The 4_4_27 issue is not sleeping. (is there any new 17 with that pattern)
I have now a stable code based mainly on
_ some code delivered by blue through pm filtering all "isomorphs" (see earlier in that thread)
_ a brute force based on the code posted here
by user zhouyundong_2012.
both are merged in a single program processing one piece of the field.
The entry is the file containing the ED patterns to process
3 parameters define
the start pattern number and the last pattern to process
the gangster to apply
The program deliver "primary puzzles" that must still be expanded as explained earlier in that thread
To-day my estimate is that
more than 10% of the task has been covered
with my own resources, the task should be finished in January 2014.
(but my deadline is mid January 2015, then end of march 2015 at best)
Anybody having cycles to share the burden is welcome
Another issue is not yet fully covered.
Starting from the "primary puzzles" we have to
expand each primary to all puzzles equivalences for the gangster used
find if 17 puzzles exist for that 4_4_27 puzzle.
For the second issue, we recently used an option of gridchecker doing much more than required, but too slow.
I am investigating to get a better option with so far 2 possibilities:
a) another process in gridchecker that must be reshaped for a friendly use
the target would be about 100 puzzles processed by second
b) a program that blue can share processing several puzzles by second.
PS: I think blue already wrote that, but the process is very sensitive
1) first and mainly to the gangster applied
2) to the pair gangster/pattern
the run time can vary from 1 to 5000 and may be more between the best and the worst pair
PS2: The way the brute force is written, it has been very easy to create a reduced code processing only bands 2 and 3 after the initial settlement of the gangster, this gives on the paper a 20% (minimum) improvement