f839bac29c
on i386 (A64), 5 cycles on amd64 (A64), and 3 cycles on ia64). gcc tends to generate very bad code for accessing floating point values as bits except when the integer accesses have the same width as the floating point values, and direct accesses to bit-fields (as is common only for long double precision) always gives such accesses. Use the expsign access method, which is good for 80-bit long doubles and hopefully no worse for 128-bit long doubles. Now the generated code is less bad. There is still unnecessary copying of the arg on amd64 and i386 and mysterious extra slowness on amd64.
91 lines
2.7 KiB
C
91 lines
2.7 KiB
C
/*-
|
|
* Copyright (c) 2008 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.
|
|
*/
|
|
|
|
#include <sys/cdefs.h>
|
|
__FBSDID("$FreeBSD$");
|
|
|
|
#include <float.h>
|
|
#include <math.h>
|
|
|
|
#include "fpmath.h"
|
|
|
|
#if LDBL_MAX_EXP != 0x4000
|
|
/* We also require the usual bias, min exp and expsign packing. */
|
|
#error "Unsupported long double format"
|
|
#endif
|
|
|
|
#define BIAS (LDBL_MAX_EXP - 1)
|
|
|
|
static const float
|
|
shift[2] = {
|
|
#if LDBL_MANT_DIG == 64
|
|
0x1.0p63, -0x1.0p63
|
|
#elif LDBL_MANT_DIG == 113
|
|
0x1.0p112, -0x1.0p112
|
|
#else
|
|
#error "Unsupported long double format"
|
|
#endif
|
|
};
|
|
static const float zero[2] = { 0.0, -0.0 };
|
|
|
|
long double
|
|
rintl(long double x)
|
|
{
|
|
union IEEEl2bits u;
|
|
uint32_t expsign;
|
|
int ex, sign;
|
|
|
|
u.e = x;
|
|
expsign = u.xbits.expsign;
|
|
ex = expsign & 0x7fff;
|
|
|
|
if (ex >= BIAS + LDBL_MANT_DIG - 1) {
|
|
if (ex == BIAS + LDBL_MAX_EXP)
|
|
return (x + x); /* Inf, NaN, or unsupported format */
|
|
return (x); /* finite and already an integer */
|
|
}
|
|
sign = expsign >> 15;
|
|
|
|
/*
|
|
* The following code assumes that intermediate results are
|
|
* evaluated in long double precision. If they are evaluated in
|
|
* greater precision, double rounding may occur, and if they are
|
|
* evaluated in less precision (as on i386), results will be
|
|
* wildly incorrect.
|
|
*/
|
|
x += shift[sign];
|
|
x -= shift[sign];
|
|
|
|
/*
|
|
* If the result is +-0, then it must have the same sign as x, but
|
|
* the above calculation doesn't always give this. Fix up the sign.
|
|
*/
|
|
if (ex < BIAS && x == 0.0L)
|
|
return (zero[sign]);
|
|
|
|
return (x);
|
|
}
|