Re: Actual F.P. implementation?

Vladimir Makarov <vmakarov@cygnus.com>
25 Nov 1999 01:49:44 -0500

          From comp.compilers

Related articles
Actual F.P. implementation? dzels901@my-deja.com (1999-11-23)
Re: Actual F.P. implementation? vmakarov@cygnus.com (Vladimir Makarov) (1999-11-25)
Re: Actual F.P. implementation? jkahrs@castor.atlas.de (Juergen Kahrs) (1999-11-25)
| List of all articles for this month |

From: Vladimir Makarov <vmakarov@cygnus.com>
Newsgroups: comp.compilers
Date: 25 Nov 1999 01:49:44 -0500
Organization: Cygnus Solutions
References: 99-11-143
Keywords: arithmetic

dzels901@my-deja.com wrote:
> Looking for sources/info. on how to implement floating point
> mul/div (also add/sub) for IEEE 754 32-bit _or_ 64-bit.


You could look at file fp-bit.c (in gcc/config) or floatlib.c (in dir.
gcc) in gcc sources http://egcs.cygnus.com/.


You can find another code in COCOM toolset (package IEEE in
Ammunition) which is placed on
http://www.linuxstart.com/~vladimir_makarov/download.html


The first implementations are simpler and faster.


The second implementation is more accurate (rounding, NaN, quite NaN,
status bits, sticky status bits, exceptions, trap masks). This
implementation can be easily used to implement 128-bit floating point
numbers because it is based on another package arithm (multiple
precision arithmetic).


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.