[flang] [llvm] [mlir] [flang][OpenMP] Add MLIR lowering for `loop ... bind` (PR #114219)

Kareem Ergawy via llvm-commits llvm-commits at lists.llvm.org
Wed Oct 30 05:36:14 PDT 2024


https://github.com/ergawy created https://github.com/llvm/llvm-project/pull/114219

Extends MLIR lowering support for the `loop` directive by adding
lowering support for the `bind` clause.

>From 23f1f4e94f93e1d508b75ba8cb5113f9779ddbee Mon Sep 17 00:00:00 2001
From: ergawy <kareem.ergawy at amd.com>
Date: Fri, 25 Oct 2024 04:00:52 -0500
Subject: [PATCH 1/4] [flang][OpenMP] Parase `bind` clause for `loop`
 direcitve.

Adds parsing for the `bind` clause. The clause was already part of the
`loop` direcitve's definition but parsing was still missing.
---
 flang/include/flang/Parser/dump-parse-tree.h  |  2 +
 flang/include/flang/Parser/parse-tree.h       |  7 ++++
 .../flang/Semantics/openmp-directive-sets.h   |  2 +
 flang/lib/Parser/openmp-parsers.cpp           |  9 +++++
 flang/lib/Semantics/check-omp-structure.cpp   | 39 ++++++++++++++++++-
 flang/lib/Semantics/check-omp-structure.h     |  1 +
 flang/lib/Semantics/resolve-directives.cpp    |  1 +
 .../Parser/OpenMP/target-loop-unparse.f90     | 16 +++++++-
 flang/test/Semantics/OpenMP/loop-bind.f90     | 33 ++++++++++++++++
 .../Semantics/OpenMP/nested-distribute.f90    |  6 +--
 flang/test/Semantics/OpenMP/nested-teams.f90  |  2 +-
 llvm/include/llvm/Frontend/OpenMP/OMP.td      |  1 +
 12 files changed, 111 insertions(+), 8 deletions(-)
 create mode 100644 flang/test/Semantics/OpenMP/loop-bind.f90

diff --git a/flang/include/flang/Parser/dump-parse-tree.h b/flang/include/flang/Parser/dump-parse-tree.h
index 31ad1b7c6ce5b5..ef6a774ec2c3d2 100644
--- a/flang/include/flang/Parser/dump-parse-tree.h
+++ b/flang/include/flang/Parser/dump-parse-tree.h
@@ -551,6 +551,8 @@ class ParseTreeDumper {
   NODE_ENUM(OmpGrainsizeClause, Prescriptiveness)
   NODE(parser, OmpNumTasksClause)
   NODE_ENUM(OmpNumTasksClause, Prescriptiveness)
+  NODE(parser, OmpBindClause)
+  NODE_ENUM(OmpBindClause, Type)
   NODE(parser, OmpProcBindClause)
   NODE_ENUM(OmpProcBindClause, Type)
   NODE_ENUM(OmpReductionClause, ReductionModifier)
diff --git a/flang/include/flang/Parser/parse-tree.h b/flang/include/flang/Parser/parse-tree.h
index 174f4c631e9d4c..d7d996ddf4239d 100644
--- a/flang/include/flang/Parser/parse-tree.h
+++ b/flang/include/flang/Parser/parse-tree.h
@@ -3716,6 +3716,13 @@ struct OmpNumTasksClause {
   std::tuple<std::optional<Prescriptiveness>, ScalarIntExpr> t;
 };
 
+// OMP 5.2 11.7.1 bind-clause ->
+//                  BIND( PARALLEL | TEAMS | THREAD )
+struct OmpBindClause {
+  ENUM_CLASS(Type, Parallel, Teams, Thread)
+  WRAPPER_CLASS_BOILERPLATE(OmpBindClause, Type);
+};
+
 // OpenMP Clauses
 struct OmpClause {
   UNION_CLASS_BOILERPLATE(OmpClause);
diff --git a/flang/include/flang/Semantics/openmp-directive-sets.h b/flang/include/flang/Semantics/openmp-directive-sets.h
index 55ef1e0ca61b9f..5e51c5c7de0e82 100644
--- a/flang/include/flang/Semantics/openmp-directive-sets.h
+++ b/flang/include/flang/Semantics/openmp-directive-sets.h
@@ -169,6 +169,7 @@ static const OmpDirectiveSet topTeamsSet{
     Directive::OMPD_teams_distribute_parallel_do,
     Directive::OMPD_teams_distribute_parallel_do_simd,
     Directive::OMPD_teams_distribute_simd,
+    Directive::OMPD_teams_loop,
 };
 
 static const OmpDirectiveSet allTeamsSet{
@@ -366,6 +367,7 @@ static const OmpDirectiveSet nestedTeamsAllowedSet{
     Directive::OMPD_distribute_parallel_do,
     Directive::OMPD_distribute_parallel_do_simd,
     Directive::OMPD_distribute_simd,
+    Directive::OMPD_loop,
     Directive::OMPD_parallel,
     Directive::OMPD_parallel_do,
     Directive::OMPD_parallel_do_simd,
diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp
index 5276e1ec1dcadd..e4a0192f20a105 100644
--- a/flang/lib/Parser/openmp-parsers.cpp
+++ b/flang/lib/Parser/openmp-parsers.cpp
@@ -428,6 +428,12 @@ TYPE_PARSER(construct<OmpLastprivateClause>(
         pure(OmpLastprivateClause::LastprivateModifier::Conditional) / ":"),
     Parser<OmpObjectList>{}))
 
