2014-04-15 19:08:46 +00:00
|
|
|
; RUN: llc < %s -march=arm64 -arm64-neon-syntax=apple -mcpu=cyclone | FileCheck %s
|
|
|
|
; RUN: llc < %s -march=arm64 -arm64-neon-syntax=apple -O0 -mcpu=cyclone | FileCheck %s --check-prefix=CHECK-FAST
|
2014-03-29 10:18:08 +00:00
|
|
|
|
|
|
|
define <16 x i8> @foo(<16 x i8> %a) nounwind optsize readnone ssp {
|
2014-04-02 14:38:58 +00:00
|
|
|
; CHECK: uaddlv.16b h0, v0
|
|
|
|
; CHECK: rshrn.8b v0, v0, #4
|
|
|
|
; CHECK: dup.16b v0, v0[0]
|
2014-03-29 10:18:08 +00:00
|
|
|
; CHECK: ret
|
2014-04-02 14:38:58 +00:00
|
|
|
|
|
|
|
; CHECK-FAST: uaddlv.16b
|
|
|
|
; CHECK-FAST: rshrn.8b
|
|
|
|
; CHECK-FAST: dup.16b
|
2014-03-29 10:18:08 +00:00
|
|
|
%tmp = tail call i32 @llvm.arm64.neon.uaddlv.i32.v16i8(<16 x i8> %a) nounwind
|
|
|
|
%tmp1 = trunc i32 %tmp to i16
|
|
|
|
%tmp2 = insertelement <8 x i16> undef, i16 %tmp1, i32 0
|
|
|
|
%tmp3 = tail call <8 x i8> @llvm.arm64.neon.rshrn.v8i8(<8 x i16> %tmp2, i32 4)
|
|
|
|
%tmp4 = shufflevector <8 x i8> %tmp3, <8 x i8> undef, <16 x i32> zeroinitializer
|
|
|
|
ret <16 x i8> %tmp4
|
|
|
|
}
|
|
|
|
|
|
|
|
declare <8 x i8> @llvm.arm64.neon.rshrn.v8i8(<8 x i16>, i32) nounwind readnone
|
|
|
|
declare i32 @llvm.arm64.neon.uaddlv.i32.v16i8(<16 x i8>) nounwind readnone
|