1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mcpu=pwr8 -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
3; RUN:   -mtriple=powerpc64le-unknown-unknown < %s | FileCheck %s
4; RUN: llc -mcpu=pwr9 -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
5; RUN:   -mtriple=powerpc64le-unknown-unknown < %s | FileCheck %s \
6; RUN:   -check-prefix=CHECK-P9
7
8define void @julia__typed_vcat_20() #0 {
9; CHECK-LABEL: julia__typed_vcat_20:
10; CHECK:       # %bb.0: # %bb
11; CHECK-NEXT:    mflr r0
12; CHECK-NEXT:    std r30, -16(r1) # 8-byte Folded Spill
13; CHECK-NEXT:    std r0, 16(r1)
14; CHECK-NEXT:    stdu r1, -48(r1)
15; CHECK-NEXT:    li r3, 1
16; CHECK-NEXT:    li r30, 0
17; CHECK-NEXT:    .p2align 4
18; CHECK-NEXT:  .LBB0_1: # %bb3
19; CHECK-NEXT:    #
20; CHECK-NEXT:    addi r3, r3, -1
21; CHECK-NEXT:    mtfprd f0, r3
22; CHECK-NEXT:    xscvsxdsp f1, f0
23; CHECK-NEXT:    bl __gnu_f2h_ieee
24; CHECK-NEXT:    nop
25; CHECK-NEXT:    clrldi r3, r3, 48
26; CHECK-NEXT:    bl __gnu_h2f_ieee
27; CHECK-NEXT:    nop
28; CHECK-NEXT:    addi r30, r30, -1
29; CHECK-NEXT:    li r3, 0
30; CHECK-NEXT:    cmpldi r30, 0
31; CHECK-NEXT:    bne+ cr0, .LBB0_1
32; CHECK-NEXT:  # %bb.2: # %bb11
33; CHECK-NEXT:    bl __gnu_f2h_ieee
34; CHECK-NEXT:    nop
35; CHECK-NEXT:    sth r3, 0(r3)
36;
37; CHECK-P9-LABEL: julia__typed_vcat_20:
38; CHECK-P9:       # %bb.0: # %bb
39; CHECK-P9-NEXT:    li r3, 0
40; CHECK-P9-NEXT:    mtctr r3
41; CHECK-P9-NEXT:    li r3, 1
42; CHECK-P9-NEXT:    .p2align 4
43; CHECK-P9-NEXT:  .LBB0_1: # %bb3
44; CHECK-P9-NEXT:    #
45; CHECK-P9-NEXT:    addi r3, r3, -1
46; CHECK-P9-NEXT:    mtfprd f0, r3
47; CHECK-P9-NEXT:    xscvsxdsp f0, f0
48; CHECK-P9-NEXT:    xscvdphp f0, f0
49; CHECK-P9-NEXT:    mffprwz r3, f0
50; CHECK-P9-NEXT:    clrlwi r3, r3, 16
51; CHECK-P9-NEXT:    mtfprwz f0, r3
52; CHECK-P9-NEXT:    li r3, 0
53; CHECK-P9-NEXT:    xscvhpdp f0, f0
54; CHECK-P9-NEXT:    bdnz .LBB0_1
55; CHECK-P9-NEXT:  # %bb.2: # %bb11
56; CHECK-P9-NEXT:    xscvdphp f0, f0
57; CHECK-P9-NEXT:    stxsihx f0, 0, r3
58bb:
59  %i = load i64, i64 addrspace(11)* null, align 8
60  %i1 = call { i64, i1 } @llvm.ssub.with.overflow.i64(i64 %i, i64 0)
61  %i2 = extractvalue { i64, i1 } %i1, 0
62  br label %bb3
63
64bb3:                                              ; preds = %bb3, %bb
65  %i4 = phi i64 [ %i10, %bb3 ], [ 1, %bb ]
66  %i5 = phi i64 [ 0, %bb3 ], [ 1, %bb ]
67  %i6 = add nsw i64 %i5, -1
68  %i7 = add i64 %i6, 0
69  %i8 = sitofp i64 %i7 to half
70  store half %i8, half addrspace(13)* undef, align 2
71  %i9 = icmp eq i64 %i4, 0
72  %i10 = add i64 %i4, 1
73  br i1 %i9, label %bb11, label %bb3
