1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature --check-globals 2; RUN: opt -S -passes=openmp-opt < %s | FileCheck %s 3; RUN: opt -S -passes=openmp-opt -openmp-opt-disable-spmdization < %s | FileCheck %s --check-prefix=CHECK-DISABLED 4; 5; void pure(void) __attribute__((pure, assume("ompx_spmd_amenable"))); 6; int no_openmp(int *) __attribute__((assume("omp_no_openmp","ompx_spmd_amenable"))); 7; 8; void sequential_loop(int *x, int N) { 9; #pragma omp target teams 10; { 11; x[0] = 0; 12; x[1] = 1; 13; x[N] = N; 14; for (int i = 2; i < N - 1; ++i) 15; x[i] = i - 1; 16; #pragma omp parallel 17; {} 18; int r1 = no_openmp(x); 19; x[r1] = r1; 20; int r2 = no_openmp(x); 21; x[r2] = r2; 22; int r3 = no_openmp(x); 23; x[r3] = r3; 24; 25; no_openmp(x); 26; pure(); 27; no_openmp(x); 28; pure(); 29; no_openmp(x); 30; pure(); 31; } 32; } 33; 34target datalayout = "e-i64:64-i128:128-v16:16-v32:32-n16:32:64" 35target triple = "nvptx64" 36 37%struct.ident_t = type { i32, i32, i32, i32, i8* } 38 39@0 = private unnamed_addr constant [23 x i8] c";unknown;unknown;0;0;;\00", align 1 40@1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @0, i32 0, i32 0) }, align 8 41@__omp_offloading_2a_fbfa7a_sequential_loop_l6_exec_mode = weak constant i8 1 42@llvm.compiler.used = appending global [1 x i8*] [i8* @__omp_offloading_2a_fbfa7a_sequential_loop_l6_exec_mode], section "llvm.metadata" 43 44; Function Attrs: convergent norecurse nounwind 45;. 46; CHECK: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" 47; CHECK: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @[[GLOB0]], i32 0, i32 0) }, align 8 48; CHECK: @[[__OMP_OFFLOADING_2A_FBFA7A_SEQUENTIAL_LOOP_L6_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 3 49; CHECK: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [1 x i8*] [i8* @__omp_offloading_2a_fbfa7a_sequential_loop_l6_exec_mode], section "llvm.metadata" 50; CHECK: @[[GLOB2:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 22, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @[[GLOB0]], i32 0, i32 0) }, align 8 51;. 52; CHECK-DISABLED: @[[GLOB0:[0-9]+]] = private unnamed_addr constant [23 x i8] c" 53; CHECK-DISABLED: @[[GLOB1:[0-9]+]] = private unnamed_addr constant [[STRUCT_IDENT_T:%.*]] { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([23 x i8], [23 x i8]* @[[GLOB0]], i32 0, i32 0) }, align 8 54; CHECK-DISABLED: @[[__OMP_OFFLOADING_2A_FBFA7A_SEQUENTIAL_LOOP_L6_EXEC_MODE:[a-zA-Z0-9_$"\\.-]+]] = weak constant i8 1 55; CHECK-DISABLED: @[[LLVM_COMPILER_USED:[a-zA-Z0-9_$"\\.-]+]] = appending global [1 x i8*] [i8* @__omp_offloading_2a_fbfa7a_sequential_loop_l6_exec_mode], section "llvm.metadata" 56; CHECK-DISABLED: @[[__OMP_OUTLINED__1_WRAPPER_ID:[a-zA-Z0-9_$"\\.-]+]] = private constant i8 undef 57;. 58define weak void @__omp_offloading_2a_fbfa7a_sequential_loop_l6(i32* %x, i64 %N) #0 { 59; CHECK-LABEL: define {{[^@]+}}@__omp_offloading_2a_fbfa7a_sequential_loop_l6 60; CHECK-SAME: (i32* [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { 61; CHECK-NEXT: entry: 62; CHECK-NEXT: [[N_ADDR_SROA_0_0_EXTRACT_TRUNC:%.