1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
2; RUN: llc -mcpu=pwr9 -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s
3
4; void foo(float *data, float d) {
5;   long i;
6;   for (i = 0; i < 8000; i++)
7;     data[i] = d;
8; }
9;
10; This loop will be unrolled by 96 and vectorized on power9.
11; icmp for loop iteration index and loop trip count(384) has LSRUse for 'reg({0,+,384})'.
12; Make sure above icmp does not impact LSR choose best formulae sets based on 'reg({(192 + %0),+,384})'
13
14define void @foo(float* nocapture %data, float %d) {
15; CHECK-LABEL: foo:
16; CHECK:       # %bb.0: # %entry
17; CHECK-NEXT:    xscvdpspn 0, 1
18; CHECK-NEXT:    li 5, 83
19; CHECK-NEXT:    addi 4, 3, 192
20; CHECK-NEXT:    mtctr 5
21; CHECK-NEXT:    xxspltw 0, 0, 0
22; CHECK-NEXT:    .p2align 4
23; CHECK-NEXT:  .LBB0_1: # %vector.body
24; CHECK-NEXT:    #
25; CHECK-NEXT:    stxv 0, -192(4)
26; CHECK-NEXT:    stxv 0, -176(4)
27; CHECK-NEXT:    stxv 0, -160(4)
28; CHECK-NEXT:    stxv 0, -144(4)
29; CHECK-NEXT:    stxv 0, -128(4)
30; CHECK-NEXT:    stxv 0, -112(4)
31; CHECK-NEXT:    stxv 0, -96(4)
32; CHECK-NEXT:    stxv 0, -80(4)
33; CHECK-NEXT:    stxv 0, -64(4)
34; CHECK-NEXT:    stxv 0, -48(4)
35; CHECK-NEXT:    stxv 0, -32(4)
36; CHECK-NEXT:    stxv 0, -16(4)
37; CHECK-NEXT:    stxv 0, 0(4)
38; CHECK-NEXT:    stxv 0, 16(4)
39; CHECK-NEXT:    stxv 0, 32(4)
40; CHECK-NEXT:    stxv 0, 48(4)
41; CHECK-NEXT:    stxv 0, 64(4)
42; CHECK-NEXT:    stxv 0, 80(4)
43; CHECK-NEXT:    stxv 0, 96(4)
44; CHECK-NEXT:    stxv 0, 112(4)
45; CHECK-NEXT:    stxv 0, 128(4)
46; CHECK-NEXT:    stxv 0, 144(4)
47; CHECK-NEXT:    stxv 0, 160(4)
48; CHECK-NEXT:    stxv 0, 176(4)
49; CHECK-NEXT:    addi 4, 4, 384
50; CHECK-NEXT:    bdnz .LBB0_1
51; CHECK-NEXT:  # %bb.2: # %for.body
52; CHECK-NEXT:    stfs 1, 31872(3)
53; CHECK-NEXT:    stfs 1, 31876(3)
54; CHECK-NEXT:    stfs 1, 31880(3)
55; CHECK-NEXT:    stfs 1, 31884(3)
56; CHECK-NEXT:    stfs 1, 31888(3)
57; CHECK-NEXT:    stfs 1, 31892(3)
58; CHECK-NEXT:    stfs 1, 31896(3)
59; CHECK-NEXT:    stfs 1, 31900(3)
60; CHECK-NEXT:    stfs 1, 31904(3)
61; CHECK-NEXT:    stfs 1, 31908(3)
62; CHECK-NEXT:    stfs 1, 31912(3)
63; CHECK-NEXT:    stfs 1, 31916(3)
64; CHECK-NEXT:    stfs 1, 31920(3)
65; CHECK-NEXT:    stfs 1, 31924(3)
66; CHECK-NEXT:    stfs 1, 31928(3)
67; CHECK-NEXT:    stfs 1, 31932(3)
68; CHECK-NEXT:    stfs 1, 31936(3)
69; CHECK-NEXT:    stfs 1, 31940(3)
70; CHECK-NEXT:    stfs 1, 31944(3)
71; CHECK-NEXT:    stfs 1, 31948(3)
72; CHECK-NEXT:    stfs 1, 31952(3)
73; CHECK-NEXT:    stfs 1, 31956(3)
74; CHECK-NEXT:    stfs 1, 31960(3)
75; CHECK-NEXT:    stfs 1, 31964(3)
76; CHECK-NEXT:    stfs 1, 31968(3)
77; CHECK-NEXT:    stfs 1, 31972(3)
78; CHECK-NEXT:    stfs 1, 31976(3)
79; CHECK-NEXT:    stfs 1, 31980(3)
80; CHECK-NEXT:    stfs 1, 31984(3)
81; CHECK-NEXT:    stfs 1, 31988(3)
82; CHECK-NEXT:    stfs 1, 31992(3)
83; CHECK-NEXT:    stfs 1, 31996(3)
84; CHECK-NEXT:    blr
85
86entry:
87  %broadcast.splatinsert16 = insertelement <4 x float> undef, float %d, i32 0
88  %broadcast.splat17 = shufflevector <4 x float> %broadcast.splatinsert16, <4 x float> undef, <4 x i32> zeroinitializer
89  %broadcast.splatinsert18 = insertelement <4 x float> undef, float %d, i32 0
90  %broadcast.splat19 = shufflevector <4 x float> %broadcast.splatinsert18, <4 x float> undef, <4 x i32> zeroinitializer
91  %broadcast.splatinsert20 = insertelement <4 x float> undef, float %d, i32 0
