glibc/sysdeps/m68k/fpu/__math.h
Ulrich Drepper ec4b0518a3 update from main archive 960919
Thu Sep 19 21:50:55 1996  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/posix/gettimeofday.c (__gettimeofday): Use localtime_r
	instead of localtime.
	Reported by Matthias Urlichs.

	* shlib-versions: Remove version number for libcrypt.
	* features.h: Define __USE_REENTRANT if _REENTRANT or _THREAD_SAFE.
	* libc-symbols.h: Define _REENTRANT while compiling libc.

	* sysdeps/unix/sysv/linux/i386/sysdep.S (__errno_location):
	Define even if !_LIBC_REENTRANT.
	* sysdeps/unix/sysv/linux/i386/sysdep.S (__errno_location):
	Likewise.

	* sysdeps/posix/cuserid.h: Remove prototype for geteuid().
	De-ANSI-fy.

	* MakeTAGS ($P/libc.pot): Generate correctly formed header.
	* po/header.pot: Correct title line.
	* po/nl.po: Update.

Thu Sep 19 18:59:55 1996  Ulrich Drepper  <drepper@cygnus.com>

	* Makeconfig (soversions.mk): Prefer shared lib version numbers
	is add ons over version in libc itself.

	* sysdeps/unix/sysv/linux/i386/sysdep.S: Include <sysdep.h>.
	* sysdeps/unix/sysv/linux/i386/sysdep.h: Prevent multiple
	inclusion.

	* libio/iofgets.c: Use __flockfile and __funlockfile instead of
 	_IO_flockfile and _IO_funlockfile resp.

	* locale/categories.def: Partly support for correct `era' handling
	in LC_TIME category.
	* locale/langinfo.h: Likewise.
	* locale/programs/ld-time.c: Likewise.
	* locale/localeinfo.h: Change comment a bit.

	* malloc/memalign.c: Don't use goto, not necessary anymore.

1996-09-18  Paul Eggert  <eggert@twinsun.com>

	* time/mktime.c (ydhms_tm_diff):  Work correctly even if year
	is negative, or if time_t is unsigned.
	* time/strftime.c (tm_diff): Work correctly even if tm_year
	is near INT_MIN.

Tue Sep 17 16:14:34 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/m68k/fpu/__math.h (__inline_mathop): Changed to generate
 	all three versions of the math function.
	(__inline_mathopf, __inline_mathopl): Removed.
	(__inline_functions): New temporary definition containing all
 	non-trivial inline functions.

Wed Sep 18 00:25:41 1996  Ulrich Drepper  <drepper@cygnus.com>

	* time/strftime.c (strftime): The T_FMT_AMPM string may be empty.

Tue Sep 17 20:27:18 1996  Ulrich Drepper  <drepper@cygnus.com>

	* math/Makefile (extra-libs-others): Use $(extra-libs) instead
	of $(extra-libc).

Tue Sep 17 17:09:44 1996  Ulrich Drepper  <drepper@cygnus.com>

	* sysdeps/stub/fpu_control.h: Correct end of file comment.

Tue Sep 17 05:39:18 1996  Ulrich Drepper  <drepper@cygnus.com>

	* stdio-common/bug3.c, stdio-common/bug4.c, stdio-common/bug5.c,
	stdio-common/test-popen.c: Remove temporary files after test.
	* stdio-common/bug5.c: Use `system' instead of `execlp'.
	Patches by Andreas Jaeger.

	* stdio-common/bug5.c: Create string for `system' argument to
	make sure the input and output file names are really correct.

Sun Sep 15 12:46:44 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* configure.in: If $os contains a hyphen add the part before the
	hyphen to $ostry.

Sun Sep 15 18:14:02 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/m68k/fpu/__math.h (__ieee754_pow, __ieee754_powf,
	__ieee754_powl): Rename local variable i to __i.
	(__ieee754_atan2, __ieee754_atan2f, __ieee754_atan2l): New inline
	functions.
	* sysdeps/m68k/fpu/e_atan2.c, sysdeps/m68k/fpu/e_atan2f.c,
	sysdeps/m68k/fpu/e_atan2l.c: New files.

1996-09-15  Paul Eggert  <eggert@twinsun.com>

	* manual/time.texi: Change `range X to Y' to `range X through
	Y', to avoid ambiguity in English.

