[rfa] libiberty: splay tree performance improvement

[Richard Henderson]

On Mon, Jun 28, 2004 at 06:23:45PM -0400, Frank Ch. Eigler wrote:
> You're right.  So does this new requirement sound acceptable,
> or else shall I fork splay-tree for libmudflap's pattern of usage?

I don't think it's reasonable as a new requirement.  Too often it's
impossible to compare data at all without allocating space somewhere
temporarily.

I do think it would be reasonable to add a splay_tree_lookup_cache
entry point that explicitly requires stable keys.


r~