[OpenMP] Refactor OMPScheduleType enum.

The OMPScheduleType enum stores the constants from libomp's internal sched_type in kmp.h and are used by several kmp API functions. The enum values have an internal structure, namely each scheduling algorithm (e.g.) exists in four variants: unordered, orderend, normerge unordered, and nomerge ordered.

This patch (basically a followup to D114940) splits the "ordered" and "nomerge" bits into separate flags, as was already done for the "monotonic" and "nonmonotonic", so we can apply bit flags operations on them. It also now contains all possible combinations according to kmp's sched_type. Deriving of the OMPScheduleType enum from clause parameters has been moved form MLIR's OpenMPToLLVMIRTranslation.cpp to OpenMPIRBuilder to make available for clang as well. Since the primary purpose of the flag is the binary interface to libomp, it has been made more private to LLVMFrontend. The primary interface for generating worksharing-loop using OpenMPIRBuilder code becomes `applyWorkshareLoop` which derives the OMPScheduleType automatically and calls the appropriate emitter function.

While this is mostly a NFC refactor, it still applies the following functional changes:
 * The logic from OpenMPToLLVMIRTranslation to derive the OMPScheduleType also applies to clang. Most notably, it now applies the nonmonotonic flag for non-static schedules by default.
 * In OpenMPToLLVMIRTranslation, the nonmonotonic default flag was previously not applied if the simd modifier was used. I assume this was a bug, since the effect was due to `loop.schedule_modifier()` returning `mlir::omp::ScheduleModifier::none` instead of `llvm::Optional::None`.
 * In OpenMPToLLVMIRTranslation, the nonmonotonic default flag was set even if ordered was specified, in breach to what the comment before citing the OpenMP specification says. I assume this was an oversight.

The ordered flag with parameter was not considered in this patch. Changes will need to be made (e.g. adding/modifying function parameters) when support for it is added. The lengthy names of the enum values can be discussed, for the moment this is avoiding reusing previously existing enum value names such as `StaticChunked` to avoid confusion.

Reviewed By: peixin

Differential Revision: https://reviews.llvm.org/D123403

Author: Meinersbur

clang/lib/CodeGen/CGStmtOpenMP.cpp

@@ -3760,9 +3760,11 @@ void CodeGenFunction::EmitOMPForDirective(const OMPForDirective &S) {
           CGM.getOpenMPRuntime().getOMPBuilder();
       llvm::OpenMPIRBuilder::InsertPointTy AllocaIP(
           AllocaInsertPt->getParent(), AllocaInsertPt->getIterator());
-      OMPBuilder.applyWorkshareLoop(Builder.getCurrentDebugLocation(), CLI,
-                                    AllocaIP, NeedsBarrier, SchedKind,
-                                    ChunkSize);
+      OMPBuilder.applyWorkshareLoop(
+          Builder.getCurrentDebugLocation(), CLI, AllocaIP, NeedsBarrier,
+          SchedKind, ChunkSize, /*HasSimdModifier=*/false,
+          /*HasMonotonicModifier=*/false, /*HasNonmonotonicModifier=*/false,
+          /*HasOrderedClause=*/false);
       return;
     }
 

clang/test/OpenMP/irbuilder_for_unsigned_auto.c

@@ -38,7 +38,7 @@
 // CHECK-NEXT:    store i32 %[[DOTCOUNT]], i32* %[[P_UPPERBOUND]], align 4
 // CHECK-NEXT:    store i32 1, i32* %[[P_STRIDE]], align 4
 // CHECK-NEXT:    %[[OMP_GLOBAL_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
-// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 38, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 1)
+// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 1073741862, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 1)
 // CHECK-NEXT:    br label %[[OMP_LOOP_PREHEADER_OUTER_COND:.+]]
 // CHECK-EMPTY:
 // CHECK-NEXT:  [[OMP_LOOP_HEADER:.*]]:

clang/test/OpenMP/irbuilder_for_unsigned_dynamic.c

@@ -38,7 +38,7 @@
 // CHECK-NEXT:    store i32 %[[DOTCOUNT]], i32* %[[P_UPPERBOUND]], align 4
 // CHECK-NEXT:    store i32 1, i32* %[[P_STRIDE]], align 4
 // CHECK-NEXT:    %[[OMP_GLOBAL_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
-// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 35, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 1)
+// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 1073741859, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 1)
 // CHECK-NEXT:    br label %[[OMP_LOOP_PREHEADER_OUTER_COND:.+]]
 // CHECK-EMPTY:
 // CHECK-NEXT:  [[OMP_LOOP_HEADER:.*]]:

