Add a couple of mips-specific string funcs.

Port the generic optimized string funcs from glibc, with some tweaks
  to cut their size a little.  The main change is making memmove
  call memcpy for forward copying to trim redundant code.
Make use of both the generic and arch-specific speed-optimized string
  funcs configurable.  Arch-specific take precedence over generic,
  and generic takes precedence over basic size-optimized uClibc funcs.

34 files changed, 3706 insertions, 2 deletions

diff --git a/libc/string/Makefile b/libc/string/Makefile
index 37a57cc26..02c86d6aa 100644
--- a/libc/string/Makefile
+++ b/libc/string/Makefile
@@ -20,10 +20,18 @@ TOPDIR=../../
 include $(TOPDIR)Rules.mak
 
 DIRS=
+
+ifeq ($(UCLIBC_HAS_STRING_GENERIC_OPT),y)
+DIRS += generic
+endif
+
+ifeq ($(UCLIBC_HAS_STRING_ARCH_OPT),y)
 ifeq ($(TARGET_ARCH),$(wildcard $(TARGET_ARCH)))
-DIRS = $(TARGET_ARCH)
+DIRS += $(TARGET_ARCH)
 endif
-ALL_SUBDIRS = arm frv i386 sh64 powerpc
+endif
+
+ALL_SUBDIRS = generic arm frv i386 sh64 powerpc mips
 
 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/generic/Makefile b/libc/string/generic/Makefile
new file mode 100644
index 000000000..3f9af72a3
--- /dev/null
+++ b/libc/string/generic/Makefile
@@ -0,0 +1,48 @@
+# Makefile for uClibc
+#
+# Copyright (C) 2000-2003 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
+
+TOPDIR=../../../
+include $(TOPDIR)Rules.mak
+
+CSRC=	memchr.c memcmp.c memcpy.c memmem.c memmove.c mempcpy.c memrchr.c \
+	memset.c rawmemchr.c strcat.c strchr.c strchrnul.c strcmp.c strcpy.c \
+	strcspn.c strlen.c strncat.c strncmp.c strncpy.c strnlen.c \
+	strrchr.c strsep.c strspn.c strstr.c strtok_r.c
+
+COBJS=$(patsubst %.c,%.o, $(CSRC))
+
+OBJS=$(COBJS)
+
+all: $(OBJS) $(LIBC)
+
+$(LIBC): ar-target
+
+ar-target: $(OBJS)
+	$(AR) $(ARFLAGS) $(LIBC) $(OBJS)
+
+# $(MOBJ): $(MSRC)
+# 	$(CC) $(CFLAGS) -DL_$* $< -c -o $*.o
+# 	$(STRIPTOOL) -x -R .note -R .comment $*.o
+
+$(COBJS): %.o : %.c
+	$(CC) $(CFLAGS) -c $< -o $@
+	$(STRIPTOOL) -x -R .note -R .comment $*.o
+
+clean:
+	$(RM) *.[oa] *~ core
+
diff --git a/libc/string/generic/bp-checks.h b/libc/string/generic/bp-checks.h
new file mode 100644
index 000000000..efbb84716
--- /dev/null
+++ b/libc/string/generic/bp-checks.h
@@ -0,0 +1,129 @@
+/* Bounded-pointer checking macros for C.
