1 // NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py UTC_ARGS: --function-signature --include-generated-funcs --replace-value-regex "__omp_offloading_[0-9a-z]+_[0-9a-z]+" "reduction_ size[.].+[.]" "pl_cond[.].+[.|,]" --prefix-filecheck-ir-name _ 2 // RUN: %clang_cc1 -fopenmp-enable-irbuilder -verify -fopenmp -fopenmp-version=45 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s 3 // expected-no-diagnostics 4 5 #ifndef HEADER 6 #define HEADER 7 8 9 extern "C" void workshareloop_unsigned_down(float *a) { 10 #pragma omp for 11 for (unsigned i = 32000000; i > 33; i -= 7) { 12 a[i] = i; 13 } 14 } 15 16 #endif // HEADER 17 18 19 20 21 22 // CHECK-LABEL: define {{[^@]+}}@workshareloop_unsigned_down 23 // CHECK-SAME: (ptr noundef [[A:%.*]]) #[[ATTR0:[0-9]+]] { 24 // CHECK-NEXT: entry: 25 // CHECK-NEXT: [[A_ADDR:%.*]] = alloca ptr, align 8 26 // CHECK-NEXT: [[I:%.*]] = alloca i32, align 4 27 // CHECK-NEXT: [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]], align 8 28 // CHECK-NEXT: [[AGG_CAPTURED1:%.*]] = alloca [[STRUCT_ANON_0:%.*]], align 4 29 // CHECK-NEXT: [[DOTCOUNT_ADDR:%.*]] = alloca i32, align 4 30 // CHECK-NEXT: [[P_LASTITER:%.*]] = alloca i32, align 4 31 // CHECK-NEXT: [[P_LOWERBOUND:%.*]] = alloca i32, align 4 32 // CHECK-NEXT: [[P_UPPERBOUND:%.*]] = alloca i32, align 4 33 // CHECK-NEXT: [[P_STRIDE:%.*]] = alloca i32, align 4 34 // CHECK-NEXT: store ptr [[A]], ptr [[A_ADDR]], align 8 35 // CHECK-NEXT: store i32 32000000, ptr [[I]], align 4 36 // CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON]], ptr [[AGG_CAPTURED]], i32 0, i32 0 37 // CHECK-NEXT: store ptr [[I]], ptr [[TMP0]], align 8 38 // CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON_0]], ptr [[AGG_CAPTURED1]], i32 0, i32 0 39 // CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[I]], align 4 40 // CHECK-NEXT: store i32 [[TMP2]], ptr [[TMP1]], align 4 41 // CHECK-NEXT: call void @__captured_stmt(ptr [[DOTCOUNT_ADDR]], ptr [[AGG_CAPTURED]]) 42 // CHECK-NEXT: [[DOTCOUNT:%.*]] = load i32, ptr [[DOTCOUNT_ADDR]], align 4 43 // CHECK-NEXT: br label [[OMP_LOOP_PREHEADER:%.*]] 44 // CHECK: omp_loop.preheader: 45 // CHECK-NEXT: store i32 0, ptr [[P_LOWERBOUND]], align 4 46 // CHECK-NEXT: [[TMP3:%.*]] = sub i32 [[DOTCOUNT]], 1 47 // CHECK-NEXT: store i32 [[TMP3]], ptr [[P_UPPERBOUND]], align 4 48 // CHECK-NEXT: store i32 1, ptr [[P_STRIDE]], align 4 49 // CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1:[0-9]+]]) 50 // CHECK-NEXT: call void @__kmpc_for_static_init_4u(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]], i32 34, ptr [[P_LASTITER]], ptr [[P_LOWERBOUND]], ptr [[P_UPPERBOUND]], ptr [[P_STRIDE]], i32 1, i32 0) 51 // CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[P_LOWERBOUND]], align 4 52 // CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[P_UPPERBOUND]], align 4 53 // CHECK-NEXT: [[TMP6:%.