merge x86_64 optimized string support

16 files changed, 2875 insertions, 15 deletions

diff --git a/libc/string/Makefile b/libc/string/Makefile
index 600f3b81d..576f915af 100644
--- a/libc/string/Makefile
+++ b/libc/string/Makefile
@@ -1,20 +1,9 @@
 # Makefile for uClibc
 #
-# Copyright (C) 2000-2003 Erik Andersen <andersen@uclibc.org>
+# Copyright (C) 2000-2005 Erik Andersen <andersen@uclibc.org>
 #
-# This program is free software; you can redistribute it and/or modify it under
-# the terms of the GNU Library General Public License as published by the Free
-# Software Foundation; either version 2 of the License, or (at your option) any
-# later version.
-#
-# This program 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 Library General Public License for more
-# details.
-#
-# You should have received a copy of the GNU Library General Public License
-# along with this program; if not, write to the Free Software Foundation, Inc.,
-# 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+# Licensed under the GNU Library General Public License version 2 or later.
+# See the COPYING.LIB file in the toplevel for more information.
 
 TOPDIR=../../
 include $(TOPDIR)Rules.mak
@@ -31,7 +20,7 @@ DIRS += $(TARGET_ARCH)
 endif
 endif
 
-ALL_SUBDIRS = generic arm frv i386 mips powerpc sh64 sparc
+ALL_SUBDIRS = generic arm frv i386 mips powerpc sh64 sparc x86_64
 
 MSRC= wstring.c
 MOBJ=  basename.o bcopy.o bzero.o dirname.o ffs.o memccpy.o memchr.o memcmp.o \
diff --git a/libc/string/x86_64/Makefile b/libc/string/x86_64/Makefile
new file mode 100644
index 000000000..2215a6025
--- /dev/null
+++ b/libc/string/x86_64/Makefile
@@ -0,0 +1,31 @@
+# Makefile for uClibc
+#
+# Copyright (C) 2000-2005 Erik Andersen <andersen@uclibc.org>
+#
+# Licensed under the GNU Library General Public License version 2 or later.
+# See the COPYING.LIB file in the toplevel for more information.
+
+TOPDIR=../../../
+include $(TOPDIR)Rules.mak
+
+CSRCS = $(wildcard *.c)
+COBJS = $(patsubst %.c,%.o,$(CSRCS))
+
+SSRCS = $(wildcard *.S)
+SOBJS = $(patsubst %.S,%.o,$(SSRCS))
+
+OBJS = $(COBJS) $(SOBJS)
+
+OBJ_LIST = ../../obj.string.$(TARGET_ARCH)
+
+all: $(OBJ_LIST)
+
+$(OBJ_LIST): $(OBJS)
+	echo $(patsubst %, string/$(TARGET_ARCH)/%, $(OBJS)) > $(OBJ_LIST)
+
+$(COBJS): %.o : %.c
+	$(CC) $(CFLAGS) -c $< -o $@
+	$(STRIPTOOL) -x -R .note -R .comment $*.o
+
+clean:
+	$(RM) *.[oa] *~ core
diff --git a/libc/string/x86_64/_glibc_inc.h b/libc/string/x86_64/_glibc_inc.h
new file mode 100644
index 000000000..f14b23c94
--- /dev/null
+++ b/libc/string/x86_64/_glibc_inc.h
@@ -0,0 +1,33 @@
+/*
+ * Setup some glibc defines so we can just drop in the
+ * asm files from glibc without any modification.
+ */
+
+#include <features.h>
+#include <bits/wordsize.h>
+
+#if __WORDSIZE == 32
+# define ENTRY_ALIGN 4
+#else
+# define ENTRY_ALIGN 2
+#endif
+
+#define ENTRY(sym) \
+	.global sym; \
+	.align  ENTRY_ALIGN; \
+	.type   sym,%function; \
+	sym:
+
+#define BP_SYM(sym) sym
+
+#define L(sym) LOC(sym)
+#define LOC(sym) \
+	.L ## sym
+
+#define END(sym) \
+	.size sym,.-sym;
+
+#undef weak_alias
+#define weak_alias(sym, alias) \
+	.weak alias; \
+    alias = sym;
diff --git a/libc/string/x86_64/bzero.S b/libc/string/x86_64/bzero.S
new file mode 100644
index 000000000..abd252e7b
--- /dev/null
+++ b/libc/string/x86_64/bzero.S
@@ -0,0 +1,3 @@
+#define memset __bzero
+#include "memset.S"
+weak_alias (__bzero, bzero)
diff --git a/libc/string/x86_64/memcpy.S b/libc/string/x86_64/memcpy.S
new file mode 100644
index 000000000..4fa38a640
--- /dev/null
+++ b/libc/string/x86_64/memcpy.S
@@ -0,0 +1,95 @@
+/* Highly optimized version for x86-64.
+   Copyright (C) 1997, 2000, 2002, 2003, 2004 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Based on i586 version contributed by Ulrich Drepper <drepper@cygnus.com>, 1997.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+/* BEWARE: `#ifdef memcpy' means that memcpy is redefined as `mempcpy',
+   and the return value is the byte after the last one copied in
+   the destination. */
+#define MEMPCPY_P (defined memcpy)
+
+        .text
+#if defined PIC && !defined NOT_IN_libc
+ENTRY (__memcpy_chk)
+	cmpq	%rdx, %rcx
+	jb	HIDDEN_JUMPTARGET (__chk_fail)
+END (__memcpy_chk)
+#endif
+ENTRY (BP_SYM (memcpy))
+	/* Cutoff for the big loop is a size of 32 bytes since otherwise
+	   the loop will never be entered.  */
+	cmpq	$32, %rdx
+	movq	%rdx, %rcx
+#if !MEMPCPY_P
+	movq	%rdi, %r10	/* Save value. */
+#endif
+
+	/* We need this in any case.  */
+	cld
+
+	jbe	1f
+
+	/* Align destination.  */
+	movq	%rdi, %rax
+	negq	%rax
+	andq	$7, %rax
+	subq	%rax, %rcx
+	xchgq	%rax, %rcx
+
+	rep; movsb
+
+	movq	%rax, %rcx
+	subq	$32, %rcx
+	js	2f
+
+	.p2align 4
+3:
+
+	/* Now correct the loop counter.  Please note that in the following
+	   code the flags are not changed anymore.  */
+	subq	$32, %rcx
+
+	movq	(%rsi), %rax
+	movq	8(%rsi), %rdx
+	movq	16(%rsi), %r8
+	movq	24(%rsi), %r9
+	movq	%rax, (%rdi)
+	movq	%rdx, 8(%rdi)
+	movq	%r8, 16(%rdi)
+	movq	%r9, 24(%rdi)
+
+	leaq	32(%rsi), %rsi
+	leaq	32(%rdi), %rdi
+
+	jns	3b
+
+	/* Correct extra loop counter modification.  */
+2:	addq	$32, %rcx
+1:	rep; movsb
+
+#if MEMPCPY_P
+	movq	%rdi, %rax		/* Set return value.  */
+#else
+	movq	%r10, %rax		/* Set return value.  */
+	
+#endif
+	ret
+
+END (BP_SYM (memcpy))
diff --git a/libc/string/x86_64/memset.S b/libc/string/x86_64/memset.S
new file mode 100644
index 000000000..d74ec8ccb
--- /dev/null
+++ b/libc/string/x86_64/memset.S
@@ -0,0 +1,138 @@
+/* memset/bzero -- set memory area to CH/0
+   Optimized version for x86-64.
+   Copyright (C) 2002, 2003, 2004, 2005 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Andreas Jaeger <aj@suse.de>.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+/* BEWARE: `#ifdef memset' means that memset is redefined as `bzero' */
+#define BZERO_P (defined memset)
+
+/* This is somehow experimental and could made dependend on the cache
+   size.  */
+#define LARGE $120000
+
+        .text
+#if !BZERO_P && defined PIC && !defined NOT_IN_libc
+ENTRY (__memset_chk)
+	cmpq	%rdx, %rcx
+	jb	HIDDEN_JUMPTARGET (__chk_fail)
+END (__memset_chk)
+#endif
+ENTRY (memset)
+#if BZERO_P
+	mov	%rsi,%rdx	/* Adjust parameter.  */
+	xorl	%esi,%esi	/* Fill with 0s.  */
+#endif
+	cmp	$0x7,%rdx	/* Check for small length.  */
+	mov	%rdi,%rcx	/* Save ptr as return value.  */
+	jbe	7f
+
+#if BZERO_P
+	mov	%rsi,%r8	/* Just copy 0.  */
+#else
+	/* Populate 8 bit data to full 64-bit.  */
+	movabs	$0x0101010101010101,%r8
+	movzbl	%sil,%eax
+	imul	%rax,%r8
+#endif
+	test	$0x7,%edi	/* Check for alignment.  */
+	je	2f
+
+	.p2align 4
+1:	/* Align ptr to 8 byte.  */
+	mov	%sil,(%rcx)
+	dec	%rdx
+	inc	%rcx
+	test	$0x7,%ecx
+	jne	1b
+
+2:	/* Check for really large regions.  */
+	mov	%rdx,%rax
+	shr	$0x6,%rax
+	je	4f
+	cmp	LARGE, %rdx
+	jae	11f
+
+	.p2align 4
+3:	/* Copy 64 bytes.  */
+	mov	%r8,(%rcx)
+	mov	%r8,0x8(%rcx)
+	mov	%r8,0x10(%rcx)
+	mov	%r8,0x18(%rcx)
+	mov	%r8,0x20(%rcx)
+	mov	%r8,0x28(%rcx)
+	mov	%r8,0x30(%rcx)
+	mov	%r8,0x38(%rcx)
+	add	$0x40,%rcx
+	dec	%rax
+	jne	3b
+
+4:	/* Copy final bytes.  */
+	and	$0x3f,%edx
+	mov	%rdx,%rax
+	shr	$0x3,%rax
+	je	6f
+
+5:	/* First in chunks of 8 bytes.  */
+	mov	%r8,(%rcx)
+	add	$0x8,%rcx
+	dec	%rax
+	jne	5b
+6:
+	and	$0x7,%edx
+7:
+	test	%rdx,%rdx
+	je	9f
+8:	/* And finally as bytes (up to 7).  */
+	mov	%sil,(%rcx)
+	inc	%rcx
+	dec	%rdx
+	jne	8b
+9:
+#if BZERO_P
+	nop
+#else
+	/* Load result (only if used as memset).  */
+	mov	%rdi,%rax	/* start address of destination is result */
+#endif
+	retq
+
+	.p2align 4
+11:	/* Copy 64 bytes without polluting the cache.  */
+	/* We could use	movntdq    %xmm0,(%rcx) here to further
+	   speed up for large cases but let's not use XMM registers.  */
+	movnti	%r8,(%rcx)
+	movnti  %r8,0x8(%rcx)
+	movnti  %r8,0x10(%rcx)
+	movnti  %r8,0x18(%rcx)
+	movnti  %r8,0x20(%rcx)
+	movnti  %r8,0x28(%rcx)
+	movnti  %r8,0x30(%rcx)
+	movnti  %r8,0x38(%rcx)
+	add	$0x40,%rcx
+	dec	%rax
+	jne	11b
+	jmp	4b
+
+END (memset)
+
+#if !BZERO_P && defined PIC && !defined NOT_IN_libc
+strong_alias (__memset_chk, __memset_zero_constant_len_parameter)
+#endif
diff --git a/libc/string/x86_64/stpcpy.S b/libc/string/x86_64/stpcpy.S
new file mode 100644
index 000000000..83294e1a8
--- /dev/null
+++ b/libc/string/x86_64/stpcpy.S
@@ -0,0 +1,6 @@
+#define USE_AS_STPCPY
+#define STRCPY __stpcpy
+
+#include "strcpy.S"
+
+weak_alias (__stpcpy, stpcpy)
diff --git a/libc/string/x86_64/strcat.S b/libc/string/x86_64/strcat.S
new file mode 100644
index 000000000..1106cb4ed
--- /dev/null
+++ b/libc/string/x86_64/strcat.S
@@ -0,0 +1,256 @@
+/* strcat(dest, src) -- Append SRC on the end of DEST.
+   Optimized for x86-64.