*]] = trunc i64 [[N]] to i32 63; CHECK-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i8 2, i1 false, i1 false) #[[ATTR4:[0-9]+]] 64; CHECK-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 65; CHECK-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] 66; CHECK: user_code.entry: 67; CHECK-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @[[GLOB1]]) #[[ATTR4]] 68; CHECK-NEXT: [[ARRAYIDX1_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 1 69; CHECK-NEXT: [[SEXT:%.*]] = shl i64 [[N]], 32 70; CHECK-NEXT: [[IDXPROM_I:%.*]] = ashr exact i64 [[SEXT]], 32 71; CHECK-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM_I]] 72; CHECK-NEXT: br label [[REGION_CHECK_TID:%.*]] 73; CHECK: region.check.tid: 74; CHECK-NEXT: [[TMP2:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block() 75; CHECK-NEXT: [[TMP3:%.*]] = icmp eq i32 [[TMP2]], 0 76; CHECK-NEXT: br i1 [[TMP3]], label [[REGION_GUARDED:%.*]], label [[REGION_BARRIER:%.*]] 77; CHECK: region.guarded: 78; CHECK-NEXT: store i32 0, i32* [[X]], align 4, !noalias !8 79; CHECK-NEXT: store i32 1, i32* [[ARRAYIDX1_I]], align 4, !noalias !8 80; CHECK-NEXT: store i32 [[N_ADDR_SROA_0_0_EXTRACT_TRUNC]], i32* [[ARRAYIDX2_I]], align 4, !noalias !8 81; CHECK-NEXT: br label [[REGION_GUARDED_END:%.*]] 82; CHECK: region.guarded.end: 83; CHECK-NEXT: br label [[REGION_BARRIER]] 84; CHECK: region.barrier: 85; CHECK-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB2]], i32 [[TMP2]]) 86; CHECK-NEXT: br label [[REGION_EXIT:%.*]] 87; CHECK: region.exit: 88; CHECK-NEXT: br label [[FOR_COND_I:%.*]] 89; CHECK: for.cond.i: 90; CHECK-NEXT: [[I_0_I:%.*]] = phi i32 [ 2, [[REGION_EXIT]] ], [ [[INC_I:%.*]], [[REGION_EXIT3:%.*]] ] 91; CHECK-NEXT: [[SUB_I:%.*]] = add nsw i32 [[N_ADDR_SROA_0_0_EXTRACT_TRUNC]], -1 92; CHECK-NEXT: [[CMP_I:%.*]] = icmp slt i32 [[I_0_I]], [[SUB_I]] 93; CHECK-NEXT: br i1 [[CMP_I]], label [[FOR_BODY_I:%.*]], label [[__OMP_OUTLINED___EXIT:%.*]] 94; CHECK: for.body.i: 95; CHECK-NEXT: [[SUB3_I:%.*]] = add nsw i32 [[I_0_I]], -1 96; CHECK-NEXT: [[IDXPROM4_I:%.*]] = zext i32 [[I_0_I]] to i64 97; CHECK-NEXT: [[ARRAYIDX5_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM4_I]] 98; CHECK-NEXT: br label [[REGION_CHECK_TID5:%.*]] 99; CHECK: region.check.tid5: 100; CHECK-NEXT: [[TMP4:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block() 101; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i32 [[TMP4]], 0 102; CHECK-NEXT: br i1 [[TMP5]], label [[REGION_GUARDED4:%.*]], label [[REGION_BARRIER2:%.*]] 103; CHECK: region.guarded4: 104; CHECK-NEXT: store i32 [[SUB3_I]], i32* [[ARRAYIDX5_I]], align 4, !noalias !8 105; CHECK-NEXT: br label [[REGION_GUARDED_END1:%.*]] 106; CHECK: region.guarded.end1: 107; CHECK-NEXT: br label [[REGION_BARRIER2]] 108; CHECK: region.barrier2: 109; CHECK-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB2]], i32 [[TMP4]]) 110; CHECK-NEXT: br label [[REGION_EXIT3]] 111; CHECK: region.exit3: 112; CHECK-NEXT: [[INC_I]] = add nuw nsw i32 [[I_0_I]], 1 113; CHECK-NEXT: br label [[FOR_COND_I]], !llvm.loop [[LOOP11:![0-9]+]] 114; CHECK: __omp_outlined__.exit: 115; CHECK-NEXT: call void @__kmpc_parallel_51(%struct.ident_t* null, i32 0, i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** null, i64 0) 116; CHECK-NEXT: [[CALL_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7:[0-9]+]], !noalias !8 117; CHECK-NEXT: [[IDXPROM6_I:%.