	(strftime): Numbers that do not have a range indicated are not padded.
	Describe E and O modifiers.
	%g, %G, %u: New formats.
	%C, %y, %Y: Describe behavior on negative years.
	%e: Fix typo (was labeled %d).
	%l, %V: Fix typo in range.
	%M, %S, %U, %w, %W: Give ranges.
	%p: Clarify how noon and midnight are handled for AM and PM.
	%s: Clarify leap second handling.
	%r: Now locale-defined.
	%C, %D, %e, %h, %n, %r, %t, %T: Say that they are POSIX.2 extensions.
	%z: Say that it is a GNU extension.
	%Z: Wording fix.

	* time/strftime.c: (strftime):
		%V: Fix mishandling of week numbers near year boundaries.
		%g, %G: New formats (suggested by Arthur David Olson).
		%U, %W: Use inline expression instead of `week' function.
		%C, %y: Handle negative years portably.
		%C, %Y: Use width 1, since values can be arbitrarily wide.
		%r: Use T_FMT_AMPM format if _NL_CURRENT is defined.
		%u: New Posix.2 format.
		%w: Width is 1, not 2.
	(iso_week_days): New function, for %V, %G, %g.
	(week): Remove; it didn't handle %V correctly.
	(__isleap): New macro.
	(mbsinit): Use arg, to pacify GCC -Wall.

1996-09-13  Paul Eggert  <eggert@twinsun.com>

	* time/strftime.c (strftime):
	If using the GNU C library, do not bother to check for
	multibyte encodings, since they're safe in formats.  Otherwise:
	  - Check for multibyte encodings when encountering any character that
	  is not in the basic execution character set of the C Standard.
	  - Use mbrlen (if available) instead of mblen, to avoid modifying
	  mblen's internal state.
	  - Do not assume that '%' cannot appear as the first character of a
	  multibyte character sequence, since this is possible when not in the
	  initial shift state.
	(HAVE_MBRLEN, MULTIBYTE_IS_FORMAT_SAFE): Define if _LIBC is defined.
	(DO_MULTIBYTE): New macro.
	(<ctype.h>): Do not include.
	(<wchar.h>): Include if HAVE_MBRLEN.
	(mbstate_t, mbrlen, mbsinit): Define if ! HAVE_MBRLEN.
	(mbstate_zero): New constant.

1996-09-12  Paul Eggert  <eggert@twinsun.com>

	* time/strftime.c (strftime):
	Use an empty zone if it can't be determined; POSIX.2 requires this.
	Use plain `int' for pad and modifier (which now contain char value).
	Use plain `int' for number_value, to print negative values correctly.
	Use plain `int' for digits; there was no need to make it unsigned.
	Initialize subfmt consistently.
	Remove incorrect code for %EC and %Ey; they aren't implemented yet.
	For %O, if there is no alternate digit, output Ascii instead of "".
	Output the `%' of an unknown format; this is most likely the right
	thing to do if a multibyte string has been misparsed.

Thu Sep 12 23:23:13 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* stdio-common/scanf7.c (main): Remove extra conversion from
	printf format string.

Thu Sep 12 23:01:16 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* stdlib/test-canon.c (tests): Rename structure member from errno
	to error, all uses changed.

Thu Sep 12 20:08:06 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/m68k/fpu/__math.h (__ldexp, __ldexpf, __ldexpl):
	Removed.
	* sysdeps/m68k/fpu/s_ldexp.c, sysdeps/m68k/fpu/s_ldexpf.c,
	sysdeps/m68k/fpu/s_ldexpl.c: Removed, use generic implementation
	instead.
	* sysdeps/m68k/fpu/s_scalbn.c, sysdeps/m68k/fpu/s_scalbnf.c,
	sysdeps/m68k/fpu/s_scalbnl.c: Replaced with old contents of
	s_ldexp.c, s_ldexpf.c and s_ldexpl.c, resp., suitably adpted.

	* sysdeps/m68k/fpu/__math.h (__frexp, __frexpf, __frexpl):
	Return value must be in [0.5, 1), not [1, 2).  Reported by Chris
	Lawrence.
	(__ilogb, __ilogbf, __ilogbl): Check for argument being zero.
	(__scalbn, __scalbnf, __scalbnl): Use second argument directly.

Thu Sep 12 19:59:24 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/libm-ieee754/e_acoshl.c: Fix typos.
	* sysdeps/libm-ieee754/s_cbrtl.c: Remove unused variable.

Thu Sep 12 19:59:24 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* sysdeps/libm-ieee754/e_acoshl.c: Fix typos.
	* sysdeps/libm-ieee754/s_cbrtl.c: Remove unused variable.

Thu Sep 12 19:56:07 1996  Andreas Schwab  <schwab@issan.informatik.uni-dortmund.de>

	* inet/herrno.c (__h_errno_location): Fix return type.

