| 1 |
Copyright 1999, 2001-2020 Free Software Foundation, Inc. |
| 2 |
Contributed by the AriC and Caramba projects, INRIA. |
| 3 |
|
| 4 |
This file is part of the GNU MPFR Library. |
| 5 |
|
| 6 |
The GNU MPFR Library is free software; you can redistribute it and/or modify |
| 7 |
it under the terms of the GNU Lesser General Public License as published by |
| 8 |
the Free Software Foundation; either version 3 of the License, or (at your |
| 9 |
option) any later version. |
| 10 |
|
| 11 |
The GNU MPFR Library is distributed in the hope that it will be useful, but |
| 12 |
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY |
| 13 |
or FITNESS FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public |
| 14 |
License for more details. |
| 15 |
|
| 16 |
You should have received a copy of the GNU Lesser General Public License |
| 17 |
along with the GNU MPFR Library; see the file COPYING.LESSER. If not, see |
| 18 |
https://www.gnu.org/licenses/ or write to the Free Software Foundation, Inc., |
| 19 |
51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA. |
| 20 |
|
| 21 |
############################################################################## |
| 22 |
|
| 23 |
Known bugs: |
| 24 |
|
| 25 |
* The overflow/underflow exceptions may be badly handled in some functions; |
| 26 |
specially when the intermediary internal results have exponent which |
| 27 |
exceeds the hardware limit (2^30 for a 32 bits CPU, and 2^62 for a 64 bits |
| 28 |
CPU) or the exact result is close to an overflow/underflow threshold. |
| 29 |
|
| 30 |
* Under Linux/x86 with the traditional FPU, some functions do not work |
| 31 |
if the FPU rounding precision has been changed to single (this is a |
| 32 |
bad practice and should be useless, but one never knows what other |
| 33 |
software will do). |
| 34 |
|
| 35 |
* Some functions do not use MPFR_SAVE_EXPO_* macros, thus do not behave |
| 36 |
correctly in a reduced exponent range. |
| 37 |
|
| 38 |
* Function hypot gives incorrect result when on the one hand the difference |
| 39 |
between parameters' exponents is near 2*MPFR_EMAX_MAX and on the other hand |
| 40 |
the output precision or the precision of the parameter with greatest |
| 41 |
absolute value is greater than 2*MPFR_EMAX_MAX-4. |
| 42 |
|
| 43 |
Potential bugs: |
| 44 |
|
| 45 |
* Possible incorrect results due to internal underflow, which can lead to |
| 46 |
a huge loss of accuracy while the error analysis doesn't take that into |
| 47 |
account. If the underflow occurs at the last function call (just before |
| 48 |
the MPFR_CAN_ROUND), the result should be correct (or MPFR gets into an |
| 49 |
infinite loop). TODO: check the code and the error analysis. |
| 50 |
|
| 51 |
* Possible bugs with huge precisions (> 2^30) and a 32-bit ABI, in particular |
| 52 |
undetected integer overflows. TODO: use the MPFR_ADD_PREC macro. |
| 53 |
|
| 54 |
* Possible bugs if the chosen exponent range does not allow to represent |
| 55 |
the range [1/16, 16]. |
| 56 |
|
| 57 |
* Possible infinite loop in some functions for particular cases: when |
| 58 |
the exact result is an exactly representable number or the middle of |
| 59 |
consecutive two such numbers. However for non-algebraic functions, it is |
| 60 |
believed that no such case exists, except the well-known cases like cos(0)=1, |
| 61 |
exp(0)=1, and so on, and the x^y function when y is an integer or y=1/2^k. |
| 62 |
|
| 63 |
* The mpfr_set_ld function may be quite slow if the long double type has an |
| 64 |
exponent of more than 15 bits. |
| 65 |
|
| 66 |
* mpfr_set_d may give wrong results on some non-IEEE architectures. |
| 67 |
|
| 68 |
* Error analysis for some functions may be incorrect (out-of-date due |
| 69 |
to modifications in the code?). |
| 70 |
|
| 71 |
* Possible use of non-portable feature (pre-C99) of the integer division |
| 72 |
with negative result. |