*]] = sext i32 [[CALL_I]] to i64 118; CHECK-NEXT: [[ARRAYIDX7_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM6_I]] 119; CHECK-NEXT: br label [[REGION_CHECK_TID10:%.*]] 120; CHECK: region.check.tid10: 121; CHECK-NEXT: [[TMP6:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block() 122; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i32 [[TMP6]], 0 123; CHECK-NEXT: br i1 [[TMP7]], label [[REGION_GUARDED9:%.*]], label [[REGION_BARRIER7:%.*]] 124; CHECK: region.guarded9: 125; CHECK-NEXT: store i32 [[CALL_I]], i32* [[ARRAYIDX7_I]], align 4, !noalias !8 126; CHECK-NEXT: br label [[REGION_GUARDED_END6:%.*]] 127; CHECK: region.guarded.end6: 128; CHECK-NEXT: br label [[REGION_BARRIER7]] 129; CHECK: region.barrier7: 130; CHECK-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB2]], i32 [[TMP6]]) 131; CHECK-NEXT: br label [[REGION_EXIT8:%.*]] 132; CHECK: region.exit8: 133; CHECK-NEXT: [[CALL8_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 134; CHECK-NEXT: [[IDXPROM9_I:%.*]] = sext i32 [[CALL8_I]] to i64 135; CHECK-NEXT: [[ARRAYIDX10_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM9_I]] 136; CHECK-NEXT: br label [[REGION_CHECK_TID15:%.*]] 137; CHECK: region.check.tid15: 138; CHECK-NEXT: [[TMP8:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block() 139; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i32 [[TMP8]], 0 140; CHECK-NEXT: br i1 [[TMP9]], label [[REGION_GUARDED14:%.*]], label [[REGION_BARRIER12:%.*]] 141; CHECK: region.guarded14: 142; CHECK-NEXT: store i32 [[CALL8_I]], i32* [[ARRAYIDX10_I]], align 4, !noalias !8 143; CHECK-NEXT: br label [[REGION_GUARDED_END11:%.*]] 144; CHECK: region.guarded.end11: 145; CHECK-NEXT: br label [[REGION_BARRIER12]] 146; CHECK: region.barrier12: 147; CHECK-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB2]], i32 [[TMP8]]) 148; CHECK-NEXT: br label [[REGION_EXIT13:%.*]] 149; CHECK: region.exit13: 150; CHECK-NEXT: [[CALL11_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 151; CHECK-NEXT: [[IDXPROM12_I:%.*]] = sext i32 [[CALL11_I]] to i64 152; CHECK-NEXT: [[ARRAYIDX13_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM12_I]] 153; CHECK-NEXT: br label [[REGION_CHECK_TID20:%.*]] 154; CHECK: region.check.tid20: 155; CHECK-NEXT: [[TMP10:%.*]] = call i32 @__kmpc_get_hardware_thread_id_in_block() 156; CHECK-NEXT: [[TMP11:%.*]] = icmp eq i32 [[TMP10]], 0 157; CHECK-NEXT: br i1 [[TMP11]], label [[REGION_GUARDED19:%.*]], label [[REGION_BARRIER17:%.*]] 158; CHECK: region.guarded19: 159; CHECK-NEXT: store i32 [[CALL11_I]], i32* [[ARRAYIDX13_I]], align 4, !noalias !8 160; CHECK-NEXT: br label [[REGION_GUARDED_END16:%.*]] 161; CHECK: region.guarded.end16: 162; CHECK-NEXT: br label [[REGION_BARRIER17]] 163; CHECK: region.barrier17: 164; CHECK-NEXT: call void @__kmpc_barrier_simple_spmd(%struct.ident_t* @[[GLOB2]], i32 [[TMP10]]) 165; CHECK-NEXT: br label [[REGION_EXIT18:%.*]] 166; CHECK: region.exit18: 167; CHECK-NEXT: [[CALL14_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 168; CHECK-NEXT: [[CALL15_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 169; CHECK-NEXT: [[CALL16_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 170; CHECK-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i8 2, i1 false) #[[ATTR4]] 171; CHECK-NEXT: ret void 172; CHECK: worker.exit: 173; CHECK-NEXT: ret void 174; 175; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_offloading_2a_fbfa7a_sequential_loop_l6 176; CHECK-DISABLED-SAME: (i32* [[X:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] { 177; CHECK-DISABLED-NEXT: entry: 178; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR:%.