search

pazusoba / core

Try to solve Puzzle & Dragons with C++
MIT License
15 stars 1 forks source link

New direction #49

Closed HenryQuan closed 2 years ago

HenryQuan commented 2 years ago

When I first started working pazusoba, I was only thinking about fancy cascading. It was really fancy, but max combo wasn't achievable most of the time because the algorithm was extremely greedy. I will try a different approach from now on.

HenryQuan commented 2 years ago

The new experiment is great, but it outputs very poor results. It is undergoing some testings at the moment. The overall performance is much better because of the lock free design.

At the same time, there is also a new approach. Solve the board using static analysis and search for the route afterward. This requires more steps, but it should work for everything. This can be a good direction as well.

HenryQuan commented 2 years ago

Max combo can be found consistently now. However, OpenMP is producing worse result now somehow. The result is not stable with OpenMP. This needs to be fixed. The next thing is profiles. This will allow the solver to solve for any leaders.

HenryQuan commented 2 years ago

The new experiment is super successful so far. Thanks to kodumaさん.