From: Simon Josefsson Date: Wed, 2 Jan 2008 15:04:31 +0000 (+0100) Subject: Update gnulib files. X-Git-Tag: gnutls_2_3_0~18 X-Git-Url: http://git.ipfire.org/gitweb.cgi?a=commitdiff_plain;h=a124812c92d6ac5a459ecdd1fffbc1bbb3fdcd9c;p=thirdparty%2Fgnutls.git Update gnulib files. --- diff --git a/lgl/dummy.c b/lgl/dummy.c deleted file mode 100644 index 5834053b05..0000000000 --- a/lgl/dummy.c +++ /dev/null @@ -1,42 +0,0 @@ -/* A dummy file, to prevent empty libraries from breaking builds. - Copyright (C) 2004, 2007 Free Software Foundation, Inc. - - This program 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. - - 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 Lesser General Public License for more details. - - You should have received a copy of the GNU Lesser General Public License - along with this program. If not, see . */ - -/* Some systems, reportedly OpenBSD and Mac OS X, refuse to create - libraries without any object files. You might get an error like: - - > ar cru .libs/libgl.a - > ar: no archive members specified - - Compiling this file, and adding its object file to the library, will - prevent the library from being empty. */ - -/* Some systems, such as Solaris with cc 5.0, refuse to work with libraries - that don't export any symbol. You might get an error like: - - > cc ... libgnu.a - > ild: (bad file) garbled symbol table in archive ../gllib/libgnu.a - - Compiling this file, and adding its object file to the library, will - prevent the library from exporting no symbols. */ - -#ifdef __sun -/* This declaration ensures that the library will export at least 1 symbol. */ -int gl_dummy_symbol; -#else -/* This declaration is solely to ensure that after preprocessing - this file is never empty. */ -typedef int dummy; -#endif diff --git a/lgl/m4/eealloc.m4 b/lgl/m4/eealloc.m4 new file mode 100644 index 0000000000..adcfd06c95 --- /dev/null +++ b/lgl/m4/eealloc.m4 @@ -0,0 +1,32 @@ +# eealloc.m4 serial 1 +dnl Copyright (C) 2003 Free Software Foundation, Inc. +dnl This file is free software; the Free Software Foundation +dnl gives unlimited permission to copy and/or distribute it, +dnl with or without modifications, as long as this notice is preserved. + +AC_DEFUN([gl_EEALLOC], +[ + AC_REQUIRE([gl_EEMALLOC]) + AC_REQUIRE([gl_EEREALLOC]) + AC_REQUIRE([AC_C_INLINE]) +]) + +AC_DEFUN([gl_EEMALLOC], +[ + _AC_FUNC_MALLOC_IF( + [gl_cv_func_malloc_0_nonnull=1], + [gl_cv_func_malloc_0_nonnull=0]) + AC_DEFINE_UNQUOTED([MALLOC_0_IS_NONNULL], $gl_cv_func_malloc_0_nonnull, + [If malloc(0) is != NULL, define this to 1. Otherwise define this + to 0.]) +]) + +AC_DEFUN([gl_EEREALLOC], +[ + _AC_FUNC_REALLOC_IF( + [gl_cv_func_realloc_0_nonnull=1], + [gl_cv_func_realloc_0_nonnull=0]) + AC_DEFINE_UNQUOTED([REALLOC_0_IS_NONNULL], $gl_cv_func_realloc_0_nonnull, + [If realloc(NULL,0) is != NULL, define this to 1. Otherwise define this + to 0.]) +]) diff --git a/lgl/m4/malloca.m4 b/lgl/m4/malloca.m4 new file mode 100644 index 0000000000..2841ae83a7 --- /dev/null +++ b/lgl/m4/malloca.m4 @@ -0,0 +1,14 @@ +# malloca.m4 serial 1 +dnl Copyright (C) 2003-2004, 2006-2007 Free Software Foundation, Inc. +dnl This file is free software; the Free Software Foundation +dnl gives unlimited permission to copy and/or distribute it, +dnl with or without modifications, as