+   Copyright (C) 2000 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Greg McGary <greg@mcgary.org>
+
+   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 _bp_checks_h_
+#define _bp_checks_h_ 1
+
+#if __BOUNDED_POINTERS__
+
+# define BOUNDS_VIOLATED (__builtin_trap (), 0)
+
+/* Verify that pointer's value >= low.  Return pointer value.  */
+# define CHECK_BOUNDS_LOW(ARG)					\
+  (((__ptrvalue (ARG) < __ptrlow (ARG)) && BOUNDS_VIOLATED),	\
+   __ptrvalue (ARG))
+
+/* Verify that pointer's value < high.  Return pointer value.  */
+# define CHECK_BOUNDS_HIGH(ARG)				\
+  (((__ptrvalue (ARG) > __ptrhigh (ARG)) && BOUNDS_VIOLATED),	\
+   __ptrvalue (ARG))
+
+# define _CHECK_N(ARG, N, COND)				\
+  (((COND)						\
+    && (__ptrvalue (ARG) < __ptrlow (ARG)		\
+	|| __ptrvalue (ARG) + (N) > __ptrhigh (ARG))	\
+    && BOUNDS_VIOLATED),				\
+   __ptrvalue (ARG))
+
+extern void *__unbounded __ubp_memchr (const void *__unbounded, int, unsigned);
+
+# define _CHECK_STRING(ARG, COND)				\
+  (((COND)							\
+    && (__ptrvalue (ARG) < __ptrlow (ARG)			\
+	|| !__ubp_memchr (__ptrvalue (ARG), '\0',			\
+		      (__ptrhigh (ARG) - __ptrvalue (ARG))))	\
+    && BOUNDS_VIOLATED),					\
+   __ptrvalue (ARG))
+
+/* Check bounds of a pointer seated to an array of N objects.  */
+# define CHECK_N(ARG, N) _CHECK_N ((ARG), (N), 1)
+/* Same as CHECK_N, but tolerate ARG == NULL.  */
+# define CHECK_N_NULL_OK(ARG, N) _CHECK_N ((ARG), (N), __ptrvalue (ARG))
+
+/* Check bounds of a pointer seated to a single object.  */
+# define CHECK_1(ARG) CHECK_N ((ARG), 1)
+/* Same as CHECK_1, but tolerate ARG == NULL.  */
+# define CHECK_1_NULL_OK(ARG) CHECK_N_NULL_OK ((ARG), 1)
+
+/* Check for NUL-terminator within string's bounds.  */
+# define CHECK_STRING(ARG) _CHECK_STRING ((ARG), 1)
+/* Same as CHECK_STRING, but tolerate ARG == NULL.  */
+# define CHECK_STRING_NULL_OK(ARG) _CHECK_STRING ((ARG), __ptrvalue (ARG))
+
+/* Check bounds of signal syscall args with type sigset_t.  */
+# define CHECK_SIGSET(SET) CHECK_N ((SET), _NSIG / (8 * sizeof *(SET)))
+/* Same as CHECK_SIGSET, but tolerate SET == NULL.  */
+# define CHECK_SIGSET_NULL_OK(SET) CHECK_N_NULL_OK ((SET), _NSIG / (8 * sizeof *(SET)))
+
+# if defined (_IOC_SIZESHIFT) && defined (_IOC_SIZEBITS)
+/* Extract the size of the ioctl data and check its bounds.  */
+#  define CHECK_IOCTL(ARG, CMD)						\
+  CHECK_N ((const char *) (ARG),					\
+	   (((CMD) >> _IOC_SIZESHIFT) & ((1 << _IOC_SIZEBITS) - 1)))
+# else
+/* We don't know the size of the ioctl data, so the best we can do
+   is check that the first byte is within bounds.  */
+#  define CHECK_IOCTL(ARG, CMD) CHECK_1 ((const char *) ARG)
+# endif
+
+/* Check bounds of `struct flock *' for the locking fcntl commands.  */
+# define CHECK_FCNTL(ARG, CMD)					\
+  (((CMD) == F_GETLK || (CMD) == F_SETLK || (CMD) == F_SETLKW)	\
+   ? CHECK_1 ((struct flock *) ARG) : (unsigned long) (ARG))
+
+/* Check bounds of an array of mincore residency-status flags that
+   cover a region of NBYTES.  Such a vector occupies one byte per page
+   of memory.  */
+# define CHECK_N_PAGES(ARG, NBYTES)				\