+// OMP 5.2 11.7.1 BIND ( PARALLEL | TEAMS | THREAD )
+TYPE_PARSER(construct<OmpBindClause>(
+    "PARALLEL" >> pure(OmpBindClause::Type::Parallel) ||
+    "TEAMS" >> pure(OmpBindClause::Type::Teams) ||
+    "THREAD" >> pure(OmpBindClause::Type::Thread)))
+
 TYPE_PARSER(
     "ACQUIRE" >> construct<OmpClause>(construct<OmpClause::Acquire>()) ||
     "ACQ_REL" >> construct<OmpClause>(construct<OmpClause::AcqRel>()) ||
@@ -442,6 +448,8 @@ TYPE_PARSER(
     "ATOMIC_DEFAULT_MEM_ORDER" >>
         construct<OmpClause>(construct<OmpClause::AtomicDefaultMemOrder>(
             parenthesized(Parser<OmpAtomicDefaultMemOrderClause>{}))) ||
+    "BIND" >> construct<OmpClause>(construct<OmpClause::Bind>(
+                  parenthesized(Parser<OmpBindClause>{}))) ||
     "COLLAPSE" >> construct<OmpClause>(construct<OmpClause::Collapse>(
                       parenthesized(scalarIntConstantExpr))) ||
     "COPYIN" >> construct<OmpClause>(construct<OmpClause::Copyin>(
@@ -623,6 +631,7 @@ TYPE_PARSER(sourced(construct<OmpLoopDirective>(first(
     "TEAMS DISTRIBUTE SIMD" >>
         pure(llvm::omp::Directive::OMPD_teams_distribute_simd),
     "TEAMS DISTRIBUTE" >> pure(llvm::omp::Directive::OMPD_teams_distribute),
+    "TEAMS LOOP" >> pure(llvm::omp::Directive::OMPD_teams_loop),
     "TILE" >> pure(llvm::omp::Directive::OMPD_tile),
     "UNROLL" >> pure(llvm::omp::Directive::OMPD_unroll)))))
 
diff --git a/flang/lib/Semantics/check-omp-structure.cpp b/flang/lib/Semantics/check-omp-structure.cpp
index 8f3eb9fefee678..f6ad1b996c9433 100644
--- a/flang/lib/Semantics/check-omp-structure.cpp
+++ b/flang/lib/Semantics/check-omp-structure.cpp
@@ -369,13 +369,47 @@ void OmpStructureChecker::HasInvalidDistributeNesting(
         "region."_err_en_US);
   }
 }
+void OmpStructureChecker::HasInvalidLoopBinding(
+    const parser::OpenMPLoopConstruct &x) {
+  const auto &beginLoopDir{std::get<parser::OmpBeginLoopDirective>(x.t)};
+  const auto &beginDir{std::get<parser::OmpLoopDirective>(beginLoopDir.t)};
+
+  auto teamsBindingChecker = [&](parser::MessageFixedText msg) {
+    const auto &clauseList{std::get<parser::OmpClauseList>(beginLoopDir.t)};
+    for (const auto &clause : clauseList.v) {
+      if (const auto *bindClause{
+              std::get_if<parser::OmpClause::Bind>(&clause.u)}) {
+        if (bindClause->v.v != parser::OmpBindClause::Type::Teams) {
+          context_.Say(beginDir.source, msg);
+        }
+      }
+    }
+  };
+
+  if (llvm::omp::Directive::OMPD_loop == beginDir.v &&
+      CurrentDirectiveIsNested() &&
+      OmpDirectiveSet{llvm::omp::OMPD_teams, llvm::omp::OMPD_target_teams}.test(
+          GetContextParent().directive)) {
+    teamsBindingChecker(
+        "`BIND(TEAMS)` must be specified since the `LOOP` region is "
+        "strictly nested inside a `TEAMS` region."_err_en_US);
+  }
+
+  if (OmpDirectiveSet{
+          llvm::omp::OMPD_teams_loop, llvm::omp::OMPD_target_teams_loop}
+          .test(beginDir.v)) {
+    teamsBindingChecker(
+        "`BIND(TEAMS)` must be specified since the `LOOP` directive is "
+        "combined with a `TEAMS` construct."_err_en_US);
+  }
+}
 
 void OmpStructureChecker::HasInvalidTeamsNesting(
     const llvm::omp::Directive &dir, const parser::CharBlock &source) {
   if (!llvm::omp::nestedTeamsAllowedSet.test(dir)) {
     context_.Say(source,
-        "Only `DISTRIBUTE` or `PARALLEL` regions are allowed to be strictly "
-        "nested inside `TEAMS` region."_err_en_US);
+        "Only `DISTRIBUTE`, `PARALLEL`, or `LOOP` regions are allowed to be "
+        "strictly nested inside `TEAMS` region."_err_en_US);
   }
 }
 
@@ -538,6 +572,7 @@ void OmpStructureChecker::Enter(const parser::OpenMPLoopConstruct &x) {
   CheckLoopItrVariableIsInt(x);
   CheckAssociatedLoopConstraints(x);
   HasInvalidDistributeNesting(x);
+  HasInvalidLoopBinding(x);
   if (CurrentDirectiveIsNested() &&
       llvm::omp::topTeamsSet.test(GetContextParent().directive)) {
     HasInvalidTeamsNesting(beginDir.v, beginDir.source);
diff --git a/flang/lib/Semantics/check-omp-structure.h b/flang/lib/Semantics/check-omp-structure.h
index 237569bc40c483..b1f19c19bd2375 100644
--- a/flang/lib/Semantics/check-omp-structure.h
+++ b/flang/lib/Semantics/check-omp-structure.h
@@ -148,6 +148,7 @@ class OmpStructureChecker
   void HasInvalidTeamsNesting(
       const llvm::omp::Directive &dir, const parser::CharBlock &source);
   void HasInvalidDistributeNesting(const parser::OpenMPLoopConstruct &x);
+  void HasInvalidLoopBinding(const parser::OpenMPLoopConstruct &x);
   // specific clause related
   bool ScheduleModifierHasType(const parser::OmpScheduleClause &,
       const parser::OmpScheduleModifierType::ModType &);
diff --git a/flang/lib/Semantics/resolve-directives.cpp b/flang/lib/Semantics/resolve-directives.cpp
index 5e3ad5f3b4773d..20f050c71a426b 100644
--- a/flang/lib/Semantics/resolve-directives.cpp
+++ b/flang/lib/Semantics/resolve-directives.cpp
@@ -1663,6 +1663,7 @@ bool OmpAttributeVisitor::Pre(const parser::OpenMPLoopConstruct &x) {
   case llvm::omp::Directive::OMPD_teams_distribute_parallel_do:
   case llvm::omp::Directive::OMPD_teams_distribute_parallel_do_simd:
   case llvm::omp::Directive::OMPD_teams_distribute_simd:
+  case llvm::omp::Directive::OMPD_teams_loop:
   case llvm::omp::Directive::OMPD_tile:
   case llvm::omp::Directive::OMPD_unroll:
     PushContext(beginDir.source, beginDir.v);
diff --git a/flang/test/Parser/OpenMP/target-loop-unparse.f90 b/flang/test/Parser/OpenMP/target-loop-unparse.f90
index 3ee2fcef075a37..b2047070496527 100644
--- a/flang/test/Parser/OpenMP/target-loop-unparse.f90
+++ b/flang/test/Parser/OpenMP/target-loop-unparse.f90
@@ -1,6 +1,8 @@
+! RUN: %flang_fc1 -fdebug-unparse -fopenmp -fopenmp-version=50 %s | \
+! RUN:   FileCheck --ignore-case %s
 