92  %broadcast.splat21 = shufflevector <4 x float> %broadcast.splatinsert20, <4 x float> undef, <4 x i32> zeroinitializer
93  %broadcast.splatinsert22 = insertelement <4 x float> undef, float %d, i32 0
94  %broadcast.splat23 = shufflevector <4 x float> %broadcast.splatinsert22, <4 x float> undef, <4 x i32> zeroinitializer
95  %broadcast.splatinsert24 = insertelement <4 x float> undef, float %d, i32 0
96  %broadcast.splat25 = shufflevector <4 x float> %broadcast.splatinsert24, <4 x float> undef, <4 x i32> zeroinitializer
97  %broadcast.splatinsert26 = insertelement <4 x float> undef, float %d, i32 0
98  %broadcast.splat27 = shufflevector <4 x float> %broadcast.splatinsert26, <4 x float> undef, <4 x i32> zeroinitializer
99  %broadcast.splatinsert28 = insertelement <4 x float> undef, float %d, i32 0
100  %broadcast.splat29 = shufflevector <4 x float> %broadcast.splatinsert28, <4 x float> undef, <4 x i32> zeroinitializer
101  %broadcast.splatinsert30 = insertelement <4 x float> undef, float %d, i32 0
102  %broadcast.splat31 = shufflevector <4 x float> %broadcast.splatinsert30, <4 x float> undef, <4 x i32> zeroinitializer
103  %broadcast.splatinsert32 = insertelement <4 x float> undef, float %d, i32 0
104  %broadcast.splat33 = shufflevector <4 x float> %broadcast.splatinsert32, <4 x float> undef, <4 x i32> zeroinitializer
105  %broadcast.splatinsert34 = insertelement <4 x float> undef, float %d, i32 0
106  %broadcast.splat35 = shufflevector <4 x float> %broadcast.splatinsert34, <4 x float> undef, <4 x i32> zeroinitializer
107  %broadcast.splatinsert36 = insertelement <4 x float> undef, float %d, i32 0
108  %broadcast.splat37 = shufflevector <4 x float> %broadcast.splatinsert36, <4 x float> undef, <4 x i32> zeroinitializer
109  %broadcast.splatinsert38 = insertelement <4 x float> undef, float %d, i32 0
110  %broadcast.splat39 = shufflevector <4 x float> %broadcast.splatinsert38, <4 x float> undef, <4 x i32> zeroinitializer
111  br label %vector.body
112
113vector.body:                                      ; preds = %vector.body, %entry
114  %index = phi i64 [ 0, %entry ], [ %index.next.1, %vector.body ]
115  %0 = getelementptr inbounds float, float* %data, i64 %index
116  %1 = bitcast float* %0 to <4 x float>*
117  store <4 x float> %broadcast.splat17, <4 x float>* %1, align 4
118  %2 = getelementptr inbounds float, float* %0, i64 4
119  %3 = bitcast float* %2 to <4 x float>*
120  store <4 x float> %broadcast.splat19, <4 x float>* %3, align 4
121  %4 = getelementptr inbounds float, float* %0, i64 8
122  %5 = bitcast float* %4 to <4 x float>*
123  store <4 x float> %broadcast.splat21, <4 x float>* %5, align 4
124  %6 = getelementptr inbounds float, float* %0, i64 12
125  %7 = bitcast float* %6 to <4 x float>*
126  store <4 x float> %broadcast.splat23, <4 x float>* %7, align 4
127  %8 = getelementptr inbounds float, float* %0, i64 16
128  %9 = bitcast float* %8 to <4 x float>*
129  store <4 x float> %broadcast.splat25, <4 x float>* %9, align 4
130  %10 = getelementptr inbounds float, float* %0, i64 20
131  %11 = bitcast float* %10 to <4 x float>*
132  store <4 x float> %broadcast.splat27, <4 x float>* %11, align 4
133  %12 = getelementptr inbounds float, float* %0, i64 24
134  %13 = bitcast float* %12 to <4 x float>*
135  store <4 x float> %broadcast.splat29, <4 x float>* %13, align 4
136  %14 = getelementptr inbounds float, float* %0, i64 28
137  %15 = bitcast float* %14 to <4 x float>*
138  store <4 x float> %broadcast.splat31, <4 x float>* %15, align 4
