David Hanley <[EMAIL PROTECTED]> writes: > Nicolas; > > IMO, fork is perfectly acceptable for the sort of > activity that you are describing. And, in fact, for > any computation that might take hours, i'd probably > prefer to write the results to a file, unless they are > *very* large.
That's probably right and maybe I will do that when I go for long-time calculations. At the moment, I am still developing, such that my computations are in the seconds and minutes range. And I do not yet have implemented an input/output possibility for the rather complicated data structures I have (finite elements on unstructured meshes in arbitrary space dimensions). > As an aside, on some systems, for some tasks, fork will be faster > than starting a thread. > > But this is lisp. I think our mentality should be to > get the job done in an elegant way that solves our > problems sufficiently fast, then move on. You're right. But this point is not of uttermost importance for me. It would have been nice if an easy possibility would have existed, though. Nicolas. > If you're not causing garbage collection, the heap > should be largely shared between the subprocesses. > > dave P.S.: I am just looking at SCL: what a pity that this work has not been done on CMUCL. (I would not mind paying for SCL, but I fear that it will lag behind CMUCL in other regions, e.g. Gerd Moellmanns pcl improvements.)
