fir.ref<f32>, !fir.ref<f32>, f32
 
 subroutine atomic_write_array1(r, n, x)
   implicit none
@@ -88,5 +88,5 @@ subroutine atomic_capture_array1(r, n, x, y)
 ! CHECK:     %[[ADD:.*]] = arith.addf %[[ARG]], %[[LOAD]] fastmath<contract> : f32
 ! CHECK:     acc.yield %[[ADD]] : f32
 ! CHECK:   }
-! CHECK:   acc.atomic.read %[[DECL_Y]]#1 = %[[R_I]] : !fir.ref<f32>, f32
+! CHECK:   acc.atomic.read %[[DECL_Y]]#1 = %[[R_I]] : !fir.ref<f32>, !fir.ref<f32>, f32
 ! CHECK: }
diff --git a/flang/test/Lower/OpenMP/atomic-capture.f90 b/flang/test/Lower/OpenMP/atomic-capture.f90
index af82e4b2a20eb2..679d22d3d7063e 100644
--- a/flang/test/Lower/OpenMP/atomic-capture.f90
+++ b/flang/test/Lower/OpenMP/atomic-capture.f90
@@ -22,7 +22,7 @@ program OmpAtomicCapture
 !CHECK: %[[TEMP:.*]] = arith.muli %[[VAL_Y_LOADED]], %[[ARG]] : i32
 !CHECK: omp.yield(%[[TEMP]] : i32)
 !CHECK: }
-!CHECK: omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, i32
+!CHECK: omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: }
     !$omp atomic hint(omp_sync_hint_uncontended) capture
         y = x * y 
@@ -36,7 +36,7 @@ program OmpAtomicCapture
 !CHECK: %[[NO_REASSOC:.*]] = hlfir.no_reassoc %[[SUB]] : i32
 !CHECK: %[[ADD:.*]] = arith.addi  %[[VAL_20]], %[[NO_REASSOC]] : i32
 !CHECK: omp.atomic.capture hint(nonspeculative) memory_order(acquire) {
-!CHECK:   omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, i32
+!CHECK:   omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK:   omp.atomic.write %[[VAL_Y_DECLARE]]#1 = %[[ADD]] : !fir.ref<i32>, i32
 !CHECK: }
 !CHECK: return
@@ -88,7 +88,7 @@ subroutine pointers_in_atomic_capture()
 !CHECK: %[[TEMP:.*]] = arith.addi %[[ARG]], %[[VAL_B]] : i32
 !CHECK: omp.yield(%[[TEMP]] : i32)
 !CHECK: }
-!CHECK: omp.atomic.read %[[VAL_B_BOX_ADDR]] = %[[VAL_A_BOX_ADDR]] : !fir.ptr<i32>, i32
+!CHECK: omp.atomic.read %[[VAL_B_BOX_ADDR]] = %[[VAL_A_BOX_ADDR]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
 !CHECK: }
 !CHECK: return
 !CHECK: }
diff --git a/flang/test/Lower/OpenMP/atomic-read.f90 b/flang/test/Lower/OpenMP/atomic-read.f90
index c3270dd6c1d670..e9bea42252faa3 100644
--- a/flang/test/Lower/OpenMP/atomic-read.f90
+++ b/flang/test/Lower/OpenMP/atomic-read.f90
@@ -25,12 +25,12 @@
 !CHECK:    %[[X_DECL:.*]]:2 = hlfir.declare %[[X_REF]] {uniq_name = "_QFEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK:    %[[Y_REF:.*]] = fir.alloca i32 {bindc_name = "y", uniq_name = "_QFEy"}
 !CHECK:    %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y_REF]] {uniq_name = "_QFEy"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-!CHECK:    omp.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1   hint(uncontended) memory_order(acquire) : !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[A_DECL]]#1 = %[[B_DECL]]#1   memory_order(relaxed) : !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[C_DECL]]#1 = %[[D_DECL]]#1   hint(contended) memory_order(seq_cst) : !fir.ref<!fir.logical<4>>, !fir.logical<4>
-!CHECK:    omp.atomic.read %[[E_DECL]]#1 = %[[F_DECL]]#1   hint(speculative) : !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   hint(nonspeculative) : !fir.ref<f32>, f32
-!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   : !fir.ref<f32>, f32
+!CHECK:    omp.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1   hint(uncontended) memory_order(acquire) : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[A_DECL]]#1 = %[[B_DECL]]#1   memory_order(relaxed) : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[C_DECL]]#1 = %[[D_DECL]]#1   hint(contended) memory_order(seq_cst) : !fir.ref<!fir.logical<4>>, !fir.ref<!fir.logical<4>>, !fir.logical<4>
+!CHECK:    omp.atomic.read %[[E_DECL]]#1 = %[[F_DECL]]#1   hint(speculative) : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   hint(nonspeculative) : !fir.ref<f32>, !fir.ref<f32>, f32
+!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   : !fir.ref<f32>, !fir.ref<f32>, f32
 
 program OmpAtomic
 
@@ -68,7 +68,7 @@ end program OmpAtomic
 !CHECK:    %[[X_POINTEE_ADDR:.*]] = fir.box_addr %[[X_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 !CHECK:    %[[Y_ADDR:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 !CHECK:    %[[Y_POINTEE_ADDR:.*]] = fir.box_addr %[[Y_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
-!CHECK:    omp.atomic.read %[[Y_POINTEE_ADDR]] = %[[X_POINTEE_ADDR]]   : !fir.ptr<i32>, i32
+!CHECK:    omp.atomic.read %[[Y_POINTEE_ADDR]] = %[[X_POINTEE_ADDR]]   : !fir.ptr<i32>, !fir.ptr<i32>, i32
 !CHECK:    %[[Y_ADDR:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 !CHECK:    %[[Y_POINTEE_ADDR:.*]] = fir.box_addr %[[Y_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 !CHECK:    %[[Y_POINTEE_VAL:.*]] = fir.load %[[Y_POINTEE_ADDR]] : !fir.ptr<i32>
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index e305e2fbde5b17..1d179c74a970e6 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -1958,8 +1958,7 @@ def OpenACC_YieldOp : OpenACC_Op<"yield", [Pure, ReturnLike, Terminator,
 // 2.12 atomic construct
 //===----------------------------------------------------------------------===//
 
-def AtomicReadOp : OpenACC_Op<"atomic.read", [AllTypesMatch<["x", "v"]>,
-                                              AtomicReadOpInterface]> {
+def AtomicReadOp : OpenACC_Op<"atomic.read", [AtomicReadOpInterface]> {
 
   let summary = "performs an atomic read";
 
@@ -1975,7 +1974,7 @@ def AtomicReadOp : OpenACC_Op<"atomic.read", [AllTypesMatch<["x", "v"]>,
                        TypeAttr:$element_type);
   let assemblyFormat = [{
     $v `=` $x
-    `:` type($x) `,` $element_type attr-dict
+    `:` type($v) `,` type($x) `,` $element_type attr-dict
   }];
   let hasVerifier = 1;
 }
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 626539cb7bde42..5fd8184fe0e0f7 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1286,7 +1286,7 @@ def TaskwaitOp : OpenMP_Op<"taskwait", clauses = [
 // two-step process.
 
 def AtomicReadOp : OpenMP_Op<"atomic.read", traits = [
-    AllTypesMatch<["x", "v"]>, AtomicReadOpInterface
+    AtomicReadOpInterface
   ], clauses = [
     OpenMP_HintClause, OpenMP_MemoryOrderClause
   ]> {
@@ -1304,7 +1304,8 @@ def AtomicReadOp : OpenMP_Op<"atomic.read", traits = [
 
   // Override clause-based assemblyFormat.
   let assemblyFormat = "$v `=` $x" # clausesReqAssemblyFormat # " oilist(" #
-    clausesOptAssemblyFormat # ") `:` type($x) `,` $element_type attr-dict";
+    clausesOptAssemblyFormat #
+    ") `:` type($v) `,` type($x) `,` $element_type attr-dict";
 
   let extraClassDeclaration = [{
     /// The number of variable operands.

>From 0f56eb52cbfe0a8bb555e72b21636953c0c48467 Mon Sep 17 00:00:00 2001
From: Kazuaki Matsumura <kmatsumura at nvidia.com>
Date: Fri, 1 Nov 2024 12:36:34 -0700
Subject: [PATCH 2/5] [flang] Generate a temp for atomic.read

---
 flang/lib/Lower/DirectivesCommon.h            | 96 +++++++++++++++++--
 .../Fir/convert-to-llvm-openmp-and-fir.fir    |  4 +-
 .../test/Lower/OpenACC/acc-atomic-capture.f90 | 62 +++++++-----
 flang/test/Lower/OpenACC/acc-atomic-read.f90  | 22 +++--
 .../Lower/OpenACC/acc-atomic-update-array.f90 |  4 +-
 flang/test/Lower/OpenMP/atomic-capture.f90    |  6 +-
 flang/test/Lower/OpenMP/atomic-read.f90       | 14 +--
 7 files changed, 155 insertions(+), 53 deletions(-)

diff --git a/flang/lib/Lower/DirectivesCommon.h b/flang/lib/Lower/DirectivesCommon.h
index 88514b16743278..8121dec7428040 100644
--- a/flang/lib/Lower/DirectivesCommon.h
+++ b/flang/lib/Lower/DirectivesCommon.h
@@ -207,9 +207,9 @@ static inline void genOmpAccAtomicWriteStatement(
 }
 
 /// Used to generate atomic.update operation which is created in existing
-/// location set by builder.
+/// location set by builder. Returns the value that is to be stored to LHS.
 template <typename AtomicListT>
-static inline void genOmpAccAtomicUpdateStatement(
+static inline mlir::Value genOmpAccAtomicUpdateStatement(
     Fortran::lower::AbstractConverter &converter, mlir::Value lhsAddr,
     mlir::Type varType, const Fortran::parser::Variable &assignmentStmtVariable,
     const Fortran::parser::Expr &assignmentStmtExpr,
@@ -337,13 +337,14 @@ static inline void genOmpAccAtomicUpdateStatement(
 
   exprValueOverrides.try_emplace(
       Fortran::semantics::GetExpr(assignmentStmtVariable), val);
+  mlir::Value convertResult;
   {
     // statement context inside the atomic block.
     converter.overrideExprValues(&exprValueOverrides);
     Fortran::lower::StatementContext atomicStmtCtx;
     mlir::Value rhsExpr = fir::getBase(converter.genExprValue(
         *Fortran::semantics::GetExpr(assignmentStmtExpr), atomicStmtCtx));
-    mlir::Value convertResult =
+    convertResult =
         firOpBuilder.createConvert(currentLocation, varType, rhsExpr);
     if constexpr (std::is_same<AtomicListT,
                                Fortran::parser::OmpAtomicClauseList>()) {
@@ -354,6 +355,7 @@ static inline void genOmpAccAtomicUpdateStatement(
     converter.resetExprOverrides();
   }
   firOpBuilder.setInsertionPointAfter(atomicUpdateOp);
+  return convertResult;
 }
 
 /// Processes an atomic construct with write clause.
@@ -414,9 +416,23 @@ void genOmpAccAtomicRead(Fortran::lower::AbstractConverter &converter,
       fir::getBase(converter.genExprAddr(fromExpr, stmtCtx));
   mlir::Value toAddress = fir::getBase(converter.genExprAddr(
       *Fortran::semantics::GetExpr(assignmentStmtVariable), stmtCtx));
-  genOmpAccAtomicCaptureStatement(converter, fromAddress, toAddress,
+  fir::FirOpBuilder &builder = converter.getFirOpBuilder();
+  mlir::Value LHSTemp = toAddress;
+  if (fromAddress.getType() != toAddress.getType()) {
+    // Create a temp for implicit type conversion.
+    LHSTemp = builder.create<fir::AllocaOp>(
+        loc, fir::unwrapRefType(fromAddress.getType()));
+  }
+  genOmpAccAtomicCaptureStatement(converter, fromAddress, LHSTemp,
                                   leftHandClauseList, rightHandClauseList,
                                   elementType, loc);
+  if (fromAddress.getType() != toAddress.getType()) {
+    // Copy the converted value of the temp to the original read LHS.
+    mlir::Value tempVal = builder.create<fir::LoadOp>(loc, LHSTemp);
+    mlir::Value convVal = builder.create<fir::ConvertOp>(
+        loc, fir::unwrapRefType(toAddress.getType()), tempVal);
+    builder.create<fir::StoreOp>(loc, convVal, toAddress);
+  }
 }
 
 /// Processes an atomic construct with update clause.
@@ -475,6 +491,41 @@ void genOmpAtomic(Fortran::lower::AbstractConverter &converter,
       &atomicClauseList, nullptr, loc);
 }
 
