1 ; RUN: opt -S -loop-vectorize -scalable-vectorization=on -mattr=+sve -mtriple aarch64-linux-gnu < %s | FileCheck %s
3 define void @invariant_load(i64 %n, i32* noalias nocapture %a, i32* nocapture readonly %b) {
4 ; CHECK-LABEL: @invariant_load
6 ; CHECK: %[[GEP:.*]] = getelementptr inbounds i32, i32* %b, i64 42
7 ; CHECK-NEXT: %[[INVLOAD:.*]] = load i32, i32* %[[GEP]]
8 ; CHECK-NEXT: %[[SPLATINS:.*]] = insertelement <vscale x 4 x i32> poison, i32 %[[INVLOAD]], i32 0
9 ; CHECK-NEXT: %[[SPLAT:.*]] = shufflevector <vscale x 4 x i32> %[[SPLATINS]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
10 ; CHECK: %[[LOAD:.*]] = load <vscale x 4 x i32>, <vscale x 4 x i32>*
11 ; CHECK-NEXT: %[[ADD:.*]] = add nsw <vscale x 4 x i32> %[[SPLAT]], %[[LOAD]]
12 ; CHECK: store <vscale x 4 x i32> %[[ADD]], <vscale x 4 x i32>*
16 for.body: ; preds = %for.body.lr.ph, %for.body
17 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
18 %arrayidx = getelementptr inbounds i32, i32* %b, i64 42
19 %0 = load i32, i32* %arrayidx, align 4
20 %arrayidx1 = getelementptr inbounds i32, i32* %b, i64 %iv
21 %1 = load i32, i32* %arrayidx1, align 4
22 %add = add nsw i32 %0, %1
23 %arrayidx2 = getelementptr inbounds i32, i32* %a, i64 %iv
24 store i32 %add, i32* %arrayidx2, align 4
25 %iv.next = add nuw nsw i64 %iv, 1
26 %exitcond.not = icmp eq i64 %iv.next, %n
27 br i1 %exitcond.not, label %for.end, label %for.body, !llvm.loop !1
29 for.end: ; preds = %for.body
33 !1 = distinct !{!1, !2, !3, !4}
34 !2 = !{!"llvm.loop.vectorize.width", i32 4}
35 !3 = !{!"llvm.loop.interleave.count", i32 1}
36 !4 = !{!"llvm.loop.vectorize.scalable.enable", i1 true}