glibc/sysdeps/m68k/m680x0/fpu/mathimpl.h

318 lines
11 KiB
C
Raw Normal View History

/* Definitions of libc internal inline math functions implemented
by the m68881/2.
Copyright (C) 1991-2020 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
2001-07-06 04:56:23 +00:00
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.
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
2001-07-06 04:56:23 +00:00
Lesser General Public License for more details.
2001-07-06 04:56:23 +00:00
You should have received a copy of the GNU Lesser General Public
2012-03-09 23:56:38 +00:00
License along with the GNU C Library. If not, see
Prefer https to http for gnu.org and fsf.org URLs Also, change sources.redhat.com to sourceware.org. This patch was automatically generated by running the following shell script, which uses GNU sed, and which avoids modifying files imported from upstream: sed -ri ' s,(http|ftp)(://(.*\.)?(gnu|fsf|sourceware)\.org($|[^.]|\.[^a-z])),https\2,g s,(http|ftp)(://(.*\.)?)sources\.redhat\.com($|[^.]|\.[^a-z]),https\2sourceware.org\4,g ' \ $(find $(git ls-files) -prune -type f \ ! -name '*.po' \ ! -name 'ChangeLog*' \ ! -path COPYING ! -path COPYING.LIB \ ! -path manual/fdl-1.3.texi ! -path manual/lgpl-2.1.texi \ ! -path manual/texinfo.tex ! -path scripts/config.guess \ ! -path scripts/config.sub ! -path scripts/install-sh \ ! -path scripts/mkinstalldirs ! -path scripts/move-if-change \ ! -path INSTALL ! -path locale/programs/charmap-kw.h \ ! -path po/libc.pot ! -path sysdeps/gnu/errlist.c \ ! '(' -name configure \ -execdir test -f configure.ac -o -f configure.in ';' ')' \ ! '(' -name preconfigure \ -execdir test -f preconfigure.ac ';' ')' \ -print) and then by running 'make dist-prepare' to regenerate files built from the altered files, and then executing the following to cleanup: chmod a+x sysdeps/unix/sysv/linux/riscv/configure # Omit irrelevant whitespace and comment-only changes, # perhaps from a slightly-different Autoconf version. git checkout -f \ sysdeps/csky/configure \ sysdeps/hppa/configure \ sysdeps/riscv/configure \ sysdeps/unix/sysv/linux/csky/configure # Omit changes that caused a pre-commit check to fail like this: # remote: *** error: sysdeps/powerpc/powerpc64/ppc-mcount.S: trailing lines git checkout -f \ sysdeps/powerpc/powerpc64/ppc-mcount.S \ sysdeps/unix/sysv/linux/s390/s390-64/syscall.S # Omit change that caused a pre-commit check to fail like this: # remote: *** error: sysdeps/sparc/sparc64/multiarch/memcpy-ultra3.S: last line does not end in newline git checkout -f sysdeps/sparc/sparc64/multiarch/memcpy-ultra3.S
2019-09-07 05:40:42 +00:00
<https://www.gnu.org/licenses/>. */
#ifndef _MATHIMPL_H
#define _MATHIMPL_H
/* This file contains the definitions of the inline math functions that
are only used internally inside libm, not visible to the user. */
#define __MATH_INLINE __extern_inline
/* 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.
