1; RUN: opt -loop-vectorize -scalable-vectorization=on -S < %s | FileCheck %s
2
3target triple = "aarch64-unknown-linux-gnu"
4
5define void @inv_store_i16(i16* noalias %dst, i16* noalias readonly %src, i64 %N) #0 {
6; CHECK-LABEL: @inv_store_i16(
7; CHECK:       vector.ph:
8; CHECK:         %[[TMP1:.*]] = insertelement <vscale x 4 x i16*> poison, i16* %dst, i32 0
9; CHECK-NEXT:    %[[SPLAT_PTRS:.*]] = shufflevector <vscale x 4 x i16*> %[[TMP1]], <vscale x 4 x i16*> poison, <vscale x 4 x i32> zeroinitializer
10; CHECK:       vector.body:
11; CHECK:         %[[VECLOAD:.*]] = load <vscale x 4 x i16>, <vscale x 4 x i16>* %{{.*}}, align 2
12; CHECK-NEXT:    call void @llvm.masked.scatter.nxv4i16.nxv4p0i16(<vscale x 4 x i16> %[[VECLOAD]], <vscale x 4 x i16*> %[[SPLAT_PTRS]], i32 2
13entry:
14  br label %for.body14
15
16for.body14:                                       ; preds = %for.body14, %entry
17  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body14 ]
18  %arrayidx = getelementptr inbounds i16, i16* %src, i64 %indvars.iv
19  %ld = load i16, i16* %arrayidx
20  store i16 %ld, i16* %dst, align 2
21  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
22  %exitcond.not = icmp eq i64 %indvars.iv.next, %N
23  br i1 %exitcond.not, label %for.inc24, label %for.body14, !llvm.loop !0
24
25for.inc24:                                        ; preds = %for.body14, %for.body
26  ret void
27}
28
29
30define void @cond_inv_store_i32(i32* noalias %dst, i32* noalias readonly %src, i64 %N) #0 {
31; CHECK-LABEL: @cond_inv_store_i32(
32; CHECK:       vector.ph:
33; CHECK:         %[[TMP1:.*]] = insertelement <vscale x 4 x i32*> poison, i32* %dst, i32 0
34; CHECK-NEXT:    %[[SPLAT_PTRS:.*]] = shufflevector <vscale x 4 x i32*> %[[TMP1]], <vscale x 4 x i32*> poison, <vscale x 4 x i32> zeroinitializer
35; CHECK:       vector.body:
36; CHECK:         %[[VECLOAD:.*]] = load <vscale x 4 x i32>, <vscale x 4 x i32>* %{{.*}}, align 4
37; CHECK-NEXT:    %[[MASK:.*]] = icmp sgt <vscale x 4 x i32> %[[VECLOAD]], shufflevector (<vscale x 4 x i32> insertelement (<vscale x 4 x i32> poison, i32 0, i32 0), <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer)
38; CHECK-NEXT:    call void @llvm.masked.scatter.nxv4i32.nxv4p0i32(<vscale x 4 x i32> %[[VECLOAD]], <vscale x 4 x i32*> %[[SPLAT_PTRS]], i32 4, <vscale x 4 x i1> %[[MASK]])
39entry:
40  br label %for.body
41
42for.body:                                         ; preds = %entry, %for.inc
43  %i.09 = phi i64 [ %inc, %for.inc ], [ 0, %entry ]
44  %arrayidx = getelementptr inbounds i32, i32* %src, i64 %i.09
45  %0 = load i32, i32* %arrayidx, align 4
46  %cmp1 = icmp sgt i32 %0, 0
47  br i1 %cmp1, label %if.then, label %for.inc
48
49if.then:                                          ; preds = %for.body
50  store i32 %0, i32* %dst, align 4
51  br label %for.inc
52
53for.inc:                                          ; preds = %for.body, %if.then
54  %inc = add nuw nsw i64 %i.09, 1
55  %exitcond.not = icmp eq i64 %inc, %N
56  br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !0
57
58for.end:                                          ; preds = %for.inc, %entry
59  ret void
60}
61
62attributes #0 = { "target-features"="+neon,+sve" }
63
64!0 = distinct !{!0, !1, !2, !3, !4, !5}
65!1 = !{!"llvm.loop.mustprogress"}
66!2 = !{!"llvm.loop.vectorize.width", i32 4}
67!3 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}
68!4 = !{!"llvm.loop.vectorize.enable", i1 true}
69!5 = !{!"llvm.loop.interleave.count", i32 1}
70
71