74
75bb11:                                             ; preds = %bb3
76  unreachable
77}
78
79declare { i64, i1 } @llvm.ssub.with.overflow.i64(i64, i64) #0
80
81define void @julia__hypot_17() #0 {
82; CHECK-LABEL: julia__hypot_17:
83; CHECK:       # %bb.0: # %bb
84; CHECK-NEXT:    mflr r0
85; CHECK-NEXT:    std r30, -16(r1) # 8-byte Folded Spill
86; CHECK-NEXT:    std r0, 16(r1)
87; CHECK-NEXT:    stdu r1, -48(r1)
88; CHECK-NEXT:    li r30, 3
89; CHECK-NEXT:    .p2align 5
90; CHECK-NEXT:  .LBB1_1: # %bb1
91; CHECK-NEXT:    #
92; CHECK-NEXT:    addi r30, r30, -1
93; CHECK-NEXT:    cmpldi r30, 0
94; CHECK-NEXT:    beq cr0, .LBB1_3
95; CHECK-NEXT:  # %bb.2: # %bb3
96; CHECK-NEXT:    #
97; CHECK-NEXT:    lhz r3, 0(0)
98; CHECK-NEXT:    bl __gnu_h2f_ieee
99; CHECK-NEXT:    nop
100; CHECK-NEXT:    fcmpu cr0, f1, f1
101; CHECK-NEXT:    bun cr0, .LBB1_1
102; CHECK-NEXT:  .LBB1_3: # %bb9
103; CHECK-NEXT:    addi r1, r1, 48
104; CHECK-NEXT:    ld r0, 16(r1)
105; CHECK-NEXT:    ld r30, -16(r1) # 8-byte Folded Reload
106; CHECK-NEXT:    mtlr r0
107; CHECK-NEXT:    blr
108;
109; CHECK-P9-LABEL: julia__hypot_17:
110; CHECK-P9:       # %bb.0: # %bb
111; CHECK-P9-NEXT:    li r3, 3
112; CHECK-P9-NEXT:    mtctr r3
113; CHECK-P9-NEXT:    li r3, 0
114; CHECK-P9-NEXT:    .p2align 5
115; CHECK-P9-NEXT:  .LBB1_1: # %bb1
116; CHECK-P9-NEXT:    #
117; CHECK-P9-NEXT:    bdzlr
118; CHECK-P9-NEXT:  # %bb.2: # %bb3
119; CHECK-P9-NEXT:    #
120; CHECK-P9-NEXT:    lxsihzx f0, 0, r3
121; CHECK-P9-NEXT:    xscvhpdp f0, f0
122; CHECK-P9-NEXT:    fcmpu cr0, f0, f0
123; CHECK-P9-NEXT:    bun cr0, .LBB1_1
124; CHECK-P9-NEXT:  # %bb.3: # %bb9
125; CHECK-P9-NEXT:    blr
126bb:
127  br label %bb1
128
129bb1:                                              ; preds = %bb3, %bb
130  %i = phi i64 [ %i4, %bb3 ], [ 2, %bb ]
131  %i2 = icmp eq i64 %i, 4
132  br i1 %i2, label %bb9, label %bb3
133
134bb3:                                              ; preds = %bb1
135  %i4 = add nuw nsw i64 %i, 1
136  %i5 = load half, half* null, align 2
137  %i6 = fpext half %i5 to float
138  %i7 = fcmp uno float %i6, 0.000000e+00
139  %i8 = or i1 %i7, false
140  br i1 %i8, label %bb1, label %bb9
141
142bb9:                                              ; preds = %bb3, %bb1
143  ret void
144}
145
146define void @func_48786() #0 {
147; CHECK-LABEL: func_48786:
148; CHECK:       # %bb.0: # %bb
149; CHECK-NEXT:    mfocrf r12, 32
150; CHECK-NEXT:    mflr r0
151; CHECK-NEXT:    std r0, 16(r1)
152; CHECK-NEXT:    stw r12, 8(r1)
153; CHECK-NEXT:    stdu r1, -48(r1)
154; CHECK-NEXT:    ld r3, 0(r3)
155; CHECK-NEXT:    std r30, 32(r1) # 8-byte Folded Spill
