| ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-globals |
| ; RUN: opt --mtriple=amdgcn-amd-amdhsa --data-layout=A5 -S -passes=openmp-opt < %s | FileCheck %s --check-prefixes=AMDGPU |
| ; RUN: opt --mtriple=nvptx64-- -S -passes=openmp-opt < %s | FileCheck %s --check-prefixes=NVPTX |
| ; RUN: opt --mtriple=amdgcn-amd-amdhsa --data-layout=A5 -S -passes=openmp-opt -openmp-opt-disable-spmdization < %s | FileCheck %s --check-prefix=AMDGPU-DISABLED1 |
| ; RUN: opt --mtriple=amdgcn-amd-amdhsa --data-layout=A5 -S -passes=openmp-opt-postlink < %s | FileCheck %s --check-prefix=AMDGPU-DISABLED2 |
| ; RUN: opt --mtriple=nvptx64-- -S -passes=openmp-opt -openmp-opt-disable-spmdization < %s | FileCheck %s --check-prefix=NVPTX-DISABLED1 |
| ; RUN: opt --mtriple=nvptx64-- -S -passes=openmp-opt-postlink < %s | FileCheck %s --check-prefix=NVPTX-DISABLED2 |
| |
| ;; void unknown(void); |
| ;; void spmd_amenable(void) __attribute__((assume("ompx_spmd_amenable"))); |
| ;; |
| ;; void sequential_loop() { |
| ;; #pragma omp target teams |
| ;; { |
| ;; for (int i = 0; i < 100; ++i) { |
| ;; #pragma omp parallel |
| ;; { |
| ;; unknown(); |
| ;; } |
| ;; } |
| ;; spmd_amenable(); |
| ;; } |
| ;; } |
| ;; |
| ;; void use(__attribute__((noescape)) int *) __attribute__((assume("ompx_spmd_amenable"))); |
| ;; |
| ;; void sequential_loop_to_stack_var() { |
| ;; #pragma omp target teams |
| ;; { |
| ;; int x; |
| ;; use(&x); |
| ;; for (int i = 0; i < 100; ++i) { |
| ;; #pragma omp parallel |
| ;; { |
| ;; unknown(); |
| ;; } |
| ;; } |
| ;; spmd_amenable(); |
| ;; } |
| ;; } |
| ;; |
| ;; void sequential_loop_to_shared_var() { |
| ;; #pragma omp target teams |
| ;; { |
| ;; int x; |
| ;; for (int i = 0; i < 100; ++i) { |
| ;; #pragma omp parallel |
| ;; { |
| ;; x++; |
| ;; unknown(); |
| ;; } |
| ;; } |
| ;; spmd_amenable(); |
| ;; } |
| ;; } |
| ;; |
| ;; void sequential_loop_to_shared_var_guarded() { |
| ;; #pragma omp target teams |
| ;; { |
| ;; int x = 42; |
| ;; for (int i = 0; i < 100; ++i) { |
| ;; #pragma omp parallel |
| ;; { |
| ;; x++; |
| ;; unknown(); |
| ;; } |
| ;; } |
| ;; spmd_amenable(); |
| ;; } |
| ;; } |
| ;; |
| ;; void do_not_spmdize_target() { |
| ;; #pragma omp target teams |
| ;; { |
| ;; // Incompatible parallel level, called both |
| ;; // from parallel and target regions |
| ;; unknown(); |
| ;; } |
| ;; } |
| ;; |
| ;; void do_not_spmdize_task() { |
| ;; #pragma omp target |
| ;; { |
| ;; #pragma omp task |
| ;; spmd_amenable(); |
| ;; #pragma omp parallel |
| ;; unknown(); |
| ;; } |
| ;; } |
| |
| %struct.ident_t = type { i32, i32, i32, i32, ptr } |
| %struct.kmp_task_t_with_privates = type { %struct.kmp_task_t } |
| %struct.kmp_task_t = type { ptr, ptr, i32, %union.kmp_cmplrdata_t, %union.kmp_cmplrdata_t } |
| %union.kmp_cmplrdata_t = type { ptr } |
| %struct.anon = type {} |
| %struct.ConfigurationEnvironmentTy = type { i8, i8, i8 } |
| %struct.KernelEnvironmentTy = type { %struct.ConfigurationEnvironmentTy, ptr, ptr } |
| |
| @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 |
| @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } |
| @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } |
| @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } |
| @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } |
| @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } |
| @__omp_offloading_fd02_2044372e_do_not_spmdize_task_l74_kernel_environment = local_unnamed_addr constant %struct.KernelEnvironmentTy { %struct.ConfigurationEnvironmentTy { i8 1, i8 0, i8 1 }, ptr @1, ptr null } |
| |
| ; AMDGPU-DISABLED: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; AMDGPU-DISABLED: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; AMDGPU-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU-DISABLED: @[[X_SHARED_1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU-DISABLED: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED: @[[__OMP_OUTLINED__7_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; NVPTX-DISABLED: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; NVPTX-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX-DISABLED: @[[X_SHARED1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX-DISABLED: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED: @[[__OMP_OUTLINED__7_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| ; AMDGPU: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; AMDGPU: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; AMDGPU: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU: @[[GLOB2:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 22, ptr @[[GLOB0]] }, align 8 |
| ; AMDGPU: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU: @[[X_SHARED_1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| ; NVPTX: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; NVPTX: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; NVPTX: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX: @[[GLOB2:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 22, ptr @[[GLOB0]] }, align 8 |
| ; NVPTX: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX: @[[X_SHARED1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| ; AMDGPU-DISABLED1: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; AMDGPU-DISABLED1: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; AMDGPU-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED1: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU-DISABLED1: @[[X_SHARED_1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU-DISABLED1: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED1: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED1: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED1: @[[__OMP_OUTLINED__7_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED1: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| ; AMDGPU-DISABLED2: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; AMDGPU-DISABLED2: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; AMDGPU-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; AMDGPU-DISABLED2: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU-DISABLED2: @[[X_SHARED_1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; AMDGPU-DISABLED2: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED2: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED2: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED2: @[[__OMP_OUTLINED__7_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; AMDGPU-DISABLED2: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| ; NVPTX-DISABLED1: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; NVPTX-DISABLED1: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; NVPTX-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED1: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED1: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX-DISABLED1: @[[X_SHARED1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX-DISABLED1: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED1: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED1: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED1: @[[__OMP_OUTLINED__7_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED1: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| ; NVPTX-DISABLED2: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" |
| ; NVPTX-DISABLED2: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, ptr @[[GLOB0]] }, align 8 |
| ; NVPTX-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_L5_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_STACK_VAR_L20_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_L35_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_SEQUENTIAL_LOOP_TO_SHARED_VAR_GUARDED_L50_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 3 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TARGET_L65_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 0, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED2: @[[__OMP_OFFLOADING_FD02_2044372E_DO_NOT_SPMDIZE_TASK_L74_KERNEL_ENVIRONMENT:[a-zA-Z0-9_$"\\.-]+]] = local_unnamed_addr constant [[STRUCT_KERNELENVIRONMENTTY:%.*]] { [[STRUCT_CONFIGURATIONENVIRONMENTTY:%.*]] { i8 0, i8 1, i8 1 }, ptr @[[GLOB1]], ptr null } |
| ; NVPTX-DISABLED2: @[[X_SHARED:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX-DISABLED2: @[[X_SHARED1:[a-zA-Z0-9_$"\\.-]+]] = internal addrspace(3) global [4 x i8] poison, align 4 |