Tue Sep 17 10:51:58 1996  Thomas Bushnell, n/BSG  <thomas@gnu.ai.mit.edu>

	* malloc/memalign.c (memalign): Only acquire __libc_malloc_lock
	for actual modifications to global state.

Fri Sep 13 01:21:36 1996  Thomas Bushnell, n/BSG  <thomas@gnu.ai.mit.edu>

	* sysdeps/mach/Makefile (includes): Add -I$(common-objpfx)mach/.
	* sysdeps/mach/hurd/Makefile (includes): Add
	-I$(common-objpfx)hurd/.  Reported by Marcus Daniels.

	* sysdeps/generic/schedbits.h (struct sched_param): Renamed from
	struct sched_params.
	* sysdeps/stub/sched_setp.c (__sched_setparam): struct
	sched_params -> struct sched_param.
	* sysdeps/stub/sched_getp.c (__sched_getparam): Likewise.
	* sysdeps/stub/sched_sets.c (__sched_setscheduler): Likewise.

Thu Sep 12 23:58:25 1996  Thomas Bushnell, n/BSG  <thomas@gnu.ai.mit.edu>

	* sysdeps/mach/libc-lock.h (__libc_cleanup_region_start): Fix
	syntax error.

	* stdio-common/Makefile: Put lockfile in routines
	unconditionally.
	Define _MT_SAFE_IO if using for libio and compiling reentrant
	libc.

	* stdio-common/vfprintf.c (__flockfile, __funlockfile): Declare
	this always, not just if _LIBC_REENTRANT.
	(__funlockfile): Don't use weak_extern for this one;
	__libc_cleanup_region_end might be defined and the use of
	__funlockfile can't be protected the way the use of __flockfile
	can be.

	* sched.h: New file.  Helper to access posix/sched.h.
Thu Sep 12 12:33:52 1996  Thomas Bushnell, n/BSG  <thomas@gnu.ai.mit.edu>

	* sysdeps/mach/hurd/dl-cache.c: Delete second copy of file
	accidentally added on.

	* sysdeps/stub/intr-msg.h: New file.

	* stdio-common/vfprintf.c: Include <libc-lock.h>.
	* stdio-common/vfscanf.c: Include <libc-lock.h>.

	* sysdeps/mach/libc-lock.h (__libc_cleanup_region_start): New
	macro.
	(__libc_cleanup_region_end): New macro.
1996-09-20 01:58:09 +00:00

