features.txt

======================
Features of CB pi v2.3
======================

1) A limit of a massive  1  billion  (1024m) digits of pi.  No other
desktop program approaches that limit.  (This  _has_  been  computed
and verified!)

2) Fairly portable.  Although some versions use a bit  of  assembly,
you  can  run  it  without  it  (accepting a performance penalty, of
course.)

3) Under DJGPP, has  the  ability  to  abort  the iteration when you
press a key.  (This could be ported to other  systems  by  modifying
the port.c file.)  Using this ability (or having the program  do  it
automatically)  allows  you  to  compute at your convenience, rather
than it running continuosly.

4)  The  ability  to  do  large  pi  runs without any virtual memory
support, only depending upon pure physical memory and disk.

5) The ability to use an external  FFT that is already tuned for the
target  system.   This  can  result  in  a  significant  performance
improvement,  especially  if  that  FFT  takes advantage of multiple
processors.

6) Able to  run  with  limited  memory.   Theoretically,  you  could
compute a billion digits with only 4m of physical memory.  (Although
it much prefers at least Digits*2  bytes of memory.)  You'd be crazy
to try, but the program could do it.

7)  Can  use memory efficient 'NTT's that require as little as 1/4th
the memory as the number of digits desired.  Can work with even less
memory.  (More is better, of course.)

8) Includes multiple pi formulas, allowing you to use two to  verify
your  result,  rather  than  just assuming the numbers are accurate.
(Hardware errors _do_ occur!   As  do the occasional latent software
bug.  I've experienced both of those!)

9) An easy to configure 'pi.ini' file to set the desired options.

10) The ability to explicitly specify where  the  program  puts  the
needed files, including on  multiple  disks.   (Although  each  file
itself must fit entirely on the specified disk.)

11) Can work on systems with or without a command line.

12) Fast for both in memory 'benchmark'ing runs, and for larger disk
based runs.