PR155818 (llvm#4006)

Author: ronlieb

clang/include/clang/Driver/Options.td

@@ -3936,10 +3936,26 @@ let Visibility = [ClangOption, CC1Option, FC1Option, FlangOption] in {
 let Group = f_Group in {
 
 def fopenmp_target_debug_EQ : Joined<["-"], "fopenmp-target-debug=">;
-def fopenmp_assume_teams_oversubscription : Flag<["-"], "fopenmp-assume-teams-oversubscription">;
-def fopenmp_assume_threads_oversubscription : Flag<["-"], "fopenmp-assume-threads-oversubscription">;
-def fno_openmp_assume_teams_oversubscription : Flag<["-"], "fno-openmp-assume-teams-oversubscription">;
-def fno_openmp_assume_threads_oversubscription : Flag<["-"], "fno-openmp-assume-threads-oversubscription">;
+def fopenmp_assume_teams_oversubscription : Flag<["-"], "fopenmp-assume-teams-oversubscription">,
+  HelpText<"Allow the optimizer to discretely increase the number of "
+           "teams. May cause ignore environment variables that set "
+           "the number of teams to be ignored. The combination of "
+           "-fopenmp-assume-teams-oversubscription "
+           "and -fopenmp-assume-threads-oversubscription "
+           "may allow the conversion of loops into sequential code by "
+           "ensuring that each team/thread executes at most one iteration.">;
+def fopenmp_assume_threads_oversubscription : Flag<["-"], "fopenmp-assume-threads-oversubscription">,
+  HelpText<"Allow the optimizer to discretely increase the number of "
+           "threads. May cause ignore environment variables that set "
+           "the number of threads to be ignored. The combination of "
+           "-fopenmp-assume-teams-oversubscription "
+           "and -fopenmp-assume-threads-oversubscription "
+           "may allow the conversion of loops into sequential code by "
+           "ensuring that each team/thread executes at most one iteration.">;
+def fno_openmp_assume_teams_oversubscription : Flag<["-"], "fno-openmp-assume-teams-oversubscription">,
+  HelpText<"Do not assume teams oversubscription.">;
+def fno_openmp_assume_threads_oversubscription : Flag<["-"], "fno-openmp-assume-threads-oversubscription">,
+  HelpText<"Do not assume threads oversubscription.">;
 def fno_openmp_assume_no_thread_state : Flag<["-"], "fno-openmp-assume-no-thread-state">, Group<f_Group>,
   Flags<[NoArgumentUnused, HelpHidden]>, Visibility<[ClangOption, CC1Option]>,
   HelpText<"Assert that a thread in a parallel region may modify an ICV">,

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

@@ -1080,11 +1080,13 @@ class OpenMPIRBuilder {
   ///                 preheader of the loop.
   /// \param LoopType Information about type of loop worksharing.
   ///                 It corresponds to type of loop workshare OpenMP pragma.
+  /// \param NoLoop   If true, no-loop code is generated.
   ///
   /// \returns Point where to insert code after the workshare construct.
   InsertPointTy applyWorkshareLoopTarget(DebugLoc DL, CanonicalLoopInfo *CLI,
                                          InsertPointTy AllocaIP,
-                                         omp::WorksharingLoopType LoopType);
+                                         omp::WorksharingLoopType LoopType,
+                                         bool NoLoop);
 
   /// Modifies the canonical loop to be a statically-scheduled workshare loop.
   ///
@@ -1204,6 +1206,7 @@ class OpenMPIRBuilder {
   ///                         present.
   /// \param LoopType Information about type of loop worksharing.
   ///                 It corresponds to type of loop workshare OpenMP pragma.
+  /// \param NoLoop If true, no-loop code is generated.
   ///
   /// \returns Point where to insert code after the workshare construct.
   LLVM_ABI InsertPointOrErrorTy applyWorkshareLoop(
@@ -1214,7 +1217,8 @@ class OpenMPIRBuilder {
       bool HasMonotonicModifier = false, bool HasNonmonotonicModifier = false,
       bool HasOrderedClause = false,
       omp::WorksharingLoopType LoopType =
-          omp::WorksharingLoopType::ForStaticLoop);
+          omp::WorksharingLoopType::ForStaticLoop,
+      bool NoLoop = false);
 