314 lines
12 KiB
C

/* Copyright (C) 1991, 92, 93, 94, 96 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
modify it under the terms of the GNU Library General Public License as
published by the Free Software Foundation; either version 2 of the
License, or (at your option) any later version.
The GNU C Library 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
Library General Public License for more details.
You should have received a copy of the GNU Library General Public
License along with the GNU C Library; see the file COPYING.LIB. If
not, write to the Free Software Foundation, Inc., 675 Mass Ave,
Cambridge, MA 02139, USA. */
#ifdef __GNUC__
#include <sys/cdefs.h>
#ifdef __NO_M81_MATH_INLINES
/* This is used when defining the functions themselves. Define them with
__ names, and with `static inline' instead of `extern inline' so the
bodies will always be used, never an external function call. */
#define __m81_u(x) __CONCAT(__,x)
#define __m81_inline static __inline
#else
#define __m81_u(x) x
#define __m81_inline extern __inline
#define __M81_MATH_INLINES 1
#endif
/* Define a const math function. */
#define __m81_defun(rettype, func, args) \
__m81_inline rettype \
__m81_u(func) args __attribute__((__const__)); \
__m81_inline rettype \
__m81_u(func) args
/* Define the three variants of a math function that has a direct
implementation in the m68k fpu. FUNC is the name for C (which will be
suffixed with f and l for the float and long double version, resp). OP
is the name of the fpu operation (without leading f). */
#define __inline_mathop(func, op) \
__m81_defun (double, func, (double __mathop_x)) \
{ \
double __result; \
__asm("f" __STRING(op) "%.x %1, %0" : "=f" (__result) : "f" (__mathop_x));\
return __result; \
} \
__m81_defun (float, func##f, (float __mathop_x)) \
{ \
float __result; \
__asm("f" __STRING(op) "%.x %1, %0" : "=f" (__result) : "f" (__mathop_x));\
return __result; \
} \
__m81_defun (long double, func##l, (long double __mathop_x)) \
{ \
long double __result; \
__asm("f" __STRING(op) "%.x %1, %0" : "=f" (__result) : "f" (__mathop_x));\
return __result; \
}
/* ieee style elementary functions */
__inline_mathop(__ieee754_acos, acos)
__inline_mathop(__ieee754_asin, asin)
__inline_mathop(__ieee754_cosh, cosh)
__inline_mathop(__ieee754_sinh, sinh)
__inline_mathop(__ieee754_exp, etox)
__inline_mathop(__ieee754_log10, log10)
__inline_mathop(__ieee754_log, logn)
__inline_mathop(__ieee754_sqrt, sqrt)
__inline_mathop(__ieee754_atanh, atanh)
__inline_mathop(__atan, atan)
__inline_mathop(__cos, cos)
__inline_mathop(__sin, sin)
__inline_mathop(__tan, tan)
__inline_mathop(__tanh, tanh)
__inline_mathop(__fabs, abs)
__inline_mathop(__sqrt, sqrt)
__inline_mathop(__rint, int)
__inline_mathop(__expm1, etoxm1)
__inline_mathop(__log1p, lognp1)
__inline_mathop(__logb, log2)
__inline_mathop(__significand, getman)
/* This macro contains the definition for the rest of the inline
functions, using __FLOAT_TYPE as the domain type and __S as the suffix
for the function names. */
#define __inline_functions(__float_type, __s) \
__m81_defun (__float_type, \
__ieee754_remainder##__s, (__float_type __x, __float_type __y)) \
{ \
__float_type __result; \
__asm("frem%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x)); \
return __result; \
} \
\
__m81_defun (__float_type, \
__ieee754_fmod##__s, (__float_type __x, __float_type __y)) \
{ \
__float_type __result; \
__asm("fmod%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x)); \
return __result; \
} \
\
__m81_defun (__float_type, \
__ieee754_atan2##__s, (__float_type __y, __float_type __x)) \
{ \
__float_type __pi, __pi_2; \
\
__asm ("fmovecr%.x %#0, %0" : "=f" (__pi)); \
__asm ("fscale%.w %#-1, %0" : "=f" (__pi_2) : "0" (__pi)); \
if (__x > 0) \
{ \
if (__y > 0) \
{ \
if (__x > __y) \
return __m81_u(__atan##__s) (__y / __x); \
else \
return __pi_2 - __m81_u(__atan##__s) (__x / __y); \
} \
else \
{ \
if (__x > -__y) \
return __m81_u(__atan##__s) (__y / __x); \
else \
return -__pi_2 - __m81_u(__atan##__s) (__x / __y); \
} \
} \
else \
{ \
if (__y > 0) \
{ \
if (-__x < __y) \
return __pi + __m81_u(__atan##__s) (__y / __x); \
else \
return __pi_2 - __m81_u(__atan##__s) (__x / __y); \
} \
else \
{ \
if (-__x > -__y) \
return -__pi + __m81_u(__atan##__s) (__y / __x); \
else \
return -__pi_2 - __m81_u(__atan##__s) (__x / __y); \
} \
} \
} \
\
__m81_inline __float_type \
__m81_u(__frexp##__s)(__float_type __value, int *__expptr) \
{ \