-! RUN: %flang_fc1 -fdebug-unparse -fopenmp %s | FileCheck --ignore-case %s
-! RUN: %flang_fc1 -fdebug-dump-parse-tree -fopenmp %s | FileCheck --check-prefix="PARSE-TREE" %s
+! RUN: %flang_fc1 -fdebug-dump-parse-tree -fopenmp -fopenmp-version=50 %s | \
+! RUN:   FileCheck --check-prefix="PARSE-TREE" %s
 
 ! Check for parsing of loop directive
 
@@ -14,6 +16,16 @@ subroutine test_loop
    j = j + 1
   end do
   !$omp end loop
+
+  !PARSE-TREE: OmpBeginLoopDirective
+  !PARSE-TREE-NEXT: OmpLoopDirective -> llvm::omp::Directive = loop
+  !PARSE-TREE-NEXT: OmpClauseList -> OmpClause -> Bind -> OmpBindClause -> Type = Thread
+  !CHECK: !$omp loop
+  !$omp loop bind(thread)
+  do i=1,10
+   j = j + 1
+  end do
+  !$omp end loop
 end subroutine
 
 subroutine test_target_loop
diff --git a/flang/test/Semantics/OpenMP/loop-bind.f90 b/flang/test/Semantics/OpenMP/loop-bind.f90
new file mode 100644
index 00000000000000..f3aa9d19fe989e
--- /dev/null
+++ b/flang/test/Semantics/OpenMP/loop-bind.f90
@@ -0,0 +1,33 @@
+! RUN: %python %S/../test_errors.py %s %flang -fopenmp -fopenmp-version=50
+
+! OpenMP Version 5.0
+! Check OpenMP construct validity for the following directives:
+! 11.7 Loop directive
+
+program main
+  integer :: i, x
+
+  !$omp teams 
+  !ERROR: `BIND(TEAMS)` must be specified since the `LOOP` region is strictly nested inside a `TEAMS` region.
+  !$omp loop bind(thread)
+  do i = 1, 10
+    x = x + 1
+  end do
+  !$omp end loop
+  !$omp end teams
+
+  !ERROR: `BIND(TEAMS)` must be specified since the `LOOP` directive is combined with a `TEAMS` construct.
+  !$omp target teams loop bind(thread)
+  do i = 1, 10
+    x = x + 1
+  end do
+  !$omp end target teams loop
+
+  !ERROR: `BIND(TEAMS)` must be specified since the `LOOP` directive is combined with a `TEAMS` construct.
+  !$omp teams loop bind(thread)
+  do i = 1, 10
+    x = x + 1
+  end do
+  !$omp end teams loop
+
+end program main
diff --git a/flang/test/Semantics/OpenMP/nested-distribute.f90 b/flang/test/Semantics/OpenMP/nested-distribute.f90
index ba8c3bf04b3377..c212763cba1df8 100644
--- a/flang/test/Semantics/OpenMP/nested-distribute.f90
+++ b/flang/test/Semantics/OpenMP/nested-distribute.f90
@@ -21,7 +21,7 @@ program main
 
   !$omp teams
    do i = 1, N
-      !ERROR: Only `DISTRIBUTE` or `PARALLEL` regions are allowed to be strictly nested inside `TEAMS` region.
+      !ERROR: Only `DISTRIBUTE`, `PARALLEL`, or `LOOP` regions are allowed to be strictly nested inside `TEAMS` region.
       !$omp task
       do k = 1, N
          a = 3.14
@@ -50,7 +50,7 @@ program main
   !$omp end parallel
 
   !$omp teams
-   !ERROR: Only `DISTRIBUTE` or `PARALLEL` regions are allowed to be strictly nested inside `TEAMS` region.
+   !ERROR: Only `DISTRIBUTE`, `PARALLEL`, or `LOOP` regions are allowed to be strictly nested inside `TEAMS` region.
    !$omp target
       !ERROR: `DISTRIBUTE` region has to be strictly nested inside `TEAMS` region.
       !$omp distribute 
@@ -82,7 +82,7 @@ program main
   !$omp end target teams
 
   !$omp teams 
-      !ERROR: Only `DISTRIBUTE` or `PARALLEL` regions are allowed to be strictly nested inside `TEAMS` region.
+      !ERROR: Only `DISTRIBUTE`, `PARALLEL`, or `LOOP` regions are allowed to be strictly nested inside `TEAMS` region.
       !$omp task
       do k = 1,10
          print *, "hello"
diff --git a/flang/test/Semantics/OpenMP/nested-teams.f90 b/flang/test/Semantics/OpenMP/nested-teams.f90
index 06eea12aba5595..b1a7c92a6906b4 100644
--- a/flang/test/Semantics/OpenMP/nested-teams.f90
+++ b/flang/test/Semantics/OpenMP/nested-teams.f90
@@ -59,7 +59,7 @@ program main
 