   /// Tile a loop nest.
   ///

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

@@ -5010,7 +5010,7 @@ static void createTargetLoopWorkshareCall(OpenMPIRBuilder *OMPBuilder,
                                           WorksharingLoopType LoopType,
                                           BasicBlock *InsertBlock, Value *Ident,
                                           Value *LoopBodyArg, Value *TripCount,
-                                          Function &LoopBodyFn) {
+                                          Function &LoopBodyFn, bool NoLoop) {
   Type *TripCountTy = TripCount->getType();
   Module &M = OMPBuilder->M;
   IRBuilder<> &Builder = OMPBuilder->Builder;
@@ -5038,16 +5038,18 @@ static void createTargetLoopWorkshareCall(OpenMPIRBuilder *OMPBuilder,
   RealArgs.push_back(ConstantInt::get(TripCountTy, 0));
   if (LoopType == WorksharingLoopType::DistributeForStaticLoop) {
     RealArgs.push_back(ConstantInt::get(TripCountTy, 0));
+    RealArgs.push_back(ConstantInt::get(Builder.getInt8Ty(), NoLoop));
+  } else {
+    RealArgs.push_back(ConstantInt::get(Builder.getInt8Ty(), 0));
   }
-  RealArgs.push_back(ConstantInt::get(Builder.getInt8Ty(), 0));
 
   Builder.CreateCall(RTLFn, RealArgs);
 }
 
 static void workshareLoopTargetCallback(
     OpenMPIRBuilder *OMPIRBuilder, CanonicalLoopInfo *CLI, Value *Ident,
     Function &OutlinedFn, const SmallVector<Instruction *, 4> &ToBeDeleted,
-    WorksharingLoopType LoopType) {
+    WorksharingLoopType LoopType, bool NoLoop) {
   IRBuilder<> &Builder = OMPIRBuilder->Builder;
   BasicBlock *Preheader = CLI->getPreheader();
   Value *TripCount = CLI->getTripCount();
@@ -5094,17 +5096,16 @@ static void workshareLoopTargetCallback(
   OutlinedFnCallInstruction->eraseFromParent();
 
   createTargetLoopWorkshareCall(OMPIRBuilder, LoopType, Preheader, Ident,
-                                LoopBodyArg, TripCount, OutlinedFn);
+                                LoopBodyArg, TripCount, OutlinedFn, NoLoop);
 
   for (auto &ToBeDeletedItem : ToBeDeleted)
     ToBeDeletedItem->eraseFromParent();
   CLI->invalidate();
 }
 
-OpenMPIRBuilder::InsertPointTy
-OpenMPIRBuilder::applyWorkshareLoopTarget(DebugLoc DL, CanonicalLoopInfo *CLI,
-                                          InsertPointTy AllocaIP,
-                                          WorksharingLoopType LoopType) {
+OpenMPIRBuilder::InsertPointTy OpenMPIRBuilder::applyWorkshareLoopTarget(
+    DebugLoc DL, CanonicalLoopInfo *CLI, InsertPointTy AllocaIP,
+    WorksharingLoopType LoopType, bool NoLoop) {
   uint32_t SrcLocStrSize;
   Constant *SrcLocStr = getOrCreateSrcLocStr(DL, SrcLocStrSize);
   Value *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
@@ -5187,7 +5188,7 @@ OpenMPIRBuilder::applyWorkshareLoopTarget(DebugLoc DL, CanonicalLoopInfo *CLI,
   OI.PostOutlineCB = [=, ToBeDeletedVec =
                              std::move(ToBeDeleted)](Function &OutlinedFn) {
     workshareLoopTargetCallback(this, CLI, Ident, OutlinedFn, ToBeDeletedVec,
-                                LoopType);
+                                LoopType, NoLoop);
   };
   addOutlineInfo(std::move(OI));
   return CLI->getAfterIP();
@@ -5198,9 +5199,9 @@ OpenMPIRBuilder::InsertPointOrErrorTy OpenMPIRBuilder::applyWorkshareLoop(
     bool NeedsBarrier, omp::ScheduleKind SchedKind, Value *ChunkSize,
     bool HasSimdModifier, bool HasMonotonicModifier,
     bool HasNonmonotonicModifier, bool HasOrderedClause,
-    WorksharingLoopType LoopType) {
+    WorksharingLoopType LoopType, bool NoLoop) {
   if (Config.isTargetDevice())
-    return applyWorkshareLoopTarget(DL, CLI, AllocaIP, LoopType);
+    return applyWorkshareLoopTarget(DL, CLI, AllocaIP, LoopType, NoLoop);
   OMPScheduleType EffectiveScheduleType = computeOpenMPScheduleType(
       SchedKind, ChunkSize, HasSimdModifier, HasMonotonicModifier,
       HasNonmonotonicModifier, HasOrderedClause);

