K312240@AEARN.BITNET (Klaus Kusche) (08/07/89)
Dear Mailing List (and especially dear Dr. J. Salsman): James Salsman wrote about "memory paged out to disk" in one of his recent mails here. I agree that it is easy to write Occam (or C, ...) code to swap memory to disk (and of course also to neighbouring nodes' memory), but all this is of no practical use as long as you don't have *demand- paged* memory, i.e. a possibility to catch and handle references to swapped-out memory with no performance penalty if the referenced memory is already there. For example, most modern Lisp garbage collectors are based on the fact that virtual memory is cheap and it doesn't matter if there is a lot of it allocated but unused most of the time --- doesn't agree well with transputers! Hence, very few people on this list will be interested in the rather trivial possibility of moving memory blocks around, but if you (or anybody else) can show us how to implement demand-paged memory on T's (or something close to it), you will surely receive the transputer mailing list award of the year!!! ************************************************************************ * Klaus Kusche * * Research Institute for Symbolic Computation * * Johannes Kepler University Tel: +43 7236 3231 67 * * A-4040 Linz Telex: (Austria) 22323 uni li a * * Austria (Europe) Fax: +43 7236 3231 30 * * * * Bitnet: K312240@AEARN * * Arpa/CS/Internet: K312240%AEARN.BITNET@CUNYVM.CUNY.EDU * * UUCP: mcvax!aearn.bitnet!K312240 * * Janet: K312240@earn.aearn (perhaps) * ************************************************************************