-
Notifications
You must be signed in to change notification settings - Fork 118
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
SAT parser improvements #1452
Comments
There are 27 places, some of which are due to different memory management in the SAT and classic parser that could be unified, or because there was no point to take care to allow it to be used with generation. In general, I think that are sparse and don't add much code complexity, and mostly serve as documentation of which code supports the SAT solver.
There are several reasons for its slowness:
To sum up, in my opinion, the speed of the SAT parser may be improved to be faster than the classic parser. It can even be changed to work with nulls. In addition, constraints may be removed to allow it to make cross-links (possibly only in certain cases according to a definition file). |
Yet another thing on the SAT parser: It naturally can lead to parallel processing per sentence if used with a parallel (multi-thread) SAT library. On 40 cores 10x speedup may be possible, depending on the SAT clauses, but of course, I don't even have a guess regarding the LG SAT parser. Here is the answer of ChatGPT when I asked it about such open-source libraries: Several open-source SAT solvers support multi-threading for speedup on modern multi-core CPUs. Here are some of the most popular ones:
These SAT solvers are all highly optimized for performance and can achieve significant speedup when running on multiple cores. However, the exact speedup will depend on the specific instance being solved, the hardware configuration being used, and the specific settings and parameters chosen for the solver. |
OK, Well then. I guess it can stay... |
I'm renaming this issue to fit the comments |
The SAT parser is slower in all situations. Supporting it no longer seems tenable.
grep -r USE_SAT_SOLVER * |wc
indicates there are over a dozen places where ifdefs could be removed.Is there any reason to keep it around?
The text was updated successfully, but these errors were encountered: