1 ; RUN: llc < %s -march=arm -mattr=+neon > %t
2 ; RUN: grep {vsli\\.8} %t | count 2
3 ; RUN: grep {vsli\\.16} %t | count 2
4 ; RUN: grep {vsli\\.32} %t | count 2
5 ; RUN: grep {vsli\\.64} %t | count 2
6 ; RUN: grep {vsri\\.8} %t | count 2
7 ; RUN: grep {vsri\\.16} %t | count 2
8 ; RUN: grep {vsri\\.32} %t | count 2
9 ; RUN: grep {vsri\\.64} %t | count 2
11 define <8 x i8> @vsli8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
12 %tmp1 = load <8 x i8>* %A
13 %tmp2 = load <8 x i8>* %B
14 %tmp3 = call <8 x i8> @llvm.arm.neon.vshiftins.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
18 define <4 x i16> @vsli16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
19 %tmp1 = load <4 x i16>* %A
20 %tmp2 = load <4 x i16>* %B
21 %tmp3 = call <4 x i16> @llvm.arm.neon.vshiftins.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >)
25 define <2 x i32> @vsli32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
26 %tmp1 = load <2 x i32>* %A
27 %tmp2 = load <2 x i32>* %B
28 %tmp3 = call <2 x i32> @llvm.arm.neon.vshiftins.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2, <2 x i32> < i32 31, i32 31 >)
32 define <1 x i64> @vsli64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
33 %tmp1 = load <1 x i64>* %A
34 %tmp2 = load <1 x i64>* %B
35 %tmp3 = call <1 x i64> @llvm.arm.neon.vshiftins.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2, <1 x i64> < i64 63 >)
39 define <16 x i8> @vsliQ8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
40 %tmp1 = load <16 x i8>* %A
41 %tmp2 = load <16 x i8>* %B
42 %tmp3 = call <16 x i8> @llvm.arm.neon.vshiftins.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2, <16 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
46 define <8 x i16> @vsliQ16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
47 %tmp1 = load <8 x i16>* %A
48 %tmp2 = load <8 x i16>* %B
49 %tmp3 = call <8 x i16> @llvm.arm.neon.vshiftins.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2, <8 x i16> < i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15 >)
53 define <4 x i32> @vsliQ32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
54 %tmp1 = load <4 x i32>* %A
55 %tmp2 = load <4 x i32>* %B
56 %tmp3 = call <4 x i32> @llvm.arm.neon.vshiftins.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2, <4 x i32> < i32 31, i32 31, i32 31, i32 31 >)
60 define <2 x i64> @vsliQ64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
61 %tmp1 = load <2 x i64>* %A
62 %tmp2 = load <2 x i64>* %B
63 %tmp3 = call <2 x i64> @llvm.arm.neon.vshiftins.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2, <2 x i64> < i64 63, i64 63 >)
67 define <8 x i8> @vsri8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
68 %tmp1 = load <8 x i8>* %A
69 %tmp2 = load <8 x i8>* %B
70 %tmp3 = call <8 x i8> @llvm.arm.neon.vshiftins.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2, <8 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
74 define <4 x i16> @vsri16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
75 %tmp1 = load <4 x i16>* %A
76 %tmp2 = load <4 x i16>* %B
77 %tmp3 = call <4 x i16> @llvm.arm.neon.vshiftins.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2, <4 x i16> < i16 -16, i16 -16, i16 -16, i16 -16 >)
81 define <2 x i32> @vsri32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
82 %tmp1 = load <2 x i32>* %A
83 %tmp2 = load <2 x i32>* %B
84 %tmp3 = call <2 x i32> @llvm.arm.neon.vshiftins.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2, <2 x i32> < i32 -32, i32 -32 >)
88 define <1 x i64> @vsri64(<1 x i64>* %A, <1 x i64>* %B) nounwind {
89 %tmp1 = load <1 x i64>* %A
90 %tmp2 = load <1 x i64>* %B
91 %tmp3 = call <1 x i64> @llvm.arm.neon.vshiftins.v1i64(<1 x i64> %tmp1, <1 x i64> %tmp2, <1 x i64> < i64 -64 >)
95 define <16 x i8> @vsriQ8(<16 x i8>* %A, <16 x i8>* %B) nounwind {
96 %tmp1 = load <16 x i8>* %A
97 %tmp2 = load <16 x i8>* %B
98 %tmp3 = call <16 x i8> @llvm.arm.neon.vshiftins.v16i8(<16 x i8> %tmp1, <16 x i8> %tmp2, <16 x i8> < i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8, i8 -8 >)
102 define <8 x i16> @vsriQ16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
103 %tmp1 = load <8 x i16>* %A
104 %tmp2 = load <8 x i16>* %B
105 %tmp3 = call <8 x i16> @llvm.arm.neon.vshiftins.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2, <8 x i16> < i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16, i16 -16 >)
109 define <4 x i32> @vsriQ32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
110 %tmp1 = load <4 x i32>* %A
111 %tmp2 = load <4 x i32>* %B
112 %tmp3 = call <4 x i32> @llvm.arm.neon.vshiftins.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2, <4 x i32> < i32 -32, i32 -32, i32 -32, i32 -32 >)
116 define <2 x i64> @vsriQ64(<2 x i64>* %A, <2 x i64>* %B) nounwind {
117 %tmp1 = load <2 x i64>* %A
118 %tmp2 = load <2 x i64>* %B
119 %tmp3 = call <2 x i64> @llvm.arm.neon.vshiftins.v2i64(<2 x i64> %tmp1, <2 x i64> %tmp2, <2 x i64> < i64 -64, i64 -64 >)
123 declare <8 x i8> @llvm.arm.neon.vshiftins.v8i8(<8 x i8>, <8 x i8>, <8 x i8>) nounwind readnone
124 declare <4 x i16> @llvm.arm.neon.vshiftins.v4i16(<4 x i16>, <4 x i16>, <4 x i16>) nounwind readnone
125 declare <2 x i32> @llvm.arm.neon.vshiftins.v2i32(<2 x i32>, <2 x i32>, <2 x i32>) nounwind readnone
126 declare <1 x i64> @llvm.arm.neon.vshiftins.v1i64(<1 x i64>, <1 x i64>, <1 x i64>) nounwind readnone
128 declare <16 x i8> @llvm.arm.neon.vshiftins.v16i8(<16 x i8>, <16 x i8>, <16 x i8>) nounwind readnone
129 declare <8 x i16> @llvm.arm.neon.vshiftins.v8i16(<8 x i16>, <8 x i16>, <8 x i16>) nounwind readnone
130 declare <4 x i32> @llvm.arm.neon.vshiftins.v4i32(<4 x i32>, <4 x i32>, <4 x i32>) nounwind readnone
131 declare <2 x i64> @llvm.arm.neon.vshiftins.v2i64(<2 x i64>, <2 x i64>, <2 x i64>) nounwind readnone