+   Copyright (C) 2002 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Andreas Jaeger <aj@suse.de>, 2002.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+
+	.text
+ENTRY (BP_SYM (strcat))
+	movq %rdi, %rcx		/* Dest. register. */
+	andl $7, %ecx		/* mask alignment bits */
+	movq %rdi, %rax		/* Duplicate destination pointer.  */
+	movq $0xfefefefefefefeff,%r8
+
+	/* First step: Find end of destination.  */
+	jz 4f			/* aligned => start loop */
+
+	neg %ecx		/* We need to align to 8 bytes.  */
+	addl $8,%ecx
+	/* Search the first bytes directly.  */
+0:	cmpb $0x0,(%rax)	/* is byte NUL? */
+	je 2f			/* yes => start copy */
+	incq %rax		/* increment pointer */
+	decl %ecx
+	jnz 0b
+
+
+
+	/* Now the source is aligned.  Scan for NUL byte.  */
+	.p2align 4
+4:
+	/* First unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found NUL => return pointer */
+
+	/* Second unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found NUL => return pointer */
+
+	/* Third unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found NUL => return pointer */
+
+	/* Fourth unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jz 4b			/* no NUL found => continue loop */
+
+	.p2align 4		/* Align, it's a jump target.  */
+3:	subq $8,%rax		/* correct pointer increment.  */
+
+	testb %cl, %cl		/* is first byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is second byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testl $0x00ff0000, %ecx /* is third byte NUL? */
+	jz 2f			/* yes => return pointer */
+	incq %rax		/* increment pointer */
+
+	testl $0xff000000, %ecx /* is fourth byte NUL? */
+	jz 2f			/* yes => return pointer */
+	incq %rax		/* increment pointer */
+
+	shrq $32, %rcx		/* look at other half.  */
+
+	testb %cl, %cl		/* is first byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is second byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testl $0xff0000, %ecx	/* is third byte NUL? */
+	jz 2f			/* yes => return pointer */
+	incq %rax		/* increment pointer */
+
+2:
+	/* Second step: Copy source to destination.  */
+
+	movq	%rsi, %rcx	/* duplicate  */
+	andl	$7,%ecx		/* mask alignment bits */
+	movq	%rax, %rdx	/* move around */
+	jz	22f		/* aligned => start loop */
+
+	neg	%ecx		/* align to 8 bytes.  */
+	addl	$8, %ecx
+	/* Align the source pointer.  */
+21:
+	movb	(%rsi), %al	/* Fetch a byte */
+	testb	%al, %al	/* Is it NUL? */
+	movb	%al, (%rdx)	/* Store it */
+	jz	24f		/* If it was NUL, done! */
+	incq	%rsi
+	incq	%rdx
+	decl	%ecx
+	jnz	21b
+
+	/* Now the sources is aligned.  Unfortunatly we cannot force
+	   to have both source and destination aligned, so ignore the
+	   alignment of the destination.  */
+	.p2align 4
+22:
+	/* 1st unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	23f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	23f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+
+	/* 2nd unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	23f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	23f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+
+	/* 3rd unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	23f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	23f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+
+	/* 4th unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	23f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	23f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+	jmp	22b		/* Next iteration.  */
+
+	/* Do the last few bytes. %rax contains the value to write.
+	   The loop is unrolled twice.  */
+	.p2align 4
+23:
+	movb	%al, (%rdx)	/* 1st byte.  */
+	testb	%al, %al	/* Is it NUL.  */
+	jz	24f		/* yes, finish.  */
+	incq	%rdx		/* Increment destination.  */
+	movb	%ah, (%rdx)	/* 2nd byte.  */
+	testb	%ah, %ah	/* Is it NUL?.  */
+	jz	24f		/* yes, finish.  */
+	incq	%rdx		/* Increment destination.  */
+	shrq	$16, %rax	/* Shift...  */
+	jmp	23b		/* and look at next two bytes in %rax.  */
+
+
+24:
+	movq	%rdi, %rax	/* Source is return value.  */
+	retq
+END (BP_SYM (strcat))
diff --git a/libc/string/x86_64/strchr.S b/libc/string/x86_64/strchr.S
new file mode 100644
index 000000000..b84b012b2
--- /dev/null
+++ b/libc/string/x86_64/strchr.S
@@ -0,0 +1,287 @@
+/* strchr (str, ch) -- Return pointer to first occurrence of CH in STR.
+   For AMD x86-64.
+   Copyright (C) 2002, 2005 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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+
+	.text
+ENTRY (BP_SYM (strchr))
+
+	/* Before we start with the main loop we process single bytes
+	   until the source pointer is aligned.  This has two reasons:
+	   1. aligned 64-bit memory access is faster
+	   and (more important)
+	   2. we process in the main loop 64 bit in one step although
+	      we don't know the end of the string.  But accessing at
+	      8-byte alignment guarantees that we never access illegal
+	      memory if this would not also be done by the trivial
+	      implementation (this is because all processor inherent
+	      boundaries are multiples of 8).  */
+
+	movq	%rdi, %rdx
+	andl	$7, %edx	/* Mask alignment bits  */
+	movq	%rdi, %rax	/* duplicate destination.  */
+	jz	1f		/* aligned => start loop */
+	neg	%edx
+	addl	$8, %edx	/* Align to 8 bytes.  */
+
+	/* Search the first bytes directly.  */
+0:	movb	(%rax), %cl	/* load byte  */
+	cmpb	%cl,%sil	/* compare byte.  */
+	je	6f		/* target found */
+	testb	%cl,%cl		/* is byte NUL? */
+	je	7f		/* yes => return NULL */
+	incq	%rax		/* increment pointer */
+	decl	%edx
+	jnz	0b
+
+
+1:
+	/* At the moment %rsi contains C.  What we need for the
+	   algorithm is C in all bytes of the register.  Avoid
+	   operations on 16 bit words because these require an
+	   prefix byte (and one more cycle).  */
+	/* Populate 8 bit data to full 64-bit.  */
+	movabs	$0x0101010101010101,%r9
+	movzbl	%sil,%edx
+	imul	%rdx,%r9
+
+	movq $0xfefefefefefefeff, %r8 /* Save magic.  */
+
+      /* We exit the loop if adding MAGIC_BITS to LONGWORD fails to
+	 change any of the hole bits of LONGWORD.
+
+	 1) Is this safe?  Will it catch all the zero bytes?
+	 Suppose there is a byte with all zeros.  Any carry bits
+	 propagating from its left will fall into the hole at its
+	 least significant bit and stop.  Since there will be no
+	 carry from its most significant bit, the LSB of the
+	 byte to the left will be unchanged, and the zero will be
+	 detected.
+
+	 2) Is this worthwhile?  Will it ignore everything except
+	 zero bytes?  Suppose every byte of QUARDWORD has a bit set
+	 somewhere.  There will be a carry into bit 8.	If bit 8
+	 is set, this will carry into bit 16.  If bit 8 is clear,
+	 one of bits 9-15 must be set, so there will be a carry
+	 into bit 16.  Similarly, there will be a carry into bit
+	 24 tec..  If one of bits 54-63 is set, there will be a carry
+	 into bit 64 (=carry flag), so all of the hole bits will
+	 be changed.
+
+	 3) But wait!  Aren't we looking for C, not zero?
+	 Good point.  So what we do is XOR LONGWORD with a longword,
+	 each of whose bytes is C.  This turns each byte that is C
+	 into a zero.  */
+
+	.p2align 4
+4:
+	/* Main Loop is unrolled 4 times.  */
+	/* First unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	xorq %r9, %rcx		/* XOR with qword c|...|c => bytes of str == c
+				   are now 0 */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found c => return pointer */
+
+	/* The quadword we looked at does not contain the value we're looking
+	   for.  Let's search now whether we have reached the end of the
+	   string.  */
+	xorq %r9, %rcx		/* restore original dword without reload */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 7f			/* highest byte is NUL => return NULL */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 7f			/* found NUL => return NULL */
+
+	/* Second unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	xorq %r9, %rcx		/* XOR with qword c|...|c => bytes of str == c
+				   are now 0 */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found c => return pointer */
+
+	/* The quadword we looked at does not contain the value we're looking
+	   for.  Let's search now whether we have reached the end of the
+	   string.  */
+	xorq %r9, %rcx		/* restore original dword without reload */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 7f			/* highest byte is NUL => return NULL */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 7f			/* found NUL => return NULL */
+	/* Third unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	xorq %r9, %rcx		/* XOR with qword c|...|c => bytes of str == c
+				   are now 0 */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found c => return pointer */
+
+	/* The quadword we looked at does not contain the value we're looking
+	   for.  Let's search now whether we have reached the end of the
+	   string.  */
+	xorq %r9, %rcx		/* restore original dword without reload */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 7f			/* highest byte is NUL => return NULL */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 7f			/* found NUL => return NULL */
+	/* Fourth unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	xorq %r9, %rcx		/* XOR with qword c|...|c => bytes of str == c
+				   are now 0 */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found c => return pointer */
+
+	/* The quadword we looked at does not contain the value we're looking
+	   for.  Let's search now whether we have reached the end of the
+	   string.  */
+	xorq %r9, %rcx		/* restore original dword without reload */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 7f			/* highest byte is NUL => return NULL */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jz 4b			/* no NUL found => restart loop */
+
+
+7:	/* Return NULL.  */
+	xorl %eax, %eax
+	retq
+
+
+	/* We now scan for the byte in which the character was matched.
+	   But we have to take care of the case that a NUL char is
+	   found before this in the dword.  Note that we XORed %rcx
+	   with the byte we're looking for, therefore the tests below look
+	   reversed.  */
+
+
+	.p2align 4		/* Align, it's a jump target.  */
+3:	movq	%r9,%rdx	/* move to %rdx so that we can access bytes */
+	subq	$8,%rax		/* correct pointer increment.  */
+	testb %cl, %cl		/* is first byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %cl		/* is first byte NUL? */
+	je 7b			/* yes => return NULL */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is second byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %ch		/* is second byte NUL? */
+	je 7b			/* yes => return NULL? */
+	incq %rax		/* increment pointer */
+
+	shrq $16, %rcx		/* make upper bytes accessible */
+	testb %cl, %cl		/* is third byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %cl		/* is third byte NUL? */
+	je 7b			/* yes => return NULL */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is fourth byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %ch		/* is fourth byte NUL? */
+	je 7b			/* yes => return NULL? */
+	incq %rax		/* increment pointer */
+
+	shrq $16, %rcx		/* make upper bytes accessible */
+	testb %cl, %cl		/* is fifth byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %cl		/* is fifth byte NUL? */
+	je 7b			/* yes => return NULL */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is sixth byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %ch		/* is sixth byte NUL? */
+	je 7b			/* yes => return NULL? */
+	incq %rax		/* increment pointer */
+
+	shrq $16, %rcx		/* make upper bytes accessible */
+	testb %cl, %cl		/* is seventh byte C? */
+	jz 6f			/* yes => return pointer */
+	cmpb %dl, %cl		/* is seventh byte NUL? */
+	je 7b			/* yes => return NULL */
+
+	/* It must be in the eigth byte and it cannot be NUL.  */
+	incq %rax
+
+6:
+	nop
+	retq
+END (BP_SYM (strchr))
+
+weak_alias (BP_SYM (strchr), BP_SYM (index))
diff --git a/libc/string/x86_64/strcmp.S b/libc/string/x86_64/strcmp.S
new file mode 100644
index 000000000..0b401a773
--- /dev/null
+++ b/libc/string/x86_64/strcmp.S
@@ -0,0 +1,41 @@
+/* Highly optimized version for x86-64.
+   Copyright (C) 1999, 2000, 2002, 2003, 2005 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Based on i686 version contributed by Ulrich Drepper
+   <drepper@cygnus.com>, 1999.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+        .text
+ENTRY (BP_SYM (strcmp))
+L(oop):	movb	(%rdi), %al
+	cmpb	(%rsi), %al
+	jne	L(neq)
+	incq	%rdi
+	incq	%rsi
+	testb	%al, %al
+	jnz	L(oop)
+
+	xorl	%eax, %eax
+	ret
+
+L(neq):	movl	$1, %eax
+	movl	$-1, %ecx
+	cmovbl	%ecx, %eax
+	ret
+END (BP_SYM (strcmp))
diff --git a/libc/string/x86_64/strcpy.S b/libc/string/x86_64/strcpy.S
new file mode 100644
index 000000000..0847e2b7c
--- /dev/null
+++ b/libc/string/x86_64/strcpy.S
@@ -0,0 +1,153 @@
+/* strcpy/stpcpy implementation for x86-64.
