Fix some problems that arise when modeling megaminx. #5

rokicki · 2016-08-16T02:33:53Z

Fixed four separate issues that arise when trying to model the megaminx.
Corners now work. Edges will need further fixes (support for permutations
with more than 20 pieces).

(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.

This was apparently broken when someone "wrapped" ksolve.

create executables called ksolve. No need to use wine as this program compiles trivially on Unix.

This needs to be configurable at runtime.

often more compact, but most importantly works correctly for puzzles that have permutations with more than 20 elements.

gives another small speed bump.

ints. If this works the next step is to change them from maps to vectors where possible.

broke some functionality.

small things on small machines.

Also, substantially sped up index calculations.

rokicki added 26 commits August 15, 2016 19:25

Fix delete to be delete[] when we new[].

211a05d

Fix iter++ that should obviously be iter2++.

b909dda

Allow ! to generate God's number information, as the docs say.

07b2bbf

This was apparently broken when someone "wrapped" ksolve.

Move the ksolve shell script to have a .sh extension so we can

d197eb6

create executables called ksolve. No need to use wine as this program compiles trivially on Unix.

Added missing flushes.

d639fa7

Slight bump in max memory we try to allocate (from 2G to 16G).

00dc9ad

This needs to be configurable at runtime.

Added some additional information to the puzzle def.

260e58f

Redid the large puzzle packing code completely. It's now much faster,

ca7a17d

often more compact, but most importantly works correctly for puzzles that have permutations with more than 20 elements.

Added -g to build command.

f87da8c

Now about 2X faster when puzzle fits in memory.

0d23dcb

Slight speed bump for large puzzles with centers.

e2c1aea

Don't leak memory so crazily on God's number explorations. Also

74b2d99

gives another small speed bump.

Removed string indexes from main datastructures and replaced them with

b3e49fe

ints. If this works the next step is to change them from maps to vectors where possible.

Eliminate some unneeded arguments.

19ee5f7

Big change; changing Position from a map to a vector. Almost certainly

1e7a465

broke some functionality.

Fixed some bugs introduced with that big rewrite.

2fb72c5

Fixed some bugs from my big changes.

ee2460d

A bug fix that only shows up on linux at the moment.

7e23d01

Fix another problem with the change from map to vector.

6531090

Added a max memory option so we can run big things on big machines and

e838632

small things on small machines.

Added verbosity option. Added computation of permutation parity.

b73343a

Made God's number calculations use permutation parity optimization.

68af894

Also, substantially sped up index calculations.

Oops; fixed a bug in table size calc.

3388a4a

Another fix?

a3368ea

Allowed counts > 2^31.

86251dc

lgarron changed the base branch from master to main June 16, 2020 01:39

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix some problems that arise when modeling megaminx. #5

Fix some problems that arise when modeling megaminx. #5

rokicki commented Aug 16, 2016

Fix some problems that arise when modeling megaminx. #5

Are you sure you want to change the base?

Fix some problems that arise when modeling megaminx. #5

Conversation

rokicki commented Aug 16, 2016