139  %16 = getelementptr inbounds float, float* %0, i64 32
140  %17 = bitcast float* %16 to <4 x float>*
141  store <4 x float> %broadcast.splat33, <4 x float>* %17, align 4
142  %18 = getelementptr inbounds float, float* %0, i64 36
143  %19 = bitcast float* %18 to <4 x float>*
144  store <4 x float> %broadcast.splat35, <4 x float>* %19, align 4
145  %20 = getelementptr inbounds float, float* %0, i64 40
146  %21 = bitcast float* %20 to <4 x float>*
147  store <4 x float> %broadcast.splat37, <4 x float>* %21, align 4
148  %22 = getelementptr inbounds float, float* %0, i64 44
149  %23 = bitcast float* %22 to <4 x float>*
150  store <4 x float> %broadcast.splat39, <4 x float>* %23, align 4
151  %index.next = add nuw nsw i64 %index, 48
152  %24 = getelementptr inbounds float, float* %data, i64 %index.next
153  %25 = bitcast float* %24 to <4 x float>*
154  store <4 x float> %broadcast.splat17, <4 x float>* %25, align 4
155  %26 = getelementptr inbounds float, float* %24, i64 4
156  %27 = bitcast float* %26 to <4 x float>*
157  store <4 x float> %broadcast.splat19, <4 x float>* %27, align 4
158  %28 = getelementptr inbounds float, float* %24, i64 8
159  %29 = bitcast float* %28 to <4 x float>*
160  store <4 x float> %broadcast.splat21, <4 x float>* %29, align 4
161  %30 = getelementptr inbounds float, float* %24, i64 12
162  %31 = bitcast float* %30 to <4 x float>*
163  store <4 x float> %broadcast.splat23, <4 x float>* %31, align 4
164  %32 = getelementptr inbounds float, float* %24, i64 16
165  %33 = bitcast float* %32 to <4 x float>*
166  store <4 x float> %broadcast.splat25, <4 x float>* %33, align 4
167  %34 = getelementptr inbounds float, float* %24, i64 20
168  %35 = bitcast float* %34 to <4 x float>*
169  store <4 x float> %broadcast.splat27, <4 x float>* %35, align 4
170  %36 = getelementptr inbounds float, float* %24, i64 24
171  %37 = bitcast float* %36 to <4 x float>*
172  store <4 x float> %broadcast.splat29, <4 x float>* %37, align 4
173  %38 = getelementptr inbounds float, float* %24, i64 28
174  %39 = bitcast float* %38 to <4 x float>*
175  store <4 x float> %broadcast.splat31, <4 x float>* %39, align 4
176  %40 = getelementptr inbounds float, float* %24, i64 32
177  %41 = bitcast float* %40 to <4 x float>*
178  store <4 x float> %broadcast.splat33, <4 x float>* %41, align 4
179  %42 = getelementptr inbounds float, float* %24, i64 36
180  %43 = bitcast float* %42 to <4 x float>*
181  store <4 x float> %broadcast.splat35, <4 x float>* %43, align 4
182  %44 = getelementptr inbounds float, float* %24, i64 40
183  %45 = bitcast float* %44 to <4 x float>*
184  store <4 x float> %broadcast.splat37, <4 x float>* %45, align 4
185  %46 = getelementptr inbounds float, float* %24, i64 44
186  %47 = bitcast float* %46 to <4 x float>*
187  store <4 x float> %broadcast.splat39, <4 x float>* %47, align 4
188  %index.next.1 = add nuw nsw i64 %index, 96
189  %48 = icmp eq i64 %index.next.1, 7968
190  br i1 %48, label %for.body, label %vector.body
191
192for.body:                                         ; preds = %vector.body
193  %arrayidx = getelementptr inbounds float, float* %data, i64 7968
194  store float %d, float* %arrayidx, align 4
195  %arrayidx.1 = getelementptr inbounds float, float* %data, i64 7969
196  store float %d, float* %arrayidx.1, align 4
197  %arrayidx.2 = getelementptr inbounds float, float* %data, i64 7970
198  store float %d, float* %arrayidx.2, align 4
199  %arrayidx.3 = getelementptr inbounds float, float* %data, i64 7971