   !$omp target
   !$omp teams
-  !ERROR: Only `DISTRIBUTE` or `PARALLEL` regions are allowed to be strictly nested inside `TEAMS` region.
+  !ERROR: Only `DISTRIBUTE`, `PARALLEL`, or `LOOP` regions are allowed to be strictly nested inside `TEAMS` region.
   !ERROR: TEAMS region can only be strictly nested within the implicit parallel region or TARGET region
   !$omp teams
   a = 3.14
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index 70179bab475779..feb8eb5c2abf49 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -73,6 +73,7 @@ def OMPC_AtomicDefaultMemOrder : Clause<"atomic_default_mem_order"> {
 }
 def OMPC_Bind : Clause<"bind"> {
   let clangClass = "OMPBindClause";
+  let flangClass = "OmpBindClause";
 }
 def OMP_CANCELLATION_CONSTRUCT_Parallel : ClauseVal<"parallel", 1, 1> {}
 def OMP_CANCELLATION_CONSTRUCT_Loop : ClauseVal<"loop", 2, 1> {}

>From 312ce293114a233b0455d368c6e5ec69e6aeb2ea Mon Sep 17 00:00:00 2001
From: ergawy <kareem.ergawy at amd.com>
Date: Mon, 28 Oct 2024 00:04:15 -0500
Subject: [PATCH 2/4] [flang][OpenMP][MLIR] Add MLIR op for `loop` directive

Adds MLIR op that corresponds to the `loop` directive.
---
 llvm/include/llvm/Frontend/OpenMP/OMP.td      | 11 +++++
 .../mlir/Dialect/OpenMP/OpenMPClauses.td      | 25 +++++++++++
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td | 44 +++++++++++++++++++
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  | 23 ++++++++++
 mlir/test/Dialect/OpenMP/invalid.mlir         | 30 +++++++++++++
 mlir/test/Dialect/OpenMP/ops.mlir             | 40 +++++++++++++++++
 6 files changed, 173 insertions(+)

diff --git a/llvm/include/llvm/Frontend/OpenMP/OMP.td b/llvm/include/llvm/Frontend/OpenMP/OMP.td
index feb8eb5c2abf49..9668ab1e90ecf3 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMP.td
+++ b/llvm/include/llvm/Frontend/OpenMP/OMP.td
@@ -71,10 +71,21 @@ def OMPC_AtomicDefaultMemOrder : Clause<"atomic_default_mem_order"> {
   let clangClass = "OMPAtomicDefaultMemOrderClause";
   let flangClass = "OmpAtomicDefaultMemOrderClause";
 }
+
+def OMP_BIND_parallel : ClauseVal<"parallel",1,1> {}
+def OMP_BIND_teams : ClauseVal<"teams",2,1> {}
+def OMP_BIND_thread : ClauseVal<"thread",3,1> { let isDefault = true; }
 def OMPC_Bind : Clause<"bind"> {
   let clangClass = "OMPBindClause";
   let flangClass = "OmpBindClause";
+  let enumClauseValue = "BindKind";
+  let allowedClauseValues = [
+    OMP_BIND_parallel,
+    OMP_BIND_teams,
+    OMP_BIND_thread
+  ];
 }
+
 def OMP_CANCELLATION_CONSTRUCT_Parallel : ClauseVal<"parallel", 1, 1> {}
 def OMP_CANCELLATION_CONSTRUCT_Loop : ClauseVal<"loop", 2, 1> {}
 def OMP_CANCELLATION_CONSTRUCT_Sections : ClauseVal<"sections", 3, 1> {}
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
index 886554f66afffc..855deab94b2f16 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPClauses.td
@@ -107,6 +107,31 @@ class OpenMP_CancelDirectiveNameClauseSkip<
 
 def OpenMP_CancelDirectiveNameClause : OpenMP_CancelDirectiveNameClauseSkip<>;
 
+//===----------------------------------------------------------------------===//
+// V5.2: [11.7.1] `bind` clause
+//===----------------------------------------------------------------------===//
+
+class OpenMP_BindClauseSkip<
+    bit traits = false, bit arguments = false, bit assemblyFormat = false,
+    bit description = false, bit extraClassDeclaration = false
+  > : OpenMP_Clause<traits, arguments, assemblyFormat, description,
+                    extraClassDeclaration> {
+  let arguments = (ins
+    OptionalAttr<BindKindAttr>:$bind_kind
+  );
+
+  let optAssemblyFormat = [{
+    `bind` `(` custom<ClauseAttr>($bind_kind) `)`
+  }];
+
+  let description = [{
+    The `bind` clause specifies the binding region of the construct on which it
+    appears.
+  }];
+}
+
+def OpenMP_BindClause : OpenMP_BindClauseSkip<>;
+
 //===----------------------------------------------------------------------===//
 // V5.2: [5.7.2] `copyprivate` clause
 //===----------------------------------------------------------------------===//
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 626539cb7bde42..4f16108a7a585f 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -382,6 +382,50 @@ def LoopNestOp : OpenMP_Op<"loop_nest", traits = [
 // 2.9.2 Workshare Loop Construct
 //===----------------------------------------------------------------------===//
 
+def LoopOp : OpenMP_Op<"loop", traits = [
+    DeclareOpInterfaceMethods<LoopWrapperInterface>, NoTerminator, SingleBlock,
+    AttrSizedOperandSegments
+  ], clauses = [
+    OpenMP_BindClause, OpenMP_PrivateClause, OpenMP_OrderClause,
+    OpenMP_ReductionClause
+  ], singleRegion = true> {
+  let summary = "loop construct";
+  let description = [{
+    A loop construct specifies that the logical iterations of the associated loops
+    may execute concurrently and permits the encountering threads to execute the
+    loop accordingly. A loop construct is a worksharing construct if its binding
+    region is the innermost enclosing parallel region. Otherwise it is not a work-
+    sharing region. The directive asserts that the iterations of the associated
+    loops may execute in any order, including concurrently. Each logical iteration
+    is executed once per instance of the loop region that is encountered by exactly
+    one thread that is a member of the binding thread set.
+
+    The body region can only contain a single block which must contain a single
+    operation, this operation must be an `omp.loop_nest`.
+
+    ```
+    omp.loop <clauses> {
+      omp.loop_nest (%i1, %i2) : index = (%c0, %c0) to (%c10, %c10) step (%c1, %c1) {
+        %a = load %arrA[%i1, %i2] : memref<?x?xf32>
+        %b = load %arrB[%i1, %i2] : memref<?x?xf32>
+        %sum = arith.addf %a, %b : f32
+        store %sum, %arrC[%i1, %i2] : memref<?x?xf32>
+        omp.yield
+      }
+    }
+    ```
+  }] # clausesDescription;
+
+  let assemblyFormat = clausesAssemblyFormat # [{
+    custom<PrivateReductionRegion>($region, $private_vars, type($private_vars),
+        $private_syms, $reduction_vars, type($reduction_vars), $reduction_byref,
+        $reduction_syms) attr-dict
+  }];
+
+  let hasVerifier = 1;
+  let hasRegionVerifier = 1;
+}
+
 def WsloopOp : OpenMP_Op<"wsloop", traits = [
     AttrSizedOperandSegments, DeclareOpInterfaceMethods<ComposableOpInterface>,
     DeclareOpInterfaceMethods<LoopWrapperInterface>, NoTerminator,
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index e1df647d6a3c71..39accbadc8e276 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1948,6 +1948,29 @@ LogicalResult LoopWrapperInterface::verifyImpl() {
   return success();
 }
 