*]] = alloca i8*, align 8 179; CHECK-DISABLED-NEXT: [[N_ADDR_SROA_0_0_EXTRACT_TRUNC:%.*]] = trunc i64 [[N]] to i32 180; CHECK-DISABLED-NEXT: [[TMP0:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1]], i8 1, i1 false, i1 true) #[[ATTR4:[0-9]+]] 181; CHECK-DISABLED-NEXT: [[BLOCK_HW_SIZE:%.*]] = call i32 @__kmpc_get_hardware_num_threads_in_block() 182; CHECK-DISABLED-NEXT: [[WARP_SIZE:%.*]] = call i32 @__kmpc_get_warp_size() 183; CHECK-DISABLED-NEXT: [[BLOCK_SIZE:%.*]] = sub i32 [[BLOCK_HW_SIZE]], [[WARP_SIZE]] 184; CHECK-DISABLED-NEXT: [[THREAD_IS_MAIN_OR_WORKER:%.*]] = icmp slt i32 [[TMP0]], [[BLOCK_SIZE]] 185; CHECK-DISABLED-NEXT: br i1 [[THREAD_IS_MAIN_OR_WORKER]], label [[IS_WORKER_CHECK:%.*]], label [[WORKER_STATE_MACHINE_FINISHED:%.*]] 186; CHECK-DISABLED: is_worker_check: 187; CHECK-DISABLED-NEXT: [[THREAD_IS_WORKER:%.*]] = icmp ne i32 [[TMP0]], -1 188; CHECK-DISABLED-NEXT: br i1 [[THREAD_IS_WORKER]], label [[WORKER_STATE_MACHINE_BEGIN:%.*]], label [[THREAD_USER_CODE_CHECK:%.*]] 189; CHECK-DISABLED: worker_state_machine.begin: 190; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) 191; CHECK-DISABLED-NEXT: [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(i8** [[WORKER_WORK_FN_ADDR]]) 192; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN:%.*]] = load i8*, i8** [[WORKER_WORK_FN_ADDR]], align 8 193; CHECK-DISABLED-NEXT: [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast i8* [[WORKER_WORK_FN]] to void (i16, i32)* 194; CHECK-DISABLED-NEXT: [[WORKER_IS_DONE:%.*]] = icmp eq i8* [[WORKER_WORK_FN]], null 195; CHECK-DISABLED-NEXT: br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]] 196; CHECK-DISABLED: worker_state_machine.finished: 197; CHECK-DISABLED-NEXT: ret void 198; CHECK-DISABLED: worker_state_machine.is_active.check: 199; CHECK-DISABLED-NEXT: br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]] 200; CHECK-DISABLED: worker_state_machine.parallel_region.check: 201; CHECK-DISABLED-NEXT: br i1 true, label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]] 202; CHECK-DISABLED: worker_state_machine.parallel_region.execute: 203; CHECK-DISABLED-NEXT: call void @__omp_outlined__1_wrapper(i16 0, i32 [[TMP0]]) 204; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]] 205; CHECK-DISABLED: worker_state_machine.parallel_region.check1: 206; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]] 207; CHECK-DISABLED: worker_state_machine.parallel_region.end: 208; CHECK-DISABLED-NEXT: call void @__kmpc_kernel_end_parallel() 209; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_DONE_BARRIER]] 210; CHECK-DISABLED: worker_state_machine.done.barrier: 211; CHECK-DISABLED-NEXT: call void @__kmpc_barrier_simple_generic(%struct.ident_t* @[[GLOB1]], i32 [[TMP0]]) 212; CHECK-DISABLED-NEXT: br label [[WORKER_STATE_MACHINE_BEGIN]] 213; CHECK-DISABLED: thread.user_code.check: 214; CHECK-DISABLED-NEXT: [[EXEC_USER_CODE:%.*]] = icmp eq i32 [[TMP0]], -1 215; CHECK-DISABLED-NEXT: br i1 [[EXEC_USER_CODE]], label [[USER_CODE_ENTRY:%.