23 lines · plain
1; RUN: llc < %s -mtriple=arm64-eabi -aarch64-neon-syntax=apple -mcpu=cyclone | FileCheck %s2; RUN: llc < %s -mtriple=arm64-eabi -aarch64-neon-syntax=apple -O0 -fast-isel -mcpu=cyclone | FileCheck %s --check-prefix=CHECK-FAST3 4define <16 x i8> @foo(<16 x i8> %a) nounwind optsize readnone ssp {5; CHECK: uaddlv.16b h0, v06; CHECK: rshrn.8b v0, v0, #47; CHECK: dup.16b v0, v0[0]8; CHECK: ret9 10; CHECK-FAST: uaddlv.16b11; CHECK-FAST: rshrn.8b12; CHECK-FAST: dup.16b13 %tmp = tail call i32 @llvm.aarch64.neon.uaddlv.i32.v16i8(<16 x i8> %a) nounwind14 %tmp1 = trunc i32 %tmp to i1615 %tmp2 = insertelement <8 x i16> undef, i16 %tmp1, i32 016 %tmp3 = tail call <8 x i8> @llvm.aarch64.neon.rshrn.v8i8(<8 x i16> %tmp2, i32 4)17 %tmp4 = shufflevector <8 x i8> %tmp3, <8 x i8> undef, <16 x i32> zeroinitializer18 ret <16 x i8> %tmp419}20 21declare <8 x i8> @llvm.aarch64.neon.rshrn.v8i8(<8 x i16>, i32) nounwind readnone22declare i32 @llvm.aarch64.neon.uaddlv.i32.v16i8(<16 x i8>) nounwind readnone23