Integer micro-benchmarks

[David Simmons]

"Steve Wart" <swart at deadspam.com> wrote in message
news:UKNF6.98935$166.1860248 at news1.rdc1.bc.home.com...
> 1.2GHz Athlon, Win2K SP2, 512Mb RAM -> David
> 333 PII, Win2K SP1, 192Mb RAM -> Steve
> [I gotta get in Redmond's good books :)]

That's not how I got the machine if that's what you're thinking.

As I mentioned in some other posts, you can build my exact machine
configuration for under $2,000 (which is amazing to me). Actually, you can
build a better machine now for that money. I built my machine a few months
ago for under $2000.

My brother is in the process of building two new machines (w/slight
variances between them), each for under $2000 with the spifiest graphics
cards for games, DVD, sound system, ATA 100 raided 7200 RPM drives, 512MB
memory, and 1.3GHz Duron with the 266Mhz FSB.  Earlier today, someone told
me you can buy the 1.7GHz Pentium processor for $350 now -- unbelievable...

As to the Win2K SP2, you just have to pay the annual MSDN fees of ~$600 for
professional subscription, or $2000 for universal subscription.  Which is
basically on par with Apple's ADC fees for being a developer -- actually its
the model Microsoft copied and improved upon.

>
> "David Simmons" <pulsar at qks.com> wrote ...
> > "Steve Wart" <swart at deadspam.com> wrote ...
> > > Here is some more silliness.
> > >
> > > If you take the #triangle method and turn it into a block closure, as
> below,
> > > in VW it takes almost 10 seconds (on my 333 PII)
> >
> > Really!  10 seconds versus 7 seconds?
> >
> > Good goobley goo. I just ran it on VW and got the times:
> >
> > 1784ms for 9,000,000 loops using <result> local to the block.
> >
> > 4102ms for 9,000,000 loops using <result> as a method temp (shared with
> the block).
> >
> How come I have about a 30% difference and yours is over 100%? Is this a
CPU
> caching thing?

I don't know. I've re-checked the runs though because this stands out as
weird.

We should look further into this anomaly to understand it because it makes
the benchmarking numbers hard to properly assess.

Maybe Eliot could answer this one?

My only guess is that using a method temp is resulting in some kind of
context being allocated every time the block is evaluated. If so, it might
be a bug in the copying-block analyzer of the compiler?

>
> > The same runs on SmallScript were:
> > 1313ms vs 1380ms respectively

By the way, the Squeak (latest non-alpha version) runs on my box yielded:

    20,940ms and 18,827ms respectively.

>
> This is the AOS VM?

Yes.

> I wonder how it would do on the .NET VM? If you are not
> at liberty to say, tap twice :)

Hmm, well, sorta. Tap once: Let's just say its markedly slower.

>
> Does SmallScript come with AOS?

Yes. Both the SmallScript compiler and a corresponding execution engine (the
core AOS Platform VM) will be free.

The SmallScript compiler and scripting/dynamic language support libraries
for the Microsoft.NET Platform will not be free. As to other frameworks,
tools, etc -- that will be determined on a case by case basis.

The AOS Platform is a quasi language independent
object-model/virtual-machine which is now in its 4th major generation since
I first designed it back at the end of 1990. That design was derived
(extensively influenced) from work I did on C based (vm) object systems
beginning back in 1986.

SmallScript began its life in 1998 as a new language that is derived from my
work in creating and evolving QKS Smalltalk from 1991 to 1998. SmallScript
includes all of the smalltalk "language" (not the ANSI Smalltalk frameworks
or Smalltalk-80 stuff) plus many other extensions.

-- Dave S.

>
> Steve