[Help-smalltalk] Re: Benchmarks

[Paolo Bonzini]

On 02/14/2011 11:51 AM, Gwenaël Casaccio wrote:
> >  4) Now here comes the problem:
> >  =======================
> >
> >  SmallInteger [
> >    foo: anInteger time: aTimeInteger [
> >      anInteger>  0 ifTrue: [
> >        ^ self  foo: anInteger - 1 time: aTimeInteger
> >      ].
> >
> >      ObjectMemory quit.
> >    ]
> >  ]

You're calling this with anInteger = 90000, and in this case I do expect 
GC to be responsible for bad performance.

However, the numbers should be very different for, say, a depth of 50 
like in your microbenchmark

  (Time millisecondsToRun: [ 1000000 timesRepeat: [
     5 recursionWithReturn: 50 ] ]) printNl.

How do gst/cog/squeak compare in this case?

Also, your benchmarks are missing one very important case, namely array 
access.  I believe this is the cause of the slowdown in the bytecode 
benchmark, especially since you proved that everything else is faster. 
:)  This cannot be helped really, because it's due to the object table.

Paolo