Tuesday, September 26, 2006
  Multicore programming: the holy grail 

This semester, due to an issue with my funding situation at college, I found work with a systems group. So officially I am off working on PL theory for a while. The thing is, after about a year of doing PL, I am convinced that as programmers or language designers one should focus on building good abstractions that compose well. All other concerns are secondary to this goal.

 

The project I am involved with one that is concerned with creating a mechanism for programming the increasing for viable multicore machines that are available in the market. The objective being with coming up with a programming paradigm that makes it “easy” to program machines that have a large number of computing units (cores, processors etc).

 

The best description of the motivation for this is described in Herb Sutter's famous article -

The Free Lunch is Over: A Fundamental Turn Towards Concurrency in Software

http://www.gotw.ca/publications/concurrency-ddj.htm

Tuesday, September 26, 2006 10:08:45 AM (Eastern Standard Time, UTC-05:00)  #    Comments [4]  | 
Thursday, September 28, 2006 4:04:44 PM (Eastern Standard Time, UTC-05:00)
"building good abstractions that compose welll. All other concerns are secondary to this goal."
Sounds positively comunist partyish. What happened to the old Roshan!!!

Will you be working on some new language or extending some existing language?
Sidharth Kuruvila
Friday, September 29, 2006 12:04:41 PM (Eastern Standard Time, UTC-05:00)
Dont know yet. I dont know about your comment about "communist partyish" - that is like saying that all people who sound forceful about something sound the same.

My point with what I was saying was that good programs or good languages are built only when the designer thinks carefully about the right components in the system and how the compose. I liken this descriptions of "design the interfaces" (in the C* languages), "design in the patterns" (in XP/agile speak), "design the combinators" (in haskell (and to some extend other functional languages))... Some approximations to this idea manifest themselves as 'x-tier architecture" or "UML design foobar" and so on.

IMO, when designing a programming paradigm, this is the most important thing. The other things that you do are comparable to "painting the bike shed".

Roshan
Thursday, October 19, 2006 7:47:21 AM (Eastern Standard Time, UTC-05:00)
Have you heard of PLINQ?

http://www.eweek.com/article2/0,1895,2009167,00.asp

I guess this would release in the near future.
Pandurang
Monday, November 06, 2006 5:34:45 AM (Eastern Standard Time, UTC-05:00)
A toolset that provides a solution to multicore programming difficulties can be found at

www.connectivelogic.co.uk

A 'technology demonstrator' of the IDE and C++ source code generator will be available end Jan 2007
Stuart Smith
Name
E-mail
Home page

Comment (Some html is allowed: a@href@title, strike) where the @ means "attribute." For example, you can use <a href="" title=""> or <blockquote cite="Scott">.  

Enter the code shown (prevents robots):
Live Comment Preview