+  ({ int _page_size_ = sysconf (_SC_PAGE_SIZE);			\
+     CHECK_N ((const char *) (ARG),				\
+	      ((NBYTES) + _page_size_ - 1) / _page_size_); })
+
+/* Return a bounded pointer with value PTR that satisfies CHECK_N (PTR, N).  */
+# define BOUNDED_N(PTR, N) 				\
+  ({ __typeof (PTR) __bounded _p_;			\
+     __ptrvalue _p_ = __ptrlow _p_ = __ptrvalue (PTR);	\
+     __ptrhigh _p_ = __ptrvalue _p_ + (N);		\
+     _p_; })
+
+#else /* !__BOUNDED_POINTERS__ */
+
+/* Do nothing if not compiling with -fbounded-pointers.  */
+
+# define BOUNDS_VIOLATED
+# define CHECK_BOUNDS_LOW(ARG) (ARG)
+# define CHECK_BOUNDS_HIGH(ARG) (ARG)
+# define CHECK_1(ARG) (ARG)
+# define CHECK_1_NULL_OK(ARG) (ARG)
+# define CHECK_N(ARG, N) (ARG)
+# define CHECK_N_NULL_OK(ARG, N) (ARG)
+# define CHECK_STRING(ARG) (ARG)
+# define CHECK_SIGSET(SET) (SET)
+# define CHECK_SIGSET_NULL_OK(SET) (SET)
+# define CHECK_IOCTL(ARG, CMD) (ARG)
+# define CHECK_FCNTL(ARG, CMD) (ARG)
+# define CHECK_N_PAGES(ARG, NBYTES) (ARG)
+# define BOUNDED_N(PTR, N) (PTR)
+
+#endif /* !__BOUNDED_POINTERS__ */
+
+#define BOUNDED_1(PTR) BOUNDED_N (PTR, 1)
+
+#endif /* _bp_checks_h_ */
diff --git a/libc/string/generic/memchr.c b/libc/string/generic/memchr.c
new file mode 100644
index 000000000..6116b8f92
--- /dev/null
+++ b/libc/string/generic/memchr.c
@@ -0,0 +1,175 @@
+/* Copyright (C) 1991,93,96,97,99,2000,2003 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Based on strlen implementation by Torbjorn Granlund (tege@sics.se),
+   with help from Dan Sahlin (dan@sics.se) and
+   commentary by Jim Blandy (jimb@ai.mit.edu);
+   adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu),
+   and implemented by Roland McGrath (roland@ai.mit.edu).
+
+   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 <string.h>
+#include <stdlib.h>
+#include <limits.h>
+
+#include "memcopy.h"
+
+#define LONG_MAX_32_BITS 2147483647
+
+#undef memchr
+
+/* Search no more than N bytes of S for C.  */
+void * memchr (const void * s, int c_in, size_t n)
+{
+  const unsigned char *char_ptr;
+  const unsigned long int *longword_ptr;
+  unsigned long int longword, magic_bits, charmask;
+  unsigned reg_char c;
+
+  c = (unsigned char) c_in;
+
+  /* Handle the first few characters by reading one character at a time.
+     Do this until CHAR_PTR is aligned on a longword boundary.  */
+  for (char_ptr = (const unsigned char *) s;
+       n > 0 && ((unsigned long int) char_ptr
+		 & (sizeof (longword) - 1)) != 0;
+       --n, ++char_ptr)
+    if (*char_ptr == c)
+      return (void *) char_ptr;
+
+  /* All these elucidatory comments refer to 4-byte longwords,
+     but the theory applies equally well to 8-byte longwords.  */
+
+  longword_ptr = (unsigned long int *) char_ptr;
+
+  /* Bits 31, 24, 16, and 8 of this number are zero.  Call these bits
+     the "holes."  Note that there is a hole just to the left of
+     each byte, with an extra at the end:
+
+     bits:  01111110 11111110 11111110 11111111
+     bytes: AAAAAAAA BBBBBBBB CCCCCCCC DDDDDDDD
+
+     The 1-bits make sure that carries propagate to the next 0-bit.
