1
2; RUN: opt -mtriple=thumbv8.1m.main -mve-tail-predication -tail-predication=enabled -mattr=+mve %s -S -o - | FileCheck %s
3
4; CHECK-LABEL: vec_mul_reduce_add
5
6; CHECK: vector.ph:
7; CHECK:  %start = call i32 @llvm.start.loop.iterations.i32
8; CHECK:  br label %vector.body
9
10; CHECK: vector.body:
11; CHECK: [[ELTS:%[^ ]+]] = phi i32 [ %N, %vector.ph ], [ [[SUB:%[^ ]+]], %vector.body ]
12; CHECK: [[VCTP:%[^ ]+]] = call <4 x i1> @llvm.arm.mve.vctp32(i32 [[ELTS]])
13; CHECK: [[SUB]] = sub i32 [[ELTS]], 4
14; CHECK: call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* {{.*}}, i32 4, <4 x i1> [[VCTP]]
15; CHECK: call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* {{.*}}, i32 4, <4 x i1> [[VCTP]],
16
17; CHECK: middle.block:
18; CHECK: [[VPSEL:%[^ ]+]] = select <4 x i1> [[VCTP]],
19; CHECK: call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[VPSEL]])
20
21define i32 @vec_mul_reduce_add(i32* noalias nocapture readonly %a, i32* noalias nocapture readonly %b, i32 %N) {
22entry:
23  %cmp8 = icmp eq i32 %N, 0
24  %0 = add i32 %N, 3
25  %1 = lshr i32 %0, 2
26  %2 = shl nuw i32 %1, 2
27  %3 = add i32 %2, -4
28  %4 = lshr i32 %3, 2
29  %5 = add nuw nsw i32 %4, 1
30  br i1 %cmp8, label %for.cond.cleanup, label %vector.ph
31
32vector.ph:                                        ; preds = %entry
33  %trip.count.minus.1 = add i32 %N, -1
34  %broadcast.splatinsert11 = insertelement <4 x i32> undef, i32 %trip.count.minus.1, i32 0
35  %broadcast.splat12 = shufflevector <4 x i32> %broadcast.splatinsert11, <4 x i32> undef, <4 x i32> zeroinitializer
36  %start = call i32 @llvm.start.loop.iterations.i32(i32 %5)
37  br label %vector.body
38
39vector.body:                                      ; preds = %vector.body, %vector.ph
40  %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
41  %lsr.iv2 = phi i32* [ %scevgep3, %vector.body ], [ %a, %vector.ph ]
42  %lsr.iv = phi i32* [ %scevgep, %vector.body ], [ %b, %vector.ph ]
43  %vec.phi = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ %9, %vector.body ]
44  %6 = phi i32 [ %start, %vector.ph ], [ %10, %vector.body ]
45  %lsr.iv24 = bitcast i32* %lsr.iv2 to <4 x i32>*
46  %lsr.iv1 = bitcast i32* %lsr.iv to <4 x i32>*
47  %broadcast.splatinsert = insertelement <4 x i32> undef, i32 %index, i32 0
48  %broadcast.splat = shufflevector <4 x i32> %broadcast.splatinsert, <4 x i32> undef, <4 x i32> zeroinitializer
49  %induction = add <4 x i32> %broadcast.splat, <i32 0, i32 1, i32 2, i32 3>
50
51  ; %7 = icmp ule <4 x i32> %induction, %broadcast.splat12
52  %7 = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 %index, i32 %N)
53
54  %wide.masked.load = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %lsr.iv24, i32 4, <4 x i1> %7, <4 x i32> undef)
55  %wide.masked.load13 = call <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>* %lsr.iv1, i32 4, <4 x i1> %7, <4 x i32> undef)
56  %8 = mul nsw <4 x i32> %wide.masked.load13, %wide.masked.load
57  %9 = add nsw <4 x i32> %8, %vec.phi
58  %index.next = add i32 %index, 4
59  %scevgep = getelementptr i32, i32* %lsr.iv, i32 4
60  %scevgep3 = getelementptr i32, i32* %lsr.iv2, i32 4
61  %10 = call i32 @llvm.loop.decrement.reg.i32.i32.i32(i32 %6, i32 1)
62  %11 = icmp ne i32 %10, 0
63  br i1 %11, label %vector.body, label %middle.block
64
65middle.block:                                     ; preds = %vector.body
66  %12 = select <4 x i1> %7, <4 x i32> %9, <4 x i32> %vec.phi
67  %13 = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> %12)
68  br label %for.cond.cleanup
69
70for.cond.cleanup:                                 ; preds = %middle.block, %entry
71  %res.0.lcssa = phi i32 [ 0, %entry ], [ %13, %middle.block ]
72  ret i32 %res.0.lcssa
73}
74
75declare <4 x i32> @llvm.masked.load.v4i32.p0v4i32(<4 x i32>*, i32 immarg, <4 x i1>, <4 x i32>)
76declare i32 @llvm.vector.reduce.add.v4i32(<4 x i32>)
77declare i32 @llvm.start.loop.iterations.i32(i32)
78declare i32 @llvm.loop.decrement.reg.i32.i32.i32(i32, i32)
79declare <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32, i32)
80