long as this notice is preserved. + +AC_DEFUN([gl_MALLOCA], +[ + dnl Use the autoconf tests for alloca(), but not the AC_SUBSTed variables + dnl @ALLOCA@ and @LTALLOCA@. + dnl gl_FUNC_ALLOCA dnl Already brought in by the module dependencies. + AC_REQUIRE([gl_EEMALLOC]) + AC_REQUIRE([AC_TYPE_LONG_LONG_INT]) +]) diff --git a/lgl/m4/memchr.m4 b/lgl/m4/memchr.m4 new file mode 100644 index 0000000000..91b8636eb8 --- /dev/null +++ b/lgl/m4/memchr.m4 @@ -0,0 +1,18 @@ +# memchr.m4 serial 4 +dnl Copyright (C) 2002, 2003, 2004 Free Software Foundation, Inc. +dnl This file is free software; the Free Software Foundation +dnl gives unlimited permission to copy and/or distribute it, +dnl with or without modifications, as long as this notice is preserved. + +AC_DEFUN([gl_FUNC_MEMCHR], +[ + AC_REPLACE_FUNCS(memchr) + if test $ac_cv_func_memchr = no; then + gl_PREREQ_MEMCHR + fi +]) + +# Prerequisites of lib/memchr.c. +AC_DEFUN([gl_PREREQ_MEMCHR], [ + AC_CHECK_HEADERS(bp-sym.h) +]) diff --git a/lgl/m4/memcmp.m4 b/lgl/m4/memcmp.m4 new file mode 100644 index 0000000000..099b141b2c --- /dev/null +++ b/lgl/m4/memcmp.m4 @@ -0,0 +1,19 @@ +# memcmp.m4 serial 12 +dnl Copyright (C) 2002, 2003, 2004, 2007 Free Software Foundation, Inc. +dnl This file is free software; the Free Software Foundation +dnl gives unlimited permission to copy and/or distribute it, +dnl with or without modifications, as long as this notice is preserved. + +AC_DEFUN([gl_FUNC_MEMCMP], +[ + AC_FUNC_MEMCMP + dnl Note: AC_FUNC_MEMCMP does AC_LIBOBJ(memcmp). + if test $ac_cv_func_memcmp_working = no; then + AC_DEFINE(memcmp, rpl_memcmp, + [Define to rpl_memcmp if the replacement function should be used.]) + gl_PREREQ_MEMCMP + fi +]) + +# Prerequisites of lib/memcmp.c. +AC_DEFUN([gl_PREREQ_MEMCMP], [:]) diff --git a/lgl/malloca.c b/lgl/malloca.c new file mode 100644 index 0000000000..778fdf1baf --- /dev/null +++ b/lgl/malloca.c @@ -0,0 +1,137 @@ +/* Safe automatic memory allocation. + Copyright (C) 2003, 2006-2007 Free Software Foundation, Inc. + Written by Bruno Haible , 2003. + + This program 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, or (at your option) + 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 Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with this program; if not, write to the Free Software Foundation, + Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ + +#include + +/* Specification. */ +#include "malloca.h" + +/* The speed critical point in this file is freea() applied to an alloca() + result: it must be fast, to match the speed of alloca(). The speed of + mmalloca() and freea() in the other case are not critical, because they + are only invoked for big memory sizes. */ + +#if HAVE_ALLOCA + +/* Store the mmalloca() results in a hash table. This is needed to reliably + distinguish a mmalloca() result and an alloca() result. + + Although it is possible that the same pointer is returned by alloca() and + by mmalloca() at different times in the same application, it does not lead + to a bug in freea(), because: + - Before a pointer returned by alloca() can point into malloc()ed memory, + the function must return, and once this has happened the programmer must + not call freea() on it anyway. + - Before a pointer returned by mmalloca() can point