Is something wrong with build 2.0?

[David Simmons]

I was just showing it off after talking about its performance. I was
really surprised to find it was slow...

On a 2.6GHz processer it does 71M loops per second - empty loop.
It calls an empty function or one which returns a member at about 250M/
sec.

That is between 1/2 and 1/4 the speed of the dynamic language jits I
am familiar with.

?
David Simmons

[David Griswold]

Did you follow the instructions at
http://www.strongtalk.org/benchmarking.html? The most common benchmarking
mistake is putting the loop in a doIt, and/or not running it multiple times
so that the compiled version gets used, since the current implementation
doesn't have on-stack replacement (it can't switch an activation from the
interpreted to the compiled form while the activation is running). You are
probably seeing interpreted performance.
-Dave

[David Simmons]

Thanks David(G). My bad.

After your appropriate chastisement ;-)

I read the benchmarking guideline. I then installed the test code as
methods. I ran it a half dozen times or so, saw it was still slow in
each. Went and ran the same benchmarks on the other vm of interest.
Came back to strongtalk workspace. Ran it again on strongtalk. And saw
that strongtalk was now jitting/optimized and getting the same perf
times as the other vm of interest.

Problem fixed. Much happier now :-)

Thank you David.

On Mar 18, 4:27 am, "David Griswold" <David.Grisw...@acm.org> wrote:
> Did you follow the instructions athttp://www.strongtalk.org/benchmarking.html? The most common benchmarking

> > David Simmons- Hide quoted text -
>
> - Show quoted text -