mlir/include/mlir/Dialect/OpenMP/OpenMPEnums.td

@@ -232,14 +232,24 @@ def TargetRegionFlagsNone : I32BitEnumAttrCaseNone<"none">;
 def TargetRegionFlagsGeneric : I32BitEnumAttrCaseBit<"generic", 0>;
 def TargetRegionFlagsSpmd : I32BitEnumAttrCaseBit<"spmd", 1>;
 def TargetRegionFlagsTripCount : I32BitEnumAttrCaseBit<"trip_count", 2>;
+def TargetRegionFlagsNoLoop : I32BitEnumAttrCaseBit<"no_loop", 3>;
 
 def TargetRegionFlags : OpenMP_BitEnumAttr<
     "TargetRegionFlags",
-    "target region property flags", [
+    "These flags describe properties of the target kernel. "
+    "TargetRegionFlagsGeneric - denotes generic kernel. "
+    "TargetRegionFlagsSpmd - denotes SPMD kernel. "
+    "TargetRegionFlagsNoLoop - denotes kernel where "
+    "num_teams * num_threads >= loop_trip_count. It allows the conversion "
+    "of loops into sequential code by ensuring that each team/thread "
+    "executes at most one iteration. "
+    "TargetRegionFlagsTripCount - checks if the loop trip count should be "
+    "calculated.", [
       TargetRegionFlagsNone,
       TargetRegionFlagsGeneric,
       TargetRegionFlagsSpmd,
-      TargetRegionFlagsTripCount
+      TargetRegionFlagsTripCount,
+      TargetRegionFlagsNoLoop
     ]>;
 
 //===----------------------------------------------------------------------===//

mlir/lib/Dialect/OpenMP/IR/OpenMPDialect.cpp

@@ -2116,6 +2116,31 @@ Operation *TargetOp::getInnermostCapturedOmpOp() {
   });
 }
 
+/// Check if we can promote SPMD kernel to No-Loop kernel.
+static bool canPromoteToNoLoop(Operation *capturedOp, TeamsOp teamsOp,
+                               WsloopOp *wsLoopOp) {
+  // num_teams clause can break no-loop teams/threads assumption.
+  if (teamsOp.getNumTeamsUpper())
+    return false;
+
+  // Reduction kernels are slower in no-loop mode.
+  if (teamsOp.getNumReductionVars())
+    return false;
+  if (wsLoopOp->getNumReductionVars())
+    return false;
+
+  // Check if the user allows the promotion of kernels to no-loop mode.
+  OffloadModuleInterface offloadMod =
+      capturedOp->getParentOfType<omp::OffloadModuleInterface>();
+  if (!offloadMod)
+    return false;
+  auto ompFlags = offloadMod.getFlags();
+  if (!ompFlags)
+    return false;
+  return ompFlags.getAssumeTeamsOversubscription() &&
+         ompFlags.getAssumeThreadsOversubscription();
+}
+
 TargetRegionFlags TargetOp::getKernelExecFlags(Operation *capturedOp) {
   // A non-null captured op is only valid if it resides inside of a TargetOp
   // and is the result of calling getInnermostCapturedOmpOp() on it.
@@ -2144,7 +2169,8 @@ TargetRegionFlags TargetOp::getKernelExecFlags(Operation *capturedOp) {
 