+//===----------------------------------------------------------------------===//
+// LoopOp
+//===----------------------------------------------------------------------===//
+
+LogicalResult LoopOp::verify() {
+  return verifyReductionVarList(*this, getReductionSyms(), getReductionVars(),
+                                getReductionByref());
+}
+
+LogicalResult LoopOp::verifyRegions() {
+  Region &region = getRegion();
+
+  // Minimal amount of checks to verify the only nested op is an
+  // `omp.loop_nest`. A more extensive vierfication is done by the
+  // `LoopWrapperInterface` trait but the difference is that `omp.loop` cannot
+  // have another nested `LoopWrapperInterface`.
+  if (range_size(region.getOps()) != 1 || !isa<LoopNestOp>(*region.op_begin()))
+    return emitError() << "`omp.loop` expected to have a single nested "
+                          "operation which is a `omp.loop_nest`";
+
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // WsloopOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/Dialect/OpenMP/invalid.mlir b/mlir/test/Dialect/OpenMP/invalid.mlir
index fd89ec31c64a60..194daf8c2edf34 100644
--- a/mlir/test/Dialect/OpenMP/invalid.mlir
+++ b/mlir/test/Dialect/OpenMP/invalid.mlir
@@ -2577,3 +2577,33 @@ func.func @omp_taskloop_invalid_composite(%lb: index, %ub: index, %step: index)
   } {omp.composite}
   return
 }
+
+// -----
+
+func.func @omp_loop_invalid_nesting(%lb : index, %ub : index, %step : index) {
+
+  // expected-error @below {{`omp.loop` expected to have a single nested operation which is a `omp.loop_nest`}}
+  omp.loop {
+    omp.simd {
+      omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+        omp.yield
+      }
+    } {omp.composite}
+  }
+
+  return
+}
+
+// -----
+
+func.func @omp_loop_invalid_nesting(%lb : index, %ub : index, %step : index) {
+
+  // expected-error @below {{custom op 'omp.loop' invalid clause value: 'dummy_value'}}
+  omp.loop bind(dummy_value) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
+  }
+
+  return
+}
diff --git a/mlir/test/Dialect/OpenMP/ops.mlir b/mlir/test/Dialect/OpenMP/ops.mlir
index 6f11b451fa00a3..47b8890b851c10 100644
--- a/mlir/test/Dialect/OpenMP/ops.mlir
+++ b/mlir/test/Dialect/OpenMP/ops.mlir
@@ -2749,3 +2749,43 @@ func.func @omp_target_private(%map1: memref<?xi32>, %map2: memref<?xi32>, %priv_
 
   return
 }
+
+// CHECK-LABEL: omp_loop
+func.func @omp_loop(%lb : index, %ub : index, %step : index) {
+  // CHECK: omp.loop {
+  omp.loop {
+    // CHECK: omp.loop_nest {{.*}} {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      // CHECK: omp.yield
+      omp.yield
+    // CHECK: }
+    }
+  // CHECK: }
+  }
+
+  // CHECK: omp.loop bind(teams) {
+  omp.loop bind(teams) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
+  // CHECK: }
+  }
+
+  // CHECK: omp.loop bind(parallel) {
+  omp.loop bind(parallel) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
+  // CHECK: }
+  }
+
+  // CHECK: omp.loop bind(thread) {
+  omp.loop bind(thread) {
+    omp.loop_nest (%iv) : index = (%lb) to (%ub) step (%step) {
+      omp.yield
+    }
+  // CHECK: }
+  }
+
+  return
+}

>From 860d1a00a58778924dea99ca5bee671e01fbd61c Mon Sep 17 00:00:00 2001
From: ergawy <kareem.ergawy at amd.com>
Date: Wed, 30 Oct 2024 04:40:14 -0500
Subject: [PATCH 3/4] [flang][OpenMP] Add basic support to lower `loop` to MLIR

Adds initial support for lowering the `loop` directive to MLIR.

The PR includes basic suport and testing for the following clauses:
 * `collapse`
 * `order`
 * `private`
 * `reduction`
---
 flang/lib/Lower/OpenMP/OpenMP.cpp             | 48 ++++++++++-
 .../test/Lower/OpenMP/Todo/loop-directive.f90 | 15 ----
 flang/test/Lower/OpenMP/loop-directive.f90    | 82 +++++++++++++++++++
 mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td |  4 +
 .../Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp  |  3 +-
 mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp  | 11 +++
 6 files changed, 146 insertions(+), 17 deletions(-)
 delete mode 100644 flang/test/Lower/OpenMP/Todo/loop-directive.f90
 create mode 100644 flang/test/Lower/OpenMP/loop-directive.f90

diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 01a40d6e2204ef..6dd1583624b1f1 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -2023,6 +2023,52 @@ static void genStandaloneDo(lower::AbstractConverter &converter,
                 llvm::omp::Directive::OMPD_do, dsp);
 }
 
