[OpenMP][NFC] Add min/max threads/teams count into the KernelEnvironment #70257
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters. Learn more about bidirectional Unicode characters
llvmbot added flang:openmp llvm:transforms clang:openmp 
Member
| @llvm/pr-subscribers-llvm-transforms @llvm/pr-subscribers-flang-openmp Author: Johannes Doerfert (jdoerfert) ChangesThe 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. Patch is 166.78 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/70257.diff 31 Files Affected:
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def index 4823c4cc6b833ec..a06ec62350afae9 100644 --- a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def +++ b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def @@ -97,7 +97,7 @@ __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) diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp index 5b24e9fe2e0c5bd..e3880eb39d82d7f 100644 --- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp +++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp @@ -4108,18 +4108,21 @@ 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)) @@ -4150,6 +4153,10 @@ OpenMPIRBuilder::createTargetInit(const LocationDescription &Loc, bool IsSPMD) { UseGenericStateMachineVal, MayUseNestedParallelismVal, IsSPMDVal, + MinThreads, + MinTeams, + MaxThreads, + MaxTeams, }); Constant *KernelEnvironmentInitializer = ConstantStruct::get( KernelEnvironment, { diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp index e50e74ea6c0d5aa..e9dbcd334321f34 100644 --- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp +++ b/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( diff --git a/llvm/test/Transforms/OpenMP/always_inline_device.ll b/llvm/test/Transforms/OpenMP/always_inline_device.ll index 5478924cf828b68..993f097d741188c 100644 --- a/llvm/test/Transforms/OpenMP/always_inline_device.ll +++ b/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 diff --git a/llvm/test/Transforms/OpenMP/custom_state_machines.ll b/llvm/test/Transforms/OpenMP/custom_state_machines.ll index 9eefb4d5fa509e6..89a298ff195493c 100644 --- a/llvm/test/Transforms/OpenMP/custom_state_machines.ll +++ b/llvm/test/Transforms/OpenMP/custom_state_machines.ll @@ -121,7 +121,7 @@ %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 @@ -129,14 +129,14 @@ @G = external global i32, align 4 @3 = private unnamed_addr constant %struct.ident_t { i32 0, i32 322, i32 2, i32 0, ptr @0 }, align 8 -@__omp_offloading_14_a36502b_no_state_machine_needed_l14_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_simple_state_machine_l22_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_simple_state_machine_pure_l77_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } -@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } +@__omp_offloading_14_a36502b_no_state_machine_needed_l14_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 } +@__omp_offloading_14_a36502b_simple_state_machine_l22_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 } +@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_l39_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 } +@__omp_offloading_14_a36502b_simple_state_machine_with_fallback_l55_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 } +@__omp_offloading_14_a36502b_simple_state_machine_no_openmp_attr_l66_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 } +@__omp_offloading_14_a36502b_simple_state_machine_pure_l77_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 } +@__omp_offloading_14_a36502b_simple_state_machine_interprocedural_nested_recursive_l92_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 } +@__omp_offloading_14_a36502b_no_state_machine_weak_callee_l112_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 } define weak void @__omp_offloading_14_a36502b_no_state_machine_needed_l14() #0 { entry: @@ -840,14 +840,14 @@ attributes #9 = { convergent nounwind readonly willreturn } ; AMDGPU: @[[GLOB2:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 2, i32 0, ptr @[[GLOB0]] }, align 8 ; AMDGPU: @[[G:[a-zA-Z0-9_$"\\.-]+]] = external global i32, align 4 ; AMDGPU: @[[GLOB3:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 322, i32 2, i32 0, ptr @[[GLOB0]] }, align 8 -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_NO_STATE_MACHINE_NEEDED_L14_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_L22_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_INTERPROCEDURAL_L39_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_WITH_FALLBACK_L55_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_NO_OPENMP_ATTR_L66_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_PURE_L77_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_INTERPROCEDURAL_NESTED_RECURSIVE_L92_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } -; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_NO_STATE_MACHINE_WEAK_CALLEE_L112_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_NO_STATE_MACHINE_NEEDED_L14_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_L22_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_INTERPROCEDURAL_L39_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_WITH_FALLBACK_L55_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_NO_OPENMP_ATTR_L66_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_PURE_L77_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_INTERPROCEDURAL_NESTED_RECURSIVE_L92_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } +; AMDGPU: @[[__OMP_OFFLOADING_14_A36502B_NO_STATE_MACHINE_WEAK_CALLEE_L112_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1, i32 0, i32 0, i32 0, i32 0 }, ptr @[[GLOB1]], ptr null } ; AMDGPU: @[[__OMP_OUTLINED__2_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef ; AMDGPU: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef ; AMDGPU: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef @@ -863,14 +863,14 @@ attributes #9 = { convergent nounwind readonly willreturn } ; NVPTX: @[[GLOB2:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 2, i32 0, ptr @[[GLOB0]] }, align 8 ; NVPTX: @[[G:[a-zA-Z0-9_$"\\.-]+]] = external global i32, align 4 ; NVPTX: @[[GLOB3:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 322, i32 2, i32 0, ptr @[[GLOB0]] }, align 8 -; NVPTX: @[[__OMP_OFFLOADING_14_A36502B_NO_STATE_MACHINE_NEEDED_L14_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } -; NVPTX: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_L22_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; NVPTX: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_INTERPROCEDURAL_L39_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; NVPTX: @[[__OMP_OFFLOADING_14_A36502B_SIMPLE_STATE_MACHINE_WITH_FALLBACK_L55_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } -; NVPTX: @[[__OMP_OFFLOADING_14_A365... [truncated] |
jhuber6 approved these changes Oct 25, 2023
Contributor
jhuber6 left a comment
jhuber6 left a comment There was a problem hiding this comment.
Trivial change, lots of tests impacted per the norm
Member Author
I take this as a good thing, honestly, even if I don't have tooling to do the update and it takes a little longer. |
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.
jdoerfert force-pushed the min_max_threads_teams_kernel_env branch from 336b4d0 to 2702498 Compare 
Collaborator
| Seems like this broke the MLIR bot, the |
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit. This suggestion is invalid because no changes were made to the code. Suggestions cannot be applied while the pull request is closed. Suggestions cannot be applied while viewing a subset of changes. Only one suggestion per line can be applied in a batch. Add this suggestion to a batch that can be applied as a single commit. Applying suggestions on deleted lines is not supported. You must change the existing code in this line in order to create a valid suggestion. Outdated suggestions cannot be applied. This suggestion has been applied or marked resolved. Suggestions cannot be applied from pending reviews. Suggestions cannot be applied on multi-line comments. Suggestions cannot be applied while the pull request is queued to merge. Suggestion cannot be applied right now. Please check back later.
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.