+   Copyright (C) 2002 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Andreas Jaeger <aj@suse.de>, 2002.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+#ifndef USE_AS_STPCPY
+# define STRCPY strcpy
+#endif
+
+	.text
+ENTRY (BP_SYM (STRCPY))
+	movq %rsi, %rcx		/* Source register. */
+	andl $7, %ecx		/* mask alignment bits */
+	movq %rdi, %rdx		/* Duplicate destination pointer.  */
+
+	jz 5f			/* aligned => start loop */
+
+	neg %ecx		/* We need to align to 8 bytes.  */
+	addl $8,%ecx
+	/* Search the first bytes directly.  */
+0:
+	movb	(%rsi), %al	/* Fetch a byte */
+	testb	%al, %al	/* Is it NUL? */
+	movb	%al, (%rdx)	/* Store it */
+	jz	4f		/* If it was NUL, done! */
+	incq	%rsi
+	incq	%rdx
+	decl	%ecx
+	jnz	0b
+
+5:
+	movq $0xfefefefefefefeff,%r8
+
+	/* Now the sources is aligned.  Unfortunatly we cannot force
+	   to have both source and destination aligned, so ignore the
+	   alignment of the destination.  */
+	.p2align 4
+1:
+	/* 1st unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	3f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	3f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+
+	/* 2nd unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	3f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	3f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+
+	/* 3rd unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	3f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	3f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+
+	/* 4th unroll.  */
+	movq	(%rsi), %rax	/* Read double word (8 bytes).  */
+	addq	$8, %rsi	/* Adjust pointer for next word.  */
+	movq	%rax, %r9	/* Save a copy for NUL finding.  */
+	addq	%r8, %r9	/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc	3f		/* highest byte is NUL => return pointer */
+	xorq	%rax, %r9	/* (word+magic)^word */
+	orq	%r8, %r9	/* set all non-carry bits */
+	incq	%r9		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+
+	jnz	3f		/* found NUL => return pointer */
+
+	movq	%rax, (%rdx)	/* Write value to destination.  */
+	addq	$8, %rdx	/* Adjust pointer.  */
+	jmp	1b		/* Next iteration.  */
+
+	/* Do the last few bytes. %rax contains the value to write.
+	   The loop is unrolled twice.  */
+	.p2align 4
+3:
+	/* Note that stpcpy needs to return with the value of the NUL
+	   byte.  */
+	movb	%al, (%rdx)	/* 1st byte.  */
+	testb	%al, %al	/* Is it NUL.  */
+	jz	4f		/* yes, finish.  */
+	incq	%rdx		/* Increment destination.  */
+	movb	%ah, (%rdx)	/* 2nd byte.  */
+	testb	%ah, %ah	/* Is it NUL?.  */
+	jz	4f		/* yes, finish.  */
+	incq	%rdx		/* Increment destination.  */
+	shrq	$16, %rax	/* Shift...  */
+	jmp	3b		/* and look at next two bytes in %rax.  */
+
+4:
+#ifdef USE_AS_STPCPY
+	movq	%rdx, %rax	/* Destination is return value.  */
+#else
+	movq	%rdi, %rax	/* Source is return value.  */
+#endif
+	retq
+END (BP_SYM (STRCPY))
diff --git a/libc/string/x86_64/strcspn.S b/libc/string/x86_64/strcspn.S
new file mode 100644
index 000000000..e1a22e4f1
--- /dev/null
+++ b/libc/string/x86_64/strcspn.S
@@ -0,0 +1,123 @@
+/* strcspn (str, ss) -- Return the length of the initial segment of STR
+			which contains no characters from SS.
+   For AMD x86-64.
+   Copyright (C) 1994-1997, 2000, 2002, 2003, 2004, 2005
+   Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Ulrich Drepper <drepper@gnu.ai.mit.edu>.
+   Bug fixes by Alan Modra <Alan@SPRI.Levels.UniSA.Edu.Au>.
+   Adopted for x86-64 by Andreas Jaeger <aj@suse.de>.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+/* BEWARE: `#ifdef strcspn' means that strcspn is redefined as `strpbrk' */
+#define STRPBRK_P (defined strcspn)
+
+	.text
+ENTRY (strcspn)
+
+	movq %rdi, %rdx		/* Save SRC.  */
+
+	/* First we create a table with flags for all possible characters.
+	   For the ASCII (7bit/8bit) or ISO-8859-X character sets which are
+	   supported by the C string functions we have 256 characters.
+	   Before inserting marks for the stop characters we clear the whole
+	   table.  */
+	movq %rdi, %r8			/* Save value.  */
+	subq $256, %rsp			/* Make space for 256 bytes.  */
+	movl $32,  %ecx			/* 32*8 bytes = 256 bytes.  */
+	movq %rsp, %rdi
+	xorl %eax, %eax			/* We store 0s.  */
+	cld
+	rep
+	stosq
+
+	movq %rsi, %rax			/* Setup skipset.  */
+
+/* For understanding the following code remember that %rcx == 0 now.
+   Although all the following instruction only modify %cl we always
+   have a correct zero-extended 64-bit value in %rcx.  */
+
+	.p2align 4
+L(2):	movb (%rax), %cl	/* get byte from skipset */
+	testb %cl, %cl		/* is NUL char? */
+	jz L(1)			/* yes => start compare loop */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in skipset table */
+
+	movb 1(%rax), %cl	/* get byte from skipset */
+	testb $0xff, %cl	/* is NUL char? */
+	jz L(1)			/* yes => start compare loop */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in skipset table */
+
+	movb 2(%rax), %cl	/* get byte from skipset */
+	testb $0xff, %cl	/* is NUL char? */
+	jz L(1)			/* yes => start compare loop */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in skipset table */
+
+	movb 3(%rax), %cl	/* get byte from skipset */
+	addq $4, %rax		/* increment skipset pointer */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in skipset table */
+	testb $0xff, %cl	/* is NUL char? */
+	jnz L(2)		/* no => process next dword from skipset */
+
+L(1):	leaq -4(%rdx), %rax	/* prepare loop */
+
+	/* We use a neat trick for the following loop.  Normally we would
+	   have to test for two termination conditions
+	   1. a character in the skipset was found
+	   and
+	   2. the end of the string was found
+	   But as a sign that the character is in the skipset we store its
+	   value in the table.  But the value of NUL is NUL so the loop
+	   terminates for NUL in every case.  */
+
+	.p2align 4
+L(3):	addq $4, %rax		/* adjust pointer for full loop round */
+
+	movb (%rax), %cl	/* get byte from string */
+	cmpb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	je L(4)			/* yes => return */
+
+	movb 1(%rax), %cl	/* get byte from string */
+	cmpb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	je L(5)			/* yes => return */
+
+	movb 2(%rax), %cl	/* get byte from string */
+	cmpb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	jz L(6)			/* yes => return */
+
+	movb 3(%rax), %cl	/* get byte from string */
+	cmpb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	jne L(3)		/* no => start loop again */
+
+	incq %rax		/* adjust pointer */
+L(6):	incq %rax
+L(5):	incq %rax
+
+L(4):	addq $256, %rsp		/* remove skipset */
+#if STRPBRK_P
+	xorl %edx,%edx
+	orb %cl, %cl		/* was last character NUL? */
+	cmovzq %rdx, %rax	/* Yes:	return NULL */
+#else	
+	subq %rdx, %rax		/* we have to return the number of valid
+				   characters, so compute distance to first
+				   non-valid character */
+#endif
+	ret
+END (strcspn)
diff --git a/libc/string/x86_64/string.c b/libc/string/x86_64/string.c
new file mode 100644
index 000000000..60caddf99
--- /dev/null
+++ b/libc/string/x86_64/string.c
@@ -0,0 +1,1454 @@
+/* Tester for string functions.
+   Copyright (C) 1995-2000, 2001 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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#ifndef _GNU_SOURCE
+#define _GNU_SOURCE
+#endif
+
+/* Make sure we don't test the optimized inline functions if we want to
+   test the real implementation.  */
+#if !defined DO_STRING_INLINES
+#undef __USE_STRING_INLINES
+#endif
+
+#include <errno.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <strings.h>
+#include <fcntl.h>
+
+#define	STREQ(a, b)	(strcmp((a), (b)) == 0)
+
+const char *it = "<UNSET>";	/* Routine name for message routines. */
+size_t errors = 0;
+
+/* Complain if condition is not true.  */
+static void
+check (int thing, int number)
+{
+  if (!thing)
+    {
+      printf("%s flunked test %d\n", it, number);
+      ++errors;
+    }
+}
+
+/* Complain if first two args don't strcmp as equal.  */
+static void
+equal (const char *a, const char *b, int number)
+{
+  check(a != NULL && b != NULL && STREQ (a, b), number);
+}
+
+char one[50];
+char two[50];
+char *cp;
+
+static void
+test_strcmp (void)
+{
+  it = "strcmp";
+  check (strcmp ("", "") == 0, 1);		/* Trivial case. */
+  check (strcmp ("a", "a") == 0, 2);		/* Identity. */
+  check (strcmp ("abc", "abc") == 0, 3);	/* Multicharacter. */
+  check (strcmp ("abc", "abcd") < 0, 4);	/* Length mismatches. */
+  check (strcmp ("abcd", "abc") > 0, 5);
+  check (strcmp ("abcd", "abce") < 0, 6);	/* Honest miscompares. */
+  check (strcmp ("abce", "abcd") > 0, 7);
+  check (strcmp ("a\203", "a") > 0, 8);		/* Tricky if char signed. */
+  check (strcmp ("a\203", "a\003") > 0, 9);
+
+  {
+    char buf1[0x40], buf2[0x40];
+    int i, j;
+    for (i=0; i < 0x10; i++)
+      for (j = 0; j < 0x10; j++)
+	{
+	  int k;
+	  for (k = 0; k < 0x3f; k++)
+	    {
+	      buf1[j] = '0' ^ (k & 4);
+	      buf2[j] = '4' ^ (k & 4);
+	    }
+	  buf1[i] = buf1[0x3f] = 0;
+	  buf2[j] = buf2[0x3f] = 0;
+	  for (k = 0; k < 0xf; k++)
+	    {
+	      int cnum = 0x10+0x10*k+0x100*j+0x1000*i;
+	      check (strcmp (buf1+i,buf2+j) == 0, cnum);
+	      buf1[i+k] = 'A' + i + k;
+	      buf1[i+k+1] = 0;
+	      check (strcmp (buf1+i,buf2+j) > 0, cnum+1);
+	      check (strcmp (buf2+j,buf1+i) < 0, cnum+2);
+	      buf2[j+k] = 'B' + i + k;
+	      buf2[j+k+1] = 0;
+	      check (strcmp (buf1+i,buf2+j) < 0, cnum+3);
+	      check (strcmp (buf2+j,buf1+i) > 0, cnum+4);
+	      buf2[j+k] = 'A' + i + k;
+	      buf1[i] = 'A' + i + 0x80;
+	      check (strcmp (buf1+i,buf2+j) > 0, cnum+5);
+	      check (strcmp (buf2+j,buf1+i) < 0, cnum+6);
+	      buf1[i] = 'A' + i;
+	    }
+	}
+  }
+}
+
+#define SIMPLE_COPY(fn, n, str, ntest) \
+  do {									      \
+    int __n;								      \
+    char *cp;								      \
+    for (__n = 0; __n < (int) sizeof (one); ++__n)			      \
+      one[__n] = 'Z';							      \
+    fn (one, str);							      \
