nuttx-update/libs/libc/machine/xtensa/arch_strcpy.S
anjiahao 603c87977f libc/machine:Add prefixes to libc functions implemented by arch
kasan does not instrument the assembly function,
so it is checked in advance before calling.
If Kasan is not turned on, the speed and space will
be almost unchanged under compiler optimization.

Signed-off-by: anjiahao <anjiahao@xiaomi.com>
2024-10-31 18:11:16 +08:00

250 lines
6.4 KiB
ArmAsm

/****************************************************************************
* libs/libc/machine/xtensa/arch_strcpy.S
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership. The
* ASF licenses this file to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance with the
* License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
****************************************************************************/
/****************************************************************************
* Included Files
****************************************************************************/
#include "xtensa_asm.h"
#include <arch/chip/core-isa.h>
#include <arch/xtensa/xtensa_abi.h>
#include "libc.h"
#ifdef LIBC_BUILD_STRCPY
/****************************************************************************
* Public Functions
****************************************************************************/
.section .text
.begin schedule
.align 4
.literal_position
.global ARCH_LIBCFUN(strcpy)
.type ARCH_LIBCFUN(strcpy), @function
ARCH_LIBCFUN(strcpy):
ENTRY(16)
/* a2 = dst, a3 = src */
mov a10, a2 # leave dst in return value register
movi a4, MASK0
movi a5, MASK1
movi a6, MASK2
movi a7, MASK3
bbsi.l a3, 0, .Lsrc1mod2
bbsi.l a3, 1, .Lsrc2mod4
.Lsrcaligned:
/* Check if the destination is aligned. */
movi a8, 3
bnone a10, a8, .Laligned
j .Ldstunaligned
.Lsrc1mod2: # src address is odd
l8ui a8, a3, 0 # get byte 0
addi a3, a3, 1 # advance src pointer
s8i a8, a10, 0 # store byte 0
beqz a8, 1f # if byte 0 is zero
addi a10, a10, 1 # advance dst pointer
bbci.l a3, 1, .Lsrcaligned # if src is now word-aligned
.Lsrc2mod4: # src address is 2 mod 4
l8ui a8, a3, 0 # get byte 0
/* 1-cycle interlock */
s8i a8, a10, 0 # store byte 0
beqz a8, 1f # if byte 0 is zero
l8ui a8, a3, 1 # get byte 0
addi a3, a3, 2 # advance src pointer
s8i a8, a10, 1 # store byte 0
addi a10, a10, 2 # advance dst pointer
bnez a8, .Lsrcaligned
1: RET(16)
/* dst is word-aligned; src is word-aligned. */
.align 4
#if XCHAL_HAVE_LOOPS
#if XCHAL_HAVE_DENSITY
/* (2 mod 4) alignment for loop instruction */
#else
/* (1 mod 4) alignment for loop instruction */
.byte 0
.byte 0
#endif
.Laligned:
#if XCHAL_HAVE_DENSITY
_movi.n a8, 0 # set up for the maximum loop count
#else
_movi a8, 0 # set up for the maximum loop count
#endif
loop a8, .Lz3 # loop forever (almost anyway)
l32i a8, a3, 0 # get word from src
addi a3, a3, 4 # advance src pointer
bnone a8, a4, .Lz0 # if byte 0 is zero
bnone a8, a5, .Lz1 # if byte 1 is zero
bnone a8, a6, .Lz2 # if byte 2 is zero
