xref: /llvm-project/llvm/test/CodeGen/PowerPC/custom-stov.ll (revision 89bbcbe285eee1287a9ea21b92e4b3d307464d44)
1*89bbcbe2SRolandF77; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
2*89bbcbe2SRolandF77; RUN: llc -mcpu=ppc64 -mtriple=powerpc64-unknown-linux-gnu \
3*89bbcbe2SRolandF77; RUN:     -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
4*89bbcbe2SRolandF77; RUN:     < %s | FileCheck %s
5*89bbcbe2SRolandF77
6*89bbcbe2SRolandF77define void @_blah() {
7*89bbcbe2SRolandF77; CHECK-LABEL: _blah:
8*89bbcbe2SRolandF77; CHECK:       # %bb.0: # %entry
9*89bbcbe2SRolandF77; CHECK-NEXT:    li r3, 0
10*89bbcbe2SRolandF77; CHECK-NEXT:    li r4, 15
11*89bbcbe2SRolandF77; CHECK-NEXT:    lvx v3, 0, r4
12*89bbcbe2SRolandF77; CHECK-NEXT:    addi r5, r1, -64
13*89bbcbe2SRolandF77; CHECK-NEXT:    lvx v4, 0, r3
14*89bbcbe2SRolandF77; CHECK-NEXT:    lvsl v2, 0, r3
15*89bbcbe2SRolandF77; CHECK-NEXT:    vperm v2, v4, v3, v2
16*89bbcbe2SRolandF77; CHECK-NEXT:    lwz r4, 16(0)
17*89bbcbe2SRolandF77; CHECK-NEXT:    stvx v2, 0, r5
18*89bbcbe2SRolandF77; CHECK-NEXT:    lhz r5, -64(r1)
19*89bbcbe2SRolandF77; CHECK-NEXT:    lhz r6, -58(r1)
20*89bbcbe2SRolandF77; CHECK-NEXT:    lhz r7, -52(r1)
21*89bbcbe2SRolandF77; CHECK-NEXT:    sth r4, -34(r1)
22*89bbcbe2SRolandF77; CHECK-NEXT:    sth r3, -36(r1)
23*89bbcbe2SRolandF77; CHECK-NEXT:    sth r3, -40(r1)
24*89bbcbe2SRolandF77; CHECK-NEXT:    sth r3, -44(r1)
25*89bbcbe2SRolandF77; CHECK-NEXT:    sth r3, -48(r1)
26*89bbcbe2SRolandF77; CHECK-NEXT:    addi r3, r1, -48
27*89bbcbe2SRolandF77; CHECK-NEXT:    sth r7, -38(r1)
28*89bbcbe2SRolandF77; CHECK-NEXT:    sth r6, -42(r1)
29*89bbcbe2SRolandF77; CHECK-NEXT:    sth r5, -46(r1)
30*89bbcbe2SRolandF77; CHECK-NEXT:    lvx v2, 0, r3
31*89bbcbe2SRolandF77; CHECK-NEXT:    addi r3, r1, -32
32*89bbcbe2SRolandF77; CHECK-NEXT:    vsldoi v3, v2, v2, 8
33*89bbcbe2SRolandF77; CHECK-NEXT:    vmaxuw v2, v2, v3
34*89bbcbe2SRolandF77; CHECK-NEXT:    vspltw v3, v2, 1
35*89bbcbe2SRolandF77; CHECK-NEXT:    vmaxuw v2, v2, v3
36*89bbcbe2SRolandF77; CHECK-NEXT:    stvx v2, 0, r3
37*89bbcbe2SRolandF77; CHECK-NEXT:    lwz r3, -32(r1)
38*89bbcbe2SRolandF77; CHECK-NEXT:    cmplwi r3, 0
39*89bbcbe2SRolandF77; CHECK-NEXT:    blr
40*89bbcbe2SRolandF77entry:
41*89bbcbe2SRolandF77  %wide.vec904 = load <12 x i16>, ptr null, align 2
42*89bbcbe2SRolandF77  %strided.vec905 = shufflevector <12 x i16> %wide.vec904, <12 x i16> zeroinitializer, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
43*89bbcbe2SRolandF77  %0 = zext <4 x i16> %strided.vec905 to <4 x i32>
44*89bbcbe2SRolandF77  %1 = tail call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> %0)
45*89bbcbe2SRolandF77  %cmp55.not823 = icmp ugt i32 1, %1
46*89bbcbe2SRolandF77  br i1 %cmp55.not823, label %for.cond.cleanup56, label %for.body57.lr.ph
47*89bbcbe2SRolandF77
48*89bbcbe2SRolandF77for.body57.lr.ph:                                 ; preds = %entry
49*89bbcbe2SRolandF77  ret void
50*89bbcbe2SRolandF77
51*89bbcbe2SRolandF77for.cond.cleanup56:                               ; preds = %entry
52*89bbcbe2SRolandF77  ret void
53*89bbcbe2SRolandF77}
54*89bbcbe2SRolandF77
55*89bbcbe2SRolandF77declare i32 @llvm.vector.reduce.umax.v4i32(<4 x i32>)
56