*]], label [[WORKER_EXIT:%.*]] 216; CHECK-DISABLED: user_code.entry: 217; CHECK-DISABLED-NEXT: [[TMP1:%.*]] = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @[[GLOB1]]) #[[ATTR4]] 218; CHECK-DISABLED-NEXT: store i32 0, i32* [[X]], align 4, !noalias !8 219; CHECK-DISABLED-NEXT: [[ARRAYIDX1_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 1 220; CHECK-DISABLED-NEXT: store i32 1, i32* [[ARRAYIDX1_I]], align 4, !noalias !8 221; CHECK-DISABLED-NEXT: [[SEXT:%.*]] = shl i64 [[N]], 32 222; CHECK-DISABLED-NEXT: [[IDXPROM_I:%.*]] = ashr exact i64 [[SEXT]], 32 223; CHECK-DISABLED-NEXT: [[ARRAYIDX2_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM_I]] 224; CHECK-DISABLED-NEXT: store i32 [[N_ADDR_SROA_0_0_EXTRACT_TRUNC]], i32* [[ARRAYIDX2_I]], align 4, !noalias !8 225; CHECK-DISABLED-NEXT: br label [[FOR_COND_I:%.*]] 226; CHECK-DISABLED: for.cond.i: 227; CHECK-DISABLED-NEXT: [[I_0_I:%.*]] = phi i32 [ 2, [[USER_CODE_ENTRY]] ], [ [[INC_I:%.*]], [[FOR_BODY_I:%.*]] ] 228; CHECK-DISABLED-NEXT: [[SUB_I:%.*]] = add nsw i32 [[N_ADDR_SROA_0_0_EXTRACT_TRUNC]], -1 229; CHECK-DISABLED-NEXT: [[CMP_I:%.*]] = icmp slt i32 [[I_0_I]], [[SUB_I]] 230; CHECK-DISABLED-NEXT: br i1 [[CMP_I]], label [[FOR_BODY_I]], label [[__OMP_OUTLINED___EXIT:%.*]] 231; CHECK-DISABLED: for.body.i: 232; CHECK-DISABLED-NEXT: [[SUB3_I:%.*]] = add nsw i32 [[I_0_I]], -1 233; CHECK-DISABLED-NEXT: [[IDXPROM4_I:%.*]] = zext i32 [[I_0_I]] to i64 234; CHECK-DISABLED-NEXT: [[ARRAYIDX5_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM4_I]] 235; CHECK-DISABLED-NEXT: store i32 [[SUB3_I]], i32* [[ARRAYIDX5_I]], align 4, !noalias !8 236; CHECK-DISABLED-NEXT: [[INC_I]] = add nuw nsw i32 [[I_0_I]], 1 237; CHECK-DISABLED-NEXT: br label [[FOR_COND_I]], !llvm.loop [[LOOP11:![0-9]+]] 238; CHECK-DISABLED: __omp_outlined__.exit: 239; CHECK-DISABLED-NEXT: call void @__kmpc_parallel_51(%struct.ident_t* null, i32 0, i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* @__omp_outlined__1_wrapper.ID, i8** null, i64 0) 240; CHECK-DISABLED-NEXT: [[CALL_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7:[0-9]+]], !noalias !8 241; CHECK-DISABLED-NEXT: [[IDXPROM6_I:%.*]] = sext i32 [[CALL_I]] to i64 242; CHECK-DISABLED-NEXT: [[ARRAYIDX7_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM6_I]] 243; CHECK-DISABLED-NEXT: store i32 [[CALL_I]], i32* [[ARRAYIDX7_I]], align 4, !noalias !8 244; CHECK-DISABLED-NEXT: [[CALL8_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 245; CHECK-DISABLED-NEXT: [[IDXPROM9_I:%.*]] = sext i32 [[CALL8_I]] to i64 246; CHECK-DISABLED-NEXT: [[ARRAYIDX10_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM9_I]] 247; CHECK-DISABLED-NEXT: store i32 [[CALL8_I]], i32* [[ARRAYIDX10_I]], align 4, !noalias !8 248; CHECK-DISABLED-NEXT: [[CALL11_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 249; CHECK-DISABLED-NEXT: [[IDXPROM12_I:%.*]] = sext i32 [[CALL11_I]] to i64 250; CHECK-DISABLED-NEXT: [[ARRAYIDX13_I:%.*]] = getelementptr inbounds i32, i32* [[X]], i64 [[IDXPROM12_I]] 251; CHECK-DISABLED-NEXT: store i32 [[CALL11_I]], i32* [[ARRAYIDX13_I]], align 4, !noalias !8 252; CHECK-DISABLED-NEXT: [[CALL14_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 253; CHECK-DISABLED-NEXT: [[CALL15_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 254; CHECK-DISABLED-NEXT: [[CALL16_I:%.