View Single Post
Old 2018-11-01, 08:02   #5
robert44444uk
 
robert44444uk's Avatar
 
Jun 2003
Oxford, UK

36458 Posts
Default

Quote:
Originally Posted by ATH View Post
I can extend it to include higher primes than 101, but the problem is calculating CRT offset. It would exceed 128 bits so I would have to switch to GMP library to calculate it instead of the 128 bit variables I use now, and then you need to compile the GMP library first in order to compile the program

It should be no problem to compile GMP in MSYS2, I do it all the time, and you have all the needed packages installed, but I do not know if you feel it is worth it. The higher you go using only the best few solutions from the lower runs, the higher risk there is that you are no where near the real minimum.
It would be worthwhile to do the GMP solution.

I agree that we are nowhere near the real minimum at higher primorials, but then again, I think a few tweaks to the program would allow for pretty decent offsets. For example if programmed to take results that are within 5 of the record at a given level, instead of 3, and then advanced those two primes at a time, then you could quickly get good results. To do that, a tweak would be required to eliminate from the results file those first few solutions that come out that are nowhere near the minimum. You could do that by calculating the theoretical expected result given an input file and only outputting those which are within 5. What do you think?

Took one offset from 2000 interval forward last night and 4 records this morning between 2600 and 2750 :)

Last fiddled with by robert44444uk on 2018-11-01 at 08:04
robert44444uk is offline   Reply With Quote