+// Replace every load of `addr` found in `expr` with a load of `newAddr`.
+static inline void
+rewriteLoadDependency(Fortran::lower::AbstractConverter &converter,
+                      mlir::Value expr, mlir::Value addr, mlir::Value newAddr) {
+  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  std::set<mlir::Operation *> dependency;
+  dependency.insert(expr.getDefiningOp());
+  while (!dependency.empty()) {
+    mlir::Operation *dep = dependency.extract(dependency.begin()).value();
+    if (!newAddr.getDefiningOp()->getParentOp()->isAncestor(dep)) {
+      // Skip it if it falls outside the parent region of the capture.
+      continue;
+    }
+    if (auto load = mlir::dyn_cast_or_null<fir::LoadOp>(dep)) {
+      if (load.getMemref().getDefiningOp() == addr.getDefiningOp()) {
+        // Found a load of `addr`. Replace it with a load of `newAddr` while
+        // converting the value.
+        firOpBuilder.setInsertionPoint(load);
+        mlir::Location loc = load.getLoc();
+        mlir::Value newLoad = firOpBuilder.create<fir::LoadOp>(loc, newAddr);
+        mlir::Value conv =
+            firOpBuilder.create<fir::ConvertOp>(loc, load.getType(), newLoad);
+        load.replaceAllUsesWith(conv);
+        load.erase();
+      }
+      continue;
+    }
+    for (mlir::Value operand : dep->getOperands()) {
+      if (operand.getDefiningOp()) {
+        dependency.insert(operand.getDefiningOp());
+      }
+    }
+  }
+}
+
 /// Processes an atomic construct with capture clause.
 template <typename AtomicT, typename AtomicListT>
 void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
@@ -529,17 +580,33 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
   firOpBuilder.createBlock(&(atomicCaptureOp->getRegion(0)));
   mlir::Block &block = atomicCaptureOp->getRegion(0).back();
   firOpBuilder.setInsertionPointToStart(&block);
+  mlir::Value captureLHS, storeLHS, LHSTemp;
+  if (Fortran::semantics::checkForSingleVariableOnRHS(stmt1)) {
+    captureLHS = stmt1LHSArg;
+    storeLHS = stmt2LHSArg;
+  } else {
+    captureLHS = stmt2LHSArg;
+    storeLHS = stmt1LHSArg;
+  }
+  LHSTemp = captureLHS;
+  if (captureLHS.getType() != storeLHS.getType()) {
+    // Create a temp for implicit type conversion.
+    LHSTemp = firOpBuilder.create<fir::AllocaOp>(
+        loc, fir::unwrapRefType(storeLHS.getType()));
+    LHSTemp.getDefiningOp()->moveBefore(atomicCaptureOp);
+  }
   if (Fortran::semantics::checkForSingleVariableOnRHS(stmt1)) {
+    mlir::Value storeValue;
     if (Fortran::semantics::checkForSymbolMatch(stmt2)) {
       // Atomic capture construct is of the form [capture-stmt, update-stmt]
       const Fortran::semantics::SomeExpr &fromExpr =
           *Fortran::semantics::GetExpr(stmt1Expr);
       mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
-          converter, stmt2LHSArg, stmt1LHSArg,
+          converter, stmt2LHSArg, LHSTemp,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, elementType, loc);
-      genOmpAccAtomicUpdateStatement<AtomicListT>(
+      storeValue = genOmpAccAtomicUpdateStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt2VarType, stmt2Var, stmt2Expr,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, loc, atomicCaptureOp);
@@ -553,14 +620,19 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
           *Fortran::semantics::GetExpr(stmt1Expr);
       mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
-          converter, stmt2LHSArg, stmt1LHSArg,
+          converter, stmt2LHSArg, LHSTemp,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, elementType, loc);
       genOmpAccAtomicWriteStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt2RHSArg,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, loc);
+      storeValue = stmt2RHSArg;
     }
+    if (captureLHS.getType() != storeLHS.getType())
+      // Replace every load of the capture LHS in the store value with the
+      // corresponding temp.
+      rewriteLoadDependency(converter, storeValue, captureLHS, LHSTemp);
   } else {
     // Atomic capture construct is of the form [update-stmt, capture-stmt]
     const Fortran::semantics::SomeExpr &fromExpr =
@@ -571,10 +643,18 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
         /*leftHandClauseList=*/nullptr,
         /*rightHandClauseList=*/nullptr, loc, atomicCaptureOp);
     genOmpAccAtomicCaptureStatement<AtomicListT>(
-        converter, stmt1LHSArg, stmt2LHSArg,
+        converter, stmt1LHSArg, LHSTemp,
         /*leftHandClauseList=*/nullptr,
         /*rightHandClauseList=*/nullptr, elementType, loc);
   }
+  if (captureLHS.getType() != storeLHS.getType()) {
+    // Copy the converted value of the temp to the original capture LHS.
+    firOpBuilder.setInsertionPointAfter(atomicCaptureOp);
+    mlir::Value tempVal = firOpBuilder.create<fir::LoadOp>(loc, LHSTemp);
+    mlir::Value convVal = firOpBuilder.create<fir::ConvertOp>(
+        loc, fir::unwrapRefType(captureLHS.getType()), tempVal);
+    firOpBuilder.create<fir::StoreOp>(loc, convVal, captureLHS);
+  }
   firOpBuilder.setInsertionPointToEnd(&block);
   if constexpr (std::is_same<AtomicListT,
                              Fortran::parser::OmpAtomicClauseList>()) {
diff --git a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
index 184abe24fe967d..168526518865b4 100644
--- a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
+++ b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
@@ -781,11 +781,11 @@ func.func @_QPsimple_reduction(%arg0: !fir.ref<!fir.array<100x!fir.logical<4>>>
 // -----
 
 // CHECK: llvm.func @_QPs
-// CHECK: omp.atomic.read %{{.*}} = %{{.*}}   : !llvm.ptr, !llvm.ptr, !llvm.struct<(f32, f32)>
+// CHECK: omp.atomic.read %{{.*}} = %{{.*}}   : !llvm.ptr, !llvm.struct<(f32, f32)>
 
 func.func @_QPs(%arg0: !fir.ref<complex<f32>> {fir.bindc_name = "x"}) {
   %0 = fir.alloca complex<f32> {bindc_name = "v", uniq_name = "_QFsEv"}
-  omp.atomic.read %0 = %arg0   : !fir.ref<complex<f32>>, !fir.ref<complex<f32>>, complex<f32>
+  omp.atomic.read %0 = %arg0   : !fir.ref<complex<f32>>, complex<f32>
   return
 }
 
diff --git a/flang/test/Lower/OpenACC/acc-atomic-capture.f90 b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
index 66b8e8c5843a81..08f866711ecd8f 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-capture.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
@@ -11,7 +11,7 @@ program acc_atomic_capture_test
 !CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %2 {uniq_name = "_QFEy"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK: %[[temp:.*]] = fir.load %[[X_DECL]]#0 : !fir.ref<i32>
 !CHECK: acc.atomic.capture {
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, i32
 !CHECK: acc.atomic.update %[[Y_DECL]]#1 : !fir.ref<i32> {
 !CHECK: ^bb0(%[[ARG:.*]]: i32):
 !CHECK: %[[result:.*]] = arith.addi %[[temp]], %[[ARG]] : i32
@@ -32,7 +32,7 @@ program acc_atomic_capture_test
 !CHECK: %[[result:.*]] = arith.muli %[[temp]], %[[ARG]] : i32
 !CHECK: acc.yield %[[result]] : i32
 !CHECK: }
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, i32
 !CHECK: }
 
     !$acc atomic capture
@@ -47,7 +47,7 @@ program acc_atomic_capture_test
 !CHECK: %[[result_noreassoc:.*]] = hlfir.no_reassoc %[[result]] : i32
 !CHECK: %[[result:.*]] = arith.addi %[[constant_20]], %[[result_noreassoc]] : i32
 !CHECK: acc.atomic.capture {
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, i32
 !CHECK: acc.atomic.write %[[Y_DECL]]#1 = %[[result]] : !fir.ref<i32>, i32
 !CHECK: }
 
@@ -82,7 +82,7 @@ subroutine pointers_in_atomic_capture()
 !CHECK: %[[result:.*]] = arith.addi %[[ARG]], %[[loaded_value]] : i32
 !CHECK: acc.yield %[[result]] : i32
 !CHECK: }
-!CHECK: acc.atomic.read %[[loaded_B_addr]] = %[[loaded_A_addr]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
+!CHECK: acc.atomic.read %[[loaded_B_addr]] = %[[loaded_A_addr]] : !fir.ptr<i32>, i32
 !CHECK: }
     integer, pointer :: a, b
     integer, target :: c, d
@@ -118,7 +118,7 @@ subroutine capture_with_convert_f32_to_i32()
 ! CHECK: %[[MUL:.*]] = arith.mulf %{{.*}}, %[[CST]] fastmath<contract> : f32
 ! CHECK: %[[CONV:.*]] = fir.convert %[[MUL]] : (f32) -> i32
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[K_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[K_DECL]]#1 : !fir.ref<i32>, i32
 ! CHECK:   acc.atomic.write %[[K_DECL]]#1 = %[[CONV]] : !fir.ref<i32>, i32
 ! CHECK: }
 
@@ -138,12 +138,21 @@ end subroutine capture_with_convert_i32_to_f64
 ! CHECK: %[[V_DECL:.*]]:2 = hlfir.declare %[[V]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ev"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: %[[X:.*]] = fir.alloca f64 {bindc_name = "x", uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"}
 ! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"} : (!fir.ref<f64>) -> (!fir.ref<f64>, !fir.ref<f64>)
-! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
-! CHECK: %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f64
+! CHECK: %[[CST:.*]] = arith.constant 1.000000e+00 : f64
+! CHECK: hlfir.assign %[[CST]] to %[[X_DECL]]#0 : f64, !fir.ref<f64>
+! CHECK: %c0_i32 = arith.constant 0 : i32
+! CHECK: hlfir.assign %c0_i32 to %[[V_DECL]]#0 : i32, !fir.ref<i32>
+! CHECK: %[[ALLOCA:.*]] = fir.alloca f64
+! CHECK: %[[LOAD1:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f64>
+! CHECK: %[[CONV1:.*]] = fir.convert %[[LOAD1]] : (f64) -> i32
+! CHECK: %[[CONV2:.*]] = fir.convert %[[CONV1]] : (i32) -> f64
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i32>, !fir.ref<f64>, f64
-! CHECK:   acc.atomic.write %[[X_DECL]]#1 = %[[CONV]] : !fir.ref<f64>, f64
+! CHECK:   acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<f64>, f64
+! CHECK:   acc.atomic.write %[[X_DECL]]#1 = %[[CONV2]] : !fir.ref<f64>, f64
 ! CHECK: }
+! CHECK: %[[LOAD2:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f64>
+! CHECK: %[[CONV3:.*]] = fir.convert %[[LOAD2]] : (f64) -> i32
+! CHECK: fir.store %[[CONV3]] to %[[V_DECL]]#1 : !fir.ref<i32>
 
 subroutine capture_with_convert_f64_to_i32()
   integer :: x
@@ -165,16 +174,20 @@ end subroutine capture_with_convert_f64_to_i32
 ! CHECK: hlfir.assign %c1_i32 to %[[X_DECL]]#0 : i32, !fir.ref<i32>
 ! CHECK: %[[CST:.*]] = arith.constant 0.000000e+00 : f64
 ! CHECK: hlfir.assign %[[CST]] to %[[V_DECL]]#0 : f64, !fir.ref<f64>
-! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<f64>
+! CHECK: %[[ALLOCA]] = fir.alloca i32
+! CHECK: %[[LOAD1:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<f64>
 ! CHECK: acc.atomic.capture {
 ! CHECK:   acc.atomic.update %[[X_DECL]]#1 : !fir.ref<i32> {
 ! CHECK:   ^bb0(%arg0: i32):
-! CHECK:     %[[MUL:.*]] = arith.mulf %[[LOAD]], %[[LOAD]] fastmath<contract> : f64
-! CHECK:     %[[CONV:.*]] = fir.convert %[[MUL]] : (f64) -> i32
-! CHECK:     acc.yield %[[CONV]] : i32
+! CHECK:     %[[MUL:.*]] = arith.mulf %[[LOAD1]], %[[LOAD1]] fastmath<contract> : f64
+! CHECK:     %[[CONV1:.*]] = fir.convert %[[MUL]] : (f64) -> i32
+! CHECK:     acc.yield %[[CONV1]] : i32
 ! CHECK:   }
-! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<f64>, !fir.ref<i32>, i32
+! CHECK:   acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<i32>, i32
 ! CHECK: }
+! CHECK: %[[LOAD2:.*]] = fir.load %[[ALLOCA]] : !fir.ref<i32>
+! CHECK: %[[CONV2:.*]] = fir.convert %[[LOAD2]] : (i32) -> f64
+! CHECK: fir.store %[[CONV2]] to %[[V_DECL]]#1 : !fir.ref<f64>
 
 subroutine capture_with_convert_i32_to_f32()
   real(4) :: x
@@ -196,16 +209,21 @@ end subroutine capture_with_convert_i32_to_f32
 ! CHECK: hlfir.assign %[[CST]] to %[[X_DECL]]#0 : f32, !fir.ref<f32>
 ! CHECK: %c0_i32 = arith.constant 0 : i32
 ! CHECK: hlfir.assign %c0_i32 to %[[V_DECL]]#0 : i32, !fir.ref<i32>
-! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
+! CHECK: %[[ALLOCA:.*]] = fir.alloca f32
+! CHECK: %[[LOAD1:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f32>
+! CHECK: %[[CONV1:.*]] = fir.convert %5 : (f32) -> i32
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i32>, !fir.ref<f32>, f32
+! CHECK:   acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<f32>, f32
 ! CHECK:   acc.atomic.update %[[X_DECL]]#1 : !fir.ref<f32> {
 ! CHECK:   ^bb0(%arg0: f32):
-! CHECK:     %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f32
-! CHECK:     %[[ADD:.*]] = arith.addf %arg0, %[[CONV]] fastmath<contract> : f32
+! CHECK:     %[[CONV2:.*]] = fir.convert %[[CONV1]] : (i32) -> f32
+! CHECK:     %[[ADD:.*]] = arith.addf %arg0, %[[CONV2]] fastmath<contract> : f32
 ! CHECK:     acc.yield %[[ADD]] : f32
 ! CHECK:   }
 ! CHECK: }
+! CHECK: %[[LOAD2:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f32>
+! CHECK: %[[CONV3:.*]] = fir.convert %[[LOAD2]] : (f32) -> i32
+! CHECK: fir.store %[[CONV3]] to %[[V_DECL]]#1 : !fir.ref<i32>
 
 subroutine array_ref_in_atomic_capture1
   integer :: x(10), v
@@ -221,7 +239,7 @@ end subroutine array_ref_in_atomic_capture1
 ! CHECK:           %[[X_DECL:.*]]:2 = hlfir.declare %[[X]](%{{.*}}) {uniq_name = "_QFarray_ref_in_atomic_capture1Ex"} : (!fir.ref<!fir.array<10xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<10xi32>>, !fir.ref<!fir.array<10xi32>>)
 ! CHECK:           %[[X_REF:.*]] = hlfir.designate %[[X_DECL]]#0 (%{{.*}})  : (!fir.ref<!fir.array<10xi32>>, index) -> !fir.ref<i32>
 ! CHECK:           acc.atomic.capture {
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, i32
 ! CHECK:             acc.atomic.update %[[X_REF]] : !fir.ref<i32> {
 ! CHECK:             ^bb0(%[[VAL_7:.*]]: i32):
 ! CHECK:               %[[VAL_8:.*]] = arith.addi %[[VAL_7]], %{{.*}} : i32
@@ -248,7 +266,7 @@ end subroutine array_ref_in_atomic_capture2
 ! CHECK:               %[[VAL_8:.*]] = arith.addi %[[VAL_7]], %{{.*}} : i32
 ! CHECK:               acc.yield %[[VAL_8]] : i32
 ! CHECK:             }
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, i32
 ! CHECK:           }
 
 subroutine comp_ref_in_atomic_capture1
@@ -269,7 +287,7 @@ end subroutine comp_ref_in_atomic_capture1
 ! CHECK:           %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcomp_ref_in_atomic_capture1Ex"} : (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>) -> (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>, !fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>)
 ! CHECK:           %[[C:.*]] = hlfir.designate %[[X_DECL]]#0{"c"}   : (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>) -> !fir.ref<i32>
 ! CHECK:           acc.atomic.capture {
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, i32
 ! CHECK:             acc.atomic.update %[[C]] : !fir.ref<i32> {
 ! CHECK:             ^bb0(%[[VAL_5:.*]]: i32):
 ! CHECK:               %[[VAL_6:.*]] = arith.addi %[[VAL_5]], %{{.*}} : i32
@@ -300,5 +318,5 @@ end subroutine comp_ref_in_atomic_capture2
 ! CHECK:               %[[VAL_6:.*]] = arith.addi %[[VAL_5]], %{{.*}} : i32
 ! CHECK:               acc.yield %[[VAL_6]] : i32
 ! CHECK:             }
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, i32
 ! CHECK:           }
diff --git a/flang/test/Lower/OpenACC/acc-atomic-read.f90 b/flang/test/Lower/OpenACC/acc-atomic-read.f90
index f2cbe6e45596a4..5491521623b9b6 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-read.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-read.f90
@@ -13,7 +13,7 @@ end program acc_atomic_test
 ! CHECK: %[[G_DECL:.*]]:2 = hlfir.declare %[[VAR_G]] {uniq_name = "_QFEg"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
 ! CHECK: %[[VAR_H:.*]] = fir.alloca f32 {bindc_name = "h", uniq_name = "_QFEh"}
 ! CHECK: %[[H_DECL:.*]]:2 = hlfir.declare %[[VAR_H]] {uniq_name = "_QFEh"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
-! CHECK: acc.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1 : !fir.ref<f32>, !fir.ref<f32>, f32
+! CHECK: acc.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1 : !fir.ref<f32>, f32
 ! CHECK: return
 ! CHECK: }
 
