The original post was on talkchess.com but it is gone now. The URL was:
http://216.25.93.108/forum/viewtopic.php?p=63058&sid=ae6cbcbf2a6bd9ac12744430604f1d9a
-------------------------------------------------------

Author: Paul Byrne
Post subject: Perft 12?
Posted: Tue Oct 24, 2006 5:03 am

Disclaimer: I've been rather erratic visiting CCC since the (move/new
forum/emoticon virus attack/whatever happened), so I haven't really
been keeping up with the fast paced world of perft computations, but here
is some information which may or may not be new... :)

After lying dormant for some time, I pulled out my old perft code (which
had the nerve to disagree with the distributed project's perft 11 result
some years back). Reran the computation and got the same conflicting
answer. Spent all of 5 minutes finding a rare bug in how en passant is
handled. Ran it again, and it now agrees with the project's stated result
of 2097651003696806.

For those curious, the computation took slightly over 20 hours. On a
16 x 1.5 GHz Itanium system. The code is in no way written for an
itanium (it isn't even 64 bit), but 16 iffy cpus beat 1 good one. :)

This was a few months ago. The result for perft 12 just finished...
drumroll? 62854969236701747 (who wins the pool?) This is a hair under
30x the perft 11 result, which seems about right. The spiffy itanium was
not available for this, so it has been done in bits and pieces on a dozen or
so computers, using a day here and a week there. Altogether about 300
cpu-days of computation over 2 or 3 months.

Is this result correct? I await verification. :) To the best of my
knowledge the code is correct. No hash tables were harmed in the
production of this number, so there is no chance of a collision. However,
as stated, a number of computers were used, so I suppose the odds of
one having a hardware problem are greater than if only one were used. I
do know which computers did which pieces, so, in theory, I could redo it
all on other machines. Or not. :)

As to perft 13... well, that will require waiting some years for faster
cheaper machines, or a significant distributed project, or perhaps a
moment of inspiration from someone out there...

-paul