Is anyone working on getting a two-tier threading architecture going
for java-linux? I haven't had much luck even in tracking down
existing c-level implementation libraries - perhaps someone else has?
I have some reservations about using the native threading
implementations of the blackdown port because one java thread == one
linux kernel thread, but linux kernel threads are a severely limited
resource.
I understand Linus' argument for keeping the current single-level linux
scheduler (for both threads and processes), but the argument breaks
down when one tries applying this order of restriction to
essential language-level features like java threads (to be silly,
would you accept a limit of only 1000 String instances?).
My application uses many (1000s) simultaneous threads, but only during
the development cycle: Essentially, the deployed system runs over a
large network of peers (say, on 100 different cpus). Each peer needs a
fairly small number (10s) of threads for natural operation. It is, um,
inconvenient to give each developer a hundred or so machines for
testing the system behavior. Altering the system to be targetted to a
small number of peers renders the results uninteresting.
So, for now, while I run my green-threaded java VM, I keep the other
CPU of my SMP machine busy running seti@home and playing mp3s ...
Cheers,
-mik
--
Michael Thome ([EMAIL PROTECTED])
----------------------------------------------------------------------
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
