1; RUN: llc < %s -mtriple armeb-eabi -mattr v7,neon -o - | FileCheck %s
2
3define void @vector_ext_2i8_to_2i64( <2 x i8>* %loadaddr, <2 x i64>* %storeaddr ) {
4; CHECK-LABEL: vector_ext_2i8_to_2i64:
5; CHECK:      vld1.16   {[[REG:d[0-9]+]][0]}, [r0:16]
6; CHECK-NEXT: vrev16.8  [[REG]], [[REG]]
7; CHECK-NEXT: vmovl.u8  [[QREG:q[0-9]+]], [[REG]]
8; CHECK-NEXT: vmovl.u16 [[QREG]], [[REG]]
9; CHECK-NEXT: vmovl.u32 [[QREG]], [[REG]]
10; CHECK-NEXT: vst1.64   {[[REG]], {{d[0-9]+}}}, [r1]
11; CHECK-NEXT: bx        lr
12  %1 = load <2 x i8>, <2 x i8>* %loadaddr
13  %2 = zext <2 x i8> %1 to <2 x i64>
14  store <2 x i64> %2, <2 x i64>* %storeaddr
15  ret void
16}
17
18define void @vector_ext_2i16_to_2i64( <2 x i16>* %loadaddr, <2 x i64>* %storeaddr ) {
19; CHECK-LABEL: vector_ext_2i16_to_2i64:
20; CHECK:      vld1.32   {[[REG:d[0-9]+]][0]}, [r0:32]
21; CHECK-NEXT: vrev32.16 [[REG]], [[REG]]
22; CHECK-NEXT: vmovl.u16 [[QREG:q[0-9]+]], [[REG]]
23; CHECK-NEXT: vmovl.u32 [[QREG]], [[REG]]
24; CHECK-NEXT: vst1.64   {[[REG]], {{d[0-9]+}}}, [r1]
25; CHECK-NEXT: bx        lr
26  %1 = load <2 x i16>, <2 x i16>* %loadaddr
27  %2 = zext <2 x i16> %1 to <2 x i64>
28  store <2 x i64> %2, <2 x i64>* %storeaddr
29  ret void
30}
31
32
33define void @vector_ext_2i8_to_2i32( <2 x i8>* %loadaddr, <2 x i32>* %storeaddr ) {
34; CHECK-LABEL: vector_ext_2i8_to_2i32:
35; CHECK:      vld1.16   {[[REG:d[0-9]+]][0]}, [r0:16]
36; CHECK-NEXT: vrev16.8  [[REG]], [[REG]]
37; CHECK-NEXT: vmovl.u8  [[QREG:q[0-9]+]], [[REG]]
38; CHECK-NEXT: vmovl.u16 [[QREG]], [[REG]]
39; CHECK-NEXT: vrev64.32 [[REG]], [[REG]]
40; CHECK-NEXT: vstr      [[REG]], [r1]
41; CHECK-NEXT: bx        lr
42  %1 = load <2 x i8>, <2 x i8>* %loadaddr
43  %2 = zext <2 x i8> %1 to <2 x i32>
44  store <2 x i32> %2, <2 x i32>* %storeaddr
45  ret void
46}
47
48define void @vector_ext_2i16_to_2i32( <2 x i16>* %loadaddr, <2 x i32>* %storeaddr ) {
49; CHECK-LABEL: vector_ext_2i16_to_2i32:
50; CHECK:      vld1.32   {[[REG:d[0-9]+]][0]}, [r0:32]
51; CHECK-NEXT: vrev32.16 [[REG]], [[REG]]
52; CHECK-NEXT: vmovl.u16 [[QREG:q[0-9]+]], [[REG]]
53; CHECK-NEXT: vrev64.32 [[REG]], [[REG]]
54; CHECK-NEXT: vstr      [[REG]], [r1]
55; CHECK-NEXT: bx        lr
56  %1 = load <2 x i16>, <2 x i16>* %loadaddr
57  %2 = zext <2 x i16> %1 to <2 x i32>
58  store <2 x i32> %2, <2 x i32>* %storeaddr
59  ret void
60}
61
62define void @vector_ext_2i8_to_2i16( <2 x i8>* %loadaddr, <2 x i16>* %storeaddr ) {
63; CHECK-LABEL: vector_ext_2i8_to_2i16:
64; CHECK:      vld1.16   {[[REG:d[0-9]+]][0]}, [r0:16]
65; CHECK-NEXT: vrev16.8  [[REG]], [[REG]]
66; CHECK-NEXT: vmovl.u8  [[QREG:q[0-9]+]], [[REG]]
67; CHECK-NEXT: vmovl.u16 [[QREG]], [[REG]]
68; CHECK-NEXT: vrev32.16 [[REG]], [[REG]]
69; CHECK-NEXT: vuzp.16   [[REG]], {{d[0-9]+}}
70; CHECK-NEXT: vrev32.16 [[REG]], {{d[0-9]+}}
71; CHECK-NEXT: vst1.32   {[[REG]][0]}, [r1:32]
72; CHECK-NEXT: bx        lr
73  %1 = load <2 x i8>, <2 x i8>* %loadaddr
74  %2 = zext <2 x i8> %1 to <2 x i16>
75  store <2 x i16> %2, <2 x i16>* %storeaddr
76  ret void
77}
78
79define void @vector_ext_4i8_to_4i32( <4 x i8>* %loadaddr, <4 x i32>* %storeaddr ) {
80; CHECK-LABEL: vector_ext_4i8_to_4i32:
81; CHECK:      vld1.32   {[[REG:d[0-9]+]][0]}, [r0:32]
82; CHECK-NEXT: vrev32.8  [[REG]], [[REG]]
83; CHECK-NEXT: vmovl.u8  [[QREG:q[0-9]+]], [[REG]]
84; CHECK-NEXT: vmovl.u16 [[QREG]], [[REG]]
85; CHECK-NEXT: vrev64.32 [[QREG]], [[QREG]]
86; CHECK-NEXT: vst1.64   {[[REG]], {{d[0-9]+}}}, [r1]
87; CHECK-NEXT: bx        lr
88  %1 = load <4 x i8>, <4 x i8>* %loadaddr
89  %2 = zext <4 x i8> %1 to <4 x i32>
90  store <4 x i32> %2, <4 x i32>* %storeaddr
91  ret void
92}
93
94define void @vector_ext_4i8_to_4i16( <4 x i8>* %loadaddr, <4 x i16>* %storeaddr ) {
95; CHECK-LABEL: vector_ext_4i8_to_4i16:
96; CHECK:      vld1.32   {[[REG:d[0-9]+]][0]}, [r0:32]
97; CHECK-NEXT: vrev32.8  [[REG]], [[REG]]
98; CHECK-NEXT: vmovl.u8  [[QREG:q[0-9]+]], [[REG]]
99; CHECK-NEXT: vrev64.16 [[REG]], [[REG]]
100; CHECK-NEXT: vstr      [[REG]], [r1]
101; CHECK-NEXT: bx        lr
102  %1 = load <4 x i8>, <4 x i8>* %loadaddr
103  %2 = zext <4 x i8> %1 to <4 x i16>
104  store <4 x i16> %2, <4 x i16>* %storeaddr
105  ret void
106}
107