@@ -39,10 +39,10 @@ subroutine atomic_read_pointer()
 ! CHECK:   %[[BOX_ADDR_X:.*]] = fir.box_addr %[[LOAD_X]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 ! CHECK:   %[[LOAD_Y:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 ! CHECK:   %[[BOX_ADDR_Y:.*]] = fir.box_addr %[[LOAD_Y]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
-! CHECK:   acc.atomic.read %[[BOX_ADDR_Y]] = %[[BOX_ADDR_X]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
+! CHECK:   acc.atomic.read %[[BOX_ADDR_Y]] = %[[BOX_ADDR_X]] : !fir.ptr<i32>, i32
 ! CHECK: }
 
-subroutine atomic_read_with_cast()
+subroutine atomic_read_with_convert()
   integer(4) :: x
   integer(8) :: y
 
@@ -50,9 +50,13 @@ subroutine atomic_read_with_cast()
   y = x
 end
 
-! CHECK-LABEL: func.func @_QPatomic_read_with_cast() {
-! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_castEx"}
-! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_castEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_castEy"}
-! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_castEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
-! CHECK: acc.atomic.read %[[Y_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i64>, !fir.ref<i32>, i32
+! CHECK-LABEL: func.func @_QPatomic_read_with_convert() {
+! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_convertEx"}
+! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_convertEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_convertEy"}
+! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_convertEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
+! CHECK: %[[ALLOCA:.*]] = fir.alloca i32
+! CHECK: acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<i32>, i32
+! CHECK: %[[LOAD:.*]] = fir.load %[[ALLOCA]] : !fir.ref<i32>
+! CHECK: %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> i64
+! CHECK: fir.store %[[CONV]] to %[[Y_DECL]]#1 : !fir.ref<i64>
diff --git a/flang/test/Lower/OpenACC/acc-atomic-update-array.f90 b/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
index f89a9ab457d499..eeb7ea29940862 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
@@ -45,7 +45,7 @@ subroutine atomic_read_array1(r, n, x)
 ! CHECK: %[[DECL_X:.*]]:2 = hlfir.declare %[[ARG2]] dummy_scope %{{[0-9]+}} {uniq_name = "_QFatomic_read_array1Ex"} : (!fir.ref<f32>, !fir.dscope) -> (!fir.ref<f32>, !fir.ref<f32>)
 ! CHECK: %[[DECL_R:.*]]:2 = hlfir.declare %[[ARG0]](%{{.*}}) dummy_scope %{{[0-9]+}} {uniq_name = "_QFatomic_read_array1Er"} : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.ref<!fir.array<?xf32>>)
 ! CHECK: %[[DES:.*]] = hlfir.designate %[[DECL_R]]#0 (%{{.*}})  : (!fir.box<!fir.array<?xf32>>, i64) -> !fir.ref<f32>
-! CHECK: acc.atomic.read %[[DECL_X]]#1 = %[[DES]] : !fir.ref<f32>, !fir.ref<f32>, f32
+! CHECK: acc.atomic.read %[[DECL_X]]#1 = %[[DES]] : !fir.ref<f32>, f32
 
 subroutine atomic_write_array1(r, n, x)
   implicit none
@@ -88,5 +88,5 @@ subroutine atomic_capture_array1(r, n, x, y)
 ! CHECK:     %[[ADD:.*]] = arith.addf %[[ARG]], %[[LOAD]] fastmath<contract> : f32
 ! CHECK:     acc.yield %[[ADD]] : f32
 ! CHECK:   }
-! CHECK:   acc.atomic.read %[[DECL_Y]]#1 = %[[R_I]] : !fir.ref<f32>, !fir.ref<f32>, f32
+! CHECK:   acc.atomic.read %[[DECL_Y]]#1 = %[[R_I]] : !fir.ref<f32>, f32
 ! CHECK: }
diff --git a/flang/test/Lower/OpenMP/atomic-capture.f90 b/flang/test/Lower/OpenMP/atomic-capture.f90
index 679d22d3d7063e..af82e4b2a20eb2 100644
--- a/flang/test/Lower/OpenMP/atomic-capture.f90
+++ b/flang/test/Lower/OpenMP/atomic-capture.f90
@@ -22,7 +22,7 @@ program OmpAtomicCapture
 !CHECK: %[[TEMP:.*]] = arith.muli %[[VAL_Y_LOADED]], %[[ARG]] : i32
 !CHECK: omp.yield(%[[TEMP]] : i32)
 !CHECK: }
-!CHECK: omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK: omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, i32
 !CHECK: }
     !$omp atomic hint(omp_sync_hint_uncontended) capture
         y = x * y 
@@ -36,7 +36,7 @@ program OmpAtomicCapture
 !CHECK: %[[NO_REASSOC:.*]] = hlfir.no_reassoc %[[SUB]] : i32
 !CHECK: %[[ADD:.*]] = arith.addi  %[[VAL_20]], %[[NO_REASSOC]] : i32
 !CHECK: omp.atomic.capture hint(nonspeculative) memory_order(acquire) {
-!CHECK:   omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:   omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, i32
 !CHECK:   omp.atomic.write %[[VAL_Y_DECLARE]]#1 = %[[ADD]] : !fir.ref<i32>, i32
 !CHECK: }
 !CHECK: return
@@ -88,7 +88,7 @@ subroutine pointers_in_atomic_capture()
 !CHECK: %[[TEMP:.*]] = arith.addi %[[ARG]], %[[VAL_B]] : i32
 !CHECK: omp.yield(%[[TEMP]] : i32)
 !CHECK: }
-!CHECK: omp.atomic.read %[[VAL_B_BOX_ADDR]] = %[[VAL_A_BOX_ADDR]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
+!CHECK: omp.atomic.read %[[VAL_B_BOX_ADDR]] = %[[VAL_A_BOX_ADDR]] : !fir.ptr<i32>, i32
 !CHECK: }
 !CHECK: return
 !CHECK: }
diff --git a/flang/test/Lower/OpenMP/atomic-read.f90 b/flang/test/Lower/OpenMP/atomic-read.f90
index e9bea42252faa3..c3270dd6c1d670 100644
--- a/flang/test/Lower/OpenMP/atomic-read.f90
+++ b/flang/test/Lower/OpenMP/atomic-read.f90
@@ -25,12 +25,12 @@
 !CHECK:    %[[X_DECL:.*]]:2 = hlfir.declare %[[X_REF]] {uniq_name = "_QFEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK:    %[[Y_REF:.*]] = fir.alloca i32 {bindc_name = "y", uniq_name = "_QFEy"}
 !CHECK:    %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y_REF]] {uniq_name = "_QFEy"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-!CHECK:    omp.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1   hint(uncontended) memory_order(acquire) : !fir.ref<i32>, !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[A_DECL]]#1 = %[[B_DECL]]#1   memory_order(relaxed) : !fir.ref<i32>, !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[C_DECL]]#1 = %[[D_DECL]]#1   hint(contended) memory_order(seq_cst) : !fir.ref<!fir.logical<4>>, !fir.ref<!fir.logical<4>>, !fir.logical<4>
-!CHECK:    omp.atomic.read %[[E_DECL]]#1 = %[[F_DECL]]#1   hint(speculative) : !fir.ref<i32>, !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   hint(nonspeculative) : !fir.ref<f32>, !fir.ref<f32>, f32
-!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   : !fir.ref<f32>, !fir.ref<f32>, f32
+!CHECK:    omp.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1   hint(uncontended) memory_order(acquire) : !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[A_DECL]]#1 = %[[B_DECL]]#1   memory_order(relaxed) : !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[C_DECL]]#1 = %[[D_DECL]]#1   hint(contended) memory_order(seq_cst) : !fir.ref<!fir.logical<4>>, !fir.logical<4>
+!CHECK:    omp.atomic.read %[[E_DECL]]#1 = %[[F_DECL]]#1   hint(speculative) : !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   hint(nonspeculative) : !fir.ref<f32>, f32
+!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   : !fir.ref<f32>, f32
 
 program OmpAtomic
 
@@ -68,7 +68,7 @@ end program OmpAtomic
 !CHECK:    %[[X_POINTEE_ADDR:.*]] = fir.box_addr %[[X_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 !CHECK:    %[[Y_ADDR:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 !CHECK:    %[[Y_POINTEE_ADDR:.*]] = fir.box_addr %[[Y_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
-!CHECK:    omp.atomic.read %[[Y_POINTEE_ADDR]] = %[[X_POINTEE_ADDR]]   : !fir.ptr<i32>, !fir.ptr<i32>, i32
+!CHECK:    omp.atomic.read %[[Y_POINTEE_ADDR]] = %[[X_POINTEE_ADDR]]   : !fir.ptr<i32>, i32
 !CHECK:    %[[Y_ADDR:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 !CHECK:    %[[Y_POINTEE_ADDR:.*]] = fir.box_addr %[[Y_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 !CHECK:    %[[Y_POINTEE_VAL:.*]] = fir.load %[[Y_POINTEE_ADDR]] : !fir.ptr<i32>

>From 1d87c40386bf7c3d8bcc6b8705999e2b80c89fde Mon Sep 17 00:00:00 2001
From: Kazuaki Matsumura <kmatsumura at nvidia.com>
Date: Fri, 1 Nov 2024 12:50:33 -0700
Subject: [PATCH 3/5] Revert the changes on OpenACCOps.td and OpenMPOps.td

---
 mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td | 5 +++--
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td   | 5 ++---
 2 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index 1d179c74a970e6..e305e2fbde5b17 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -1958,7 +1958,8 @@ def OpenACC_YieldOp : OpenACC_Op<"yield", [Pure, ReturnLike, Terminator,
 // 2.12 atomic construct
 //===----------------------------------------------------------------------===//
 
-def AtomicReadOp : OpenACC_Op<"atomic.read", [AtomicReadOpInterface]> {
+def AtomicReadOp : OpenACC_Op<"atomic.read", [AllTypesMatch<["x", "v"]>,
+                                              AtomicReadOpInterface]> {
 
   let summary = "performs an atomic read";
 
@@ -1974,7 +1975,7 @@ def AtomicReadOp : OpenACC_Op<"atomic.read", [AtomicReadOpInterface]> {
                        TypeAttr:$element_type);
   let assemblyFormat = [{
     $v `=` $x
-    `:` type($v) `,` type($x) `,` $element_type attr-dict
+    `:` type($x) `,` $element_type attr-dict
   }];
   let hasVerifier = 1;
 }
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 5fd8184fe0e0f7..626539cb7bde42 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1286,7 +1286,7 @@ def TaskwaitOp : OpenMP_Op<"taskwait", clauses = [
 // two-step process.
 
 def AtomicReadOp : OpenMP_Op<"atomic.read", traits = [
-    AtomicReadOpInterface
+    AllTypesMatch<["x", "v"]>, AtomicReadOpInterface
   ], clauses = [
     OpenMP_HintClause, OpenMP_MemoryOrderClause
   ]> {
@@ -1304,8 +1304,7 @@ def AtomicReadOp : OpenMP_Op<"atomic.read", traits = [
 
   // Override clause-based assemblyFormat.
   let assemblyFormat = "$v `=` $x" # clausesReqAssemblyFormat # " oilist(" #
-    clausesOptAssemblyFormat #
-    ") `:` type($v) `,` type($x) `,` $element_type attr-dict";
+    clausesOptAssemblyFormat # ") `:` type($x) `,` $element_type attr-dict";
 
   let extraClassDeclaration = [{
     /// The number of variable operands.

>From bbaf50b26e3a87f48c83a87198907ab0d13382a4 Mon Sep 17 00:00:00 2001
From: Kazuaki Matsumura <kmatsumura at nvidia.com>
Date: Mon, 4 Nov 2024 10:32:04 -0800
Subject: [PATCH 4/5] Push the third approach: assume a converted reference
 type on the LHS of atomic.read

---
 flang/lib/Lower/DirectivesCommon.h            | 127 ++++--------------
 .../test/Lower/OpenACC/acc-atomic-capture.f90 |  14 +-
 flang/test/Lower/OpenACC/acc-atomic-read.f90  |   7 +-
 3 files changed, 30 insertions(+), 118 deletions(-)

diff --git a/flang/lib/Lower/DirectivesCommon.h b/flang/lib/Lower/DirectivesCommon.h
index 8121dec7428040..02a63696ce3603 100644
--- a/flang/lib/Lower/DirectivesCommon.h
+++ b/flang/lib/Lower/DirectivesCommon.h
@@ -139,12 +139,22 @@ static inline void genOmpAccAtomicCaptureStatement(
     mlir::Value toAddress,
     [[maybe_unused]] const AtomicListT *leftHandClauseList,
     [[maybe_unused]] const AtomicListT *rightHandClauseList,
-    mlir::Type elementType, mlir::Location loc) {
+    mlir::Location loc) {
   // Generate `atomic.read` operation for atomic assigment statements
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+  mlir::Type elementType = fir::unwrapRefType(toAddress.getType());
 
   processOmpAtomicTODO<AtomicListT>(elementType, loc);
 
+  if (toAddress.getType() != fromAddress.getType()) {
+    // Allow for implicit type conversion. The `toAddress` operand will have
+    // the same reference type as `fromAddress`, but its value should be
+    // interpreted as `elementType`.
+    mlir::Value convertOp = firOpBuilder.create<fir::ConvertOp>(
+        loc, fromAddress.getType(), toAddress);
+    convertOp.getDefiningOp()->moveAfter(toAddress.getDefiningOp());
+    toAddress = convertOp;
+  }
   if constexpr (std::is_same<AtomicListT,
                              Fortran::parser::OmpAtomicClauseList>()) {
     // If no hint clause is specified, the effect is as if
@@ -207,9 +217,9 @@ static inline void genOmpAccAtomicWriteStatement(
 }
 
 /// Used to generate atomic.update operation which is created in existing
-/// location set by builder. Returns the value that is to be stored to LHS.
+/// location set by builder.
 template <typename AtomicListT>
-static inline mlir::Value genOmpAccAtomicUpdateStatement(
+static inline void genOmpAccAtomicUpdateStatement(
     Fortran::lower::AbstractConverter &converter, mlir::Value lhsAddr,
     mlir::Type varType, const Fortran::parser::Variable &assignmentStmtVariable,
     const Fortran::parser::Expr &assignmentStmtExpr,
@@ -337,14 +347,13 @@ static inline mlir::Value genOmpAccAtomicUpdateStatement(
 
   exprValueOverrides.try_emplace(
       Fortran::semantics::GetExpr(assignmentStmtVariable), val);
-  mlir::Value convertResult;
   {
     // statement context inside the atomic block.
     converter.overrideExprValues(&exprValueOverrides);
     Fortran::lower::StatementContext atomicStmtCtx;
     mlir::Value rhsExpr = fir::getBase(converter.genExprValue(
         *Fortran::semantics::GetExpr(assignmentStmtExpr), atomicStmtCtx));
-    convertResult =
+    mlir::Value convertResult =
         firOpBuilder.createConvert(currentLocation, varType, rhsExpr);
     if constexpr (std::is_same<AtomicListT,
                                Fortran::parser::OmpAtomicClauseList>()) {
@@ -355,7 +364,6 @@ static inline mlir::Value genOmpAccAtomicUpdateStatement(
     converter.resetExprOverrides();
   }
   firOpBuilder.setInsertionPointAfter(atomicUpdateOp);
-  return convertResult;
 }
 
