glibc/sysdeps/aarch64/memrchr.S
Adhemerval Zanella 22999b2f0f string: Add libc_hidden_proto for memrchr
Although static linker can optimize it to local call, it follows the
internal scheme to provide hidden proto and definitions.

Reviewed-by: Carlos Eduardo Seo <carlos.seo@linaro.org>
2023-02-08 17:13:58 -03:00

129 lines
3.0 KiB
ArmAsm

/* memrchr - find the last occurrence of a byte in a memory block
Copyright (C) 2015-2023 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 cntin x2
#define result x0
#define src x3
#define cntrem x4
#define synd x5
#define shift x6
#define tmp x7
#define end x8
#define endm1 x9
#define vrepchr v0
#define qdata q1
#define vdata v1
#define vhas_chr v2
#define vend v3
#define dend d3
/*
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 (__memrchr)
PTR_ARG (0)
SIZE_ARG (2)
add end, srcin, cntin
sub endm1, end, 1
bic src, endm1, 15
cbz cntin, L(nomatch)
ld1 {vdata.16b}, [src]
dup vrepchr.16b, chrin
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
neg shift, end, lsl 2
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
lsl synd, synd, shift
cbz synd, L(start_loop)
clz synd, synd
sub result, endm1, synd, lsr 2
cmp cntin, synd, lsr 2
csel result, result, xzr, hi
ret
nop
L(start_loop):
subs cntrem, src, srcin
b.ls L(nomatch)
/* Make sure that it won't overread by a 16-byte chunk */
sub cntrem, cntrem, 1
tbz cntrem, 4, L(loop32_2)
add src, src, 16
.p2align 5
L(loop32):
ldr qdata, [src, -32]!
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
umaxp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
fmov synd, dend
cbnz synd, L(end)
L(loop32_2):
ldr qdata, [src, -16]
subs cntrem, cntrem, 32
cmeq vhas_chr.16b, vdata.16b, vrepchr.16b
b.lo L(end_2)
umaxp vend.16b, vhas_chr.16b, vhas_chr.16b /* 128->64 */
fmov synd, dend
cbz synd, L(loop32)
L(end_2):
sub src, src, 16
L(end):
shrn vend.8b, vhas_chr.8h, 4 /* 128->64 */
fmov synd, dend
add tmp, src, 15
#ifdef __AARCH64EB__
rbit synd, synd
#endif
clz synd, synd
sub tmp, tmp, synd, lsr 2
cmp tmp, srcin
csel result, tmp, xzr, hs
ret
L(nomatch):
mov result, 0
ret
END (__memrchr)
libc_hidden_def (__memrchr)
weak_alias (__memrchr, memrchr)
libc_hidden_builtin_def (memrchr)