+     The 0-bits provide holes for carries to fall into.  */
+
+  if (sizeof (longword) != 4 && sizeof (longword) != 8)
+    abort ();
+
+#if LONG_MAX <= LONG_MAX_32_BITS
+  magic_bits = 0x7efefeff;
+#else
+  magic_bits = ((unsigned long int) 0x7efefefe << 32) | 0xfefefeff;
+#endif
+
+  /* Set up a longword, each of whose bytes is C.  */
+  charmask = c | (c << 8);
+  charmask |= charmask << 16;
+#if LONG_MAX > LONG_MAX_32_BITS
+  charmask |= charmask << 32;
+#endif
+
+  /* Instead of the traditional loop which tests each character,
+     we will test a longword at a time.  The tricky part is testing
+     if *any of the four* bytes in the longword in question are zero.  */
+  while (n >= sizeof (longword))
+    {
+      /* We tentatively 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 LONGWORD 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.  If one of bits 24-30 is set, there will be a carry
+	 into bit 31, so all of the hole bits will be changed.
+
+	 The one misfire occurs when bits 24-30 are clear and bit
+	 31 is set; in this case, the hole at bit 31 is not
+	 changed.  If we had access to the processor carry flag,
+	 we could close this loophole by putting the fourth hole
+	 at bit 32!
+
+	 So it ignores everything except 128's, when they're aligned
+	 properly.
+
+	 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.  */
+
+      longword = *longword_ptr++ ^ charmask;
+
+      /* Add MAGIC_BITS to LONGWORD.  */
+      if ((((longword + magic_bits)
+
+	    /* Set those bits that were unchanged by the addition.  */
+	    ^ ~longword)
+
+	   /* Look at only the hole bits.  If any of the hole bits
+	      are unchanged, most likely one of the bytes was a
+	      zero.  */
+	   & ~magic_bits) != 0)
+	{
+	  /* Which of the bytes was C?  If none of them were, it was
+	     a misfire; continue the search.  */
+
+	  const unsigned char *cp = (const unsigned char *) (longword_ptr - 1);
+
+	  if (cp[0] == c)
+	    return (void *) cp;
+	  if (cp[1] == c)
+	    return (void *) &cp[1];
+	  if (cp[2] == c)
+	    return (void *) &cp[2];
+	  if (cp[3] == c)
+	    return (void *) &cp[3];
+#if LONG_MAX > 2147483647
+	  if (cp[4] == c)
+	    return (void *) &cp[4];
+	  if (cp[5] == c)
+	    return (void *) &cp[5];
+	  if (cp[6] == c)
+	    return (void *) &cp[6];
+	  if (cp[7] == c)
+	    return (void *) &cp[7];
+#endif
+	}
+
+      n -= sizeof (longword);
+    }
+
+  char_ptr = (const unsigned char *) longword_ptr;
+
+  while (n-- > 0)
+    {
+      if (*char_ptr == c)
+	return (void *) char_ptr;
+      else
+	++char_ptr;
+    }
+
+  return 0;
+}
diff --git a/libc/string/generic/memcmp.c b/libc/string/generic/memcmp.c
new file mode 100644
index 000000000..d9c3a7cab
--- /dev/null
+++ b/libc/string/generic/memcmp.c
@@ -0,0 +1,340 @@
+/* Copyright (C) 1991,1993,1995,1997,1998,2003,2004
+   Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Torbjorn Granlund (tege@sics.se).
+
+   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 <string.h>
+
+#undef memcmp
+
+#include "memcopy.h"
+
+#include <endian.h>
+
+#if __BYTE_ORDER == __BIG_ENDIAN
+# define WORDS_BIGENDIAN
+#endif
+
+#ifdef WORDS_BIGENDIAN
+# define CMP_LT_OR_GT(a, b) ((a) > (b) ? 1 : -1)
+#else
+# define CMP_LT_OR_GT(a, b) memcmp_bytes ((a), (b))
+#endif
+
+/* BE VERY CAREFUL IF YOU CHANGE THIS CODE!  */
+
+/* The strategy of this memcmp is:
+
+   1. Compare bytes until one of the block pointers is aligned.
+
+   2. Compare using memcmp_common_alignment or
+      memcmp_not_common_alignment, regarding the alignment of the other
+      block after the initial byte operations.  The maximum number of
+      full words (of type op_t) are compared in this way.