+    for (cp = one, __n = 0; __n < n; ++__n, ++cp)			      \
+      check (*cp == '0' + (n % 10), ntest);				      \
+    check (*cp == '\0', ntest);						      \
+  } while (0)
+
+static void
+test_strcpy (void)
+{
+  int i;
+  it = "strcpy";
+  check (strcpy (one, "abcd") == one, 1); /* Returned value. */
+  equal (one, "abcd", 2);		/* Basic test. */
+
+  (void) strcpy (one, "x");
+  equal (one, "x", 3);			/* Writeover. */
+  equal (one+2, "cd", 4);		/* Wrote too much? */
+
+  (void) strcpy (two, "hi there");
+  (void) strcpy (one, two);
+  equal (one, "hi there", 5);		/* Basic test encore. */
+  equal (two, "hi there", 6);		/* Stomped on source? */
+
+  (void) strcpy (one, "");
+  equal (one, "", 7);			/* Boundary condition. */
+
+  for (i = 0; i < 16; i++)
+    {
+      (void) strcpy (one + i, "hi there");	/* Unaligned destination. */
+      equal (one + i, "hi there", 8 + (i * 2));
+      (void) strcpy (two, one + i);		/* Unaligned source. */
+      equal (two, "hi there", 9 + (i * 2));
+    }
+
+  SIMPLE_COPY(strcpy, 0, "", 41);
+  SIMPLE_COPY(strcpy, 1, "1", 42);
+  SIMPLE_COPY(strcpy, 2, "22", 43);
+  SIMPLE_COPY(strcpy, 3, "333", 44);
+  SIMPLE_COPY(strcpy, 4, "4444", 45);
+  SIMPLE_COPY(strcpy, 5, "55555", 46);
+  SIMPLE_COPY(strcpy, 6, "666666", 47);
+  SIMPLE_COPY(strcpy, 7, "7777777", 48);
+  SIMPLE_COPY(strcpy, 8, "88888888", 49);
+  SIMPLE_COPY(strcpy, 9, "999999999", 50);
+  SIMPLE_COPY(strcpy, 10, "0000000000", 51);
+  SIMPLE_COPY(strcpy, 11, "11111111111", 52);
+  SIMPLE_COPY(strcpy, 12, "222222222222", 53);
+  SIMPLE_COPY(strcpy, 13, "3333333333333", 54);
+  SIMPLE_COPY(strcpy, 14, "44444444444444", 55);
+  SIMPLE_COPY(strcpy, 15, "555555555555555", 56);
+  SIMPLE_COPY(strcpy, 16, "6666666666666666", 57);
+}
+
+static void
+test_stpcpy (void)
+{
+  it = "stpcpy";
+  check ((stpcpy (one, "a") - one) == 1, 1);
+  equal (one, "a", 2);
+
+  check ((stpcpy (one, "ab") - one) == 2, 3);
+  equal (one, "ab", 4);
+
+  check ((stpcpy (one, "abc") - one) == 3, 5);
+  equal (one, "abc", 6);
+
+  check ((stpcpy (one, "abcd") - one) == 4, 7);
+  equal (one, "abcd", 8);
+
+  check ((stpcpy (one, "abcde") - one) == 5, 9);
+  equal (one, "abcde", 10);
+
+  check ((stpcpy (one, "abcdef") - one) == 6, 11);
+  equal (one, "abcdef", 12);
+
+  check ((stpcpy (one, "abcdefg") - one) == 7, 13);
+  equal (one, "abcdefg", 14);
+
+  check ((stpcpy (one, "abcdefgh") - one) == 8, 15);
+  equal (one, "abcdefgh", 16);
+
+  check ((stpcpy (one, "abcdefghi") - one) == 9, 17);
+  equal (one, "abcdefghi", 18);
+
+  check ((stpcpy (one, "x") - one) == 1, 19);
+  equal (one, "x", 20);			/* Writeover. */
+  equal (one+2, "cdefghi", 21);		/* Wrote too much? */
+
+  check ((stpcpy (one, "xx") - one) == 2, 22);
+  equal (one, "xx", 23);		/* Writeover. */
+  equal (one+3, "defghi", 24);		/* Wrote too much? */
+
+  check ((stpcpy (one, "xxx") - one) == 3, 25);
+  equal (one, "xxx", 26);		/* Writeover. */
+  equal (one+4, "efghi", 27);		/* Wrote too much? */
+
+  check ((stpcpy (one, "xxxx") - one) == 4, 28);
+  equal (one, "xxxx", 29);		/* Writeover. */
+  equal (one+5, "fghi", 30);		/* Wrote too much? */
+
+  check ((stpcpy (one, "xxxxx") - one) == 5, 31);
+  equal (one, "xxxxx", 32);		/* Writeover. */
+  equal (one+6, "ghi", 33);		/* Wrote too much? */
+
+  check ((stpcpy (one, "xxxxxx") - one) == 6, 34);
+  equal (one, "xxxxxx", 35);		/* Writeover. */
+  equal (one+7, "hi", 36);		/* Wrote too much? */
+
+  check ((stpcpy (one, "xxxxxxx") - one) == 7, 37);
+  equal (one, "xxxxxxx", 38);		/* Writeover. */
+  equal (one+8, "i", 39);		/* Wrote too much? */
+
+  check ((stpcpy (stpcpy (stpcpy (one, "a"), "b"), "c") - one) == 3, 40);
+  equal (one, "abc", 41);
+  equal (one + 4, "xxx", 42);
+
+  SIMPLE_COPY(stpcpy, 0, "", 43);
+  SIMPLE_COPY(stpcpy, 1, "1", 44);
+  SIMPLE_COPY(stpcpy, 2, "22", 45);
+  SIMPLE_COPY(stpcpy, 3, "333", 46);
+  SIMPLE_COPY(stpcpy, 4, "4444", 47);
+  SIMPLE_COPY(stpcpy, 5, "55555", 48);
+  SIMPLE_COPY(stpcpy, 6, "666666", 49);
+  SIMPLE_COPY(stpcpy, 7, "7777777", 50);
+  SIMPLE_COPY(stpcpy, 8, "88888888", 51);
+  SIMPLE_COPY(stpcpy, 9, "999999999", 52);
+  SIMPLE_COPY(stpcpy, 10, "0000000000", 53);
+  SIMPLE_COPY(stpcpy, 11, "11111111111", 54);
+  SIMPLE_COPY(stpcpy, 12, "222222222222", 55);
+  SIMPLE_COPY(stpcpy, 13, "3333333333333", 56);
+  SIMPLE_COPY(stpcpy, 14, "44444444444444", 57);
+  SIMPLE_COPY(stpcpy, 15, "555555555555555", 58);
+  SIMPLE_COPY(stpcpy, 16, "6666666666666666", 59);
+}
+
+static void
+test_stpncpy (void)
+{
+  it = "stpncpy";
+  memset (one, 'x', sizeof (one));
+  check (stpncpy (one, "abc", 2) == one + 2, 1);
+  check (stpncpy (one, "abc", 3) == one + 3, 2);
+  check (stpncpy (one, "abc", 4) == one + 3, 3);
+  check (one[3] == '\0' && one[4] == 'x', 4);
+  check (stpncpy (one, "abcd", 5) == one + 4, 5);
+  check (one[4] == '\0' && one[5] == 'x', 6);
+  check (stpncpy (one, "abcd", 6) == one + 4, 7);
+  check (one[4] == '\0' && one[5] == '\0' && one[6] == 'x', 8);
+}
+
+static void
+test_strcat (void)
+{
+  it = "strcat";
+  (void) strcpy (one, "ijk");
+  check (strcat (one, "lmn") == one, 1); /* Returned value. */
+  equal (one, "ijklmn", 2);		/* Basic test. */
+
+  (void) strcpy (one, "x");
+  (void) strcat (one, "yz");
+  equal (one, "xyz", 3);			/* Writeover. */
+  equal (one+4, "mn", 4);			/* Wrote too much? */
+
+  (void) strcpy (one, "gh");
+  (void) strcpy (two, "ef");
+  (void) strcat (one, two);
+  equal (one, "ghef", 5);			/* Basic test encore. */
+  equal (two, "ef", 6);			/* Stomped on source? */
+
+  (void) strcpy (one, "");
+  (void) strcat (one, "");
+  equal (one, "", 7);			/* Boundary conditions. */
+  (void) strcpy (one, "ab");
+  (void) strcat (one, "");
+  equal (one, "ab", 8);
+  (void) strcpy (one, "");
+  (void) strcat (one, "cd");
+  equal (one, "cd", 9);
+}
+
+static void
+test_strncat (void)
+{
+  /* First test it as strcat, with big counts, then test the count
+     mechanism.  */
+  it = "strncat";
+  (void) strcpy (one, "ijk");
+  check (strncat (one, "lmn", 99) == one, 1);	/* Returned value. */
+  equal (one, "ijklmn", 2);		/* Basic test. */
+
+  (void) strcpy (one, "x");
+  (void) strncat (one, "yz", 99);
+  equal (one, "xyz", 3);		/* Writeover. */
+  equal (one+4, "mn", 4);		/* Wrote too much? */
+
+  (void) strcpy (one, "gh");
+  (void) strcpy (two, "ef");
+  (void) strncat (one, two, 99);
+  equal (one, "ghef", 5);			/* Basic test encore. */
+  equal (two, "ef", 6);			/* Stomped on source? */
+
+  (void) strcpy (one, "");
+  (void) strncat (one, "", 99);
+  equal (one, "", 7);			/* Boundary conditions. */
+  (void) strcpy (one, "ab");
+  (void) strncat (one, "", 99);
+  equal (one, "ab", 8);
+  (void) strcpy (one, "");
+  (void) strncat (one, "cd", 99);
+  equal (one, "cd", 9);
+
+  (void) strcpy (one, "ab");
+  (void) strncat (one, "cdef", 2);
+  equal (one, "abcd", 10);			/* Count-limited. */
+
+  (void) strncat (one, "gh", 0);
+  equal (one, "abcd", 11);			/* Zero count. */
+
+  (void) strncat (one, "gh", 2);
+  equal (one, "abcdgh", 12);		/* Count and length equal. */
+
+  (void) strncat (one, "ij", (size_t)-1);	/* set sign bit in count */
+  equal (one, "abcdghij", 13);
+}
+
+static void
+test_strncmp (void)
+{
+  /* First test as strcmp with big counts, then test count code.  */
+  it = "strncmp";
+  check (strncmp ("", "", 99) == 0, 1);	/* Trivial case. */
+  check (strncmp ("a", "a", 99) == 0, 2);	/* Identity. */
+  check (strncmp ("abc", "abc", 99) == 0, 3);	/* Multicharacter. */
+  check (strncmp ("abc", "abcd", 99) < 0, 4);	/* Length unequal. */
+  check (strncmp ("abcd", "abc", 99) > 0, 5);
+  check (strncmp ("abcd", "abce", 99) < 0, 6);	/* Honestly unequal. */
+  check (strncmp ("abce", "abcd", 99) > 0, 7);
+  check (strncmp ("a\203", "a", 2) > 0, 8);	/* Tricky if '\203' < 0 */
+  check (strncmp ("a\203", "a\003", 2) > 0, 9);
+  check (strncmp ("abce", "abcd", 3) == 0, 10);	/* Count limited. */
+  check (strncmp ("abce", "abc", 3) == 0, 11);	/* Count == length. */
+  check (strncmp ("abcd", "abce", 4) < 0, 12);	/* Nudging limit. */
+  check (strncmp ("abc", "def", 0) == 0, 13);	/* Zero count. */
+  check (strncmp ("abc", "", (size_t)-1) > 0, 14); /* set sign bit in count */
+  check (strncmp ("abc", "abc", (size_t)-2) == 0, 15);
+}
+
+static void
+test_strncpy (void)
+{
+  /* Testing is a bit different because of odd semantics.  */
+  it = "strncpy";
+  check (strncpy (one, "abc", 4) == one, 1);	/* Returned value. */
+  equal (one, "abc", 2);			/* Did the copy go right? */
+
+  (void) strcpy (one, "abcdefgh");
+  (void) strncpy (one, "xyz", 2);
+  equal (one, "xycdefgh", 3);			/* Copy cut by count. */
+
+  (void) strcpy (one, "abcdefgh");
+  (void) strncpy (one, "xyz", 3);		/* Copy cut just before NUL. */
+  equal (one, "xyzdefgh", 4);
+
+  (void) strcpy (one, "abcdefgh");
+  (void) strncpy (one, "xyz", 4);		/* Copy just includes NUL. */
+  equal (one, "xyz", 5);
+  equal (one+4, "efgh", 6);			/* Wrote too much? */
+
+  (void) strcpy (one, "abcdefgh");
+  (void) strncpy (one, "xyz", 5);		/* Copy includes padding. */
+  equal (one, "xyz", 7);
+  equal (one+4, "", 8);
+  equal (one+5, "fgh", 9);
+
+  (void) strcpy (one, "abc");
+  (void) strncpy (one, "xyz", 0);		/* Zero-length copy. */
+  equal (one, "abc", 10);
+
+  (void) strncpy (one, "", 2);		/* Zero-length source. */
+  equal (one, "", 11);
+  equal (one+1, "", 12);
+  equal (one+2, "c", 13);
+
+  (void) strcpy (one, "hi there");
+  (void) strncpy (two, one, 9);
+  equal (two, "hi there", 14);		/* Just paranoia. */
+  equal (one, "hi there", 15);		/* Stomped on source? */
+}
+
+static void
+test_strlen (void)
+{
+  it = "strlen";
+  check (strlen ("") == 0, 1);		/* Empty. */
+  check (strlen ("a") == 1, 2);		/* Single char. */
+  check (strlen ("abcd") == 4, 3);	/* Multiple chars. */
+  {
+    char buf[4096];
+    int i;
+    char *p;
+    for (i=0; i < 0x100; i++)
+      {