 /// Processes an atomic construct with write clause.
@@ -411,28 +419,12 @@ void genOmpAccAtomicRead(Fortran::lower::AbstractConverter &converter,
   Fortran::lower::StatementContext stmtCtx;
   const Fortran::semantics::SomeExpr &fromExpr =
       *Fortran::semantics::GetExpr(assignmentStmtExpr);
-  mlir::Type elementType = converter.genType(fromExpr);
   mlir::Value fromAddress =
       fir::getBase(converter.genExprAddr(fromExpr, stmtCtx));
   mlir::Value toAddress = fir::getBase(converter.genExprAddr(
       *Fortran::semantics::GetExpr(assignmentStmtVariable), stmtCtx));
-  fir::FirOpBuilder &builder = converter.getFirOpBuilder();
-  mlir::Value LHSTemp = toAddress;
-  if (fromAddress.getType() != toAddress.getType()) {
-    // Create a temp for implicit type conversion.
-    LHSTemp = builder.create<fir::AllocaOp>(
-        loc, fir::unwrapRefType(fromAddress.getType()));
-  }
-  genOmpAccAtomicCaptureStatement(converter, fromAddress, LHSTemp,
-                                  leftHandClauseList, rightHandClauseList,
-                                  elementType, loc);
-  if (fromAddress.getType() != toAddress.getType()) {
-    // Copy the converted value of the temp to the original read LHS.
-    mlir::Value tempVal = builder.create<fir::LoadOp>(loc, LHSTemp);
-    mlir::Value convVal = builder.create<fir::ConvertOp>(
-        loc, fir::unwrapRefType(toAddress.getType()), tempVal);
-    builder.create<fir::StoreOp>(loc, convVal, toAddress);
-  }
+  genOmpAccAtomicCaptureStatement(converter, fromAddress, toAddress,
+                                  leftHandClauseList, rightHandClauseList, loc);
 }
 
 /// Processes an atomic construct with update clause.
@@ -491,41 +483,6 @@ void genOmpAtomic(Fortran::lower::AbstractConverter &converter,
       &atomicClauseList, nullptr, loc);
 }
 
-// Replace every load of `addr` found in `expr` with a load of `newAddr`.
-static inline void
-rewriteLoadDependency(Fortran::lower::AbstractConverter &converter,
-                      mlir::Value expr, mlir::Value addr, mlir::Value newAddr) {
-  fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  std::set<mlir::Operation *> dependency;
-  dependency.insert(expr.getDefiningOp());
-  while (!dependency.empty()) {
-    mlir::Operation *dep = dependency.extract(dependency.begin()).value();
-    if (!newAddr.getDefiningOp()->getParentOp()->isAncestor(dep)) {
-      // Skip it if it falls outside the parent region of the capture.
-      continue;
-    }
-    if (auto load = mlir::dyn_cast_or_null<fir::LoadOp>(dep)) {
-      if (load.getMemref().getDefiningOp() == addr.getDefiningOp()) {
-        // Found a load of `addr`. Replace it with a load of `newAddr` while
-        // converting the value.
-        firOpBuilder.setInsertionPoint(load);
-        mlir::Location loc = load.getLoc();
-        mlir::Value newLoad = firOpBuilder.create<fir::LoadOp>(loc, newAddr);
-        mlir::Value conv =
-            firOpBuilder.create<fir::ConvertOp>(loc, load.getType(), newLoad);
-        load.replaceAllUsesWith(conv);
-        load.erase();
-      }
-      continue;
-    }
-    for (mlir::Value operand : dep->getOperands()) {
-      if (operand.getDefiningOp()) {
-        dependency.insert(operand.getDefiningOp());
-      }
-    }
-  }
-}
-
 /// Processes an atomic construct with capture clause.
 template <typename AtomicT, typename AtomicListT>
 void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
@@ -580,33 +537,14 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
   firOpBuilder.createBlock(&(atomicCaptureOp->getRegion(0)));
   mlir::Block &block = atomicCaptureOp->getRegion(0).back();
   firOpBuilder.setInsertionPointToStart(&block);
-  mlir::Value captureLHS, storeLHS, LHSTemp;
-  if (Fortran::semantics::checkForSingleVariableOnRHS(stmt1)) {
-    captureLHS = stmt1LHSArg;
-    storeLHS = stmt2LHSArg;
-  } else {
-    captureLHS = stmt2LHSArg;
-    storeLHS = stmt1LHSArg;
-  }
-  LHSTemp = captureLHS;
-  if (captureLHS.getType() != storeLHS.getType()) {
-    // Create a temp for implicit type conversion.
-    LHSTemp = firOpBuilder.create<fir::AllocaOp>(
-        loc, fir::unwrapRefType(storeLHS.getType()));
-    LHSTemp.getDefiningOp()->moveBefore(atomicCaptureOp);
-  }
   if (Fortran::semantics::checkForSingleVariableOnRHS(stmt1)) {
-    mlir::Value storeValue;
     if (Fortran::semantics::checkForSymbolMatch(stmt2)) {
       // Atomic capture construct is of the form [capture-stmt, update-stmt]
-      const Fortran::semantics::SomeExpr &fromExpr =
-          *Fortran::semantics::GetExpr(stmt1Expr);
-      mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
-          converter, stmt2LHSArg, LHSTemp,
+          converter, stmt2LHSArg, stmt1LHSArg,
           /*leftHandClauseList=*/nullptr,
-          /*rightHandClauseList=*/nullptr, elementType, loc);
-      storeValue = genOmpAccAtomicUpdateStatement<AtomicListT>(
+          /*rightHandClauseList=*/nullptr, loc);
+      genOmpAccAtomicUpdateStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt2VarType, stmt2Var, stmt2Expr,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, loc, atomicCaptureOp);
@@ -616,44 +554,25 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
       mlir::Value stmt2RHSArg =
           fir::getBase(converter.genExprValue(assign2.rhs, stmtCtx));
       firOpBuilder.setInsertionPointToStart(&block);
-      const Fortran::semantics::SomeExpr &fromExpr =
-          *Fortran::semantics::GetExpr(stmt1Expr);
-      mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
-          converter, stmt2LHSArg, LHSTemp,
+          converter, stmt2LHSArg, stmt1LHSArg,
           /*leftHandClauseList=*/nullptr,
-          /*rightHandClauseList=*/nullptr, elementType, loc);
+          /*rightHandClauseList=*/nullptr, loc);
       genOmpAccAtomicWriteStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt2RHSArg,
           /*leftHandClauseList=*/nullptr,
           /*rightHandClauseList=*/nullptr, loc);
-      storeValue = stmt2RHSArg;
     }
-    if (captureLHS.getType() != storeLHS.getType())
-      // Replace every load of the capture LHS in the store value with the
-      // corresponding temp.
-      rewriteLoadDependency(converter, storeValue, captureLHS, LHSTemp);
   } else {
     // Atomic capture construct is of the form [update-stmt, capture-stmt]
-    const Fortran::semantics::SomeExpr &fromExpr =
-        *Fortran::semantics::GetExpr(stmt2Expr);
-    mlir::Type elementType = converter.genType(fromExpr);
     genOmpAccAtomicUpdateStatement<AtomicListT>(
         converter, stmt1LHSArg, stmt1VarType, stmt1Var, stmt1Expr,
         /*leftHandClauseList=*/nullptr,
         /*rightHandClauseList=*/nullptr, loc, atomicCaptureOp);
     genOmpAccAtomicCaptureStatement<AtomicListT>(
-        converter, stmt1LHSArg, LHSTemp,
+        converter, stmt1LHSArg, stmt2LHSArg,
         /*leftHandClauseList=*/nullptr,
-        /*rightHandClauseList=*/nullptr, elementType, loc);
-  }
-  if (captureLHS.getType() != storeLHS.getType()) {
-    // Copy the converted value of the temp to the original capture LHS.
-    firOpBuilder.setInsertionPointAfter(atomicCaptureOp);
-    mlir::Value tempVal = firOpBuilder.create<fir::LoadOp>(loc, LHSTemp);
-    mlir::Value convVal = firOpBuilder.create<fir::ConvertOp>(
-        loc, fir::unwrapRefType(captureLHS.getType()), tempVal);
-    firOpBuilder.create<fir::StoreOp>(loc, convVal, captureLHS);
+        /*rightHandClauseList=*/nullptr, loc);
   }
   firOpBuilder.setInsertionPointToEnd(&block);
   if constexpr (std::is_same<AtomicListT,
diff --git a/flang/test/Lower/OpenACC/acc-atomic-capture.f90 b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
index 08f866711ecd8f..51700717f7f710 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-capture.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
@@ -138,21 +138,17 @@ end subroutine capture_with_convert_i32_to_f64
 ! CHECK: %[[V_DECL:.*]]:2 = hlfir.declare %[[V]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ev"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: %[[X:.*]] = fir.alloca f64 {bindc_name = "x", uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"}
 ! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"} : (!fir.ref<f64>) -> (!fir.ref<f64>, !fir.ref<f64>)
+! CHECK: %[[X_CONV:.*]] = fir.convert %[[X_DECL]]#1 : (!fir.ref<f64>) -> !fir.ref<i32>
 ! CHECK: %[[CST:.*]] = arith.constant 1.000000e+00 : f64
 ! CHECK: hlfir.assign %[[CST]] to %[[X_DECL]]#0 : f64, !fir.ref<f64>
 ! CHECK: %c0_i32 = arith.constant 0 : i32
 ! CHECK: hlfir.assign %c0_i32 to %[[V_DECL]]#0 : i32, !fir.ref<i32>
-! CHECK: %[[ALLOCA:.*]] = fir.alloca f64
-! CHECK: %[[LOAD1:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f64>
-! CHECK: %[[CONV1:.*]] = fir.convert %[[LOAD1]] : (f64) -> i32
-! CHECK: %[[CONV2:.*]] = fir.convert %[[CONV1]] : (i32) -> f64
+! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
+! CHECK: %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f64
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<f64>, f64
-! CHECK:   acc.atomic.write %[[X_DECL]]#1 = %[[CONV2]] : !fir.ref<f64>, f64
+! CHECK:   acc.atomic.read %[[V_DECL]] = %[[X_CONV]]#1 : !fir.ref<f64>, i32
+! CHECK:   acc.atomic.write %[[X_DECL]]#1 = %[[CONV]] : !fir.ref<f64>, f64
 ! CHECK: }
-! CHECK: %[[LOAD2:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f64>
-! CHECK: %[[CONV3:.*]] = fir.convert %[[LOAD2]] : (f64) -> i32
-! CHECK: fir.store %[[CONV3]] to %[[V_DECL]]#1 : !fir.ref<i32>
 
 subroutine capture_with_convert_f64_to_i32()
   integer :: x
diff --git a/flang/test/Lower/OpenACC/acc-atomic-read.f90 b/flang/test/Lower/OpenACC/acc-atomic-read.f90
index 5491521623b9b6..33cf17c21f0495 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-read.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-read.f90
@@ -53,10 +53,7 @@ subroutine atomic_read_with_convert()
 ! CHECK-LABEL: func.func @_QPatomic_read_with_convert() {
 ! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_convertEx"}
 ! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_convertEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[X_CONV:.*]] = fir.convert %[[X_DECL]]#1 : (!fir.ref<i32>) -> !fir.ref<i64>
 ! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_convertEy"}
 ! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_convertEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
-! CHECK: %[[ALLOCA:.*]] = fir.alloca i32
-! CHECK: acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<i32>, i32
-! CHECK: %[[LOAD:.*]] = fir.load %[[ALLOCA]] : !fir.ref<i32>
-! CHECK: %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> i64
-! CHECK: fir.store %[[CONV]] to %[[Y_DECL]]#1 : !fir.ref<i64>
+! CHECK: acc.atomic.read %[[Y_DECL]]#1 = %[[X_CONV]] : !fir.ref<i64>, i32

>From 16e63ef6914d857eb0c68769d4b3e953dd6d2dff Mon Sep 17 00:00:00 2001
From: Kazuaki Matsumura <kmatsumura at nvidia.com>
Date: Mon, 4 Nov 2024 11:20:00 -0800
Subject: [PATCH 5/5] Revert commits up to the first approach

---
 flang/lib/Lower/DirectivesCommon.h            |  31 +++---
 .../Fir/convert-to-llvm-openmp-and-fir.fir    |   4 +-
 .../test/Lower/OpenACC/acc-atomic-capture.f90 |  48 ++++----
 flang/test/Lower/OpenACC/acc-atomic-read.f90  |  19 ++--
 .../Lower/OpenACC/acc-atomic-update-array.f90 |   4 +-
 flang/test/Lower/OpenMP/atomic-capture.f90    |   6 +-
 flang/test/Lower/OpenMP/atomic-read.f90       |  14 +--
 .../mlir/Dialect/OpenACC/OpenACCOps.td        |   5 +-
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td |   5 +-
 .../OpenMPToLLVM/convert-to-llvmir.mlir       |   2 +-
 mlir/test/Dialect/OpenACC/invalid.mlir        |  14 +--
 mlir/test/Dialect/OpenACC/ops.mlir            |  16 +--
 mlir/test/Dialect/OpenMP/invalid.mlir         |  40 +++----
 mlir/test/Dialect/OpenMP/ops.mlir             | 104 +++++++++---------
 .../Target/LLVMIR/openmp-llvm-invalid.mlir    |   4 +-
 mlir/test/Target/LLVMIR/openmp-llvm.mlir      |  98 ++++++++---------
 16 files changed, 202 insertions(+), 212 deletions(-)

diff --git a/flang/lib/Lower/DirectivesCommon.h b/flang/lib/Lower/DirectivesCommon.h
index 02a63696ce3603..88514b16743278 100644
--- a/flang/lib/Lower/DirectivesCommon.h
+++ b/flang/lib/Lower/DirectivesCommon.h
@@ -139,22 +139,12 @@ static inline void genOmpAccAtomicCaptureStatement(
     mlir::Value toAddress,
     [[maybe_unused]] const AtomicListT *leftHandClauseList,
     [[maybe_unused]] const AtomicListT *rightHandClauseList,
-    mlir::Location loc) {
+    mlir::Type elementType, mlir::Location loc) {
   // Generate `atomic.read` operation for atomic assigment statements
   fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
-  mlir::Type elementType = fir::unwrapRefType(toAddress.getType());
 
   processOmpAtomicTODO<AtomicListT>(elementType, loc);
 