| ; NVPTX-DISABLED2: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED2: @[[__OMP_OUTLINED__3_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED2: @[[__OMP_OUTLINED__5_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED2: @[[__OMP_OUTLINED__7_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ; NVPTX-DISABLED2: @[[__OMP_OUTLINED__9_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef |
| ;. |
| define weak void @__omp_offloading_fd02_2044372e_sequential_loop_l5() #0 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; AMDGPU-SAME: () #[[ATTR0:[0-9]+]] { |
| ; AMDGPU-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; NVPTX-SAME: () #[[ATTR0:[0-9]+]] { |
| ; NVPTX-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; AMDGPU-DISABLED1-SAME: () #[[ATTR0:[0-9]+]] { |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; AMDGPU-DISABLED2-SAME: () #[[ATTR0:[0-9]+]] { |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; NVPTX-DISABLED1-SAME: () #[[ATTR0:[0-9]+]] { |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; NVPTX-DISABLED2-SAME: () #[[ATTR0:[0-9]+]] { |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; AMDGPU-DISABLED-SAME: () #[[ATTR0:[0-9]+]] { |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5 |
| ; NVPTX-DISABLED-SAME: () #[[ATTR0:[0-9]+]] { |
| ; NVPTX-DISABLED-NEXT: call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ; NVPTX-DISABLED-NEXT: ret void |
| call void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() |
| ret void |
| } |
| |
| define internal void @__omp_offloading_fd02_2044372e_sequential_loop_l5__debug() { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; AMDGPU-SAME: () #[[ATTR1:[0-9]+]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU: common.ret: |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: user_code.entry: |
| ; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; AMDGPU-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; NVPTX-SAME: () #[[ATTR1:[0-9]+]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX: common.ret: |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: user_code.entry: |
| ; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; NVPTX-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; NVPTX-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; AMDGPU-DISABLED1-SAME: () #[[ATTR1:[0-9]+]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: is_worker_check: |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.begin: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.finished: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED1: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED1: thread.user_code.check: |
| ; AMDGPU-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED1: common.ret: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: user_code.entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; AMDGPU-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; AMDGPU-DISABLED2-SAME: () #[[ATTR1:[0-9]+]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: is_worker_check: |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.begin: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.finished: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED2: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED2: thread.user_code.check: |
| ; AMDGPU-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED2: common.ret: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: user_code.entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; AMDGPU-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; NVPTX-DISABLED1-SAME: () #[[ATTR1:[0-9]+]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: is_worker_check: |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.begin: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.finished: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED1: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED1: thread.user_code.check: |
| ; NVPTX-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED1: common.ret: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: user_code.entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; NVPTX-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; NVPTX-DISABLED2-SAME: () #[[ATTR1:[0-9]+]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: is_worker_check: |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.begin: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.finished: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED2: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED2: thread.user_code.check: |
| ; NVPTX-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED2: common.ret: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: user_code.entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; NVPTX-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; AMDGPU-DISABLED-SAME: () #[[ATTR1:[0-9]+]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: is_worker_check: |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.begin: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.finished: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__1_wrapper.ID |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED: thread.user_code.check: |
| ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED: common.ret: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: user_code.entry: |
| ; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED-NEXT: br label [[COMMON_RET]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_l5__debug |
| ; NVPTX-DISABLED-SAME: () #[[ATTR1:[0-9]+]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED: is_worker_check: |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.begin: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.finished: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__1_wrapper.ID |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED: thread.user_code.check: |
| ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED: common.ret: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: user_code.entry: |
| ; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4:[0-9]+]] |
| ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18:![0-9]+]] |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED-NEXT: br label [[COMMON_RET]] |
| entry: |
| %.zero.addr = alloca i32, align 4 |
| %.threadid_temp. = alloca i32, align 4 |
| %0 = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_l5_kernel_environment) |
| %exec_user_code = icmp eq i32 %0, -1 |
| br i1 %exec_user_code, label %user_code.entry, label %common.ret |
| |
| common.ret: ; preds = %entry, %user_code.entry |
| ret void |
| |
| user_code.entry: ; preds = %entry |
| %1 = call i32 @__kmpc_global_thread_num(ptr @1) |
| store i32 0, ptr %.zero.addr, align 4 |
| store i32 %1, ptr %.threadid_temp., align 4, !tbaa !18 |
| call void @__omp_outlined__(ptr %.threadid_temp., ptr %.zero.addr) #6 |
| call void @__kmpc_target_deinit() |
| br label %common.ret |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__(ptr noalias %.global_tid., ptr noalias %.bound_tid.) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU: for.cond: |
| ; AMDGPU-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU: for.cond.cleanup: |
| ; AMDGPU-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: for.body: |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX: for.cond: |
| ; NVPTX-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX: for.cond.cleanup: |
| ; NVPTX-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: for.body: |
| ; NVPTX-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED1: for.cond: |
| ; AMDGPU-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED1: for.cond.cleanup: |
| ; AMDGPU-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: for.body: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED2: for.cond: |
| ; AMDGPU-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED2: for.cond.cleanup: |
| ; AMDGPU-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: for.body: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED1: for.cond: |
| ; NVPTX-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED1: for.cond.cleanup: |
| ; NVPTX-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: for.body: |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED2: for.cond: |
