Re: call_outs

Yes, thats right. Its the "standard" case which would not benefit much from the proposed optimization since buckets should be small. There is another extreme case, having objects with `() lfun being all equal with same __hash(). That would be the worst case and the optimization would not help at all.

Regarding benchmarks: The last column in the output seems meaningless in many benchmarks. Since its a float before printing, maybe some precision should be added.

Are those benchmarks run automatically somewhere, maybe xenoclient?

On Thu, 7 Apr 2011, Martin Nilsson (Opera Mini - AFK!) @ Pike (-) developers forum wrote:

Well, you can bias the other way by adding a call out benchmark using lambdas, right?