>Ted Wright <wright at en.com> is widely believed to have written:
>> > > 0 tinyBenchmarks =  '118,408,880 bytecodes/sec; 3,085,422 sends/sec'
>...
>> 0 tinyBenchmarks. '102,236,421 bytecodes/sec; 2994,238 sends/sec'

Tim Rowledge <tim at sumeru.stanford.edu> responded...
>All meaningless. Run some real benchmarks, or failing that (since I've
>never seen any 'real' benchmarks worth a cent) at least try the Green
>book benchmarks. A copy (that ran ok last time I bothered) can be found
>at http://sumeru.stanford.edu/tim/pooters/SqFiles/deltas/sqBenchmarks.cs

I disagree that they are meaningless, but you're right that they don't tell the whole story.  Your suggested benchmarks are certainly useful, and I've also just added a suite of 7 macro benchmarks in the latest spate of updates.  Take a look at SystemDictionary>>macroBenchmarks.

Interestingly, a great macro benchmark that leverages off the old standard is...

	ContextPart runSimulated: [0 tinyBencmarks].

It's a trivially simple meta version of tinyBenchmarks, but it shows that, while J3 shows a 3x and 6x improvement in bytecode speed and send speed, its overall gain is currently typically closer to 2x.

	- Dan

(...another fun one is to run tinyBenchmarks in the InterpreterSimulator ;-)