1 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2 ; RUN: llc -O3 -tail-predication=enabled -mtriple=thumbv8.1m.main -mattr=+mve,+mve.fp %s -o - | FileCheck %s
4 target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
5 target triple = "thumbv8.1m-arm-none-eabi"
7 ; Tests that LSR will not interfere with the VCTP intrinsic,
8 ; and that this loop will correctly become tail-predicated.
10 define arm_aapcs_vfpcc float @vctpi32(ptr %0, i32 %1) {
11 ; CHECK-LABEL: vctpi32:
13 ; CHECK-NEXT: push {r4, lr}
14 ; CHECK-NEXT: mvn r3, #31
15 ; CHECK-NEXT: vmov.32 q2[0], r0
16 ; CHECK-NEXT: movs r4, #0
17 ; CHECK-NEXT: subs r2, r1, #1
18 ; CHECK-NEXT: vadd.i32 q2, q2, r3
19 ; CHECK-NEXT: vidup.u32 q1, r4, #8
20 ; CHECK-NEXT: vmov r0, s8
21 ; CHECK-NEXT: vadd.i32 q1, q1, r0
22 ; CHECK-NEXT: vmov.i32 q0, #0x0
23 ; CHECK-NEXT: dlstp.32 lr, r2
24 ; CHECK-NEXT: .LBB0_1: @ =>This Inner Loop Header: Depth=1
25 ; CHECK-NEXT: vldrw.u32 q2, [q1, #32]!
26 ; CHECK-NEXT: vadd.f32 q0, q0, q2
27 ; CHECK-NEXT: letp lr, .LBB0_1
28 ; CHECK-NEXT: @ %bb.2:
29 ; CHECK-NEXT: bl vecAddAcrossF32Mve
30 ; CHECK-NEXT: vmov s0, r0
31 ; CHECK-NEXT: vcvt.f32.s32 s0, s0
32 ; CHECK-NEXT: vabs.f32 s0, s0
33 ; CHECK-NEXT: pop {r4, pc}
34 %3 = tail call { <4 x i32>, i32 } @llvm.arm.mve.vidup.v4i32(i32 0, i32 8)
35 %4 = extractvalue { <4 x i32>, i32 } %3, 0
36 %5 = add nsw i32 %1, -1
37 %6 = ptrtoint ptr %0 to i32
38 %7 = insertelement <4 x i32> undef, i32 %6, i32 0
39 %8 = add <4 x i32> %7, <i32 -32, i32 undef, i32 undef, i32 undef>
40 %9 = shufflevector <4 x i32> %8, <4 x i32> undef, <4 x i32> zeroinitializer
41 %10 = add <4 x i32> %4, %9
45 %12 = phi i32 [ %5, %2 ], [ %20, %11 ]
46 %13 = phi <4 x float> [ zeroinitializer, %2 ], [ %19, %11 ]
47 %14 = phi <4 x i32> [ %10, %2 ], [ %17, %11 ]
48 %15 = tail call <4 x i1> @llvm.arm.mve.vctp32(i32 %12)
49 %16 = tail call { <4 x float>, <4 x i32> } @llvm.arm.mve.vldr.gather.base.wb.predicated.v4f32.v4i32.v4i1(<4 x i32> %14, i32 32, <4 x i1> %15)
50 %17 = extractvalue { <4 x float>, <4 x i32> } %16, 1
51 %18 = extractvalue { <4 x float>, <4 x i32> } %16, 0
52 %19 = tail call <4 x float> @llvm.arm.mve.add.predicated.v4f32.v4i1(<4 x float> %13, <4 x float> %18, <4 x i1> %15, <4 x float> %13)
53 %20 = add nsw i32 %12, -4
54 %21 = icmp sgt i32 %12, 4
55 br i1 %21, label %11, label %22
58 %23 = tail call arm_aapcs_vfpcc i32 @vecAddAcrossF32Mve(<4 x float> %19)
59 %24 = sitofp i32 %23 to float
60 %25 = tail call float @llvm.fabs.f32(float %24)
64 declare { <4 x i32>, i32 } @llvm.arm.mve.vidup.v4i32(i32, i32)
65 declare <4 x i1> @llvm.arm.mve.vctp32(i32)
66 declare { <4 x float>, <4 x i32> } @llvm.arm.mve.vldr.gather.base.wb.predicated.v4f32.v4i32.v4i1(<4 x i32>, i32, <4 x i1>)
67 declare <4 x float> @llvm.arm.mve.add.predicated.v4f32.v4i1(<4 x float>, <4 x float>, <4 x i1>, <4 x float>)
68 declare arm_aapcs_vfpcc i32 @vecAddAcrossF32Mve(...)
69 declare float @llvm.fabs.f32(float)