1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt %s -loop-vectorize -force-vector-interleave=1 -force-vector-width=4 -S | FileCheck %s 3 4define void @signed(ptr %x, ptr %y, i32 %n) { 5; CHECK-LABEL: @signed( 6; CHECK-NEXT: entry: 7; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[N:%.*]], 0 8; CHECK-NEXT: br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]] 9; CHECK: for.body.preheader: 10; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 11; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 12; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] 13; CHECK: vector.memcheck: 14; CHECK-NEXT: [[TMP0:%.*]] = shl nuw nsw i64 [[WIDE_TRIP_COUNT]], 2 15; CHECK-NEXT: [[UGLYGEP:%.*]] = getelementptr i8, ptr [[Y:%.*]], i64 [[TMP0]] 16; CHECK-NEXT: [[UGLYGEP1:%.*]] = getelementptr i8, ptr [[X:%.*]], i64 [[TMP0]] 17; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[Y]], [[UGLYGEP1]] 18; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[X]], [[UGLYGEP]] 19; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] 20; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] 21; CHECK: vector.ph: 22; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 23; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 24; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 25; CHECK: vector.body: 26; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 27; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 0 28; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[TMP1]] 29; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0 30; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4, !alias.scope !0 31; CHECK-NEXT: [[TMP4:%.*]] = call <4 x i32> @llvm.fptosi.sat.v4i32.v4f32(<4 x float> [[WIDE_LOAD]]) 32; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[Y]], i64 [[TMP1]] 33; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0 34; CHECK-NEXT: store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4, !alias.scope !3, !noalias !0 35; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 36; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 37; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] 38; CHECK: middle.block: 39; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 40; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]] 41; CHECK: scalar.ph: 42; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ] 43; CHECK-NEXT: br label [[FOR_BODY:%.*]] 44; CHECK: for.cond.cleanup.loopexit: 45; CHECK-NEXT: br label [[FOR_COND_CLEANUP]] 46; CHECK: for.cond.cleanup: 47; CHECK-NEXT: ret void 48; CHECK: for.body: 49; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 50; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDVARS_IV]] 51; CHECK-NEXT: [[TMP8:%.*]] = load float, ptr [[ARRAYIDX]], align 4 52; CHECK-NEXT: [[TMP9:%.*]] = tail call i32 @llvm.fptosi.sat.i32.f32(float [[TMP8]]) 53; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[Y]], i64 [[INDVARS_IV]] 54; CHECK-NEXT: store i32 [[TMP9]], ptr [[ARRAYIDX2]], align 4 55; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 56; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 57; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]] 58; 59entry: 60 %cmp6 = icmp sgt i32 %n, 0 61 br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup 62 63for.body.preheader: ; preds = %entry 64 %wide.trip.count = zext i32 %n to i64 65 br label %for.body 66 67for.cond.cleanup: ; preds = %for.body, %entry 68 ret void 69 70for.body: ; preds = %for.body.preheader, %for.body 71 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 72 %arrayidx = getelementptr inbounds float, ptr %x, i64 %indvars.iv 73 %0 = load float, ptr %arrayidx, align 4 74 %1 = tail call i32 @llvm.fptosi.sat.i32.f32(float %0) 75 %arrayidx2 = getelementptr inbounds i32, ptr %y, i64 %indvars.iv 76 store i32 %1, ptr %arrayidx2, align 4 77 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 78 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count 79 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body 80} 81 82define void @unsigned(ptr %x, ptr %y, i32 %n) { 83; CHECK-LABEL: @unsigned( 84; CHECK-NEXT: entry: 85; CHECK-NEXT: [[CMP6:%.*]] = icmp sgt i32 [[N:%.*]], 0 86; CHECK-NEXT: br i1 [[CMP6]], label [[FOR_BODY_PREHEADER:%.*]], label [[FOR_COND_CLEANUP:%.*]] 87; CHECK: for.body.preheader: 88; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[N]] to i64 89; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[WIDE_TRIP_COUNT]], 4 90; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]] 91; CHECK: vector.memcheck: 92; CHECK-NEXT: [[TMP0:%.*]] = shl nuw nsw i64 [[WIDE_TRIP_COUNT]], 2 93; CHECK-NEXT: [[UGLYGEP:%.*]] = getelementptr i8, ptr [[Y:%.*]], i64 [[TMP0]] 94; CHECK-NEXT: [[UGLYGEP1:%.*]] = getelementptr i8, ptr [[X:%.*]], i64 [[TMP0]] 95; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[Y]], [[UGLYGEP1]] 96; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[X]], [[UGLYGEP]] 97; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]] 98; CHECK-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]] 99; CHECK: vector.ph: 100; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[WIDE_TRIP_COUNT]], 4 101; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[WIDE_TRIP_COUNT]], [[N_MOD_VF]] 102; CHECK-NEXT: br label [[VECTOR_BODY:%.*]] 103; CHECK: vector.body: 104; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ] 105; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 0 106; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[TMP1]] 107; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0 108; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4, !alias.scope !8 109; CHECK-NEXT: [[TMP4:%.*]] = call <4 x i32> @llvm.fptoui.sat.v4i32.v4f32(<4 x float> [[WIDE_LOAD]]) 110; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[Y]], i64 [[TMP1]] 111; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0 112; CHECK-NEXT: store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4, !alias.scope !11, !noalias !8 113; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 114; CHECK-NEXT: [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] 115; CHECK-NEXT: br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]] 116; CHECK: middle.block: 117; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]] 118; CHECK-NEXT: br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]] 119; CHECK: scalar.ph: 120; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ] 121; CHECK-NEXT: br label [[FOR_BODY:%.*]] 122; CHECK: for.cond.cleanup.loopexit: 123; CHECK-NEXT: br label [[FOR_COND_CLEANUP]] 124; CHECK: for.cond.cleanup: 125; CHECK-NEXT: ret void 126; CHECK: for.body: 127; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 128; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDVARS_IV]] 129; CHECK-NEXT: [[TMP8:%.*]] = load float, ptr [[ARRAYIDX]], align 4 130; CHECK-NEXT: [[TMP9:%.*]] = tail call i32 @llvm.fptoui.sat.i32.f32(float [[TMP8]]) 131; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[Y]], i64 [[INDVARS_IV]] 132; CHECK-NEXT: store i32 [[TMP9]], ptr [[ARRAYIDX2]], align 4 133; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 134; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]] 135; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]] 136; 137entry: 138 %cmp6 = icmp sgt i32 %n, 0 139 br i1 %cmp6, label %for.body.preheader, label %for.cond.cleanup 140 141for.body.preheader: ; preds = %entry 142 %wide.trip.count = zext i32 %n to i64 143 br label %for.body 144 145for.cond.cleanup: ; preds = %for.body, %entry 146 ret void 147 148for.body: ; preds = %for.body.preheader, %for.body 149 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 150 %arrayidx = getelementptr inbounds float, ptr %x, i64 %indvars.iv 151 %0 = load float, ptr %arrayidx, align 4 152 %1 = tail call i32 @llvm.fptoui.sat.i32.f32(float %0) 153 %arrayidx2 = getelementptr inbounds i32, ptr %y, i64 %indvars.iv 154 store i32 %1, ptr %arrayidx2, align 4 155 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 156 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count 157 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body 158} 159 160declare i32 @llvm.fptosi.sat.i32.f32(float) 161declare i32 @llvm.fptoui.sat.i32.f32(float) 162