Re: [Ocaml-lib-devel] New Library 1: XList

[Nicolas C. <war...@fr...>]

> > - while talking about performance, could someone do some test program
for it
> > ? I would like to test for example the improvement of inlining the
setcdr
>
> I have a "beta" implementation of something like the Perl Benchmark
> module:
>
>  http://www.bagley.org/~doug/ocaml/benchmark/
>
> It can help to answer such questions as: "is it faster to do it this
> way or that way?"
>
> I've attached an example below, which indicates, as one might expect,
> that inlining helps with byte code, but not so much with optimized
> code.

Thanks Doug, do you volunteer for further timing contests ? :-)

Thanks that's what I wanted to hear. I think then we should inline such
calls ( I myself like the bytecode, and you can't always easily switch to
native when you're using some features like Dynlink ). Definition proposal
for inlined calls : we should inline calls that does not require more than 1
or 2 easy lines of code, and are called many times.

BTW, I was also wondering about the local-loop definitions against a
separate-function, but since the loop is only called once, that shouldn't
make even a small difference.

What about the Brian's sort performances ?

Nicolas Cannasse