*]] = call i32 @no_openmp(i32* nonnull [[X]]) #[[ATTR7]], !noalias !8 255; CHECK-DISABLED-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* nonnull @[[GLOB1]], i8 1, i1 true) #[[ATTR4]] 256; CHECK-DISABLED-NEXT: ret void 257; CHECK-DISABLED: worker.exit: 258; CHECK-DISABLED-NEXT: ret void 259; 260entry: 261 %N.addr.sroa.0.0.extract.trunc = trunc i64 %N to i32 262 %0 = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i8 1, i1 true, i1 true) #3 263 %exec_user_code = icmp eq i32 %0, -1 264 br i1 %exec_user_code, label %user_code.entry, label %worker.exit 265 266user_code.entry: ; preds = %entry 267 %1 = call i32 @__kmpc_global_thread_num(%struct.ident_t* nonnull @1) 268 store i32 0, i32* %x, align 4, !noalias !8 269 %arrayidx1.i = getelementptr inbounds i32, i32* %x, i64 1 270 store i32 1, i32* %arrayidx1.i, align 4, !noalias !8 271 %sext = shl i64 %N, 32 272 %idxprom.i = ashr exact i64 %sext, 32 273 %arrayidx2.i = getelementptr inbounds i32, i32* %x, i64 %idxprom.i 274 store i32 %N.addr.sroa.0.0.extract.trunc, i32* %arrayidx2.i, align 4, !noalias !8 275 br label %for.cond.i 276 277for.cond.i: ; preds = %for.body.i, %user_code.entry 278 %i.0.i = phi i32 [ 2, %user_code.entry ], [ %inc.i, %for.body.i ] 279 %sub.i = add nsw i32 %N.addr.sroa.0.0.extract.trunc, -1 280 %cmp.i = icmp slt i32 %i.0.i, %sub.i 281 br i1 %cmp.i, label %for.body.i, label %__omp_outlined__.exit 282 283for.body.i: ; preds = %for.cond.i 284 %sub3.i = add nsw i32 %i.0.i, -1 285 %idxprom4.i = zext i32 %i.0.i to i64 286 %arrayidx5.i = getelementptr inbounds i32, i32* %x, i64 %idxprom4.i 287 store i32 %sub3.i, i32* %arrayidx5.i, align 4, !noalias !8 288 %inc.i = add nuw nsw i32 %i.0.i, 1 289 br label %for.cond.i, !llvm.loop !11 290 291__omp_outlined__.exit: ; preds = %for.cond.i 292 call void @__kmpc_parallel_51(%struct.ident_t* null, i32 0, i32 1, i32 -1, i32 -1, i8* bitcast (void (i32*, i32*)* @__omp_outlined__1 to i8*), i8* bitcast (void (i16, i32)* @__omp_outlined__1_wrapper to i8*), i8** null, i64 0) 293 %call.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8 294 %idxprom6.i = sext i32 %call.i to i64 295 %arrayidx7.i = getelementptr inbounds i32, i32* %x, i64 %idxprom6.i 296 store i32 %call.i, i32* %arrayidx7.i, align 4, !noalias !8 297 %call8.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8 298 %idxprom9.i = sext i32 %call8.i to i64 299 %arrayidx10.i = getelementptr inbounds i32, i32* %x, i64 %idxprom9.i 300 store i32 %call8.i, i32* %arrayidx10.i, align 4, !noalias !8 301 %call11.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8 302 %idxprom12.i = sext i32 %call11.i to i64 303 %arrayidx13.i = getelementptr inbounds i32, i32* %x, i64 %idxprom12.i 304 store i32 %call11.i, i32* %arrayidx13.i, align 4, !noalias !8 305 %call14.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8 306 %call15.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8 307 %call16.i = call i32 @no_openmp(i32* nonnull %x) #5, !noalias !8 308 call void @__kmpc_target_deinit(%struct.ident_t* nonnull @1, i8 1, i1 true) #3 309 ret void 310 311worker.exit: ; preds = %entry 312 ret void 313} 314 315define internal void @__omp_outlined__1(i32* noalias %.global_tid., i32* noalias %.bound_tid.) { 316; CHECK-LABEL: define {{[^@]+}}@__omp_outlined__1 317; CHECK-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) { 318; CHECK-NEXT: ret void 319; 320; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1 321; CHECK-DISABLED-SAME: (i32* noalias [[DOTGLOBAL_TID_:%.