1; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 5 2; REQUIRES: asserts 3; RUN: opt -passes=loop-distribute -enable-loop-distribute \ 4; RUN: -debug-only=loop-distribute -disable-output 2>&1 %s | FileCheck %s 5 6define void @f(ptr noalias %a, ptr noalias %b, ptr noalias %c, ptr noalias %d, i64 %stride) { 7; CHECK-LABEL: 'f' 8; CHECK-NEXT: LDist: Found a candidate loop: for.body 9; CHECK-NEXT: LDist: Backward dependences: 10; CHECK-NEXT: Backward: 11; CHECK-NEXT: %load.a = load i32, ptr %gep.a, align 4 -> 12; CHECK-NEXT: store i32 %mul.a, ptr %gep.a.plus4, align 4 13; CHECK-NEXT: LDist: Seeded partitions: 14; CHECK-NEXT: LDist: Partition 0: (cycle) 15; CHECK-NEXT: for.body: %load.a = load i32, ptr %gep.a, align 4 16; CHECK-NEXT: for.body: %load.b = load i32, ptr %gep.b, align 4 17; CHECK-NEXT: for.body: store i32 %mul.a, ptr %gep.a.plus4, align 4 18; CHECK-NEXT: LDist: Partition 1: 19; CHECK-NEXT: for.body: %loadD = load i32, ptr %gep.d, align 4 20; CHECK-NEXT: LDist: Partition 2: 21; CHECK-NEXT: for.body: %load.strided.a = load i32, ptr %gep.strided.a, align 4 22; CHECK-NEXT: LDist: Partition 3: 23; CHECK-NEXT: for.body: store i32 %mul.c, ptr %gep.c, align 4 24; CHECK-NEXT: LDist: Merged partitions: 25; CHECK-NEXT: LDist: Partition 0: (cycle) 26; CHECK-NEXT: for.body: %load.a = load i32, ptr %gep.a, align 4 27; CHECK-NEXT: for.body: %load.b = load i32, ptr %gep.b, align 4 28; CHECK-NEXT: for.body: store i32 %mul.a, ptr %gep.a.plus4, align 4 29; CHECK-NEXT: LDist: Partition 1: 30; CHECK-NEXT: for.body: %loadD = load i32, ptr %gep.d, align 4 31; CHECK-NEXT: for.body: %load.strided.a = load i32, ptr %gep.strided.a, align 4 32; CHECK-NEXT: for.body: store i32 %mul.c, ptr %gep.c, align 4 33; CHECK-NEXT: LDist: Populated partitions: 34; CHECK-NEXT: LDist: Partition 0: (cycle) 35; CHECK-NEXT: for.body: %load.a = load i32, ptr %gep.a, align 4 36; CHECK-NEXT: for.body: %load.b = load i32, ptr %gep.b, align 4 37; CHECK-NEXT: for.body: store i32 %mul.a, ptr %gep.a.plus4, align 4 38; CHECK-NEXT: for.body: br i1 %exitcond, label %exit, label %for.body 39; CHECK-NEXT: for.body: %exitcond = icmp eq i64 %add, 20 40; CHECK-NEXT: for.body: %add = add nuw nsw i64 %ind, 1 41; CHECK-NEXT: for.body: %ind = phi i64 [ 0, %entry ], [ %add, %for.body ] 42; CHECK-NEXT: for.body: %mul.a = mul i32 %load.b, %load.a 43; CHECK-NEXT: for.body: %gep.a.plus4 = getelementptr inbounds i32, ptr %a, i64 %add 44; CHECK-NEXT: for.body: %gep.b = getelementptr inbounds i32, ptr %b, i64 %ind 45; CHECK-NEXT: for.body: %gep.a = getelementptr inbounds i32, ptr %a, i64 %ind 46; CHECK-NEXT: LDist: Partition 1: 47; CHECK-NEXT: for.body: %loadD = load i32, ptr %gep.d, align 4 48; CHECK-NEXT: for.body: %load.strided.a = load i32, ptr %gep.strided.a, align 4 49; CHECK-NEXT: for.body: store i32 %mul.c, ptr %gep.c, align 4 50; CHECK-NEXT: for.body: br i1 %exitcond, label %exit, label %for.body 51; CHECK-NEXT: for.body: %exitcond = icmp eq i64 %add, 20 52; CHECK-NEXT: for.body: %add = add nuw nsw i64 %ind, 1 53; CHECK-NEXT: for.body: %ind = phi i64 [ 0, %entry ], [ %add, %for.body ] 54; CHECK-NEXT: for.body: %mul.c = mul i32 %loadD, %load.strided.a 55; CHECK-NEXT: for.body: %gep.c = getelementptr inbounds i32, ptr %c, i64 %ind 56; CHECK-NEXT: for.body: %gep.strided.a = getelementptr inbounds i32, ptr %a, i64 %mul 57; CHECK-NEXT: for.body: %mul = mul i64 %ind, %stride 58; CHECK-NEXT: for.body: %gep.d = getelementptr inbounds i32, ptr %d, i64 %ind 59; CHECK-NEXT: LDist: Distributing loop: for.body 60; CHECK-NEXT: LDist: Pointers: 61; CHECK-NEXT: LDist: After removing unused Instrs: 62; CHECK-NEXT: LDist: Partition 0: 63; CHECK-NEXT: for.body.ldist1: ; preds = %for.body.ldist1, %for.body.ph.ldist1 64; CHECK-NEXT: %ind.ldist1 = phi i64 [ 0, %for.body.ph.ldist1 ], [ %add.ldist1, %for.body.ldist1 ] 65; CHECK-NEXT: %gep.a.ldist1 = getelementptr inbounds i32, ptr %a, i64 %ind.ldist1 66; CHECK-NEXT: %load.a.ldist1 = load i32, ptr %gep.a.ldist1, align 4 67; CHECK-NEXT: %gep.b.ldist1 = getelementptr inbounds i32, ptr %b, i64 %ind.ldist1 68; CHECK-NEXT: %load.b.ldist1 = load i32, ptr %gep.b.ldist1, align 4 69; CHECK-NEXT: %mul.a.ldist1 = mul i32 %load.b.ldist1, %load.a.ldist1 70; CHECK-NEXT: %add.ldist1 = add nuw nsw i64 %ind.ldist1, 1 71; CHECK-NEXT: %gep.a.plus4.ldist1 = getelementptr inbounds i32, ptr %a, i64 %add.ldist1 72; CHECK-NEXT: store i32 %mul.a.ldist1, ptr %gep.a.plus4.ldist1, align 4 73; CHECK-NEXT: %exitcond.ldist1 = icmp eq i64 %add.ldist1, 20 74; CHECK-NEXT: br i1 %exitcond.ldist1, label %for.body.ph, label %for.body.ldist1 75; CHECK-NEXT: LDist: Partition 1: 76; CHECK-NEXT: for.body: ; preds = %for.body, %for.body.ph 77; CHECK-NEXT: %ind = phi i64 [ 0, %for.body.ph ], [ %add, %for.body ] 78; CHECK-NEXT: %add = add nuw nsw i64 %ind, 1 79; CHECK-NEXT: %gep.d = getelementptr inbounds i32, ptr %d, i64 %ind 80; CHECK-NEXT: %loadD = load i32, ptr %gep.d, align 4 81; CHECK-NEXT: %mul = mul i64 %ind, %stride 82; CHECK-NEXT: %gep.strided.a = getelementptr inbounds i32, ptr %a, i64 %mul 83; CHECK-NEXT: %load.strided.a = load i32, ptr %gep.strided.a, align 4 84; CHECK-NEXT: %mul.c = mul i32 %loadD, %load.strided.a 85; CHECK-NEXT: %gep.c = getelementptr inbounds i32, ptr %c, i64 %ind 86; CHECK-NEXT: store i32 %mul.c, ptr %gep.c, align 4 87; CHECK-NEXT: %exitcond = icmp eq i64 %add, 20 88; CHECK-NEXT: br i1 %exitcond, label %exit.loopexit1, label %for.body 89; 90entry: 91 br label %for.body 92 93for.body: ; preds = %for.body, %entry 94 %ind = phi i64 [ 0, %entry ], [ %add, %for.body ] 95 %gep.a = getelementptr inbounds i32, ptr %a, i64 %ind 96 %load.a = load i32, ptr %gep.a, align 4 97 %gep.b = getelementptr inbounds i32, ptr %b, i64 %ind 98 %load.b = load i32, ptr %gep.b, align 4 99 %mul.a = mul i32 %load.b, %load.a 100 %add = add nuw nsw i64 %ind, 1 101 %gep.a.plus4 = getelementptr inbounds i32, ptr %a, i64 %add 102 store i32 %mul.a, ptr %gep.a.plus4, align 4 103 %gep.d = getelementptr inbounds i32, ptr %d, i64 %ind 104 %loadD = load i32, ptr %gep.d, align 4 105 %mul = mul i64 %ind, %stride 106 %gep.strided.a = getelementptr inbounds i32, ptr %a, i64 %mul 107 %load.strided.a = load i32, ptr %gep.strided.a, align 4 108 %mul.c = mul i32 %loadD, %load.strided.a 109 %gep.c = getelementptr inbounds i32, ptr %c, i64 %ind 110 store i32 %mul.c, ptr %gep.c, align 4 111 %exitcond = icmp eq i64 %add, 20 112 br i1 %exitcond, label %exit, label %for.body 113 114exit: ; preds = %for.body 115 ret void 116} 117