[OpenMP][NFC] Add min/max threads/teams count into the KernelEnvironm…

…ent (#70257)

The runtime needs to know about the acceptable launch bounds, especially
if the compiler (middle- or backend) assumed those bounds. While this
patch does not yet inform the runtime, it stores the bounds in a place
that can/will be accessed and is associated with the kernel.

llvm/include/llvm/Frontend/OpenMP/OMPKinds.def

@@ -95,7 +95,7 @@ __OMP_STRUCT_TYPE(AsyncInfo, __tgt_async_info, false, Int8Ptr)
 __OMP_STRUCT_TYPE(DependInfo, kmp_dep_info, false, SizeTy, SizeTy, Int8)
 __OMP_STRUCT_TYPE(Task, kmp_task_ompbuilder_t, false, VoidPtr, VoidPtr, Int32, VoidPtr, VoidPtr)
 __OMP_STRUCT_TYPE(ConfigurationEnvironment, ConfigurationEnvironmentTy, false,
-                  Int8, Int8, Int8)
+                  Int8, Int8, Int8, Int32, Int32, Int32, Int32)
 __OMP_STRUCT_TYPE(DynamicEnvironment, DynamicEnvironmentTy, false, Int16)
 __OMP_STRUCT_TYPE(KernelEnvironment, KernelEnvironmentTy, false,
                   ConfigurationEnvironment, IdentPtr, DynamicEnvironmentPtr)

llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp

@@ -4071,18 +4071,21 @@ OpenMPIRBuilder::createTargetInit(const LocationDescription &Loc, bool IsSPMD) {
   uint32_t SrcLocStrSize;
   Constant *SrcLocStr = getOrCreateSrcLocStr(Loc, SrcLocStrSize);
   Constant *Ident = getOrCreateIdent(SrcLocStr, SrcLocStrSize);
-  ConstantInt *IsSPMDVal = ConstantInt::getSigned(
-      IntegerType::getInt8Ty(Int8->getContext()),
-      IsSPMD ? OMP_TGT_EXEC_MODE_SPMD : OMP_TGT_EXEC_MODE_GENERIC);
-  ConstantInt *UseGenericStateMachineVal = ConstantInt::getSigned(
-      IntegerType::getInt8Ty(Int8->getContext()), !IsSPMD);
-  ConstantInt *MayUseNestedParallelismVal =
-      ConstantInt::getSigned(IntegerType::getInt8Ty(Int8->getContext()), true);
-  ConstantInt *DebugIndentionLevelVal =
-      ConstantInt::getSigned(IntegerType::getInt16Ty(Int8->getContext()), 0);
+  Constant *IsSPMDVal = ConstantInt::getSigned(
+      Int8, IsSPMD ? OMP_TGT_EXEC_MODE_SPMD : OMP_TGT_EXEC_MODE_GENERIC);
+  Constant *UseGenericStateMachineVal = ConstantInt::getSigned(Int8, !IsSPMD);
+  Constant *MayUseNestedParallelismVal = ConstantInt::getSigned(Int8, true);
+  Constant *DebugIndentionLevelVal = ConstantInt::getSigned(Int16, 0);
 
-  // We need to strip the debug prefix to get the correct kernel name.
   Function *Kernel = Builder.GetInsertBlock()->getParent();
+  auto [MinThreadsVal, MaxThreadsVal] = readThreadBoundsForKernel(*Kernel);
+  auto [MinTeamsVal, MaxTeamsVal] = readTeamBoundsForKernel(*Kernel);
+  Constant *MinThreads = ConstantInt::getSigned(Int32, MinThreadsVal);
+  Constant *MaxThreads = ConstantInt::getSigned(Int32, MaxThreadsVal);
+  Constant *MinTeams = ConstantInt::getSigned(Int32, MinTeamsVal);
+  Constant *MaxTeams = ConstantInt::getSigned(Int32, MaxTeamsVal);
+
+  // We need to strip the debug prefix to get the correct kernel name.
   StringRef KernelName = Kernel->getName();
   const std::string DebugPrefix = "_debug__";
   if (KernelName.ends_with(DebugPrefix))
@@ -4113,6 +4116,10 @@ OpenMPIRBuilder::createTargetInit(const LocationDescription &Loc, bool IsSPMD) {
                                     UseGenericStateMachineVal,
                                     MayUseNestedParallelismVal,
                                     IsSPMDVal,
+                                    MinThreads,
+                                    MaxThreads,
+                                    MinTeams,
+                                    MaxTeams,
                                 });
   Constant *KernelEnvironmentInitializer = ConstantStruct::get(
       KernelEnvironment, {

llvm/lib/Transforms/IPO/OpenMPOpt.cpp

@@ -191,6 +191,10 @@ namespace KernelInfo {
 //   uint8_t UseGenericStateMachine;
 //   uint8_t MayUseNestedParallelism;
 //   llvm::omp::OMPTgtExecModeFlags ExecMode;
+//   int32_t MinThreads;
+//   int32_t MaxThreads;
+//   int32_t MinTeams;
+//   int32_t MaxTeams;
 // };
 