| ; NVPTX-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED2: for.cond.cleanup: |
| ; NVPTX-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: for.body: |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED: for.cond: |
| ; AMDGPU-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED: for.cond.cleanup: |
| ; AMDGPU-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: for.body: |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__ |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED: for.cond: |
| ; NVPTX-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED: for.cond.cleanup: |
| ; NVPTX-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7:[0-9]+]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: for.body: |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP22:![0-9]+]] |
| entry: |
| %captured_vars_addrs = alloca [0 x ptr], align 8 |
| br label %for.cond |
| |
| for.cond: ; preds = %for.body, %entry |
| %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ] |
| %cmp = icmp slt i32 %i.0, 100 |
| br i1 %cmp, label %for.body, label %for.cond.cleanup |
| |
| for.cond.cleanup: ; preds = %for.cond |
| call void @spmd_amenable() #10 |
| ret void |
| |
| for.body: ; preds = %for.cond |
| %0 = load i32, ptr %.global_tid., align 4, !tbaa !18 |
| call void @__kmpc_parallel_51(ptr @1, i32 %0, i32 1, i32 -1, i32 -1, ptr @__omp_outlined__1, ptr @__omp_outlined__1_wrapper, ptr %captured_vars_addrs, i64 0) |
| %inc = add nsw i32 %i.0, 1 |
| br label %for.cond, !llvm.loop !22 |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__1(ptr noalias %.global_tid., ptr noalias %.bound_tid.) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: call void @unknown() #[[ATTR8:[0-9]+]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| call void @unknown() #11 |
| ret void |
| } |
| |
| ; Function Attrs: convergent norecurse nounwind |
| define internal void @__omp_outlined__1_wrapper(i16 zeroext %0, i32 %1) #3 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; AMDGPU-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; AMDGPU-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; NVPTX-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; NVPTX-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper |
| ; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__1(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| %.addr1 = alloca i32, align 4 |
| %.zero.addr = alloca i32, align 4 |
| %global_args = alloca ptr, align 8 |
| store i32 %1, ptr %.addr1, align 4, !tbaa !18 |
| store i32 0, ptr %.zero.addr, align 4 |
| call void @__kmpc_get_shared_variables(ptr %global_args) |
| call void @__omp_outlined__1(ptr %.addr1, ptr %.zero.addr) #6 |
| ret void |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20() #0 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; AMDGPU-SAME: () #[[ATTR0]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU: common.ret: |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: user_code.entry: |
| ; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; NVPTX-SAME: () #[[ATTR0]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX: common.ret: |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: user_code.entry: |
| ; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; AMDGPU-DISABLED1-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: is_worker_check: |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.begin: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.finished: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED1: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED1: thread.user_code.check: |
| ; AMDGPU-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED1: common.ret: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: user_code.entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; AMDGPU-DISABLED2-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: is_worker_check: |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.begin: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.finished: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED2: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED2: thread.user_code.check: |
| ; AMDGPU-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED2: common.ret: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: user_code.entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; NVPTX-DISABLED1-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: is_worker_check: |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.begin: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.finished: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED1: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED1: thread.user_code.check: |
| ; NVPTX-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED1: common.ret: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: user_code.entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; NVPTX-DISABLED2-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: is_worker_check: |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.begin: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.finished: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED2: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED2: thread.user_code.check: |
| ; NVPTX-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED2: common.ret: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: user_code.entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: is_worker_check: |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.begin: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.finished: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__3_wrapper.ID |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED: thread.user_code.check: |
| ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED: common.ret: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: user_code.entry: |
| ; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED-NEXT: br label [[COMMON_RET]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20 |
| ; NVPTX-DISABLED-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED: is_worker_check: |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.begin: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.finished: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__3_wrapper.ID |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__3_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED: thread.user_code.check: |
| ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED: common.ret: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: user_code.entry: |
| ; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__2(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED-NEXT: br label [[COMMON_RET]] |
| entry: |
| %.zero.addr = alloca i32, align 4 |
| %.threadid_temp. = alloca i32, align 4 |
| %0 = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_stack_var_l20_kernel_environment) |
| %exec_user_code = icmp eq i32 %0, -1 |
| br i1 %exec_user_code, label %user_code.entry, label %common.ret |
| |
| common.ret: ; preds = %entry, %user_code.entry |
| ret void |
| |
| user_code.entry: ; preds = %entry |
| %1 = call i32 @__kmpc_global_thread_num(ptr @1) |
| store i32 0, ptr %.zero.addr, align 4 |
| store i32 %1, ptr %.threadid_temp., align 4, !tbaa !18 |
| call void @__omp_outlined__2(ptr %.threadid_temp., ptr %.zero.addr) #6 |
| call void @__kmpc_target_deinit() |
| br label %common.ret |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__2(ptr noalias %.global_tid., ptr noalias %.bound_tid.) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4, addrspace(5) |
| ; AMDGPU-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-NEXT: [[MALLOC_CAST:%.*]] = addrspacecast ptr addrspace(5) [[X_H2S]] to ptr |
| ; AMDGPU-NEXT: call void @use(ptr nocapture [[MALLOC_CAST]]) #[[ATTR7]] |
| ; AMDGPU-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU: for.cond: |
| ; AMDGPU-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU: for.cond.cleanup: |
| ; AMDGPU-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: for.body: |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4 |
| ; NVPTX-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-NEXT: call void @use(ptr nocapture [[X_H2S]]) #[[ATTR7]] |
