Impressive results Seth!
I have been looking at different approaches myself and have been testing those at small integer sizes (<10^20) to see how often they deliver maximum prime gaps (a test of the m efficiency if you will). I have to say that my findings are inconclusive.
Skipping m values will result in missing max prime gaps. So I am sticking to much slower search routines that include all m (m = all odd multipliers in center = m * p# / d).
If you do want to limit the numbers of m to test, my observation is that m = not squarefree gives more maximum prime gaps than searching for m = prime or m = (squarefree and not prime).
If you do want to increase merit, I would suggest the following:
I do find that adding an extra q to the center value (center = q * m * p# / d) delivers larger merits. Note: m and q are both prime.
Just my 2 cents.
Kind regards
Michiel