*]] = sub i32 [[TMP5]], [[TMP4]] 54 // CHECK-NEXT: [[TMP7:%.*]] = add i32 [[TMP6]], 1 55 // CHECK-NEXT: br label [[OMP_LOOP_HEADER:%.*]] 56 // CHECK: omp_loop.header: 57 // CHECK-NEXT: [[OMP_LOOP_IV:%.*]] = phi i32 [ 0, [[OMP_LOOP_PREHEADER]] ], [ [[OMP_LOOP_NEXT:%.*]], [[OMP_LOOP_INC:%.*]] ] 58 // CHECK-NEXT: br label [[OMP_LOOP_COND:%.*]] 59 // CHECK: omp_loop.cond: 60 // CHECK-NEXT: [[OMP_LOOP_CMP:%.*]] = icmp ult i32 [[OMP_LOOP_IV]], [[TMP7]] 61 // CHECK-NEXT: br i1 [[OMP_LOOP_CMP]], label [[OMP_LOOP_BODY:%.*]], label [[OMP_LOOP_EXIT:%.*]] 62 // CHECK: omp_loop.body: 63 // CHECK-NEXT: [[TMP8:%.*]] = add i32 [[OMP_LOOP_IV]], [[TMP4]] 64 // CHECK-NEXT: call void @__captured_stmt.1(ptr [[I]], i32 [[TMP8]], ptr [[AGG_CAPTURED1]]) 65 // CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[I]], align 4 66 // CHECK-NEXT: [[CONV:%.*]] = uitofp i32 [[TMP9]] to float 67 // CHECK-NEXT: [[TMP10:%.*]] = load ptr, ptr [[A_ADDR]], align 8 68 // CHECK-NEXT: [[TMP11:%.*]] = load i32, ptr [[I]], align 4 69 // CHECK-NEXT: [[IDXPROM:%.*]] = zext i32 [[TMP11]] to i64 70 // CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i64 [[IDXPROM]] 71 // CHECK-NEXT: store float [[CONV]], ptr [[ARRAYIDX]], align 4 72 // CHECK-NEXT: br label [[OMP_LOOP_INC]] 73 // CHECK: omp_loop.inc: 74 // CHECK-NEXT: [[OMP_LOOP_NEXT]] = add nuw i32 [[OMP_LOOP_IV]], 1 75 // CHECK-NEXT: br label [[OMP_LOOP_HEADER]] 76 // CHECK: omp_loop.exit: 77 // CHECK-NEXT: call void @__kmpc_for_static_fini(ptr @[[GLOB1]], i32 [[OMP_GLOBAL_THREAD_NUM]]) 78 // CHECK-NEXT: [[OMP_GLOBAL_THREAD_NUM2:%.*]] = call i32 @__kmpc_global_thread_num(ptr @[[GLOB1]]) 79 // CHECK-NEXT: call void @__kmpc_barrier(ptr @[[GLOB2:[0-9]+]], i32 [[OMP_GLOBAL_THREAD_NUM2]]) 80 // CHECK-NEXT: br label [[OMP_LOOP_AFTER:%.*]] 81 // CHECK: omp_loop.after: 82 // CHECK-NEXT: ret void 83 // 84 // 85 // CHECK-LABEL: define {{[^@]+}}@__captured_stmt 86 // CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[DISTANCE:%.*]], ptr noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1:[0-9]+]] { 87 // CHECK-NEXT: entry: 88 // CHECK-NEXT: [[DISTANCE_ADDR:%.*]] = alloca ptr, align 8 89 // CHECK-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca ptr, align 8 90 // CHECK-NEXT: [[DOTSTART:%.*]] = alloca i32, align 4 91 // CHECK-NEXT: [[DOTSTOP:%.*]] = alloca i32, align 4 92 // CHECK-NEXT: [[DOTSTEP:%.*]] = alloca i32, align 4 93 // CHECK-NEXT: store ptr [[DISTANCE]], ptr [[DISTANCE_ADDR]], align 8 94 // CHECK-NEXT: store ptr [[__CONTEXT]], ptr [[__CONTEXT_ADDR]], align 8 95 // CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[__CONTEXT_ADDR]], align 8 96 // CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON:%.*]], ptr [[TMP0]], i32 0, i32 0 97 // CHECK-NEXT: [[TMP2:%.