 Sun Invites IBM, Cray To Work On New Computer Language
started 12/19/2003; 1:02:08 PM - last post 12/20/2003; 9:48:37 AM
|
|
Chris Rathman - Sun Invites IBM, Cray To Work On New Computer Language 
12/19/2003; 1:02:08 PM (reads: 9541, responses: 6)
|
|
| Sun Invites IBM, Cray To Work On New Computer Language |
|
via Java Lobby
Whether IBM and Cray will agree to collaborate on the effort is unclear. Both companies have their own software plans that include developing new languages and operating systems as part of their competing work on the High Productivity Computing Systems (HPCS) project under the Defense Advanced Research Projects Agency (Darpa).
"Java has made it easy to program using a small number of threads. But in this [technical computing] world you have to handle thousands or hundreds of thousands of threads. We need the right language constructs to do that,"
Ok, it's obvious that Java threads are not scalable, but I don't know that Sun's Hero project is a language project so much as it is an attempt at a VM specification. The Darpa High Productivity Computing Systems concentrates on massively parallel architectures, so hopefully they'll tap into languages like Erlang and Oz for inspiration.
Posted to general by Chris Rathman on 12/19/03; 1:05:27 PM
|
|
|
|
Ehud Lamm - Re: Sun Invites IBM, Cray To Work On New Computer Language
12/19/2003; 1:08:17 PM (reads: 699, responses: 0)
|
|
|
Is this April 1st already?
|
|
Dan - Re: Sun Invites IBM, Cray To Work On New Computer Language
12/19/2003; 1:13:43 PM (reads: 696, responses: 0)
|
|
|
I suppose they could tap into languages like Erlang and Oz (or Occam) for, well, languages. I'm not sure it's necessary to go and reinvent the wheel again for this.
|
|
Chris Rathman - Re: Sun Invites IBM, Cray To Work On New Computer Language
12/19/2003; 1:19:01 PM (reads: 690, responses: 0)
|
|
Is this April 1st already? Didn't the last April Fool's joke about a VM result in Parrot (and Dan's still busy on that one).
I'm not sure it's necessary to go and reinvent the wheel again for this. Since money is no object in these sorts of things, reinvention is probably the first requirement. Though I did see that the DOD learned their lesson from Ada and will require compatibility (retrofitting) with Fortran for the new design.
|
|
Ehud Lamm - Re: Sun Invites IBM, Cray To Work On New Computer Language
12/19/2003; 3:40:45 PM (reads: 647, responses: 0)
|
|
|
The TechWeb article is an example of technical journalism at its worst. Does anyone have access to technical details about this thing?
The article mentions machine independent bytecodes (is this the right choice for HPC?) and the "expertise" Sun gained from Java. This is amusing, but doesn't really tell us what's the score. Are they thinking about "machine dsecription" kind of things (see Ramsey's work), optimizing VMs? Something else?
The TechWeb article ends by talking about threads, and COP. So yeah, looking at erlang may be a good idea. But is threading really the problem for scientific computing? What kinds of scientific computing are they looking into?
|
|
Chris Rathman - Re: Sun Invites IBM, Cray To Work On New Computer Language
12/19/2003; 4:07:33 PM (reads: 653, responses: 0)
|
|
|
Don't know much more than what the articles mention. I gather that the project aims to make the tools hardware independent - a goal usually not associated with such projects. The project and intended uses is described described on the Darpa HPCS website.
I gather that Sun wants to design the VM and influence the resulting language to be something not far from Java. IBM and Cray are more concentrated on the hardware side and seem to be leaning toward messaging schemes to get the parallelism. Programming language design is not the primary mission, but it sounds like they are exploring programming languages in the process.
|
|
Peter Van Roy - Re: Sun Invites IBM, Cray To Work On New Computer Language
12/20/2003; 9:48:37 AM (reads: 543, responses: 0)
|
|
|
The ICITIES project did large-scale simulations of agent communities
to study the structure and evolution of the WWW. They used Mozart
on a cluster machine for these simulations. There is a talk by Seif Haridi,
Parallel
Agent Based Simulation on PC Cluster,
that gives an overview of the problem, the simulation architecture, and
the results.
They measured significant performance improvement because
of fine-grained concurrency and dataflow synchronization.
They also noticed the importance of first-class components
and component-oriented programming when doing many different kinds of
simulations in a simulation framework.
The use of Oz worked out quite well because of its cheap threads, dataflow,
and because components ('functors') are values in the language. SICS did
some performance optimization of Mozart as part of this project (e.g.,
improving thread scheduling and dictionary performance).
|
|
|
|