Source-Changes-HG archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

[src/trunk]: src/lib/libc/arch/m68k/gen ldexp(denormal, exp>1023) would gener...



details:   https://anonhg.NetBSD.org/src/rev/39263fb615b6
branches:  trunk
changeset: 475925:39263fb615b6
user:      mycroft <mycroft%NetBSD.org@localhost>
date:      Sun Aug 29 22:50:25 1999 +0000

description:
ldexp(denormal, exp>1023) would generate the wrong result in all non-overflow
cases.  Totally rework this code to fix this bug *and* make it faster.

diffstat:

 lib/libc/arch/m68k/gen/ldexp_040.c |  184 ++++++++++++++++++------------------
 1 files changed, 93 insertions(+), 91 deletions(-)

diffs (239 lines):

diff -r 88e0f353c5a0 -r 39263fb615b6 lib/libc/arch/m68k/gen/ldexp_040.c
--- a/lib/libc/arch/m68k/gen/ldexp_040.c        Sun Aug 29 22:45:41 1999 +0000
+++ b/lib/libc/arch/m68k/gen/ldexp_040.c        Sun Aug 29 22:50:25 1999 +0000
@@ -1,12 +1,11 @@
-/*     $NetBSD: ldexp_040.c,v 1.2 1999/08/29 19:42:54 mycroft Exp $    */
+/*     $NetBSD: ldexp_040.c,v 1.3 1999/08/29 22:50:25 mycroft Exp $    */
 
-/*
- * Copyright (c) 1992, 1993
- *     The Regents of the University of California.  All rights reserved.
+/*-
+ * Copyright (c) 1999 The NetBSD Foundation, Inc.
+ * All rights reserved.
  *
- * This software was developed by the Computer Systems Engineering group
- * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
- * contributed to Berkeley.
+ * This code is derived from software contributed to The NetBSD Foundation
+ * by Charles M. Hannum.
  *
  * Redistribution and use in source and binary forms, with or without
  * modification, are permitted provided that the following conditions
@@ -18,34 +17,28 @@
  *    documentation and/or other materials provided with the distribution.
  * 3. All advertising materials mentioning features or use of this software
  *    must display the following acknowledgement:
- *     This product includes software developed by the University of
- *     California, Berkeley and its contributors.
- * 4. Neither the name of the University nor the names of its contributors
- *    may be used to endorse or promote products derived from this software
- *    without specific prior written permission.
+ *        This product includes software developed by the NetBSD
+ *        Foundation, Inc. and its contributors.
+ * 4. Neither the name of The NetBSD Foundation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
  *
- * THIS SOFTWARE IS PROVIDED BY THE REGENTS 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 REGENTS 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.
- *
- * from: Header: ldexp.c,v 1.1 91/07/07 04:28:19 torek Exp
+ * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. 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 FOUNDATION 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 defined(LIBC_SCCS) && !defined(lint)
-#if 0
-static const char sccsid[] = "@(#)ldexp.c      8.1 (Berkeley) 6/4/93";
-#else
-__RCSID("$NetBSD: ldexp_040.c,v 1.2 1999/08/29 19:42:54 mycroft Exp $");
-#endif
+__RCSID("$NetBSD: ldexp_040.c,v 1.3 1999/08/29 22:50:25 mycroft Exp $");
 #endif /* LIBC_SCCS and not lint */
 
 #include <sys/types.h>
@@ -54,8 +47,7 @@
 #include <math.h>
 
 /*
- * double ldexp(double val, int exp)
- * returns: val * (2**exp)
+ * Multiply the given value by 2^exp.
  */
 double
 ldexp(val, exp)
@@ -68,85 +60,95 @@
                struct ieee_double s;
        } u, mul;
 
-       /*
-        * If input is zero, or no change, just return input.
-        * Likewise, if input is Inf or NaN, just return it.
-        */
        u.v = val;
        oldexp = u.s.dbl_exp;
-       if (val == 0.0 || exp == 0 || oldexp == DBL_EXP_INFNAN)
-               return (val);
 