+	p = (char *) ((unsigned long int)(buf + 0xff) & ~0xff) + i;
+	strcpy (p, "OK");
+	strcpy (p+3, "BAD/WRONG");
+	check (strlen (p) == 2, 4+i);
+      }
+   }
+}
+
+static void
+test_strnlen (void)
+{
+  it = "strnlen";
+  check (strnlen ("", 10) == 0, 1);		/* Empty. */
+  check (strnlen ("a", 10) == 1, 2);		/* Single char. */
+  check (strnlen ("abcd", 10) == 4, 3);	/* Multiple chars. */
+  check (strnlen ("foo", (size_t)-1) == 3, 4);	/* limits of n. */
+
+  {
+    char buf[4096];
+    int i;
+    char *p;
+    for (i=0; i < 0x100; i++)
+      {
+	p = (char *) ((unsigned long int)(buf + 0xff) & ~0xff) + i;
+	strcpy (p, "OK");
+	strcpy (p+3, "BAD/WRONG");
+	check (strnlen (p, 100) == 2, 5+i);
+      }
+   }
+}
+
+static void
+test_strchr (void)
+{
+  it = "strchr";
+  check (strchr ("abcd", 'z') == NULL, 1);	/* Not found. */
+  (void) strcpy (one, "abcd");
+  check (strchr (one, 'c') == one+2, 2);	/* Basic test. */
+  check (strchr (one, 'd') == one+3, 3);	/* End of string. */
+  check (strchr (one, 'a') == one, 4);		/* Beginning. */
+  check (strchr (one, '\0') == one+4, 5);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  check (strchr (one, 'b') == one+1, 6);	/* Finding first. */
+  (void) strcpy (one, "");
+  check (strchr (one, 'b') == NULL, 7);		/* Empty string. */
+  check (strchr (one, '\0') == one, 8);		/* NUL in empty string. */
+  {
+    char buf[4096];
+    int i;
+    char *p;
+    for (i=0; i < 0x100; i++)
+      {
+	p = (char *) ((unsigned long int) (buf + 0xff) & ~0xff) + i;
+	strcpy (p, "OK");
+	strcpy (p+3, "BAD/WRONG");
+	check (strchr (p, '/') == NULL, 9+i);
+      }
+   }
+}
+
+static void
+test_strchrnul (void)
+{
+  const char *os;
+  it = "strchrnul";
+  cp = strchrnul ((os = "abcd"), 'z');
+  check (*cp == '\0', 1);			/* Not found. */
+  check (cp == os + 4, 2);
+  (void) strcpy (one, "abcd");
+  check (strchrnul (one, 'c') == one+2, 3);	/* Basic test. */
+  check (strchrnul (one, 'd') == one+3, 4);	/* End of string. */
+  check (strchrnul (one, 'a') == one, 5);	/* Beginning. */
+  check (strchrnul (one, '\0') == one+4, 6);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  check (strchrnul (one, 'b') == one+1, 7);	/* Finding first. */
+  (void) strcpy (one, "");
+  check (strchrnul (one, 'b') == one, 8);	/* Empty string. */
+  check (strchrnul (one, '\0') == one, 9);	/* NUL in empty string. */
+  {
+    char buf[4096];
+    int i;
+    char *p;
+    for (i=0; i < 0x100; i++)
+      {
+	p = (char *) ((unsigned long int) (buf + 0xff) & ~0xff) + i;
+	strcpy (p, "OK");
+	strcpy (p+3, "BAD/WRONG");
+	cp = strchrnul (p, '/');
+	check (*cp == '\0', 9+2*i);
+	check (cp == p+2, 10+2*i);
+      }
+   }
+}
+
+static void
+test_rawmemchr (void)
+{
+  it = "rawmemchr";
+  (void) strcpy (one, "abcd");
+  check (rawmemchr (one, 'c') == one+2, 1);	/* Basic test. */
+  check (rawmemchr (one, 'd') == one+3, 2);	/* End of string. */
+  check (rawmemchr (one, 'a') == one, 3);		/* Beginning. */
+  check (rawmemchr (one, '\0') == one+4, 4);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  check (rawmemchr (one, 'b') == one+1, 5);	/* Finding first. */
+  (void) strcpy (one, "");
+  check (rawmemchr (one, '\0') == one, 6);	/* NUL in empty string. */
+  {
+    char buf[4096];
+    int i;
+    char *p;
+    for (i=0; i < 0x100; i++)
+      {
+	p = (char *) ((unsigned long int) (buf + 0xff) & ~0xff) + i;
+	strcpy (p, "OK");
+	strcpy (p+3, "BAD/WRONG");
+	check (rawmemchr (p, 'R') == p+8, 6+i);
+      }
+   }
+}
+
+static void
+test_index (void)
+{
+  it = "index";
+  check (index ("abcd", 'z') == NULL, 1);	/* Not found. */
+  (void) strcpy (one, "abcd");
+  check (index (one, 'c') == one+2, 2);	/* Basic test. */
+  check (index (one, 'd') == one+3, 3);	/* End of string. */
+  check (index (one, 'a') == one, 4);	/* Beginning. */
+  check (index (one, '\0') == one+4, 5);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  check (index (one, 'b') == one+1, 6);	/* Finding first. */
+  (void) strcpy (one, "");
+  check (index (one, 'b') == NULL, 7);	/* Empty string. */
+  check (index (one, '\0') == one, 8);	/* NUL in empty string. */
+}
+
+static void
+test_strrchr (void)
+{
+  it = "strrchr";
+  check (strrchr ("abcd", 'z') == NULL, 1);	/* Not found. */
+  (void) strcpy (one, "abcd");
+  check (strrchr (one, 'c') == one+2, 2);	/* Basic test. */
+  check (strrchr (one, 'd') == one+3, 3);	/* End of string. */
+  check (strrchr (one, 'a') == one, 4);		/* Beginning. */
+  check (strrchr (one, '\0') == one+4, 5);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  check (strrchr (one, 'b') == one+3, 6);	/* Finding last. */
+  (void) strcpy (one, "");
+  check (strrchr (one, 'b') == NULL, 7);	/* Empty string. */
+  check (strrchr (one, '\0') == one, 8);	/* NUL in empty string. */
+  {
+    char buf[4096];
+    int i;
+    char *p;
+    for (i=0; i < 0x100; i++)
+      {
+	p = (char *) ((unsigned long int) (buf + 0xff) & ~0xff) + i;
+	strcpy (p, "OK");
+	strcpy (p+3, "BAD/WRONG");
+	check (strrchr (p, '/') == NULL, 9+i);
+      }
+   }
+}
+
+static void
+test_memrchr (void)
+{
+  size_t l;
+  it = "memrchr";
+  check (memrchr ("abcd", 'z', 5) == NULL, 1);	/* Not found. */
+  (void) strcpy (one, "abcd");
+  l = strlen (one) + 1;
+  check (memrchr (one, 'c', l) == one+2, 2);	/* Basic test. */
+  check (memrchr (one, 'd', l) == one+3, 3);	/* End of string. */
+  check (memrchr (one, 'a', l) == one, 4);		/* Beginning. */
+  check (memrchr (one, '\0', l) == one+4, 5);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  l = strlen (one) + 1;
+  check (memrchr (one, 'b', l) == one+3, 6);	/* Finding last. */
+  (void) strcpy (one, "");
+  l = strlen (one) + 1;
+  check (memrchr (one, 'b', l) == NULL, 7);	/* Empty string. */
+  check (memrchr (one, '\0', l) == one, 8);	/* NUL in empty string. */
+
+  /* now test all possible alignment and length combinations to catch
+     bugs due to unrolled loops (assuming unrolling is limited to no
+     more than 128 byte chunks: */
+  {
+    char buf[128 + sizeof(long)];
+    long align, len, i, pos;
+
+    for (align = 0; align < (long) sizeof(long); ++align) {
+      for (len = 0; len < (long) (sizeof(buf) - align); ++len) {
+	for (i = 0; i < len; ++i)
+	  buf[align + i] = 'x';		/* don't depend on memset... */
+
+	for (pos = len - 1; pos >= 0; --pos) {
+#if 0
+	  printf("align %d, len %d, pos %d\n", align, len, pos);
+#endif
+	  check(memrchr(buf + align, 'x', len) == buf + align + pos, 9);
+	  check(memrchr(buf + align + pos + 1, 'x', len - (pos + 1)) == NULL,
+		10);
+	  buf[align + pos] = '-';
+	}
+      }
+    }
+  }
+}
+
+static void
+test_rindex (void)
+{
+  it = "rindex";
+  check (rindex ("abcd", 'z') == NULL, 1);	/* Not found. */
+  (void) strcpy (one, "abcd");
+  check (rindex (one, 'c') == one+2, 2);	/* Basic test. */
+  check (rindex (one, 'd') == one+3, 3);	/* End of string. */
+  check (rindex (one, 'a') == one, 4);	/* Beginning. */
+  check (rindex (one, '\0') == one+4, 5);	/* Finding NUL. */
+  (void) strcpy (one, "ababa");
+  check (rindex (one, 'b') == one+3, 6);	/* Finding last. */
+  (void) strcpy (one, "");
+  check (rindex (one, 'b') == NULL, 7);	/* Empty string. */
+  check (rindex (one, '\0') == one, 8);	/* NUL in empty string. */
+}
+
+static void
+test_strpbrk (void)
+{
+  it = "strpbrk";
+  check(strpbrk("abcd", "z") == NULL, 1);	/* Not found. */
+  (void) strcpy(one, "abcd");
+  check(strpbrk(one, "c") == one+2, 2);	/* Basic test. */
+  check(strpbrk(one, "d") == one+3, 3);	/* End of string. */
+  check(strpbrk(one, "a") == one, 4);	/* Beginning. */
+  check(strpbrk(one, "") == NULL, 5);	/* Empty search list. */
+  check(strpbrk(one, "cb") == one+1, 6);	/* Multiple search. */
+  (void) strcpy(one, "abcabdea");
+  check(strpbrk(one, "b") == one+1, 7);	/* Finding first. */
+  check(strpbrk(one, "cb") == one+1, 8);	/* With multiple search. */
+  check(strpbrk(one, "db") == one+1, 9);	/* Another variant. */
+  (void) strcpy(one, "");
+  check(strpbrk(one, "bc") == NULL, 10);	/* Empty string. */
+  (void) strcpy(one, "");
+  check(strpbrk(one, "bcd") == NULL, 11);	/* Empty string. */
+  (void) strcpy(one, "");
+  check(strpbrk(one, "bcde") == NULL, 12);	/* Empty string. */
+  check(strpbrk(one, "") == NULL, 13);	/* Both strings empty. */
+  (void) strcpy(one, "abcabdea");
+  check(strpbrk(one, "befg") == one+1, 14);	/* Finding first. */
+  check(strpbrk(one, "cbr") == one+1, 15);	/* With multiple search. */
+  check(strpbrk(one, "db") == one+1, 16);	/* Another variant. */
+  check(strpbrk(one, "efgh") == one+6, 17);	/* And yet another. */
+}
+
+static void
+test_strstr (void)
+{
+  it = "strstr";
+  check(strstr("abcd", "z") == NULL, 1);	/* Not found. */
+  check(strstr("abcd", "abx") == NULL, 2);	/* Dead end. */
+  (void) strcpy(one, "abcd");
+  check(strstr(one, "c") == one+2, 3);	/* Basic test. */
+  check(strstr(one, "bc") == one+1, 4);	/* Multichar. */
+  check(strstr(one, "d") == one+3, 5);	/* End of string. */
+  check(strstr(one, "cd") == one+2, 6);	/* Tail of string. */
+  check(strstr(one, "abc") == one, 7);	/* Beginning. */
+  check(strstr(one, "abcd") == one, 8);	/* Exact match. */
+  check(strstr(one, "abcde") == NULL, 9);	/* Too long. */
+  check(strstr(one, "de") == NULL, 10);	/* Past end. */
+  check(strstr(one, "") == one, 11);	/* Finding empty. */
+  (void) strcpy(one, "ababa");
+  check(strstr(one, "ba") == one+1, 12);	/* Finding first. */
+  (void) strcpy(one, "");
+  check(strstr(one, "b") == NULL, 13);	/* Empty string. */
+  check(strstr(one, "") == one, 14);	/* Empty in empty string. */
+  (void) strcpy(one, "bcbca");
+  check(strstr(one, "bca") == one+2, 15);	/* False start. */
+  (void) strcpy(one, "bbbcabbca");
+  check(strstr(one, "bbca") == one+1, 16);	/* With overlap. */
+}
+
+static void
+test_strspn (void)
+{
+  it = "strspn";
+  check(strspn("abcba", "abc") == 5, 1);	/* Whole string. */
+  check(strspn("abcba", "ab") == 2, 2);	/* Partial. */