s32i a8, a10, 0 # store word to dst
bnone a8, a7, .Lz3 # if byte 3 is zero
addi a10, a10, 4 # advance dst pointer
#else /* !XCHAL_HAVE_LOOPS */
1: addi a10, a10, 4 # advance dst pointer
.Laligned:
l32i a8, a3, 0 # get word from src
addi a3, a3, 4 # advance src pointer
bnone a8, a4, .Lz0 # if byte 0 is zero
bnone a8, a5, .Lz1 # if byte 1 is zero
bnone a8, a6, .Lz2 # if byte 2 is zero
s32i a8, a10, 0 # store word to dst
bany a8, a7, 1b # if byte 3 is zero
#endif /* !XCHAL_HAVE_LOOPS */
.Lz3: /* Byte 3 is zero. */
RET(16)
.Lz0: /* Byte 0 is zero. */
#if XCHAL_HAVE_BE
movi a8, 0
#endif
s8i a8, a10, 0
RET(16)
.Lz1: /* Byte 1 is zero. */
#if XCHAL_HAVE_BE
extui a8, a8, 16, 16
#endif
s16i a8, a10, 0
RET(16)
.Lz2: /* Byte 2 is zero. */
#if XCHAL_HAVE_BE
extui a8, a8, 16, 16
#endif
s16i a8, a10, 0
movi a8, 0
s8i a8, a10, 2
RET(16)
#if 1
/* For now just use byte copy loop for the unaligned destination case. */
.align 4
#if XCHAL_HAVE_LOOPS
#if XCHAL_HAVE_DENSITY
/* (2 mod 4) alignment for loop instruction */
#else
/* (1 mod 4) alignment for loop instruction */
.byte 0
.byte 0
#endif
#endif
.Ldstunaligned:
#if XCHAL_HAVE_LOOPS
#if XCHAL_HAVE_DENSITY
_movi.n a8, 0 # set up for the maximum loop count
#else
_movi a8, 0 # set up for the maximum loop count
#endif
loop a8, 2f # loop forever (almost anyway)
#endif
1: l8ui a8, a3, 0
addi a3, a3, 1
s8i a8, a10, 0
addi a10, a10, 1
#if XCHAL_HAVE_LOOPS
beqz a8, 2f
#else
bnez a8, 1b
#endif
2: RET(16)
#else /* 0 */
/* This code is not functional yet. */
.Ldstunaligned:
l32i a9, a2, 0 # load word from dst
#if XCHAL_HAVE_BE
ssa8b a9 # rotate by dst alignment so that
src a9, a9, a9 # shift in loop will put back in place
ssa8l a9 # shift left by byte*8
#else
ssa8l a9 # rotate by dst alignment so that
src a9, a9, a9 # shift in loop will put back in place
ssa8b a9 # shift left by 32-byte*8
#endif
/* dst is word-aligned; src is unaligned. */
.Ldstunalignedloop:
l32i a8, a3, 0 # get word from src
/* 1-cycle interlock */
bnone a8, a4, .Lu0 # if byte 0 is zero
bnone a8, a5, .Lu1 # if byte 1 is zero
bnone a8, a6, .Lu2 # if byte 2 is zero
src a9, a8, a9 # combine last word and this word
s32i a9, a10, 0 # store word to dst
bnone a8, a7, .Lu3 # if byte 3 is nonzero, iterate
l32i a9, a3, 4 # get word from src
addi a3, a3, 8 # advance src pointer
bnone a9, a4, .Lu4 # if byte 0 is zero
bnone a9, a5, .Lu5 # if byte 1 is zero
bnone a9, a6, .Lu6 # if byte 2 is zero
src a8, a9, a8 # combine last word and this word
s32i a8, a10, 4 # store word to dst
addi a10, a10, 8 # advance dst pointer
bany a8, a7, .Ldstunalignedloop # if byte 3 is nonzero, iterate
/* Byte 7 is zero. */
.Lu7: RET(16)
.Lu0: /* Byte 0 is zero. */
#if XCHAL_HAVE_BE
movi a8, 0
#endif
s8i a8, a10, 0
RET(16)
.Lu1: /* Byte 1 is zero. */
#if XCHAL_HAVE_BE
extui a8, a8, 16, 16
#endif
s16i a8, a10, 0
RET(16)
.Lu2: /* Byte 2 is zero. */
s16i a8, a10, 0
movi a8, 0
s8i a8, a10, 2
RET(16)
#endif /* 0 */
.end schedule
.size ARCH_LIBCFUN(strcpy), . - ARCH_LIBCFUN(strcpy)
#endif