On Wed, 2007-02-14 at 00:18 -0500, Chris King wrote:
> Hi,
> 
> I got curious about fthreads and decided to try implementing the
> Shootout's "cheap concurrency" test [1].  However I'm encountering
> very odd problems.

BTW: there is a 'performance' target in the build system which
builds a custom driver and spawns threads and sends them messages.

This test make 500,000 threads and sends each one a message.
On my AMD64 this takes 2 seconds: this is a transaction
rate of 500K/sec. This easy solves the so-called '10K' problem,
in fact it makes it into a joke.

The shootout test is therefore entirely useless as a performance
measurement.

In fact .. most of the overhead in this performance
test is probably nothing to do with Felix threads:
the objects are created with 'malloc', and, since they're
roots, the current GC is using an STL Map to hold the
registrations, which is O(log n).

If the test make 500K threads and delivered a lot more
messages the transaction rate would probably go up.


-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net

-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys-and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Felix-language mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/felix-language

Reply via email to