*]] = load ptr, ptr [[TMP1]], align 8 98 // CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[TMP2]], align 4 99 // CHECK-NEXT: store i32 [[TMP3]], ptr [[DOTSTART]], align 4 100 // CHECK-NEXT: store i32 33, ptr [[DOTSTOP]], align 4 101 // CHECK-NEXT: store i32 -7, ptr [[DOTSTEP]], align 4 102 // CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr [[DOTSTART]], align 4 103 // CHECK-NEXT: [[TMP5:%.*]] = load i32, ptr [[DOTSTOP]], align 4 104 // CHECK-NEXT: [[CMP:%.*]] = icmp ugt i32 [[TMP4]], [[TMP5]] 105 // CHECK-NEXT: br i1 [[CMP]], label [[COND_TRUE:%.*]], label [[COND_FALSE:%.*]] 106 // CHECK: cond.true: 107 // CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr [[DOTSTART]], align 4 108 // CHECK-NEXT: [[TMP7:%.*]] = load i32, ptr [[DOTSTOP]], align 4 109 // CHECK-NEXT: [[SUB:%.*]] = sub i32 [[TMP6]], [[TMP7]] 110 // CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr [[DOTSTEP]], align 4 111 // CHECK-NEXT: [[SUB1:%.*]] = sub nsw i32 0, [[TMP8]] 112 // CHECK-NEXT: [[SUB2:%.*]] = sub i32 [[SUB1]], 1 113 // CHECK-NEXT: [[ADD:%.*]] = add i32 [[SUB]], [[SUB2]] 114 // CHECK-NEXT: [[TMP9:%.*]] = load i32, ptr [[DOTSTEP]], align 4 115 // CHECK-NEXT: [[SUB3:%.*]] = sub nsw i32 0, [[TMP9]] 116 // CHECK-NEXT: [[DIV:%.*]] = udiv i32 [[ADD]], [[SUB3]] 117 // CHECK-NEXT: br label [[COND_END:%.*]] 118 // CHECK: cond.false: 119 // CHECK-NEXT: br label [[COND_END]] 120 // CHECK: cond.end: 121 // CHECK-NEXT: [[COND:%.*]] = phi i32 [ [[DIV]], [[COND_TRUE]] ], [ 0, [[COND_FALSE]] ] 122 // CHECK-NEXT: [[TMP10:%.*]] = load ptr, ptr [[DISTANCE_ADDR]], align 8 123 // CHECK-NEXT: store i32 [[COND]], ptr [[TMP10]], align 4 124 // CHECK-NEXT: ret void 125 // 126 // 127 // CHECK-LABEL: define {{[^@]+}}@__captured_stmt.1 128 // CHECK-SAME: (ptr noundef nonnull align 4 dereferenceable(4) [[LOOPVAR:%.*]], i32 noundef [[LOGICAL:%.*]], ptr noalias noundef [[__CONTEXT:%.*]]) #[[ATTR1]] { 129 // CHECK-NEXT: entry: 130 // CHECK-NEXT: [[LOOPVAR_ADDR:%.*]] = alloca ptr, align 8 131 // CHECK-NEXT: [[LOGICAL_ADDR:%.*]] = alloca i32, align 4 132 // CHECK-NEXT: [[__CONTEXT_ADDR:%.*]] = alloca ptr, align 8 133 // CHECK-NEXT: store ptr [[LOOPVAR]], ptr [[LOOPVAR_ADDR]], align 8 134 // CHECK-NEXT: store i32 [[LOGICAL]], ptr [[LOGICAL_ADDR]], align 4 135 // CHECK-NEXT: store ptr [[__CONTEXT]], ptr [[__CONTEXT_ADDR]], align 8 136 // CHECK-NEXT: [[TMP0:%.*]] = load ptr, ptr [[__CONTEXT_ADDR]], align 8 137 // CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds nuw [[STRUCT_ANON_0:%.*]], ptr [[TMP0]], i32 0, i32 0 138 // CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr [[TMP1]], align 4 139 // CHECK-NEXT: [[TMP3:%.*]] = load i32, ptr [[LOGICAL_ADDR]], align 4 140 // CHECK-NEXT: [[MUL:%.*]] = mul i32 -7, [[TMP3]] 141 // CHECK-NEXT: [[ADD:%.*]] = add i32 [[TMP2]], [[MUL]] 142 // CHECK-NEXT: [[TMP4:%.*]] = load ptr, ptr [[LOOPVAR_ADDR]], align 8 143 // CHECK-NEXT: store i32 [[ADD]], ptr [[TMP4]], align 4 144 // CHECK-NEXT: ret void 145 // 146