1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=thumbv8.1m.main -mattr=+mve -tail-predication=enabled %s -o - | FileCheck %s
3
4define dso_local arm_aapcs_vfpcc i32 @minmaxval4(i32* nocapture readonly %x, i32* nocapture %minp) {
5; CHECK-LABEL: minmaxval4:
6; CHECK:       @ %bb.0: @ %entry
7; CHECK-NEXT:    push {r7, lr}
8; CHECK-NEXT:    vmov.i32 q0, #0x80000000
9; CHECK-NEXT:    vmvn.i32 q1, #0x80000000
10; CHECK-NEXT:    movs r2, #10
11; CHECK-NEXT:    dlstp.32 lr, r2
12; CHECK-NEXT:  .LBB0_1: @ %vector.body
13; CHECK-NEXT:    @ =>This Inner Loop Header: Depth=1
14; CHECK-NEXT:    vldrw.u32 q2, [r0], #16
15; CHECK-NEXT:    vpt.s32 gt, q2, q0
16; CHECK-NEXT:    vmovt q0, q2
17; CHECK-NEXT:    vpt.s32 gt, q1, q2
18; CHECK-NEXT:    vmovt q1, q2
19; CHECK-NEXT:    letp lr, .LBB0_1
20; CHECK-NEXT:  @ %bb.2: @ %middle.block
21; CHECK-NEXT:    mvn r0, #-2147483648
22; CHECK-NEXT:    vminv.s32 r0, q1
23; CHECK-NEXT:    str r0, [r1]
24; CHECK-NEXT:    mov.w r0, #-2147483648
25; CHECK-NEXT:    vmaxv.s32 r0, q0
26; CHECK-NEXT:    pop {r7, pc}
27entry:
28  br label %vector.body
29
30vector.body:                                      ; preds = %vector.body, %entry
31  %index = phi i32 [ 0, %entry ], [ %index.next, %vector.body ]
32  %vec.phi = phi <4 x i32> [ <i32 2147483647, i32 2147483647, i32 2147483647, i32 2147483647>, %entry ], [ %5, %vector.body ]
33  %vec.phi29 = phi <4 x i32> [ <i32 -2147483648, i32 -2147483648, i32 -2147483648, i32 -2147483648>, %entry ], [ %7, %vector.body ]
34  %0 = getelementptr inbounds i32, i32* %x, i32 %index
35  %active.lane.mask = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %index, i32 10)
36  %1 = bitcast i32* %0 to <4 x i32>*
37  %wide.masked.load = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %1, i32 4, <4 x i1> %active.lane.mask, <4 x i32> undef)
38  %2 = icmp sgt <4 x i32> %wide.masked.load, %vec.phi29
39  %3 = icmp slt <4 x i32> %wide.masked.load, %vec.phi
40  %4 = and <4 x i1> %active.lane.mask, %3
41  %5 = select <4 x i1> %4, <4 x i32> %wide.masked.load, <4 x i32> %vec.phi
42  %6 = and <4 x i1> %active.lane.mask, %2
43  %7 = select <4 x i1> %6, <4 x i32> %wide.masked.load, <4 x i32> %vec.phi29
44  %index.next = add i32 %index, 4
45  %8 = icmp eq i32 %index.next, 12
46  br i1 %8, label %middle.block, label %vector.body
47
48middle.block:                                     ; preds = %vector.body
49  %9 = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> %7)
50  %10 = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> %5)
51  store i32 %10, i32* %minp, align 4
52  ret i32 %9
53}
54
55declare <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32, i32) #1
56declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32 immarg, <4 x i1>, <4 x i32>) #2
57declare i32 @llvm.vector.reduce.smin.v4i32(<4 x i32>) #3
58declare i32 @llvm.vector.reduce.smax.v4i32(<4 x i32>) #3
59
60