xref: /llvm-project/llvm/test/CodeGen/PowerPC/pr52894.ll (revision b922a3621116b404d868af8b74cab25ab78555be)
1; RUN: llc -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
2; RUN:   -mcpu=pwr8 -mtriple=powerpc64le < %s | FileCheck %s
3; RUN: llc -verify-machineinstrs -ppc-asm-full-reg-names -ppc-vsr-nums-as-vr \
4; RUN:   -mcpu=pwr8 -mtriple=powerpc64 < %s | FileCheck %s
5%struct.d = type { [131072 x i32] }
6
7@a = dso_local local_unnamed_addr global [4096 x i32] zeroinitializer, align 4
8
9; Function Attrs: mustprogress uwtable
10define dso_local void @_Z1g1dILi17EE(ptr nocapture noundef readnone byval(%struct.d) align 8 %0) local_unnamed_addr #0 {
11; CHECK-LABEL: _Z1g1dILi17EE:
12; CHECK-NOT:    mtfprd f0, r4
13; CHECK-NOT:    stdx r3, r1, r4
14; CHECK-NOT:    mffprd r4, f0
15; CHECK-NOT:    mtfprd f0, r4
16; CHECK-NOT:    ldx r3, r1, r4
17; CHECK-NOT:    mffprd r4, f0
18; CHECK-NOT:    mtfprd f0, r4
19; CHECK-NOT:    stdx r3, r1, r4
20; CHECK-NOT:    mffprd r4, f0
21; CHECK-NOT:    mtfprd f0, r4
22; CHECK-NOT:    ldx r3, r1, r4
23; CHECK-NOT:    mffprd r4, f0
24entry:
25  %c = alloca %struct.d, align 8
26  call void @llvm.lifetime.start.p0(i64 524288, ptr nonnull %c) #3
27  br label %vector.body
28
29vector.body:                                      ; preds = %vector.body.1, %entry
30  %index = phi i64 [ 0, %entry ], [ %index.next.1, %vector.body.1 ]
31  %vec.ind = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %entry ], [ %vec.ind.next.1, %vector.body.1 ]
32  %step.add = add <4 x i32> %vec.ind, <i32 4, i32 4, i32 4, i32 4>
33  %step.add24 = add <4 x i32> %vec.ind, <i32 8, i32 8, i32 8, i32 8>
34  %step.add25 = add <4 x i32> %vec.ind, <i32 12, i32 12, i32 12, i32 12>
35  %step.add26 = add <4 x i32> %vec.ind, <i32 16, i32 16, i32 16, i32 16>
36  %step.add27 = add <4 x i32> %vec.ind, <i32 20, i32 20, i32 20, i32 20>
37  %step.add28 = add <4 x i32> %vec.ind, <i32 24, i32 24, i32 24, i32 24>
38  %step.add29 = add <4 x i32> %vec.ind, <i32 28, i32 28, i32 28, i32 28>
39  %step.add30 = add <4 x i32> %vec.ind, <i32 32, i32 32, i32 32, i32 32>
40  %step.add31 = add <4 x i32> %vec.ind, <i32 36, i32 36, i32 36, i32 36>
41  %step.add32 = add <4 x i32> %vec.ind, <i32 40, i32 40, i32 40, i32 40>
42  %step.add33 = add <4 x i32> %vec.ind, <i32 44, i32 44, i32 44, i32 44>
43  %1 = getelementptr inbounds [4096 x i32], ptr @a, i64 0, i64 %index
44  store <4 x i32> %vec.ind, ptr %1, align 4
45  %2 = getelementptr inbounds i32, ptr %1, i64 4
46  store <4 x i32> %step.add, ptr %2, align 4
47  %3 = getelementptr inbounds i32, ptr %1, i64 8
48  store <4 x i32> %step.add24, ptr %3, align 4
49  %4 = getelementptr inbounds i32, ptr %1, i64 12
50  store <4 x i32> %step.add25, ptr %4, align 4
51  %5 = getelementptr inbounds i32, ptr %1, i64 16
52  store <4 x i32> %step.add26, ptr %5, align 4
53  %6 = getelementptr inbounds i32, ptr %1, i64 20
54  store <4 x i32> %step.add27, ptr %6, align 4
55  %7 = getelementptr inbounds i32, ptr %1, i64 24
56  store <4 x i32> %step.add28, ptr %7, align 4
57  %8 = getelementptr inbounds i32, ptr %1, i64 28
58  store <4 x i32> %step.add29, ptr %8, align 4
59  %9 = getelementptr inbounds i32, ptr %1, i64 32
60  store <4 x i32> %step.add30, ptr %9, align 4
61  %10 = getelementptr inbounds i32, ptr %1, i64 36
62  store <4 x i32> %step.add31, ptr %10, align 4
63  %11 = getelementptr inbounds i32, ptr %1, i64 40
64  store <4 x i32> %step.add32, ptr %11, align 4
65  %12 = getelementptr inbounds i32, ptr %1, i64 44
66  store <4 x i32> %step.add33, ptr %12, align 4
67  %index.next = add nuw nsw i64 %index, 48
68  %13 = icmp eq i64 %index.next, 4080
69  br i1 %13, label %for.body, label %vector.body.1
70
71vector.body.1:                                    ; preds = %vector.body