   // Detect target-teams-distribute-parallel-wsloop[-simd].
   if (numWrappers == 2) {
-    if (!isa<WsloopOp>(innermostWrapper))
+    WsloopOp *wsloopOp = dyn_cast<WsloopOp>(innermostWrapper);
+    if (!wsloopOp)
       return TargetRegionFlags::generic;
 
     innermostWrapper = std::next(innermostWrapper);
@@ -2155,12 +2181,17 @@ TargetRegionFlags TargetOp::getKernelExecFlags(Operation *capturedOp) {
     if (!isa_and_present<ParallelOp>(parallelOp))
       return TargetRegionFlags::generic;
 
-    Operation *teamsOp = parallelOp->getParentOp();
-    if (!isa_and_present<TeamsOp>(teamsOp))
+    TeamsOp teamsOp = dyn_cast<TeamsOp>(parallelOp->getParentOp());
+    if (!teamsOp)
       return TargetRegionFlags::generic;
 
-    if (teamsOp->getParentOp() == targetOp.getOperation())
-      return TargetRegionFlags::spmd | TargetRegionFlags::trip_count;
+    if (teamsOp->getParentOp() == targetOp.getOperation()) {
+      TargetRegionFlags result =
+          TargetRegionFlags::spmd | TargetRegionFlags::trip_count;
+      if (canPromoteToNoLoop(capturedOp, teamsOp, wsloopOp))
+        result = result | TargetRegionFlags::no_loop;
+      return result;
+    }
   }
   // Detect target-teams-distribute[-simd] and target-teams-loop.
   else if (isa<DistributeOp, LoopOp>(innermostWrapper)) {

mlir/lib/Target/LLVMIR/Dialect/OpenMP/OpenMPToLLVMIRTranslation.cpp

@@ -2593,13 +2593,34 @@ convertOmpWsloop(Operation &opInst, llvm::IRBuilderBase &builder,
   }
 
   builder.SetInsertPoint(*regionBlock, (*regionBlock)->begin());
+
+  // Check if we can generate no-loop kernel
+  bool noLoopMode = false;
+  omp::TargetOp targetOp = wsloopOp->getParentOfType<mlir::omp::TargetOp>();
+  if (targetOp) {
+    Operation *targetCapturedOp = targetOp.getInnermostCapturedOmpOp();
+    // We need this check because, without it, noLoopMode would be set to true
+    // for every omp.wsloop nested inside a no-loop SPMD target region, even if
+    // that loop is not the top-level SPMD one.
+    if (loopOp == targetCapturedOp) {
+      omp::TargetRegionFlags kernelFlags =
+          targetOp.getKernelExecFlags(targetCapturedOp);
+      if (omp::bitEnumContainsAll(kernelFlags,
+                                  omp::TargetRegionFlags::spmd |
+                                      omp::TargetRegionFlags::no_loop) &&
+          !omp::bitEnumContainsAny(kernelFlags,
+                                   omp::TargetRegionFlags::generic))
+        noLoopMode = true;
+    }
+  }
+
   llvm::OpenMPIRBuilder::InsertPointOrErrorTy wsloopIP =
       ompBuilder->applyWorkshareLoop(
           ompLoc.DL, loopInfo, allocaIP, loopNeedsBarrier,
           convertToScheduleKind(schedule), chunk, isSimd,
           scheduleMod == omp::ScheduleModifier::monotonic,
           scheduleMod == omp::ScheduleModifier::nonmonotonic, isOrdered,
-          workshareLoopType);
+          workshareLoopType, noLoopMode);
 
