1 ; RUN: llc < %s -march=arm -mattr=+neon | FileCheck %s
3 define <8 x i16> @vshlls8(<8 x i8>* %A) nounwind {
6 %tmp1 = load <8 x i8>* %A
7 %tmp2 = call <8 x i16> @llvm.arm.neon.vshiftls.v8i16(<8 x i8> %tmp1, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
11 define <4 x i32> @vshlls16(<4 x i16>* %A) nounwind {
14 %tmp1 = load <4 x i16>* %A
15 %tmp2 = call <4 x i32> @llvm.arm.neon.vshiftls.v4i32(<4 x i16> %tmp1, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >)
19 define <2 x i64> @vshlls32(<2 x i32>* %A) nounwind {
22 %tmp1 = load <2 x i32>* %A
23 %tmp2 = call <2 x i64> @llvm.arm.neon.vshiftls.v2i64(<2 x i32> %tmp1, <2 x i32> < i32 31, i32 31 >)
27 define <8 x i16> @vshllu8(<8 x i8>* %A) nounwind {
30 %tmp1 = load <8 x i8>* %A
31 %tmp2 = call <8 x i16> @llvm.arm.neon.vshiftlu.v8i16(<8 x i8> %tmp1, <8 x i8> < i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7 >)
35 define <4 x i32> @vshllu16(<4 x i16>* %A) nounwind {
38 %tmp1 = load <4 x i16>* %A
39 %tmp2 = call <4 x i32> @llvm.arm.neon.vshiftlu.v4i32(<4 x i16> %tmp1, <4 x i16> < i16 15, i16 15, i16 15, i16 15 >)
43 define <2 x i64> @vshllu32(<2 x i32>* %A) nounwind {
46 %tmp1 = load <2 x i32>* %A
47 %tmp2 = call <2 x i64> @llvm.arm.neon.vshiftlu.v2i64(<2 x i32> %tmp1, <2 x i32> < i32 31, i32 31 >)
51 ; The following tests use the maximum shift count, so the signedness is
52 ; irrelevant. Test both signed and unsigned versions.
53 define <8 x i16> @vshlli8(<8 x i8>* %A) nounwind {
56 %tmp1 = load <8 x i8>* %A
57 %tmp2 = call <8 x i16> @llvm.arm.neon.vshiftls.v8i16(<8 x i8> %tmp1, <8 x i8> < i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8 >)
61 define <4 x i32> @vshlli16(<4 x i16>* %A) nounwind {
64 %tmp1 = load <4 x i16>* %A
65 %tmp2 = call <4 x i32> @llvm.arm.neon.vshiftlu.v4i32(<4 x i16> %tmp1, <4 x i16> < i16 16, i16 16, i16 16, i16 16 >)
69 define <2 x i64> @vshlli32(<2 x i32>* %A) nounwind {
72 %tmp1 = load <2 x i32>* %A
73 %tmp2 = call <2 x i64> @llvm.arm.neon.vshiftls.v2i64(<2 x i32> %tmp1, <2 x i32> < i32 32, i32 32 >)
77 declare <8 x i16> @llvm.arm.neon.vshiftls.v8i16(<8 x i8>, <8 x i8>) nounwind readnone
78 declare <4 x i32> @llvm.arm.neon.vshiftls.v4i32(<4 x i16>, <4 x i16>) nounwind readnone
79 declare <2 x i64> @llvm.arm.neon.vshiftls.v2i64(<2 x i32>, <2 x i32>) nounwind readnone
81 declare <8 x i16> @llvm.arm.neon.vshiftlu.v8i16(<8 x i8>, <8 x i8>) nounwind readnone
82 declare <4 x i32> @llvm.arm.neon.vshiftlu.v4i32(<4 x i16>, <4 x i16>) nounwind readnone
83 declare <2 x i64> @llvm.arm.neon.vshiftlu.v2i64(<2 x i32>, <2 x i32>) nounwind readnone