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:64:64:64:32-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: [[TMP0:%.*]] = shl i32 [[N:%.*]], 1 19; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[TMP0]], 0 20; CHECK-NEXT: br i1 [[CMP1]], label [[FOR_END:%.*]], label [[FOR_BODY_PREHEADER:%.*]] 21; CHECK: for.body.preheader: 22; CHECK-NEXT: [[TMP1:%.*]] = add i32 [[TMP0]], -1 23; CHECK-NEXT: [[TMP2:%.*]] = lshr i32 [[TMP1]], 1 24; CHECK-NEXT: [[TMP3:%.*]] = shl i32 [[TMP2]], 3 25; CHECK-NEXT: [[TMP4:%.*]] = add i32 [[TMP3]], 8 26; CHECK-NEXT: call void @llvm.memset.p0.i32(ptr align 4 [[F:%.*]], i8 0, i32 [[TMP4]], i1 false) 27; CHECK-NEXT: br label [[FOR_BODY:%.*]] 28; CHECK: for.body: 29; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 30; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[F]], i32 [[INDVARS_IV]] 31; CHECK-NEXT: [[TMP5:%.*]] = or disjoint i32 [[INDVARS_IV]], 1 32; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[F]], i32 [[TMP5]] 33; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 2 34; CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[INDVARS_IV_NEXT]], [[TMP0]] 35; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT:%.*]] 36; CHECK: for.end.loopexit: 37; CHECK-NEXT: br label [[FOR_END]] 38; CHECK: for.end: 39; CHECK-NEXT: ret void 40; 41entry: 42 %0 = shl i32 %n, 1 43 %cmp1 = icmp eq i32 %0, 0 44 br i1 %cmp1, label %for.end, label %for.body.preheader 45 46for.body.preheader: ; preds = %entry 47 br label %for.body 48 49for.body: ; preds = %for.body.preheader, %for.body 50 %indvars.iv = phi i32 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 51 %arrayidx = getelementptr inbounds i32, ptr %f, i32 %indvars.iv 52 store i32 0, ptr %arrayidx, align 4 53 %1 = or disjoint i32 %indvars.iv, 1 54 %arrayidx2 = getelementptr inbounds i32, ptr %f, i32 %1 55 store i32 0, ptr %arrayidx2, align 4 56 %indvars.iv.next = add nuw nsw i32 %indvars.iv, 2 57 %cmp = icmp ult i32 %indvars.iv.next, %0 58 br i1 %cmp, label %for.body, label %for.end.loopexit 59 60for.end.loopexit: ; preds = %for.body 61 br label %for.end 62 63for.end: ; preds = %for.end.loopexit, %entry 64 ret void 65} 66 67;void test_pattern(int *f, unsigned n) { 68; for (unsigned i = 0; i < 2 * n; i += 2) { 69; f[i] = 2; 70; f[i+1] = 2; 71; } 72;} 73define void @test_pattern(ptr %f, i32 %n) nounwind ssp { 74; CHECK-LABEL: @test_pattern( 75; CHECK-NEXT: entry: 76; CHECK-NEXT: [[MUL:%.*]] = shl i32 [[N:%.*]], 1 77; CHECK-NEXT: [[CMP1:%.*]] = icmp eq i32 [[MUL]], 0 78; CHECK-NEXT: br i1 [[CMP1]], label [[FOR_END:%.*]], label [[FOR_BODY_PREHEADER:%.*]] 79; CHECK: for.body.preheader: 80; CHECK-NEXT: [[TMP0:%.*]] = add i32 [[MUL]], -1 81; CHECK-NEXT: [[TMP1:%.*]] = lshr i32 [[TMP0]], 1 82; CHECK-NEXT: [[TMP2:%.*]] = shl i32 [[TMP1]], 3 83; CHECK-NEXT: [[TMP3:%.*]] = add i32 [[TMP2]], 8 84; CHECK-NEXT: call void @memset_pattern16(ptr [[F:%.*]], ptr @.memset_pattern, i32 [[TMP3]]) 85; CHECK-NEXT: br label [[FOR_BODY:%.*]] 86; CHECK: for.body: 87; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 88; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[F]], i32 [[INDVARS_IV]] 89; CHECK-NEXT: [[X1:%.*]] = or disjoint i32 [[INDVARS_IV]], 1 90; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[F]], i32 [[X1]] 91; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i32 [[INDVARS_IV]], 2 92; CHECK-NEXT: [[CMP:%.*]] = icmp ult i32 [[INDVARS_IV_NEXT]], [[MUL]] 93; CHECK-NEXT: br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END_LOOPEXIT:%.*]] 94; CHECK: for.end.loopexit: 95; CHECK-NEXT: br label [[FOR_END]] 96; CHECK: for.end: 97; CHECK-NEXT: ret void 98; 99entry: 100 %mul = shl i32 %n, 1 101 %cmp1 = icmp eq i32 %mul, 0 102 br i1 %cmp1, label %for.end, label %for.body.preheader 103 104for.body.preheader: ; preds = %entry 105 br label %for.body 106 107for.body: ; preds = %for.body.preheader, %for.body 108 %indvars.iv = phi i32 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ] 109 %arrayidx = getelementptr inbounds i32, ptr %f, i32 %indvars.iv 110 store i32 2, ptr %arrayidx, align 4 111 %x1 = or disjoint i32 %indvars.iv, 1 112 %arrayidx2 = getelementptr inbounds i32, ptr %f, i32 %x1 113 store i32 2, ptr %arrayidx2, align 4 114 %indvars.iv.next = add nuw nsw i32 %indvars.iv, 2 115 %cmp = icmp ult i32 %indvars.iv.next, %mul 116 br i1 %cmp, label %for.body, label %for.end.loopexit 117 118for.end.loopexit: ; preds = %for.body 119 br label %for.end 120 121for.end: ; preds = %for.end.loopexit, %entry 122 ret void 123} 124