Re: Moving pthread_(cond|mutex) to libc

[christos%astron.com@localhost (Christos Zoulas)]

In article <20130124160140.GF1055%homeworld.netbsd.org@localhost>,
Emmanuel Dreyfus  <manu%netbsd.org@localhost> wrote:
>On Thu, Jan 24, 2013 at 04:42:46PM +0100, Joerg Sonnenberger wrote:
>> I don't see any such consensus. Such move penalises non-thread
>> applications and is objectable for that alone.
>
>Right. Then please go back there and propose anyther way to fix the problem:
>http://mail-index.netbsd.org/tech-pkg/2013/01/19/msg010635.html

I think that the next step is to actually measure the time difference
between the current stub implementation and the proposed change.
Perhaps the overhead of the constructors is tolerable, and the
overhead of the wait routines can be mitigated by checking the
current number of threads in the application and returning if the
application is not threaded. The question is what workload? Perhaps
a full system build would be good. Any other suggestions or ideas?

christos