1 ; RUN: llc -asm-verbose=false < %s | FileCheck %s
3 target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
4 target triple = "armv8.2a-arm-unknown-eabihf"
6 define dso_local void @vec8(half* nocapture readonly %V, i32 %N) local_unnamed_addr #0 {
8 ; CHECK-NEXT: vld1.16 {d16, d17}, [r0]!
9 ; CHECK-NEXT: subs r1, r1, #8
10 ; CHECK-NEXT: bne .LBB0_1
15 %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
16 %0 = getelementptr inbounds half, half* %V, i32 %index
17 %1 = bitcast half* %0 to <8 x half>*
18 %wide.load = load volatile <8 x half>, <8 x half>* %1, align 2
19 %index.next = add i32 %index, 8
20 %cmp = icmp eq i32 %index.next, %N
21 br i1 %cmp, label %byeblock, label %vector.body
27 define dso_local void @vec4(half* nocapture readonly %V, i32 %N) local_unnamed_addr #0 {
29 ; CHECK-NEXT: vld1.16 {d16}, [r0]!
30 ; CHECK-NEXT: subs r1, r1, #4
31 ; CHECK-NEXT: bne .LBB1_1
36 %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
37 %0 = getelementptr inbounds half, half* %V, i32 %index
38 %1 = bitcast half* %0 to <4 x half>*
39 %wide.load = load volatile <4 x half>, <4 x half>* %1, align 2
40 %index.next = add i32 %index, 4
41 %cmp = icmp eq i32 %index.next, %N
42 br i1 %cmp, label %byeblock, label %vector.body
48 attributes #0 = { norecurse nounwind readonly "no-infs-fp-math"="true" "no-nans-fp-math"="true" "no-signed-zeros-fp-math"="true" "no-trapping-math"="true" "target-cpu"="generic" "target-features"="+armv8.2-a,+fullfp16,+strict-align,-thumb-mode" "unsafe-fp-math"="true" "use-soft-float"="false" }