1 ; RUN: llvm-as < %s | llc -march=arm -mattr=+neon > %t
2 ; RUN: grep {vpaddl\\.s8} %t | count 2
3 ; RUN: grep {vpaddl\\.s16} %t | count 2
4 ; RUN: grep {vpaddl\\.s32} %t | count 2
5 ; RUN: grep {vpaddl\\.u8} %t | count 2
6 ; RUN: grep {vpaddl\\.u16} %t | count 2
7 ; RUN: grep {vpaddl\\.u32} %t | count 2
9 define <4 x i16> @vpaddls8(<8 x i8>* %A) nounwind {
10 %tmp1 = load <8 x i8>* %A
11 %tmp2 = call <4 x i16> @llvm.arm.neon.vpaddls.v4i16.v8i8(<8 x i8> %tmp1)
15 define <2 x i32> @vpaddls16(<4 x i16>* %A) nounwind {
16 %tmp1 = load <4 x i16>* %A
17 %tmp2 = call <2 x i32> @llvm.arm.neon.vpaddls.v2i32.v4i16(<4 x i16> %tmp1)
21 define <1 x i64> @vpaddls32(<2 x i32>* %A) nounwind {
22 %tmp1 = load <2 x i32>* %A
23 %tmp2 = call <1 x i64> @llvm.arm.neon.vpaddls.v1i64.v2i32(<2 x i32> %tmp1)
27 define <4 x i16> @vpaddlu8(<8 x i8>* %A) nounwind {
28 %tmp1 = load <8 x i8>* %A
29 %tmp2 = call <4 x i16> @llvm.arm.neon.vpaddlu.v4i16.v8i8(<8 x i8> %tmp1)
33 define <2 x i32> @vpaddlu16(<4 x i16>* %A) nounwind {
34 %tmp1 = load <4 x i16>* %A
35 %tmp2 = call <2 x i32> @llvm.arm.neon.vpaddlu.v2i32.v4i16(<4 x i16> %tmp1)
39 define <1 x i64> @vpaddlu32(<2 x i32>* %A) nounwind {
40 %tmp1 = load <2 x i32>* %A
41 %tmp2 = call <1 x i64> @llvm.arm.neon.vpaddlu.v1i64.v2i32(<2 x i32> %tmp1)
45 define <8 x i16> @vpaddlQs8(<16 x i8>* %A) nounwind {
46 %tmp1 = load <16 x i8>* %A
47 %tmp2 = call <8 x i16> @llvm.arm.neon.vpaddls.v8i16.v16i8(<16 x i8> %tmp1)
51 define <4 x i32> @vpaddlQs16(<8 x i16>* %A) nounwind {
52 %tmp1 = load <8 x i16>* %A
53 %tmp2 = call <4 x i32> @llvm.arm.neon.vpaddls.v4i32.v8i16(<8 x i16> %tmp1)
57 define <2 x i64> @vpaddlQs32(<4 x i32>* %A) nounwind {
58 %tmp1 = load <4 x i32>* %A
59 %tmp2 = call <2 x i64> @llvm.arm.neon.vpaddls.v2i64.v4i32(<4 x i32> %tmp1)
63 define <8 x i16> @vpaddlQu8(<16 x i8>* %A) nounwind {
64 %tmp1 = load <16 x i8>* %A
65 %tmp2 = call <8 x i16> @llvm.arm.neon.vpaddlu.v8i16.v16i8(<16 x i8> %tmp1)
69 define <4 x i32> @vpaddlQu16(<8 x i16>* %A) nounwind {
70 %tmp1 = load <8 x i16>* %A
71 %tmp2 = call <4 x i32> @llvm.arm.neon.vpaddlu.v4i32.v8i16(<8 x i16> %tmp1)
75 define <2 x i64> @vpaddlQu32(<4 x i32>* %A) nounwind {
76 %tmp1 = load <4 x i32>* %A
77 %tmp2 = call <2 x i64> @llvm.arm.neon.vpaddlu.v2i64.v4i32(<4 x i32> %tmp1)
81 declare <4 x i16> @llvm.arm.neon.vpaddls.v4i16.v8i8(<8 x i8>) nounwind readnone
82 declare <2 x i32> @llvm.arm.neon.vpaddls.v2i32.v4i16(<4 x i16>) nounwind readnone
83 declare <1 x i64> @llvm.arm.neon.vpaddls.v1i64.v2i32(<2 x i32>) nounwind readnone
85 declare <4 x i16> @llvm.arm.neon.vpaddlu.v4i16.v8i8(<8 x i8>) nounwind readnone
86 declare <2 x i32> @llvm.arm.neon.vpaddlu.v2i32.v4i16(<4 x i16>) nounwind readnone
87 declare <1 x i64> @llvm.arm.neon.vpaddlu.v1i64.v2i32(<2 x i32>) nounwind readnone
89 declare <8 x i16> @llvm.arm.neon.vpaddls.v8i16.v16i8(<16 x i8>) nounwind readnone
90 declare <4 x i32> @llvm.arm.neon.vpaddls.v4i32.v8i16(<8 x i16>) nounwind readnone
91 declare <2 x i64> @llvm.arm.neon.vpaddls.v2i64.v4i32(<4 x i32>) nounwind readnone
93 declare <8 x i16> @llvm.arm.neon.vpaddlu.v8i16.v16i8(<16 x i8>) nounwind readnone
94 declare <4 x i32> @llvm.arm.neon.vpaddlu.v4i32.v8i16(<8 x i16>) nounwind readnone
95 declare <2 x i64> @llvm.arm.neon.vpaddlu.v2i64.v4i32(<4 x i32>) nounwind readnone