clang/test/OpenMP/irbuilder_for_unsigned_dynamic_chunked.c

@@ -38,7 +38,7 @@
 // CHECK-NEXT:    store i32 %[[DOTCOUNT]], i32* %[[P_UPPERBOUND]], align 4
 // CHECK-NEXT:    store i32 1, i32* %[[P_STRIDE]], align 4
 // CHECK-NEXT:    %[[OMP_GLOBAL_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
-// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 35, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 5)
+// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 1073741859, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 5)
 // CHECK-NEXT:    br label %[[OMP_LOOP_PREHEADER_OUTER_COND:.+]]
 // CHECK-EMPTY:
 // CHECK-NEXT:  [[OMP_LOOP_HEADER:.*]]:

clang/test/OpenMP/irbuilder_for_unsigned_runtime.c

@@ -38,7 +38,7 @@
 // CHECK-NEXT:    store i32 %[[DOTCOUNT]], i32* %[[P_UPPERBOUND]], align 4
 // CHECK-NEXT:    store i32 1, i32* %[[P_STRIDE]], align 4
 // CHECK-NEXT:    %[[OMP_GLOBAL_THREAD_NUM:.+]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* @1)
-// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 37, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 1)
+// CHECK-NEXT:    call void @__kmpc_dispatch_init_4u(%struct.ident_t* @1, i32 %[[OMP_GLOBAL_THREAD_NUM]], i32 1073741861, i32 1, i32 %[[DOTCOUNT]], i32 1, i32 1)
 // CHECK-NEXT:    br label %[[OMP_LOOP_PREHEADER_OUTER_COND:.+]]
 // CHECK-EMPTY:
 // CHECK-NEXT:  [[OMP_LOOP_HEADER:.*]]:

