4ec986eab3
registers as volatile. Instructions that *wrote* to FP state were already marked volatile, but apparently gcc has license to move non-volatile asms past volatile asms. This broke amd64's feupdateenv at -O2 due to a WAR conflict between fnstsw and fldenv there.
245 lines
5.2 KiB
C
245 lines
5.2 KiB
C
/*-
|
|
* Copyright (c) 2004 David Schultz <das@FreeBSD.ORG>
|
|
* All rights reserved.
|
|
*
|
|
* 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.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``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 THE AUTHOR OR CONTRIBUTORS 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.
|
|
*
|
|
* $FreeBSD$
|
|
*/
|
|
|
|
#ifndef _FENV_H_
|
|
#define _FENV_H_
|
|
|
|
#include <sys/_types.h>
|
|
|
|
typedef __uint64_t fenv_t;
|
|
typedef __uint64_t fexcept_t;
|
|
|
|
/* Exception flags */
|
|
#define FE_INVALID 0x00000200
|
|
#define FE_DIVBYZERO 0x00000040
|
|
#define FE_OVERFLOW 0x00000100
|
|
#define FE_UNDERFLOW 0x00000080
|
|
#define FE_INEXACT 0x00000020
|
|
#define FE_ALL_EXCEPT (FE_DIVBYZERO | FE_INEXACT | \
|
|
FE_INVALID | FE_OVERFLOW | FE_UNDERFLOW)
|
|
|
|
/*
|
|
* Rounding modes
|
|
*
|
|
* We can't just use the hardware bit values here, because that would
|
|
* make FE_UPWARD and FE_DOWNWARD negative, which is not allowed.
|
|
*/
|
|
#define FE_TONEAREST 0x0
|
|
#define FE_TOWARDZERO 0x1
|
|
#define FE_UPWARD 0x2
|
|
#define FE_DOWNWARD 0x3
|
|
#define _ROUND_MASK (FE_TONEAREST | FE_DOWNWARD | \
|
|
FE_UPWARD | FE_TOWARDZERO)
|
|
#define _ROUND_SHIFT 30
|
|
|
|
__BEGIN_DECLS
|
|
|
|
/* Default floating-point environment */
|
|
extern const fenv_t __fe_dfl_env;
|
|
#define FE_DFL_ENV (&__fe_dfl_env)
|
|
|
|
/* We need to be able to map status flag positions to mask flag positions */
|
|
#define _FPUSW_SHIFT 18
|
|
#define _ENABLE_MASK (FE_ALL_EXCEPT << _FPUSW_SHIFT)
|
|
|
|
#define __ldxfsr(__r) __asm __volatile("ldx %0, %%fsr" : : "m" (__r))
|
|
#define __stxfsr(__r) __asm __volatile("stx %%fsr, %0" : "=m" (*(__r)))
|
|
|
|
static __inline int
|
|
feclearexcept(int __excepts)
|
|
{
|
|
fexcept_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
__r &= ~__excepts;
|
|
__ldxfsr(__r);
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
fegetexceptflag(fexcept_t *__flagp, int __excepts)
|
|
{
|
|
fexcept_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
*__flagp = __r & __excepts;
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
fesetexceptflag(const fexcept_t *__flagp, int __excepts)
|
|
{
|
|
fexcept_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
__r &= ~__excepts;
|
|
__r |= *__flagp & __excepts;
|
|
__ldxfsr(__r);
|
|
return (0);
|
|
}
|
|
|
|
/*
|
|
* In contrast with the ia64 platform, it seems to be worthwhile to
|
|
* inline this function on sparc64 even when the arguments are not
|
|
* compile-time constants. Perhaps this depends on the register window.
|
|
*/
|
|
static __inline int
|
|
feraiseexcept(int __excepts)
|
|
{
|
|
volatile double d;
|
|
|
|
/*
|
|
* With a compiler that supports the FENV_ACCESS pragma
|
|
* properly, simple expressions like '0.0 / 0.0' should
|
|
* be sufficient to generate traps. Unfortunately, we
|
|
* need to bring a volatile variable into the equation
|
|
* to prevent incorrect optimizations.
|
|
*/
|
|
if (__excepts & FE_INVALID) {
|
|
d = 0.0;
|
|
d = 0.0 / d;
|
|
}
|
|
if (__excepts & FE_DIVBYZERO) {
|
|
d = 0.0;
|
|
d = 1.0 / d;
|
|
}
|
|
if (__excepts & FE_OVERFLOW) {
|
|
d = 0x1.ffp1023;
|
|
d *= 2.0;
|
|
}
|
|
if (__excepts & FE_UNDERFLOW) {
|
|
d = 0x1p-1022;
|
|
d /= 0x1p1023;
|
|
}
|
|
if (__excepts & FE_INEXACT) {
|
|
d = 0x1p-1022;
|
|
d += 1.0;
|
|
}
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
fetestexcept(int __excepts)
|
|
{
|
|
fexcept_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
return (__r & __excepts);
|
|
}
|
|
|
|
static __inline int
|
|
fegetround(void)
|
|
{
|
|
fenv_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
return ((__r >> _ROUND_SHIFT) & _ROUND_MASK);
|
|
}
|
|
|
|
static __inline int
|
|
fesetround(int __round)
|
|
{
|
|
fenv_t __r;
|
|
|
|
if (__round & ~_ROUND_MASK)
|
|
return (-1);
|
|
__stxfsr(&__r);
|
|
__r &= ~(_ROUND_MASK << _ROUND_SHIFT);
|
|
__r |= __round << _ROUND_SHIFT;
|
|
__ldxfsr(__r);
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
fegetenv(fenv_t *__envp)
|
|
{
|
|
|
|
__stxfsr(__envp);
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
feholdexcept(fenv_t *__envp)
|
|
{
|
|
fenv_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
*__envp = __r;
|
|
__r &= ~(FE_ALL_EXCEPT | _ENABLE_MASK);
|
|
__ldxfsr(__r);
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
fesetenv(const fenv_t *__envp)
|
|
{
|
|
|
|
__ldxfsr(*__envp);
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
feupdateenv(const fenv_t *__envp)
|
|
{
|
|
fexcept_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
__ldxfsr(*__envp);
|
|
feraiseexcept(__r & FE_ALL_EXCEPT);
|
|
return (0);
|
|
}
|
|
|
|
#if __BSD_VISIBLE
|
|
|
|
static __inline int
|
|
fesetmask(int __mask)
|
|
{
|
|
fenv_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
__r &= ~_ENABLE_MASK;
|
|
__r |= __mask << _FPUSW_SHIFT;
|
|
__ldxfsr(__r);
|
|
return (0);
|
|
}
|
|
|
|
static __inline int
|
|
fegetmask(void)
|
|
{
|
|
fenv_t __r;
|
|
|
|
__stxfsr(&__r);
|
|
return ((__r & _ENABLE_MASK) >> _FPUSW_SHIFT);
|
|
}
|
|
|
|
#endif /* __BSD_VISIBLE */
|
|
|
|
__END_DECLS
|
|
|
|
#endif /* !_FENV_H_ */
|