72  %vec.ind.next = add <4 x i32> %vec.ind, <i32 48, i32 48, i32 48, i32 48>
73  %step.add.1 = add <4 x i32> %vec.ind, <i32 52, i32 52, i32 52, i32 52>
74  %step.add24.1 = add <4 x i32> %vec.ind, <i32 56, i32 56, i32 56, i32 56>
75  %step.add25.1 = add <4 x i32> %vec.ind, <i32 60, i32 60, i32 60, i32 60>
76  %step.add26.1 = add <4 x i32> %vec.ind, <i32 64, i32 64, i32 64, i32 64>
77  %step.add27.1 = add <4 x i32> %vec.ind, <i32 68, i32 68, i32 68, i32 68>
78  %step.add28.1 = add <4 x i32> %vec.ind, <i32 72, i32 72, i32 72, i32 72>
79  %step.add29.1 = add <4 x i32> %vec.ind, <i32 76, i32 76, i32 76, i32 76>
80  %step.add30.1 = add <4 x i32> %vec.ind, <i32 80, i32 80, i32 80, i32 80>
81  %step.add31.1 = add <4 x i32> %vec.ind, <i32 84, i32 84, i32 84, i32 84>
82  %step.add32.1 = add <4 x i32> %vec.ind, <i32 88, i32 88, i32 88, i32 88>
83  %step.add33.1 = add <4 x i32> %vec.ind, <i32 92, i32 92, i32 92, i32 92>
84  %14 = getelementptr inbounds [4096 x i32], ptr @a, i64 0, i64 %index.next
85  store <4 x i32> %vec.ind.next, ptr %14, align 4
86  %15 = getelementptr inbounds i32, ptr %14, i64 4
87  store <4 x i32> %step.add.1, ptr %15, align 4
88  %16 = getelementptr inbounds i32, ptr %14, i64 8
89  store <4 x i32> %step.add24.1, ptr %16, align 4
90  %17 = getelementptr inbounds i32, ptr %14, i64 12
91  store <4 x i32> %step.add25.1, ptr %17, align 4
92  %18 = getelementptr inbounds i32, ptr %14, i64 16
93  store <4 x i32> %step.add26.1, ptr %18, align 4
94  %19 = getelementptr inbounds i32, ptr %14, i64 20
95  store <4 x i32> %step.add27.1, ptr %19, align 4
96  %20 = getelementptr inbounds i32, ptr %14, i64 24
97  store <4 x i32> %step.add28.1, ptr %20, align 4
98  %21 = getelementptr inbounds i32, ptr %14, i64 28
99  store <4 x i32> %step.add29.1, ptr %21, align 4
100  %22 = getelementptr inbounds i32, ptr %14, i64 32
101  store <4 x i32> %step.add30.1, ptr %22, align 4
102  %23 = getelementptr inbounds i32, ptr %14, i64 36
103  store <4 x i32> %step.add31.1, ptr %23, align 4
104  %24 = getelementptr inbounds i32, ptr %14, i64 40
105  store <4 x i32> %step.add32.1, ptr %24, align 4
106  %25 = getelementptr inbounds i32, ptr %14, i64 44
107  store <4 x i32> %step.add33.1, ptr %25, align 4
108  %index.next.1 = add nuw nsw i64 %index, 96
109  %vec.ind.next.1 = add <4 x i32> %vec.ind, <i32 96, i32 96, i32 96, i32 96>
110  br label %vector.body
111
112vector.body40:                                    ; preds = %vector.body40.1, %for.body
113  %index41 = phi i64 [ 0, %for.body ], [ %index.next56.1, %vector.body40.1 ]
114  %vec.ind42 = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %for.body ], [ %vec.ind.next55.1, %vector.body40.1 ]
115  %step.add43 = add <4 x i32> %vec.ind42, <i32 4, i32 4, i32 4, i32 4>
116  %step.add44 = add <4 x i32> %vec.ind42, <i32 8, i32 8, i32 8, i32 8>
117  %step.add45 = add <4 x i32> %vec.ind42, <i32 12, i32 12, i32 12, i32 12>
118  %step.add46 = add <4 x i32> %vec.ind42, <i32 16, i32 16, i32 16, i32 16>
119  %step.add47 = add <4 x i32> %vec.ind42, <i32 20, i32 20, i32 20, i32 20>
120  %step.add48 = add <4 x i32> %vec.ind42, <i32 24, i32 24, i32 24, i32 24>
121  %step.add49 = add <4 x i32> %vec.ind42, <i32 28, i32 28, i32 28, i32 28>
122  %step.add50 = add <4 x i32> %vec.ind42, <i32 32, i32 32, i32 32, i32 32>
123  %step.add51 = add <4 x i32> %vec.ind42, <i32 36, i32 36, i32 36, i32 36>
124  %step.add52 = add <4 x i32> %vec.ind42, <i32 40, i32 40, i32 40, i32 40>
125  %step.add53 = add <4 x i32> %vec.ind42, <i32 44, i32 44, i32 44, i32 44>
126  %26 = getelementptr inbounds [4096 x i32], ptr @a, i64 0, i64 %index41
127  store <4 x i32> %vec.ind42, ptr %26, align 4
128  %27 = getelementptr inbounds i32, ptr %26, i64 4
129  store <4 x i32> %step.add43, ptr %27, align 4
130  %28 = getelementptr inbounds i32, ptr %26, i64 8
