[BACK]Return to s_fmaf.c CVS log [TXT][DIR] Up to [cvs.NetBSD.org] / src / lib / libm / src

File: [cvs.NetBSD.org] / src / lib / libm / src / s_fmaf.c (download)

Revision 1.3, Sat May 6 18:02:52 2017 UTC (6 years, 10 months ago) by christos
Branch: MAIN
CVS Tags: prg-localcount2-base3, prg-localcount2-base2, phil-wifi-base, phil-wifi-20200421, phil-wifi-20200411, phil-wifi-20200406, phil-wifi-20191119, phil-wifi-20190609, phil-wifi, pgoyette-compat-merge-20190127, pgoyette-compat-base, pgoyette-compat-20190127, pgoyette-compat-20190118, pgoyette-compat-1226, pgoyette-compat-1126, pgoyette-compat-1020, pgoyette-compat-0930, pgoyette-compat-0906, pgoyette-compat-0728, pgoyette-compat-0625, pgoyette-compat-0521, pgoyette-compat-0502, pgoyette-compat-0422, pgoyette-compat-0415, pgoyette-compat-0407, pgoyette-compat-0330, pgoyette-compat-0322, pgoyette-compat-0315, pgoyette-compat, perseant-stdc-iso10646-base, perseant-stdc-iso10646, netbsd-9-base, netbsd-9-3-RELEASE, netbsd-9-2-RELEASE, netbsd-9-1-RELEASE, netbsd-9-0-RELEASE, netbsd-9-0-RC2, netbsd-9-0-RC1, netbsd-9, netbsd-8-base, netbsd-8-2-RELEASE, netbsd-8-1-RELEASE, netbsd-8-1-RC1, netbsd-8-0-RELEASE, netbsd-8-0-RC2, netbsd-8-0-RC1, netbsd-8, netbsd-10-base, netbsd-10-0-RELEASE, netbsd-10-0-RC6, netbsd-10-0-RC5, netbsd-10-0-RC4, netbsd-10-0-RC3, netbsd-10-0-RC2, netbsd-10-0-RC1, netbsd-10, matt-nb8-mediatek-base, matt-nb8-mediatek, is-mlppp-base, is-mlppp, cjep_sun2x-base1, cjep_sun2x-base, cjep_sun2x, cjep_staticlib_x-base1, cjep_staticlib_x-base, cjep_staticlib_x, HEAD
Changes since 1.2: +4 -2 lines

add missing namespace.h

/*	$NetBSD: s_fmaf.c,v 1.3 2017/05/06 18:02:52 christos Exp $	*/

/*-
 * Copyright (c) 2005-2011 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>
#if 0
__FBSDID("$FreeBSD: src/lib/msun/src/s_fmaf.c,v 1.3 2011/10/15 04:16:58 das Exp $");
#else
__RCSID("$NetBSD: s_fmaf.c,v 1.3 2017/05/06 18:02:52 christos Exp $");
#endif

#include "namespace.h"

#include <fenv.h>

#include "math.h"
#include "math_private.h"

/*
 * Fused multiply-add: Compute x * y + z with a single rounding error.
 *
 * A double has more than twice as much precision than a float, so
 * direct double-precision arithmetic suffices, except where double
 * rounding occurs.
 */
float
fmaf(float x, float y, float z)
{
	double xy, result;
	uint32_t hr, lr;

	xy = (double)x * y;
	result = xy + z;
	EXTRACT_WORDS(hr, lr, result);
	/* Common case: The double precision result is fine. */
	if ((lr & 0x1fffffff) != 0x10000000 ||	/* not a halfway case */
	    (hr & 0x7ff00000) == 0x7ff00000 ||	/* NaN */
	    result - xy == z ||			/* exact */
	    fegetround() != FE_TONEAREST)	/* not round-to-nearest */
		return (result);

	/*
	 * If result is inexact, and exactly halfway between two float values,
	 * we need to adjust the low-order bit in the direction of the error.
	 */
	fesetround(FE_TOWARDZERO);
	volatile double vxy = xy;  /* XXX work around gcc CSE bug */
	double adjusted_result = vxy + z;
	fesetround(FE_TONEAREST);
	if (result == adjusted_result)
		SET_LOW_WORD(adjusted_result, lr + 1);
	return (adjusted_result);
}