into the stack, it + must be freed. The only function that can free it is freea(), and + when freea() frees it, it also removes it from the hash table. */ + +#define MAGIC_NUMBER 0x1415fb4a +#define MAGIC_SIZE sizeof (int) +/* This is how the header info would look like without any alignment + considerations. */ +struct preliminary_header { void *next; char room[MAGIC_SIZE]; }; +/* But the header's size must be a multiple of sa_alignment_max. */ +#define HEADER_SIZE \ + (((sizeof (struct preliminary_header) + sa_alignment_max - 1) / sa_alignment_max) * sa_alignment_max) +struct header { void *next; char room[HEADER_SIZE - sizeof (struct preliminary_header) + MAGIC_SIZE]; }; +/* Verify that HEADER_SIZE == sizeof (struct header). */ +typedef int verify1[2 * (HEADER_SIZE == sizeof (struct header)) - 1]; +/* We make the hash table quite big, so that during lookups the probability + of empty hash buckets is quite high. There is no need to make the hash + table resizable, because when the hash table gets filled so much that the + lookup becomes slow, it means that the application has memory leaks. */ +#define HASH_TABLE_SIZE 257 +static void * mmalloca_results[HASH_TABLE_SIZE]; + +#endif + +void * +mmalloca (size_t n) +{ +#if HAVE_ALLOCA + /* Allocate one more word, that serves as an indicator for malloc()ed + memory, so that freea() of an alloca() result is fast. */ + size_t nplus = n + HEADER_SIZE; + + if (nplus >= n) + { + char *p = (char *) malloc (nplus); + + if (p != NULL) + { + size_t slot; + + p += HEADER_SIZE; + + /* Put a magic number into the indicator word. */ + ((int *) p)[-1] = MAGIC_NUMBER; + + /* Enter p into the hash table. */ + slot = (unsigned long) p % HASH_TABLE_SIZE; + ((struct header *) (p - HEADER_SIZE))->next = mmalloca_results[slot]; + mmalloca_results[slot] = p; + + return p; + } + } + /* Out of memory. */ + return NULL; +#else +# if !MALLOC_0_IS_NONNULL + if (n == 0) + n = 1; +# endif + return malloc (n); +#endif +} + +#if HAVE_ALLOCA +void +freea (void *p) +{ + /* mmalloca() may have returned NULL. */ + if (p != NULL) + { + /* Attempt to quickly distinguish the mmalloca() result - which has + a magic indicator word - and the alloca() result - which has an + uninitialized indicator word. It is for this test that sa_increment + additional bytes are allocated in the alloca() case. */ + if (((int *) p)[-1] == MAGIC_NUMBER) + { + /* Looks like a mmalloca() result. To see whether it really is one, + perform a lookup in the hash table. */ + size_t slot = (unsigned long) p % HASH_TABLE_SIZE; + void **chain = &mmalloca_results[slot]; + for (; *chain != NULL;) + { + if (*chain == p) + { + /* Found it. Remove it from the hash table and free it. */ + char *p_begin = (char *) p - HEADER_SIZE; + *chain = ((struct header *) p_begin)->next; + free (p_begin); + return; + } + chain = &((struct header *) ((char *) *chain - HEADER_SIZE))->next; + } + } + /* At this point, we know it was not a mmalloca() result. */ + } +} +#endif diff --git a/lgl/malloca.h b/lgl/malloca.h new file mode 100644 index 0000000000..15448a34e3 --- /dev/null +++ b/lgl/malloca.h @@ -0,0 +1,134 @@ +/* Safe automatic memory allocation. + Copyright (C) 2003-2007 Free Software Foundation, Inc. + Written by Bruno Haible , 2003. + + This program 