131  store <4 x i32> %step.add44, ptr %28, align 4
132  %29 = getelementptr inbounds i32, ptr %26, i64 12
133  store <4 x i32> %step.add45, ptr %29, align 4
134  %30 = getelementptr inbounds i32, ptr %26, i64 16
135  store <4 x i32> %step.add46, ptr %30, align 4
136  %31 = getelementptr inbounds i32, ptr %26, i64 20
137  store <4 x i32> %step.add47, ptr %31, align 4
138  %32 = getelementptr inbounds i32, ptr %26, i64 24
139  store <4 x i32> %step.add48, ptr %32, align 4
140  %33 = getelementptr inbounds i32, ptr %26, i64 28
141  store <4 x i32> %step.add49, ptr %33, align 4
142  %34 = getelementptr inbounds i32, ptr %26, i64 32
143  store <4 x i32> %step.add50, ptr %34, align 4
144  %35 = getelementptr inbounds i32, ptr %26, i64 36
145  store <4 x i32> %step.add51, ptr %35, align 4
146  %36 = getelementptr inbounds i32, ptr %26, i64 40
147  store <4 x i32> %step.add52, ptr %36, align 4
148  %37 = getelementptr inbounds i32, ptr %26, i64 44
149  store <4 x i32> %step.add53, ptr %37, align 4
150  %index.next56 = add nuw nsw i64 %index41, 48
151  %38 = icmp eq i64 %index.next56, 4080
152  br i1 %38, label %for.body5, label %vector.body40.1
153
154vector.body40.1:                                  ; preds = %vector.body40
155  %vec.ind.next55 = add <4 x i32> %vec.ind42, <i32 48, i32 48, i32 48, i32 48>
156  %step.add43.1 = add <4 x i32> %vec.ind42, <i32 52, i32 52, i32 52, i32 52>
157  %step.add44.1 = add <4 x i32> %vec.ind42, <i32 56, i32 56, i32 56, i32 56>
158  %step.add45.1 = add <4 x i32> %vec.ind42, <i32 60, i32 60, i32 60, i32 60>
159  %step.add46.1 = add <4 x i32> %vec.ind42, <i32 64, i32 64, i32 64, i32 64>
160  %step.add47.1 = add <4 x i32> %vec.ind42, <i32 68, i32 68, i32 68, i32 68>
161  %step.add48.1 = add <4 x i32> %vec.ind42, <i32 72, i32 72, i32 72, i32 72>
162  %step.add49.1 = add <4 x i32> %vec.ind42, <i32 76, i32 76, i32 76, i32 76>
163  %step.add50.1 = add <4 x i32> %vec.ind42, <i32 80, i32 80, i32 80, i32 80>
164  %step.add51.1 = add <4 x i32> %vec.ind42, <i32 84, i32 84, i32 84, i32 84>
165  %step.add52.1 = add <4 x i32> %vec.ind42, <i32 88, i32 88, i32 88, i32 88>
166  %step.add53.1 = add <4 x i32> %vec.ind42, <i32 92, i32 92, i32 92, i32 92>
167  %39 = getelementptr inbounds [4096 x i32], ptr @a, i64 0, i64 %index.next56
168  store <4 x i32> %vec.ind.next55, ptr %39, align 4
169  %40 = getelementptr inbounds i32, ptr %39, i64 4
170  store <4 x i32> %step.add43.1, ptr %40, align 4
171  %41 = getelementptr inbounds i32, ptr %39, i64 8
172  store <4 x i32> %step.add44.1, ptr %41, align 4
173  %42 = getelementptr inbounds i32, ptr %39, i64 12
174  store <4 x i32> %step.add45.1, ptr %42, align 4
175  %43 = getelementptr inbounds i32, ptr %39, i64 16
176  store <4 x i32> %step.add46.1, ptr %43, align 4
177  %44 = getelementptr inbounds i32, ptr %39, i64 20
178  store <4 x i32> %step.add47.1, ptr %44, align 4
179  %45 = getelementptr inbounds i32, ptr %39, i64 24
180  store <4 x i32> %step.add48.1, ptr %45, align 4
181  %46 = getelementptr inbounds i32, ptr %39, i64 28
182  store <4 x i32> %step.add49.1, ptr %46, align 4
183  %47 = getelementptr inbounds i32, ptr %39, i64 32
184  store <4 x i32> %step.add50.1, ptr %47, align 4
185  %48 = getelementptr inbounds i32, ptr %39, i64 36
186  store <4 x i32> %step.add51.1, ptr %48, align 4
187  %49 = getelementptr inbounds i32, ptr %39, i64 40
188  store <4 x i32> %step.add52.1, ptr %49, align 4
189  %50 = getelementptr inbounds i32, ptr %39, i64 44
190  store <4 x i32> %step.add53.1, ptr %50, align 4
191  %index.next56.1 = add nuw nsw i64 %index41, 96
192  %vec.ind.next55.1 = add <4 x i32> %vec.ind42, <i32 96, i32 96, i32 96, i32 96>
193  br label %vector.body40
194
195for.body:                                         ; preds = %vector.body
