1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
3 define <8 x i8> @vpaddi8(<8 x i8>* %A, <8 x i8>* %B) nounwind {
6 %tmp1 = load <8 x i8>* %A
7 %tmp2 = load <8 x i8>* %B
8 %tmp3 = call <8 x i8> @llvm.arm.neon.vpadd.v8i8(<8 x i8> %tmp1, <8 x i8> %tmp2)
12 define <4 x i16> @vpaddi16(<4 x i16>* %A, <4 x i16>* %B) nounwind {
15 %tmp1 = load <4 x i16>* %A
16 %tmp2 = load <4 x i16>* %B
17 %tmp3 = call <4 x i16> @llvm.arm.neon.vpadd.v4i16(<4 x i16> %tmp1, <4 x i16> %tmp2)
21 define <2 x i32> @vpaddi32(<2 x i32>* %A, <2 x i32>* %B) nounwind {
24 %tmp1 = load <2 x i32>* %A
25 %tmp2 = load <2 x i32>* %B
26 %tmp3 = call <2 x i32> @llvm.arm.neon.vpadd.v2i32(<2 x i32> %tmp1, <2 x i32> %tmp2)
30 define <2 x float> @vpaddf32(<2 x float>* %A, <2 x float>* %B) nounwind {
33 %tmp1 = load <2 x float>* %A
34 %tmp2 = load <2 x float>* %B
35 %tmp3 = call <2 x float> @llvm.arm.neon.vpadd.v2f32(<2 x float> %tmp1, <2 x float> %tmp2)
39 declare <8 x i8> @llvm.arm.neon.vpadd.v8i8(<8 x i8>, <8 x i8>) nounwind readnone
40 declare <4 x i16> @llvm.arm.neon.vpadd.v4i16(<4 x i16>, <4 x i16>) nounwind readnone
41 declare <2 x i32> @llvm.arm.neon.vpadd.v2i32(<2 x i32>, <2 x i32>) nounwind readnone
43 declare <2 x float> @llvm.arm.neon.vpadd.v2f32(<2 x float>, <2 x float>) nounwind readnone
45 define <4 x i16> @vpaddls8(<8 x i8>* %A) nounwind {
48 %tmp1 = load <8 x i8>* %A
49 %tmp2 = call <4 x i16> @llvm.arm.neon.vpaddls.v4i16.v8i8(<8 x i8> %tmp1)
53 define <2 x i32> @vpaddls16(<4 x i16>* %A) nounwind {
56 %tmp1 = load <4 x i16>* %A
57 %tmp2 = call <2 x i32> @llvm.arm.neon.vpaddls.v2i32.v4i16(<4 x i16> %tmp1)
61 define <1 x i64> @vpaddls32(<2 x i32>* %A) nounwind {
64 %tmp1 = load <2 x i32>* %A
65 %tmp2 = call <1 x i64> @llvm.arm.neon.vpaddls.v1i64.v2i32(<2 x i32> %tmp1)
69 define <4 x i16> @vpaddlu8(<8 x i8>* %A) nounwind {
72 %tmp1 = load <8 x i8>* %A
73 %tmp2 = call <4 x i16> @llvm.arm.neon.vpaddlu.v4i16.v8i8(<8 x i8> %tmp1)
77 define <2 x i32> @vpaddlu16(<4 x i16>* %A) nounwind {
80 %tmp1 = load <4 x i16>* %A
81 %tmp2 = call <2 x i32> @llvm.arm.neon.vpaddlu.v2i32.v4i16(<4 x i16> %tmp1)
85 define <1 x i64> @vpaddlu32(<2 x i32>* %A) nounwind {
88 %tmp1 = load <2 x i32>* %A
89 %tmp2 = call <1 x i64> @llvm.arm.neon.vpaddlu.v1i64.v2i32(<2 x i32> %tmp1)
93 define <8 x i16> @vpaddlQs8(<16 x i8>* %A) nounwind {
96 %tmp1 = load <16 x i8>* %A
97 %tmp2 = call <8 x i16> @llvm.arm.neon.vpaddls.v8i16.v16i8(<16 x i8> %tmp1)
101 define <4 x i32> @vpaddlQs16(<8 x i16>* %A) nounwind {
104 %tmp1 = load <8 x i16>* %A
105 %tmp2 = call <4 x i32> @llvm.arm.neon.vpaddls.v4i32.v8i16(<8 x i16> %tmp1)
109 define <2 x i64> @vpaddlQs32(<4 x i32>* %A) nounwind {
112 %tmp1 = load <4 x i32>* %A
113 %tmp2 = call <2 x i64> @llvm.arm.neon.vpaddls.v2i64.v4i32(<4 x i32> %tmp1)
117 define <8 x i16> @vpaddlQu8(<16 x i8>* %A) nounwind {
120 %tmp1 = load <16 x i8>* %A
121 %tmp2 = call <8 x i16> @llvm.arm.neon.vpaddlu.v8i16.v16i8(<16 x i8> %tmp1)
125 define <4 x i32> @vpaddlQu16(<8 x i16>* %A) nounwind {
128 %tmp1 = load <8 x i16>* %A
129 %tmp2 = call <4 x i32> @llvm.arm.neon.vpaddlu.v4i32.v8i16(<8 x i16> %tmp1)
133 define <2 x i64> @vpaddlQu32(<4 x i32>* %A) nounwind {
136 %tmp1 = load <4 x i32>* %A
137 %tmp2 = call <2 x i64> @llvm.arm.neon.vpaddlu.v2i64.v4i32(<4 x i32> %tmp1)
141 declare <4 x i16> @llvm.arm.neon.vpaddls.v4i16.v8i8(<8 x i8>) nounwind readnone
142 declare <2 x i32> @llvm.arm.neon.vpaddls.v2i32.v4i16(<4 x i16>) nounwind readnone
143 declare <1 x i64> @llvm.arm.neon.vpaddls.v1i64.v2i32(<2 x i32>) nounwind readnone
145 declare <4 x i16> @llvm.arm.neon.vpaddlu.v4i16.v8i8(<8 x i8>) nounwind readnone
146 declare <2 x i32> @llvm.arm.neon.vpaddlu.v2i32.v4i16(<4 x i16>) nounwind readnone
147 declare <1 x i64> @llvm.arm.neon.vpaddlu.v1i64.v2i32(<2 x i32>) nounwind readnone
149 declare <8 x i16> @llvm.arm.neon.vpaddls.v8i16.v16i8(<16 x i8>) nounwind readnone
150 declare <4 x i32> @llvm.arm.neon.vpaddls.v4i32.v8i16(<8 x i16>) nounwind readnone
151 declare <2 x i64> @llvm.arm.neon.vpaddls.v2i64.v4i32(<4 x i32>) nounwind readnone
153 declare <8 x i16> @llvm.arm.neon.vpaddlu.v8i16.v16i8(<16 x i8>) nounwind readnone
154 declare <4 x i32> @llvm.arm.neon.vpaddlu.v4i32.v8i16(<8 x i16>) nounwind readnone
155 declare <2 x i64> @llvm.arm.neon.vpaddlu.v2i64.v4i32(<4 x i32>) nounwind readnone