-  if (toAddress.getType() != fromAddress.getType()) {
-    // Allow for implicit type conversion. The `toAddress` operand will have
-    // the same reference type as `fromAddress`, but its value should be
-    // interpreted as `elementType`.
-    mlir::Value convertOp = firOpBuilder.create<fir::ConvertOp>(
-        loc, fromAddress.getType(), toAddress);
-    convertOp.getDefiningOp()->moveAfter(toAddress.getDefiningOp());
-    toAddress = convertOp;
-  }
   if constexpr (std::is_same<AtomicListT,
                              Fortran::parser::OmpAtomicClauseList>()) {
     // If no hint clause is specified, the effect is as if
@@ -419,12 +409,14 @@ void genOmpAccAtomicRead(Fortran::lower::AbstractConverter &converter,
   Fortran::lower::StatementContext stmtCtx;
   const Fortran::semantics::SomeExpr &fromExpr =
       *Fortran::semantics::GetExpr(assignmentStmtExpr);
+  mlir::Type elementType = converter.genType(fromExpr);
   mlir::Value fromAddress =
       fir::getBase(converter.genExprAddr(fromExpr, stmtCtx));
   mlir::Value toAddress = fir::getBase(converter.genExprAddr(
       *Fortran::semantics::GetExpr(assignmentStmtVariable), stmtCtx));
   genOmpAccAtomicCaptureStatement(converter, fromAddress, toAddress,
-                                  leftHandClauseList, rightHandClauseList, loc);
+                                  leftHandClauseList, rightHandClauseList,
+                                  elementType, loc);
 }
 
 /// Processes an atomic construct with update clause.
@@ -540,10 +532,13 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
   if (Fortran::semantics::checkForSingleVariableOnRHS(stmt1)) {
     if (Fortran::semantics::checkForSymbolMatch(stmt2)) {
       // Atomic capture construct is of the form [capture-stmt, update-stmt]
+      const Fortran::semantics::SomeExpr &fromExpr =
+          *Fortran::semantics::GetExpr(stmt1Expr);
+      mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt1LHSArg,
           /*leftHandClauseList=*/nullptr,
-          /*rightHandClauseList=*/nullptr, loc);
+          /*rightHandClauseList=*/nullptr, elementType, loc);
       genOmpAccAtomicUpdateStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt2VarType, stmt2Var, stmt2Expr,
           /*leftHandClauseList=*/nullptr,
@@ -554,10 +549,13 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
       mlir::Value stmt2RHSArg =
           fir::getBase(converter.genExprValue(assign2.rhs, stmtCtx));
       firOpBuilder.setInsertionPointToStart(&block);
+      const Fortran::semantics::SomeExpr &fromExpr =
+          *Fortran::semantics::GetExpr(stmt1Expr);
+      mlir::Type elementType = converter.genType(fromExpr);
       genOmpAccAtomicCaptureStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt1LHSArg,
           /*leftHandClauseList=*/nullptr,
-          /*rightHandClauseList=*/nullptr, loc);
+          /*rightHandClauseList=*/nullptr, elementType, loc);
       genOmpAccAtomicWriteStatement<AtomicListT>(
           converter, stmt2LHSArg, stmt2RHSArg,
           /*leftHandClauseList=*/nullptr,
@@ -565,6 +563,9 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
     }
   } else {
     // Atomic capture construct is of the form [update-stmt, capture-stmt]
+    const Fortran::semantics::SomeExpr &fromExpr =
+        *Fortran::semantics::GetExpr(stmt2Expr);
+    mlir::Type elementType = converter.genType(fromExpr);
     genOmpAccAtomicUpdateStatement<AtomicListT>(
         converter, stmt1LHSArg, stmt1VarType, stmt1Var, stmt1Expr,
         /*leftHandClauseList=*/nullptr,
@@ -572,7 +573,7 @@ void genOmpAccAtomicCapture(Fortran::lower::AbstractConverter &converter,
     genOmpAccAtomicCaptureStatement<AtomicListT>(
         converter, stmt1LHSArg, stmt2LHSArg,
         /*leftHandClauseList=*/nullptr,
-        /*rightHandClauseList=*/nullptr, loc);
+        /*rightHandClauseList=*/nullptr, elementType, loc);
   }
   firOpBuilder.setInsertionPointToEnd(&block);
   if constexpr (std::is_same<AtomicListT,
diff --git a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
index 168526518865b4..184abe24fe967d 100644
--- a/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
+++ b/flang/test/Fir/convert-to-llvm-openmp-and-fir.fir
@@ -781,11 +781,11 @@ func.func @_QPsimple_reduction(%arg0: !fir.ref<!fir.array<100x!fir.logical<4>>>
 // -----
 
 // CHECK: llvm.func @_QPs
-// CHECK: omp.atomic.read %{{.*}} = %{{.*}}   : !llvm.ptr, !llvm.struct<(f32, f32)>
+// CHECK: omp.atomic.read %{{.*}} = %{{.*}}   : !llvm.ptr, !llvm.ptr, !llvm.struct<(f32, f32)>
 
 func.func @_QPs(%arg0: !fir.ref<complex<f32>> {fir.bindc_name = "x"}) {
   %0 = fir.alloca complex<f32> {bindc_name = "v", uniq_name = "_QFsEv"}
-  omp.atomic.read %0 = %arg0   : !fir.ref<complex<f32>>, complex<f32>
+  omp.atomic.read %0 = %arg0   : !fir.ref<complex<f32>>, !fir.ref<complex<f32>>, complex<f32>
   return
 }
 
diff --git a/flang/test/Lower/OpenACC/acc-atomic-capture.f90 b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
index 51700717f7f710..797d322ca7ef13 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-capture.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-capture.f90
@@ -11,7 +11,7 @@ program acc_atomic_capture_test
 !CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %2 {uniq_name = "_QFEy"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK: %[[temp:.*]] = fir.load %[[X_DECL]]#0 : !fir.ref<i32>
 !CHECK: acc.atomic.capture {
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, i32
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: acc.atomic.update %[[Y_DECL]]#1 : !fir.ref<i32> {
 !CHECK: ^bb0(%[[ARG:.*]]: i32):
 !CHECK: %[[result:.*]] = arith.addi %[[temp]], %[[ARG]] : i32
@@ -32,7 +32,7 @@ program acc_atomic_capture_test
 !CHECK: %[[result:.*]] = arith.muli %[[temp]], %[[ARG]] : i32
 !CHECK: acc.yield %[[result]] : i32
 !CHECK: }
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, i32
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: }
 
     !$acc atomic capture
@@ -47,7 +47,7 @@ program acc_atomic_capture_test
 !CHECK: %[[result_noreassoc:.*]] = hlfir.no_reassoc %[[result]] : i32
 !CHECK: %[[result:.*]] = arith.addi %[[constant_20]], %[[result_noreassoc]] : i32
 !CHECK: acc.atomic.capture {
-!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, i32
+!CHECK: acc.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: acc.atomic.write %[[Y_DECL]]#1 = %[[result]] : !fir.ref<i32>, i32
 !CHECK: }
 
@@ -82,7 +82,7 @@ subroutine pointers_in_atomic_capture()
 !CHECK: %[[result:.*]] = arith.addi %[[ARG]], %[[loaded_value]] : i32
 !CHECK: acc.yield %[[result]] : i32
 !CHECK: }
-!CHECK: acc.atomic.read %[[loaded_B_addr]] = %[[loaded_A_addr]] : !fir.ptr<i32>, i32
+!CHECK: acc.atomic.read %[[loaded_B_addr]] = %[[loaded_A_addr]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
 !CHECK: }
     integer, pointer :: a, b
     integer, target :: c, d
@@ -118,7 +118,7 @@ subroutine capture_with_convert_f32_to_i32()
 ! CHECK: %[[MUL:.*]] = arith.mulf %{{.*}}, %[[CST]] fastmath<contract> : f32
 ! CHECK: %[[CONV:.*]] = fir.convert %[[MUL]] : (f32) -> i32
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[K_DECL]]#1 : !fir.ref<i32>, i32
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[K_DECL]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:   acc.atomic.write %[[K_DECL]]#1 = %[[CONV]] : !fir.ref<i32>, i32
 ! CHECK: }
 
@@ -138,7 +138,6 @@ end subroutine capture_with_convert_i32_to_f64
 ! CHECK: %[[V_DECL:.*]]:2 = hlfir.declare %[[V]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ev"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 ! CHECK: %[[X:.*]] = fir.alloca f64 {bindc_name = "x", uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"}
 ! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcapture_with_convert_i32_to_f64Ex"} : (!fir.ref<f64>) -> (!fir.ref<f64>, !fir.ref<f64>)
-! CHECK: %[[X_CONV:.*]] = fir.convert %[[X_DECL]]#1 : (!fir.ref<f64>) -> !fir.ref<i32>
 ! CHECK: %[[CST:.*]] = arith.constant 1.000000e+00 : f64
 ! CHECK: hlfir.assign %[[CST]] to %[[X_DECL]]#0 : f64, !fir.ref<f64>
 ! CHECK: %c0_i32 = arith.constant 0 : i32
@@ -146,7 +145,7 @@ end subroutine capture_with_convert_i32_to_f64
 ! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
 ! CHECK: %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f64
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[V_DECL]] = %[[X_CONV]]#1 : !fir.ref<f64>, i32
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i32>, !fir.ref<f64>, f64
 ! CHECK:   acc.atomic.write %[[X_DECL]]#1 = %[[CONV]] : !fir.ref<f64>, f64
 ! CHECK: }
 
@@ -170,20 +169,16 @@ end subroutine capture_with_convert_f64_to_i32
 ! CHECK: hlfir.assign %c1_i32 to %[[X_DECL]]#0 : i32, !fir.ref<i32>
 ! CHECK: %[[CST:.*]] = arith.constant 0.000000e+00 : f64
 ! CHECK: hlfir.assign %[[CST]] to %[[V_DECL]]#0 : f64, !fir.ref<f64>
-! CHECK: %[[ALLOCA]] = fir.alloca i32
-! CHECK: %[[LOAD1:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<f64>
+! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<f64>
 ! CHECK: acc.atomic.capture {
 ! CHECK:   acc.atomic.update %[[X_DECL]]#1 : !fir.ref<i32> {
 ! CHECK:   ^bb0(%arg0: i32):
-! CHECK:     %[[MUL:.*]] = arith.mulf %[[LOAD1]], %[[LOAD1]] fastmath<contract> : f64
-! CHECK:     %[[CONV1:.*]] = fir.convert %[[MUL]] : (f64) -> i32
-! CHECK:     acc.yield %[[CONV1]] : i32
+! CHECK:     %[[MUL:.*]] = arith.mulf %[[LOAD]], %[[LOAD]] fastmath<contract> : f64
+! CHECK:     %[[CONV:.*]] = fir.convert %[[MUL]] : (f64) -> i32
+! CHECK:     acc.yield %[[CONV]] : i32
 ! CHECK:   }
-! CHECK:   acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<i32>, i32
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<f64>, !fir.ref<i32>, i32
 ! CHECK: }
-! CHECK: %[[LOAD2:.*]] = fir.load %[[ALLOCA]] : !fir.ref<i32>
-! CHECK: %[[CONV2:.*]] = fir.convert %[[LOAD2]] : (i32) -> f64
-! CHECK: fir.store %[[CONV2]] to %[[V_DECL]]#1 : !fir.ref<f64>
 
 subroutine capture_with_convert_i32_to_f32()
   real(4) :: x
@@ -205,21 +200,16 @@ end subroutine capture_with_convert_i32_to_f32
 ! CHECK: hlfir.assign %[[CST]] to %[[X_DECL]]#0 : f32, !fir.ref<f32>
 ! CHECK: %c0_i32 = arith.constant 0 : i32
 ! CHECK: hlfir.assign %c0_i32 to %[[V_DECL]]#0 : i32, !fir.ref<i32>
-! CHECK: %[[ALLOCA:.*]] = fir.alloca f32
-! CHECK: %[[LOAD1:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f32>
-! CHECK: %[[CONV1:.*]] = fir.convert %5 : (f32) -> i32
+! CHECK: %[[LOAD:.*]] = fir.load %[[V_DECL]]#0 : !fir.ref<i32>
 ! CHECK: acc.atomic.capture {
-! CHECK:   acc.atomic.read %[[ALLOCA]] = %[[X_DECL]]#1 : !fir.ref<f32>, f32
+! CHECK:   acc.atomic.read %[[V_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i32>, !fir.ref<f32>, f32
 ! CHECK:   acc.atomic.update %[[X_DECL]]#1 : !fir.ref<f32> {
 ! CHECK:   ^bb0(%arg0: f32):
-! CHECK:     %[[CONV2:.*]] = fir.convert %[[CONV1]] : (i32) -> f32
-! CHECK:     %[[ADD:.*]] = arith.addf %arg0, %[[CONV2]] fastmath<contract> : f32
+! CHECK:     %[[CONV:.*]] = fir.convert %[[LOAD]] : (i32) -> f32
+! CHECK:     %[[ADD:.*]] = arith.addf %arg0, %[[CONV]] fastmath<contract> : f32
 ! CHECK:     acc.yield %[[ADD]] : f32
 ! CHECK:   }
 ! CHECK: }
-! CHECK: %[[LOAD2:.*]] = fir.load %[[ALLOCA]] : !fir.ref<f32>
-! CHECK: %[[CONV3:.*]] = fir.convert %[[LOAD2]] : (f32) -> i32
-! CHECK: fir.store %[[CONV3]] to %[[V_DECL]]#1 : !fir.ref<i32>
 
 subroutine array_ref_in_atomic_capture1
   integer :: x(10), v
@@ -235,7 +225,7 @@ end subroutine array_ref_in_atomic_capture1
 ! CHECK:           %[[X_DECL:.*]]:2 = hlfir.declare %[[X]](%{{.*}}) {uniq_name = "_QFarray_ref_in_atomic_capture1Ex"} : (!fir.ref<!fir.array<10xi32>>, !fir.shape<1>) -> (!fir.ref<!fir.array<10xi32>>, !fir.ref<!fir.array<10xi32>>)
 ! CHECK:           %[[X_REF:.*]] = hlfir.designate %[[X_DECL]]#0 (%{{.*}})  : (!fir.ref<!fir.array<10xi32>>, index) -> !fir.ref<i32>
 ! CHECK:           acc.atomic.capture {
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:             acc.atomic.update %[[X_REF]] : !fir.ref<i32> {
 ! CHECK:             ^bb0(%[[VAL_7:.*]]: i32):
 ! CHECK:               %[[VAL_8:.*]] = arith.addi %[[VAL_7]], %{{.*}} : i32
@@ -262,7 +252,7 @@ end subroutine array_ref_in_atomic_capture2
 ! CHECK:               %[[VAL_8:.*]] = arith.addi %[[VAL_7]], %{{.*}} : i32
 ! CHECK:               acc.yield %[[VAL_8]] : i32
 ! CHECK:             }
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[X_REF]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:           }
 
 subroutine comp_ref_in_atomic_capture1
@@ -283,7 +273,7 @@ end subroutine comp_ref_in_atomic_capture1
 ! CHECK:           %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFcomp_ref_in_atomic_capture1Ex"} : (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>) -> (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>, !fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>)
 ! CHECK:           %[[C:.*]] = hlfir.designate %[[X_DECL]]#0{"c"}   : (!fir.ref<!fir.type<_QFcomp_ref_in_atomic_capture1Tt1{c:i32}>>) -> !fir.ref<i32>
 ! CHECK:           acc.atomic.capture {
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:             acc.atomic.update %[[C]] : !fir.ref<i32> {
 ! CHECK:             ^bb0(%[[VAL_5:.*]]: i32):
 ! CHECK:               %[[VAL_6:.*]] = arith.addi %[[VAL_5]], %{{.*}} : i32
@@ -314,5 +304,5 @@ end subroutine comp_ref_in_atomic_capture2
 ! CHECK:               %[[VAL_6:.*]] = arith.addi %[[VAL_5]], %{{.*}} : i32
 ! CHECK:               acc.yield %[[VAL_6]] : i32
 ! CHECK:             }
-! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, i32
+! CHECK:             acc.atomic.read %[[V_DECL]]#1 = %[[C]] : !fir.ref<i32>, !fir.ref<i32>, i32
 ! CHECK:           }
diff --git a/flang/test/Lower/OpenACC/acc-atomic-read.f90 b/flang/test/Lower/OpenACC/acc-atomic-read.f90
index 33cf17c21f0495..f2cbe6e45596a4 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-read.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-read.f90
@@ -13,7 +13,7 @@ end program acc_atomic_test
 ! CHECK: %[[G_DECL:.*]]:2 = hlfir.declare %[[VAR_G]] {uniq_name = "_QFEg"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
 ! CHECK: %[[VAR_H:.*]] = fir.alloca f32 {bindc_name = "h", uniq_name = "_QFEh"}
 ! CHECK: %[[H_DECL:.*]]:2 = hlfir.declare %[[VAR_H]] {uniq_name = "_QFEh"} : (!fir.ref<f32>) -> (!fir.ref<f32>, !fir.ref<f32>)
-! CHECK: acc.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1 : !fir.ref<f32>, f32
+! CHECK: acc.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1 : !fir.ref<f32>, !fir.ref<f32>, f32
 ! CHECK: return
 ! CHECK: }
 
