Quote:
Originally Posted by Greenbank
n=89 weight=452294
n=89 sticks out somewhat! And I've already taken n=88 :)

See my benchmarks for n=89, comparing dodeca_2_5 and 3_0 ( just for [0T,1T] ):
dodeca_2_5:
Number of Prp tests=1225
Time=46 sec.
dodeca_3_0
Number of Prp tests=45391
Time=23 sec.
So it isn't very important the number of prp tests, see my mpz_powm() instruction for n=89 bits modulus takes about say 50000 clock cycles, so if we are checking 45391 prp tests, then that take about 2*10^9 clock cycles or about 1.3 sec. for my pc ( 1.7 GHz ), so the prp testing time is about 5% of the total time, 95% is the sieving time!
That was the reason why I've decreased all parameters in 2_5 version of dodeca program.
I've also defined the weight of n ( for dodeca numbers ). See post #28 in this thread.