PIPS
Macros | Functions
memchr.c File Reference
#include <config.h>
#include <string.h>
#include <stddef.h>
#include <limits.h>
+ Include dependency graph for memchr.c:

Go to the source code of this file.

Macros

#define reg_char   char
 Copyright (C) 1991, 1993, 1996-1997, 1999-2000, 2003-2004, 2006, 2008-2014 Free Software Foundation, Inc. More...
 
#define BP_SYM(sym)   sym
 
#define __memchr   memchr
 

Functions

void * __memchr (void const *s, int c_in, size_t n)
 Search no more than N bytes of S for C. More...
 

Macro Definition Documentation

◆ __memchr

#define __memchr   memchr

Definition at line 54 of file memchr.c.

◆ BP_SYM

#define BP_SYM (   sym)    sym

Definition at line 45 of file memchr.c.

◆ reg_char

#define reg_char   char

Copyright (C) 1991, 1993, 1996-1997, 1999-2000, 2003-2004, 2006, 2008-2014 Free Software Foundation, Inc.

Based on strlen implementation by Torbjorn Granlund (tege@.nosp@m.sics.nosp@m..se), with help from Dan Sahlin (dan@s.nosp@m.ics..nosp@m.se) and commentary by Jim Blandy (jimb@.nosp@m.ai.m.nosp@m.it.ed.nosp@m.u); adaptation to memchr suggested by Dick Karpinski (dick@.nosp@m.cca..nosp@m.ucsf..nosp@m.edu), and implemented by Roland McGrath (rolan.nosp@m.d@ai.nosp@m..mit..nosp@m.edu).

NOTE: The canonical source of this file is maintained with the GNU C Library. Bugs can be reported to bug-g.nosp@m.libc.nosp@m.@prep.nosp@m..ai..nosp@m.mit.e.nosp@m.du.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3 of the License, or 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 General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.

Definition at line 37 of file memchr.c.

Function Documentation

◆ __memchr()

void* __memchr ( void const *  s,
int  c_in,
size_t  n 
)

Search no more than N bytes of S for C.


On 32-bit hardware, choosing longword to be a 32-bit unsigned long instead of a 64-bit uintmax_t tends to give better performance. On 64-bit hardware, unsigned long is generally 64 bits already. Change this typedef to experiment with performance.

Handle the first few bytes by reading one byte at a time. Do this until CHAR_PTR is aligned on a longword boundary.

All these elucidatory comments refer to 4-byte longwords, but the theory applies equally well to any size longwords.

Compute auxiliary longword values: repeated_one is a value which has a 1 in every byte. repeated_c has c in every byte.

Instead of the traditional loop which tests each byte, 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 equal to c. We first use an xor with repeated_c. This reduces the task to testing whether any of the four bytes in longword1 is zero.

We compute tmp = ((longword1 - repeated_one) & ~longword1) & (repeated_one << 7). That is, we perform the following operations:

  1. Subtract repeated_one.
  2. & ~longword1.
  3. & a mask consisting of 0x80 in every byte. Consider what happens in each byte:
  • If a byte of longword1 is zero, step 1 and 2 transform it into 0xff, and step 3 transforms it into 0x80. A carry can also be propagated to more significant bytes.
  • If a byte of longword1 is nonzero, let its lowest 1 bit be at position k (0 <= k <= 7); so the lowest k bits are 0. After step 1, the byte ends in a single bit of value 0 and k bits of value 1. After step 2, the result is just k bits of value 1: 2^k - 1. After step 3, the result is 0. And no carry is produced. So, if longword1 has only non-zero bytes, tmp is zero. Whereas if longword1 has a zero byte, call j the position of the least significant zero byte. Then the result has a zero at positions 0, ..., j-1 and a 0x80 at position j. We cannot predict the result at the more significant bytes (positions j+1..3), but it does not matter since we already have a non-zero bit at position 8*j+7.

So, the test whether any byte in longword1 is zero is equivalent to testing whether tmp is nonzero.

At this point, we know that either n < sizeof (longword), or one of the sizeof (longword) bytes starting at char_ptr is == c. On little-endian machines, we could determine the first such byte without any further memory accesses, just by looking at the tmp result from the last loop iteration. But this does not work on big-endian machines. Choose code that works in both cases.

Definition at line 59 of file memchr.c.

