It might be a FAQ question, but anyway: are there any guidelines forQuote:> [...]
>Normal programmers, including the implementors of JVMs that want to support
>real concurrency on Solaris 2.x, code to the threads API at layer (2) and
>should never mess with LWPs, except perhaps to give the library code a hint
>about how many it should allocate.
defining the optimal number of LWPs for one's program? Is it the
number of CPUs in the box? I don't think that thread or pthread
libraries fire up LWPs dynamically, so one have to do it on one's own.
Am I wrong here?