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, or (at your option) + 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 Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with this program; if not, write to the Free Software Foundation, + Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. */ + +#ifndef _MALLOCA_H +#define _MALLOCA_H + +#include +#include +#include + + +#ifdef __cplusplus +extern "C" { +#endif + + +/* safe_alloca(N) is equivalent to alloca(N) when it is safe to call + alloca(N); otherwise it returns NULL. It either returns N bytes of + memory allocated on the stack, that lasts until the function returns, + or NULL. + Use of safe_alloca should be avoided: + - inside arguments of function calls - undefined behaviour, + - in inline functions - the allocation may actually last until the + calling function returns. +*/ +#if HAVE_ALLOCA +/* The OS usually guarantees only one guard page at the bottom of the stack, + and a page size can be as small as 4096 bytes. So we cannot safely + allocate anything larger than 4096 bytes. Also care for the possibility + of a few compiler-allocated temporary stack slots. + This must be a macro, not an inline function. */ +# define safe_alloca(N) ((N) < 4032 ? alloca (N) : NULL) +#else +# define safe_alloca(N) ((void) (N), NULL) +#endif + +/* malloca(N) is a safe variant of alloca(N). It allocates N bytes of + memory allocated on the stack, that must be freed using freea() before + the function returns. Upon failure, it returns NULL. */ +#if HAVE_ALLOCA +# define malloca(N) \ + ((N) < 4032 - sa_increment \ + ? (void *) ((char *) alloca ((N) + sa_increment) + sa_increment) \ + : mmalloca (N)) +#else +# define malloca(N) \ + mmalloca (N) +#endif +extern void * mmalloca (size_t n); + +/* Free a block of memory allocated through malloca(). */ +#if HAVE_ALLOCA +extern void freea (void *p); +#else +# define freea free +#endif + +/* nmalloca(N,S) is an overflow-safe variant of malloca (N * S). + It allocates an array of N objects, each with S bytes of memory, + on the stack. S must be positive and N must be nonnegative. + The array must be freed using freea() before the function returns. */ +#if 1 +/* Cf. the definition of xalloc_oversized. */ +# define nmalloca(n, s) \ + ((n) > (size_t) (sizeof (ptrdiff_t) <= sizeof (size_t) ? -1 : -2) / (s) \ + ? NULL \ + : malloca ((n) * (s))) +#else +extern void * nmalloca (size_t n, size_t s); +#endif + + +#ifdef __cplusplus +} +#endif + + +/* ------------------- Auxiliary, non-public definitions ------------------- */ + +/* Determine the alignment of a type at compile time. */ +#if defined __GNUC__ +# define sa_alignof __alignof__ +#elif defined __cplusplus + template struct sa_alignof_helper { char __slot1; type __slot2; }; +# define sa_alignof(type) offsetof (sa_alignof_helper, __slot2) +#elif defined __hpux + /* Work around a HP-UX 10.20 cc bug with enums constants defined as offsetof + values. */ +# define sa_alignof(type) (sizeof (type) <= 4 ? 4 : 8) +#elif defined _AIX + /* Work around an AIX 3.2.5 xlc bug with enums constants defined as offsetof + values. */ +# define sa_alignof(type) (sizeof (type) <= 4 ? 