hi
does gambit-c have any strategy when one day those 80 cores cpus are available? maybe I am thinking about this too earlier.
just find this page on register: http://www.theregister.co.uk/2007/08/20/tilera_tile64_chip/
Afficher les réponses par date
On Tue, Aug 21, 2007 at 06:34:39AM +0800, naruto canada wrote :
hi
Hi
does gambit-c have any strategy when one day those 80 cores cpus are available? maybe I am thinking about this too earlier.
One of those strategies simply consists in running 80 instances of Gambit/Termite, hoping that each one will run on its own core. But this doesn't make programming easier, as it then depends on *your* abitility to write parallel code.
But I'm sure that the day some of us use an 80 core CPU daily, gambit will somehow handle that. Or at least an (expt 2 n) core. May this not make you write naive exponential algorithms by laziness :)
P!
80 instances of *termite* ?
You would have one instance of Termite and express your code as Termite nodes (which, as you say, still requires an ability to write parallel code, but Termite does help). I don't know how Termite supports multiple processors, but it does, whether it creates the 80 Gambit instances for you or directly supports SMP. I played around with Termite a long time ago and it didn't seem to create multiple Gambit instances, but I forget how I tested that. How does Termite implement multiple processor support?
So go ahead and write Erlang-style code with Termite. The style of code is very unlikely to change while the implementation of it probably will (for good reasons), but that will be transparent to us (hopefully).
James
On 8/21/07, Adrien Pierard pierarda@iro.umontreal.ca wrote:
On Tue, Aug 21, 2007 at 06:34:39AM +0800, naruto canada wrote :
hi
Hi
does gambit-c have any strategy when one day those 80 cores cpus are available? maybe I am thinking about this too earlier.
One of those strategies simply consists in running 80 instances of Gambit/Termite, hoping that each one will run on its own core. But this doesn't make programming easier, as it then depends on *your* abitility to write parallel code.
But I'm sure that the day some of us use an 80 core CPU daily, gambit will somehow handle that. Or at least an (expt 2 n) core. May this not make you write naive exponential algorithms by laziness :)
P!
-- "I am not a Church numeral; I am a free variable!" (The Scheme Underground) For every complex problem, there is a solution that is simple, neat, and wrong. _______________________________________________ Gambit-list mailing list Gambit-list@iro.umontreal.ca https://webmail.iro.umontreal.ca/mailman/listinfo/gambit-list