1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -S -mtriple=x86_64-pc-linux-gnu -mcpu=generic -passes=slp-vectorizer -pass-remarks-output=%t < %s | FileCheck %s 3; RUN: FileCheck --input-file=%t --check-prefix=YAML %s 4 5define i32 @foo(ptr nocapture readonly %diff) #0 { 6; CHECK-LABEL: @foo( 7; CHECK-NEXT: entry: 8; CHECK-NEXT: [[M2:%.*]] = alloca [8 x [8 x i32]], align 16 9; CHECK-NEXT: br label [[FOR_BODY:%.*]] 10; CHECK: for.body: 11; CHECK-NEXT: [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ] 12; CHECK-NEXT: [[A_088:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[ADD24:%.*]], [[FOR_BODY]] ] 13; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[INDVARS_IV]], 3 14; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[DIFF:%.*]], i64 [[TMP0]] 15; CHECK-NEXT: [[TMP1:%.*]] = or disjoint i64 [[TMP0]], 4 16; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[DIFF]], i64 [[TMP1]] 17; CHECK-NEXT: [[ARRAYIDX6:%.*]] = getelementptr inbounds [8 x [8 x i32]], ptr [[M2]], i64 0, i64 [[INDVARS_IV]], i64 0 18; CHECK-NEXT: [[TMP2:%.*]] = load <2 x i32>, ptr [[ARRAYIDX]], align 4 19; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr [[ARRAYIDX2]], align 4 20; CHECK-NEXT: [[TMP4:%.*]] = add nsw <2 x i32> [[TMP3]], [[TMP2]] 21; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x i32> [[TMP4]], i32 0 22; CHECK-NEXT: [[ADD10:%.*]] = add nsw i32 [[TMP5]], [[A_088]] 23; CHECK-NEXT: store <2 x i32> [[TMP4]], ptr [[ARRAYIDX6]], align 16 24; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i32> [[TMP4]], i32 1 25; CHECK-NEXT: [[ADD24]] = add nsw i32 [[ADD10]], [[TMP6]] 26; CHECK-NEXT: [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1 27; CHECK-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 8 28; CHECK-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]] 29; CHECK: for.end: 30; CHECK-NEXT: ret i32 [[ADD24]] 31; 32entry: 33 %m2 = alloca [8 x [8 x i32]], align 16 34 br label %for.body 35 36for.body: ; preds = %for.body, %entry 37 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ] 38 %a.088 = phi i32 [ 0, %entry ], [ %add24, %for.body ] 39 %0 = shl i64 %indvars.iv, 3 40 %arrayidx = getelementptr inbounds i32, ptr %diff, i64 %0 41 %1 = load i32, ptr %arrayidx, align 4 42 %2 = or disjoint i64 %0, 4 43 %arrayidx2 = getelementptr inbounds i32, ptr %diff, i64 %2 44 %3 = load i32, ptr %arrayidx2, align 4 45 %add3 = add nsw i32 %3, %1 46 %arrayidx6 = getelementptr inbounds [8 x [8 x i32]], ptr %m2, i64 0, i64 %indvars.iv, i64 0 47 store i32 %add3, ptr %arrayidx6, align 16 48 %add10 = add nsw i32 %add3, %a.088 49 %4 = or disjoint i64 %0, 1 50 %arrayidx13 = getelementptr inbounds i32, ptr %diff, i64 %4 51 %5 = load i32, ptr %arrayidx13, align 4 52 %6 = or disjoint i64 %0, 5 53 %arrayidx16 = getelementptr inbounds i32, ptr %diff, i64 %6 54 %7 = load i32, ptr %arrayidx16, align 4 55 %add17 = add nsw i32 %7, %5 56 %arrayidx20 = getelementptr inbounds [8 x [8 x i32]], ptr %m2, i64 0, i64 %indvars.iv, i64 1 57 store i32 %add17, ptr %arrayidx20, align 4 58 %add24 = add nsw i32 %add10, %add17 59 60 ; YAML: Pass: slp-vectorizer 61 ; YAML-NEXT: Name: StoresVectorized 62 ; YAML-NEXT: Function: foo 63 ; YAML-NEXT: Args: 64 ; YAML-NEXT: - String: 'Stores SLP vectorized with cost ' 65 ; YAML-NEXT: - Cost: '-1' 66 ; YAML-NEXT: - String: ' and with tree size ' 67 ; YAML-NEXT: - TreeSize: '4' 68 69 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 70 %exitcond = icmp eq i64 %indvars.iv.next, 8 71 br i1 %exitcond, label %for.end, label %for.body 72 73for.end: ; preds = %for.body 74 ret i32 %add24 75} 76 77