Note: GCC 6 objects to __attribute__ ((__leaf__)) on static functions. */
#define __m81_u(x) __CONCAT(__,x)
#define __m81_inline static __inline
#define __m81_nth(fn) __NTH (fn)
/* Define a math function. */
#define __m81_defun(rettype, func, args, attrs) \
__m81_inline rettype attrs \
__m81_nth (__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, attrs) \
__inline_mathop1(double, func, op, attrs) \
__inline_mathop1(float, __CONCAT(func,f), op, attrs) \
__inline_mathop1(long double, __CONCAT(func,l), op, attrs)
#define __inline_mathop1(float_type,func, op, attrs) \
__m81_defun (float_type, func, (float_type __mathop_x), attrs) \
{ \
float_type __result; \
__asm __volatile__ ("f" __STRING(op) "%.x %1, %0" \
: "=f" (__result) : "f" (__mathop_x)); \
return __result; \
}
__inline_mathop(__atan, atan,)
__inline_mathop(__cos, cos,)
__inline_mathop(__sin, sin,)
__inline_mathop(__tan, tan,)
__inline_mathop(__tanh, tanh,)
__inline_mathop(__fabs, abs, __attribute__ ((__const__)))
__inline_mathop(__rint, int,)
__inline_mathop(__expm1, etoxm1,)
__inline_mathop(__log1p, lognp1,)
__inline_mathop(__significand, getman,)
__inline_mathop(__trunc, intrz, __attribute__ ((__const__)))
/* This macro contains the definition for the rest of the inline
functions, using FLOAT_TYPE as the domain type and M as a macro
that adds the suffix for the function names. */
#define __inline_functions(float_type, m) \
__m81_defun (float_type, m(__floor), (float_type __x), \
__attribute__ ((__const__))) \
{ \
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, m(__ceil), (float_type __x), \
__attribute__ ((__const__))) \
{ \
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; \
}
#define __CONCAT_d(arg) arg
#define __CONCAT_f(arg) arg ## f
#define __CONCAT_l(arg) arg ## l
__inline_functions(double, __CONCAT_d)
__inline_functions(float, __CONCAT_f)
__inline_functions(long double, __CONCAT_l)
#undef __inline_functions
# define __inline_functions(float_type, m) \
__m81_defun (int, m(__isinf), (float_type __value), \
__attribute__ ((__const__))) \
{ \
/* 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, m(__finite), (float_type __value), \
__attribute__ ((__const__))) \
{ \
/* 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 (float_type, m(__scalbn), \
(float_type __x, int __n),) \
{ \
float_type __result; \
__asm __volatile__ ("fscale%.l %1, %0" : "=f" (__result) \
: "dmi" (__n), "0" (__x)); \
return __result; \
}
__inline_functions(double, __CONCAT_d)
__inline_functions(float, __CONCAT_f)
__inline_functions(long double, __CONCAT_l)
#undef __inline_functions
# define __inline_functions(float_type, m) \
__m81_defun (int, m(__isnan), (float_type __value), \
__attribute__ ((__const__))) \
{ \
char __result; \
__asm ("ftst%.x %1\n" \
"fsun %0" : "=dm" (__result) : "f" (__value)); \
return __result; \
}
__inline_functions(double, __CONCAT_d)
__inline_functions(float, __CONCAT_f)
__inline_functions(long double, __CONCAT_l)
#undef __inline_functions
# define __inline_functions(float_type, m) \
__m81_defun (float_type, m(__scalbln), \
(float_type __x, long int __n),) \
{ \
return m(__scalbn) (__x, __n); \
} \
\
__m81_defun (float_type, m(__nearbyint), (float_type __x),) \
{ \
float_type __result; \
unsigned long int __ctrl_reg; \
__asm __volatile__ ("fmove%.l %!, %0" : "=dm" (__ctrl_reg)); \
/* Temporarily disable the inexact exception. */ \
__asm __volatile__ ("fmove%.l %0, %!" : /* No outputs. */ \
: "dmi" (__ctrl_reg & ~0x200)); \
__asm __volatile__ ("fint%.x %1, %0" : "=f" (__result) : "f" (__x)); \
__asm __volatile__ ("fmove%.l %0, %!" : /* No outputs. */ \
: "dmi" (__ctrl_reg)); \
return __result; \
} \
\
__m81_defun (long int, m(__lrint), (float_type __x),) \
{ \
long int __result; \
__asm __volatile__ ("fmove%.l %1, %0" : "=dm" (__result) : "f" (__x)); \
return __result; \
}
__inline_functions (double, __CONCAT_d)
__inline_functions (float, __CONCAT_f)
__inline_functions (long double, __CONCAT_l)
#undef __inline_functions
#define __inline_functions(float_type, m) \
__m81_inline void \
__m81_nth (__m81_u(m(__sincos)) \
(float_type __x, float_type *__sinx, float_type *__cosx)) \
{ \
__asm __volatile__ ("fsincos%.x %2,%1:%0" \
: "=f" (*__sinx), "=f" (*__cosx) : "f" (__x)); \
}
__inline_functions (double, __CONCAT_d)
__inline_functions (float, __CONCAT_f)
__inline_functions (long double, __CONCAT_l)
#undef __inline_functions
#undef __CONCAT_d
#undef __CONCAT_f
#undef __CONCAT_l
