Em Ter, 2010-05-18 às 12:58 -0700, Alex Elsayed escreveu:
> You are imposing a false dichotomy here. Neither 'green' threads nor kernel
> threads preclude each other. In fact, it can be convincingly argued that they
> work _best_ when combined. Please look at the GSoC proposal for hybrid
> threading on the Parrot list.
While I agree that there isn't a dichotomy, the point here is more in
the lines of:
1) Green threads are usually related to the requirement of serialized
access to data so you can share all data in the thread without
resorting to locks for every value.
2) If that requirement is dropped, once only data that is explicitly
marked as shared can be seen by both threads, the point for green
threads is moot, since the OS threads are always going to be better
performant then a manually implemented scheduler.
My original idea was pointing in creating a "shared memory space" that
would be seen by every green thread in the same os thread, where some
lines would be drawn to allow OS threading with different memory spaces
- message passing would be used to communicate between two different
"memory spaces".
But what we might be getting here is at the point where we don't need
green threads at all... I'm still not sure about one point or another,
tho..
daniel
