Monday, February 22, 2010

Forks and Threads

I was looking at what I wanted for concurrency again, and decided against almost everything I so smartly thought of before. Given how chips are build these days, with support for processes, threads, and some interprocess communication the only plausible thing to do is:
  1. Support fork, create a copy of the whole process, and make use of the fact that modern chips have copy-on-write triggers.
  2. Support light thread creation, which should be mapped to some eval f a0 ... an construct, i.e., evaluation of a distinct function.
  3. Possibly support very light threads, for example by just having simple round-robin primitives in the language.
Not the academically nicest solution, but probably the best I can -easily- do.

Bye Erlang

No comments:

Post a Comment