diff options
| author | Wilco Dijkstra <wilco.dijkstra@arm.com> | 2023-01-11 13:52:23 +0000 |
|---|---|---|
| committer | Wilco Dijkstra <wilco.dijkstra@arm.com> | 2023-01-17 15:09:18 +0000 |
| commit | 09ebd8549b2ce5a3a6c0c7c5f3e62227faf50a99 (patch) | |
| tree | 622dd725d9427e8afc3af67c02bb0eff7bf48810 /sysdeps/aarch64 | |
| parent | 51541a229740801882490177fa178e49264b13fb (diff) | |
| download | glibc-09ebd8549b2ce5a3a6c0c7c5f3e62227faf50a99.zip glibc-09ebd8549b2ce5a3a6c0c7c5f3e62227faf50a99.tar.gz glibc-09ebd8549b2ce5a3a6c0c7c5f3e62227faf50a99.tar.bz2 | |
AArch64: Improve strchrnul
Unroll the main loop, which improves performance slightly.
Reviewed-by: Szabolcs Nagy <szabolcs.nagy@arm.com>
Diffstat (limited to 'sysdeps/aarch64')
| -rw-r--r-- | sysdeps/aarch64/strchrnul.S | 12 |
1 files changed, 10 insertions, 2 deletions
diff --git a/sysdeps/aarch64/strchrnul.S b/sysdeps/aarch64/strchrnul.S index 4ca1e58..aa8c9a4 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 |