newlib-cygwin/newlib/libc/string/strstr.c

201 lines
6.0 KiB
C
Raw Normal View History

Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
/* Optimized strstr function.
Copyright (c) 2018 Arm Ltd. All rights reserved.
SPDX-License-Identifier: BSD-3-Clause
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
3. The name of the company may not be used to endorse or promote
products derived from this software without specific prior written
permission.
THIS SOFTWARE IS PROVIDED BY ARM LTD ``AS IS'' AND ANY EXPRESS OR IMPLIED
WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
IN NO EVENT SHALL ARM LTD BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED
TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */
2000-02-18 03:39:52 +08:00
/*
FUNCTION
<<strstr>>---find string segment
INDEX
strstr
SYNOPSIS
2000-02-18 03:39:52 +08:00
#include <string.h>
char *strstr(const char *<[s1]>, const char *<[s2]>);
DESCRIPTION
Locates the first occurrence in the string pointed to by <[s1]> of
2000-02-18 03:39:52 +08:00
the sequence of characters in the string pointed to by <[s2]>
(excluding the terminating null character).
2000-02-18 03:39:52 +08:00
RETURNS
Returns a pointer to the located string segment, or a null
pointer if the string <[s2]> is not found. If <[s2]> points to
a string with zero length, <[s1]> is returned.
2000-02-18 03:39:52 +08:00
PORTABILITY
<<strstr>> is ANSI C.
<<strstr>> requires no supporting OS subroutines.
QUICKREF
strstr ansi pure
*/
#include <string.h>
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
#include <limits.h>
#if defined(PREFER_SIZE_OVER_SPEED) || defined(__OPTIMIZE_SIZE__) \
|| CHAR_BIT > 8
2000-02-18 03:39:52 +08:00
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
/* Small and efficient strstr implementation. */
2000-02-18 03:39:52 +08:00
char *
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
strstr (const char *hs, const char *ne)
2000-02-18 03:39:52 +08:00
{
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
size_t i;
int c = ne[0];
2000-02-18 03:39:52 +08:00
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
if (c == 0)
return (char*)hs;
2000-02-18 03:39:52 +08:00
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
for ( ; hs[0] != '\0'; hs++)
{
if (hs[0] != c)
continue;
for (i = 1; ne[i] != 0; i++)
if (hs[i] != ne[i])
break;
if (ne[i] == '\0')
return (char*)hs;
2000-02-18 03:39:52 +08:00
}
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
return NULL;
}
#else /* compilation for speed */
# define RETURN_TYPE char *
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
# define AVAILABLE(h, h_l, j, n_l) (((j) <= (h_l) - (n_l)) \
|| ((h_l) += strnlen ((h) + (h_l), (n_l) | 2048), ((j) <= (h_l) - (n_l))))
# include "str-two-way.h"
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
/* Number of bits used to index shift table. */
#define SHIFT_TABLE_BITS 6
static inline char *
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
strstr2 (const unsigned char *hs, const unsigned char *ne)
{
uint32_t h1 = (ne[0] << 16) | ne[1];
uint32_t h2 = 0;
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
for (int c = hs[0]; h1 != h2 && c != 0; c = *++hs)
h2 = (h2 << 16) | c;
return h1 == h2 ? (char *)hs - 2 : NULL;
}
static inline char *
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
strstr3 (const unsigned char *hs, const unsigned char *ne)
{
uint32_t h1 = (ne[0] << 24) | (ne[1] << 16) | (ne[2] << 8);
uint32_t h2 = 0;
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
for (int c = hs[0]; h1 != h2 && c != 0; c = *++hs)
h2 = (h2 | c) << 8;
return h1 == h2 ? (char *)hs - 3 : NULL;
}
static inline char *
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
strstr4 (const unsigned char *hs, const unsigned char *ne)
{
uint32_t h1 = (ne[0] << 24) | (ne[1] << 16) | (ne[2] << 8) | ne[3];
uint32_t h2 = 0;
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
for (int c = hs[0]; c != 0 && h1 != h2; c = *++hs)
h2 = (h2 << 8) | c;
return h1 == h2 ? (char *)hs - 4 : NULL;
}
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
/* Extremely fast strstr algorithm with guaranteed linear-time performance.
