Sept 10: 1998. And this is what the artist says about his opus:
I used the same algorithm and formula as Bellard, except that I wrote my
program in highly optimized assembly language, which made it about 10 times
faster than Bellard's program, and I publicized PiHex on the web, and asked
for donations of idle computer time, so I now have (at last count) 342
computers running PiHex, in contrast to the dozen or so that Bellard used.
Also, all the computers running PiHex are windows-based PCs, while
Bellard used primarily unix boxes.
I could write something about it, but I think that it would be of more
interest to assembly-language programmers than anyone else.
Colin Percival
============================================================================
He did it!
Colin Percival, 17, now fully enrolled SFU student, finished his computations
to find the 5-trillionth binary digit of Pi and thus established a world record
for finding the largest known digit of Pi. It was the first time in history
that distributed computation was used on an attempt to break the world record;
previously, supercomputers had been used for this task. The algorithm for
computing the nth binary digit of Pi can be nicely parallelized; Colin used
a network of 25 computers, distributed all over the globe. The total running
time was 4 months. (On a serial machine of comparable speed, it would have
taken 17 months.)
Colin is now planning to compute the 40-trillionth digit; on the same
computing network, using the same algorithm, this task will take 4 years.
The whole story was published in the Burnaby Newsleader last Sunday 6th;
I'll get up a copy and put it in the lab shortly. Also, Colin might contribute
a short note on his computations on our Web site.
Jens Happe
===========================================================================
From Colin_Percival@sfu.ca Thu Sep 10 14:51:14 1998
1. The actual digits:
Starting at the 1.25 trillionth hexit, Pi is 07E45733CC790B5B5979. The
fourth bit of this (a zero) is the five trillionth bit.
2. The dates the computation began and ended.
March 21 - August 21.
3. A check was done (as in Bellard's case).
Of course -- I don't trust myself enough not to check it. The
calculations were done twice, 4 bits apart (like in the BBP paper), and
compared. They matched perfectly apart from rounding errors.
Also, before any computer was allowed to do any work, it had to pass a
self-test, which consisted of calculating the one hundred millionth hexit
of Pi.
4. The number of machines used.
25 computers (out of about 200 participating in PiHex) worked on the five
trillionth bit. These ranged from a Pentium 60 to a dual Pentium II 333.
All this information, plus the unanalyzed data (ie, partial results) are
available from the PiHex web site at http://www.cecm.sfu.ca/projects/pihex/.
Colin Percival