Re: Objective-C and Smalltalk; speed of message send

[Richard Frith-Macdonald]

On 10 Aug 2004, at 09:32, Marcel Weiher wrote:

> On 8 Aug 2004, at 05:08, Lyndon Tremblay wrote:
>
> > "5 timesRepeat: [ Transcript show: 0 tinyBenchmarks; cr ]"
> >
> > Both compiled at -O3,
> > C Function - 1938968 calls/sec
> > Objective-C Message - 824973 sends/sec
> >
> > This is under a system with stable load, tested multiple times. I've 
> > had the
> > Squeak test at "144m bytecodes/sec; 7.7m sends/sec", so perhaps an 
> > active
> > image had an effect here. Um, tell me what I have missed, or what is 
> > really
> > going on. I am surprised enough to not believe it is a fair test.
>
> Apart from what Richard said, it seems you are primarily measuring the 
> speed of the getrusage() system call that clock() tends to be based 
> on...

Actually no ... if you examine the code carefully, you will see that 
clock() is used to time operations (one call to clock() before the 
operation and one call after), but that the only operation actually 
used for presenting a result is the fist one which takes over a second 
to execute ...  so the execution time of clock() is insignificant.

> Actual message passing speed for a constant receiver is around 50 ns 
> on 1 Ghz G4, with a third of that going to inter-module calling 
> overhead.  So around 20 million message sends per second.  Squeak on 
> the same system is significantly slower.

That's interesting.
Is that with the apple or gnu runtime?
Is that repeated uses of the same message to objects of the same class? 
 I ask because the apple runtime caches those and is much faster if you 
benchmark that than if you benchmark against a large variety of 
different methods.  Of course, you can argue that performance is most 
important in a tight loop sending one message repeatedly ...