Just wanted to elaborate a little on Erlang since it made an appearance and it is my favorite language of the moment.
1. Concurrency is certainly an Erlang feature. However, it is not what makes it unique. Scala does this too using Actors. And Scala even runs on the JVM as a first-class language. Haskell also gives you superb concurrent programming capability. In fact, in Haskell it is absurdly simple to spread over cores using the Control.Concurrent library. 2. 'Shared Nothing' architecture is not a factor of concurrency but really what allows Erlang to scale exceedingly well. But even this is not what makes Erlang unique. What makes Erlang unique is that resiliency is a first-class citizen. By resiliency I mean its support for Process Monitoring (If anything collapses, simply kill it, make a new one, figure out where it left off from the universal stack and go! Similarly, if you notice something fishy going on with your code - hot swap it with an update. Again the Erlang stack assists. Joe Armstrong's philosophy was simple - large systems screw up. When they do, give the engineers the maximum possible chance of saving the situation. That makes Erlang unparalled. Cheers, Zubin. On Tue, Jul 15, 2008 at 7:49 PM, Saifi Khan <[EMAIL PROTECTED]> wrote: > On Sat, 21 Jun 2008, Namita Iyer wrote: > > > Date: Sat, 21 Jun 2008 17:12:23 +0530 > > From: Namita Iyer <[EMAIL PROTECTED] <namita%40twincling.org>> > > Reply-To: [email protected] <twincling%40yahoogroups.com> > > To: [email protected] <twincling%40yahoogroups.com> > > Subject: [twincling] memory barriers and locking primitives > > > > Has anyone used memory barriers with multicore processors ? > > (embedded systems programmers?) > > > > In what scenarios do they work better than locking primitives ? > > > > Namita > > > > Hi Namita: > > i did read couple of papers of Herb Sutter > and off course programming in Erlang :-) > > One of the approach that seemed to have worked very well > for accomplishing concurrency is a 'shared nothing' model > instead of 'shared memory' model esp. in distributed computing. > > Erlang follows this model, also known as 'Actor' model. > http://en.wikipedia.org/wiki/Actor_model > > Can you share your thoughts about the concept of Memory barriers > and how a C/C++ code might be structured in absence of > Critical sections and mutexes ? > > thanks > Saifi. > > [Non-text portions of this message have been removed]
