1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mtriple=powerpc64-unknown-freebsd13.0 -verify-machineinstrs \
3; RUN:   -mcpu=ppc64 -ppc-asm-full-reg-names < %s | FileCheck %s
4@a = local_unnamed_addr global float* null, align 8
5
6; Function Attrs: nounwind
7define void @d() local_unnamed_addr #0 {
8; CHECK-LABEL: d:
9; CHECK:       # %bb.0: # %entry
10; CHECK-NEXT:    mflr r0
11; CHECK-NEXT:    std r0, 16(r1)
12; CHECK-NEXT:    stdu r1, -208(r1)
13; CHECK-NEXT:    addis r3, r2, .LC0@toc@ha
14; CHECK-NEXT:    std r29, 184(r1) # 8-byte Folded Spill
15; CHECK-NEXT:    ld r3, .LC0@toc@l(r3)
16; CHECK-NEXT:    std r30, 192(r1) # 8-byte Folded Spill
17; CHECK-NEXT:    ld r29, 0(r3)
18; CHECK-NEXT:    bl c
19; CHECK-NEXT:    nop
20; CHECK-NEXT:    mr r30, r3
21; CHECK-NEXT:    bl b
22; CHECK-NEXT:    nop
23; CHECK-NEXT:    cmpwi r30, 1
24; CHECK-NEXT:    blt cr0, .LBB0_9
25; CHECK-NEXT:  # %bb.1: # %for.body.preheader
26; CHECK-NEXT:    cmplwi r30, 4
27; CHECK-NEXT:    clrldi r4, r30, 32
28; CHECK-NEXT:    li r5, 0
29; CHECK-NEXT:    blt cr0, .LBB0_7
30; CHECK-NEXT:  # %bb.2: # %vector.memcheck
31; CHECK-NEXT:    rldic r6, r30, 2, 30
32; CHECK-NEXT:    add r7, r3, r6
33; CHECK-NEXT:    cmpld r29, r7
34; CHECK-NEXT:    add r6, r29, r6
35; CHECK-NEXT:    bc 4, lt, .LBB0_4
36; CHECK-NEXT:  # %bb.3: # %vector.memcheck
37; CHECK-NEXT:    cmpld r3, r6
38; CHECK-NEXT:    bc 12, lt, .LBB0_7
39; CHECK-NEXT:  .LBB0_4: # %vector.ph
40; CHECK-NEXT:    rlwinm r5, r4, 0, 0, 29
41; CHECK-NEXT:    li r7, 15
42; CHECK-NEXT:    addi r6, r5, -4
43; CHECK-NEXT:    addi r8, r1, 144
44; CHECK-NEXT:    rldicl r6, r6, 62, 2
45; CHECK-NEXT:    addi r9, r1, 128
46; CHECK-NEXT:    addi r6, r6, 1
47; CHECK-NEXT:    addi r10, r1, 160
48; CHECK-NEXT:    mtctr r6
49; CHECK-NEXT:    li r6, 0
50; CHECK-NEXT:    addi r11, r1, 112
51; CHECK-NEXT:  .LBB0_5: # %vector.body
52; CHECK-NEXT:    #
53; CHECK-NEXT:    add r12, r3, r6
54; CHECK-NEXT:    lvx v3, r3, r6
55; CHECK-NEXT:    lvx v5, r12, r7
56; CHECK-NEXT:    add r12, r29, r6
57; CHECK-NEXT:    lvsl v2, r3, r6
58; CHECK-NEXT:    vperm v2, v3, v5, v2
59; CHECK-NEXT:    lvx v3, r29, r6
60; CHECK-NEXT:    lvx v5, r12, r7
61; CHECK-NEXT:    lvsl v4, r29, r6
62; CHECK-NEXT:    stvx v2, 0, r8
63; CHECK-NEXT:    vperm v2, v3, v5, v4
64; CHECK-NEXT:    stvx v2, 0, r9
65; CHECK-NEXT:    lfs f0, 156(r1)
66; CHECK-NEXT:    lfs f1, 140(r1)
67; CHECK-NEXT:    fdivs f0, f1, f0
68; CHECK-NEXT:    lfs f1, 136(r1)
69; CHECK-NEXT:    stfs f0, 172(r1)
70; CHECK-NEXT:    lfs f0, 152(r1)
71; CHECK-NEXT:    fdivs f0, f1, f0
72; CHECK-NEXT:    lfs f1, 132(r1)
73; CHECK-NEXT:    stfs f0, 168(r1)
74; CHECK-NEXT:    lfs f0, 148(r1)
75; CHECK-NEXT:    fdivs f0, f1, f0
76; CHECK-NEXT:    lfs f1, 128(r1)
77; CHECK-NEXT:    stfs f0, 164(r1)
78; CHECK-NEXT:    lfs f0, 144(r1)
79; CHECK-NEXT:    fdivs f0, f1, f0
80; CHECK-NEXT:    stfs f0, 160(r1)
81; CHECK-NEXT:    lvx v2, 0, r10
82; CHECK-NEXT:    stvx v2, 0, r11
83; CHECK-NEXT:    ld r0, 112(r1)
84; CHECK-NEXT:    stdx r0, r29, r6
85; CHECK-NEXT:    addi r6, r6, 16
86; CHECK-NEXT:    ld r0, 120(r1)
87; CHECK-NEXT:    std r0, 8(r12)
88; CHECK-NEXT:    bdnz .LBB0_5
89; CHECK-NEXT:  # %bb.6: # %middle.block
90; CHECK-NEXT:    cmpld r5, r4
91; CHECK-NEXT:    beq cr0, .LBB0_9
92; CHECK-NEXT:  .LBB0_7: # %for.body.preheader18
93; CHECK-NEXT:    sldi r6, r5, 2
94; CHECK-NEXT:    sub r5, r4, r5
95; CHECK-NEXT:    addi r6, r6, -4
96; CHECK-NEXT:    add r3, r3, r6
97; CHECK-NEXT:    add r4, r29, r6
98; CHECK-NEXT:    mtctr r5
99; CHECK-NEXT:  .LBB0_8: # %for.body
100; CHECK-NEXT:    #
101; CHECK-NEXT:    lfsu f0, 4(r4)
102; CHECK-NEXT:    lfsu f1, 4(r3)
103; CHECK-NEXT:    fdivs f0, f0, f1
104; CHECK-NEXT:    stfs f0, 0(r4)