@@ -39,10 +39,10 @@ subroutine atomic_read_pointer()
 ! CHECK:   %[[BOX_ADDR_X:.*]] = fir.box_addr %[[LOAD_X]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 ! CHECK:   %[[LOAD_Y:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 ! CHECK:   %[[BOX_ADDR_Y:.*]] = fir.box_addr %[[LOAD_Y]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
-! CHECK:   acc.atomic.read %[[BOX_ADDR_Y]] = %[[BOX_ADDR_X]] : !fir.ptr<i32>, i32
+! CHECK:   acc.atomic.read %[[BOX_ADDR_Y]] = %[[BOX_ADDR_X]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
 ! CHECK: }
 
-subroutine atomic_read_with_convert()
+subroutine atomic_read_with_cast()
   integer(4) :: x
   integer(8) :: y
 
@@ -50,10 +50,9 @@ subroutine atomic_read_with_convert()
   y = x
 end
 
-! CHECK-LABEL: func.func @_QPatomic_read_with_convert() {
-! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_convertEx"}
-! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_convertEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-! CHECK: %[[X_CONV:.*]] = fir.convert %[[X_DECL]]#1 : (!fir.ref<i32>) -> !fir.ref<i64>
-! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_convertEy"}
-! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_convertEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
-! CHECK: acc.atomic.read %[[Y_DECL]]#1 = %[[X_CONV]] : !fir.ref<i64>, i32
+! CHECK-LABEL: func.func @_QPatomic_read_with_cast() {
+! CHECK: %[[X:.*]] = fir.alloca i32 {bindc_name = "x", uniq_name = "_QFatomic_read_with_castEx"}
+! CHECK: %[[X_DECL:.*]]:2 = hlfir.declare %[[X]] {uniq_name = "_QFatomic_read_with_castEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
+! CHECK: %[[Y:.*]] = fir.alloca i64 {bindc_name = "y", uniq_name = "_QFatomic_read_with_castEy"}
+! CHECK: %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y]] {uniq_name = "_QFatomic_read_with_castEy"} : (!fir.ref<i64>) -> (!fir.ref<i64>, !fir.ref<i64>)
+! CHECK: acc.atomic.read %[[Y_DECL]]#1 = %[[X_DECL]]#1 : !fir.ref<i64>, !fir.ref<i32>, i32
diff --git a/flang/test/Lower/OpenACC/acc-atomic-update-array.f90 b/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
index eeb7ea29940862..f89a9ab457d499 100644
--- a/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
+++ b/flang/test/Lower/OpenACC/acc-atomic-update-array.f90
@@ -45,7 +45,7 @@ subroutine atomic_read_array1(r, n, x)
 ! CHECK: %[[DECL_X:.*]]:2 = hlfir.declare %[[ARG2]] dummy_scope %{{[0-9]+}} {uniq_name = "_QFatomic_read_array1Ex"} : (!fir.ref<f32>, !fir.dscope) -> (!fir.ref<f32>, !fir.ref<f32>)
 ! CHECK: %[[DECL_R:.*]]:2 = hlfir.declare %[[ARG0]](%{{.*}}) dummy_scope %{{[0-9]+}} {uniq_name = "_QFatomic_read_array1Er"} : (!fir.ref<!fir.array<?xf32>>, !fir.shape<1>, !fir.dscope) -> (!fir.box<!fir.array<?xf32>>, !fir.ref<!fir.array<?xf32>>)
 ! CHECK: %[[DES:.*]] = hlfir.designate %[[DECL_R]]#0 (%{{.*}})  : (!fir.box<!fir.array<?xf32>>, i64) -> !fir.ref<f32>
-! CHECK: acc.atomic.read %[[DECL_X]]#1 = %[[DES]] : !fir.ref<f32>, f32
+! CHECK: acc.atomic.read %[[DECL_X]]#1 = %[[DES]] : !fir.ref<f32>, !fir.ref<f32>, f32
 
 subroutine atomic_write_array1(r, n, x)
   implicit none
@@ -88,5 +88,5 @@ subroutine atomic_capture_array1(r, n, x, y)
 ! CHECK:     %[[ADD:.*]] = arith.addf %[[ARG]], %[[LOAD]] fastmath<contract> : f32
 ! CHECK:     acc.yield %[[ADD]] : f32
 ! CHECK:   }
-! CHECK:   acc.atomic.read %[[DECL_Y]]#1 = %[[R_I]] : !fir.ref<f32>, f32
+! CHECK:   acc.atomic.read %[[DECL_Y]]#1 = %[[R_I]] : !fir.ref<f32>, !fir.ref<f32>, f32
 ! CHECK: }
diff --git a/flang/test/Lower/OpenMP/atomic-capture.f90 b/flang/test/Lower/OpenMP/atomic-capture.f90
index af82e4b2a20eb2..679d22d3d7063e 100644
--- a/flang/test/Lower/OpenMP/atomic-capture.f90
+++ b/flang/test/Lower/OpenMP/atomic-capture.f90
@@ -22,7 +22,7 @@ program OmpAtomicCapture
 !CHECK: %[[TEMP:.*]] = arith.muli %[[VAL_Y_LOADED]], %[[ARG]] : i32
 !CHECK: omp.yield(%[[TEMP]] : i32)
 !CHECK: }
-!CHECK: omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, i32
+!CHECK: omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK: }
     !$omp atomic hint(omp_sync_hint_uncontended) capture
         y = x * y 
@@ -36,7 +36,7 @@ program OmpAtomicCapture
 !CHECK: %[[NO_REASSOC:.*]] = hlfir.no_reassoc %[[SUB]] : i32
 !CHECK: %[[ADD:.*]] = arith.addi  %[[VAL_20]], %[[NO_REASSOC]] : i32
 !CHECK: omp.atomic.capture hint(nonspeculative) memory_order(acquire) {
-!CHECK:   omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, i32
+!CHECK:   omp.atomic.read %[[VAL_X_DECLARE]]#1 = %[[VAL_Y_DECLARE]]#1 : !fir.ref<i32>, !fir.ref<i32>, i32
 !CHECK:   omp.atomic.write %[[VAL_Y_DECLARE]]#1 = %[[ADD]] : !fir.ref<i32>, i32
 !CHECK: }
 !CHECK: return
@@ -88,7 +88,7 @@ subroutine pointers_in_atomic_capture()
 !CHECK: %[[TEMP:.*]] = arith.addi %[[ARG]], %[[VAL_B]] : i32
 !CHECK: omp.yield(%[[TEMP]] : i32)
 !CHECK: }
-!CHECK: omp.atomic.read %[[VAL_B_BOX_ADDR]] = %[[VAL_A_BOX_ADDR]] : !fir.ptr<i32>, i32
+!CHECK: omp.atomic.read %[[VAL_B_BOX_ADDR]] = %[[VAL_A_BOX_ADDR]] : !fir.ptr<i32>, !fir.ptr<i32>, i32
 !CHECK: }
 !CHECK: return
 !CHECK: }
diff --git a/flang/test/Lower/OpenMP/atomic-read.f90 b/flang/test/Lower/OpenMP/atomic-read.f90
index c3270dd6c1d670..e9bea42252faa3 100644
--- a/flang/test/Lower/OpenMP/atomic-read.f90
+++ b/flang/test/Lower/OpenMP/atomic-read.f90
@@ -25,12 +25,12 @@
 !CHECK:    %[[X_DECL:.*]]:2 = hlfir.declare %[[X_REF]] {uniq_name = "_QFEx"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
 !CHECK:    %[[Y_REF:.*]] = fir.alloca i32 {bindc_name = "y", uniq_name = "_QFEy"}
 !CHECK:    %[[Y_DECL:.*]]:2 = hlfir.declare %[[Y_REF]] {uniq_name = "_QFEy"} : (!fir.ref<i32>) -> (!fir.ref<i32>, !fir.ref<i32>)
-!CHECK:    omp.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1   hint(uncontended) memory_order(acquire) : !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[A_DECL]]#1 = %[[B_DECL]]#1   memory_order(relaxed) : !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[C_DECL]]#1 = %[[D_DECL]]#1   hint(contended) memory_order(seq_cst) : !fir.ref<!fir.logical<4>>, !fir.logical<4>
-!CHECK:    omp.atomic.read %[[E_DECL]]#1 = %[[F_DECL]]#1   hint(speculative) : !fir.ref<i32>, i32
-!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   hint(nonspeculative) : !fir.ref<f32>, f32
-!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   : !fir.ref<f32>, f32
+!CHECK:    omp.atomic.read %[[X_DECL]]#1 = %[[Y_DECL]]#1   hint(uncontended) memory_order(acquire) : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[A_DECL]]#1 = %[[B_DECL]]#1   memory_order(relaxed) : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[C_DECL]]#1 = %[[D_DECL]]#1   hint(contended) memory_order(seq_cst) : !fir.ref<!fir.logical<4>>, !fir.ref<!fir.logical<4>>, !fir.logical<4>
+!CHECK:    omp.atomic.read %[[E_DECL]]#1 = %[[F_DECL]]#1   hint(speculative) : !fir.ref<i32>, !fir.ref<i32>, i32
+!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   hint(nonspeculative) : !fir.ref<f32>, !fir.ref<f32>, f32
+!CHECK:    omp.atomic.read %[[G_DECL]]#1 = %[[H_DECL]]#1   : !fir.ref<f32>, !fir.ref<f32>, f32
 
 program OmpAtomic
 
@@ -68,7 +68,7 @@ end program OmpAtomic
 !CHECK:    %[[X_POINTEE_ADDR:.*]] = fir.box_addr %[[X_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 !CHECK:    %[[Y_ADDR:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 !CHECK:    %[[Y_POINTEE_ADDR:.*]] = fir.box_addr %[[Y_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
-!CHECK:    omp.atomic.read %[[Y_POINTEE_ADDR]] = %[[X_POINTEE_ADDR]]   : !fir.ptr<i32>, i32
+!CHECK:    omp.atomic.read %[[Y_POINTEE_ADDR]] = %[[X_POINTEE_ADDR]]   : !fir.ptr<i32>, !fir.ptr<i32>, i32
 !CHECK:    %[[Y_ADDR:.*]] = fir.load %[[Y_DECL]]#0 : !fir.ref<!fir.box<!fir.ptr<i32>>>
 !CHECK:    %[[Y_POINTEE_ADDR:.*]] = fir.box_addr %[[Y_ADDR]] : (!fir.box<!fir.ptr<i32>>) -> !fir.ptr<i32>
 !CHECK:    %[[Y_POINTEE_VAL:.*]] = fir.load %[[Y_POINTEE_ADDR]] : !fir.ptr<i32>
diff --git a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
index e305e2fbde5b17..1d179c74a970e6 100644
--- a/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
+++ b/mlir/include/mlir/Dialect/OpenACC/OpenACCOps.td
@@ -1958,8 +1958,7 @@ def OpenACC_YieldOp : OpenACC_Op<"yield", [Pure, ReturnLike, Terminator,
 // 2.12 atomic construct
 //===----------------------------------------------------------------------===//
 
-def AtomicReadOp : OpenACC_Op<"atomic.read", [AllTypesMatch<["x", "v"]>,
-                                              AtomicReadOpInterface]> {
+def AtomicReadOp : OpenACC_Op<"atomic.read", [AtomicReadOpInterface]> {
 
   let summary = "performs an atomic read";
 
@@ -1975,7 +1974,7 @@ def AtomicReadOp : OpenACC_Op<"atomic.read", [AllTypesMatch<["x", "v"]>,
                        TypeAttr:$element_type);
   let assemblyFormat = [{
     $v `=` $x
-    `:` type($x) `,` $element_type attr-dict
+    `:` type($v) `,` type($x) `,` $element_type attr-dict
   }];
   let hasVerifier = 1;
 }
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 626539cb7bde42..5fd8184fe0e0f7 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -1286,7 +1286,7 @@ def TaskwaitOp : OpenMP_Op<"taskwait", clauses = [
 // two-step process.
 
 def AtomicReadOp : OpenMP_Op<"atomic.read", traits = [
-    AllTypesMatch<["x", "v"]>, AtomicReadOpInterface
+    AtomicReadOpInterface
   ], clauses = [
     OpenMP_HintClause, OpenMP_MemoryOrderClause
   ]> {
@@ -1304,7 +1304,8 @@ def AtomicReadOp : OpenMP_Op<"atomic.read", traits = [
 
