Probably Dance

I can program and like games

Tag: hash table

Fibonacci Hashing: The Optimization that the World Forgot (or: a Better Alternative to Integer Modulo)

I recently posted a blog post about a new hash table, and whenever I do something like that, I learn at least one new thing from my comments. In my last comment section Rich Geldreich talks about his hash table which uses “Fibonacci Hashing”, which I hadn’t heard of before. I have worked a lot on hash tables, so I thought I have at least heard of all the big important tricks and techniques, but I also know that there are so many small tweaks and improvements that you can’t possibly know them all. I thought this might be another neat small trick to add to the collection.

Turns out I was wrong. This is a big one. And everyone should be using it. Hash tables should not be prime number sized and they should not use an integer modulo to map hashes into slots. Fibonacci hashing is just better. Yet somehow nobody is using it and lots of big hash tables (including all the big implementations of std::unordered_map) are much slower than they should be because they don’t use Fibonacci Hashing. So let’s figure this out.

Read the rest of this entry »

A new fast hash table in response to Google’s new fast hash table

Hi, I wrote my new favorite hash table. This came about because last year I wrote the fastest hash table (I still make that claim) and this year one of the organizers of the C++Now conference asked me to give a talk. My problem was that Google had also announced a new fast hash table last year, and I wasn’t sure if mine would compare well against theirs.

The main benefit of Google’s hash table over mine was that Google’s has less memory overhead: It has a higher max_load_factor (meaning how full can the table get before it grows to a bigger array) and it has only 1 byte overhead per entry, where the overhead of my table depended on the alignment of your data. (if your data is 8 byte aligned, you’ll have 8 bytes overhead)

So I spent months working on that conference talk, trying to find something that would be a good response to Google’s hash table. Surprisingly enough I ended up with a chaining hash table that is almost as fast as my hash table from last year, while having even less memory overhead than Google’s hash table and which has this really nice property of having stable performance: Every hash table has some performance pitfalls, but this one has fewer than most and will cause problems less often than others will. So what that does is that it’s a hash table that’s really easy to recommend.

Read the rest of this entry »

I Wrote The Fastest Hashtable

I had to get there eventually. I had a blog post called “I Wrote a Fast Hashtable” and another blog post called “I Wrote a Faster Hashtable.” Now I finally wrote the fastest hashtable. And by that I mean that I have the fastest lookups of any hashtable I could find, while my inserts and erases are also really fast. (but not the fastest)

The trick is to use Robin Hood hashing with an upper limit on the number of probes. If an element has to be more than X positions away from its ideal position, you grow the table and hope that with a bigger table every element can be close to where it wants to be. Turns out that this works really well. X can be relatively small which allows some nice optimizations for the inner loop of a hashtable lookup.

If you just want to try it, here is a download link. Or scroll down to the bottom of the blog post to the section “Source Code and Usage.” If you want more details read on.

Read the rest of this entry »