[apple/libc.git] / x86_64 / string / strlen.s

/*
 * Copyright (c) 2005-2012 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 * 
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 * 
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 *
 * This file implements strlen( ) for the x86_64 architecture.
 */

.globl _strlen

/*****************************************************************************
 *  Macros                                                                   *
 *****************************************************************************/

.macro EstablishFrame
	push      %rbp
	mov       %rsp,      %rbp
.endm

.macro ClearFrameAndReturn
	pop       %rbp
	ret
.endm

/*****************************************************************************
 *  Entrypoint                                                               *
 *****************************************************************************/

.text
.align 5
_strlen:
//	size_t strlen(const char *s);
//
//	returns the length of the string s (i.e. the distance in bytes from
//	s to the first NUL byte following s).  We look for NUL bytes using
//	pcmpeqb on 16-byte aligned blocks.  Although this may read past the
//	end of the string, because all access is aligned, it will never
//	read past the end of the string across a page boundary, or even
//	accross a cacheline.
	EstablishFrame
	mov       %rdi,	     %rcx
	mov       %rdi,      %rdx

//	Load the 16-byte block containing the first byte of the string, and
//	compare each byte to zero.  If any NUL bytes are present in this
//	block, the corresponding *bit* in esi will be set to 1.
	and       $-16,      %rdi
	pxor      %xmm0,     %xmm0
	pcmpeqb  (%rdi),     %xmm0
	pmovmskb  %xmm0,     %esi

//	The 16 bytes that we checked for NUL included some bytes preceeding
//	the start of the string, if s is not 16-byte aligned.  We create a
//	mask based on the alignment of s which covers only those bits
//	corresponding to bytes that do not preceed s, and check for NULs
//	only in those bits.  If we do not find one, we jump to our main
//	search loop.
	and       $0xf,      %rcx
	or        $-1,       %rax
	shl       %cl,       %rax
	and       %eax,      %esi
	jz        L_loop

L_foundNUL:
//	The last 16-byte block that we searched contained at least one NUL.
//	We use bsf to identify the first NUL, and compute the distance from
//	that byte to the start of the string.
	bsf       %esi,      %eax
	sub       %rdx,      %rdi
	add       %rdi,      %rax
	ClearFrameAndReturn

.align 4
L_loop:
//	Main search loop: check for NUL in a 16-byte block, continuing
//	loop until one is found.
	add       $16,       %rdi
	pxor      %xmm0,     %xmm0
	pcmpeqb  (%rdi),     %xmm0
	pmovmskb  %xmm0,     %esi
	test      %esi,      %esi
	jz        L_loop
	jmp       L_foundNUL
Commit	Line	Data
8e029c65	1	/*
6465356a	2	* Copyright (c) 2005-2012 Apple Computer, Inc. All rights reserved.
8e029c65	3	*
6465356a A	4	* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
	5	*
	6	* This file contains Original Code and/or Modifications of Original Code
	7	* as defined in and that are subject to the Apple Public Source License
	8	* Version 2.0 (the 'License'). You may not use this file except in
	9	* compliance with the License. The rights granted to you under the License
	10	* may not be used to create, or enable the creation or redistribution of,
	11	* unlawful or unlicensed copies of an Apple operating system, or to
	12	* circumvent, violate, or enable the circumvention or violation of, any
	13	* terms of an Apple operating system software license agreement.
	14	*
	15	* Please obtain a copy of the License at
	16	* http://www.opensource.apple.com/apsl/ and read it before using this file.
	17	*
	18	* The Original Code and all software distributed under the License are
	19	* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
8e029c65 A	20	* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
8e029c65 A	21	* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
6465356a A	22	* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
	23	* Please see the License for the specific language governing rights and
	24	* limitations under the License.
	25	*
	26	* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
8e029c65	27	*
6465356a	28	* This file implements strlen( ) for the x86_64 architecture.
8e029c65 A	29	*/
8e029c65 A	30
6465356a	31	.globl _strlen
8e029c65	32
6465356a A	33	/*****************************************************************************
	34	* Macros *
	35	*****************************************************************************/
8e029c65	36
6465356a A	37	.macro EstablishFrame
	38	push %rbp
	39	mov %rsp, %rbp
	40	.endm
	41
	42	.macro ClearFrameAndReturn
	43	pop %rbp
8e029c65	44	ret
6465356a A	45	.endm
	46
	47	/*****************************************************************************
	48	* Entrypoint *
	49	*****************************************************************************/
	50
	51	.text
	52	.align 5
	53	_strlen:
	54	// size_t strlen(const char *s);
	55	//
	56	// returns the length of the string s (i.e. the distance in bytes from
	57	// s to the first NUL byte following s). We look for NUL bytes using
	58	// pcmpeqb on 16-byte aligned blocks. Although this may read past the
	59	// end of the string, because all access is aligned, it will never
	60	// read past the end of the string across a page boundary, or even
	61	// accross a cacheline.
	62	EstablishFrame
	63	mov %rdi, %rcx
	64	mov %rdi, %rdx
	65
	66	// Load the 16-byte block containing the first byte of the string, and
	67	// compare each byte to zero. If any NUL bytes are present in this
	68	// block, the corresponding bit in esi will be set to 1.
	69	and $-16, %rdi
	70	pxor %xmm0, %xmm0
	71	pcmpeqb (%rdi), %xmm0
	72	pmovmskb %xmm0, %esi
	73
	74	// The 16 bytes that we checked for NUL included some bytes preceeding
	75	// the start of the string, if s is not 16-byte aligned. We create a
	76	// mask based on the alignment of s which covers only those bits
	77	// corresponding to bytes that do not preceed s, and check for NULs
	78	// only in those bits. If we do not find one, we jump to our main
	79	// search loop.
	80	and $0xf, %rcx
	81	or $-1, %rax
	82	shl %cl, %rax
	83	and %eax, %esi
	84	jz L_loop
8e029c65	85
6465356a A	86	L_foundNUL:
	87	// The last 16-byte block that we searched contained at least one NUL.
	88	// We use bsf to identify the first NUL, and compute the distance from
	89	// that byte to the start of the string.
	90	bsf %esi, %eax
	91	sub %rdx, %rdi
	92	add %rdi, %rax
	93	ClearFrameAndReturn
8e029c65	94
6465356a A	95	.align 4
	96	L_loop:
	97	// Main search loop: check for NUL in a 16-byte block, continuing
	98	// loop until one is found.
	99	add $16, %rdi
	100	pxor %xmm0, %xmm0
	101	pcmpeqb (%rdi), %xmm0
	102	pmovmskb %xmm0, %esi
	103	test %esi, %esi
	104	jz L_loop
	105	jmp L_foundNUL