Re: NPT vs. NGPT benchmarks

On 09/22/2002 at 03:38:36 PM MST, Dan Kegel <[email protected]> wrote:
>
> Would the NGPT folks care to comment on Ulrich's thread
> benchmark comparing his NPT to NGPT and LinuxThreads?
> It's at
>   http://marc.theaimsgroup.com/?l=linux-kernel&m=103269598000900&w=2
>
> Also, I now contrast the two libraries at
>   http://www.kegel.com/c10k.html#threaded
> and would appreciate any suggestions or corrections.
> - Dan
> _______________________________________________

The NGPT team is reviewing the posting.  Unfortunately, the post does not
show the benchmark that was used nor does it indicate what kind of
processor configuration was used.  As everyone using NGPT knows, NGPT is
highly configurable and the default settings do not always give you what's
best for a particular application.  I suspect (it would be nice to know)
that the benchmark was probably run on uniprocessor using the defaults (at
least from any NGPT standpoint).  If that is truly the case, the numbers
would not surprise me.  NGPT is designed to take advantage of and excel in
a server environment and is optimized for it.  That is not to say that it
can't be run in a client environment, you just need to configure it
properly for such an environment.


Regards,
     Bill Abt
     Senior Software Engineer
     Next Generation POSIX Threading for Linux
     IBM Cambridge, MA, USA 02142
     Ext: +(00)1 603-673-3440
     T/L: 253-9938
     Cell: +(00)1 617-803-7514
     [email protected] or [email protected]
     http://oss.software.ibm.com/developerworks/opensource/pthreads