156; CHECK-NEXT:    # implicit-def: $x30
157; CHECK-NEXT:    cmpdi r3, 0
158; CHECK-NEXT:    crnot 4*cr2+lt, eq
159; CHECK-NEXT:    bc 12, 4*cr5+lt, .LBB2_3
160; CHECK-NEXT:    .p2align 4
161; CHECK-NEXT:  .LBB2_1: # %bb4
162; CHECK-NEXT:    lhz r3, 0(r3)
163; CHECK-NEXT:    bl __gnu_h2f_ieee
164; CHECK-NEXT:    nop
165; CHECK-NEXT:    bc 4, 4*cr2+lt, .LBB2_6
166; CHECK-NEXT:  # %bb.2: # %bb8
167; CHECK-NEXT:    bl __gnu_f2h_ieee
168; CHECK-NEXT:    nop
169; CHECK-NEXT:    sth r3, 0(0)
170; CHECK-NEXT:  .LBB2_3: # %bb10
171; CHECK-NEXT:    #
172; CHECK-NEXT:    cmpldi r30, 0
173; CHECK-NEXT:    beq cr0, .LBB2_5
174; CHECK-NEXT:  # %bb.4: # %bb12
175; CHECK-NEXT:    #
176; CHECK-NEXT:    addi r30, r30, 1
177; CHECK-NEXT:    bc 4, 4*cr5+lt, .LBB2_1
178; CHECK-NEXT:    b .LBB2_3
179; CHECK-NEXT:  .LBB2_5: # %bb14
180; CHECK-NEXT:    ld r30, 32(r1) # 8-byte Folded Reload
181; CHECK-NEXT:    addi r1, r1, 48
182; CHECK-NEXT:    ld r0, 16(r1)
183; CHECK-NEXT:    lwz r12, 8(r1)
184; CHECK-NEXT:    mtocrf 32, r12
185; CHECK-NEXT:    mtlr r0
186; CHECK-NEXT:    blr
187; CHECK-NEXT:  .LBB2_6: # %bb15
188;
189; CHECK-P9-LABEL: func_48786:
190; CHECK-P9:       # %bb.0: # %bb
191; CHECK-P9-NEXT:    ld r3, 0(r3)
192; CHECK-P9-NEXT:    cmpdi r3, 0
193; CHECK-P9-NEXT:    mtctr r3
194; CHECK-P9-NEXT:    li r3, 0
195; CHECK-P9-NEXT:    crnot 4*cr5+lt, eq
196; CHECK-P9-NEXT:    b .LBB2_2
197; CHECK-P9-NEXT:    .p2align 5
198; CHECK-P9-NEXT:  .LBB2_1: # %bb10
199; CHECK-P9-NEXT:    #
200; CHECK-P9-NEXT:    bdzlr
201; CHECK-P9-NEXT:  .LBB2_2: # %bb2
202; CHECK-P9-NEXT:    #
203; CHECK-P9-NEXT:    bc 12, 4*cr5+lt, .LBB2_1
204; CHECK-P9-NEXT:  # %bb.3: # %bb4
205; CHECK-P9-NEXT:    #
206; CHECK-P9-NEXT:    lxsihzx f0, 0, r3
207; CHECK-P9-NEXT:    xscvhpdp f0, f0
208; CHECK-P9-NEXT:    bc 4, 4*cr5+lt, .LBB2_5
209; CHECK-P9-NEXT:  # %bb.4: # %bb8
210; CHECK-P9-NEXT:    #
211; CHECK-P9-NEXT:    xscvdphp f0, f0
212; CHECK-P9-NEXT:    stxsihx f0, 0, r3
213; CHECK-P9-NEXT:    b .LBB2_1
214; CHECK-P9-NEXT:  .LBB2_5: # %bb15
215bb:
216  %i = load i64, i64 addrspace(11)* undef, align 8
217  %i1 = load i64, i64 addrspace(11)* undef, align 8
218  br label %bb2
219
220bb2:                                              ; preds = %bb12, %bb
221  %i3 = phi i64 [ undef, %bb ], [ %i13, %bb12 ]
222  br i1 undef, label %bb10, label %bb4
223
224bb4:                                              ; preds = %bb2
225  switch i32 undef, label %bb9 [
226    i32 1426063360, label %bb5
227    i32 1275068416, label %bb5
228  ]
229
230bb5:                                              ; preds = %bb4, %bb4
231  %i6 = load half, half addrspace(13)* undef, align 2
