1 ; RUN: llvm-as < %s | llc -march=arm -mattr=+neon > %t
2 ; RUN: grep {vqrshrn\\.s16} %t | count 1
3 ; RUN: grep {vqrshrn\\.s32} %t | count 1
4 ; RUN: grep {vqrshrn\\.s64} %t | count 1
5 ; RUN: grep {vqrshrn\\.u16} %t | count 1
6 ; RUN: grep {vqrshrn\\.u32} %t | count 1
7 ; RUN: grep {vqrshrn\\.u64} %t | count 1
8 ; RUN: grep {vqrshrun\\.s16} %t | count 1
9 ; RUN: grep {vqrshrun\\.s32} %t | count 1
10 ; RUN: grep {vqrshrun\\.s64} %t | count 1
12 define <8 x i8> @vqrshrns8(<8 x i16>* %A) nounwind {
13 %tmp1 = load <8 x i16>* %A
14 %tmp2 = call <8 x i8> @llvm.arm.neon.vqrshiftns.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >)
18 define <4 x i16> @vqrshrns16(<4 x i32>* %A) nounwind {
19 %tmp1 = load <4 x i32>* %A
20 %tmp2 = call <4 x i16> @llvm.arm.neon.vqrshiftns.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >)
24 define <2 x i32> @vqrshrns32(<2 x i64>* %A) nounwind {
25 %tmp1 = load <2 x i64>* %A
26 %tmp2 = call <2 x i32> @llvm.arm.neon.vqrshiftns.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >)
30 define <8 x i8> @vqrshrnu8(<8 x i16>* %A) nounwind {
31 %tmp1 = load <8 x i16>* %A
32 %tmp2 = call <8 x i8> @llvm.arm.neon.vqrshiftnu.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >)
36 define <4 x i16> @vqrshrnu16(<4 x i32>* %A) nounwind {
37 %tmp1 = load <4 x i32>* %A
38 %tmp2 = call <4 x i16> @llvm.arm.neon.vqrshiftnu.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >)
42 define <2 x i32> @vqrshrnu32(<2 x i64>* %A) nounwind {
43 %tmp1 = load <2 x i64>* %A
44 %tmp2 = call <2 x i32> @llvm.arm.neon.vqrshiftnu.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >)
48 define <8 x i8> @vqrshruns8(<8 x i16>* %A) nounwind {
49 %tmp1 = load <8 x i16>* %A
50 %tmp2 = call <8 x i8> @llvm.arm.neon.vqrshiftnsu.v8i8(<8 x i16> %tmp1, <8 x i16> < i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8, i16 -8 >)
54 define <4 x i16> @vqrshruns16(<4 x i32>* %A) nounwind {
55 %tmp1 = load <4 x i32>* %A
56 %tmp2 = call <4 x i16> @llvm.arm.neon.vqrshiftnsu.v4i16(<4 x i32> %tmp1, <4 x i32> < i32 -16, i32 -16, i32 -16, i32 -16 >)
60 define <2 x i32> @vqrshruns32(<2 x i64>* %A) nounwind {
61 %tmp1 = load <2 x i64>* %A
62 %tmp2 = call <2 x i32> @llvm.arm.neon.vqrshiftnsu.v2i32(<2 x i64> %tmp1, <2 x i64> < i64 -32, i64 -32 >)
66 declare <8 x i8> @llvm.arm.neon.vqrshiftns.v8i8(<8 x i16>, <8 x i16>) nounwind readnone
67 declare <4 x i16> @llvm.arm.neon.vqrshiftns.v4i16(<4 x i32>, <4 x i32>) nounwind readnone
68 declare <2 x i32> @llvm.arm.neon.vqrshiftns.v2i32(<2 x i64>, <2 x i64>) nounwind readnone
70 declare <8 x i8> @llvm.arm.neon.vqrshiftnu.v8i8(<8 x i16>, <8 x i16>) nounwind readnone
71 declare <4 x i16> @llvm.arm.neon.vqrshiftnu.v4i16(<4 x i32>, <4 x i32>) nounwind readnone
72 declare <2 x i32> @llvm.arm.neon.vqrshiftnu.v2i32(<2 x i64>, <2 x i64>) nounwind readnone
74 declare <8 x i8> @llvm.arm.neon.vqrshiftnsu.v8i8(<8 x i16>, <8 x i16>) nounwind readnone
75 declare <4 x i16> @llvm.arm.neon.vqrshiftnsu.v4i16(<4 x i32>, <4 x i32>) nounwind readnone
76 declare <2 x i32> @llvm.arm.neon.vqrshiftnsu.v2i32(<2 x i64>, <2 x i64>) nounwind readnone