| ; NVPTX-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX: for.cond: |
| ; NVPTX-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX: for.cond.cleanup: |
| ; NVPTX-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: for.body: |
| ; NVPTX-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4, addrspace(5) |
| ; AMDGPU-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[MALLOC_CAST:%.*]] = addrspacecast ptr addrspace(5) [[X_H2S]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: call void @use(ptr nocapture [[MALLOC_CAST]]) #[[ATTR7]] |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED1: for.cond: |
| ; AMDGPU-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED1: for.cond.cleanup: |
| ; AMDGPU-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: for.body: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4, addrspace(5) |
| ; AMDGPU-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[MALLOC_CAST:%.*]] = addrspacecast ptr addrspace(5) [[X_H2S]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: call void @use(ptr nocapture [[MALLOC_CAST]]) #[[ATTR7]] |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED2: for.cond: |
| ; AMDGPU-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED2: for.cond.cleanup: |
| ; AMDGPU-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: for.body: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-DISABLED1-NEXT: call void @use(ptr nocapture [[X_H2S]]) #[[ATTR7]] |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED1: for.cond: |
| ; NVPTX-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED1: for.cond.cleanup: |
| ; NVPTX-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: for.body: |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-DISABLED2-NEXT: call void @use(ptr nocapture [[X_H2S]]) #[[ATTR7]] |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED2: for.cond: |
| ; NVPTX-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED2: for.cond.cleanup: |
| ; NVPTX-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: for.body: |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4, addrspace(5) |
| ; AMDGPU-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[MALLOC_CAST:%.*]] = addrspacecast ptr addrspace(5) [[X_H2S]] to ptr |
| ; AMDGPU-DISABLED-NEXT: call void @use(ptr nocapture [[MALLOC_CAST]]) #[[ATTR7]] |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED: for.cond: |
| ; AMDGPU-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED: for.cond.cleanup: |
| ; AMDGPU-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: for.body: |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; AMDGPU-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__2 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[X_H2S:%.*]] = alloca i8, i64 4, align 4 |
| ; NVPTX-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [0 x ptr], align 8 |
| ; NVPTX-DISABLED-NEXT: call void @use(ptr nocapture [[X_H2S]]) #[[ATTR7]] |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED: for.cond: |
| ; NVPTX-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED: for.cond.cleanup: |
| ; NVPTX-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: for.body: |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 0) |
| ; NVPTX-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]] |
| entry: |
| %captured_vars_addrs = alloca [0 x ptr], align 8 |
| %x = call align 4 ptr @__kmpc_alloc_shared(i64 4) |
| call void @use(ptr nocapture %x) #10 |
| br label %for.cond |
| |
| for.cond: ; preds = %for.body, %entry |
| %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ] |
| %cmp = icmp slt i32 %i.0, 100 |
| br i1 %cmp, label %for.body, label %for.cond.cleanup |
| |
| for.cond.cleanup: ; preds = %for.cond |
| call void @spmd_amenable() #10 |
| call void @__kmpc_free_shared(ptr %x, i64 4) |
| ret void |
| |
| for.body: ; preds = %for.cond |
| %0 = load i32, ptr %.global_tid., align 4, !tbaa !18 |
| call void @__kmpc_parallel_51(ptr @1, i32 %0, i32 1, i32 -1, i32 -1, ptr @__omp_outlined__3, ptr @__omp_outlined__3_wrapper, ptr %captured_vars_addrs, i64 0) |
| %inc = add nsw i32 %i.0, 1 |
| br label %for.cond, !llvm.loop !25 |
| } |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__3(ptr noalias %.global_tid., ptr noalias %.bound_tid.) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| call void @unknown() #11 |
| ret void |
| } |
| |
| ; Function Attrs: convergent norecurse nounwind |
| define internal void @__omp_outlined__3_wrapper(i16 zeroext %0, i32 %1) #3 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; AMDGPU-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; AMDGPU-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; NVPTX-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; NVPTX-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__3_wrapper |
| ; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__3(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| %.addr1 = alloca i32, align 4 |
| %.zero.addr = alloca i32, align 4 |
| %global_args = alloca ptr, align 8 |
| store i32 %1, ptr %.addr1, align 4, !tbaa !18 |
| store i32 0, ptr %.zero.addr, align 4 |
| call void @__kmpc_get_shared_variables(ptr %global_args) |
| call void @__omp_outlined__3(ptr %.addr1, ptr %.zero.addr) #6 |
| ret void |
| } |
| |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35() #0 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; AMDGPU-SAME: () #[[ATTR0]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU: common.ret: |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: user_code.entry: |
| ; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; NVPTX-SAME: () #[[ATTR0]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX: common.ret: |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: user_code.entry: |
| ; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; AMDGPU-DISABLED1-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: is_worker_check: |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.begin: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.finished: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED1: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED1: thread.user_code.check: |
| ; AMDGPU-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED1: common.ret: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: user_code.entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; AMDGPU-DISABLED2-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: is_worker_check: |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.begin: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.finished: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED2: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED2: thread.user_code.check: |
| ; AMDGPU-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED2: common.ret: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: user_code.entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; NVPTX-DISABLED1-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: is_worker_check: |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.begin: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.finished: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED1: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED1: thread.user_code.check: |
| ; NVPTX-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED1: common.ret: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: user_code.entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; NVPTX-DISABLED2-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: is_worker_check: |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.begin: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.finished: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED2: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED2: thread.user_code.check: |
| ; NVPTX-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED2: common.ret: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: user_code.entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: is_worker_check: |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.begin: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.finished: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__5_wrapper.ID |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED: thread.user_code.check: |
| ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED: common.ret: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: user_code.entry: |
| ; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED-NEXT: br label [[COMMON_RET]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35 |
| ; NVPTX-DISABLED-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED: is_worker_check: |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.begin: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.finished: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__5_wrapper.ID |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED: thread.user_code.check: |
| ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED: common.ret: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: user_code.entry: |
| ; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__4(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED-NEXT: br label [[COMMON_RET]] |
| entry: |
| %.zero.addr = alloca i32, align 4 |
| %.threadid_temp. = alloca i32, align 4 |
| %0 = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_l35_kernel_environment) |
| %exec_user_code = icmp eq i32 %0, -1 |
| br i1 %exec_user_code, label %user_code.entry, label %common.ret |
| |
| common.ret: ; preds = %entry, %user_code.entry |
| ret void |
| |
| user_code.entry: ; preds = %entry |
| %1 = call i32 @__kmpc_global_thread_num(ptr @1) |
| store i32 0, ptr %.zero.addr, align 4 |
| store i32 %1, ptr %.threadid_temp., align 4, !tbaa !18 |
| call void @__omp_outlined__4(ptr %.threadid_temp., ptr %.zero.addr) #6 |
| call void @__kmpc_target_deinit() |
| br label %common.ret |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__4(ptr noalias %.global_tid., ptr noalias %.bound_tid.) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU: for.cond: |
| ; AMDGPU-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU: for.cond.cleanup: |
| ; AMDGPU-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: for.body: |
| ; AMDGPU-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX: for.cond: |
| ; NVPTX-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX: for.cond.cleanup: |
| ; NVPTX-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: for.body: |
| ; NVPTX-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; NVPTX-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED1: for.cond: |
| ; AMDGPU-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED1: for.cond.cleanup: |
| ; AMDGPU-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: for.body: |
| ; AMDGPU-DISABLED1-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED2: for.cond: |
| ; AMDGPU-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED2: for.cond.cleanup: |
| ; AMDGPU-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: for.body: |
| ; AMDGPU-DISABLED2-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED1: for.cond: |
| ; NVPTX-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED1: for.cond.cleanup: |
| ; NVPTX-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: for.body: |
| ; NVPTX-DISABLED1-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED2: for.cond: |
| ; NVPTX-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED2: for.cond.cleanup: |
| ; NVPTX-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: for.body: |
| ; NVPTX-DISABLED2-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED: for.cond: |
| ; AMDGPU-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED: for.cond.cleanup: |
| ; AMDGPU-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: for.body: |
| ; AMDGPU-DISABLED-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__4 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED: for.cond: |
| ; NVPTX-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED: for.cond.cleanup: |
| ; NVPTX-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: for.body: |
| ; NVPTX-DISABLED-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26:![0-9]+]] |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP28:![0-9]+]] |
| entry: |
| %captured_vars_addrs = alloca [1 x ptr], align 8 |
| %x = call align 4 ptr @__kmpc_alloc_shared(i64 4) |
| br label %for.cond |
| |
| for.cond: ; preds = %for.body, %entry |
| %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ] |
| %cmp = icmp slt i32 %i.0, 100 |
| br i1 %cmp, label %for.body, label %for.cond.cleanup |
| |
| for.cond.cleanup: ; preds = %for.cond |
| call void @spmd_amenable() #10 |
| call void @__kmpc_free_shared(ptr %x, i64 4) |
| ret void |
| |
| for.body: ; preds = %for.cond |
| store ptr %x, ptr %captured_vars_addrs, align 8, !tbaa !26 |
| %0 = load i32, ptr %.global_tid., align 4, !tbaa !18 |
| call void @__kmpc_parallel_51(ptr @1, i32 %0, i32 1, i32 -1, i32 -1, ptr @__omp_outlined__5, ptr @__omp_outlined__5_wrapper, ptr %captured_vars_addrs, i64 1) |
| %inc = add nsw i32 %i.0, 1 |
| br label %for.cond, !llvm.loop !28 |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__5(ptr noalias %.global_tid., ptr noalias %.bound_tid., ptr nonnull align 4 dereferenceable(4) %x) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-DISABLED1-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-DISABLED2-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-DISABLED1-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-DISABLED2-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-DISABLED-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @unknown() #[[ATTR8]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-DISABLED-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @unknown() #[[ATTR8]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| %0 = load i32, ptr %x, align 4, !tbaa !18 |
| %inc = add nsw i32 %0, 1 |
| store i32 %inc, ptr %x, align 4, !tbaa !18 |
| call void @unknown() #11 |
| ret void |
| } |
| |
| ; Function Attrs: convergent norecurse nounwind |
| define internal void @__omp_outlined__5_wrapper(i16 zeroext %0, i32 %1) #3 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; AMDGPU-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED1-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; AMDGPU-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED2-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; NVPTX-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED1-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; NVPTX-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED2-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__5_wrapper |
| ; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-DISABLED-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__5(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| %.addr1 = alloca i32, align 4 |
| %.zero.addr = alloca i32, align 4 |
| %global_args = alloca ptr, align 8 |
| store i32 %1, ptr %.addr1, align 4, !tbaa !18 |
| store i32 0, ptr %.zero.addr, align 4 |
| call void @__kmpc_get_shared_variables(ptr %global_args) |
| %2 = load ptr, ptr %global_args, align 8 |
| %3 = load ptr, ptr %2, align 8, !tbaa !26 |
| call void @__omp_outlined__5(ptr %.addr1, ptr %.zero.addr, ptr %3) #6 |
| ret void |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define weak void @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50() #0 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; AMDGPU-SAME: () #[[ATTR0]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU: common.ret: |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: user_code.entry: |
| ; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; NVPTX-SAME: () #[[ATTR0]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX: common.ret: |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: user_code.entry: |
| ; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; AMDGPU-DISABLED1-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: is_worker_check: |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.begin: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.finished: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED1: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED1: thread.user_code.check: |
