Robustness of the built in associative array

[Dave]

In article <e0360o$28ep$1 at digitaldaemon.com>, Nick says...
>
>(*) It is uncommon for decent hash functions. It seems the builtin AA is
>optimized to also work with very bad hashers (in which case it approaches a pure
>binary tree.) I guess this isn't a bad feature for a generic AA, but it is also
>some of the reason why my custom hash map outperformed it (I didn't use btrees
>at all.)
>

Just curious - what portion of the perf. diff. was that? What data types? And in
your estimation what accounts for the rest of the diff. (e.g. are you
pre-allocating a bunch of space for the data (as opposed to the key) or not
using the GC)?

Thanks,

- Dave