[stringtemplate-interest] ST4 faster then ST3, some quick numbers

[Oliver Zeigermann]

I was using caliper for the micro benchmarking of a simple template
language that allows both compiled and interpreted mode. As there is
quite a bit of dynamic (runtime) stuff that can't be directly compiled
in template languages the speedup was only between 2 and 5.

2011/1/20 Terence Parr <parrt at cs.usfca.edu>:
> apparently 10,000 is the magic number of iterations to warm up the compilation. Can you do it 10000 times and *then* run your tests?  Take a look in the benchmark dir actually. you'll see my tools based upon reading cliff click and josh bloch on micro benchmarking.
>
> still, great news :)  once we get over compilation, I wonder how much faster just exec time is.
>
> Ter
> On Jan 19, 2011, at 6:03 PM, Collin Fagan wrote:
>
>> Hi Everybody,
>>
>> I did a quick benchmark comparing ST3 and ST4.
>>
>> ST3 - Time: 34.6 milliseconds
>> ST4 - Time: 18.3 milliseconds
>>
>> These numbers are the average time it takes to generate 10 'builder' classes from 10 Swing classes averaged over 1000 trials. I tried to only benchmarked the ST code and ignored all the I/O and reflection. The only exception is that in both cases the template is loaded from a file using ST API calls. So it looks like that means ST4 is 47% faster then ST3 according to my meatball benchmarks.
>>
>> !YAY for ST4!
>>
>> Collin
>>
>>
>>
>>
>>
>>
>>
>> _______________________________________________
>> stringtemplate-interest mailing list
>> stringtemplate-interest at antlr.org
>> http://www.antlr.org/mailman/listinfo/stringtemplate-interest
>
> _______________________________________________
> stringtemplate-interest mailing list
> stringtemplate-interest at antlr.org
> http://www.antlr.org/mailman/listinfo/stringtemplate-interest
>