>
> Java is performing 28 GCs while running the benchmark (spending some
> 60 ms in GC), reducing the heap each time by 500 KB.  This means, if I
> understand the statistics, the benchmark is generating and throwing
> away objects worth of 14 MB.  I'd guess that's some 300.000 objects.
>
> So don't assume there're no objects involved :-)
>
How did you get those results, I used both -Xloggc:log.gc and
later -verbose:gc to
no avail.
Benoit
PS
attached you'll find the test I used , this is just the original java bench
adapted too the
new regex package.


begin 666 regtest.java
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-*"D["@D)?0H)?0I]"@``
`
end