Re: [openpgp] SHA-x performance

[Werner Koch <[e-mail]>]

On Tue, 11 Aug 2015 19:48, [e-mail] said:

> asm, C, Visual Basic, whatever), it's still going to be way faster than you
> ever need.  The only difference will be whether it's 100x faster than required
> or 200x faster.  

Depends on what you want to do: To feed a big pipe you will need a
reasonable fast CPU anyway because there are other things to do beside
crypto.  Thus it matters for those applications whether you can you
still use a cheaper CPU.

> Where it matters is IoT implementations, 180MHz STM32's and the like.  For
> example on a Cortex A7 (which is way more powerful than most IoT devices use,

Right that is a different class but fortunately we can run our protocols
easily there.  I concur that a MUST algorithm in OpenPGP should work
well on small devices.  This is also the reason why one algorithm
_might_ not fit all devices.

bench.cr.yp.to has a lot data points but AFAICS it start with an A8 and
it any case there is too much data for a useful discussion.  Do you have
a suggestion on what CPUs from low to high end to do benchmarks so to
check which SHA variant is suitable?  Although, I assume that SHA-256
will likely be the best over all CPUs, having some concrete data points
may help to convince other folks.


Shalom-Salam,

   Werner

-- 
Die Gedanken sind frei.  Ausnahmen regelt ein Bundesgesetz.