1 ; RUN: llc < %s -march=arm -mattr=+neon > %t
2 ; RUN: grep {vqdmulh\\.s16} %t | count 2
3 ; RUN: grep {vqdmulh\\.s32} %t | count 2
4 ; RUN: grep {vqrdmulh\\.s16} %t | count 2
5 ; RUN: grep {vqrdmulh\\.s32} %t | count 2
7 define <4 x i16> @vqdmulhs16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
8 %tmp1 = load <4 x i16>* %A
9 %tmp2 = load <4 x i16>* %B
10 %tmp3 = call <4 x i16> @llvm.arm.neon.vqdmulh.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
14 define <2 x i32> @vqdmulhs32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
15 %tmp1 = load <2 x i32>* %A
16 %tmp2 = load <2 x i32>* %B
17 %tmp3 = call <2 x i32> @llvm.arm.neon.vqdmulh.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
21 define <8 x i16> @vqdmulhQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
22 %tmp1 = load <8 x i16>* %A
23 %tmp2 = load <8 x i16>* %B
24 %tmp3 = call <8 x i16> @llvm.arm.neon.vqdmulh.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
28 define <4 x i32> @vqdmulhQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
29 %tmp1 = load <4 x i32>* %A
30 %tmp2 = load <4 x i32>* %B
31 %tmp3 = call <4 x i32> @llvm.arm.neon.vqdmulh.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
35 declare <4 x i16> @llvm.arm.neon.vqdmulh.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
36 declare <2 x i32> @llvm.arm.neon.vqdmulh.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
38 declare <8 x i16> @llvm.arm.neon.vqdmulh.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
39 declare <4 x i32> @llvm.arm.neon.vqdmulh.v4i32(<4 x i32>, <4 x i32>) nounwind readnone
41 define <4 x i16> @vqrdmulhs16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
42 %tmp1 = load <4 x i16>* %A
43 %tmp2 = load <4 x i16>* %B
44 %tmp3 = call <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
48 define <2 x i32> @vqrdmulhs32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
49 %tmp1 = load <2 x i32>* %A
50 %tmp2 = load <2 x i32>* %B
51 %tmp3 = call <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
55 define <8 x i16> @vqrdmulhQs16(<8 x i16>* %A, <8 x i16>* %B) nounwind {
56 %tmp1 = load <8 x i16>* %A
57 %tmp2 = load <8 x i16>* %B
58 %tmp3 = call <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16> %tmp1, <8 x i16> %tmp2)
62 define <4 x i32> @vqrdmulhQs32(<4 x i32>* %A, <4 x i32>* %B) nounwind {
63 %tmp1 = load <4 x i32>* %A
64 %tmp2 = load <4 x i32>* %B
65 %tmp3 = call <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32> %tmp1, <4 x i32> %tmp2)
69 declare <4 x i16> @llvm.arm.neon.vqrdmulh.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
70 declare <2 x i32> @llvm.arm.neon.vqrdmulh.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
72 declare <8 x i16> @llvm.arm.neon.vqrdmulh.v8i16(<8 x i16>, <8 x i16>) nounwind readnone
73 declare <4 x i32> @llvm.arm.neon.vqrdmulh.v4i32(<4 x i32>, <4 x i32>) nounwind readnone