+  check(strspn("abc", "qx") == 0, 3);	/* None. */
+  check(strspn("", "ab") == 0, 4);	/* Null string. */
+  check(strspn("abc", "") == 0, 5);	/* Null search list. */
+}
+
+static void
+test_strcspn (void)
+{
+  it = "strcspn";
+  check(strcspn("abcba", "qx") == 5, 1);	/* Whole string. */
+  check(strcspn("abcba", "cx") == 2, 2);	/* Partial. */
+  check(strcspn("abc", "abc") == 0, 3);	/* None. */
+  check(strcspn("", "ab") == 0, 4);	/* Null string. */
+  check(strcspn("abc", "") == 3, 5);	/* Null search list. */
+}
+
+static void
+test_strtok (void)
+{
+  it = "strtok";
+  (void) strcpy(one, "first, second, third");
+  equal(strtok(one, ", "), "first", 1);	/* Basic test. */
+  equal(one, "first", 2);
+  equal(strtok((char *)NULL, ", "), "second", 3);
+  equal(strtok((char *)NULL, ", "), "third", 4);
+  check(strtok((char *)NULL, ", ") == NULL, 5);
+  (void) strcpy(one, ", first, ");
+  equal(strtok(one, ", "), "first", 6);	/* Extra delims, 1 tok. */
+  check(strtok((char *)NULL, ", ") == NULL, 7);
+  (void) strcpy(one, "1a, 1b; 2a, 2b");
+  equal(strtok(one, ", "), "1a", 8);	/* Changing delim lists. */
+  equal(strtok((char *)NULL, "; "), "1b", 9);
+  equal(strtok((char *)NULL, ", "), "2a", 10);
+  (void) strcpy(two, "x-y");
+  equal(strtok(two, "-"), "x", 11);	/* New string before done. */
+  equal(strtok((char *)NULL, "-"), "y", 12);
+  check(strtok((char *)NULL, "-") == NULL, 13);
+  (void) strcpy(one, "a,b, c,, ,d");
+  equal(strtok(one, ", "), "a", 14);	/* Different separators. */
+  equal(strtok((char *)NULL, ", "), "b", 15);
+  equal(strtok((char *)NULL, " ,"), "c", 16);	/* Permute list too. */
+  equal(strtok((char *)NULL, " ,"), "d", 17);
+  check(strtok((char *)NULL, ", ") == NULL, 18);
+  check(strtok((char *)NULL, ", ") == NULL, 19);	/* Persistence. */
+  (void) strcpy(one, ", ");
+  check(strtok(one, ", ") == NULL, 20);	/* No tokens. */
+  (void) strcpy(one, "");
+  check(strtok(one, ", ") == NULL, 21);	/* Empty string. */
+  (void) strcpy(one, "abc");
+  equal(strtok(one, ", "), "abc", 22);	/* No delimiters. */
+  check(strtok((char *)NULL, ", ") == NULL, 23);
+  (void) strcpy(one, "abc");
+  equal(strtok(one, ""), "abc", 24);	/* Empty delimiter list. */
+  check(strtok((char *)NULL, "") == NULL, 25);
+  (void) strcpy(one, "abcdefgh");
+  (void) strcpy(one, "a,b,c");
+  equal(strtok(one, ","), "a", 26);	/* Basics again... */
+  equal(strtok((char *)NULL, ","), "b", 27);
+  equal(strtok((char *)NULL, ","), "c", 28);
+  check(strtok((char *)NULL, ",") == NULL, 29);
+  equal(one+6, "gh", 30);			/* Stomped past end? */
+  equal(one, "a", 31);			/* Stomped old tokens? */
+  equal(one+2, "b", 32);
+  equal(one+4, "c", 33);
+}
+
+static void
+test_strtok_r (void)
+{
+  it = "strtok_r";
+  (void) strcpy(one, "first, second, third");
+  cp = NULL;	/* Always initialize cp to make sure it doesn't point to some old data.  */
+  equal(strtok_r(one, ", ", &cp), "first", 1);	/* Basic test. */
+  equal(one, "first", 2);
+  equal(strtok_r((char *)NULL, ", ", &cp), "second", 3);
+  equal(strtok_r((char *)NULL, ", ", &cp), "third", 4);
+  check(strtok_r((char *)NULL, ", ", &cp) == NULL, 5);
+  (void) strcpy(one, ", first, ");
+  cp = NULL;
+  equal(strtok_r(one, ", ", &cp), "first", 6);	/* Extra delims, 1 tok. */
+  check(strtok_r((char *)NULL, ", ", &cp) == NULL, 7);
+  (void) strcpy(one, "1a, 1b; 2a, 2b");
+  cp = NULL;
+  equal(strtok_r(one, ", ", &cp), "1a", 8);	/* Changing delim lists. */
+  equal(strtok_r((char *)NULL, "; ", &cp), "1b", 9);
+  equal(strtok_r((char *)NULL, ", ", &cp), "2a", 10);
+  (void) strcpy(two, "x-y");
+  cp = NULL;
+  equal(strtok_r(two, "-", &cp), "x", 11);	/* New string before done. */
+  equal(strtok_r((char *)NULL, "-", &cp), "y", 12);
+  check(strtok_r((char *)NULL, "-", &cp) == NULL, 13);
+  (void) strcpy(one, "a,b, c,, ,d");
+  cp = NULL;
+  equal(strtok_r(one, ", ", &cp), "a", 14);	/* Different separators. */
+  equal(strtok_r((char *)NULL, ", ", &cp), "b", 15);
+  equal(strtok_r((char *)NULL, " ,", &cp), "c", 16);	/* Permute list too. */
+  equal(strtok_r((char *)NULL, " ,", &cp), "d", 17);
+  check(strtok_r((char *)NULL, ", ", &cp) == NULL, 18);
+  check(strtok_r((char *)NULL, ", ", &cp) == NULL, 19);	/* Persistence. */
+  (void) strcpy(one, ", ");
+  cp = NULL;
+  check(strtok_r(one, ", ", &cp) == NULL, 20);	/* No tokens. */
+  (void) strcpy(one, "");
+  cp = NULL;
+  check(strtok_r(one, ", ", &cp) == NULL, 21);	/* Empty string. */
+  check(strtok_r((char *)NULL, ", ", &cp) == NULL, 22);	/* Persistence. */
+  (void) strcpy(one, "abc");
+  cp = NULL;
+  equal(strtok_r(one, ", ", &cp), "abc", 23);	/* No delimiters. */
+  check(strtok_r((char *)NULL, ", ", &cp) == NULL, 24);
+  (void) strcpy(one, "abc");
+  cp = NULL;
+  equal(strtok_r(one, "", &cp), "abc", 25);	/* Empty delimiter list. */
+  check(strtok_r((char *)NULL, "", &cp) == NULL, 26);
+  (void) strcpy(one, "abcdefgh");
+  (void) strcpy(one, "a,b,c");
+  cp = NULL;
+  equal(strtok_r(one, ",", &cp), "a", 27);	/* Basics again... */
+  equal(strtok_r((char *)NULL, ",", &cp), "b", 28);
+  equal(strtok_r((char *)NULL, ",", &cp), "c", 29);
+  check(strtok_r((char *)NULL, ",", &cp) == NULL, 30);
+  equal(one+6, "gh", 31);			/* Stomped past end? */
+  equal(one, "a", 32);			/* Stomped old tokens? */
+  equal(one+2, "b", 33);
+  equal(one+4, "c", 34);
+}
+
+static void
+test_strsep (void)
+{
+  char *ptr;
+  it = "strsep";
+  cp = strcpy(one, "first, second, third");
+  equal(strsep(&cp, ", "), "first", 1);	/* Basic test. */
+  equal(one, "first", 2);
+  equal(strsep(&cp, ", "), "", 3);
+  equal(strsep(&cp, ", "), "second", 4);
+  equal(strsep(&cp, ", "), "", 5);
+  equal(strsep(&cp, ", "), "third", 6);
+  check(strsep(&cp, ", ") == NULL, 7);
+  cp = strcpy(one, ", first, ");
+  equal(strsep(&cp, ", "), "", 8);
+  equal(strsep(&cp, ", "), "", 9);
+  equal(strsep(&cp, ", "), "first", 10);	/* Extra delims, 1 tok. */
+  equal(strsep(&cp, ", "), "", 11);
+  equal(strsep(&cp, ", "), "", 12);
+  check(strsep(&cp, ", ") == NULL, 13);
+  cp = strcpy(one, "1a, 1b; 2a, 2b");
+  equal(strsep(&cp, ", "), "1a", 14);	/* Changing delim lists. */
+  equal(strsep(&cp, ", "), "", 15);
+  equal(strsep(&cp, "; "), "1b", 16);
+  equal(strsep(&cp, ", "), "", 17);
+  equal(strsep(&cp, ", "), "2a", 18);
+  cp = strcpy(two, "x-y");
+  equal(strsep(&cp, "-"), "x", 19);	/* New string before done. */
+  equal(strsep(&cp, "-"), "y", 20);
+  check(strsep(&cp, "-") == NULL, 21);
+  cp = strcpy(one, "a,b, c,, ,d ");
+  equal(strsep(&cp, ", "), "a", 22);	/* Different separators. */
+  equal(strsep(&cp, ", "), "b", 23);
+  equal(strsep(&cp, " ,"), "", 24);
+  equal(strsep(&cp, " ,"), "c", 25);	/* Permute list too. */
+  equal(strsep(&cp, " ,"), "", 26);
+  equal(strsep(&cp, " ,"), "", 27);
+  equal(strsep(&cp, " ,"), "", 28);
+  equal(strsep(&cp, " ,"), "d", 29);
+  equal(strsep(&cp, " ,"), "", 30);
+  check(strsep(&cp, ", ") == NULL, 31);
+  check(strsep(&cp, ", ") == NULL, 32);	/* Persistence. */
+  cp = strcpy(one, ", ");
+  equal(strsep(&cp, ", "), "", 33);
+  equal(strsep(&cp, ", "), "", 34);
+  equal(strsep(&cp, ", "), "", 35);
+  check(strsep(&cp, ", ") == NULL, 36);	/* No tokens. */
+  cp = strcpy(one, "");
+  equal(strsep(&cp, ", "), "", 37);
+  check(strsep(&cp, ", ") == NULL, 38);	/* Empty string. */
+  cp = strcpy(one, "abc");
+  equal(strsep(&cp, ", "), "abc", 39);	/* No delimiters. */
+  check(strsep(&cp, ", ") == NULL, 40);
+  cp = strcpy(one, "abc");
+  equal(strsep(&cp, ""), "abc", 41);	/* Empty delimiter list. */
+  check(strsep(&cp, "") == NULL, 42);
+  (void) strcpy(one, "abcdefgh");
+  cp = strcpy(one, "a,b,c");
+  equal(strsep(&cp, ","), "a", 43);	/* Basics again... */
+  equal(strsep(&cp, ","), "b", 44);
+  equal(strsep(&cp, ","), "c", 45);
+  check(strsep(&cp, ",") == NULL, 46);
+  equal(one+6, "gh", 47);		/* Stomped past end? */
+  equal(one, "a", 48);			/* Stomped old tokens? */
+  equal(one+2, "b", 49);
+  equal(one+4, "c", 50);
+
+  {
+    char text[] = "This,is,a,test";
+    char *list = strdupa (text);
+    equal (strsep (&list, ","), "This", 51);
+    equal (strsep (&list, ","), "is", 52);
+    equal (strsep (&list, ","), "a", 53);
+    equal (strsep (&list, ","), "test", 54);
+    check (strsep (&list, ",") == NULL, 55);
+  }
+
+  cp = strcpy(one, "a,b, c,, ,d,");
+  equal(strsep(&cp, ","), "a", 56);	/* Different separators. */
+  equal(strsep(&cp, ","), "b", 57);
+  equal(strsep(&cp, ","), " c", 58);	/* Permute list too. */
+  equal(strsep(&cp, ","), "", 59);
+  equal(strsep(&cp, ","), " ", 60);
+  equal(strsep(&cp, ","), "d", 61);
+  equal(strsep(&cp, ","), "", 62);
+  check(strsep(&cp, ",") == NULL, 63);
+  check(strsep(&cp, ",") == NULL, 64);	/* Persistence. */
+
+  cp = strcpy(one, "a,b, c,, ,d,");
+  equal(strsep(&cp, "xy,"), "a", 65);	/* Different separators. */
+  equal(strsep(&cp, "x,y"), "b", 66);
+  equal(strsep(&cp, ",xy"), " c", 67);	/* Permute list too. */
+  equal(strsep(&cp, "xy,"), "", 68);
+  equal(strsep(&cp, "x,y"), " ", 69);
+  equal(strsep(&cp, ",xy"), "d", 70);
+  equal(strsep(&cp, "xy,"), "", 71);
+  check(strsep(&cp, "x,y") == NULL, 72);
+  check(strsep(&cp, ",xy") == NULL, 73);	/* Persistence. */
+
+  cp = strcpy(one, "ABC");
+  one[4] = ':';
+  equal(strsep(&cp, "C"), "AB", 74);	/* Access beyond NUL.  */
+  ptr = strsep(&cp, ":");
+  equal(ptr, "", 75);
+  check(ptr == one + 3, 76);
+  check(cp == NULL, 77);
+
+  cp = strcpy(one, "ABC");
+  one[4] = ':';
+  equal(strsep(&cp, "CD"), "AB", 78);	/* Access beyond NUL.  */
+  ptr = strsep(&cp, ":.");
+  equal(ptr, "", 79);
+  check(ptr == one + 3, 80);
+
+  cp = strcpy(one, "ABC");		/* No token in string.  */
+  equal(strsep(&cp, ","), "ABC", 81);
+  check(cp == NULL, 82);
+
+  *one = '\0';				/* Empty string. */
+  cp = one;
+  ptr = strsep(&cp, ",");
+  equal(ptr, "", 83);
+  check(ptr == one, 84);
+  check(cp == NULL, 85);
+
