1 ; RUN: llc < %s -march=arm -mattr=+neon > %t
2 ; RUN: grep {vpadal\\.s8} %t | count 2
3 ; RUN: grep {vpadal\\.s16} %t | count 2
4 ; RUN: grep {vpadal\\.s32} %t | count 2
5 ; RUN: grep {vpadal\\.u8} %t | count 2
6 ; RUN: grep {vpadal\\.u16} %t | count 2
7 ; RUN: grep {vpadal\\.u32} %t | count 2
9 define <4 x i16> @vpadals8(<4 x i16>* %A, <8 x i8>* %B) nounwind {
10 %tmp1 = load <4 x i16>* %A
11 %tmp2 = load <8 x i8>* %B
12 %tmp3 = call <4 x i16> @llvm.arm.neon.vpadals.v4i16.v8i8(<4 x i16> %tmp1, <8 x i8> %tmp2)
16 define <2 x i32> @vpadals16(<2 x i32>* %A, <4 x i16>* %B) nounwind {
17 %tmp1 = load <2 x i32>* %A
18 %tmp2 = load <4 x i16>* %B
19 %tmp3 = call <2 x i32> @llvm.arm.neon.vpadals.v2i32.v4i16(<2 x i32> %tmp1, <4 x i16> %tmp2)
23 define <1 x i64> @vpadals32(<1 x i64>* %A, <2 x i32>* %B) nounwind {
24 %tmp1 = load <1 x i64>* %A
25 %tmp2 = load <2 x i32>* %B
26 %tmp3 = call <1 x i64> @llvm.arm.neon.vpadals.v1i64.v2i32(<1 x i64> %tmp1, <2 x i32> %tmp2)
30 define <4 x i16> @vpadalu8(<4 x i16>* %A, <8 x i8>* %B) nounwind {
31 %tmp1 = load <4 x i16>* %A
32 %tmp2 = load <8 x i8>* %B
33 %tmp3 = call <4 x i16> @llvm.arm.neon.vpadalu.v4i16.v8i8(<4 x i16> %tmp1, <8 x i8> %tmp2)
37 define <2 x i32> @vpadalu16(<2 x i32>* %A, <4 x i16>* %B) nounwind {
38 %tmp1 = load <2 x i32>* %A
39 %tmp2 = load <4 x i16>* %B
40 %tmp3 = call <2 x i32> @llvm.arm.neon.vpadalu.v2i32.v4i16(<2 x i32> %tmp1, <4 x i16> %tmp2)
44 define <1 x i64> @vpadalu32(<1 x i64>* %A, <2 x i32>* %B) nounwind {
45 %tmp1 = load <1 x i64>* %A
46 %tmp2 = load <2 x i32>* %B
47 %tmp3 = call <1 x i64> @llvm.arm.neon.vpadalu.v1i64.v2i32(<1 x i64> %tmp1, <2 x i32> %tmp2)
51 define <8 x i16> @vpadalQs8(<8 x i16>* %A, <16 x i8>* %B) nounwind {
52 %tmp1 = load <8 x i16>* %A
53 %tmp2 = load <16 x i8>* %B
54 %tmp3 = call <8 x i16> @llvm.arm.neon.vpadals.v8i16.v16i8(<8 x i16> %tmp1, <16 x i8> %tmp2)
58 define <4 x i32> @vpadalQs16(<4 x i32>* %A, <8 x i16>* %B) nounwind {
59 %tmp1 = load <4 x i32>* %A
60 %tmp2 = load <8 x i16>* %B
61 %tmp3 = call <4 x i32> @llvm.arm.neon.vpadals.v4i32.v8i16(<4 x i32> %tmp1, <8 x i16> %tmp2)
65 define <2 x i64> @vpadalQs32(<2 x i64>* %A, <4 x i32>* %B) nounwind {
66 %tmp1 = load <2 x i64>* %A
67 %tmp2 = load <4 x i32>* %B
68 %tmp3 = call <2 x i64> @llvm.arm.neon.vpadals.v2i64.v4i32(<2 x i64> %tmp1, <4 x i32> %tmp2)
72 define <8 x i16> @vpadalQu8(<8 x i16>* %A, <16 x i8>* %B) nounwind {
73 %tmp1 = load <8 x i16>* %A
74 %tmp2 = load <16 x i8>* %B
75 %tmp3 = call <8 x i16> @llvm.arm.neon.vpadalu.v8i16.v16i8(<8 x i16> %tmp1, <16 x i8> %tmp2)
79 define <4 x i32> @vpadalQu16(<4 x i32>* %A, <8 x i16>* %B) nounwind {
80 %tmp1 = load <4 x i32>* %A
81 %tmp2 = load <8 x i16>* %B
82 %tmp3 = call <4 x i32> @llvm.arm.neon.vpadalu.v4i32.v8i16(<4 x i32> %tmp1, <8 x i16> %tmp2)
86 define <2 x i64> @vpadalQu32(<2 x i64>* %A, <4 x i32>* %B) nounwind {
87 %tmp1 = load <2 x i64>* %A
88 %tmp2 = load <4 x i32>* %B
89 %tmp3 = call <2 x i64> @llvm.arm.neon.vpadalu.v2i64.v4i32(<2 x i64> %tmp1, <4 x i32> %tmp2)
93 declare <4 x i16> @llvm.arm.neon.vpadals.v4i16.v8i8(<4 x i16>, <8 x i8>) nounwind readnone
94 declare <2 x i32> @llvm.arm.neon.vpadals.v2i32.v4i16(<2 x i32>, <4 x i16>) nounwind readnone
95 declare <1 x i64> @llvm.arm.neon.vpadals.v1i64.v2i32(<1 x i64>, <2 x i32>) nounwind readnone
97 declare <4 x i16> @llvm.arm.neon.vpadalu.v4i16.v8i8(<4 x i16>, <8 x i8>) nounwind readnone
98 declare <2 x i32> @llvm.arm.neon.vpadalu.v2i32.v4i16(<2 x i32>, <4 x i16>) nounwind readnone
99 declare <1 x i64> @llvm.arm.neon.vpadalu.v1i64.v2i32(<1 x i64>, <2 x i32>) nounwind readnone
101 declare <8 x i16> @llvm.arm.neon.vpadals.v8i16.v16i8(<8 x i16>, <16 x i8>) nounwind readnone
102 declare <4 x i32> @llvm.arm.neon.vpadals.v4i32.v8i16(<4 x i32>, <8 x i16>) nounwind readnone
103 declare <2 x i64> @llvm.arm.neon.vpadals.v2i64.v4i32(<2 x i64>, <4 x i32>) nounwind readnone
105 declare <8 x i16> @llvm.arm.neon.vpadalu.v8i16.v16i8(<8 x i16>, <16 x i8>) nounwind readnone
106 declare <4 x i32> @llvm.arm.neon.vpadalu.v4i32.v8i16(<4 x i32>, <8 x i16>) nounwind readnone
107 declare <2 x i64> @llvm.arm.neon.vpadalu.v2i64.v4i32(<2 x i64>, <4 x i32>) nounwind readnone