Jacob Christian Munch-Andersen

I don't particularly like xxHash, see this issue to learn about why: https://github.com/Cyan4973/xxHash/issues/180 The designer doesn't really know what they are doing. xxHash3 is practically okay-ish, just good enough for...

I didn't think of this until now, maybe you considered this construction obvious, but I just thought that if we did parallel tracks we would run out of registers, and...

Required context: [https://peter.website/meow-hash-cryptanalysis](https://peter.website/meow-hash-cryptanalysis) Thank you for doing this. I'm impressed by the work you put into this, far beyond just producing proof of a break. Thoughts and comments in no...

1. I think whole shifted input thing is out in any future designs, it didn't do what it was supposed to do, and performance-wise it only works on X86. I...

@cmuratori I'd estimate the combined chance that a non-adversarial change would match the attack pattern and actually succeed to produce the collision scenario is below 2^-128, but only just. I...

Did a take on a disclaimer. I have worded it with a higher priority for authentication customers than hash table customers as the hash table case requires some pretty serious...

Maybe include a link to the analysis, seems weird not to have that.

It is internal cancellations needed to cancel the entire change, so basically what it says it that something unlikely has to happen 11 times for a full internal collision to...

I found a cool trick while playing around with a new function, with 3-reg instructions you can switch what registers get used for what lanes during computation, without using any...

And how does that work out if you want a C compatible library?