Quote:
Originally Posted by Kosmaj
My first tests indicate that ver.4 is about 30% slower than ver.3. Please tell me, can I use ver.3 to test n=103 for k<2^60 (by removing the n>=100 condition from the source).

In some cases it is possible because it is using sieve_limit=1000. Yes you can change ver.3, by removing that condition.
Note that in all of the new programs now you can change sieve_limit in the source and to do this you have to change also number_of_primes, because there is an error checking for this in the program, you can also change block_length and magic_constant, and of course magic_constant>=block_length>=sieve_limit has to be true . But note that in hexadeca program you can change only sieve_limit ( and number_of_primes ), don't change block_length otherwise the program won't search for the whole interval for n=76 ( there is no magic_constant for this hexadeca ).
I've changed the source in dodeca_4_0 to use sieve_limit=2000 and number_of_primes=303 but this was still slower by 7% for n=99. I don't know why perhaps because we are using in that part mpz_t integers and not long variables.