Re: [PATCH] Add strchrnul

["Yaakov (Cygwin/X)" <yselkowitz at users dot sourceforge dot net>]

On Mon, 2011-03-28 at 10:10 +0200, Corinna Vinschen wrote:
> Isn't this implementation a bit simple?  It's as fast as the underlying
> strchr function in case of a hit, but it suffers quite a performance hit
> in case of a miss, because the string has to be scanned twice.
> 
> So I'm wondering, shouldn't we better adapt the strchr implementation
> and just return a pointer to s where strchr returns NULL?

There's always a tradeoff, simplicity vs. code duplication.  I presume
you mean something like the attached.  I'm just not sure about the !
PREFER_SIZE_OVER_SPEED block of the code.


Yaakov

/*
FUNCTION
	<<strchrnul>>---search for character in string

INDEX
	strchrnul

ANSI_SYNOPSIS
	#include <string.h>
	char * strchrnul(const char *<[string]>, int <[c]>);

TRAD_SYNOPSIS
	#include <string.h>
	char * strchrnul(<[string]>, <[c]>);
	const char *<[string]>;
	int <[c]>;

DESCRIPTION
	This function finds the first occurence of <[c]> (converted to
	a char) in the string pointed to by <[string]> (including the
	terminating null character).

RETURNS
	Returns a pointer to the located character, or a pointer
	to the concluding null byte if <[c]> does not occur in <[string]>.

PORTABILITY
<<strchrnul>> is a GNU extension.

<<strchrnul>> requires no supporting OS subroutines.

QUICKREF
	strchrnul
*/

#include <string.h>
#include <limits.h>

/* Nonzero if X is not aligned on a "long" boundary.  */
#define UNALIGNED(X) ((long)X & (sizeof (long) - 1))

/* How many bytes are loaded each iteration of the word copy loop.  */
#define LBLOCKSIZE (sizeof (long))

#if LONG_MAX == 2147483647L
#define DETECTNULL(X) (((X) - 0x01010101) & ~(X) & 0x80808080)
#else
#if LONG_MAX == 9223372036854775807L
/* Nonzero if X (a long int) contains a NULL byte. */
#define DETECTNULL(X) (((X) - 0x0101010101010101) & ~(X) & 0x8080808080808080)
#else
#error long int is not a 32bit or 64bit type.
#endif
#endif

/* DETECTCHAR returns nonzero if (long)X contains the byte used
   to fill (long)MASK. */
#define DETECTCHAR(X,MASK) (DETECTNULL(X ^ MASK))

char *
_DEFUN (strchrnul, (s1, i),
	_CONST char *s1 _AND
	int i)
{
  _CONST unsigned char *s = (_CONST unsigned char *)s1;
  unsigned char c = i;

#if !defined(PREFER_SIZE_OVER_SPEED) && !defined(__OPTIMIZE_SIZE__)
  unsigned long mask,j;
  unsigned long *aligned_addr;

  /* Special case for finding 0.  */
  if (!c)
    {
      while (UNALIGNED (s))
        {
          if (!*s)
            return (char *) s;
          s++;
        }
      /* Operate a word at a time.  */
      aligned_addr = (unsigned long *) s;
      while (!DETECTNULL (*aligned_addr))
        aligned_addr++;
      /* Found the end of string.  */
      s = (const unsigned char *) aligned_addr;
      while (*s)
        s++;
      return (char *) s;
    }

  /* All other bytes.  Align the pointer, then search a long at a time.  */
  while (UNALIGNED (s))
    {
      if (!*s || *s == c)
        return (char *) s;
      s++;
    }

  mask = c;
  for (j = 8; j < LBLOCKSIZE * 8; j <<= 1)
    mask = (mask << j) | mask;

  aligned_addr = (unsigned long *) s;
  while (!DETECTNULL (*aligned_addr) && !DETECTCHAR (*aligned_addr, mask))
    aligned_addr++;

  /* The block of bytes currently pointed to by aligned_addr
     contains either a null or the target char, or both.  We
     catch it using the bytewise search.  */

  s = (unsigned char *) aligned_addr;

#endif /* not PREFER_SIZE_OVER_SPEED */

  while (*s && *s != c)
    s++;
  return (char *)s;
}