        /*
-        * Compute new exponent and check for over/under flow.
-        * Underflow, unfortunately, could mean switching to denormal.
-        * If result out of range, set ERANGE and return 0 if too small
-        * or Inf if too big, with the same sign as the input value.
+        * If input is zero, Inf or NaN, just return it.
+        */
+       if (u.v == 0.0 || oldexp == DBL_EXP_INFNAN)
+               return (val);
+
+       if (oldexp == 0) {
+               /*
+                * u.v is denormal.  We must adjust it so that the exponent
+                * arithmetic below will work.
+                */
+               if (exp <= DBL_EXP_BIAS) {
+                       /*
+                        * Optimization: if the scaling can be done in a single
+                        * multiply, or underflows, just do it now.
+                        */
+                       if (exp <= -DBL_FRACBITS) {
+                               errno = ERANGE;
+                               return (0.0);
+                       }
+                       mul.v = 1.0;
+                       mul.s.dbl_exp = exp + DBL_EXP_BIAS;
+                       u.v *= mul.v;
+                       if (u.v == 0.0) {
+                               errno = ERANGE;
+                               return (0.0);
+                       }
+                       return (u.v);
+               } else {
+                       /*
+                        * We know that exp is very large, and therefore the
+                        * result cannot be denormal (though it may be Inf).
+                        * Shift u.v by just enough to make it normal.
+                        */
+                       mul.v = 1.0;
+                       mul.s.dbl_exp = DBL_FRACBITS + DBL_EXP_BIAS;
+                       u.v *= mul.v;
+                       exp -= DBL_FRACBITS;
+                       oldexp = u.s.dbl_exp;
+               }
+       }
+
+       /*
+        * u.v is now normalized and oldexp has been adjusted if necessary.
+        * Calculate the new exponent and check for underflow and overflow.
         */
        newexp = oldexp + exp;
+
        if (newexp >= DBL_EXP_INFNAN) {
-               /* u.s.dbl_sign = val < 0.0; -- already set */
+               /*
+                * The result overflowed; return +/-Inf.
+                */
                u.s.dbl_exp = DBL_EXP_INFNAN;
-               u.s.dbl_frach = u.s.dbl_fracl = 0;
+               u.s.dbl_frach = 0;
+               u.s.dbl_fracl = 0;
                errno = ERANGE;
-               return (u.v);           /* Inf */
-       }
-       if (newexp <= 0) {
+               return (u.v);
+       } else if (newexp <= 0) {
                /*
-                * The output number is either a denormal or underflows
-                * (see comments in machine/ieee.h).
+                * The output number is either denormal or underflows (see
+                * comments in machine/ieee.h).
                 */
                if (newexp <= -DBL_FRACBITS) {
                        errno = ERANGE;
                        return (0.0);
                }
                /*
-                * We are going to produce a denorm.  Our `exp' argument
-                * might be as small as -2097, and we cannot compute
-                * 2^-2097, so we may have to do this as many as three
-                * steps (not just two, as for positive `exp's below).
+                * Denormalize the result.  We do this with a multiply.  If exp
+                * is very large, it won't fit in a double, so we have to
+                * adjust the exponent first.  This is safe because we know
+                * that u.v is normal at this point.
                 */
+               if (exp <= -DBL_EXP_BIAS) {
+                       u.s.dbl_exp = 1;
+                       exp += oldexp - 1;
+               }
                mul.v = 1.0;
-               while (exp <= -DBL_EXP_BIAS) {
-                       mul.s.dbl_exp = 1;
-                       val *= mul.v;
-                       exp += DBL_EXP_BIAS - 1;
-               }
                mul.s.dbl_exp = exp + DBL_EXP_BIAS;
-               val *= mul.v;
-               return (val);
-       }
-
-       /*
-        * Newexp is positive.
-        *
-        * If oldexp is zero, we are starting with a denorm, and simply
-        * adjusting the exponent will produce bogus answers.  We need
-        * to fix that first.
-        */
-       if (oldexp == 0) {
+               u.v *= mul.v;
+               return (u.v);
+       } else {
                /*
-                * Multiply by 2^mulexp to make the number normalizable.
-                * We cannot multiply by more than 2^1023, but `exp'
-                * argument might be as large as 2046.  A single
-                * adjustment, however, will normalize the number even
-                * for huge `exp's, and then we can use exponent
-                * arithmetic just as for normal `double's.
+                * The result is normal; just replace the old exponent with the
+                * new one.
                 */
-               mulexp = exp <= DBL_EXP_BIAS ? exp : DBL_EXP_BIAS;
-               mul.v = 1.0;
-               mul.s.dbl_exp = mulexp + DBL_EXP_BIAS;
-               val *= mul.v;
-               if (mulexp == exp)
-                       return (val);
-               u.v = val;
-               newexp -= mulexp;
+               u.s.dbl_exp = newexp;
+               return (u.v);
        }
-
-       /*
-        * Both oldexp and newexp are positive; just replace the
-        * old exponent with the new one.
-        */
-       u.s.dbl_exp = newexp;
-       return (u.v);
 }



Home | Main Index | Thread Index | Old Index