[tahoe-dev] Running Tahoe on ARM plugs

Zooko O'Whielacronx zooko at zooko.com
Fri Feb 18 23:29:55 PST 2011


> It appears that this lil guy has hardware AES-128 support [1] that
> reportedly more than doubles AES performance [2]

Oh, good find!

In addition to reducing the time taken to compute AES-128 by a factor
of 2.5 (according to [2]), it also presumably reduces the load on the
CPU during that time since the AES-specific circuits are doing most of
the work, right? So it could free up more than just 2.5 times as many
CPU cycles for other purposes.

So I wonder if a Davies-Meyer based on AES-128 would have both (a) a
very good chance that it has 2nd-preimage-resistance, and (b)
dramatically better performance on SheevaPlugs than a dedicated hash
function which would have to run in software.

> but unfortunately it
> appears like the crypto processor is only accessible from kernel mode
...
> so the distro, or the user, would
> have to patch them in - and the number of people who are going to roll
> their own custom patched kernel is pretty small compared to the number
> of people who might theoretically want to run Tahoe on a plug).

Well, if people aren't willing and able to do that, then they could
run Davies-Meyer-AES-128 in software. I wonder how efficient that
would be.

> Perhaps some of the DSP instructions could be used to speed up the FEC
> encoding/decoding? I haven't investigated this area at all yet.

My assumption is that FEC will be a minimal cost that it won't be
worth optimizing. :-) But I would like to find out real numbers!

Regards,

Zooko


More information about the tahoe-dev mailing list