[RC5] Re: PiHex

Colin Percival Colin_Percival at sfu.ca
Mon Jan 25 10:16:50 EST 1999


  I've been lurking here for a while, but now that PiHex has come up again,
it's probably time for me to provide some facts.

  PiHex is _not_ computing all the digits of Pi (up to a point, of course).
 PiHex uses the BBP algorithm and Bellard's formula to compute a small
bitstring a long way off.  Thus we compuuted the five trillionth bit (aka 5
billionth for the english) and a few around it, but not all the 500GB of
intervening bits.
  As a result, the memory/disk/communications overhead of PiHex is small,
basically the same as d.net.
  In fact, the differences between PiHex and d.net are few:  PiHex has to
get all the results back before we are done, while d.net only needs the
'winning' key;  PiHex (as with all computations of Pi) computes everything
twice (in different ways) to guard against errors;  Each 'block' creates a
small amount (~80 bytes) of data which must be kept for later use;  PiHex
is not yet automated, although I am working on it.
  Before anyone asks, no, I do not want to incorporate PiHex into d.net.
PiHex will be ending after we finish computing the quadillionth bit (+
about 40 bits around it) of Pi, at which point I will go on to do other
things, specifically a 'classical' computation of Pi (computing all the
digits up to a point).  Even if I were willing to give up PiHex's autonomy,
I would like to keep on doing everything by myself just because it forces
my to learn the tools necessary for my next project.
  As for whether a classical computation of Pi is possible over a
distributed system, wait and find out. ;-)

Colin Percival

On Sun, 24 Jan 1999 17:10:12 GMT, kai_rode at bigfoot.com wrote:
>Is PiHex suited for distributed computing?

--
To unsubscribe, send 'unsubscribe rc5' to majordomo at lists.distributed.net
rc5-digest subscribers replace rc5 with rc5-digest



More information about the rc5 mailing list