1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -passes=loop-idiom < %s -S | FileCheck %s 3target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64" 4 5; CHECK: @.memset_pattern = private unnamed_addr constant [4 x i32] [i32 2, i32 2, i32 2, i32 2], align 16 6 7target triple = "x86_64-apple-darwin10.0.0" 8 9;void test(int *f, unsigned n) { 10; for (unsigned i = 0; i < 2 * n; i += 2) { 11; f[i] = 0; 12; f[i+1] = 0; 13; } 14;} 15define void @test(ptr %f, i32 %n) nounwind ssp { 16; CHECK-LABEL: @test( 17; CHECK-NEXT: entry: 18; CHECK-NEXT: [[MUL:%.*]] = shl i32 [[N:%.*]], 1 19; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[MUL]], 0 20; CHECK-NEXT: br i1 [[CMP1]], label [[FOR_END:%.*]], label [[FOR_BODY_PREHEADER:%.*]] 21; CHECK: for.body.preheader: 22; CHECK-NEXT: [[TMP0:%.*]] = zext i32 [[MUL]] to i64 23; CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[TMP0]], -1 24; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 1 25; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[TMP2]], 3 26; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], 8 27; CHECK-NEXT: call void @llvm.memset.p0.i64(ptr align 4 [[F:%.*]], i8 0, i64 [[TMP4]], i1 false) 28; CHECK-NEXT: br label [[FOR_BODY:%.*]] 29; CHECK: for.body: 30; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 31; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[F]], i64 [[INDVARS_IV]] 32; CHECK-NEXT: [[TMP5:%.*]] = or disjoint i64 [[INDVARS_IV]], 1 33; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[F]], i64 [[TMP5]] 34; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 2 35; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], [[TMP0]] 36; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT:%.*]] 37; CHECK: for.end.loopexit: 38; CHECK-NEXT: br label [[FOR_END]] 39; CHECK: for.end: 40; CHECK-NEXT: ret void 41; 42entry: 43 %mul = shl i32 %n, 1 44 %cmp1 = icmp eq i32 %mul, 0 45 br i1 %cmp1, label %for.end, label %for.body.preheader 46 47for.body.preheader: ; preds = %entry 48 %0 = zext i32 %mul to i64 49 br label %for.body 50 51for.body: ; preds = %for.body.preheader, %for.body 52 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 53 %arrayidx = getelementptr inbounds i32, ptr %f, i64 %indvars.iv 54 store i32 0, ptr %arrayidx, align 4 55 %1 = or disjoint i64 %indvars.iv, 1 56 %arrayidx2 = getelementptr inbounds i32, ptr %f, i64 %1 57 store i32 0, ptr %arrayidx2, align 4 58 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 2 59 %cmp = icmp ult i64 %indvars.iv.next, %0 60 br i1 %cmp, label %for.body, label %for.end.loopexit 61 62for.end.loopexit: ; preds = %for.body 63 br label %for.end 64 65for.end: ; preds = %for.end.loopexit, %entry 66 ret void 67} 68 69;void test_pattern(int *f, unsigned n) { 70; for (unsigned i = 0; i < 2 * n; i += 2) { 71; f[i] = 2; 72; f[i+1] = 2; 73; } 74;} 75define void @test_pattern(ptr %f, i32 %n) nounwind ssp { 76; CHECK-LABEL: @test_pattern( 77; CHECK-NEXT: entry: 78; CHECK-NEXT: [[MUL:%.*]] = shl i32 [[N:%.*]], 1 79; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[MUL]], 0 80; CHECK-NEXT: br i1 [[CMP1]], label [[FOR_END:%.*]], label [[FOR_BODY_PREHEADER:%.*]] 81; CHECK: for.body.preheader: 82; CHECK-NEXT: [[TMP0:%.*]] = zext i32 [[MUL]] to i64 83; CHECK-NEXT: [[TMP1:%.*]] = add nsw i64 [[TMP0]], -1 84; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 1 85; CHECK-NEXT: [[TMP3:%.*]] = shl i64 [[TMP2]], 3 86; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], 8 87; CHECK-NEXT: call void @memset_pattern16(ptr [[F:%.*]], ptr @.memset_pattern, i64 [[TMP4]]) 88; CHECK-NEXT: br label [[FOR_BODY:%.*]] 89; CHECK: for.body: 90; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 91; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[F]], i64 [[INDVARS_IV]] 92; CHECK-NEXT: [[TMP5:%.*]] = or disjoint i64 [[INDVARS_IV]], 1 93; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[F]], i64 [[TMP5]] 94; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 2 95; CHECK-NEXT: [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], [[TMP0]] 96; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT:%.*]] 97; CHECK: for.end.loopexit: 98; CHECK-NEXT: br label [[FOR_END]] 99; CHECK: for.end: 100; CHECK-NEXT: ret void 101; 102entry: 103 %mul = shl i32 %n, 1 104 %cmp1 = icmp eq i32 %mul, 0 105 br i1 %cmp1, label %for.end, label %for.body.preheader 106 107for.body.preheader: ; preds = %entry 108 %0 = zext i32 %mul to i64 109 br label %for.body 110 111for.body: ; preds = %for.body.preheader, %for.body 112 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 113 %arrayidx = getelementptr inbounds i32, ptr %f, i64 %indvars.iv 114 store i32 2, ptr %arrayidx, align 4 115 %1 = or disjoint i64 %indvars.iv, 1 116 %arrayidx2 = getelementptr inbounds i32, ptr %f, i64 %1 117 store i32 2, ptr %arrayidx2, align 4 118 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 2 119 %cmp = icmp ult i64 %indvars.iv.next, %0 120 br i1 %cmp, label %for.body, label %for.end.loopexit 121 122for.end.loopexit: ; preds = %for.body 123 br label %for.end 124 125for.end: ; preds = %for.end.loopexit, %entry 126 ret void 127} 128