*]], i32* noalias [[DOTBOUND_TID_:%.*]]) { 322; CHECK-DISABLED-NEXT: ret void 323; 324 ret void 325} 326define internal void @__omp_outlined__1_wrapper(i16 zeroext %0, i32 %1) { 327; CHECK-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper 328; CHECK-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) { 329; CHECK-NEXT: ret void 330; 331; CHECK-DISABLED-LABEL: define {{[^@]+}}@__omp_outlined__1_wrapper 332; CHECK-DISABLED-SAME: (i16 zeroext [[TMP0:%.*]], i32 [[TMP1:%.*]]) { 333; CHECK-DISABLED-NEXT: ret void 334; 335 ret void 336} 337 338declare void @__kmpc_parallel_51(%struct.ident_t*, i32, i32, i32, i32, i8*, i8*, i8**, i64) 339 340declare i32 @__kmpc_target_init(%struct.ident_t*, i8, i1, i1) 341 342; Function Attrs: convergent 343declare i32 @no_openmp(i32*) #1 344 345; Function Attrs: convergent nounwind readonly willreturn 346declare void @pure() #2 347 348; Function Attrs: nounwind 349declare i32 @__kmpc_global_thread_num(%struct.ident_t*) #3 350 351declare void @__kmpc_target_deinit(%struct.ident_t*, i8, i1) 352 353; Function Attrs: inaccessiblememonly nofree nosync nounwind willreturn 354declare void @llvm.experimental.noalias.scope.decl(metadata) #4 355 356attributes #0 = { convergent norecurse nounwind "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 357attributes #1 = { convergent "frame-pointer"="all" "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 358attributes #2 = { convergent nounwind readonly willreturn "frame-pointer"="all" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 359attributes #3 = { nounwind } 360attributes #4 = { inaccessiblememonly nofree nosync nounwind willreturn } 361attributes #5 = { convergent nounwind "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" } 362 363!omp_offload.info = !{!0} 364!nvvm.annotations = !{!1} 365!llvm.module.flags = !{!2, !3, !4, !5, !6} 366!llvm.ident = !{!7} 367 368!0 = !{i32 0, i32 42, i32 16513658, !"sequential_loop", i32 6, i32 0} 369!1 = !{void (i32*, i64)* @__omp_offloading_2a_fbfa7a_sequential_loop_l6, !"kernel", i32 1} 370!2 = !{i32 1, !"wchar_size", i32 4} 371!3 = !{i32 7, !"openmp", i32 50} 372!4 = !{i32 7, !"openmp-device", i32 50} 373!5 = !{i32 7, !"PIC Level", i32 2} 374!6 = !{i32 7, !"frame-pointer", i32 2} 375!7 = !{!"clang version 14.0.0"} 376!8 = !{!9} 377!9 = distinct !{!9, !10, !"__omp_outlined__: %__context"} 378!10 = distinct !{!10, !"__omp_outlined__"} 379!11 = distinct !{!11, !12} 380!12 = !{!"llvm.loop.mustprogress"} 381;. 382; CHECK: attributes #[[ATTR0]] = { convergent norecurse nounwind "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 383; CHECK: attributes #[[ATTR1:[0-9]+]] = { alwaysinline } 384; CHECK: attributes #[[ATTR2:[0-9]+]] = { convergent "frame-pointer"="all" "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 385; CHECK: attributes #[[ATTR3:[0-9]+]] = { convergent nounwind readonly willreturn "frame-pointer"="all" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 386; CHECK: attributes #[[ATTR4]] = { nounwind } 387; CHECK: attributes #[[ATTR5:[0-9]+]] = { inaccessiblememonly nofree nosync nounwind willreturn } 388; CHECK: attributes #[[ATTR6:[0-9]+]] = { convergent nounwind } 389; CHECK: attributes #[[ATTR7]] = { convergent nounwind "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" } 390;. 