[cfe-dev] RFC: change string hash function in libc++

[Chandler Carruth]

One point that I think would be interesting to share (if you have any
information):

On Fri, Dec 2, 2011 at 2:42 PM, Craig Silverstein <csilvers at google.com>wrote:

> In favor of
> Murmurhash: it's been around for longer and is already used in other
> STL implementations.  In favor of Cityhash: it performs a bit better
> than Murmurhash, on average.
>

What are the numbers behind these "performs better"?

In particular, I'd like to better understand the tradeoffs between the
following:

1) code size
2) time-to-hash small strings (1 to 16 bytes, 16 to 128 bytes, 128 bytes to
1024 bytes. everything larger is "large")
3) time-to-hash large strings
4) resistance to collisions
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20111202/051f3c85/attachment.html>