SSE4.2 strstr/strcasestr for x86-64.

This patch implements SSE4.2 strstr/strcasestr, using Knuth-Morris-Pratt
string searching algorithm.

ChangeLog

@@ -1,3 +1,20 @@
+2009-07-20  H.J. Lu  <hongjiu.lu@intel.com>
+
+	* string/strcasestr.c (STRCASESTR): New macro.
+	(__strcasestr): Renamed to ..
+	(STRCASESTR): ...this.
+	* string/strstr.c (STRSTR): New macro.
+	(strstr): Renamed to ..
+	(STRSTR): ...this.
+	* sysdeps/x86_64/multiarch/Makefile (sysdep_routines): Add
+	strstr-c strcasestr-c
+	(CFLAGS-strstr.c): New.
+	(CFLAGS-strcasestr.c): Likewise.
+	* sysdeps/x86_64/multiarch/strcasestr-c.c: New file.
+	* sysdeps/x86_64/multiarch/strcasestr.c: New file.
+	* sysdeps/x86_64/multiarch/strstr-c.c: New file.
+	* sysdeps/x86_64/multiarch/strstr.c: New file.
+
 2009-07-20  Ulrich Drepper  <drepper@redhat.com>
 
 	* locale/localeinfo.h (LIMAGIC): Update value for LC_CTYPE.

string/strcasestr.c

@@ -1,5 +1,6 @@
 /* Return the offset of one string within another.
-   Copyright (C) 1994, 1996-2000, 2004, 2008 Free Software Foundation, Inc.
+   Copyright (C) 1994, 1996-2000, 2004, 2008, 2009
+   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
@@ -52,11 +53,16 @@
 #undef strcasestr
 #undef __strcasestr
 
+#ifndef STRCASESTR
+#define STRCASESTR __strcasestr
+#endif
+
+
 /* Find the first occurrence of NEEDLE in HAYSTACK, using
    case-insensitive comparison.  This function gives unspecified
    results in multibyte locales.  */
 char *
-__strcasestr (const char *haystack_start, const char *needle_start)
+STRCASESTR (const char *haystack_start, const char *needle_start)
 {
   const char *haystack = haystack_start;
   const char *needle = needle_start;

string/strstr.c

@@ -1,5 +1,6 @@
 /* Return the offset of one string within another.
-   Copyright (C) 1994,1996,1997,2000,2001,2003,2008 Free Software Foundation, Inc.
+   Copyright (C) 1994,1996,1997,2000,2001,2003,2008,2009
+   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
@@ -40,11 +41,15 @@
 
 #undef strstr
 
+#ifndef STRSTR
+#define STRSTR strstr
+#endif
+
 /* Return the first occurrence of NEEDLE in HAYSTACK.  Return HAYSTACK
    if NEEDLE is empty, otherwise NULL if NEEDLE is not found in
    HAYSTACK.  */
 char *
-strstr (const char *haystack_start, const char *needle_start)
+STRSTR (const char *haystack_start, const char *needle_start)
 {
   const char *haystack = haystack_start;
   const char *needle = needle_start;

sysdeps/x86_64/multiarch/Makefile

@@ -6,9 +6,11 @@ endif
 ifeq ($(subdir),string)
 sysdep_routines += stpncpy-c strncpy-c strncmp-c
 ifeq (yes,$(config-cflags-sse4))
-sysdep_routines += strcspn-c strpbrk-c strspn-c
+sysdep_routines += strcspn-c strpbrk-c strspn-c strstr-c strcasestr-c
 CFLAGS-strcspn-c.c += -msse4
 CFLAGS-strpbrk-c.c += -msse4
 CFLAGS-strspn-c.c += -msse4
+CFLAGS-strstr.c += -msse4
+CFLAGS-strcasestr.c += -msse4
 endif
 endif

sysdeps/x86_64/multiarch/strcasestr-c.c

@@ -0,0 +1,18 @@
+#include "init-arch.h"
+
+#define STRCASESTR __strcasestr_sse2
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name) \
+  __hidden_ver1 (__strcasestr_sse2, __GI_strcasestr, __strcasestr_sse2);
+
+#include "string/strcasestr.c"
+
+extern char *__strcasestr_sse42 (const char *, const char *);
+
+#if 1
+libc_ifunc (__strcasestr,
+	    HAS_SSE4_2 ? __strcasestr_sse42 : __strcasestr_sse2);
+#else
+libc_ifunc (__strcasestr,
+	    0 ? __strcasestr_sse42 : __strcasestr_sse2);
+#endif

sysdeps/x86_64/multiarch/strcasestr.c

@@ -0,0 +1,3 @@
+#define USE_AS_STRCASESTR
+#define STRSTR_SSE42 __strcasestr_sse42
+#include "strstr.c"

sysdeps/x86_64/multiarch/strstr-c.c

@@ -0,0 +1,12 @@
+#include "init-arch.h"
+
+#define STRSTR __strstr_sse2
+#undef libc_hidden_builtin_def
+#define libc_hidden_builtin_def(name) \
+  __hidden_ver1 (__strstr_sse2, __GI_strstr, __strstr_sse2);
+
+#include "string/strstr.c"
+
+extern char *__strstr_sse42 (const char *, const char *);
+
+libc_ifunc (strstr, HAS_SSE4_2 ? __strstr_sse42 : __strstr_sse2);