391; CHECK-DISABLED: attributes #[[ATTR0]] = { convergent norecurse nounwind "frame-pointer"="all" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 392; CHECK-DISABLED: attributes #[[ATTR1:[0-9]+]] = { alwaysinline } 393; CHECK-DISABLED: attributes #[[ATTR2:[0-9]+]] = { convergent "frame-pointer"="all" "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 394; CHECK-DISABLED: attributes #[[ATTR3:[0-9]+]] = { convergent nounwind readonly willreturn "frame-pointer"="all" "llvm.assume"="ompx_spmd_amenable" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="sm_53" "target-features"="+ptx32,+sm_53" } 395; CHECK-DISABLED: attributes #[[ATTR4]] = { nounwind } 396; CHECK-DISABLED: attributes #[[ATTR5:[0-9]+]] = { inaccessiblememonly nofree nosync nounwind willreturn } 397; CHECK-DISABLED: attributes #[[ATTR6:[0-9]+]] = { convergent nounwind } 398; CHECK-DISABLED: attributes #[[ATTR7]] = { convergent nounwind "llvm.assume"="omp_no_openmp,ompx_spmd_amenable" } 399;. 400; CHECK: [[META0:![0-9]+]] = !{i32 0, i32 42, i32 16513658, !"sequential_loop", i32 6, i32 0} 401; CHECK: [[META1:![0-9]+]] = !{void (i32*, i64)* @__omp_offloading_2a_fbfa7a_sequential_loop_l6, !"kernel", i32 1} 402; CHECK: [[META2:![0-9]+]] = !{i32 1, !"wchar_size", i32 4} 403; CHECK: [[META3:![0-9]+]] = !{i32 7, !"openmp", i32 50} 404; CHECK: [[META4:![0-9]+]] = !{i32 7, !"openmp-device", i32 50} 405; CHECK: [[META5:![0-9]+]] = !{i32 7, !"PIC Level", i32 2} 406; CHECK: [[META6:![0-9]+]] = !{i32 7, !"frame-pointer", i32 2} 407; CHECK: [[META7:![0-9]+]] = !{!"clang version 14.0.0"} 408; CHECK: [[META8:![0-9]+]] = !{!9} 409; CHECK: [[META9:![0-9]+]] = distinct !{!9, !10, !"__omp_outlined__: %__context"} 410; CHECK: [[META10:![0-9]+]] = distinct !{!10, !"__omp_outlined__"} 411; CHECK: [[LOOP11]] = distinct !{!11, !12} 412; CHECK: [[META12:![0-9]+]] = !{!"llvm.loop.mustprogress"} 413;. 414; CHECK-DISABLED: [[META0:![0-9]+]] = !{i32 0, i32 42, i32 16513658, !"sequential_loop", i32 6, i32 0} 415; CHECK-DISABLED: [[META1:![0-9]+]] = !{void (i32*, i64)* @__omp_offloading_2a_fbfa7a_sequential_loop_l6, !"kernel", i32 1} 416; CHECK-DISABLED: [[META2:![0-9]+]] = !{i32 1, !"wchar_size", i32 4} 417; CHECK-DISABLED: [[META3:![0-9]+]] = !{i32 7, !"openmp", i32 50} 418; CHECK-DISABLED: [[META4:![0-9]+]] = !{i32 7, !"openmp-device", i32 50} 419; CHECK-DISABLED: [[META5:![0-9]+]] = !{i32 7, !"PIC Level", i32 2} 420; CHECK-DISABLED: [[META6:![0-9]+]] = !{i32 7, !"frame-pointer", i32 2} 421; CHECK-DISABLED: [[META7:![0-9]+]] = !{!"clang version 14.0.0"} 422; CHECK-DISABLED: [[META8:![0-9]+]] = !{!9} 423; CHECK-DISABLED: [[META9:![0-9]+]] = distinct !{!9, !10, !"__omp_outlined__: %__context"} 424; CHECK-DISABLED: [[META10:![0-9]+]] = distinct !{!10, !"__omp_outlined__"} 425; CHECK-DISABLED: [[LOOP11]] = distinct !{!11, !12} 426; CHECK-DISABLED: [[META12:![0-9]+]] = !{!"llvm.loop.mustprogress"} 427;. 428