+static void genLoopClauses(
+    lower::AbstractConverter &converter, semantics::SemanticsContext &semaCtx,
+    const List<Clause> &clauses, mlir::Location loc,
+    mlir::omp::LoopOperands &clauseOps,
+    llvm::SmallVectorImpl<const semantics::Symbol *> &reductionSyms) {
+  ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processOrder(clauseOps);
+  cp.processReduction(loc, clauseOps, reductionSyms);
+  cp.processTODO<clause::Bind, clause::Lastprivate>(
+      loc, llvm::omp::Directive::OMPD_loop);
+}
+
+static void genLoopOp(lower::AbstractConverter &converter,
+                      lower::SymMap &symTable,
+                      semantics::SemanticsContext &semaCtx,
+                      lower::pft::Evaluation &eval, mlir::Location loc,
+                      const ConstructQueue &queue,
+                      ConstructQueue::const_iterator item) {
+  mlir::omp::LoopOperands loopClauseOps;
+  llvm::SmallVector<const semantics::Symbol *> loopReductionSyms;
+  genLoopClauses(converter, semaCtx, item->clauses, loc, loopClauseOps,
+                 loopReductionSyms);
+
+  DataSharingProcessor dsp(converter, semaCtx, item->clauses, eval,
+                           /*shouldCollectPreDeterminedSymbols=*/true,
+                           /*useDelayedPrivatization=*/true, &symTable);
+  dsp.processStep1(&loopClauseOps);
+
+  mlir::omp::LoopNestOperands loopNestClauseOps;
+  llvm::SmallVector<const semantics::Symbol *> iv;
+  genLoopNestClauses(converter, semaCtx, eval, item->clauses, loc,
+                     loopNestClauseOps, iv);
+
+  EntryBlockArgs loopArgs;
+  loopArgs.priv.syms = dsp.getDelayedPrivSymbols();
+  loopArgs.priv.vars = loopClauseOps.privateVars;
+  loopArgs.reduction.syms = loopReductionSyms;
+  loopArgs.reduction.vars = loopClauseOps.reductionVars;
+
+  auto loopOp =
+      genWrapperOp<mlir::omp::LoopOp>(converter, loc, loopClauseOps, loopArgs);
+  genLoopNestOp(converter, symTable, semaCtx, eval, loc, queue, item,
+                loopNestClauseOps, iv, {{loopOp, loopArgs}},
+                llvm::omp::Directive::OMPD_loop, dsp);
+}
+
 static void genStandaloneParallel(lower::AbstractConverter &converter,
                                   lower::SymMap &symTable,
                                   semantics::SemanticsContext &semaCtx,
@@ -2459,7 +2505,7 @@ static void genOMPDispatch(lower::AbstractConverter &converter,
     genStandaloneDo(converter, symTable, semaCtx, eval, loc, queue, item);
     break;
   case llvm::omp::Directive::OMPD_loop:
-    TODO(loc, "Unhandled directive " + llvm::omp::getOpenMPDirectiveName(dir));
+    genLoopOp(converter, symTable, semaCtx, eval, loc, queue, item);
     break;
   case llvm::omp::Directive::OMPD_masked:
     genMaskedOp(converter, symTable, semaCtx, eval, loc, queue, item);
diff --git a/flang/test/Lower/OpenMP/Todo/loop-directive.f90 b/flang/test/Lower/OpenMP/Todo/loop-directive.f90
deleted file mode 100644
index f1aea70458aa6c..00000000000000
--- a/flang/test/Lower/OpenMP/Todo/loop-directive.f90
+++ /dev/null
@@ -1,15 +0,0 @@
-! This test checks lowering of OpenMP loop Directive.
-
-! RUN: %not_todo_cmd bbc -emit-fir -fopenmp -o - %s 2>&1 | FileCheck %s
-! RUN: %not_todo_cmd %flang_fc1 -emit-fir -fopenmp -o - %s 2>&1 | FileCheck %s
-
-! CHECK: not yet implemented: Unhandled directive loop
-subroutine test_loop()
-  integer :: i, j = 1
-  !$omp loop
-  do i=1,10
-   j = j + 1
-  end do
-  !$omp end loop
-end subroutine
-
diff --git a/flang/test/Lower/OpenMP/loop-directive.f90 b/flang/test/Lower/OpenMP/loop-directive.f90
new file mode 100644
index 00000000000000..cfef8f84c3668f
--- /dev/null
+++ b/flang/test/Lower/OpenMP/loop-directive.f90
@@ -0,0 +1,82 @@
+! This test checks lowering of OpenMP loop Directive.
+
+! RUN: bbc -emit-hlfir -fopenmp -fopenmp-version=50 -o - %s 2>&1 | FileCheck %s
+! RUN: %flang_fc1 -emit-hlfir -fopenmp -fopenmp-version=50 -o - %s 2>&1 | FileCheck %s
+
+! CHECK: omp.declare_reduction @[[RED:add_reduction_i32]] : i32
+! CHECK: omp.private {type = private} @[[DUMMY_PRIV:.*test_privateEdummy_private.*]] : !fir.ref<i32>
+! CHECK: omp.private {type = private} @[[I_PRIV:.*test_no_clausesEi.*]] : !fir.ref<i32>
+
+! CHECK-LABEL: func.func @_QPtest_no_clauses
+subroutine test_no_clauses()
+  integer :: i, j, dummy = 1
+
+  ! CHECK: omp.loop private(@[[I_PRIV]] %{{.*}}#0 -> %[[ARG:.*]] : !fir.ref<i32>) {
+  ! CHECK:   omp.loop_nest (%[[IV:.*]]) : i32 = (%{{.*}}) to (%{{.*}}) {{.*}} {
+  ! CHECK:     %[[ARG_DECL:.*]]:2 = hlfir.declare %[[ARG]]
+  ! CHECK:     fir.store %[[IV]] to %[[ARG_DECL]]#1 : !fir.ref<i32>
+  ! CHECK:   }
+  ! CHECK: }
+  !$omp loop
+  do i=1,10
+   dummy = dummy + 1
+  end do
+  !$omp end loop
+end subroutine
+
+! CHECK-LABEL: func.func @_QPtest_collapse
+subroutine test_collapse()
+  integer :: i, j, dummy = 1
+  ! CHECK: omp.loop private(@{{.*}} %{{.*}}#0 -> %{{.*}}, @{{.*}} %{{.*}}#0 -> %{{.*}} : {{.*}}) {
+  ! CHECK:   omp.loop_nest (%{{.*}}, %{{.*}}) : i32 {{.*}} {
+  ! CHECK:   }
+  ! CHECK: }
+  !$omp loop collapse(2)
+  do i=1,10
+    do j=2,20
+     dummy = dummy + 1
+    end do
+  end do
+  !$omp end loop
+end subroutine
+
+! CHECK-LABEL: func.func @_QPtest_private
+subroutine test_private()
+  integer :: i, dummy = 1
+  ! CHECK: omp.loop private(@[[DUMMY_PRIV]] %{{.*}}#0 -> %{{.*}}, @{{.*}} %{{.*}}#0 -> %{{.*}} : {{.*}}) {
+  ! CHECK:   omp.loop_nest (%{{.*}}) : i32 = (%{{.*}}) to (%{{.*}}) {{.*}} {
+  ! CHECK:   }
+  ! CHECK: }
+  !$omp loop private(dummy)
+  do i=1,10
+   dummy = dummy + 1
+  end do
+  !$omp end loop
+end subroutine
+
+
+! CHECK-LABEL: func.func @_QPtest_order
+subroutine test_order()
+  integer :: i, dummy = 1
+  ! CHECK: omp.loop order(reproducible:concurrent) private(@{{.*}} %{{.*}}#0 -> %{{.*}} : {{.*}}) {
+  ! CHECK: }
+  !$omp loop order(concurrent)
+  do i=1,10
+   dummy = dummy + 1
+  end do
+  !$omp end loop
+end subroutine
+
+! CHECK-LABEL: func.func @_QPtest_reduction
+subroutine test_reduction()
+  integer :: i, dummy = 1
+
+  ! CHECK: omp.loop private(@{{.*}} %{{.*}}#0 -> %{{.*}} : !{{.*}}) reduction
+  ! CHECK-SAME:  (@[[RED]] %{{.*}}#0 -> %{{.*}} : !{{.*}}) {
+  ! CHECK: }
+  !$omp loop reduction(+:dummy)
+  do i=1,10
+   dummy = dummy + 1
+  end do
+  !$omp end loop
+end subroutine
diff --git a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
index 4f16108a7a585f..89c34e1410f6bd 100644
--- a/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
+++ b/mlir/include/mlir/Dialect/OpenMP/OpenMPOps.td
@@ -422,6 +422,10 @@ def LoopOp : OpenMP_Op<"loop", traits = [
         $reduction_syms) attr-dict
   }];
 
