The Freecell Solver project, which develops a free and open-source software framework for solving layouts of Freecell (also known as “FreeCell”), and several similar Solitaire variants is seeking assistance in the form of a provision of direct or indirect access to high performance computing (HPC) hardware, namely computers with large amounts of accessible RAM (128 GB or so or more), large amounts of hard disk space (about 10 GB - less than RAM - should be enough), and good reliability.
If you can provide us with such resources to further our progress in researching Patience card games, please contact Shlomi Fish.
Freecell has been an active source of research and poses interesting problems in the domain of artificial intelligence. The Freecell Frequently Asked Questions (FAQ) list contains a lot of information about Freecell from the results of Freecell researchers.
Generalised Freecell (where the number of cards in a sequence in a deal is variable beyond Ace-to-King) has been shown to be NP-Complete ( "Malte Helmert, Complexity results for standard benchmark domains in planning, Artificial Intelligence Journal 143(2):219-262, 2003." ).
In order to run the solver on several hard deals (which for limited memory are intractable), and see if there is a solution or not, we need a large amount of memory. We have done improvements in the memory consumption of the solver, but the problem domain still necessitates a large amount. As a result, access to computers with a large amount of RAM will be beneficial for our cause.
We have been kindly provided with access to such computers, but the have proven to be too unreliable for our needs, and caused the invoked processes to stop prematurely, and as a result, we are seeking some more reliable machines.
By proving whether currently intractable Freecell deals are solvable or impossible, we can reach better conclusions about the statistics of solvable or impossible deals. Furthermore, some of the statistics collected have been generated by non-open-source or in-house computerised solvers and as such are less dependable than our open-source solver.