4 : 8) +#else +# define sa_alignof(type) offsetof (struct { char __slot1; type __slot2; }, __slot2) +#endif + +enum +{ +/* The desired alignment of memory allocations is the maximum alignment + among all elementary types. */ + sa_alignment_long = sa_alignof (long), + sa_alignment_double = sa_alignof (double), +#if HAVE_LONG_LONG_INT + sa_alignment_longlong = sa_alignof (long long), +#endif + sa_alignment_longdouble = sa_alignof (long double), + sa_alignment_max = ((sa_alignment_long - 1) | (sa_alignment_double - 1) +#if HAVE_LONG_LONG_INT + | (sa_alignment_longlong - 1) +#endif + | (sa_alignment_longdouble - 1) + ) + 1, +/* The increment that guarantees room for a magic word must be >= sizeof (int) + and a multiple of sa_alignment_max. */ + sa_increment = ((sizeof (int) + sa_alignment_max - 1) / sa_alignment_max) * sa_alignment_max +}; + +#endif /* _MALLOCA_H */ diff --git a/lgl/malloca.valgrind b/lgl/malloca.valgrind new file mode 100644 index 0000000000..52f0a50f57 --- /dev/null +++ b/lgl/malloca.valgrind @@ -0,0 +1,7 @@ +# Suppress a valgrind message about use of uninitialized memory in freea(). +# This use is OK because it provides only a speedup. +{ + freea + Memcheck:Cond + fun:freea +} diff --git a/lgl/memchr.c b/lgl/memchr.c new file mode 100644 index 0000000000..96729cfb7d --- /dev/null +++ b/lgl/memchr.c @@ -0,0 +1,200 @@ +/* Copyright (C) 1991, 1993, 1996, 1997, 1999, 2000, 2003, 2004, 2006 Free + Software Foundation, Inc. + + Based on strlen implementation by Torbjorn Granlund (tege@sics.se), + with help from Dan Sahlin (dan@sics.se) and + commentary by Jim Blandy (jimb@ai.mit.edu); + adaptation to memchr suggested by Dick Karpinski (dick@cca.ucsf.edu), + and implemented by Roland McGrath (roland@ai.mit.edu). + +NOTE: The canonical source of this file is maintained with the GNU C Library. +Bugs can be reported to bug-glibc@prep.ai.mit.edu. + +This program 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 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 Lesser General Public License for more details. + +You should have received a copy of the GNU Lesser General Public License +along with this program. If not, see . */ + +#ifndef _LIBC +# include +#endif + +#include + +#include + +#if defined _LIBC +# include +#else +# define reg_char char +#endif + +#include + +#if HAVE_BP_SYM_H || defined _LIBC +# include +#else +# define BP_SYM(sym) sym +#endif + +#undef memchr +#undef __memchr + +/* Search no more than N bytes of S for C. */ +void * +__memchr (void const *s, int c_in, size_t n) +{ + const unsigned char *char_ptr; + const unsigned long int *longword_ptr; + unsigned long int longword, magic_bits, charmask; + unsigned reg_char c; + int i; + + c = (unsigned char) c_in; + + /* 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 = (const unsigned char *) s; + n > 0 && (size_t) char_ptr % sizeof longword != 0; + --n, ++char_ptr) + if (*char_ptr == c) + return (void *) char_ptr; + + /* All these elucidatory comments refer to 4-byte longwords, + but the theory applies equally well to any size longwords. */ + + longword_ptr = (const 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. */ + + /* Set MAGIC_BITS to be this pattern of 1 and 0 bits. + Set CHARMASK to be a longword, each of whose bytes is C. */ + + magic_bits = 0xfefefefe; + charmask = c | (c << 8); + charmask |= charmask << 16; +#if 0xffffffffU < ULONG_MAX + magic_bits |= magic_bits << 32; + charmask |= charmask << 32; + if (8 < sizeof longword) + for (i = 64; i < sizeof longword * 8; i *= 2) + { + magic_bits |= magic_bits << i; + charmask |= charmask << i; + } +#endif + magic_bits = (ULONG_MAX >> 1) & (magic_bits | 1); + + /* 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 (n >= sizeof longword) + { + /* 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. + + 3) But wait! Aren't we looking for C, not zero? + Good point. So what we do is XOR LONGWORD with a longword, + each of whose bytes is C. This turns each byte that is C + into a zero. */ + + longword = *longword_ptr++ ^ charmask; + + /* Add MAGIC_BITS to LONGWORD. */ + if ((((longword + magic_bits) + + /* Set those bits that were unchanged by the addition. */ + ^ ~longword) + + /* Look at only the hole bits. If any of the hole bits + are unchanged, most likely one of the bytes was a + zero. */ + & ~magic_bits) != 0) + { + /* Which of the bytes was C? If none of them were, it was + a misfire; continue the search. */ + + const unsigned char *cp = (const unsigned char *) (longword_ptr - 1); + + if (cp[0] == c) + return (void *) cp; + if (cp[1] == c) + return (void *) &cp[1]; + if (cp[2] == c) + return (void *) &cp[2]; + if (cp[3] == c) + return (void *) &cp[3]; + if (4 < sizeof longword && cp[4] == c) + return (void *) &cp[4]; + if (5 < sizeof longword && cp[5] == c) + return (void *) &cp[5]; + if (6 < sizeof longword && cp[6] == c) + return (void *) &cp[6]; + if (7 < sizeof longword && cp[7] == c) + return (void *) &cp[7]; + if (8 < sizeof longword) + for (i = 8; i < sizeof longword; i++) + if (cp[i] == c) + return (void *) &cp[i]; + } + + n -= sizeof longword; + } + + char_ptr = (const unsigned char *) longword_ptr; + + while (n-- > 0) + { + if (*char_ptr == c) + return (void *) char_ptr; + else + ++char_ptr; + } + + return 0; +} +#ifdef weak_alias +weak_alias (__memchr, BP_SYM (memchr)) +#endif diff --git a/lgl/memcmp.c b/lgl/memcmp.c new file mode 100644 index 0000000000..fa340631f4 --- /dev/null +++ b/lgl/memcmp.c @@ -0,0 +1,361 @@ +/* Copyright (C) 1991, 1993, 1995, 1997, 1998, 2003, 2006 Free Software + Foundation, Inc. + + Contributed by Torbjorn Granlund (tege@sics.se). + + NOTE: The canonical source of this file is maintained with the GNU C Library. + Bugs can be reported to bug-glibc@prep.ai.mit.edu. + + This program 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 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 Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public License + along with this program. If not, see . */ + +#ifndef _LIBC +# include +#endif + +#include + +#undef memcmp + +#ifdef _LIBC + +# include +# include + +# if __BYTE_ORDER == __BIG_ENDIAN +# define WORDS_BIGENDIAN +# endif + +#else /* Not in the GNU C library. */ + +# include + +/* Type to use for aligned memory operations. + This should normally be the biggest type supported by a single load + and store. Must be an unsigned type. */ +# define op_t unsigned long int +# define OPSIZ (sizeof(op_t)) + +/* Threshold value for when to enter the unrolled loops. */ +# define OP_T_THRES 16 + +/* Type to use for unaligned operations. */ +typedef unsigned char byte; + +# ifndef WORDS_BIGENDIAN +# define MERGE(w0, sh_1, w1, sh_2) (((w0) >> (sh_1)) | ((w1) << (sh_2))) +# else +# define MERGE(w0, sh_1, w1, sh_2) (((w0) << (sh_1)) | ((w1) >> (sh_2))) +# endif + +#endif /* In the GNU C library. */ + +#ifdef WORDS_BIGENDIAN +# define CMP_LT_OR_GT(a, b) ((a) > (b) ? 