Fix some problems that arise when modeling megaminx. #5
Commits on Aug 16, 2016
-
-
-
Fix two issues in indexing. First, a bunch of ints should be long longs
(when orientation counts might exceed 2^31 options). Also, the unpackVector was assuming all values are >0, but that's not the case for orientations, so added an offset of 1 on pack and subtracted that offset on unpack.
Commits on Aug 17, 2016
-
Allow ! to generate God's number information, as the docs say.
This was apparently broken when someone "wrapped" ksolve.
Commits on Jul 9, 2018
-
Move the ksolve shell script to have a .sh extension so we can
create executables called ksolve. No need to use wine as this program compiles trivially on Unix.
-
Commits on Jul 27, 2018
-
Slight bump in max memory we try to allocate (from 2G to 16G).
This needs to be configurable at runtime.
-
-
Redid the large puzzle packing code completely. It's now much faster,
often more compact, but most importantly works correctly for puzzles that have permutations with more than 20 elements.
-
-
Commits on Jul 28, 2018
Commits on Jul 29, 2018
-
Don't leak memory so crazily on God's number explorations. Also
gives another small speed bump.
-
Removed string indexes from main datastructures and replaced them with
ints. If this works the next step is to change them from maps to vectors where possible.
-
-
Big change; changing Position from a map to a vector. Almost certainly
broke some functionality.
-
Commits on Aug 3, 2018
-
-
Added a max memory option so we can run big things on big machines and
small things on small machines.
-
Made God's number calculations use permutation parity optimization.
Also, substantially sped up index calculations.
-
-
-