Subject: Re: New hash algorithm - FNV - for use in the kernel
To: Darren Reed <>
From: Darren Reed <>
List: tech-kern
Date: 11/27/2001 21:13:38
In some email I received from Darren Reed, sie wrote:
> In some email I received from David Laight, sie wrote:
> > > There is a potential issue however; FNV uses multiply (*) and xor (^),
> > > whereas a lot of the existing hashes on strings and buffers just use 
> > > addition (+).
> > 
> > Using multiply will (probably) stuff the performance of sytems that
> > don't have fast hardware multiply (eg sparc32).
> Only pre-sparc-v8 has no hardware multiply, from my code generation tests
> using SUNWspro CC.  The SPARC chip in the SS2 is a version 8 (or at least
> the powerup is O:-) and runs "gcc -mv8" code without a hitch.

I take that back.

The microsparc (as in the Classic & LX) appears to be the first chip with
hardware multiply as that's where 'gcc -mv8' of a simple "do lots of
multiplies" program outperforms (quite significantly) without the '-mv8'.

So sun4 & sun4c do not have hardware multiply but the rest should.