1 ; RUN: llc < %s -mtriple=thumbv7-none-linux-gnueabi -mcpu=cortex-a9 -mattr=+neon,+neonfp | FileCheck %s
2 ; PR14824. The test is presented by Jiangning Liu. If the ld/st optimization algorithm is changed, this test case may fail.
3 ; Also if the machine code for ld/st optimizor is changed, this test case may fail. If so, remove this test.
5 define void @sample_test(ptr %secondSource, ptr %source, ptr %dest) nounwind {
12 %s0 = load <8 x i64> , ptr %source, align 64
13 %arrayidx64 = getelementptr inbounds <8 x i64>, ptr %source, i32 6
14 %s120 = load <8 x i64> , ptr %arrayidx64, align 64
15 %s122 = bitcast <8 x i64> %s120 to i512
16 %data.i.i677.48.extract.shift = lshr i512 %s122, 384
17 %data.i.i677.48.extract.trunc = trunc i512 %data.i.i677.48.extract.shift to i64
18 %s123 = insertelement <8 x i64> undef, i64 %data.i.i677.48.extract.trunc, i32 0
19 %data.i.i677.32.extract.shift = lshr i512 %s122, 256
20 %data.i.i677.32.extract.trunc = trunc i512 %data.i.i677.32.extract.shift to i64
21 %s124 = insertelement <8 x i64> %s123, i64 %data.i.i677.32.extract.trunc, i32 1
22 %data.i.i677.16.extract.shift = lshr i512 %s122, 128
23 %data.i.i677.16.extract.trunc = trunc i512 %data.i.i677.16.extract.shift to i64
24 %s125 = insertelement <8 x i64> %s124, i64 %data.i.i677.16.extract.trunc, i32 2
25 %data.i.i677.56.extract.shift = lshr i512 %s122, 448
26 %data.i.i677.56.extract.trunc = trunc i512 %data.i.i677.56.extract.shift to i64
27 %s126 = insertelement <8 x i64> %s125, i64 %data.i.i677.56.extract.trunc, i32 3
28 %data.i.i677.24.extract.shift = lshr i512 %s122, 192
29 %data.i.i677.24.extract.trunc = trunc i512 %data.i.i677.24.extract.shift to i64
30 %s127 = insertelement <8 x i64> %s126, i64 %data.i.i677.24.extract.trunc, i32 4
31 %s128 = insertelement <8 x i64> %s127, i64 %data.i.i677.32.extract.trunc, i32 5
32 %s129 = insertelement <8 x i64> %s128, i64 %data.i.i677.16.extract.trunc, i32 6
33 %s130 = insertelement <8 x i64> %s129, i64 %data.i.i677.56.extract.trunc, i32 7
36 %s1 = load <8 x i64> , ptr %secondSource, align 64
37 %arrayidx67 = getelementptr inbounds <8 x i64>, ptr %secondSource, i32 6
38 %s121 = load <8 x i64> , ptr %arrayidx67, align 64
39 %s131 = bitcast <8 x i64> %s121 to i512
40 %data.i1.i676.48.extract.shift = lshr i512 %s131, 384
41 %data.i1.i676.48.extract.trunc = trunc i512 %data.i1.i676.48.extract.shift to i64
42 %s132 = insertelement <8 x i64> undef, i64 %data.i1.i676.48.extract.trunc, i32 0
43 %data.i1.i676.32.extract.shift = lshr i512 %s131, 256
44 %data.i1.i676.32.extract.trunc = trunc i512 %data.i1.i676.32.extract.shift to i64
45 %s133 = insertelement <8 x i64> %s132, i64 %data.i1.i676.32.extract.trunc, i32 1
46 %data.i1.i676.16.extract.shift = lshr i512 %s131, 128
47 %data.i1.i676.16.extract.trunc = trunc i512 %data.i1.i676.16.extract.shift to i64
48 %s134 = insertelement <8 x i64> %s133, i64 %data.i1.i676.16.extract.trunc, i32 2
49 %data.i1.i676.56.extract.shift = lshr i512 %s131, 448
50 %data.i1.i676.56.extract.trunc = trunc i512 %data.i1.i676.56.extract.shift to i64
51 %s135 = insertelement <8 x i64> %s134, i64 %data.i1.i676.56.extract.trunc, i32 3
52 %data.i1.i676.24.extract.shift = lshr i512 %s131, 192
53 %data.i1.i676.24.extract.trunc = trunc i512 %data.i1.i676.24.extract.shift to i64
54 %s136 = insertelement <8 x i64> %s135, i64 %data.i1.i676.24.extract.trunc, i32 4
55 %s137 = insertelement <8 x i64> %s136, i64 %data.i1.i676.32.extract.trunc, i32 5
56 %s138 = insertelement <8 x i64> %s137, i64 %data.i1.i676.16.extract.trunc, i32 6
57 %s139 = insertelement <8 x i64> %s138, i64 %data.i1.i676.56.extract.trunc, i32 7
59 ; Operations about %Source and %secondSource
60 %vecinit28.i.i699 = shufflevector <8 x i64> %s139, <8 x i64> %s130, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 12, i32 undef, i32 undef, i32 undef>
61 %vecinit35.i.i700 = shufflevector <8 x i64> %vecinit28.i.i699, <8 x i64> %s139, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 13, i32 undef, i32 undef>
62 %vecinit42.i.i701 = shufflevector <8 x i64> %vecinit35.i.i700, <8 x i64> %s139, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 14, i32 undef>
63 %vecinit49.i.i702 = shufflevector <8 x i64> %vecinit42.i.i701, <8 x i64> %s130, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 15>
64 %arrayidx72 = getelementptr inbounds <8 x i64>, ptr %dest, i32 6
65 store <8 x i64> %vecinit49.i.i702, ptr %arrayidx72, align 64
66 %arrayidx78 = getelementptr inbounds <8 x i64>, ptr %secondSource, i32 7
67 %s141 = load <8 x i64> , ptr %arrayidx78, align 64
68 %s151 = bitcast <8 x i64> %s141 to i512
69 %data.i1.i649.32.extract.shift = lshr i512 %s151, 256
70 %data.i1.i649.32.extract.trunc = trunc i512 %data.i1.i649.32.extract.shift to i64
71 %s152 = insertelement <8 x i64> undef, i64 %data.i1.i649.32.extract.trunc, i32 0
72 %s153 = insertelement <8 x i64> %s152, i64 %data.i1.i649.32.extract.trunc, i32 1
73 %data.i1.i649.16.extract.shift = lshr i512 %s151, 128
74 %data.i1.i649.16.extract.trunc = trunc i512 %data.i1.i649.16.extract.shift to i64
75 %s154 = insertelement <8 x i64> %s153, i64 %data.i1.i649.16.extract.trunc, i32 2
76 %data.i1.i649.8.extract.shift = lshr i512 %s151, 64
77 %data.i1.i649.8.extract.trunc = trunc i512 %data.i1.i649.8.extract.shift to i64
78 %s155 = insertelement <8 x i64> %s154, i64 %data.i1.i649.8.extract.trunc, i32 3
79 %arrayidx83 = getelementptr inbounds <8 x i64>, ptr %dest, i32 7
80 store <8 x i64> %s155, ptr %arrayidx83, align 64