llvm/include/llvm/Frontend/OpenMP/OMPConstants.h

@@ -74,34 +74,114 @@ enum class IdentFlag {
 
 /// \note This needs to be kept in sync with kmp.h enum sched_type.
 /// Todo: Update kmp.h to include this file, and remove the enums in kmp.h
-///       To complete this, more enum values will need to be moved here.
 enum class OMPScheduleType {
-  StaticChunked = 33,
-  Static = 34, // static unspecialized
-  DynamicChunked = 35,
-  GuidedChunked = 36, // guided unspecialized
-  Runtime = 37,
-  Auto = 38, // auto
-
-  StaticBalancedChunked = 45, // static with chunk adjustment (e.g., simd)
-  GuidedSimd = 46,            // guided with chunk adjustment
-  RuntimeSimd = 47,           // runtime with chunk adjustment
-
-  OrderedStaticChunked = 65,
-  OrderedStatic = 66, // ordered static unspecialized
-  OrderedDynamicChunked = 67,
-  OrderedGuidedChunked = 68,
-  OrderedRuntime = 69,
-  OrderedAuto = 70, // ordered auto
-
-  DistributeChunked = 91, // distribute static chunked
-  Distribute = 92,        // distribute static unspecialized
-
-  ModifierMonotonic =
-      (1 << 29), // Set if the monotonic schedule modifier was present
-  ModifierNonmonotonic =
-      (1 << 30), // Set if the nonmonotonic schedule modifier was present
-  ModifierMask = ModifierMonotonic | ModifierNonmonotonic,
+  // For typed comparisons, not a valid schedule
+  None = 0,
+
+  // Schedule algorithms
+  BaseStaticChunked = 1,
+  BaseStatic = 2,
+  BaseDynamicChunked = 3,
+  BaseGuidedChunked = 4,
+  BaseRuntime = 5,
+  BaseAuto = 6,
+  BaseTrapezoidal = 7,
+  BaseGreedy = 8,
+  BaseBalanced = 9,
+  BaseGuidedIterativeChunked = 10,
+  BaseGuidedAnalyticalChunked = 11,
+  BaseSteal = 12,
+
+  // with chunk adjustment (e.g., simd)
+  BaseStaticBalancedChunked = 13,
+  BaseGuidedSimd = 14,
+  BaseRuntimeSimd = 15,
+
+  // static schedules algorithims for distribute
+  BaseDistributeChunked = 27,
+  BaseDistribute = 28,
+
+  // Modifier flags to be combined with schedule algorithms
+  ModifierUnordered = (1 << 5),
+  ModifierOrdered = (1 << 6),
+  ModifierNomerge = (1 << 7),
+  ModifierMonotonic = (1 << 29),
+  ModifierNonmonotonic = (1 << 30),
+
+  // Masks combining multiple flags
+  OrderingMask = ModifierUnordered | ModifierOrdered | ModifierNomerge,
+  MonotonicityMask = ModifierMonotonic | ModifierNonmonotonic,
+  ModifierMask = OrderingMask | MonotonicityMask,
+
+  // valid schedule type values, without monotonicity flags
+  UnorderedStaticChunked = BaseStaticChunked | ModifierUnordered,   // 33
+  UnorderedStatic = BaseStatic | ModifierUnordered,                 // 34
+  UnorderedDynamicChunked = BaseDynamicChunked | ModifierUnordered, // 35
+  UnorderedGuidedChunked = BaseGuidedChunked | ModifierUnordered,   // 36
+  UnorderedRuntime = BaseRuntime | ModifierUnordered,               // 37
+  UnorderedAuto = BaseAuto | ModifierUnordered,                     // 38
+  UnorderedTrapezoidal = BaseTrapezoidal | ModifierUnordered,       // 39
+  UnorderedGreedy = BaseGreedy | ModifierUnordered,                 // 40
+  UnorderedBalanced = BaseBalanced | ModifierUnordered,             // 41
+  UnorderedGuidedIterativeChunked =
+      BaseGuidedIterativeChunked | ModifierUnordered, // 42
+  UnorderedGuidedAnalyticalChunked =
+      BaseGuidedAnalyticalChunked | ModifierUnordered, // 43
+  UnorderedSteal = BaseSteal | ModifierUnordered,      // 44
+
+  UnorderedStaticBalancedChunked =
+      BaseStaticBalancedChunked | ModifierUnordered,          // 45
+  UnorderedGuidedSimd = BaseGuidedSimd | ModifierUnordered,   // 46
+  UnorderedRuntimeSimd = BaseRuntimeSimd | ModifierUnordered, // 47
+
+  OrderedStaticChunked = BaseStaticChunked | ModifierOrdered,   // 65
+  OrderedStatic = BaseStatic | ModifierOrdered,                 // 66
+  OrderedDynamicChunked = BaseDynamicChunked | ModifierOrdered, // 67
+  OrderedGuidedChunked = BaseGuidedChunked | ModifierOrdered,   // 68
+  OrderedRuntime = BaseRuntime | ModifierOrdered,               // 69
+  OrderedAuto = BaseAuto | ModifierOrdered,                     // 70
+  OrderdTrapezoidal = BaseTrapezoidal | ModifierOrdered,        // 71
+
+  OrderedDistributeChunked = BaseDistributeChunked | ModifierOrdered, // 91
+  OrderedDistribute = BaseDistribute | ModifierOrdered,               // 92
+
+  NomergeUnorderedStaticChunked =
+      BaseStaticChunked | ModifierUnordered | ModifierNomerge, // 161
+  NomergeUnorderedStatic =
+      BaseStatic | ModifierUnordered | ModifierNomerge, // 162
+  NomergeUnorderedDynamicChunked =
+      BaseDynamicChunked | ModifierUnordered | ModifierNomerge, // 163
+  NomergeUnorderedGuidedChunked =
+      BaseGuidedChunked | ModifierUnordered | ModifierNomerge, // 164
+  NomergeUnorderedRuntime =
+      BaseRuntime | ModifierUnordered | ModifierNomerge,                 // 165
+  NomergeUnorderedAuto = BaseAuto | ModifierUnordered | ModifierNomerge, // 166
+  NomergeUnorderedTrapezoidal =
+      BaseTrapezoidal | ModifierUnordered | ModifierNomerge, // 167
+  NomergeUnorderedGreedy =
+      BaseGreedy | ModifierUnordered | ModifierNomerge, // 168
+  NomergeUnorderedBalanced =
+      BaseBalanced | ModifierUnordered | ModifierNomerge, // 169
+  NomergeUnorderedGuidedIterativeChunked =
+      BaseGuidedIterativeChunked | ModifierUnordered | ModifierNomerge, // 170
+  NomergeUnorderedGuidedAnalyticalChunked =
+      BaseGuidedAnalyticalChunked | ModifierUnordered | ModifierNomerge, // 171
+  NomergeUnorderedSteal =
+      BaseSteal | ModifierUnordered | ModifierNomerge, // 172
+
+  NomergeOrderedStaticChunked =
+      BaseStaticChunked | ModifierOrdered | ModifierNomerge,             // 193
+  NomergeOrderedStatic = BaseStatic | ModifierOrdered | ModifierNomerge, // 194
+  NomergeOrderedDynamicChunked =
+      BaseDynamicChunked | ModifierOrdered | ModifierNomerge, // 195
+  NomergeOrderedGuidedChunked =
+      BaseGuidedChunked | ModifierOrdered | ModifierNomerge, // 196
+  NomergeOrderedRuntime =
+      BaseRuntime | ModifierOrdered | ModifierNomerge,               // 197
+  NomergeOrderedAuto = BaseAuto | ModifierOrdered | ModifierNomerge, // 198
+  NomergeOrderedTrapezoidal =
+      BaseTrapezoidal | ModifierOrdered | ModifierNomerge, // 199
+
   LLVM_MARK_AS_BITMASK_ENUM(/* LargestValue */ ModifierMask)
 };
 

llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h

@@ -344,6 +344,7 @@ class OpenMPIRBuilder {
                                    ArrayRef<CanonicalLoopInfo *> Loops,
                                    InsertPointTy ComputeIP);
 
+private:
   /// Modifies the canonical loop to be a statically-scheduled workshare loop.
   ///
   /// This takes a \p LoopInfo representing a canonical loop, such as the one
@@ -403,17 +404,15 @@ class OpenMPIRBuilder {
   ///                     the loop.
   /// \param Chunk    The size of loop chunk considered as a unit when
   ///                 scheduling. If \p nullptr, defaults to 1.
-  /// \param Ordered  Indicates whether the ordered clause is specified without
-  ///                 parameter.
   ///
   /// \returns Point where to insert code after the workshare construct.
   InsertPointTy applyDynamicWorkshareLoop(DebugLoc DL, CanonicalLoopInfo *CLI,
                                           InsertPointTy AllocaIP,
                                           omp::OMPScheduleType SchedType,
                                           bool NeedsBarrier,
-                                          Value *Chunk = nullptr,
-                                          bool Ordered = false);
+                                          Value *Chunk = nullptr);
 
+public:
   /// Modifies the canonical loop to be a workshare loop.
   ///
   /// This takes a \p LoopInfo representing a canonical loop, such as the one
@@ -436,13 +435,23 @@ class OpenMPIRBuilder {
   ///                     the loop.
   /// \param SchedKind Scheduling algorithm to use.
   /// \param ChunkSize The chunk size for the inner loop.
+  /// \param HasSimdModifier Whether the simd modifier is present in the
+  ///                        schedule clause.
+  /// \param HasMonotonicModifier Whether the monotonic modifier is present in
+  ///                             the schedule clause.
+  /// \param HasNonmonotonicModifier Whether the nonmonotonic modifier is
+  ///                                present in the schedule clause.
+  /// \param HasOrderedClause Whether the (parameterless) ordered clause is
+  ///                         present.
   ///
   /// \returns Point where to insert code after the workshare construct.
   InsertPointTy applyWorkshareLoop(
       DebugLoc DL, CanonicalLoopInfo *CLI, InsertPointTy AllocaIP,
       bool NeedsBarrier,
       llvm::omp::ScheduleKind SchedKind = llvm::omp::OMP_SCHEDULE_Default,
-      Value *ChunkSize = nullptr);
+      Value *ChunkSize = nullptr, bool HasSimdModifier = false,
+      bool HasMonotonicModifier = false, bool HasNonmonotonicModifier = false,
+      bool HasOrderedClause = false);
 
   /// Tile a loop nest.
   ///