1 : -1) +#else +# define CMP_LT_OR_GT(a, b) memcmp_bytes ((a), (b)) +#endif + +/* BE VERY CAREFUL IF YOU CHANGE THIS CODE! */ + +/* The strategy of this memcmp is: + + 1. Compare bytes until one of the block pointers is aligned. + + 2. Compare using memcmp_common_alignment or + memcmp_not_common_alignment, regarding the alignment of the other + block after the initial byte operations. The maximum number of + full words (of type op_t) are compared in this way. + + 3. Compare the few remaining bytes. */ + +#ifndef WORDS_BIGENDIAN +/* memcmp_bytes -- Compare A and B bytewise in the byte order of the machine. + A and B are known to be different. + This is needed only on little-endian machines. */ + +# ifdef __GNUC__ +__inline +# endif +static int +memcmp_bytes (long unsigned int a, long unsigned int b) +{ + long int srcp1 = (long int) &a; + long int srcp2 = (long int) &b; + op_t a0, b0; + + do + { + a0 = ((byte *) srcp1)[0]; + b0 = ((byte *) srcp2)[0]; + srcp1 += 1; + srcp2 += 1; + } + while (a0 == b0); + return a0 - b0; +} +#endif + +/* memcmp_common_alignment -- Compare blocks at SRCP1 and SRCP2 with LEN `op_t' + objects (not LEN bytes!). Both SRCP1 and SRCP2 should be aligned for + memory operations on `op_t's. */ +#ifdef __GNUC__ +__inline +#endif +static int +memcmp_common_alignment (long int srcp1, long int srcp2, size_t len) +{ + op_t a0, a1; + op_t b0, b1; + + switch (len % 4) + { + default: /* Avoid warning about uninitialized local variables. */ + case 2: + a0 = ((op_t *) srcp1)[0]; + b0 = ((op_t *) srcp2)[0]; + srcp1 -= 2 * OPSIZ; + srcp2 -= 2 * OPSIZ; + len += 2; + goto do1; + case 3: + a1 = ((op_t *) srcp1)[0]; + b1 = ((op_t *) srcp2)[0]; + srcp1 -= OPSIZ; + srcp2 -= OPSIZ; + len += 1; + goto do2; + case 0: + if (OP_T_THRES <= 3 * OPSIZ && len == 0) + return 0; + a0 = ((op_t *) srcp1)[0]; + b0 = ((op_t *) srcp2)[0]; + goto do3; + case 1: + a1 = ((op_t *) srcp1)[0]; + b1 = ((op_t *) srcp2)[0]; + srcp1 += OPSIZ; + srcp2 += OPSIZ; + len -= 1; + if (OP_T_THRES <= 3 * OPSIZ && len == 0) + goto do0; + /* Fall through. */ + } + + do + { + a0 = ((op_t *) srcp1)[0]; + b0 = ((op_t *) srcp2)[0]; + if (a1 != b1) + return CMP_LT_OR_GT (a1, b1); + + do3: + a1 = ((op_t *) srcp1)[1]; + b1 = ((op_t *) srcp2)[1]; + if (a0 != b0) + return CMP_LT_OR_GT (a0, b0); + + do2: + a0 = ((op_t *) srcp1)[2]; + b0 = ((op_t *) srcp2)[2]; + if (a1 != b1) + return CMP_LT_OR_GT (a1, b1); + + do1: + a1 = ((op_t *) srcp1)[3]; + b1 = ((op_t *) srcp2)[3]; + if (a0 != b0) + return CMP_LT_OR_GT (a0, b0); + + srcp1 += 4 * OPSIZ; + srcp2 += 4 * OPSIZ; + len -= 4; + } + while (len != 0); + + /* This is the right position for do0. Please don't move + it into the loop. */ + do0: + if (a1 != b1) + return CMP_LT_OR_GT (a1, b1); + return 0; +} + +/* memcmp_not_common_alignment -- Compare blocks at SRCP1 and SRCP2 with LEN + `op_t' objects (not LEN bytes!). SRCP2 should be aligned for memory + operations on `op_t', but SRCP1 *should be unaligned*. */ +#ifdef __GNUC__ +__inline +#endif +static int +memcmp_not_common_alignment (long int srcp1, long int srcp2, size_t len) +{ + op_t a0, a1, a2, a3; + op_t