 // struct DynamicEnvironmentTy {
@@ -217,6 +221,10 @@ KERNEL_ENVIRONMENT_IDX(Ident, 1)
 KERNEL_ENVIRONMENT_CONFIGURATION_IDX(UseGenericStateMachine, 0)
 KERNEL_ENVIRONMENT_CONFIGURATION_IDX(MayUseNestedParallelism, 1)
 KERNEL_ENVIRONMENT_CONFIGURATION_IDX(ExecMode, 2)
+KERNEL_ENVIRONMENT_CONFIGURATION_IDX(MinThreads, 3)
+KERNEL_ENVIRONMENT_CONFIGURATION_IDX(MaxThreads, 4)
+KERNEL_ENVIRONMENT_CONFIGURATION_IDX(MinTeams, 5)
+KERNEL_ENVIRONMENT_CONFIGURATION_IDX(MaxTeams, 6)
 
 #undef KERNEL_ENVIRONMENT_CONFIGURATION_IDX
 
@@ -241,6 +249,10 @@ KERNEL_ENVIRONMENT_GETTER(Configuration, ConstantStruct)
 KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(UseGenericStateMachine)
 KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(MayUseNestedParallelism)
 KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(ExecMode)
+KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(MinThreads)
+KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(MaxThreads)
+KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(MinTeams)
+KERNEL_ENVIRONMENT_CONFIGURATION_GETTER(MaxTeams)
 
 #undef KERNEL_ENVIRONMENT_CONFIGURATION_GETTER
 
@@ -3636,6 +3648,10 @@ struct AAKernelInfoFunction : AAKernelInfo {
   KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(UseGenericStateMachine)
   KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(MayUseNestedParallelism)
   KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(ExecMode)
+  KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(MinThreads)
+  KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(MaxThreads)
+  KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(MinTeams)
+  KERNEL_ENVIRONMENT_CONFIGURATION_SETTER(MaxTeams)
 
 #undef KERNEL_ENVIRONMENT_CONFIGURATION_SETTER
 
@@ -3723,6 +3739,19 @@ struct AAKernelInfoFunction : AAKernelInfo {
     else
       setExecModeOfKernelEnvironment(AssumedExecModeC);
 
+    auto *Int32Ty = Type::getInt32Ty(Fn->getContext());
+    auto [MinThreads, MaxThreads] =
+        OpenMPIRBuilder::readThreadBoundsForKernel(*Fn);
+    if (MinThreads)
+      setMinThreadsOfKernelEnvironment(ConstantInt::get(Int32Ty, MinThreads));
+    if (MaxThreads)
+      setMaxThreadsOfKernelEnvironment(ConstantInt::get(Int32Ty, MaxThreads));
+    auto [MinTeams, MaxTeams] = OpenMPIRBuilder::readTeamBoundsForKernel(*Fn);
+    if (MinTeams)
+      setMinTeamsOfKernelEnvironment(ConstantInt::get(Int32Ty, MinTeams));
+    if (MaxTeams)
+      setMaxTeamsOfKernelEnvironment(ConstantInt::get(Int32Ty, MaxTeams));
+
     ConstantInt *MayUseNestedParallelismC =
         KernelInfo::getMayUseNestedParallelismFromKernelEnvironment(KernelEnvC);
     ConstantInt *AssumedMayUseNestedParallelismC = ConstantInt::get(

llvm/test/Transforms/OpenMP/always_inline_device.ll

@@ -3,19 +3,19 @@
 
 %struct.ident_t = type { i32, i32, i32, i32, ptr }
 %struct.KernelEnvironmentTy = type { %struct.ConfigurationEnvironmentTy, ptr, ptr }
-%struct.ConfigurationEnvironmentTy = type { i8, i8, i8 }
+%struct.ConfigurationEnvironmentTy = type { i8, i8, i8, i32, i32, i32, i32 }
 @0 = private unnamed_addr constant [23 x i8] c";unknown;unknown;0;0;;\00", align 1
 @1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, ptr @0 }, align 8
 @G = external global i8
 
-@kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null }
+@kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @1, ptr null }
 
 ; Function Attrs: convergent norecurse nounwind
 ;.
 ; CHECK: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c"
 ; CHECK: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8
 ; CHECK: @[[G:[a-zA-Z0-9_$"\\.-]+]] = external global i8
-; CHECK: @[[KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 3 }, ptr @[[GLOB1]], ptr null }
+; CHECK: @[[KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 3, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null }
 ;.
 define weak void @__omp_offloading_fd02_c0934fc2_foo_l4() #0 {
 ; CHECK: Function Attrs: norecurse nounwind