| ; AMDGPU-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED1: common.ret: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: user_code.entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; AMDGPU-DISABLED2-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: is_worker_check: |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.begin: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.finished: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED2: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED2: thread.user_code.check: |
| ; AMDGPU-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED2: common.ret: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: user_code.entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; NVPTX-DISABLED1-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: is_worker_check: |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.begin: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.finished: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED1: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED1: thread.user_code.check: |
| ; NVPTX-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED1: common.ret: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: user_code.entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; NVPTX-DISABLED2-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: is_worker_check: |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.begin: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.finished: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.check1: |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED2: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED2: thread.user_code.check: |
| ; NVPTX-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED2: common.ret: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: user_code.entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: is_worker_check: |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.begin: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.finished: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.check: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__7_wrapper.ID |
| ; AMDGPU-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.execute: |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; AMDGPU-DISABLED: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED: thread.user_code.check: |
| ; AMDGPU-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED: common.ret: |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: user_code.entry: |
| ; AMDGPU-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED-NEXT: br label [[COMMON_RET]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50 |
| ; NVPTX-DISABLED-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED: is_worker_check: |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.begin: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.finished: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.check: |
| ; NVPTX-DISABLED-NEXT: [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__7_wrapper.ID |
| ; NVPTX-DISABLED-NEXT: br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.execute: |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-DISABLED-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] |
| ; NVPTX-DISABLED: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED: thread.user_code.check: |
| ; NVPTX-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED: common.ret: |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: user_code.entry: |
| ; NVPTX-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: store i32 [[TMP1]], ptr [[DOTTHREADID_TEMP_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__6(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR5]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED-NEXT: br label [[COMMON_RET]] |
| entry: |
| %.zero.addr = alloca i32, align 4 |
| %.threadid_temp. = alloca i32, align 4 |
| %0 = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_sequential_loop_to_shared_var_guarded_l50_kernel_environment) |
| %exec_user_code = icmp eq i32 %0, -1 |
| br i1 %exec_user_code, label %user_code.entry, label %common.ret |
| |
| common.ret: ; preds = %entry, %user_code.entry |
| ret void |
| |
| user_code.entry: ; preds = %entry |
| %1 = call i32 @__kmpc_global_thread_num(ptr @1) |
| store i32 0, ptr %.zero.addr, align 4 |
| store i32 %1, ptr %.threadid_temp., align 4, !tbaa !18 |
| call void @__omp_outlined__6(ptr %.threadid_temp., ptr %.zero.addr) #6 |
| call void @__kmpc_target_deinit() |
| br label %common.ret |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__6(ptr noalias %.global_tid., ptr noalias %.bound_tid.) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-NEXT: br label [[REGION_CHECK_TID:%.*]] |
| ; AMDGPU: region.check.tid: |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = call fastcc i32 @__kmpc_get_hardware_thread_id_in_block() |
| ; AMDGPU-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 0 |
| ; AMDGPU-NEXT: br i1 [[TMP1]], label [[REGION_GUARDED:%.*]], label [[REGION_BARRIER:%.*]] |
| ; AMDGPU: region.guarded: |
| ; AMDGPU-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: br label [[REGION_GUARDED_END:%.*]] |
| ; AMDGPU: region.guarded.end: |
| ; AMDGPU-NEXT: br label [[REGION_BARRIER]] |
| ; AMDGPU: region.barrier: |
| ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_spmd(ptr @[[GLOB2]], i32 [[TMP0]]) |
| ; AMDGPU-NEXT: br label [[REGION_EXIT:%.*]] |
| ; AMDGPU: region.exit: |
| ; AMDGPU-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU: for.cond: |
| ; AMDGPU-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[REGION_EXIT]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU: for.cond.cleanup: |
| ; AMDGPU-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: for.body: |
| ; AMDGPU-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-NEXT: br label [[REGION_CHECK_TID:%.*]] |
| ; NVPTX: region.check.tid: |
| ; NVPTX-NEXT: [[TMP0:%.*]] = call fastcc i32 @__kmpc_get_hardware_thread_id_in_block() |
| ; NVPTX-NEXT: [[TMP1:%.*]] = icmp eq i32 [[TMP0]], 0 |
| ; NVPTX-NEXT: br i1 [[TMP1]], label [[REGION_GUARDED:%.*]], label [[REGION_BARRIER:%.*]] |
| ; NVPTX: region.guarded: |
| ; NVPTX-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: br label [[REGION_GUARDED_END:%.*]] |
| ; NVPTX: region.guarded.end: |
| ; NVPTX-NEXT: br label [[REGION_BARRIER]] |
| ; NVPTX: region.barrier: |
| ; NVPTX-NEXT: call void @__kmpc_barrier_simple_spmd(ptr @[[GLOB2]], i32 [[TMP0]]) |
| ; NVPTX-NEXT: br label [[REGION_EXIT:%.*]] |
| ; NVPTX: region.exit: |
| ; NVPTX-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX: for.cond: |
| ; NVPTX-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[REGION_EXIT]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX: for.cond.cleanup: |
| ; NVPTX-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: for.body: |
| ; NVPTX-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-NEXT: [[TMP2:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP2]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-DISABLED1-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED1: for.cond: |
| ; AMDGPU-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED1: for.cond.cleanup: |
| ; AMDGPU-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: for.body: |
| ; AMDGPU-DISABLED1-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-DISABLED2-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED2: for.cond: |
| ; AMDGPU-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED2: for.cond.cleanup: |
| ; AMDGPU-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: for.body: |
| ; AMDGPU-DISABLED2-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-DISABLED1-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED1: for.cond: |
| ; NVPTX-DISABLED1-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED1-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED1: for.cond.cleanup: |
| ; NVPTX-DISABLED1-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: for.body: |
| ; NVPTX-DISABLED1-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-DISABLED1-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED1-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-DISABLED2-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED2: for.cond: |
| ; NVPTX-DISABLED2-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED2-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED2: for.cond.cleanup: |
| ; NVPTX-DISABLED2-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: for.body: |
