[LLVMdev] FoldingSet #collisions comparison

Gregory Petrosyan gregory.petrosyan at gmail.com
Wed Feb 10 16:57:13 PST 2010


On Wed, Feb 10, 2010 at 04:54:36PM -0800, Chris Lattner wrote:
> Thanks for doing the evaluation!  It sounds like we should stay with SFH.
> In addition to fewer collisions, it is cheaper to compute.

While I agree that staying with SFH makes sense, I should note that in
microbenchmarks, MMH wins by a big margin (see the original patch).

	Gregory



More information about the llvm-dev mailing list