b0, b1, b2, b3; + op_t x; + int shl, shr; + + /* Calculate how to shift a word read at the memory operation + aligned srcp1 to make it aligned for comparison. */ + + shl = 8 * (srcp1 % OPSIZ); + shr = 8 * OPSIZ - shl; + + /* Make SRCP1 aligned by rounding it down to the beginning of the `op_t' + it points in the middle of. */ + srcp1 &= -OPSIZ; + + switch (len % 4) + { + default: /* Avoid warning about uninitialized local variables. */ + case 2: + a1 = ((op_t *) srcp1)[0]; + a2 = ((op_t *) srcp1)[1]; + b2 = ((op_t *) srcp2)[0]; + srcp1 -= 1 * OPSIZ; + srcp2 -= 2 * OPSIZ; + len += 2; + goto do1; + case 3: + a0 = ((op_t *) srcp1)[0]; + a1 = ((op_t *) srcp1)[1]; + b1 = ((op_t *) srcp2)[0]; + srcp2 -= 1 * OPSIZ; + len += 1; + goto do2; + case 0: + if (OP_T_THRES <= 3 * OPSIZ && len == 0) + return 0; + a3 = ((op_t *) srcp1)[0]; + a0 = ((op_t *) srcp1)[1]; + b0 = ((op_t *) srcp2)[0]; + srcp1 += 1 * OPSIZ; + goto do3; + case 1: + a2 = ((op_t *) srcp1)[0]; + a3 = ((op_t *) srcp1)[1]; + b3 = ((op_t *) srcp2)[0]; + srcp1 += 2 * OPSIZ; + srcp2 += 1 * OPSIZ; + len -= 1; + if (OP_T_THRES <= 3 * OPSIZ && len == 0) + goto do0; + /* Fall through. */ + } + + do + { + a0 = ((op_t *) srcp1)[0]; + b0 = ((op_t *) srcp2)[0]; + x = MERGE(a2, shl, a3, shr); + if (x != b3) + return CMP_LT_OR_GT (x, b3); + + do3: + a1 = ((op_t *) srcp1)[1]; + b1 = ((op_t *) srcp2)[1]; + x = MERGE(a3, shl, a0, shr); + if (x != b0) + return CMP_LT_OR_GT (x, b0); + + do2: + a2 = ((op_t *) srcp1)[2]; + b2 = ((op_t *) srcp2)[2]; + x = MERGE(a0, shl, a1, shr); + if (x != b1) + return CMP_LT_OR_GT (x, b1); + + do1: + a3 = ((op_t *) srcp1)[3]; + b3 = ((op_t *) srcp2)[3]; + x = MERGE(a1, shl, a2, shr); + if (x != b2) + return CMP_LT_OR_GT (x, b2); + + srcp1 += 4 * OPSIZ; + srcp2 += 4 * OPSIZ; + len -= 4; + } + while (len != 0); + + /* This is the right position for do0. Please don't move + it into the loop. */ + do0: + x = MERGE(a2, shl, a3, shr); + if (x != b3) + return CMP_LT_OR_GT (x, b3); + return 0; +} + +int +rpl_memcmp (const void *s1, const void *s2, size_t len) +{ + op_t a0; + op_t b0; + long int srcp1 = (long int) s1; + long int srcp2 = (long int) s2; + op_t res; + + if (len >= OP_T_THRES) + { + /* There are at least some bytes to compare. No need to test + for LEN == 0 in this alignment loop. */ + while (srcp2 % OPSIZ != 0) + { + a0 = ((byte *) srcp1)[0]; + b0 = ((byte *) srcp2)[0]; + srcp1 += 1; + srcp2 += 1; + res = a0 - b0; + if (res != 0) + return res; + len -= 1; + } + + /* SRCP2 is now aligned for memory operations on `op_t'. + SRCP1 alignment determines if we can do a simple, + aligned compare or need to shuffle bits. */ + + if (srcp1 % OPSIZ == 0) + res = memcmp_common_alignment (srcp1, srcp2, len / OPSIZ); + else + res = memcmp_not_common_alignment (srcp1, srcp2, len / OPSIZ); + if (res != 0) + return res; + + /* Number of bytes remaining in the interval [0..OPSIZ-1]. */ + srcp1 += len & -OPSIZ; + srcp2 += len & -OPSIZ; + len %= OPSIZ; + } + + /* There are just a few bytes to compare. Use byte memory operations. */ + while (len != 0) + { + a0 = ((byte *) srcp1)[0]; + b0 = ((byte *) srcp2)[0]; + srcp1 += 1; + srcp2 += 1; + res = a0 - b0; + if (res != 0) + return res; + len -= 1; + } + + return 0; +} + +#ifdef weak_alias +# undef bcmp +weak_alias (memcmp, bcmp) +#endif