200  store float %d, float* %arrayidx.3, align 4
201  %arrayidx.4 = getelementptr inbounds float, float* %data, i64 7972
202  store float %d, float* %arrayidx.4, align 4
203  %arrayidx.5 = getelementptr inbounds float, float* %data, i64 7973
204  store float %d, float* %arrayidx.5, align 4
205  %arrayidx.6 = getelementptr inbounds float, float* %data, i64 7974
206  store float %d, float* %arrayidx.6, align 4
207  %arrayidx.7 = getelementptr inbounds float, float* %data, i64 7975
208  store float %d, float* %arrayidx.7, align 4
209  %arrayidx.8 = getelementptr inbounds float, float* %data, i64 7976
210  store float %d, float* %arrayidx.8, align 4
211  %arrayidx.9 = getelementptr inbounds float, float* %data, i64 7977
212  store float %d, float* %arrayidx.9, align 4
213  %arrayidx.10 = getelementptr inbounds float, float* %data, i64 7978
214  store float %d, float* %arrayidx.10, align 4
215  %arrayidx.11 = getelementptr inbounds float, float* %data, i64 7979
216  store float %d, float* %arrayidx.11, align 4
217  %arrayidx.12 = getelementptr inbounds float, float* %data, i64 7980
218  store float %d, float* %arrayidx.12, align 4
219  %arrayidx.13 = getelementptr inbounds float, float* %data, i64 7981
220  store float %d, float* %arrayidx.13, align 4
221  %arrayidx.14 = getelementptr inbounds float, float* %data, i64 7982
222  store float %d, float* %arrayidx.14, align 4
223  %arrayidx.15 = getelementptr inbounds float, float* %data, i64 7983
224  store float %d, float* %arrayidx.15, align 4
225  %arrayidx.16 = getelementptr inbounds float, float* %data, i64 7984
226  store float %d, float* %arrayidx.16, align 4
227  %arrayidx.17 = getelementptr inbounds float, float* %data, i64 7985
228  store float %d, float* %arrayidx.17, align 4
229  %arrayidx.18 = getelementptr inbounds float, float* %data, i64 7986
230  store float %d, float* %arrayidx.18, align 4
231  %arrayidx.19 = getelementptr inbounds float, float* %data, i64 7987
232  store float %d, float* %arrayidx.19, align 4
233  %arrayidx.20 = getelementptr inbounds float, float* %data, i64 7988
234  store float %d, float* %arrayidx.20, align 4
235  %arrayidx.21 = getelementptr inbounds float, float* %data, i64 7989
236  store float %d, float* %arrayidx.21, align 4
237  %arrayidx.22 = getelementptr inbounds float, float* %data, i64 7990
238  store float %d, float* %arrayidx.22, align 4
239  %arrayidx.23 = getelementptr inbounds float, float* %data, i64 7991
240  store float %d, float* %arrayidx.23, align 4
241  %arrayidx.24 = getelementptr inbounds float, float* %data, i64 7992
242  store float %d, float* %arrayidx.24, align 4
243  %arrayidx.25 = getelementptr inbounds float, float* %data, i64 7993
244  store float %d, float* %arrayidx.25, align 4
245  %arrayidx.26 = getelementptr inbounds float, float* %data, i64 7994
246  store float %d, float* %arrayidx.26, align 4
247  %arrayidx.27 = getelementptr inbounds float, float* %data, i64 7995
248  store float %d, float* %arrayidx.27, align 4
249  %arrayidx.28 = getelementptr inbounds float, float* %data, i64 7996
250  store float %d, float* %arrayidx.28, align 4
251  %arrayidx.29 = getelementptr inbounds float, float* %data, i64 7997
252  store float %d, float* %arrayidx.29, align 4
253  %arrayidx.30 = getelementptr inbounds float, float* %data, i64 7998
254  store float %d, float* %arrayidx.30, align 4
255  %arrayidx.31 = getelementptr inbounds float, float* %data, i64 7999
256  store float %d, float* %arrayidx.31, align 4
257  ret void
258}
259