[svlug] Re: Thread Programming and Java

[Jeffrey B. Siegal]

Tin Le wrote:
> Agreed.  But there would still be a large number of threads in the
> system, whether they are active or not.  Linux does not deal well with
> lots of threads.

But, as I said, the issue raised by IBM is *runnable* threads.  I do not
believe there is any performance problem with having many sleeping
threads.  Certainly IBM didn't find one (I don't think they looked).

> Secondly, because of the way thread is implemented [in
> Linux], there is hard limitation on how many you can have - roughly 4000
> threads max.

I'm not positive but I think this has been removed in newer x86 kernels,
and probably never existed on non-x86 kernels (the limitation arises
from the size of the hardware task table on the x86).