+  *one = '\0';				/* Empty string and no token. */
+  cp = one;
+  ptr = strsep(&cp, "");
+  equal(ptr, "", 86);
+  check(ptr == one , 87);
+  check(cp == NULL, 88);
+}
+
+static void
+test_memcmp (void)
+{
+  it = "memcmp";
+  check(memcmp("a", "a", 1) == 0, 1);		/* Identity. */
+  check(memcmp("abc", "abc", 3) == 0, 2);	/* Multicharacter. */
+  check(memcmp("abcd", "abce", 4) < 0, 3);	/* Honestly unequal. */
+  check(memcmp("abce", "abcd", 4) > 0, 4);
+  check(memcmp("alph", "beta", 4) < 0, 5);
+  check(memcmp("a\203", "a\003", 2) > 0, 6);
+  check(memcmp("abce", "abcd", 3) == 0, 7);	/* Count limited. */
+  check(memcmp("abc", "def", 0) == 0, 8);	/* Zero count. */
+}
+
+static void
+test_memchr (void)
+{
+  it = "memchr";
+  check(memchr("abcd", 'z', 4) == NULL, 1);	/* Not found. */
+  (void) strcpy(one, "abcd");
+  check(memchr(one, 'c', 4) == one+2, 2);	/* Basic test. */
+  check(memchr(one, ~0xff|'c', 4) == one+2, 2);	/* ignore highorder bits. */
+  check(memchr(one, 'd', 4) == one+3, 3);	/* End of string. */
+  check(memchr(one, 'a', 4) == one, 4);	/* Beginning. */
+  check(memchr(one, '\0', 5) == one+4, 5);	/* Finding NUL. */
+  (void) strcpy(one, "ababa");
+  check(memchr(one, 'b', 5) == one+1, 6);	/* Finding first. */
+  check(memchr(one, 'b', 0) == NULL, 7);	/* Zero count. */
+  check(memchr(one, 'a', 1) == one, 8);	/* Singleton case. */
+  (void) strcpy(one, "a\203b");
+  check(memchr(one, 0203, 3) == one+1, 9);	/* Unsignedness. */
+
+  /* now test all possible alignment and length combinations to catch
+     bugs due to unrolled loops (assuming unrolling is limited to no
+     more than 128 byte chunks: */
+  {
+    char buf[128 + sizeof(long)];
+    long align, len, i, pos;
+
+    for (align = 0; align < (long) sizeof(long); ++align) {
+      for (len = 0; len < (long) (sizeof(buf) - align); ++len) {
+	for (i = 0; i < len; ++i) {
+	  buf[align + i] = 'x';		/* don't depend on memset... */
+	}
+	for (pos = 0; pos < len; ++pos) {
+#if 0
+	  printf("align %d, len %d, pos %d\n", align, len, pos);
+#endif
+	  check(memchr(buf + align, 'x', len) == buf + align + pos, 10);
+	  check(memchr(buf + align, 'x', pos) == NULL, 11);
+	  buf[align + pos] = '-';
+	}
+      }
+    }
+  }
+}
+
+static void
+test_memcpy (void)
+{
+  int i;
+  it = "memcpy";
+  check(memcpy(one, "abc", 4) == one, 1);	/* Returned value. */
+  equal(one, "abc", 2);			/* Did the copy go right? */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) memcpy(one+1, "xyz", 2);
+  equal(one, "axydefgh", 3);		/* Basic test. */
+
+  (void) strcpy(one, "abc");
+  (void) memcpy(one, "xyz", 0);
+  equal(one, "abc", 4);			/* Zero-length copy. */
+
+  (void) strcpy(one, "hi there");
+  (void) strcpy(two, "foo");
+  (void) memcpy(two, one, 9);
+  equal(two, "hi there", 5);		/* Just paranoia. */
+  equal(one, "hi there", 6);		/* Stomped on source? */
+
+  for (i = 0; i < 16; i++)
+    {
+      const char *x = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
+      strcpy (one, x);
+      check (memcpy (one + i, "hi there", 9) == one + i,
+	     7 + (i * 6));		/* Unaligned destination. */
+      check (memcmp (one, x, i) == 0, 8 + (i * 6));  /* Wrote under? */
+      equal (one + i, "hi there", 9 + (i * 6));
+      check (one[i + 9] == 'x', 10 + (i * 6));       /* Wrote over? */
+      check (memcpy (two, one + i, 9) == two,
+	     11 + (i * 6));		/* Unaligned source. */
+      equal (two, "hi there", 12 + (i * 6));
+    }
+}
+
+static void
+test_mempcpy (void)
+{
+  int i;
+  it = "mempcpy";
+  check(mempcpy(one, "abc", 4) == one + 4, 1);	/* Returned value. */
+  equal(one, "abc", 2);			/* Did the copy go right? */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) mempcpy(one+1, "xyz", 2);
+  equal(one, "axydefgh", 3);		/* Basic test. */
+
+  (void) strcpy(one, "abc");
+  (void) mempcpy(one, "xyz", 0);
+  equal(one, "abc", 4);			/* Zero-length copy. */
+
+  (void) strcpy(one, "hi there");
+  (void) strcpy(two, "foo");
+  (void) mempcpy(two, one, 9);
+  equal(two, "hi there", 5);		/* Just paranoia. */
+  equal(one, "hi there", 6);		/* Stomped on source? */
+
+  for (i = 0; i < 16; i++)
+    {
+      const char *x = "xxxxxxxxxxxxxxxxxxxxxxxxxxxxxx";
+      strcpy (one, x);
+      check (mempcpy (one + i, "hi there", 9) == one + i + 9,
+	     7 + (i * 6));		/* Unaligned destination. */
+      check (memcmp (one, x, i) == 0, 8 + (i * 6));  /* Wrote under? */
+      equal (one + i, "hi there", 9 + (i * 6));
+      check (one[i + 9] == 'x', 10 + (i * 6));       /* Wrote over? */
+      check (mempcpy (two, one + i, 9) == two + 9,
+	     11 + (i * 6));		/* Unaligned source. */
+      equal (two, "hi there", 12 + (i * 6));
+    }
+}
+
+static void
+test_memmove (void)
+{
+  it = "memmove";
+  check(memmove(one, "abc", 4) == one, 1);	/* Returned value. */
+  equal(one, "abc", 2);			/* Did the copy go right? */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) memmove(one+1, "xyz", 2);
+  equal(one, "axydefgh", 3);		/* Basic test. */
+
+  (void) strcpy(one, "abc");
+  (void) memmove(one, "xyz", 0);
+  equal(one, "abc", 4);			/* Zero-length copy. */
+
+  (void) strcpy(one, "hi there");
+  (void) strcpy(two, "foo");
+  (void) memmove(two, one, 9);
+  equal(two, "hi there", 5);		/* Just paranoia. */
+  equal(one, "hi there", 6);		/* Stomped on source? */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) memmove(one+1, one, 9);
+  equal(one, "aabcdefgh", 7);		/* Overlap, right-to-left. */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) memmove(one+1, one+2, 7);
+  equal(one, "acdefgh", 8);		/* Overlap, left-to-right. */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) memmove(one, one, 9);
+  equal(one, "abcdefgh", 9);		/* 100% overlap. */
+}
+
+static void
+test_memccpy (void)
+{
+  /* First test like memcpy, then the search part The SVID, the only
+     place where memccpy is mentioned, says overlap might fail, so we
+     don't try it.  Besides, it's hard to see the rationale for a
+     non-left-to-right memccpy.  */
+  it = "memccpy";
+  check(memccpy(one, "abc", 'q', 4) == NULL, 1);	/* Returned value. */
+  equal(one, "abc", 2);			/* Did the copy go right? */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) memccpy(one+1, "xyz", 'q', 2);
+  equal(one, "axydefgh", 3);		/* Basic test. */
+
+  (void) strcpy(one, "abc");
+  (void) memccpy(one, "xyz", 'q', 0);
+  equal(one, "abc", 4);			/* Zero-length copy. */
+
+  (void) strcpy(one, "hi there");
+  (void) strcpy(two, "foo");
+  (void) memccpy(two, one, 'q', 9);
+  equal(two, "hi there", 5);		/* Just paranoia. */
+  equal(one, "hi there", 6);		/* Stomped on source? */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) strcpy(two, "horsefeathers");
+  check(memccpy(two, one, 'f', 9) == two+6, 7);	/* Returned value. */
+  equal(one, "abcdefgh", 8);		/* Source intact? */
+  equal(two, "abcdefeathers", 9);		/* Copy correct? */
+
+  (void) strcpy(one, "abcd");
+  (void) strcpy(two, "bumblebee");
+  check(memccpy(two, one, 'a', 4) == two+1, 10);	/* First char. */
+  equal(two, "aumblebee", 11);
+  check(memccpy(two, one, 'd', 4) == two+4, 12);	/* Last char. */
+  equal(two, "abcdlebee", 13);
+  (void) strcpy(one, "xyz");
+  check(memccpy(two, one, 'x', 1) == two+1, 14);	/* Singleton. */
+  equal(two, "xbcdlebee", 15);
+}
+
+static void
+test_memset (void)
+{
+  int i;
+
+  it = "memset";
+  (void) strcpy(one, "abcdefgh");
+  check(memset(one+1, 'x', 3) == one+1, 1);	/* Return value. */
+  equal(one, "axxxefgh", 2);		/* Basic test. */
+
+  (void) memset(one+2, 'y', 0);
+  equal(one, "axxxefgh", 3);		/* Zero-length set. */
+
+  (void) memset(one+5, 0, 1);
+  equal(one, "axxxe", 4);			/* Zero fill. */
+  equal(one+6, "gh", 5);			/* And the leftover. */
+
+  (void) memset(one+2, 010045, 1);
+  equal(one, "ax\045xe", 6);		/* Unsigned char convert. */
+
+  /* Non-8bit fill character.  */
+  memset (one, 0x101, sizeof (one));
+  for (i = 0; i < (int) sizeof (one); ++i)
+    check (one[i] == '\01', 7);
+
+  /* Test for more complex versions of memset, for all alignments and
+     lengths up to 256. This test takes a little while, perhaps it should
+     be made weaker?  */
+  {
+    char data[512];
+    int j;
+    int k;
+    int c;
+
+    for (i = 0; i < 512; i++)
+      data[i] = 'x';
+    for (c = 0; c <= 'y'; c += 'y')  /* check for memset(,0,) and
+					memset(,'y',) */
+      for (j = 0; j < 256; j++)
+	for (i = 0; i < 256; i++)
+	  {
+	    memset (data + i, c, j);
+	    for (k = 0; k < i; k++)
+	      if (data[k] != 'x')
+		goto fail;
+	    for (k = i; k < i+j; k++)
+	      {
+		if (data[k] != c)
+		  goto fail;
+		data[k] = 'x';
+	      }
+	    for (k = i+j; k < 512; k++)
+	      if (data[k] != 'x')
+		goto fail;
+	    continue;
+
+	  fail:
+	    check (0, 8 + i + j * 256 + (c != 0) * 256 * 256);
+	  }
+  }
+}
+
+static void
+test_bcopy (void)
+{
+  /* Much like memcpy.  Berklix manual is silent about overlap, so
+     don't test it.  */
+  it = "bcopy";
+  (void) bcopy("abc", one, 4);
+  equal(one, "abc", 1);			/* Simple copy. */
+
+  (void) strcpy(one, "abcdefgh");
+  (void) bcopy("xyz", one+1, 2);
+  equal(one, "axydefgh", 2);		/* Basic test. */
+
+  (void) strcpy(one, "abc");
+  (void) bcopy("xyz", one, 0);
+  equal(one, "abc", 3);			/* Zero-length copy. */
+
+  (void) strcpy(one, "hi there");
+  (void) strcpy(two, "foo");
+  (void) bcopy(one, two, 9);
+  equal(two, "hi there", 4);		/* Just paranoia. */
+  equal(one, "hi there", 5);		/* Stomped on source? */
+}
+
+static void
+test_bzero (void)
+{
+  it = "bzero";
+  (void) strcpy(one, "abcdef");
+  bzero(one+2, 2);
+  equal(one, "ab", 1);			/* Basic test. */
+  equal(one+3, "", 2);
+  equal(one+4, "ef", 3);
+
+  (void) strcpy(one, "abcdef");
+  bzero(one+2, 0);
+  equal(one, "abcdef", 4);		/* Zero-length copy. */
+}
+
+static void
+test_strndup (void)
+{
+  char *p, *q;
+  it = "strndup";
+  p = strndup("abcdef", 12);
+  check(p != NULL, 1);