| ; NVPTX-DISABLED2-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-DISABLED2-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED2-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; AMDGPU-DISABLED-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; AMDGPU-DISABLED: for.cond: |
| ; AMDGPU-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; AMDGPU-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; AMDGPU-DISABLED: for.cond.cleanup: |
| ; AMDGPU-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; AMDGPU-DISABLED: for.body: |
| ; AMDGPU-DISABLED-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared.1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; AMDGPU-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; AMDGPU-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__6 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]]) #[[ATTR1]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[CAPTURED_VARS_ADDRS:%.*]] = alloca [1 x ptr], align 8 |
| ; NVPTX-DISABLED-NEXT: store i32 42, ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND:%.*]] |
| ; NVPTX-DISABLED: for.cond: |
| ; NVPTX-DISABLED-NEXT: [[I_0:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INC:%.*]], [[FOR_BODY:%.*]] ] |
| ; NVPTX-DISABLED-NEXT: [[CMP:%.*]] = icmp slt i32 [[I_0]], 100 |
| ; NVPTX-DISABLED-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_COND_CLEANUP:%.*]] |
| ; NVPTX-DISABLED: for.cond.cleanup: |
| ; NVPTX-DISABLED-NEXT: call void @spmd_amenable() #[[ATTR7]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED: for.body: |
| ; NVPTX-DISABLED-NEXT: store ptr addrspacecast (ptr addrspace(3) @x_shared1 to ptr), ptr [[CAPTURED_VARS_ADDRS]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[DOTGLOBAL_TID_]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_parallel_51(ptr @[[GLOB1]], i32 [[TMP0]], i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper.ID, ptr [[CAPTURED_VARS_ADDRS]], i64 1) |
| ; NVPTX-DISABLED-NEXT: [[INC]] = add nsw i32 [[I_0]], 1 |
| ; NVPTX-DISABLED-NEXT: br label [[FOR_COND]], !llvm.loop [[LOOP29:![0-9]+]] |
| entry: |
| %captured_vars_addrs = alloca [1 x ptr], align 8 |
| %x = call align 4 ptr @__kmpc_alloc_shared(i64 4) |
| store i32 42, ptr %x, align 4, !tbaa !18 |
| br label %for.cond |
| |
| for.cond: ; preds = %for.body, %entry |
| %i.0 = phi i32 [ 0, %entry ], [ %inc, %for.body ] |
| %cmp = icmp slt i32 %i.0, 100 |
| br i1 %cmp, label %for.body, label %for.cond.cleanup |
| |
| for.cond.cleanup: ; preds = %for.cond |
| call void @spmd_amenable() #10 |
| call void @__kmpc_free_shared(ptr %x, i64 4) |
| ret void |
| |
| for.body: ; preds = %for.cond |
| store ptr %x, ptr %captured_vars_addrs, align 8, !tbaa !26 |
| %0 = load i32, ptr %.global_tid., align 4, !tbaa !18 |
| call void @__kmpc_parallel_51(ptr @1, i32 %0, i32 1, i32 -1, i32 -1, ptr @__omp_outlined__7, ptr @__omp_outlined__7_wrapper, ptr %captured_vars_addrs, i64 1) |
| %inc = add nsw i32 %i.0, 1 |
| br label %for.cond, !llvm.loop !29 |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define internal void @__omp_outlined__7(ptr noalias %.global_tid., ptr noalias %.bound_tid., ptr nonnull align 4 dereferenceable(4) %x) { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; AMDGPU-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; NVPTX-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; AMDGPU-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-DISABLED1-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED1-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; AMDGPU-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-DISABLED2-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED2-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; NVPTX-DISABLED1-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-DISABLED1-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED1-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; NVPTX-DISABLED2-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-DISABLED2-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED2-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; AMDGPU-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; AMDGPU-DISABLED-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; AMDGPU-DISABLED-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7 |
| ; NVPTX-DISABLED-SAME: (ptr noalias [[DOTGLOBAL_TID_:%.*]], ptr noalias [[DOTBOUND_TID_:%.*]], ptr nonnull align 4 dereferenceable(4) [[X:%.*]]) { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[TMP0:%.*]] = load i32, ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: [[INC:%.*]] = add nsw i32 [[TMP0]], 1 |
| ; NVPTX-DISABLED-NEXT: store i32 [[INC]], ptr [[X]], align 4, !tbaa [[TBAA18]] |
| ; NVPTX-DISABLED-NEXT: call void @unknowni32p(ptr [[X]]) #[[ATTR8]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| %0 = load i32, ptr %x, align 4, !tbaa !18 |
| %inc = add nsw i32 %0, 1 |
| store i32 %inc, ptr %x, align 4, !tbaa !18 |
| call void @unknowni32p(ptr %x) #11 |
| ret void |
| } |
| |
| ; Function Attrs: convergent norecurse nounwind |
| define internal void @__omp_outlined__7_wrapper(i16 zeroext %0, i32 %1) #3 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; AMDGPU-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: ret void |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; NVPTX-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; AMDGPU-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED1-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; AMDGPU-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED2-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; NVPTX-DISABLED1-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED1-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; NVPTX-DISABLED2-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED2-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; AMDGPU-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; AMDGPU-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; AMDGPU-DISABLED-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; AMDGPU-DISABLED-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; AMDGPU-DISABLED-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED-NEXT: ret void |
| ; NVPTX-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__7_wrapper |
| ; NVPTX-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) #[[ATTR2]] { |
| ; NVPTX-DISABLED-NEXT: entry: |
| ; NVPTX-DISABLED-NEXT: [[DOTADDR1:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED-NEXT: [[GLOBAL_ARGS:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED-NEXT: call void @__kmpc_get_shared_variables(ptr [[GLOBAL_ARGS]]) |
| ; NVPTX-DISABLED-NEXT: [[TMP2:%.*]] = load ptr, ptr [[GLOBAL_ARGS]], align 8 |
| ; NVPTX-DISABLED-NEXT: [[TMP3:%.*]] = load ptr, ptr [[TMP2]], align 8, !tbaa [[TBAA26]] |
| ; NVPTX-DISABLED-NEXT: call void @__omp_outlined__7(ptr [[DOTADDR1]], ptr [[DOTZERO_ADDR]], ptr [[TMP3]]) #[[ATTR4]] |
| ; NVPTX-DISABLED-NEXT: ret void |
| entry: |
| %.addr1 = alloca i32, align 4 |
| %.zero.addr = alloca i32, align 4 |
| %global_args = alloca ptr, align 8 |
| store i32 %1, ptr %.addr1, align 4, !tbaa !18 |
| store i32 0, ptr %.zero.addr, align 4 |
| call void @__kmpc_get_shared_variables(ptr %global_args) |
| %2 = load ptr, ptr %global_args, align 8 |
| %3 = load ptr, ptr %2, align 8, !tbaa !26 |
| call void @__omp_outlined__7(ptr %.addr1, ptr %.zero.addr, ptr %3) #6 |
| ret void |
| } |
| |
| ; Function Attrs: alwaysinline convergent norecurse nounwind |
| define weak void @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65() #0 { |
| ; AMDGPU-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; AMDGPU-SAME: () #[[ATTR0]] { |
| ; AMDGPU-NEXT: entry: |
| ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; AMDGPU-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU: is_worker_check: |
| ; AMDGPU-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU: worker_state_machine.begin: |
| ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU: worker_state_machine.finished: |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: worker_state_machine.is_active.check: |
| ; AMDGPU-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU: worker_state_machine.parallel_region.end: |
| ; AMDGPU-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU: worker_state_machine.done.barrier: |
