1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
|
Copyright 1999, 2001, 2002, 2003, 2004, 2005 Free Software Foundation.
Contributed by the Spaces project, INRIA Lorraine.
This file is part of the MPFR Library.
The MPFR Library is free software; you can redistribute it and/or 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 MPFR 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 Lesser General Public
License for more details.
You should have received a copy of the GNU Lesser General Public License
along with the MPFR Library; see the file COPYING.LIB. If not, write to
the Free Software Foundation, Inc., 59 Temple Place - Suite 330, Boston,
MA 02111-1307, USA.
##############################################################################
Probably many bugs.
Known bugs:
* The overflows/underflows are partially implemented in some functions.
For instance, mpfr_pow (z, x, y, rnd) fails for very small x and some
values of y.
* The mpfr_set_ld function assumes that the long double type has an
exponent of at most 15 bits.
* When MPFR is compiled with Visual C++ under MS Windows, some tests may
fail, because of the way special IEEE-754 double values are tested.
Potential bugs:
* Possible integer overflows on some machines. Some types are used
inconsistently. Possible bugs with huge precisions (> 2^30).
* Possible bugs if the chosen exponent range does not allow to represent
some numbers such as 1 or 1/2.
* Possible infinite loop in some functions for particular cases: when
the exact result is an exactly representable number or the middle of
consecutive two such numbers.
* mpfr_set_d may give wrong results on some architectures.
* Error analysis for some functions may be incorrect (out-of-date due
to modifications in the code?).
* mpfr_hypot may fail for x very large, y very small and a very large
target precision. Other functions may be affected by similar problems.
Problems due to compiler bugs:
* on some architectures (for example alpha-dec-osf), gcc 3.3 wrongly
compares "long double" floating-point numbers, with optimization level
1 or higher. This bug can be detected by the following program:
#include <stdio.h>
#include <stdlib.h>
#include <float.h>
int
main ()
{
long double d;
d = 1.0; while (d < LDBL_MAX / 2.0) d += d;
if (d == (long double) 0.0)
printf ("d equals 0.0\n");
}
This results in a problem in the mpfr_set_ld function. A workaround is
to compile set_ld.c with -O0 (no optimization).
|