+
+   3. Compare the few remaining bytes.  */
+
+#ifndef WORDS_BIGENDIAN
+/* memcmp_bytes -- Compare A and B bytewise in the byte order of the machine.
+   A and B are known to be different.
+   This is needed only on little-endian machines.  */
+
+static int memcmp_bytes __P((op_t, op_t));
+
+# ifdef  __GNUC__
+__inline
+# endif
+static int
+memcmp_bytes (a, b)
+     op_t a, b;
+{
+  long int srcp1 = (long int) &a;
+  long int srcp2 = (long int) &b;
+  op_t a0, b0;
+
+  do
+    {
+      a0 = ((byte *) srcp1)[0];
+      b0 = ((byte *) srcp2)[0];
+      srcp1 += 1;
+      srcp2 += 1;
+    }
+  while (a0 == b0);
+  return a0 - b0;
+}
+#endif
+
+static int memcmp_common_alignment __P((long, long, size_t));
+
+/* memcmp_common_alignment -- Compare blocks at SRCP1 and SRCP2 with LEN `op_t'
+   objects (not LEN bytes!).  Both SRCP1 and SRCP2 should be aligned for
+   memory operations on `op_t's.  */
+static int
+memcmp_common_alignment (srcp1, srcp2, len)
+     long int srcp1;
+     long int srcp2;
+     size_t len;
+{
+  op_t a0, a1;
+  op_t b0, b1;
+
+  switch (len % 4)
+    {
+    default: /* Avoid warning about uninitialized local variables.  */
+    case 2:
+      a0 = ((op_t *) srcp1)[0];
+      b0 = ((op_t *) srcp2)[0];
+      srcp1 -= 2 * OPSIZ;
+      srcp2 -= 2 * OPSIZ;
+      len += 2;
+      goto do1;
+    case 3:
+      a1 = ((op_t *) srcp1)[0];
+      b1 = ((op_t *) srcp2)[0];
+      srcp1 -= OPSIZ;
+      srcp2 -= OPSIZ;
+      len += 1;
+      goto do2;
+    case 0:
+      if (OP_T_THRES <= 3 * OPSIZ && len == 0)
+	return 0;
+      a0 = ((op_t *) srcp1)[0];
+      b0 = ((op_t *) srcp2)[0];
+      goto do3;
+    case 1:
+      a1 = ((op_t *) srcp1)[0];
+      b1 = ((op_t *) srcp2)[0];
+      srcp1 += OPSIZ;
+      srcp2 += OPSIZ;
+      len -= 1;
+      if (OP_T_THRES <= 3 * OPSIZ && len == 0)
+	goto do0;
+      /* Fall through.  */
+    }
+
+  do
+    {
+      a0 = ((op_t *) srcp1)[0];
+      b0 = ((op_t *) srcp2)[0];
+      if (a1 != b1)
+	return CMP_LT_OR_GT (a1, b1);
+
+    do3:
+      a1 = ((op_t *) srcp1)[1];
+      b1 = ((op_t *) srcp2)[1];
+      if (a0 != b0)
+	return CMP_LT_OR_GT (a0, b0);
+
+    do2:
+      a0 = ((op_t *) srcp1)[2];
+      b0 = ((op_t *) srcp2)[2];
+      if (a1 != b1)
+	return CMP_LT_OR_GT (a1, b1);
+
+    do1:
+      a1 = ((op_t *) srcp1)[3];
+      b1 = ((op_t *) srcp2)[3];
+      if (a0 != b0)
+	return CMP_LT_OR_GT (a0, b0);
+
+      srcp1 += 4 * OPSIZ;
+      srcp2 += 4 * OPSIZ;
+      len -= 4;
+    }
+  while (len != 0);
+
+  /* This is the right position for do0.  Please don't move
+     it into the loop.  */
+ do0:
+  if (a1 != b1)
+    return CMP_LT_OR_GT (a1, b1);
+  return 0;
+}
+
+static int memcmp_not_common_alignment __P((long, long, size_t));
+
+/* memcmp_not_common_alignment -- Compare blocks at SRCP1 and SRCP2 with LEN
+   `op_t' objects (not LEN bytes!).  SRCP2 should be aligned for memory
+   operations on `op_t', but SRCP1 *should be unaligned*.  */
+static int
+memcmp_not_common_alignment (srcp1, srcp2, len)
+     long int srcp1;
+     long int srcp2;
+     size_t len;
+{
+  op_t a0, a1, a2, a3;
+  op_t b0, b1, b2, b3;
+  op_t x;
+  int shl, shr;
+
+  /* Calculate how to shift a word read at the memory operation
+     aligned srcp1 to make it aligned for comparison.  */
+
+  shl = 8 * (srcp1 % OPSIZ);
+  shr = 8 * OPSIZ - shl;
+
+  /* Make SRCP1 aligned by rounding it down to the beginning of the `op_t'
+     it points in the middle of.  */
+  srcp1 &= -OPSIZ;
+
+  switch (len % 4)
+    {
+    default: /* Avoid warning about uninitialized local variables.  */
+    case 2:
+      a1 = ((op_t *) srcp1)[0];
+      a2 = ((op_t *) srcp1)[1];
+      b2 = ((op_t *) srcp2)[0];
+      srcp1 -= 1 * OPSIZ;
+      srcp2 -= 2 * OPSIZ;
+      len += 2;
+      goto do1;
+    case 3:
+      a0 = ((op_t *) srcp1)[0];
+      a1 = ((op_t *) srcp1)[1];
+      b1 = ((op_t *) srcp2)[0];
+      srcp2 -= 1 * OPSIZ;
+      len += 1;
+      goto do2;
+    case 0:
+      if (OP_T_THRES <= 3 * OPSIZ && len == 0)
+	return 0;
+      a3 = ((op_t *) srcp1)[0];
+      a0 = ((op_t *) srcp1)[1];
+      b0 = ((op_t *) srcp2)[0];
+      srcp1 += 1 * OPSIZ;
+      goto do3;
+    case 1:
+      a2 = ((op_t *) srcp1)[0];
+      a3 = ((op_t *) srcp1)[1];
+      b3 = ((op_t *) srcp2)[0];
+      srcp1 += 2 * OPSIZ;
+      srcp2 += 1 * OPSIZ;
+      len -= 1;
+      if (OP_T_THRES <= 3 * OPSIZ && len == 0)
+	goto do0;
+      /* Fall through.  */
+    }
+
+  do
+    {
+      a0 = ((op_t *) srcp1)[0];
+      b0 = ((op_t *) srcp2)[0];
+      x = MERGE(a2, shl, a3, shr);
+      if (x != b3)
+	return CMP_LT_OR_GT (x, b3);
+
+    do3:
+      a1 = ((op_t *) srcp1)[1];
+      b1 = ((op_t *) srcp2)[1];
+      x = MERGE(a3, shl, a0, shr);
+      if (x != b0)
+	return CMP_LT_OR_GT (x, b0);
+
+    do2:
+      a2 = ((op_t *) srcp1)[2];
+      b2 = ((op_t *) srcp2)[2];
+      x = MERGE(a0, shl, a1, shr);
+      if (x != b1)
+	return CMP_LT_OR_GT (x, b1);
+
+    do1:
+      a3 = ((op_t *) srcp1)[3];
+      b3 = ((op_t *) srcp2)[3];
+      x = MERGE(a1, shl, a2, shr);
+      if (x != b2)
+	return CMP_LT_OR_GT (x, b2);
+
+      srcp1 += 4 * OPSIZ;
+      srcp2 += 4 * OPSIZ;
+      len -= 4;
+    }
+  while (len != 0);
+
+  /* This is the right position for do0.  Please don't move
+     it into the loop.  */
+ do0:
+  x = MERGE(a2, shl, a3, shr);
+  if (x != b3)
+    return CMP_LT_OR_GT (x, b3);
+  return 0;
+}
+
+int
+memcmp (s1, s2, len)
+     const __ptr_t s1;
+     const __ptr_t s2;
+     size_t len;
+{
+  op_t a0;
+  op_t b0;
+  long int srcp1 = (long int) s1;
+  long int srcp2 = (long int) s2;
+  op_t res;
+
+  if (len >= OP_T_THRES)
+    {
+      /* There are at least some bytes to compare.  No need to test
+	 for LEN == 0 in this alignment loop.  */
+      while (srcp2 % OPSIZ != 0)
+	{
+	  a0 = ((byte *) srcp1)[0];
+	  b0 = ((byte *) srcp2)[0];
+	  srcp1 += 1;
+	  srcp2 += 1;
+	  res = a0 - b0;
+	  if (res != 0)
+	    return res;
+	  len -= 1;
+	}
+
+      /* SRCP2 is now aligned for memory operations on `op_t'.
