French software whiz claims record
A French software engineer said on Friday he was claiming a world record for calculating Pi, the constant that has fascinated mathematicians for millennia.
Paris - A French software engineer said on Friday he was claiming a world record for calculating Pi, the constant that has fascinated mathematicians for millennia.
Fabrice Bellard told AFP he used an inexpensive desktop computer -- and not a supercomputer used in past records -- to calculate Pi to nearly 2.7 trillion decimal places.
That is around 123 billion digits more than the previous record set last August by Japanese professor Daisuke Takahashi, he said.
Takahashi, using a T2K Open Supercomputer, took 29 hours to crunch Pi to 2.577 billion digits.
Bellard took 131 days, comprising 103 for the computation in binary digits, 13 days for verification, 12 days to convert the binary digits to a base of 10 and three final days to check the conversion.
The gear cost "a bit less than EUR 2,000" (USD 3,000), Bellard, who earns a living as a software consultant in digital television in Paris, said in an email exchange.
"It is a completely standard PC. The only unusual thing is that it has five 1.5-teraoctet hard disks. Mainstream PCs generally have only one 1-teraoctet disk."
Bellard has placed on his website details of the achievement, including the use of a high-powered mathematical engine called the Chudnovsky algorithm that chewed through the computation.
Extracts of the 2,699,999,990,000-digit outcome have been published so that they can be compared to preceding records in order to gain independent verification, Bellard told AFP.
Files containing the digits are also being offered to any outside organism keen on hosting the record, he said.
Pi, the ratio of a circle's circumference to its diameter, kicks off with 3.14159... in a string whose digits are believed never to repeat or end.
Bellard said he was "not especially interested" in Pi's digits but more in taking up the gauntlet of writing the software to carry out the arithmetic.
"Optimising these algorithms to get good performance is a difficult programming challenge," he wrote.