232  %i7 = icmp ult i64 0, %i1
233  br i1 %i7, label %bb8, label %bb15
234
235bb8:                                              ; preds = %bb5
236  store half %i6, half addrspace(13)* null, align 2
237  br label %bb10
238
239bb9:                                              ; preds = %bb4
240  unreachable
241
242bb10:                                             ; preds = %bb8, %bb2
243  %i11 = icmp eq i64 %i3, 0
244  br i1 %i11, label %bb14, label %bb12
245
246bb12:                                             ; preds = %bb10
247  %i13 = add i64 %i3, 1
248  br label %bb2
249
250bb14:                                             ; preds = %bb10
251  ret void
252
253bb15:                                             ; preds = %bb5
254  unreachable
255}
256
257define void @func_48785(half %arg) #0 {
258; CHECK-LABEL: func_48785:
259; CHECK:       # %bb.0: # %bb
260; CHECK-NEXT:    mflr r0
261; CHECK-NEXT:    std r29, -32(r1) # 8-byte Folded Spill
262; CHECK-NEXT:    std r30, -24(r1) # 8-byte Folded Spill
263; CHECK-NEXT:    stfd f31, -8(r1) # 8-byte Folded Spill
264; CHECK-NEXT:    std r0, 16(r1)
265; CHECK-NEXT:    stdu r1, -64(r1)
266; CHECK-NEXT:    fmr f31, f1
267; CHECK-NEXT:    li r30, 0
268; CHECK-NEXT:    li r29, 0
269; CHECK-NEXT:    .p2align 5
270; CHECK-NEXT:  .LBB3_1: # %bb1
271; CHECK-NEXT:    #
272; CHECK-NEXT:    fmr f1, f31
273; CHECK-NEXT:    bl __gnu_f2h_ieee
274; CHECK-NEXT:    nop
275; CHECK-NEXT:    addi r29, r29, -12
276; CHECK-NEXT:    sth r3, 0(r30)
277; CHECK-NEXT:    addi r30, r30, 24
278; CHECK-NEXT:    cmpldi r29, 0
279; CHECK-NEXT:    bne+ cr0, .LBB3_1
280; CHECK-NEXT:  # %bb.2: # %bb5
281;
282; CHECK-P9-LABEL: func_48785:
283; CHECK-P9:       # %bb.0: # %bb
284; CHECK-P9-NEXT:    li r3, 1
285; CHECK-P9-NEXT:    rldic r3, r3, 62, 1
286; CHECK-P9-NEXT:    mtctr r3
287; CHECK-P9-NEXT:    li r3, 0
288; CHECK-P9-NEXT:    .p2align 4
289; CHECK-P9-NEXT:  .LBB3_1: # %bb1
290; CHECK-P9-NEXT:    #
291; CHECK-P9-NEXT:    xscvdphp f0, f1
292; CHECK-P9-NEXT:    stxsihx f0, 0, r3
293; CHECK-P9-NEXT:    addi r3, r3, 24
294; CHECK-P9-NEXT:    bdnz .LBB3_1
295; CHECK-P9-NEXT:  # %bb.2: # %bb5
296bb:
297  br label %bb1
298
299bb1:                                              ; preds = %bb1, %bb
300  %i = phi i64 [ 0, %bb ], [ %i3, %bb1 ]
301  %i2 = getelementptr inbounds half, half addrspace(13)* null, i64 %i
302  store half %arg, half addrspace(13)* %i2, align 2
303  %i3 = add i64 %i, 12
304  %i4 = icmp eq i64 %i3, 0
305  br i1 %i4, label %bb5, label %bb1
306
307bb5:                                              ; preds = %bb1
308  unreachable
309}
310attributes #0 = { nounwind }
311