+  if (p != NULL)
+    {
+      equal(p, "abcdef", 2);
+      q = strndup(p + 1, 2);
+      check(q != NULL, 3);
+      if (q != NULL)
+	equal(q, "bc", 4);
+      free (q);
+    }
+  free (p);
+  p = strndup("abc def", 3);
+  check(p != NULL, 5);
+  if (p != NULL)
+    equal(p, "abc", 6);
+  free (p);
+}
+
+static void
+test_bcmp (void)
+{
+  it = "bcmp";
+  check(bcmp("a", "a", 1) == 0, 1);	/* Identity. */
+  check(bcmp("abc", "abc", 3) == 0, 2);	/* Multicharacter. */
+  check(bcmp("abcd", "abce", 4) != 0, 3);	/* Honestly unequal. */
+  check(bcmp("abce", "abcd", 4) != 0, 4);
+  check(bcmp("alph", "beta", 4) != 0, 5);
+  check(bcmp("abce", "abcd", 3) == 0, 6);	/* Count limited. */
+  check(bcmp("abc", "def", 0) == 0, 8);	/* Zero count. */
+}
+
+static void
+test_strerror (void)
+{
+  it = "strerror";
+  check(strerror(EDOM) != 0, 1);
+  check(strerror(ERANGE) != 0, 2);
+  check(strerror(ENOENT) != 0, 3);
+}
+
+int
+main (void)
+{
+  int status;
+
+  /* Test strcmp first because we use it to test other things.  */
+  test_strcmp ();
+
+  /* Test strcpy next because we need it to set up other tests.  */
+  test_strcpy ();
+
+  /* A closely related function is stpcpy.  */
+  test_stpcpy ();
+
+  /* stpncpy.  */
+  test_stpncpy ();
+
+  /* strcat.  */
+  test_strcat ();
+
+  /* strncat.  */
+  test_strncat ();
+
+  /* strncmp.  */
+  test_strncmp ();
+
+  /* strncpy.  */
+  test_strncpy ();
+
+  /* strlen.  */
+  test_strlen ();
+
+  /* strnlen.  */
+  test_strnlen ();
+
+  /* strchr.  */
+  test_strchr ();
+
+  /* strchrnul.  */
+  test_strchrnul ();
+
+  /* rawmemchr.  */
+  test_rawmemchr ();
+
+  /* index - just like strchr.  */
+  test_index ();
+
+  /* strrchr.  */
+  test_strrchr ();
+
+  /* memrchr.  */
+  test_memrchr ();
+
+  /* rindex - just like strrchr.  */
+  test_rindex ();
+
+  /* strpbrk - somewhat like strchr.  */
+  test_strpbrk ();
+
+  /* strstr - somewhat like strchr.  */
+  test_strstr ();
+
+  /* strspn.  */
+  test_strspn ();
+
+  /* strcspn.  */
+  test_strcspn ();
+
+  /* strtok - the hard one.  */
+  test_strtok ();
+
+  /* strtok_r.  */
+  test_strtok_r ();
+
+  /* strsep.  */
+  test_strsep ();
+
+  /* memcmp.  */
+  test_memcmp ();
+
+  /* memchr.  */
+  test_memchr ();
+
+  /* memcpy - need not work for overlap.  */
+  test_memcpy ();
+
+  /* memmove - must work on overlap.  */
+  test_memmove ();
+
+  /* mempcpy */
+  test_mempcpy ();
+
+  /* memccpy.  */
+  test_memccpy ();
+
+  /* memset.  */
+  test_memset ();
+
+  /* bcopy.  */
+  test_bcopy ();
+
+  /* bzero.  */
+  test_bzero ();
+
+  /* bcmp - somewhat like memcmp.  */
+  test_bcmp ();
+
+  /* strndup.  */
+  test_strndup ();
+
+  /* strerror - VERY system-dependent.  */
+  test_strerror ();
+
+
+  if (errors == 0)
+    {
+      status = EXIT_SUCCESS;
+      puts("No errors.");
+    }
+  else
+    {
+      status = EXIT_FAILURE;
+      printf("%lu errors.\n", (unsigned long)errors);
+    }
+
+  return status;
+}
diff --git a/libc/string/x86_64/strlen.S b/libc/string/x86_64/strlen.S
new file mode 100644
index 000000000..36b63031e
--- /dev/null
+++ b/libc/string/x86_64/strlen.S
@@ -0,0 +1,135 @@
+/* strlen(str) -- determine the length of the string STR.
+   Copyright (C) 2002, 2003 Free Software Foundation, Inc.
+   Based on i486 version contributed by Ulrich Drepper <drepper@redhat.com>.
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+
+	.text
+ENTRY (strlen)
+	movq %rdi, %rcx		/* Duplicate source pointer. */
+	andl $7, %ecx		/* mask alignment bits */
+	movq %rdi, %rax		/* duplicate destination.  */
+	jz 1f			/* aligned => start loop */
+
+	neg %ecx		/* We need to align to 8 bytes.  */
+	addl $8,%ecx
+	/* Search the first bytes directly.  */
+0:	cmpb $0x0,(%rax)	/* is byte NUL? */
+	je 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+	decl %ecx
+	jnz 0b
+
+1:	movq $0xfefefefefefefeff,%r8 /* Save magic.  */
+
+	.p2align 4		/* Align loop.  */
+4:	/* Main Loop is unrolled 4 times.  */
+	/* First unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found NUL => return pointer */
+
+	/* Second unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found NUL => return pointer */
+
+	/* Third unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jnz 3f			/* found NUL => return pointer */
+
+	/* Fourth unroll.  */
+	movq (%rax), %rcx	/* get double word (= 8 bytes) in question */
+	addq $8,%rax		/* adjust pointer for next word */
+	movq %r8, %rdx		/* magic value */
+	addq %rcx, %rdx		/* add the magic value to the word.  We get
+				   carry bits reported for each byte which
+				   is *not* 0 */
+	jnc 3f			/* highest byte is NUL => return pointer */
+	xorq %rcx, %rdx		/* (word+magic)^word */
+	orq %r8, %rdx		/* set all non-carry bits */
+	incq %rdx		/* add 1: if one carry bit was *not* set
+				   the addition will not result in 0.  */
+	jz 4b			/* no NUL found => continue loop */
+
+	.p2align 4		/* Align, it's a jump target.  */
+3:	subq $8,%rax		/* correct pointer increment.  */
+
+	testb %cl, %cl		/* is first byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is second byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testl $0x00ff0000, %ecx /* is third byte NUL? */
+	jz 2f			/* yes => return pointer */
+	incq %rax		/* increment pointer */
+
+	testl $0xff000000, %ecx /* is fourth byte NUL? */
+	jz 2f			/* yes => return pointer */
+	incq %rax		/* increment pointer */
+
+	shrq $32, %rcx		/* look at other half.  */
+
+	testb %cl, %cl		/* is first byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testb %ch, %ch		/* is second byte NUL? */
+	jz 2f			/* yes => return */
+	incq %rax		/* increment pointer */
+
+	testl $0xff0000, %ecx	/* is third byte NUL? */
+	jz 2f			/* yes => return pointer */
+	incq %rax		/* increment pointer */
+2:
+	subq %rdi, %rax		/* compute difference to string start */
+	ret
+END (strlen)
diff --git a/libc/string/x86_64/strpbrk.S b/libc/string/x86_64/strpbrk.S
new file mode 100644
index 000000000..c49cd2c06
--- /dev/null
+++ b/libc/string/x86_64/strpbrk.S
@@ -0,0 +1,2 @@
+#define strcspn strpbrk
+#include "strcspn.S"
diff --git a/libc/string/x86_64/strspn.S b/libc/string/x86_64/strspn.S
new file mode 100644
index 000000000..245d8e699
--- /dev/null
+++ b/libc/string/x86_64/strspn.S
@@ -0,0 +1,114 @@
+/* strspn (str, ss) -- Return the length of the initial segment of STR
+			which contains only characters from SS.
+   For AMD x86-64.
+   Copyright (C) 1994-1997, 2000, 2002, 2003, 2004, 2005
+   Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Ulrich Drepper <drepper@gnu.ai.mit.edu>.
+   Bug fixes by Alan Modra <Alan@SPRI.Levels.UniSA.Edu.Au>.
+   Adopted for x86-64 by Andreas Jaeger <aj@suse.de>.
+
+   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, write to the Free
+   Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
+   02111-1307 USA.  */
+
+#include "_glibc_inc.h"
+
+	.text
+ENTRY (strspn)
+
+	movq %rdi, %rdx		/* Save SRC.  */
+
+	/* First we create a table with flags for all possible characters.
+	   For the ASCII (7bit/8bit) or ISO-8859-X character sets which are
+	   supported by the C string functions we have 256 characters.
+	   Before inserting marks for the stop characters we clear the whole
+	   table.  */
+	movq %rdi, %r8			/* Save value.  */
+	subq $256, %rsp			/* Make space for 256 bytes.  */
+	movl $32,  %ecx			/* 32*8 bytes = 256 bytes.  */
+	movq %rsp, %rdi
+	xorl %eax, %eax			/* We store 0s.  */
+	cld
+	rep
+	stosq
+
+	movq %rsi, %rax			/* Setup stopset.  */
+
+/* For understanding the following code remember that %rcx == 0 now.
+   Although all the following instruction only modify %cl we always
+   have a correct zero-extended 64-bit value in %rcx.  */
+
+	.p2align 4
+L(2):	movb (%rax), %cl	/* get byte from stopset */
+	testb %cl, %cl		/* is NUL char? */
+	jz L(1)			/* yes => start compare loop */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in stopset table */
+
+	movb 1(%rax), %cl	/* get byte from stopset */
+	testb $0xff, %cl	/* is NUL char? */
+	jz L(1)			/* yes => start compare loop */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in stopset table */
+
+	movb 2(%rax), %cl	/* get byte from stopset */
+	testb $0xff, %cl	/* is NUL char? */
+	jz L(1)			/* yes => start compare loop */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in stopset table */
+
+	movb 3(%rax), %cl	/* get byte from stopset */
+	addq $4, %rax		/* increment stopset pointer */
+	movb %cl, (%rsp,%rcx)	/* set corresponding byte in stopset table */
+	testb $0xff, %cl	/* is NUL char? */
+	jnz L(2)		/* no => process next dword from stopset */
+
+L(1):	leaq -4(%rdx), %rax	/* prepare loop */
+
+	/* We use a neat trick for the following loop.  Normally we would
+	   have to test for two termination conditions
+	   1. a character in the stopset was found
+	   and
+	   2. the end of the string was found
+	   But as a sign that the character is in the stopset we store its
+	   value in the table.  But the value of NUL is NUL so the loop
+	   terminates for NUL in every case.  */
+
+	.p2align 4
+L(3):	addq $4, %rax		/* adjust pointer for full loop round */
+
+	movb (%rax), %cl	/* get byte from string */
+	testb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	jz L(4)			/* no => return */
+
+	movb 1(%rax), %cl	/* get byte from string */
+	testb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	jz L(5)			/* no => return */
+
+	movb 2(%rax), %cl	/* get byte from string */
+	testb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	jz L(6)			/* no => return */
+
+	movb 3(%rax), %cl	/* get byte from string */
+	testb %cl, (%rsp,%rcx)	/* is it contained in skipset? */
+	jnz L(3)		/* yes => start loop again */
+
+	incq %rax		/* adjust pointer */
+L(6):	incq %rax
+L(5):	incq %rax
+
+L(4):	addq $256, %rsp		/* remove stopset */
+	subq %rdx, %rax		/* we have to return the number of valid
+				   characters, so compute distance to first
+				   non-valid character */
+	ret
+END (strspn)