Janko,
> How are we in Smalltalk on that field? Anyone doing some research as well?
Yes, I am working on this. Back in 1992 I built a machine with 64
nodes (using the already obsolete 68020 plus a T222 transputer for
communication) to research parallel Smalltalk. What I am developing
now is more like the J-Machine from MIT:
http://cva.stanford.edu/projects/j-machine/Note the interesting results they got with their ConcurrentSmalltalk
(which had a Lisp syntax, as you would expect from an MIT project).
They developed active messages, wormhole routing, concurrent
aggregates and lots of other good things.
Another system, also called ConcurrentSmalltalk and developed at the
same time, took a different route, but their results were equally
interesting:
http://portal.acm.org/citation.cfm?id=38765.38844http://portal.acm.org/citation.cfm?id=28730http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.35.1907The article you mentioned says that message passing is being hidden as
a low level implementation detail, but also that Smalltalk has message
passing instead of subroutine calls. If you are trying to program in
C, dealing explicitly with MPI (the Message Passing Interface library)
is so painful that it is good to have some tool to do it for you. But
we Smalltalkers like to write our messages, unless it is an extra
layer with an awkward syntax like in Croquet. So I feel we have an
advantage over everyone else.
-- Jecel
Locked