60 {
61  /* On 32-bit hardware, choosing longword to be a 32-bit unsigned
62  long instead of a 64-bit uintmax_t tends to give better
63  performance. On 64-bit hardware, unsigned long is generally 64
64  bits already. Change this typedef to experiment with
65  performance. */
66  typedef unsigned long int longword;
67 
68  const unsigned char *char_ptr;
69  const longword *longword_ptr;
70  longword repeated_one;
71  longword repeated_c;
72  unsigned reg_char c;
73 
74  c = (unsigned char) c_in;
75 
76  /* Handle the first few bytes by reading one byte at a time.
77  Do this until CHAR_PTR is aligned on a longword boundary. */
78  for (char_ptr = (const unsigned char *) s;
79  n > 0 && (size_t) char_ptr % sizeof (longword) != 0;
80  --n, ++char_ptr)
81  if (*char_ptr == c)
82  return (void *) char_ptr;
83 
84  longword_ptr = (const longword *) char_ptr;
85 
86  /* All these elucidatory comments refer to 4-byte longwords,
87  but the theory applies equally well to any size longwords. */
88 
89  /* Compute auxiliary longword values:
90  repeated_one is a value which has a 1 in every byte.
91  repeated_c has c in every byte. */
92  repeated_one = 0x01010101;
93  repeated_c = c | (c << 8);
94  repeated_c |= repeated_c << 16;
95  if (0xffffffffU < (longword) -1)
96  {
97  repeated_one |= repeated_one << 31 << 1;
98  repeated_c |= repeated_c << 31 << 1;
99  if (8 < sizeof (longword))
100  {
101  size_t i;
102 
103  for (i = 64; i < sizeof (longword) * 8; i *= 2)
104  {
105  repeated_one |= repeated_one << i;
106  repeated_c |= repeated_c << i;
107  }
108  }
109  }
110 
111  /* Instead of the traditional loop which tests each byte, we will test a
112  longword at a time. The tricky part is testing if *any of the four*
113  bytes in the longword in question are equal to c. We first use an xor
114  with repeated_c. This reduces the task to testing whether *any of the
115  four* bytes in longword1 is zero.
116 
117  We compute tmp =
118  ((longword1 - repeated_one) & ~longword1) & (repeated_one << 7).
119  That is, we perform the following operations:
120  1. Subtract repeated_one.
121  2. & ~longword1.
122  3. & a mask consisting of 0x80 in every byte.
123  Consider what happens in each byte:
124  - If a byte of longword1 is zero, step 1 and 2 transform it into 0xff,
125  and step 3 transforms it into 0x80. A carry can also be propagated
126  to more significant bytes.
127  - If a byte of longword1 is nonzero, let its lowest 1 bit be at
128  position k (0 <= k <= 7); so the lowest k bits are 0. After step 1,
129  the byte ends in a single bit of value 0 and k bits of value 1.
130  After step 2, the result is just k bits of value 1: 2^k - 1. After
131  step 3, the result is 0. And no carry is produced.
132  So, if longword1 has only non-zero bytes, tmp is zero.
133  Whereas if longword1 has a zero byte, call j the position of the least
134  significant zero byte. Then the result has a zero at positions 0, ...,
135  j-1 and a 0x80 at position j. We cannot predict the result at the more
136  significant bytes (positions j+1..3), but it does not matter since we
137  already have a non-zero bit at position 8*j+7.
138 
139  So, the test whether any byte in longword1 is zero is equivalent to
140  testing whether tmp is nonzero. */
141 
142  while (n >= sizeof (longword))
143  {
144  longword longword1 = *longword_ptr ^ repeated_c;
145 
146  if ((((longword1 - repeated_one) & ~longword1)
147  & (repeated_one << 7)) != 0)
148  break;
149  longword_ptr++;
150  n -= sizeof (longword);
151  }
152 
153  char_ptr = (const unsigned char *) longword_ptr;
154 
155  /* At this point, we know that either n < sizeof (longword), or one of the
156  sizeof (longword) bytes starting at char_ptr is == c. On little-endian
157  machines, we could determine the first such byte without any further
158  memory accesses, just by looking at the tmp result from the last loop
159  iteration. But this does not work on big-endian machines. Choose code
160  that works in both cases. */
161 
162  for (; n > 0; --n, ++char_ptr)
163  {
164  if (*char_ptr == c)
165  return (void *) char_ptr;
166  }
167 
168  return NULL;
169 }
c_in
FILE * c_in
Definition: c_syntax.h:291
reg_char
#define reg_char
Copyright (C) 1991, 1993, 1996-1997, 1999-2000, 2003-2004, 2006, 2008-2014 Free Software Foundation,...
Definition: memchr.c:37
size_t
size_t size_t
Definition: properties.c:413

References c_in, and reg_char.