Re: [patch 2/7] lib/hashmod: Add modulo based hash mechanism

From: Thomas Gleixner
Date: Thu Apr 28 2016 - 19:27:58 EST


On Thu, 28 Apr 2016, Linus Torvalds wrote:
>
> I'd really hate to add *another* ad-hoc hash when the previous ad-hoc
> hash has been shown to be bad.

I completely agree.

I'm not a hashing wizard and I completely failed to understand why
hash_long/ptr are so horrible for the various test cases I ran.

So my ad hoc test was to use the only hash function I truly understand. It was
state of the art in my university days :) And surprise, surprise it worked
really well.

My main focus was really to solve this futex issue which plages various people
and not to dive into hashing theory for a few weeks.

I'll try to dig up some time to analyze the hash_long failure unless someone
familiar with the problem is beating me to it.

Thanks,

tglx