[PATCH] Optimized strnlen

Jakub Jelinek jakub@redhat.com
Tue Aug 21 13:30:00 GMT 2001


Hi!

The strnlen implementation you've checked in is very slow, in fact it beats
the C strnlen attached below only on very very small maxlen/strlens and only
by a few ticks.
This patch moves strnlen to sysdeps so that architectures can optimize it as
they like (e.g. if they know memchr will not pre-load words into cache and
it is a win, they can implement it using memchr, or whatever).
Attached is a program I used for testing, together with results from IA-64
(X) and i686 (Y). From this log, it is clear that modelling strnlen on top
of memchr makes no sense on i686, it is always way slower to use memchr
(both the i486+ inline memchr and i486/memchr.S).

2001-08-21  Jakub Jelinek  <jakub@redhat.com>

	* string/strnlen.c: Remove.
	* sysdeps/generic/strnlen.c: New.
	* sysdeps/i386/i486/bits/string.h (strnlen): Remove.

--- libc/string/strnlen.c.jj	Tue Aug 21 11:46:17 2001
+++ libc/string/strnlen.c	Tue Aug 21 22:33:03 2001
@@ -1,44 +0,0 @@
-/* Find the length of STRING, but scan at most MAXLEN characters.
-   Copyright (C) 1996, 1997, 1998, 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.  */
-
-#include <string.h>
-
-/* Find the length of S, but scan at most MAXLEN characters.  If no
-   '\0' terminator is found in that many characters, return MAXLEN.  */
-
-size_t
-__strnlen (const char *s, size_t maxlen)
-{
-  size_t len = 0;
-
-  while (s[len] != '\0' && maxlen > 0)
-    {
-      if (s[++len] == '\0' || --maxlen == 0)
-	return len;
-      if (s[++len] == '\0' || --maxlen == 0)
-	return len;
-      if (s[++len] == '\0' || --maxlen == 0)
-	return len;
-      ++len;
-      --maxlen;
-    }
-
-  return len;
-}
-weak_alias (__strnlen, strnlen)
--- libc/sysdeps/generic/strnlen.c.jj	Tue Aug 21 22:32:55 2001
+++ libc/sysdeps/generic/strnlen.c	Tue Aug 21 22:32:49 2001
@@ -0,0 +1,157 @@
+/* Find the length of STRING, but scan at most MAXLEN characters.
+   Copyright (C) 1991, 1993, 1997, 2000, 2001 Free Software Foundation, Inc.
+   Contributed by Jakub Jelinek <jakub@redhat.com>.
+
+   Based on strlen written by Torbjorn Granlund (tege@sics.se),
+   with help from Dan Sahlin (dan@sics.se);
+   commentary by Jim Blandy (jimb@ai.mit.edu).
+
+   The GNU C Library 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.
+
+   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
+   Library General Public License for more details.
+
+   You should have received a copy of the GNU Library General Public
+   License along with the GNU C Library; see the file COPYING.LIB.  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>
+
+/* Find the length of S, but scan at most MAXLEN characters.  If no
+   '\0' terminator is found in that many characters, return MAXLEN.  */
+size_t
+__strnlen (const char *str, size_t maxlen)
+{
+  const char *char_ptr, *end_ptr = str + maxlen;
+  const unsigned long int *longword_ptr;
+  unsigned long int longword, magic_bits, himagic, lomagic;
+
+  if (maxlen == 0)
+    return 0;
+
+  /* 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 = str; ((unsigned long int) char_ptr
+			& (sizeof (longword) - 1)) != 0;
+       ++char_ptr)
+    if (*char_ptr == '\0')
+      {
+	if (char_ptr > end_ptr)
+	  char_ptr = end_ptr;
+	return char_ptr - str;
+      }
+
+  /* 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.  */
+  magic_bits = 0x7efefeffL;
+  himagic = 0x80808080L;
+  lomagic = 0x01010101L;
+  if (sizeof (longword) > 4)
+    {
+      /* 64-bit version of the magic.  */
+      /* Do the shift in two steps to avoid a warning if long has 32 bits.  */
+      magic_bits = ((0x7efefefeL << 16) << 16) | 0xfefefeffL;
+      himagic = ((himagic << 16) << 16) | himagic;
+      lomagic = ((lomagic << 16) << 16) | lomagic;
+    }
+  if (sizeof (longword) > 8)
+    abort ();
+
+  /* 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 (longword_ptr < (unsigned long int *) end_ptr)
+    {
+      /* 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.  */
+
+      longword = *longword_ptr++;
+
+      if ((longword - lomagic) & himagic)
+	{
+	  /* Which of the bytes was the zero?  If none of them were, it was
+	     a misfire; continue the search.  */
+
+	  const char *cp = (const char *) (longword_ptr - 1);
+
+	  char_ptr = cp;
+	  if (cp[0] == 0)
+	    break;
+	  char_ptr = cp + 1;
+	  if (cp[1] == 0)
+	    break;
+	  char_ptr = cp + 2;
+	  if (cp[2] == 0)
+	    break;
+	  char_ptr = cp + 3;
+	  if (cp[3] == 0)
+	    break;
+	  if (sizeof (longword) > 4)
+	    {
+	      char_ptr = cp + 4;
+	      if (cp[4] == 0)
+		break;
+	      char_ptr = cp + 5;
+	      if (cp[5] == 0)
+		break;
+	      char_ptr = cp + 6;
+	      if (cp[6] == 0)
+		break;
+	      char_ptr = cp + 7;
+	      if (cp[7] == 0)
+		break;
+	    }
+	}
+      char_ptr = end_ptr;
+    }
+
+  if (char_ptr > end_ptr)
+    char_ptr = end_ptr;
+  return char_ptr - str;
+}
+weak_alias (__strnlen, strnlen)
--- libc/sysdeps/i386/i486/bits/string.h.jj	Tue Aug 21 22:27:42 2001
+++ libc/sysdeps/i386/i486/bits/string.h	Tue Aug 21 22:41:47 2001
@@ -540,18 +540,6 @@ __strlen_g (__const char *__str)
 }
 
 
-
-#if defined __USE_GNU
-#define _HAVE_STRING_ARCH_strnlen	1
-__STRING_INLINE size_t
-strnlen (__const char *__string, size_t __maxlen)
-{
-  __const char *__end = (__const char *) memchr (__string, '\0', __maxlen);
-  return __end ? (size_t) (__end - __string) : __maxlen;
-}
-#endif
-
-
 /* Copy SRC to DEST.  */
 #define _HAVE_STRING_ARCH_strcpy 1
 #define strcpy(dest, src) \

	Jakub
-------------- next part --------------
A non-text attachment was scrubbed...
Name: strnlen.tar.bz2
Type: application/x-bzip2
Size: 6332 bytes
Desc: not available
URL: <http://sourceware.org/pipermail/libc-hacker/attachments/20010821/5c67d8af/attachment.bz2>


More information about the Libc-hacker mailing list