   if (failed(handleError(wsloopIP, opInst)))
     return failure();
@@ -5683,6 +5704,12 @@ initTargetDefaultAttrs(omp::TargetOp targetOp, Operation *capturedOp,
                 ? llvm::omp::OMP_TGT_EXEC_MODE_GENERIC_SPMD
                 : llvm::omp::OMP_TGT_EXEC_MODE_GENERIC
           : llvm::omp::OMP_TGT_EXEC_MODE_SPMD;
+  if (omp::bitEnumContainsAll(kernelFlags,
+                              omp::TargetRegionFlags::spmd |
+                                  omp::TargetRegionFlags::no_loop) &&
+      !omp::bitEnumContainsAny(kernelFlags, omp::TargetRegionFlags::generic))
+    attrs.ExecFlags = llvm::omp::OMP_TGT_EXEC_MODE_SPMD_NO_LOOP;
+
   attrs.MinTeams = minTeamsVal;
   attrs.MaxTeams.front() = maxTeamsVal;
   attrs.MinThreads = 1;

offload/test/offloading/fortran/target-no-loop.f90

@@ -0,0 +1,96 @@
+! REQUIRES: flang
+
+! RUN: %libomptarget-compile-fortran-generic -O3  -fopenmp-assume-threads-oversubscription -fopenmp-assume-teams-oversubscription
+! RUN: env LIBOMPTARGET_INFO=16 OMP_NUM_TEAMS=16 OMP_TEAMS_THREAD_LIMIT=16 %libomptarget-run-generic 2>&1 | %fcheck-generic
+function check_errors(array) result (errors)
+   integer, intent(in) :: array(1024)
+   integer :: errors
+   integer :: i
+   errors = 0
+   do i = 1, 1024
+      if ( array( i) .ne. (i) ) then
+         errors = errors + 1
+      end if
+   end do
+end function
+
+program main
+   use omp_lib
+   implicit none
+   integer :: i,j,red
+   integer :: array(1024), errors = 0
+   array = 1
+
+   ! No-loop kernel
+   !$omp target teams distribute parallel do
+   do i = 1, 1024
+      array(i) = i
+   end do
+   errors = errors + check_errors(array)
+
+   ! SPMD kernel (num_teams clause blocks promotion to no-loop)
+   array = 1
+   !$omp target teams distribute parallel do num_teams(3)
+   do i = 1, 1024
+      array(i) = i
+   end do
+
+   errors = errors + check_errors(array)
+
+   ! No-loop kernel
+   array = 1
+   !$omp target teams distribute parallel do num_threads(64)
+   do i = 1, 1024
+      array(i) = i
+    end do
+
+   errors = errors + check_errors(array)
+
+   ! SPMD kernel
+   array = 1
+   !$omp target parallel do
+   do i = 1, 1024
+      array(i) = i
+   end do
+
+   errors = errors + check_errors(array)
+
+   ! Generic kernel
+   array = 1
+   !$omp target teams distribute
+   do i = 1, 1024
+      array(i) = i
+   end do
+
+   errors = errors + check_errors(array)
+
+   ! SPMD kernel (reduction clause blocks promotion to no-loop)
+   array = 1
+   red =0
+   !$omp target teams distribute parallel do reduction(+:red)
+   do i = 1, 1024
+      red = red + array(i)
+   end do
+
+   if (red .ne. 1024) then
+      errors = errors + 1
+   end if
+
+   print *,"number of errors: ", errors
+
+end program main
+
+! CHECK:  "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}} SPMD-No-Loop mode
+! CHECK:  info: #Args: 3 Teams x Thrds:   64x  16
+! CHECK:  "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}} SPMD mode
+! CHECK:  info: #Args: 3 Teams x Thrds:   3x  16 {{.*}}
+! CHECK:  "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}} SPMD-No-Loop mode
+! CHECK:  info: #Args: 3 Teams x Thrds:   64x  16 {{.*}}
+! CHECK:  "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}} SPMD mode
+! CHECK:  info: #Args: 3 Teams x Thrds:   1x  16
+! CHECK:  "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}} Generic mode
+! CHECK:  info: #Args: 3 Teams x Thrds:   16x  16 {{.*}}
+! CHECK:  "PluginInterface" device {{[0-9]+}} info: Launching kernel {{.*}} SPMD mode
+! CHECK:  info: #Args: 4 Teams x Thrds:   16x  16 {{.*}}
+! CHECK:  number of errors: 0
+