Okay, good to know. What kind of info would you need to get into things?

Also, any idea why the report says that my Bootstrap.main(String[]) is called over 1000 times? Seems a bit odd to me. This method literally does a little string parsing, then calls starts Forge in a Thread (and waits for that thread to finish.)


On Tue, Jan 29, 2013 at 5:33 PM, Stuart Douglas <stuart.w.douglas@gmail.com> wrote:
Tracing is pretty useless at identifying slow parts, as the overhead it adds distorts the profile too much. The only thing it is really good for is identifying methods that are called too often.

Sampling is generally much better.



Stuart

Lincoln Baxter, III wrote:
What kind of information do you need? I was unable to get YourKit to
dump a .snapshot file for some reason. If that's what you need, I can
try again. Which profile mode? This was done using tracing.

Thanks!


On Tue, Jan 29, 2013 at 4:32 PM, Stuart Douglas
<stuart.w.douglas@gmail.com <mailto:stuart.w.douglas@gmail.com>> wrote:

    The screenshot does not really tell us much. We would need to see
    the actual profile information.

    Stuart


    Lincoln Baxter, III wrote:

        Hi Jozef, Stuart, and Weld-devs,

        In Forge 2 we are using Weld extensively, and one of the things
        we do is
        start up many instances simultaneously.

        We may have anywhere from one to one-hundred or more weld instances.
        Currently we have only seen around 10-12 instances, and
        performance is
        "Okay", but in theory, we could see hundreds of instances, at which
        point, performance starts to be a concern. We're working around this
        problem by disabling CDI support on some internal addons, but...
        it's
        not really reasonable to expect that everyone will do this.

        Which means... we need to figure out how to shave as much time
        off the
        bootstrap as possible. Currently each weld instance takes
        anywhere from
        80ms to 450ms to start (not really sure why such variation yet,) and
        we'd hopefully like to get that down even lower, around 10-20ms.
        Classloading time only would be optimal, but obviously difficult
        to achieve.


        How can we get the most speed out of Weld? Most of our
        deployments have
        only ~15 bean classes at most. It seems like a lot of time
        (~30-40%) is
        being spent in the Google concurrent collections.

        (Screenshot attached.)

        Thanks,

        --
        Lincoln Baxter, III
        http://ocpsoft.org
        "Simpler is better."




--
Lincoln Baxter, III
http://ocpsoft.org
"Simpler is better."



--
Lincoln Baxter, III
http://ocpsoft.org
"Simpler is better."