   // Override clause-based assemblyFormat.
   let assemblyFormat = "$v `=` $x" # clausesReqAssemblyFormat # " oilist(" #
-    clausesOptAssemblyFormat # ") `:` type($x) `,` $element_type attr-dict";
+    clausesOptAssemblyFormat #
+    ") `:` type($v) `,` type($x) `,` $element_type attr-dict";
 
   let extraClassDeclaration = [{
     /// The number of variable operands.
diff --git a/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir b/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
index e496c66f86522b..4f37dd16b44dd3 100644
--- a/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
+++ b/mlir/test/Conversion/OpenMPToLLVM/convert-to-llvmir.mlir
@@ -120,7 +120,7 @@ func.func @atomic_write(%a: !llvm.ptr) -> () {
 // CHECK: (%[[ARG0:.*]]: !llvm.ptr, %[[ARG1:.*]]: !llvm.ptr)
 // CHECK: omp.atomic.read %[[ARG1]] = %[[ARG0]] hint(contended) memory_order(acquire) : !llvm.ptr
 func.func @atomic_read(%a: !llvm.ptr, %b: !llvm.ptr) -> () {
-  omp.atomic.read %b = %a memory_order(acquire) hint(contended) : !llvm.ptr, i32
+  omp.atomic.read %b = %a memory_order(acquire) hint(contended) : !llvm.ptr, !llvm.ptr, i32
   return
 }
 
diff --git a/mlir/test/Dialect/OpenACC/invalid.mlir b/mlir/test/Dialect/OpenACC/invalid.mlir
index 96edb585ae21a2..0cc65f7b30d98d 100644
--- a/mlir/test/Dialect/OpenACC/invalid.mlir
+++ b/mlir/test/Dialect/OpenACC/invalid.mlir
@@ -611,7 +611,7 @@ func.func @acc_atomic_update(%x: memref<i32>, %expr: i32) {
 func.func @acc_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
   // expected-error @below {{expected three operations in atomic.capture region}}
   acc.atomic.capture {
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     acc.terminator
   }
   return
@@ -622,8 +622,8 @@ func.func @acc_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
 func.func @acc_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
   acc.atomic.capture {
     // expected-error @below {{invalid sequence of operations in the capture region}}
-    acc.atomic.read %v = %x : memref<i32>, i32
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     acc.terminator
   }
   return
@@ -699,7 +699,7 @@ func.func @acc_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
   acc.atomic.capture {
     // expected-error @below {{invalid sequence of operations in the capture region}}
     acc.atomic.write %x = %expr : memref<i32>, i32
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     acc.terminator
   }
   return
@@ -715,7 +715,7 @@ func.func @acc_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>,
       %newval = llvm.add %xval, %expr : i32
       acc.yield %newval : i32
     }
-    acc.atomic.read %v = %y : memref<i32>, i32
+    acc.atomic.read %v = %y : memref<i32>, memref<i32>, i32
     acc.terminator
   }
 }
@@ -725,7 +725,7 @@ func.func @acc_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>,
 func.func @acc_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>, %expr: i32) {
   acc.atomic.capture {
     // expected-error @below {{captured variable in atomic.read must be updated in second operation}}
-    acc.atomic.read %v = %y : memref<i32>, i32
+    acc.atomic.read %v = %y : memref<i32>, memref<i32>, i32
     acc.atomic.update %x : memref<i32> {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -740,7 +740,7 @@ func.func @acc_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>,
 func.func @acc_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>, %expr: i32) {
   acc.atomic.capture {
     // expected-error @below {{captured variable in atomic.read must be updated in second operation}}
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     acc.atomic.write %y = %expr : memref<i32>, i32
     acc.terminator
   }
diff --git a/mlir/test/Dialect/OpenACC/ops.mlir b/mlir/test/Dialect/OpenACC/ops.mlir
index 2ef2178cb2b63a..3ed81b5117f227 100644
--- a/mlir/test/Dialect/OpenACC/ops.mlir
+++ b/mlir/test/Dialect/OpenACC/ops.mlir
@@ -1703,8 +1703,8 @@ acc.set default_async(%i32Value : i32)
 // CHECK-LABEL: func.func @acc_atomic_read
 // CHECK-SAME: (%[[v:.*]]: memref<i32>, %[[x:.*]]: memref<i32>)
 func.func @acc_atomic_read(%v: memref<i32>, %x: memref<i32>) {
-  // CHECK: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
-  acc.atomic.read %v = %x : memref<i32>, i32
+  // CHECK: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
+  acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -1806,7 +1806,7 @@ func.func @acc_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   acc.yield %[[newval]] : i32
   // CHECK-NEXT: }
-  // CHECK-NEXT: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   acc.atomic.capture {
     acc.atomic.update %x : memref<i32> {
@@ -1814,10 +1814,10 @@ func.func @acc_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       acc.yield %newval : i32
     }
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   // CHECK: acc.atomic.capture {
-  // CHECK-NEXT: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: acc.atomic.update %[[x]] : memref<i32>
   // CHECK-NEXT: (%[[xval:.*]]: i32):
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
@@ -1825,7 +1825,7 @@ func.func @acc_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT: }
   // CHECK-NEXT: }
   acc.atomic.capture {
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     acc.atomic.update %x : memref<i32> {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -1833,11 +1833,11 @@ func.func @acc_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
     }
   }
   // CHECK: acc.atomic.capture {
-  // CHECK-NEXT: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: acc.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: acc.atomic.write %[[x]] = %[[expr]] : memref<i32>, i32
   // CHECK-NEXT: }
   acc.atomic.capture {
-    acc.atomic.read %v = %x : memref<i32>, i32
+    acc.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     acc.atomic.write %x = %expr : memref<i32>, i32
   }
 
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index fd89ec31c64a60..db941d401d52dc 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -891,7 +891,7 @@ func.func @omp_ordered5(%arg1 : i32, %arg2 : i32, %arg3 : i32, %vec0 : i64, %vec
 
 func.func @omp_atomic_read1(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{the hints omp_sync_hint_nonspeculative and omp_sync_hint_speculative cannot be combined.}}
-  omp.atomic.read %v = %x hint(speculative, nonspeculative) : memref<i32>, i32
+  omp.atomic.read %v = %x hint(speculative, nonspeculative) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -899,7 +899,7 @@ func.func @omp_atomic_read1(%x: memref<i32>, %v: memref<i32>) {
 
 func.func @omp_atomic_read2(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{invalid clause value: 'xyz'}}
-  omp.atomic.read %v = %x memory_order(xyz) : memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(xyz) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -907,7 +907,7 @@ func.func @omp_atomic_read2(%x: memref<i32>, %v: memref<i32>) {
 
 func.func @omp_atomic_read3(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{memory-order must not be acq_rel or release for atomic reads}}
-  omp.atomic.read %v = %x memory_order(acq_rel) : memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(acq_rel) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -915,7 +915,7 @@ func.func @omp_atomic_read3(%x: memref<i32>, %v: memref<i32>) {
 
 func.func @omp_atomic_read4(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{memory-order must not be acq_rel or release for atomic reads}}
-  omp.atomic.read %v = %x memory_order(release) : memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(release) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -923,7 +923,7 @@ func.func @omp_atomic_read4(%x: memref<i32>, %v: memref<i32>) {
 
 func.func @omp_atomic_read5(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{`memory_order` clause can appear at most once in the expansion of the oilist directive}}
-  omp.atomic.read %v = %x memory_order(acquire) memory_order(relaxed) : memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(acquire) memory_order(relaxed) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -931,7 +931,7 @@ func.func @omp_atomic_read5(%x: memref<i32>, %v: memref<i32>) {
 
 func.func @omp_atomic_read6(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{`hint` clause can appear at most once in the expansion of the oilist directive}}
-  omp.atomic.read %v =  %x hint(speculative) hint(contended) : memref<i32>, i32
+  omp.atomic.read %v =  %x hint(speculative) hint(contended) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -939,7 +939,7 @@ func.func @omp_atomic_read6(%x: memref<i32>, %v: memref<i32>) {
 
 func.func @omp_atomic_read6(%x: memref<i32>, %v: memref<i32>) {
   // expected-error @below {{read and write must not be to the same location for atomic reads}}
-  omp.atomic.read %x =  %x hint(speculative) : memref<i32>, i32
+  omp.atomic.read %x =  %x hint(speculative) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -1137,7 +1137,7 @@ func.func @omp_atomic_update(%x: memref<i32>, %expr: i32) {
 func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
   // expected-error @below {{expected three operations in atomic.capture region}}
   omp.atomic.capture {
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     omp.terminator
   }
   return
@@ -1148,8 +1148,8 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
 func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
   omp.atomic.capture {
     // expected-error @below {{invalid sequence of operations in the capture region}}
-    omp.atomic.read %v = %x : memref<i32>, i32
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     omp.terminator
   }
   return
@@ -1225,7 +1225,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
   omp.atomic.capture {
     // expected-error @below {{invalid sequence of operations in the capture region}}
     omp.atomic.write %x = %expr : memref<i32>, i32
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     omp.terminator
   }
   return
@@ -1241,7 +1241,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>,
       %newval = llvm.add %xval, %expr : i32
       omp.yield (%newval : i32)
     }
-    omp.atomic.read %v = %y : memref<i32>, i32
+    omp.atomic.read %v = %y : memref<i32>, memref<i32>, i32
     omp.terminator
   }
 }
@@ -1251,7 +1251,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>,
 func.func @omp_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>, %expr: i32) {
   omp.atomic.capture {
     // expected-error @below {{captured variable in atomic.read must be updated in second operation}}
-    omp.atomic.read %v = %y : memref<i32>, i32
+    omp.atomic.read %v = %y : memref<i32>, memref<i32>, i32
     omp.atomic.update %x : memref<i32> {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -1266,7 +1266,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>,
 func.func @omp_atomic_capture(%x: memref<i32>, %y: memref<i32>, %v: memref<i32>, %expr: i32) {
   omp.atomic.capture {
     // expected-error @below {{captured variable in atomic.read must be updated in second operation}}
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     omp.atomic.write %y = %expr : memref<i32>, i32
     omp.terminator
   }
@@ -1282,7 +1282,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   return
 }
@@ -1297,7 +1297,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   return
 }
@@ -1312,7 +1312,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   return
 }
@@ -1327,7 +1327,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   return
 }
@@ -1342,7 +1342,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   return
 }
@@ -1357,7 +1357,7 @@ func.func @omp_atomic_capture(%x: memref<i32>, %v: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x memory_order(seq_cst) : memref<i32>, i32
+    omp.atomic.read %v = %x memory_order(seq_cst) : memref<i32>, memref<i32>, i32
   }
   return
 }
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index 6f11b451fa00a3..b606f9eb708cf3 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -1282,20 +1282,20 @@ func.func @omp_ordered(%arg1 : i32, %arg2 : i32, %arg3 : i32,
 // CHECK-LABEL: omp_atomic_read
 // CHECK-SAME: (%[[v:.*]]: memref<i32>, %[[x:.*]]: memref<i32>)
 func.func @omp_atomic_read(%v: memref<i32>, %x: memref<i32>) {
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
-  omp.atomic.read %v = %x : memref<i32>, i32
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(seq_cst) : memref<i32>, i32
-  omp.atomic.read %v = %x memory_order(seq_cst) : memref<i32>, i32
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(acquire) : memref<i32>, i32
-  omp.atomic.read %v = %x memory_order(acquire) : memref<i32>, i32
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(relaxed) : memref<i32>, i32
-  omp.atomic.read %v = %x memory_order(relaxed) : memref<i32>, i32
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] hint(contended, nonspeculative) : memref<i32>, i32
-  omp.atomic.read %v = %x hint(nonspeculative, contended) : memref<i32>, i32
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] hint(contended, speculative) memory_order(seq_cst) : memref<i32>, i32
-  omp.atomic.read %v = %x hint(speculative, contended) memory_order(seq_cst) : memref<i32>, i32
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(seq_cst) : memref<i32>, i32
-  omp.atomic.read %v = %x hint(none) memory_order(seq_cst) : memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(seq_cst) : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(seq_cst) : memref<i32>, memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(acquire) : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(acquire) : memref<i32>, memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(relaxed) : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x memory_order(relaxed) : memref<i32>, memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] hint(contended, nonspeculative) : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x hint(nonspeculative, contended) : memref<i32>, memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] hint(contended, speculative) memory_order(seq_cst) : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x hint(speculative, contended) memory_order(seq_cst) : memref<i32>, memref<i32>, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] memory_order(seq_cst) : memref<i32>, memref<i32>, i32
+  omp.atomic.read %v = %x hint(none) memory_order(seq_cst) : memref<i32>, memref<i32>, i32
   return
 }
 
@@ -1531,7 +1531,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture{
     omp.atomic.update %x : memref<i32> {
@@ -1539,10 +1539,10 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
   // CHECK: omp.atomic.capture {
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: omp.atomic.update %[[x]] : memref<i32>
   // CHECK-NEXT: (%[[xval:.*]]: i32):
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
@@ -1550,7 +1550,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT: }
   // CHECK-NEXT: }
   omp.atomic.capture{
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     omp.atomic.update %x : memref<i32> {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -1558,11 +1558,11 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
     }
   }
   // CHECK: omp.atomic.capture {
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: omp.atomic.write %[[x]] = %[[expr]] : memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture{
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
     omp.atomic.write %x = %expr : memref<i32>, i32
   }
 
@@ -1572,7 +1572,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(none) {
     omp.atomic.update %x : memref<i32> {
@@ -1580,7 +1580,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(uncontended) {
@@ -1589,7 +1589,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(uncontended) {
     omp.atomic.update %x : memref<i32> {
@@ -1597,7 +1597,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(contended) {
@@ -1606,7 +1606,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(contended) {
     omp.atomic.update %x : memref<i32> {
@@ -1614,7 +1614,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(nonspeculative) {
@@ -1623,7 +1623,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(nonspeculative) {
     omp.atomic.update %x : memref<i32> {
@@ -1631,7 +1631,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(speculative) {
@@ -1640,7 +1640,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(speculative) {
     omp.atomic.update %x : memref<i32> {
@@ -1648,7 +1648,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(uncontended, nonspeculative) {
@@ -1657,7 +1657,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(uncontended, nonspeculative) {
     omp.atomic.update %x : memref<i32> {
@@ -1665,7 +1665,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(contended, nonspeculative) {
@@ -1674,7 +1674,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(contended, nonspeculative) {
     omp.atomic.update %x : memref<i32> {
@@ -1682,7 +1682,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(uncontended, speculative) {
@@ -1691,7 +1691,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(uncontended, speculative) {
     omp.atomic.update %x : memref<i32> {
@@ -1699,7 +1699,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(contended, speculative) {
@@ -1708,7 +1708,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(contended, speculative) {
     omp.atomic.update %x : memref<i32> {
@@ -1716,7 +1716,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture memory_order(seq_cst) {
@@ -1725,7 +1725,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture memory_order(seq_cst) {
     omp.atomic.update %x : memref<i32> {
@@ -1733,7 +1733,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture memory_order(acq_rel) {
@@ -1742,7 +1742,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture memory_order(acq_rel) {
     omp.atomic.update %x : memref<i32> {
@@ -1750,7 +1750,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture memory_order(acquire) {
@@ -1759,7 +1759,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture memory_order(acquire) {
     omp.atomic.update %x : memref<i32> {
@@ -1767,7 +1767,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture memory_order(release) {
@@ -1776,7 +1776,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture memory_order(release) {
     omp.atomic.update %x : memref<i32> {
@@ -1784,7 +1784,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture memory_order(relaxed) {
@@ -1793,7 +1793,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture memory_order(relaxed) {
     omp.atomic.update %x : memref<i32> {
@@ -1801,7 +1801,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   // CHECK: omp.atomic.capture hint(contended, speculative) memory_order(seq_cst) {
@@ -1810,7 +1810,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
   // CHECK-NEXT:   %[[newval:.*]] = llvm.add %[[xval]], %[[expr]] : i32
   // CHECK-NEXT:   omp.yield(%[[newval]] : i32)
   // CHECK-NEXT: }
-  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, i32
+  // CHECK-NEXT: omp.atomic.read %[[v]] = %[[x]] : memref<i32>, memref<i32>, i32
   // CHECK-NEXT: }
   omp.atomic.capture hint(contended, speculative) memory_order(seq_cst) {
     omp.atomic.update %x : memref<i32> {
@@ -1818,7 +1818,7 @@ func.func @omp_atomic_capture(%v: memref<i32>, %x: memref<i32>, %expr: i32) {
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : memref<i32>, i32
+    omp.atomic.read %v = %x : memref<i32>, memref<i32>, i32
   }
 
