arm64/lib/strlen.S

1 /* SPDX-License-Identifier: GPL-2.0-only */
3  * Copyright (c) 2013-2021, Arm Limited.
6  * https://github.com/ARM-software/optimized-routines/blob/98e4d6a5c13c8e54/string/aarch64/strlen.S
11 #include <asm/mte-def.h>
15  * ARMv8-a, AArch64, unaligned accesses, min page size 4k.
36 	/* NUL detection works on the principle that (X - 1) & (~X) & 0x80
37 	   (=> (X - 1) & ~(X | 0x7f)) is non-zero iff a byte is zero, and
39 	   (X - 1) & 0x80 is zero for non-NUL ASCII characters, but gives
48  * (16-byte) granularity, and we must ensure that no access straddles this
58 	   of the string for a NUL character.  In order to do an unaligned ldp
60 	   byte we calculate the length from the 2 8-byte words using
64 	   If the string is longer than 16 bytes, we align src so don't need
65 	   further page cross checks, and process 32 bytes per iteration
66 	   using the fast NUL check.  If we encounter non-ASCII characters,
70 	   address, remove any characters before the string, and continue
76 	   checking for larger page sizes - the cost of setting up the correct
83 	and	tmp1, srcin, MIN_PAGE_SIZE - 1
85 	cmp	tmp1, MIN_PAGE_SIZE - 16
89 	/* For big-endian, carry propagation (if the final byte in the
90 	   string is 0x01) means we cannot use has_nul1/2 directly.
91 	   Since we expect strings to be small and early-exit,
92 	   byte-swap the data now so has_null1/2 will be correct.  */
114 	/* The inner loop processes 32 bytes per iteration and uses the fast
115 	   NUL check.  If we encounter non-ASCII characters, use a second
148 	/* For big-endian, carry propagation (if the final byte in the
149 	   string is 0x01) means we cannot use has_nul1/2 directly.  The
150 	   easiest way to get the correct byte is to byte-swap the data
190 	   srcin to 0x7f, so we ignore any NUL bytes before the string.
196 	mov	tmp4, -1
198 	/* Big-endian.	Early bytes are at MSB.	 */
201 	/* Little-endian.  Early bytes are at LSB.  */