[733] in Software_Announce
apfloat 2.40 installed in outland for linux
daemon@ATHENA.MIT.EDU (Ken T Takusagawa)
Thu Jun 17 12:54:17 2004
Date: Thu, 17 Jun 2004 12:53:25 -0400 (EDT)
From: Ken T Takusagawa <kenta@MIT.EDU>
To: software-announce@MIT.EDU
cc: bug-outland@MIT.EDU
apfloat 2.40 has been installed for i386_rh9 and i386_rhel3
in the outland locker.
http://www.apfloat.org/
apfloat is a C++ library for arbitrary precision arithmetic.
It uses FFT-style transforms to perform multiplications of
N-digit numbers in less than quadratic time, probably O(N
log N). (In contrast, the programs bc, dc, and gp all use
slow quadratic-time multiplication.)
The include files are installed in
/mit/outland/include/apfloat . The library itself is
/mit/outland/lib/apfloat.a . You will also need to link
with -lpthread.
A demo program, "pi", has been installed at
/mit/outland/bin/pi . This program calculates pi to an
arbitrary number of digits in a arbitary base (2-16) using
one of 3 different algorithms. Ten million digits is easily
achievable in a few minutes ("pi 10000000").
The source to the demo program, as well as other useful
source codes and demo programs, are available at
/mit/outland/arch/i386_rh9/build/apfloat .
I have applied a home-grown patch for "rawprint" output,
which allows output in bases greater than 16. The patch is
at /mit/outland/src/apfloat/patch .
--ken