196  store i32 4080, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4080), align 4
197  store i32 4081, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4081), align 4
198  store i32 4082, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4082), align 4
199  store i32 4083, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4083), align 4
200  store i32 4084, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4084), align 4
201  store i32 4085, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4085), align 4
202  store i32 4086, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4086), align 4
203  store i32 4087, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4087), align 4
204  store i32 4088, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4088), align 4
205  store i32 4089, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4089), align 4
206  store i32 4090, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4090), align 4
207  store i32 4091, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4091), align 4
208  store i32 4092, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4092), align 4
209  store i32 4093, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4093), align 4
210  store i32 4094, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4094), align 4
211  store i32 4095, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4095), align 4
212  call void @_ZN1dILi17EE1eEv(ptr noundef nonnull align 4 dereferenceable(524288) %c)
213  br label %vector.body40
214
215for.body5:                                        ; preds = %vector.body40
216  store i32 4080, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4080), align 4
217  store i32 4081, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4081), align 4
218  store i32 4082, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4082), align 4
219  store i32 4083, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4083), align 4
220  store i32 4084, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4084), align 4
221  store i32 4085, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4085), align 4
222  store i32 4086, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4086), align 4
223  store i32 4087, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4087), align 4
224  store i32 4088, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4088), align 4
225  store i32 4089, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4089), align 4
226  store i32 4090, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4090), align 4
227  store i32 4091, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4091), align 4
228  store i32 4092, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4092), align 4
229  store i32 4093, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4093), align 4
230  store i32 4094, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4094), align 4
231  store i32 4095, ptr getelementptr inbounds ([4096 x i32], ptr @a, i64 0, i64 4095), align 4
232  call void @_Z1h1dILi17EE(ptr noundef nonnull byval(%struct.d) align 8 %c)
233  call void @llvm.lifetime.end.p0(i64 524288, ptr nonnull %c) #3
234  ret void
235}
236
237; Function Attrs: argmemonly mustprogress nocallback nofree nosync nounwind willreturn
238declare void @llvm.lifetime.start.p0(i64 immarg, ptr nocapture) #1
239
240; Function Attrs: argmemonly mustprogress nocallback nofree nosync nounwind willreturn
241declare void @llvm.lifetime.end.p0(i64 immarg, ptr nocapture) #1
242
243declare void @_ZN1dILi17EE1eEv(ptr noundef nonnull align 4 dereferenceable(524288)) local_unnamed_addr #2
244
245declare void @_Z1h1dILi17EE(ptr noundef byval(%struct.d) align 8) local_unnamed_addr #2
246attributes #0 = { nounwind }
247