From 7d628ab7cf11d887f5d2887ccabb0f8095fb4b25 Mon Sep 17 00:00:00 2001 From: Wilco Dijkstra Date: Wed, 11 Jan 2023 13:52:23 +0000 Subject: [PATCH] AArch64: Improve strchrnul Unroll the main loop, which improves performance slightly. Reviewed-by: Szabolcs Nagy (cherry picked from commit 09ebd8549b2ce5a3a6c0c7c5f3e62227faf50a99) --- sysdeps/aarch64/strchrnul.S | 12 ++++++++++-- 1 file changed, 10 insertions(+), 2 deletions(-) diff --git a/sysdeps/aarch64/strchrnul.S b/sysdeps/aarch64/strchrnul.S index 9de680b63e..9856fc5887 100644 --- a/sysdeps/aarch64/strchrnul.S +++ b/sysdeps/aarch64/strchrnul.S @@ -70,14 +70,22 @@ ENTRY (__strchrnul) .p2align 4 L(loop): - ldr qdata, [src, 16]! + ldr qdata, [src, 16] + cmeq vhas_chr.16b, vdata.16b, vrepchr.16b + cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b + umaxp vend.16b, vhas_chr.16b, vhas_chr.16b + fmov tmp1, dend + cbnz tmp1, L(end) + ldr qdata, [src, 32]! cmeq vhas_chr.16b, vdata.16b, vrepchr.16b cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b umaxp vend.16b, vhas_chr.16b, vhas_chr.16b fmov tmp1, dend cbz tmp1, L(loop) - + sub src, src, 16 +L(end): shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */ + add src, src, 16 fmov tmp1, dend #ifndef __AARCH64EB__ rbit tmp1, tmp1 -- 2.47.2