Saturday, July 11, 2020

Rev it up

Under the direction of Mark "rogue" Rodenkirch in mersenneforum, on June 28 I was able to run a probable-prime search/verification program that executes some six times faster than what I can do in Mathematica. Considering that Mathematica is what I have been using for my large-Leyland-prime search for almost five years now, that's an awful lot of wasted time!

In order to more efficiently use this program (OpenPFGW; OS X version, pfgw64), a prime-sieving algorithm (OS X version, xyyxsieve) was recommended as a precursor, in order to trim the available candidate numbers to a much-lesser amount. In Mathematica such a routine is incorporated in its PrimeQ function but PrimeQ will not go beyond its built-in limit. So, on July 2 I ran my first prime sieve.

This is a game changer for me. On Wednesday afternoon, a storm knocked out the power here for longer than I had computer battery-backup.


As a result, my Leyland prime search for interval #9 was interrupted. It was scheduled for completion July 25 but recent experience suggested that it would not actually have finished until a week into August. So I used the heavenly portent to convert my remaining search-space into a format that I could use for xyyxsieve and, subsequently, pfgw64. A lot of ongoing manual fiddling and such was needed for the conversions but this allowed me to get used to the new processes. This morning it completed the searches!

No comments:

Post a Comment