glibc/sysdeps/aarch64/strchrnul.S
Danila Kutenin 3c99806989 aarch64: Optimize string functions with shrn instruction
We found that string functions were using AND+ADDP
to find the nibble/syndrome mask but there is an easier
opportunity through `SHRN dst.8b, src.8h, 4` (shift
right every 2 bytes by 4 and narrow to 1 byte) and has
same latency on all SIMD ARMv8 targets as ADDP. There
are also possible gaps for memcmp but that's for
another patch.

We see 10-20% savings for small-mid size cases (<=128)
which are primary cases for general workloads.
2022-07-06 09:26:20 +01:00

91 lines
2.3 KiB
ArmAsm

/* strchrnul - find a character or nul in a string
Copyright (C) 2014-2022 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library. If not, see
<https://www.gnu.org/licenses/>. */
#include <sysdep.h>
/* Assumptions:
*
* ARMv8-a, AArch64, Advanced SIMD.
* MTE compatible.
*/
#define srcin x0
#define chrin w1
#define result x0
#define src x2
#define tmp1 x1
#define tmp2 x3
#define vrepchr v0
#define vdata v1
#define qdata q1
#define vhas_nul v2
#define vhas_chr v3
#define vend v4
#define dend d4
/*
Core algorithm:
For each 16-byte chunk we calculate a 64-bit nibble mask value with four bits
per byte. We take 4 bits of every comparison byte with shift right and narrow
by 4 instruction. Since the bits in the nibble mask reflect the order in
which things occur in the original string, counting leading zeros identifies
exactly which byte matched. */
ENTRY (__strchrnul)
PTR_ARG (0)
bic src, srcin, 15
dup vrepchr.16b, chrin
ld1 {vdata.16b}, [src]
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
cmhs vhas_chr.16b, vhas_chr.16b, vdata.16b
lsl tmp2, srcin, 2
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov tmp1, dend
lsr tmp1, tmp1, tmp2 /* Mask padding bits. */
cbz tmp1, L(loop)
rbit tmp1, tmp1
clz tmp1, tmp1
add result, srcin, tmp1, lsr 2
ret
.p2align 4
L(loop):
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
cbz tmp1, L(loop)
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov tmp1, dend
#ifndef __AARCH64EB__
rbit tmp1, tmp1
#endif
clz tmp1, tmp1
add result, src, tmp1, lsr 2
ret
END(__strchrnul)
weak_alias (__strchrnul, strchrnul)