/* 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, attrs) \
__inline_mathop1(double, func, op, attrs) \
__inline_mathop1(float, __CONCAT(func,f), op, attrs) \
__inline_mathop1(long double, __CONCAT(func,l), op, attrs)
#define __inline_mathop1(float_type,func, op, attrs) \
__m81_defun (float_type, func, (float_type __mathop_x), attrs) \
{ \
float_type __result; \
__asm __volatile__ ("f" __STRING(op) "%.x %1, %0" \
: "=f" (__result) : "f" (__mathop_x)); \
return __result; \
}
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__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_exp2, twotox,)
__inline_mathop (__ieee754_exp10, tentox,)
__inline_mathop (__ieee754_log10, log10,)
__inline_mathop (__ieee754_log2, log2,)
__inline_mathop (__ieee754_log, logn,)
__inline_mathop (__ieee754_sqrt, sqrt,)
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__inline_mathop (__ieee754_atanh, atanh,)
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__m81_defun (double, __ieee754_remainder, (double __x, double __y),)
{
double __result;
__asm ("frem%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x));
return __result;
}
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__m81_defun (float, __ieee754_remainderf, (float __x, float __y),)
{
float __result;
__asm ("frem%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x));
return __result;
}
__m81_defun (long double,
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__ieee754_remainderl, (long double __x, long double __y),)
{
long double __result;
__asm ("frem%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x));
return __result;
}
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__m81_defun (double, __ieee754_fmod, (double __x, double __y),)
{
double __result;
__asm ("fmod%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x));
return __result;
}
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__m81_defun (float, __ieee754_fmodf, (float __x, float __y),)
{
float __result;
__asm ("fmod%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x));
return __result;
}
__m81_defun (long double,
Fix m68k bits/mathinline.h attributes (bug 22631). m68k bits/mathinline.h declares various functions with const attributes. These are inappropriate for functions that have results depending on the rounding mode; the machine-independent bits/mathcalls.h only uses const attributes for a very few functions with no rounding mode dependence, and the m68k header should do likewise. GCC uses pure for such functions with -frounding-math, resulting in GCC mainline warning for conflicts with between the header and the built-in attributes and glibc failing to build for m68k with GCC mainline. This patch fixes the attributes to avoid using const except when bits/mathcalls.h does so. (There are a few functions where maybe bits/mathcalls.h could do so but doesn't, but keeping the headers in sync in this regard seems to be the safe approach.) Tested compilation with build-many-glibcs.py with GCC mainline. [BZ #22631] * sysdeps/m68k/m680x0/fpu/bits/mathinline.h (__m81_defun): Add argument for attrubutes. All callers changed. (__inline_mathop1): Likewise. All callers changed. (__inline_mathop): Likewise. All callers changed. [__USE_MISC] (scalbn): Use __inline_forward instead of __inline_forward_c. [__USE_ISOC99] (scalbln): Likewise. [__USE_ISOC99] (nearbyint): Likewise. [__USE_ISOC99] (lrint): Likewise. [__USE_MISC] (scalbnf): Likewise. [__USE_ISOC99] (scalblnf): Likewise. [__USE_ISOC99] (nearbyintf): Likewise. [__USE_ISOC99] (lrintf): Likewise. [__USE_MISC] (scalbnl): Likewise. [__USE_ISOC99] (scalblnl): Likewise. [__USE_ISOC99] (nearbyintl): Likewise. [__USE_ISOC99] (lrintl): Likewise. * sysdeps/m68k/m680x0/fpu/mathimpl.h: All callers of __inline_mathop and __m81_defun changed.
2017-12-19 02:02:26 +00:00
__ieee754_fmodl, (long double __x, long double __y),)
{
long double __result;
__asm ("fmod%.x %1, %0" : "=f" (__result) : "f" (__y), "0" (__x));
return __result;
}
/* Get the m68881 condition codes, to quickly check multiple conditions. */
static __inline__ unsigned long
__m81_test (long double __val)
{
unsigned long __fpsr;
__asm ("ftst%.x %1; fmove%.l %/fpsr,%0" : "=dm" (__fpsr) : "f" (__val));
return __fpsr;
}
/* Bit values returned by __m81_test. */
#define __M81_COND_NAN (1 << 24)
#define __M81_COND_INF (2 << 24)
#define __M81_COND_ZERO (4 << 24)
#define __M81_COND_NEG (8 << 24)
#endif /* _MATHIMPL_H */