Re: [boost] [review] Review of Unordered library begins today December 7

From: Daniel James (daniel_james_at_[hidden])
Date: 2007-12-19 16:52:31

On 19/12/2007, Jamie Allsop <ja11sop_at_[hidden]> wrote:
> First apologies for the belated review and its brevity - I really had
> hoped I'd have time to be more thorough, but here goes.

Thanks for reviewing the library - suggestions are always welcome,
especially help with improving the documentation. I'll look into your
suggestions soon (they all look worthwhile).

> One thing that would be nice are some performance details, perhaps
> comparisons against some other library implementations using the same
> has function. Just a thought.

A problem is that some implementations require better quality hash
functions because they use a power of 2 for the number of buckets. So
it's probably worth trying a some different hash functions.

> I haven't done a lot of work with hash containers before (I
> haven't needed to). I DO need to make heavy use of them now so this
> library is very timely. I've unfortunately not been able to get time to
> replace our existing implementation with this one and do a performance
> comparison.

I did some informal benchmarking a while ago and found the gcc version
to be a little faster (sorry, I didn't record the figures). I think
this was partly because it doesn't fully support allocators. I'm a bit
weary of publishing artificial speed comparisons so some serious real
life comparisons would be very useful.

I'll see if I can dig out my benchmarks, they used boost headers as
test data so they weren't completely unrealistic.

Daniel