Small needles up to size 4 use a dedicated linear search. Longer needles
up to size 254 use Sunday's Quick-Search algorithm. Due to its simplicity
it has the best average performance of string matching algorithms on almost
all inputs. It uses a bad-character shift table to skip past mismatches.
By limiting the needle length to 254, the shift table can be reduced to 8
bits per entry, lowering preprocessing overhead and minimizing cache effects.
The limit also implies the worst-case performance is linear.
Even larger needles are processed by the linear-time Two-Way algorithm.
*/
char *
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
strstr (const char *haystack, const char *needle)
{
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
const unsigned char *hs = (const unsigned char *) haystack;
const unsigned char *ne = (const unsigned char *) needle;
/* Handle short needle special cases first. */
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
if (ne[0] == '\0')
return (char *) hs;
if (ne[1] == '\0')
return (char*)strchr (hs, ne[0]);
if (ne[2] == '\0')
return strstr2 (hs, ne);
if (ne[3] == '\0')
return strstr3 (hs, ne);
if (ne[4] == '\0')
return strstr4 (hs, ne);
size_t ne_len = strlen (ne);
size_t hs_len = strnlen (hs, ne_len | 512);
/* Ensure haystack length is >= needle length. */
if (hs_len < ne_len)
return NULL;
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
/* Use the Quick-Search algorithm for needle lengths less than 255. */
if (__builtin_expect (ne_len < 255, 1))
{
uint8_t shift[1 << SHIFT_TABLE_BITS];
const unsigned char *end = hs + hs_len - ne_len;
/* Initialize bad character shift hash table. */
memset (shift, ne_len + 1, sizeof (shift));
for (int i = 0; i < ne_len; i++)
shift[ne[i] % sizeof (shift)] = ne_len - i;
do
{
hs--;
/* Search by skipping past bad characters. */
size_t tmp = shift[hs[ne_len] % sizeof (shift)];
for (hs += tmp; hs <= end; hs += tmp)
{
tmp = shift[hs[ne_len] % sizeof (shift)];
if (memcmp (hs, ne, ne_len) == 0)
return (char*) hs;
}
if (end[ne_len] == 0)
return NULL;
end += strnlen (end + ne_len, 2048);
}
while (hs <= end);
return NULL;
}
/* Use Two-Way algorithm for very long needles. */
return two_way_long_needle (hs, hs_len, ne, ne_len);
2000-02-18 03:39:52 +08:00
}
Improve performance of strstr v3: Add support for read ahead using strnlen, giving an additional 25% speedup on large inputs (both short and long needles). This patch significantly improves performance of strstr by using Sunday's Quick-Search algorithm. Due to its simplicity it has the best average performance of string matching algorithms on almost all inputs. It uses a bad-character shift table to skip past mismatches. The needle length is limited to 254 - this reduces the shift table memory 4 to 8 times, lowering preprocessing overhead and minimizing cache effects. The limit also implies its worst-case performance is linear. Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE has been improved to use strnlen to read the input in chunks. This results in a 2.5 times speedup for large needles, reducing the performance drop when the Quick-Search algorithm can't be used. The code for 1-4 byte needles has been simplified and now uses unsigned char. Since the optimized code relies on 8-bit chars, we defer to the size-optimized implementation if CHAR_BIT > 8. The performance gain of finding a set of randomly chosen words of size 8 in 256 bytes of English text is 14 times on AArch64. For longer haystacks the gain is well over 20 times. The size-optimized strstr has also been rewritten from scratch to improve performance. On the same test the performance gain is 69%. Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test (https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c). --
2018-10-19 01:29:36 +08:00
#endif /* compilation for speed */