Re: Reducing AES tables spaces

["Arnaud Lacombe" <lacombar%gmail.com@localhost>]

On Tue, Nov 18, 2008 at 9:37 PM, Matt Thomas <matt%3am-software.com@localhost> 
wrote:
> Here's my AES implementation.  It's small (5K of code and rodata on i386)
> and quick.  I've been meaning to replace sys/crypto/rijndael with it.
> I consider the existing code to be most writeonly code and wanted to
> write an implementation I could understand.
>
> http://www.netbsd.org/~matt/aes.pax
>
I just had a look at it, it is exactly what I propose, except I didn't
re-invent the wheel. You just use a central block processing function
and use loop rather (partly-)unrolled code.

> Instead of hand tweaking things, I let the compiler decide what to do.
>
what do the compiler decide in this case ?

> The heart of the algorithm is only 143 lines and half of that is comments.
>
sure, if you don't count key setup as part of the algorithm, the line
count drop. The current code is in the same order of magnitude. It
should not be hard to adapt it to use a central block processing
function (even pretty easy).

 - Arnaud