| ; AMDGPU-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU: thread.user_code.check: |
| ; AMDGPU-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU: common.ret: |
| ; AMDGPU-NEXT: ret void |
| ; AMDGPU: user_code.entry: |
| ; AMDGPU-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: call void @__omp_outlined__8(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; NVPTX-SAME: () #[[ATTR0]] { |
| ; NVPTX-NEXT: entry: |
| ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; NVPTX-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX: is_worker_check: |
| ; NVPTX-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX: worker_state_machine.begin: |
| ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX: worker_state_machine.finished: |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: worker_state_machine.is_active.check: |
| ; NVPTX-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX: worker_state_machine.parallel_region.end: |
| ; NVPTX-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX: worker_state_machine.done.barrier: |
| ; NVPTX-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX: thread.user_code.check: |
| ; NVPTX-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX: common.ret: |
| ; NVPTX-NEXT: ret void |
| ; NVPTX: user_code.entry: |
| ; NVPTX-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-NEXT: call void @__omp_outlined__8(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; AMDGPU-DISABLED1-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED1-NEXT: entry: |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: is_worker_check: |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.begin: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.finished: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-DISABLED1-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED1: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED1: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED1: thread.user_code.check: |
| ; AMDGPU-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED1: common.ret: |
| ; AMDGPU-DISABLED1-NEXT: ret void |
| ; AMDGPU-DISABLED1: user_code.entry: |
| ; AMDGPU-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__omp_outlined__8(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; AMDGPU-DISABLED2-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED2-NEXT: entry: |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: is_worker_check: |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; AMDGPU-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; AMDGPU-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.begin: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]]) |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8 |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; AMDGPU-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.finished: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: worker_state_machine.is_active.check: |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.fallback.execute: |
| ; AMDGPU-DISABLED2-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; AMDGPU-DISABLED2: worker_state_machine.parallel_region.end: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; AMDGPU-DISABLED2: worker_state_machine.done.barrier: |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; AMDGPU-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; AMDGPU-DISABLED2: thread.user_code.check: |
| ; AMDGPU-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; AMDGPU-DISABLED2: common.ret: |
| ; AMDGPU-DISABLED2-NEXT: ret void |
| ; AMDGPU-DISABLED2: user_code.entry: |
| ; AMDGPU-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__omp_outlined__8(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; AMDGPU-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; AMDGPU-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED1-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; NVPTX-DISABLED1-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED1-NEXT: entry: |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED1-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED1-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: is_worker_check: |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED1-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED1-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED1-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.begin: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED1-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.finished: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED1-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-DISABLED1-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED1: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED1: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED1-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED1: thread.user_code.check: |
| ; NVPTX-DISABLED1-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED1-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED1: common.ret: |
| ; NVPTX-DISABLED1-NEXT: ret void |
| ; NVPTX-DISABLED1: user_code.entry: |
| ; NVPTX-DISABLED1-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: call void @__omp_outlined__8(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED1-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED1-NEXT: br label [[COMMON_RET]] |
| ; |
| ; NVPTX-DISABLED2-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; NVPTX-DISABLED2-SAME: () #[[ATTR0]] { |
| ; NVPTX-DISABLED2-NEXT: entry: |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8 |
| ; NVPTX-DISABLED2-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; NVPTX-DISABLED2-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: is_worker_check: |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() |
| ; NVPTX-DISABLED2-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() |
| ; NVPTX-DISABLED2-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] |
| ; NVPTX-DISABLED2-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.begin: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]]) |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8 |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr |
| ; NVPTX-DISABLED2-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.finished: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: worker_state_machine.is_active.check: |
| ; NVPTX-DISABLED2-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.fallback.execute: |
| ; NVPTX-DISABLED2-NEXT: call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] |
| ; NVPTX-DISABLED2: worker_state_machine.parallel_region.end: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_kernel_end_parallel() |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] |
| ; NVPTX-DISABLED2: worker_state_machine.done.barrier: |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]]) |
| ; NVPTX-DISABLED2-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] |
| ; NVPTX-DISABLED2: thread.user_code.check: |
| ; NVPTX-DISABLED2-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 |
| ; NVPTX-DISABLED2-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[COMMON_RET:%.*]] |
| ; NVPTX-DISABLED2: common.ret: |
| ; NVPTX-DISABLED2-NEXT: ret void |
| ; NVPTX-DISABLED2: user_code.entry: |
| ; NVPTX-DISABLED2-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: call void @__omp_outlined__8(ptr [[DOTTHREADID_TEMP_]], ptr [[DOTZERO_ADDR]]) #[[ATTR4]] |
| ; NVPTX-DISABLED2-NEXT: call void @__kmpc_target_deinit() |
| ; NVPTX-DISABLED2-NEXT: br label [[COMMON_RET]] |
| ; |
| ; AMDGPU-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65 |
| ; AMDGPU-DISABLED-SAME: () #[[ATTR0]] { |
| ; AMDGPU-DISABLED-NEXT: entry: |
| ; AMDGPU-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca ptr, align 8, addrspace(5) |
| ; AMDGPU-DISABLED-NEXT: [[DOTZERO_ADDR:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[DOTTHREADID_TEMP_:%.*]] = alloca i32, align 4 |
| ; AMDGPU-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(ptr @__omp_offloading_fd02_2044372e_do_not_spmdize_target_l65_kernel_environment) |
| ; AMDGPU-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 |
| ; AMDGPU-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] |
| ; AMDGPU-DISABLED: is_worker_check: |
|