Thomas Joad wrote:
Gecode does all its allocation through malloc, but doesn't reimplement it. It just "bundles" everything at a single place so that memory management can be reimplemented easily. So this would actually be an option, to replace the Gecode allocation by a scheme that is GC aware.Do you foresee that happening in the near future? Since you work on both projects, you obviously have a better sense of what the undertaking would involve.
We have no current plans to implement that. The changes to Gecode wouldn't be so big, I think, but we have no infrastructure in Alice for cooperation with the GC.
I like Alice very much -- I don't want to give up its static typing for the dynamism of Mozart, for example, or to leave SML to dip to C+ + level to directly call the Gecode API. But I was a bit worried that even constant garbage collection had no effect on memory usage.
It seems that this is a bug, although I cannot currently see what exactly goes wrong here. Clearly, some memory is kept alive, as triggering the GC should really be enough to keep memory usage low. The original problem should be completely solvable using Space.discard, and it shouldn't be necessary at all to call the GC manually.
Cheers,
Guido
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ alice-users mailing list [email protected] http://www.ps.uni-sb.de/mailman/listinfo/alice-users