+  let builders = [
+    OpBuilder<(ins CArg<"const LoopOperands &">:$clauses)>
+  ];
+
   let hasVerifier = 1;
   let hasRegionVerifier = 1;
 }
diff --git a/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp b/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
index aa824a95b1574a..32db173c9c34a9 100644
--- a/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
+++ b/mlir/lib/Conversion/OpenMPToLLVM/OpenMPToLLVM.cpp
@@ -238,7 +238,7 @@ void mlir::configureOpenMPToLLVMConversionLegality(
       omp::ParallelOp, omp::PrivateClauseOp, omp::SectionOp, omp::SectionsOp,
       omp::SimdOp, omp::SingleOp, omp::TargetDataOp, omp::TargetOp,
       omp::TaskgroupOp, omp::TaskloopOp, omp::TaskOp, omp::TeamsOp,
-      omp::WsloopOp>([&](Operation *op) {
+      omp::WsloopOp, omp::LoopOp>([&](Operation *op) {
     return std::all_of(op->getRegions().begin(), op->getRegions().end(),
                        [&](Region &region) {
                          return typeConverter.isLegal(&region);
@@ -284,6 +284,7 @@ void mlir::populateOpenMPToLLVMConversionPatterns(LLVMTypeConverter &converter,
       RegionOpConversion<omp::TargetOp>, RegionOpConversion<omp::TaskgroupOp>,
       RegionOpConversion<omp::TaskloopOp>, RegionOpConversion<omp::TaskOp>,
       RegionOpConversion<omp::TeamsOp>, RegionOpConversion<omp::WsloopOp>,
+      RegionOpConversion<omp::LoopOp>,
       RegionOpWithVarOperandsConversion<omp::AtomicUpdateOp>>(converter);
 }
 
diff --git a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
index 39accbadc8e276..c2fc1b8aa554b6 100644
--- a/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
+++ b/mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp
@@ -1952,6 +1952,17 @@ LogicalResult LoopWrapperInterface::verifyImpl() {
 // LoopOp
 //===----------------------------------------------------------------------===//
 
+void LoopOp::build(OpBuilder &builder, OperationState &state,
+                   const LoopOperands &clauses) {
+  MLIRContext *ctx = builder.getContext();
+
+  LoopOp::build(builder, state, clauses.bindKind, clauses.privateVars,
+                makeArrayAttr(ctx, clauses.privateSyms), clauses.order,
+                clauses.orderMod, clauses.reductionVars,
+                makeDenseBoolArrayAttr(ctx, clauses.reductionByref),
+                makeArrayAttr(ctx, clauses.reductionSyms));
+}
+
 LogicalResult LoopOp::verify() {
   return verifyReductionVarList(*this, getReductionSyms(), getReductionVars(),
                                 getReductionByref());

>From ecbfdcdc1f562164f48791ec891831aa19e5d265 Mon Sep 17 00:00:00 2001
From: ergawy <kareem.ergawy at amd.com>
Date: Wed, 30 Oct 2024 07:33:39 -0500
Subject: [PATCH 4/4] [flang][OpenMP] Add MLIR lowering for `loop ... bind`

Extends MLIR lowering support for the `loop` directive by adding
lowering support for the `bind` clause.
---
 flang/lib/Lower/OpenMP/ClauseProcessor.cpp | 28 ++++++++++++++++++++++
 flang/lib/Lower/OpenMP/ClauseProcessor.h   |  1 +
 flang/lib/Lower/OpenMP/Clauses.cpp         | 15 ++++++++++--
 flang/lib/Lower/OpenMP/OpenMP.cpp          |  4 ++--
 flang/test/Lower/OpenMP/loop-directive.f90 | 12 ++++++++++
 5 files changed, 56 insertions(+), 4 deletions(-)

diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
index 7c254ce673855a..3e8d9c0ff161ee 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.cpp
@@ -98,6 +98,25 @@ genAllocateClause(lower::AbstractConverter &converter,
   genObjectList(objects, converter, allocateOperands);
 }
 