__float_type __mantissa, __exponent; \
int __iexponent; \
if (__value == 0.0) \
{ \
*__expptr = 0; \
return __value; \
} \
__asm("fgetexp%.x %1, %0" : "=f" (__exponent) : "f" (__value)); \
__iexponent = (int) __exponent + 1; \
*__expptr = __iexponent; \
__asm("fscale%.l %2, %0" : "=f" (__mantissa) \
: "0" (__value), "dmi" (-__iexponent)); \
return __mantissa; \
} \
\
__m81_defun (__float_type, __floor##__s, (__float_type __x)) \
{ \
__float_type __result; \
unsigned long int __ctrl_reg; \
__asm __volatile__ ("fmove%.l %!, %0" : "=dm" (__ctrl_reg)); \
/* Set rounding towards negative infinity. */ \
__asm __volatile__ ("fmove%.l %0, %!" : /* No outputs. */ \
: "dmi" ((__ctrl_reg & ~0x10) | 0x20)); \
/* Convert X to an integer, using -Inf rounding. */ \
__asm __volatile__ ("fint%.x %1, %0" : "=f" (__result) : "f" (__x)); \
/* Restore the previous rounding mode. */ \
__asm __volatile__ ("fmove%.l %0, %!" : /* No outputs. */ \
: "dmi" (__ctrl_reg)); \
return __result; \
} \
\
__m81_defun (__float_type, \
__ieee754_pow##__s, (__float_type __x, __float_type __y)) \
{ \
__float_type __result; \
if (__x == 0.0) \
{ \
if (__y <= 0.0) \
__result = 0.0 / 0.0; \
else \
__result = 0.0; \
} \
else if (__y == 0.0 || __x == 1.0) \
__result = 1.0; \
else if (__y == 1.0) \
__result = __x; \
else if (__y == 2.0) \
__result = __x * __x; \
else if (__x == 10.0) \
__asm("ftentox%.x %1, %0" : "=f" (__result) : "f" (__y)); \
else if (__x == 2.0) \
__asm("ftwotox%.x %1, %0" : "=f" (__result) : "f" (__y)); \
else if (__x < 0.0) \
{ \
__float_type __temp = __m81_u (__rint##__s) (__y); \
if (__y == __temp) \
{ \
int __i = (int) __y; \
__result = (__m81_u(__ieee754_exp##__s) \
(__y * __m81_u(__ieee754_log##__s) (-__x))); \
if (__i & 1) \
__result = -__result; \
} \
else \
__result = 0.0 / 0.0; \
} \
else \
__result = (__m81_u(__ieee754_exp##__s) \
(__y * __m81_u(__ieee754_log##__s) (__x))); \
return __result; \
} \
\
__m81_defun (__float_type, __ceil##__s, (__float_type __x)) \
{ \
__float_type __result; \
unsigned long int __ctrl_reg; \
__asm __volatile__ ("fmove%.l %!, %0" : "=dm" (__ctrl_reg)); \
/* Set rounding towards positive infinity. */ \
__asm __volatile__ ("fmove%.l %0, %!" : /* No outputs. */ \
: "dmi" (__ctrl_reg | 0x30)); \
/* Convert X to an integer, using +Inf rounding. */ \
__asm __volatile__ ("fint%.x %1, %0" : "=f" (__result) : "f" (__x)); \
/* Restore the previous rounding mode. */ \
__asm __volatile__ ("fmove%.l %0, %!" : /* No outputs. */ \
: "dmi" (__ctrl_reg)); \
return __result; \
} \
\
__m81_inline __float_type \
__m81_u(__modf##__s)(__float_type __value, __float_type *__iptr) \
{ \
__float_type __modf_int; \
__asm ("fintrz%.x %1, %0" : "=f" (__modf_int) : "f" (__value)); \
*__iptr = __modf_int; \
return __value - __modf_int; \
} \
\
__m81_defun (int, __isinf##__s, (__float_type __value)) \
{ \
/* There is no branch-condition for infinity, \
so we must extract and examine the condition codes manually. */ \
unsigned long int __fpsr; \
__asm("ftst%.x %1\n" \
"fmove%.l %/fpsr, %0" : "=dm" (__fpsr) : "f" (__value)); \
return (__fpsr & (2 << 24)) ? (__fpsr & (8 << 24) ? -1 : 1) : 0; \
} \
\
__m81_defun (int, __isnan##__s, (__float_type __value)) \
{ \
char __result; \
__asm("ftst%.x %1\n" \
"fsun %0" : "=dm" (__result) : "f" (__value)); \
return __result; \
} \
\
__m81_defun (int, __finite##__s, (__float_type __value)) \
{ \
/* There is no branch-condition for infinity, so we must extract and \
examine the condition codes manually. */ \
unsigned long int __fpsr; \
__asm ("ftst%.x %1\n" \
"fmove%.l %/fpsr, %0" : "=dm" (__fpsr) : "f" (__value)); \
return (__fpsr & (3 << 24)) == 0; \
} \
\
__m81_defun (int, __ilogb##__s, (__float_type __x)) \
{ \
__float_type __result; \
if (__x == 0.0) \
return 0x80000001; \
__asm("fgetexp%.x %1, %0" : "=f" (__result) : "f" (__x)); \
return (int) __result; \
} \
\
__m81_defun (__float_type, \
__ieee754_scalb##__s, (__float_type __x, __float_type __n)) \
{ \
__float_type __result; \
__asm ("fscale%.x %1, %0" : "=f" (__result) : "f" (__n), "0" (__x)); \
return __result; \
} \
\
__m81_defun (__float_type, __scalbn##__s, (__float_type __x, int __n)) \
{ \
__float_type __result; \
__asm ("fscale%.l %1, %0" : "=f" (__result) : "dmi" (__n), "0" (__x)); \
return __result; \
}
/* This defines the three variants of the inline functions. */
__inline_functions (double, )
__inline_functions (float, f)
__inline_functions (long double, l)
#undef __inline_functions
#endif /* GCC. */