   return
@@ -2420,14 +2420,14 @@ func.func @omp_requires_multiple() -> ()
 // CHECK-LABEL: @opaque_pointers_atomic_rwu
 // CHECK-SAME: (%[[v:.*]]: !llvm.ptr, %[[x:.*]]: !llvm.ptr)
 func.func @opaque_pointers_atomic_rwu(%v: !llvm.ptr, %x: !llvm.ptr) {
-  // CHECK: omp.atomic.read %[[v]] = %[[x]] : !llvm.ptr, i32
+  // CHECK: omp.atomic.read %[[v]] = %[[x]] : !llvm.ptr, !llvm.ptr, i32
   // CHECK: %[[VAL:.*]] = llvm.load %[[x]] : !llvm.ptr -> i32
   // CHECK: omp.atomic.write %[[v]] = %[[VAL]] : !llvm.ptr, i32
   // CHECK: omp.atomic.update %[[x]] : !llvm.ptr {
   // CHECK-NEXT: ^{{[[:alnum:]]+}}(%[[XVAL:.*]]: i32):
   // CHECK-NEXT:   omp.yield(%[[XVAL]] : i32)
   // CHECK-NEXT: }
-  omp.atomic.read %v = %x : !llvm.ptr, i32
+  omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   %val = llvm.load %x : !llvm.ptr -> i32
   omp.atomic.write %v = %val : !llvm.ptr, i32
   omp.atomic.update %x : !llvm.ptr {
diff --git a/mlir/test/Target/LLVMIR/openmp-llvm-invalid.mlir b/mlir/test/Target/LLVMIR/openmp-llvm-invalid.mlir
index 3c6ca1ef0c6bf5..41bc5c4ba525fc 100644
--- a/mlir/test/Target/LLVMIR/openmp-llvm-invalid.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-llvm-invalid.mlir
@@ -35,7 +35,7 @@ llvm.func @omp_atomic_update_multiple_step_update(%x: !llvm.ptr, %expr: i32) {
 llvm.func @omp_atomic_update_multiple_step_update(%x: !llvm.ptr, %v: !llvm.ptr, %expr: i32) {
   // expected-error @+1 {{LLVM Translation failed for operation: omp.atomic.capture}}
   omp.atomic.capture memory_order(seq_cst) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     // expected-error @+1 {{no atomic update operation with region argument as operand found inside atomic.update region}}
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
@@ -52,7 +52,7 @@ llvm.func @omp_atomic_update_multiple_step_update(%x: !llvm.ptr, %v: !llvm.ptr,
 // update operation
 llvm.func @omp_atomic_update_multiple_step_update(%x: !llvm.ptr, %v: !llvm.ptr, %expr: i32) {
   omp.atomic.capture memory_order(seq_cst) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %t1 = llvm.mul %xval, %expr : i32
diff --git a/mlir/test/Target/LLVMIR/openmp-llvm.mlir b/mlir/test/Target/LLVMIR/openmp-llvm.mlir
index 49f9f3562c78b5..e68102ecc8d474 100644
--- a/mlir/test/Target/LLVMIR/openmp-llvm.mlir
+++ b/mlir/test/Target/LLVMIR/openmp-llvm.mlir
@@ -1348,21 +1348,21 @@ llvm.func @omp_atomic_read(%arg0 : !llvm.ptr, %arg1 : !llvm.ptr) -> () {
 
   // CHECK: %[[X1:.*]] = load atomic i32, ptr %[[ARG0]] monotonic, align 4
   // CHECK: store i32 %[[X1]], ptr %[[ARG1]], align 4
-  omp.atomic.read %arg1 = %arg0 : !llvm.ptr, i32
+  omp.atomic.read %arg1 = %arg0 : !llvm.ptr, !llvm.ptr, i32
 
   // CHECK: %[[X2:.*]] = load atomic i32, ptr %[[ARG0]] seq_cst, align 4
   // CHECK: call void @__kmpc_flush(ptr @{{.*}})
   // CHECK: store i32 %[[X2]], ptr %[[ARG1]], align 4
-  omp.atomic.read %arg1 = %arg0 memory_order(seq_cst) : !llvm.ptr, i32
+  omp.atomic.read %arg1 = %arg0 memory_order(seq_cst) : !llvm.ptr, !llvm.ptr, i32
 
   // CHECK: %[[X3:.*]] = load atomic i32, ptr %[[ARG0]] acquire, align 4
   // CHECK: call void @__kmpc_flush(ptr @{{.*}})
   // CHECK: store i32 %[[X3]], ptr %[[ARG1]], align 4
-  omp.atomic.read %arg1 = %arg0 memory_order(acquire) : !llvm.ptr, i32
+  omp.atomic.read %arg1 = %arg0 memory_order(acquire) : !llvm.ptr, !llvm.ptr, i32
 
   // CHECK: %[[X4:.*]] = load atomic i32, ptr %[[ARG0]] monotonic, align 4
   // CHECK: store i32 %[[X4]], ptr %[[ARG1]], align 4
-  omp.atomic.read %arg1 = %arg0 memory_order(relaxed) : !llvm.ptr, i32
+  omp.atomic.read %arg1 = %arg0 memory_order(relaxed) : !llvm.ptr, !llvm.ptr, i32
   llvm.return
 }
 
@@ -1518,7 +1518,7 @@ llvm.func @_QPomp_atomic_capture_complex() {
         %20 = llvm.insertvalue %17, %19[1] : !llvm.struct<(f32, f32)>
         omp.yield(%20 : !llvm.struct<(f32, f32)>)
       }
-      omp.atomic.read %1 = %3 : !llvm.ptr, !llvm.struct<(f32, f32)>
+      omp.atomic.read %1 = %3 : !llvm.ptr, !llvm.ptr, !llvm.struct<(f32, f32)>
     }
     llvm.return
 }
@@ -1541,7 +1541,7 @@ llvm.func @omp_atomic_read_complex(){
     %1 = llvm.alloca %0 x !llvm.struct<(f32, f32)> {bindc_name = "ib"} : (i64) -> !llvm.ptr
     %2 = llvm.mlir.constant(1 : i64) : i64
     %3 = llvm.alloca %2 x !llvm.struct<(f32, f32)> {bindc_name = "ia"} : (i64) -> !llvm.ptr
-    omp.atomic.read %1 = %3 : !llvm.ptr, !llvm.struct<(f32, f32)>
+    omp.atomic.read %1 = %3 : !llvm.ptr, !llvm.ptr, !llvm.struct<(f32, f32)>
     llvm.return
 }
 
@@ -1646,7 +1646,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.add %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[res:.*]] = atomicrmw sub ptr %[[x]], i32 %[[expr]] monotonic
@@ -1658,7 +1658,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.sub %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[res:.*]] = atomicrmw and ptr %[[x]], i32 %[[expr]] monotonic
@@ -1670,7 +1670,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.and %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[res:.*]] = atomicrmw or ptr %[[x]], i32 %[[expr]] monotonic
@@ -1682,7 +1682,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.or %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[res:.*]] = atomicrmw xor ptr %[[x]], i32 %[[expr]] monotonic
@@ -1694,7 +1694,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.xor %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1709,7 +1709,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.mul %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1724,7 +1724,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.sdiv %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1739,7 +1739,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.udiv %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1754,7 +1754,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.shl %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1769,7 +1769,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.lshr %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1784,7 +1784,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.ashr %xval, %expr : i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1799,7 +1799,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = "llvm.intr.smax"(%xval, %expr) : (i32, i32) -> i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1814,7 +1814,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = "llvm.intr.smin"(%xval, %expr) : (i32, i32) -> i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1829,7 +1829,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = "llvm.intr.umax"(%xval, %expr) : (i32, i32) -> i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1844,7 +1844,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = "llvm.intr.umin"(%xval, %expr) : (i32, i32) -> i32
       omp.yield(%newval : i32)
     }
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1859,7 +1859,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.fadd %xval, %exprf : f32
       omp.yield(%newval : f32)
     }
-    omp.atomic.read %vf = %xf : !llvm.ptr, f32
+    omp.atomic.read %vf = %xf : !llvm.ptr, !llvm.ptr, f32
   }
 
   // CHECK: %[[xval:.*]] = phi i32
@@ -1874,7 +1874,7 @@ llvm.func @omp_atomic_capture_prefix_update(
       %newval = llvm.fsub %xval, %exprf : f32
       omp.yield(%newval : f32)
     }
-    omp.atomic.read %vf = %xf : !llvm.ptr, f32
+    omp.atomic.read %vf = %xf : !llvm.ptr, !llvm.ptr, f32
   }
 
   llvm.return
@@ -1890,7 +1890,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %[[res:.*]] = atomicrmw add ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -1901,7 +1901,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %[[res:.*]] = atomicrmw sub ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.sub %xval, %expr : i32
@@ -1912,7 +1912,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %[[res:.*]] = atomicrmw and ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.and %xval, %expr : i32
@@ -1923,7 +1923,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %[[res:.*]] = atomicrmw or ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.or %xval, %expr : i32
@@ -1934,7 +1934,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %[[res:.*]] = atomicrmw xor ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.xor %xval, %expr : i32
@@ -1949,7 +1949,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.mul %xval, %expr : i32
@@ -1964,7 +1964,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.sdiv %xval, %expr : i32
@@ -1979,7 +1979,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.udiv %xval, %expr : i32
@@ -1994,7 +1994,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.shl %xval, %expr : i32
@@ -2009,7 +2009,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.lshr %xval, %expr : i32
@@ -2024,7 +2024,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.ashr %xval, %expr : i32
@@ -2039,7 +2039,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = "llvm.intr.smax"(%xval, %expr) : (i32, i32) -> i32
@@ -2054,7 +2054,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = "llvm.intr.smin"(%xval, %expr) : (i32, i32) -> i32
@@ -2069,7 +2069,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = "llvm.intr.umax"(%xval, %expr) : (i32, i32) -> i32
@@ -2084,7 +2084,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK-NEXT: %{{.*}} = cmpxchg ptr %[[x]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = "llvm.intr.umin"(%xval, %expr) : (i32, i32) -> i32
@@ -2100,7 +2100,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %{{.*}} = cmpxchg ptr %[[xf]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store float %[[xvalf]], ptr %[[vf]]
   omp.atomic.capture {
-    omp.atomic.read %vf = %xf : !llvm.ptr, f32
+    omp.atomic.read %vf = %xf : !llvm.ptr, !llvm.ptr, f32
     omp.atomic.update %xf : !llvm.ptr {
     ^bb0(%xval: f32):
       %newval = llvm.fadd %xval, %exprf : f32
@@ -2116,7 +2116,7 @@ llvm.func @omp_atomic_capture_postfix_update(
   // CHECK: %{{.*}} = cmpxchg ptr %[[xf]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store float %[[xvalf]], ptr %[[vf]]
   omp.atomic.capture {
-    omp.atomic.read %vf = %xf : !llvm.ptr, f32
+    omp.atomic.read %vf = %xf : !llvm.ptr, !llvm.ptr, f32
     omp.atomic.update %xf : !llvm.ptr {
     ^bb0(%xval: f32):
       %newval = llvm.fsub %xval, %exprf : f32
@@ -2136,7 +2136,7 @@ llvm.func @omp_atomic_capture_misc(
   // CHECK: %[[xval:.*]] = atomicrmw xchg ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[xval]], ptr %[[v]]
   omp.atomic.capture{
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.write %x = %expr : !llvm.ptr, i32
   }
 
@@ -2147,14 +2147,14 @@ llvm.func @omp_atomic_capture_misc(
   // CHECK: %{{.*}} = cmpxchg ptr %[[xf]], i32 %[[xval]], i32 %[[newval_]] monotonic monotonic
   // CHECK: store float %[[xvalf]], ptr %[[vf]]
   omp.atomic.capture{
-    omp.atomic.read %vf = %xf : !llvm.ptr, f32
+    omp.atomic.read %vf = %xf : !llvm.ptr, !llvm.ptr, f32
     omp.atomic.write %xf = %exprf : !llvm.ptr, f32
   }
 
   // CHECK: %[[res:.*]] = atomicrmw add ptr %[[x]], i32 %[[expr]] seq_cst
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture memory_order(seq_cst) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -2165,7 +2165,7 @@ llvm.func @omp_atomic_capture_misc(
   // CHECK: %[[res:.*]] = atomicrmw add ptr %[[x]], i32 %[[expr]] acquire
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture memory_order(acquire) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -2176,7 +2176,7 @@ llvm.func @omp_atomic_capture_misc(
   // CHECK: %[[res:.*]] = atomicrmw add ptr %[[x]], i32 %[[expr]] release
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture memory_order(release) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -2187,7 +2187,7 @@ llvm.func @omp_atomic_capture_misc(
   // CHECK: %[[res:.*]] = atomicrmw add ptr %[[x]], i32 %[[expr]] monotonic
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture memory_order(relaxed) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -2198,7 +2198,7 @@ llvm.func @omp_atomic_capture_misc(
   // CHECK: %[[res:.*]] = atomicrmw add ptr %[[x]], i32 %[[expr]] acq_rel
   // CHECK: store i32 %[[res]], ptr %[[v]]
   omp.atomic.capture memory_order(acq_rel) {
-    omp.atomic.read %v = %x : !llvm.ptr, i32
+    omp.atomic.read %v = %x : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %x : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32
@@ -2771,12 +2771,12 @@ llvm.func @omp_taskgroup_task(%x: i32, %y: i32, %zaddr: !llvm.ptr) {
 llvm.func @omp_opaque_pointers(%arg0 : !llvm.ptr, %arg1: !llvm.ptr, %expr: i32) -> () {
   // CHECK: %[[X1:.*]] = load atomic i32, ptr %[[ARG0]] monotonic, align 4
   // CHECK: store i32 %[[X1]], ptr %[[ARG1]], align 4
-  omp.atomic.read %arg1 = %arg0 : !llvm.ptr, i32
+  omp.atomic.read %arg1 = %arg0 : !llvm.ptr, !llvm.ptr, i32
 
   // CHECK: %[[RES:.*]] = atomicrmw add ptr %[[ARG1]], i32 %[[EXPR]] acq_rel
   // CHECK: store i32 %[[RES]], ptr %[[ARG0]]
   omp.atomic.capture memory_order(acq_rel) {
-    omp.atomic.read %arg0 = %arg1 : !llvm.ptr, i32
+    omp.atomic.read %arg0 = %arg1 : !llvm.ptr, !llvm.ptr, i32
     omp.atomic.update %arg1 : !llvm.ptr {
     ^bb0(%xval: i32):
       %newval = llvm.add %xval, %expr : i32



More information about the flang-commits mailing list