1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=slp-vectorizer -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s 3 4define i32 @foo(ptr nocapture readonly %diff) #0 { 5; CHECK-LABEL: @foo( 6; CHECK-NEXT: entry: 7; CHECK-NEXT: [[M2:%.*]] = alloca [8 x [8 x i32]], align 16 8; CHECK-NEXT: br label [[FOR_BODY:%.*]] 9; CHECK: for.body: 10; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 11; CHECK-NEXT: [[A_088:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[OP_RDX:%.*]], [[FOR_BODY]] ] 12; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[INDVARS_IV]], 3 13; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[DIFF:%.*]], i64 [[TMP0]] 14; CHECK-NEXT: [[TMP1:%.*]] = or disjoint i64 [[TMP0]], 4 15; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[DIFF]], i64 [[TMP1]] 16; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds [8 x [8 x i32]], ptr [[M2]], i64 0, i64 [[INDVARS_IV]], i64 0 17; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[ARRAYIDX]], align 4 18; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr [[ARRAYIDX2]], align 4 19; CHECK-NEXT: [[TMP4:%.*]] = add nsw <4 x i32> [[TMP3]], [[TMP2]] 20; CHECK-NEXT: store <4 x i32> [[TMP4]], ptr [[ARRAYIDX6]], align 16 21; CHECK-NEXT: [[TMP5:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP4]]) 22; CHECK-NEXT: [[OP_RDX]] = add i32 [[TMP5]], [[A_088]] 23; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 24; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 8 25; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]] 26; CHECK: for.end: 27; CHECK-NEXT: call void @ff(ptr [[M2]]) 28; CHECK-NEXT: ret i32 [[OP_RDX]] 29; 30entry: 31 %m2 = alloca [8 x [8 x i32]], align 16 32 br label %for.body 33 34for.body: ; preds = %for.body, %entry 35 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] 36 %a.088 = phi i32 [ 0, %entry ], [ %add52, %for.body ] 37 %0 = shl i64 %indvars.iv, 3 38 %arrayidx = getelementptr inbounds i32, ptr %diff, i64 %0 39 %1 = load i32, ptr %arrayidx, align 4 40 %2 = or disjoint i64 %0, 4 41 %arrayidx2 = getelementptr inbounds i32, ptr %diff, i64 %2 42 %3 = load i32, ptr %arrayidx2, align 4 43 %add3 = add nsw i32 %3, %1 44 %arrayidx6 = getelementptr inbounds [8 x [8 x i32]], ptr %m2, i64 0, i64 %indvars.iv, i64 0 45 store i32 %add3, ptr %arrayidx6, align 16 46 %add10 = add nsw i32 %add3, %a.088 47 %4 = or disjoint i64 %0, 1 48 %arrayidx13 = getelementptr inbounds i32, ptr %diff, i64 %4 49 %5 = load i32, ptr %arrayidx13, align 4 50 %6 = or disjoint i64 %0, 5 51 %arrayidx16 = getelementptr inbounds i32, ptr %diff, i64 %6 52 %7 = load i32, ptr %arrayidx16, align 4 53 %add17 = add nsw i32 %7, %5 54 %arrayidx20 = getelementptr inbounds [8 x [8 x i32]], ptr %m2, i64 0, i64 %indvars.iv, i64 1 55 store i32 %add17, ptr %arrayidx20, align 4 56 %add24 = add nsw i32 %add10, %add17 57 %8 = or disjoint i64 %0, 2 58 %arrayidx27 = getelementptr inbounds i32, ptr %diff, i64 %8 59 %9 = load i32, ptr %arrayidx27, align 4 60 %10 = or disjoint i64 %0, 6 61 %arrayidx30 = getelementptr inbounds i32, ptr %diff, i64 %10 62 %11 = load i32, ptr %arrayidx30, align 4 63 %add31 = add nsw i32 %11, %9 64 %arrayidx34 = getelementptr inbounds [8 x [8 x i32]], ptr %m2, i64 0, i64 %indvars.iv, i64 2 65 store i32 %add31, ptr %arrayidx34, align 8 66 %add38 = add nsw i32 %add24, %add31 67 %12 = or disjoint i64 %0, 3 68 %arrayidx41 = getelementptr inbounds i32, ptr %diff, i64 %12 69 %13 = load i32, ptr %arrayidx41, align 4 70 %14 = or disjoint i64 %0, 7 71 %arrayidx44 = getelementptr inbounds i32, ptr %diff, i64 %14 72 %15 = load i32, ptr %arrayidx44, align 4 73 %add45 = add nsw i32 %15, %13 74 %arrayidx48 = getelementptr inbounds [8 x [8 x i32]], ptr %m2, i64 0, i64 %indvars.iv, i64 3 75 store i32 %add45, ptr %arrayidx48, align 4 76 %add52 = add nsw i32 %add38, %add45 77 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 78 %exitcond = icmp eq i64 %indvars.iv.next, 8 79 br i1 %exitcond, label %for.end, label %for.body 80 81for.end: ; preds = %for.body 82 call void @ff(ptr %m2) #1 83 ret i32 %add52 84} 85 86declare void @ff(ptr) #2 87 88 89