105; CHECK-NEXT:    bdnz .LBB0_8
106; CHECK-NEXT:  .LBB0_9: # %for.end
107; CHECK-NEXT:    ld r30, 192(r1) # 8-byte Folded Reload
108; CHECK-NEXT:    ld r29, 184(r1) # 8-byte Folded Reload
109; CHECK-NEXT:    addi r1, r1, 208
110; CHECK-NEXT:    ld r0, 16(r1)
111; CHECK-NEXT:    mtlr r0
112; CHECK-NEXT:    blr
113entry:
114  %0 = load float*, float** @a, align 8
115  %call = call signext i32 bitcast (i32 (...)* @c to i32 ()*)() #2
116  %call1 = call float* bitcast (float* (...)* @b to float* ()*)() #2
117  %cmp11 = icmp sgt i32 %call, 0
118  br i1 %cmp11, label %for.body.preheader, label %for.end
119
120for.body.preheader:                               ; preds = %entry
121  %wide.trip.count = zext i32 %call to i64
122  %min.iters.check = icmp ult i32 %call, 4
123  br i1 %min.iters.check, label %for.body.preheader18, label %vector.memcheck
124
125vector.memcheck:                                  ; preds = %for.body.preheader
126  %scevgep = getelementptr float, float* %0, i64 %wide.trip.count
127  %scevgep15 = getelementptr float, float* %call1, i64 %wide.trip.count
128  %bound0 = icmp ult float* %0, %scevgep15
129  %bound1 = icmp ult float* %call1, %scevgep
130  %found.conflict = and i1 %bound0, %bound1
131  br i1 %found.conflict, label %for.body.preheader18, label %vector.ph
132
133vector.ph:                                        ; preds = %vector.memcheck
134  %n.vec = and i64 %wide.trip.count, 4294967292
135  br label %vector.body
136
137vector.body:                                      ; preds = %vector.body, %vector.ph
138  %index = phi i64 [ 0, %vector.ph ], [ %index.next, %vector.body ]
139  %1 = getelementptr inbounds float, float* %call1, i64 %index
140  %2 = bitcast float* %1 to <4 x float>*
141  %wide.load = load <4 x float>, <4 x float>* %2, align 4
142  %3 = getelementptr inbounds float, float* %0, i64 %index
143  %4 = bitcast float* %3 to <4 x float>*
144  %wide.load17 = load <4 x float>, <4 x float>* %4, align 4
145  %5 = fdiv reassoc nsz arcp afn <4 x float> %wide.load17, %wide.load
146  %6 = bitcast float* %3 to <4 x float>*
147  store <4 x float> %5, <4 x float>* %6, align 4
148  %index.next = add i64 %index, 4
149  %7 = icmp eq i64 %index.next, %n.vec
150  br i1 %7, label %middle.block, label %vector.body
151
152middle.block:                                     ; preds = %vector.body
153  %cmp.n = icmp eq i64 %n.vec, %wide.trip.count
154  br i1 %cmp.n, label %for.end, label %for.body.preheader18
155
156for.body.preheader18:                             ; preds = %middle.block, %vector.memcheck, %for.body.preheader
157  %indvars.iv.ph = phi i64 [ 0, %vector.memcheck ], [ 0, %for.body.preheader ], [ %n.vec, %middle.block ]
158  br label %for.body
159
160for.body:                                         ; preds = %for.body.preheader18, %for.body
161  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ %indvars.iv.ph, %for.body.preheader18 ]
162  %arrayidx = getelementptr inbounds float, float* %call1, i64 %indvars.iv
163  %8 = load float, float* %arrayidx, align 4
164  %arrayidx3 = getelementptr inbounds float, float* %0, i64 %indvars.iv
165  %9 = load float, float* %arrayidx3, align 4
166  %div = fdiv reassoc nsz arcp afn float %9, %8
167  store float %div, float* %arrayidx3, align 4
168  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
169  %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count
170  br i1 %exitcond.not, label %for.end, label %for.body
171
172for.end:                                          ; preds = %for.body, %middle.block, %entry
173  ret void
174}
175
176declare signext i32 @c(...) local_unnamed_addr #1
177
178declare float* @b(...) local_unnamed_addr #1
179
180attributes #0 = { nounwind }
181