+static mlir::omp::ClauseBindKindAttr
+genBindKindAttr(fir::FirOpBuilder &firOpBuilder,
+                const omp::clause::Bind &clause) {
+  mlir::omp::ClauseBindKind bindKind;
+  switch (clause.v) {
+  case omp::clause::Bind::Binding::Teams:
+    bindKind = mlir::omp::ClauseBindKind::Teams;
+    break;
+  case omp::clause::Bind::Binding::Parallel:
+    bindKind = mlir::omp::ClauseBindKind::Parallel;
+    break;
+  case omp::clause::Bind::Binding::Thread:
+    bindKind = mlir::omp::ClauseBindKind::Thread;
+    break;
+  }
+  return mlir::omp::ClauseBindKindAttr::get(firOpBuilder.getContext(),
+                                            bindKind);
+}
+
 static mlir::omp::ClauseProcBindKindAttr
 genProcBindKindAttr(fir::FirOpBuilder &firOpBuilder,
                     const omp::clause::ProcBind &clause) {
@@ -197,6 +216,15 @@ static void convertLoopBounds(lower::AbstractConverter &converter,
 // ClauseProcessor unique clauses
 //===----------------------------------------------------------------------===//
 
+bool ClauseProcessor::processBind(mlir::omp::BindClauseOps &result) const {
+  if (auto *clause = findUniqueClause<omp::clause::Bind>()) {
+    fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder();
+    result.bindKind = genBindKindAttr(firOpBuilder, *clause);
+    return true;
+  }
+  return false;
+}
+
 bool ClauseProcessor::processCollapse(
     mlir::Location currentLocation, lower::pft::Evaluation &eval,
     mlir::omp::LoopRelatedClauseOps &result,
diff --git a/flang/lib/Lower/OpenMP/ClauseProcessor.h b/flang/lib/Lower/OpenMP/ClauseProcessor.h
index f34121c70d0b44..772e5415d2da27 100644
--- a/flang/lib/Lower/OpenMP/ClauseProcessor.h
+++ b/flang/lib/Lower/OpenMP/ClauseProcessor.h
@@ -53,6 +53,7 @@ class ClauseProcessor {
       : converter(converter), semaCtx(semaCtx), clauses(clauses) {}
 
   // 'Unique' clauses: They can appear at most once in the clause list.
+  bool processBind(mlir::omp::BindClauseOps &result) const;
   bool
   processCollapse(mlir::Location currentLocation, lower::pft::Evaluation &eval,
                   mlir::omp::LoopRelatedClauseOps &result,
diff --git a/flang/lib/Lower/OpenMP/Clauses.cpp b/flang/lib/Lower/OpenMP/Clauses.cpp
index 9483f643acd55a..71e2bcce27e183 100644
--- a/flang/lib/Lower/OpenMP/Clauses.cpp
+++ b/flang/lib/Lower/OpenMP/Clauses.cpp
@@ -461,8 +461,19 @@ AtomicDefaultMemOrder make(const parser::OmpClause::AtomicDefaultMemOrder &inp,
 
 Bind make(const parser::OmpClause::Bind &inp,
           semantics::SemanticsContext &semaCtx) {
-  // inp -> empty
-  llvm_unreachable("Empty: bind");
+  // inp.v -> parser::OmpBindClause
+  using wrapped = parser::OmpBindClause;
+
+  CLAUSET_ENUM_CONVERT( //
+      convert, wrapped::Type, Bind::Binding,
+      // clang-format off
+      MS(Teams, Teams)
+      MS(Parallel, Parallel)
+      MS(Thread, Thread)
+      // clang-format on
+  );
+
+  return Bind{/*Binding=*/convert(inp.v.v)};
 }
 
 // CancellationConstructType: empty
diff --git a/flang/lib/Lower/OpenMP/OpenMP.cpp b/flang/lib/Lower/OpenMP/OpenMP.cpp
index 6dd1583624b1f1..6b35ad81f2a0ca 100644
--- a/flang/lib/Lower/OpenMP/OpenMP.cpp
+++ b/flang/lib/Lower/OpenMP/OpenMP.cpp
@@ -2029,10 +2029,10 @@ static void genLoopClauses(
     mlir::omp::LoopOperands &clauseOps,
     llvm::SmallVectorImpl<const semantics::Symbol *> &reductionSyms) {
   ClauseProcessor cp(converter, semaCtx, clauses);
+  cp.processBind(clauseOps);
   cp.processOrder(clauseOps);
   cp.processReduction(loc, clauseOps, reductionSyms);
-  cp.processTODO<clause::Bind, clause::Lastprivate>(
-      loc, llvm::omp::Directive::OMPD_loop);
+  cp.processTODO<clause::Lastprivate>(loc, llvm::omp::Directive::OMPD_loop);
 }
 
 static void genLoopOp(lower::AbstractConverter &converter,
diff --git a/flang/test/Lower/OpenMP/loop-directive.f90 b/flang/test/Lower/OpenMP/loop-directive.f90
index cfef8f84c3668f..59bf20c060f4ee 100644
--- a/flang/test/Lower/OpenMP/loop-directive.f90
+++ b/flang/test/Lower/OpenMP/loop-directive.f90
@@ -80,3 +80,15 @@ subroutine test_reduction()
   end do
   !$omp end loop
 end subroutine
+
+! CHECK-LABEL: func.func @_QPtest_bind
+subroutine test_bind()
+  integer :: i, dummy = 1
+  ! CHECK: omp.loop bind(thread) private(@{{.*}} %{{.*}}#0 -> %{{.*}} : {{.*}}) {
+  ! CHECK: }
+  !$omp loop bind(thread)
+  do i=1,10
+   dummy = dummy + 1
+  end do
+  !$omp end loop
+end subroutine



More information about the llvm-commits mailing list