Re: RamFAST & GS/OS cache

[toddpw@cco.caltech.edu (Todd P. Whitesel)]

behrenss@hphalle6.informatik.tu-muenchen.de (Soenke Behrens) writes:

> I think the RF should support GS/OS caching.

Until the FST's can use cached blocks directly and without copying them, the
RF is better off doing it this way. The only place this runs into trouble is
when large files kick the directory blocks out of the RF cache (and I have
observed this on occasion) -- and that could be fixed pretty easily if Drew's
driver told the RF controller what had cache priority and what didn't.

Drew is typically pretty lazy about the GS/OS driver (it's a teeny thing, and
looks like he wants the controller to do all the work) but if he doesn't
enhance it soon I'll probably start bugging him about it too.

Todd Whitesel
toddpw @ cco.caltech.edu