[armedbear-devel] [j-devel] Improving startup time: sanity check

[Erik Huelsmann]

> On Wed, Oct 28, 2009 at 11:20 PM, Alessio Stalla
> <alessiostalla at gmail.com> wrote:
>> The test might be wrong
>
> It was, indeed, wrong. It measured class loading time too for
> reflection, but that can't be eliminated, no matter what instantiation
> technique you use.
>
> New results on 50000 iterations:
>
> REFLECTION: 9660480097
> NEW: 88353767
> % SLOWER: 10933
>
> REFLECTION: 9602410291
> NEW: 78459331
> % SLOWER: 12238
>
> REFLECTION: 9925574038
> NEW: 86923015
> % SLOWER: 11418
>
> ~100 times slower is still impressive.

I'm actually very impressed by your idea. I had a similar idea, but
got stuck while trying to figure out how to call the static function.
Your solution is - ofcourse - the only way to go!

I'm also very excited about the route you choose to go: I think I can
implement this in one evening into the current code base: All that's
needed is a change to loadCompiledFunction and the generation of a new
method for each class file during the compilation phase somewhere in
compiler-pass2.

Someone else might even be able to do it!

Bye,

Erik.