+	 SRCP1 alignment determines if we can do a simple,
+	 aligned compare or need to shuffle bits.  */
+
+      if (srcp1 % OPSIZ == 0)
+	res = memcmp_common_alignment (srcp1, srcp2, len / OPSIZ);
+      else
+	res = memcmp_not_common_alignment (srcp1, srcp2, len / OPSIZ);
+      if (res != 0)
+	return res;
+
+      /* Number of bytes remaining in the interval [0..OPSIZ-1].  */
+      srcp1 += len & -OPSIZ;
+      srcp2 += len & -OPSIZ;
+      len %= OPSIZ;
+    }
+
+  /* There are just a few bytes to compare.  Use byte memory operations.  */
+  while (len != 0)
+    {
+      a0 = ((byte *) srcp1)[0];
+      b0 = ((byte *) srcp2)[0];
+      srcp1 += 1;
+      srcp2 += 1;
+      res = a0 - b0;
+      if (res != 0)
+	return res;
+      len -= 1;
+    }
+
+  return 0;
+}
+#ifdef weak_alias
+# undef bcmp
+weak_alias (memcmp, bcmp)
+#endif
diff --git a/libc/string/generic/memcopy.h b/libc/string/generic/memcopy.h
new file mode 100644
index 000000000..df1ba9a97
--- /dev/null
+++ b/libc/string/generic/memcopy.h
@@ -0,0 +1,150 @@
+/* memcopy.h -- definitions for memory copy functions.  Generic C version.
+   Copyright (C) 1991, 1992, 1993, 1997 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+   Contributed by Torbjorn Granlund (tege@sics.se).
+
+   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.  */
+
+/* The strategy of the memory functions is:
+
+     1. Copy bytes until the destination pointer is aligned.
+
+     2. Copy words in unrolled loops.  If the source and destination
+     are not aligned in the same way, use word memory operations,
+     but shift and merge two read words before writing.
+
+     3. Copy the few remaining bytes.
+
+   This is fast on processors that have at least 10 registers for
+   allocation by GCC, and that can access memory at reg+const in one
+   instruction.
+
+   I made an "exhaustive" test of this memmove when I wrote it,
+   exhaustive in the sense that I tried all alignment and length
+   combinations, with and without overlap.  */
+
+#include <sys/cdefs.h>
+#include <endian.h>
+
+/* The macros defined in this file are:
+
+   BYTE_COPY_FWD(dst_beg_ptr, src_beg_ptr, nbytes_to_copy)
+
+   BYTE_COPY_BWD(dst_end_ptr, src_end_ptr, nbytes_to_copy)
+
+   WORD_COPY_FWD(dst_beg_ptr, src_beg_ptr, nbytes_remaining, nbytes_to_copy)
+
+   WORD_COPY_BWD(dst_end_ptr, src_end_ptr, nbytes_remaining, nbytes_to_copy)
+
+   MERGE(old_word, sh_1, new_word, sh_2)
+     [I fail to understand.  I feel stupid.  --roland]
+*/
+
+/* Type to use for aligned memory operations.
+   This should normally be the biggest type supported by a single load
+   and store.  */
+#define	op_t	unsigned long int
+#define OPSIZ	(sizeof(op_t))
+
+/* Type to use for unaligned operations.  */
+typedef unsigned char byte;
+