[viff-devel] Optimizing preprocessing

[Marcel Keller]

Hi Janus,

> I am not opposed to your suggestion. However I would like to point out 
> that in VIFF you compute on shares and not field elements!. Computing 
> directly on the field elements is hacking the abstractions of VIFF.

I don't think so. Since I work with VIFF, it treats field elements as 
they would be shares with threshold 0. You can multiply and add field 
elements and shares. Furthermore, if you open to shares and multiply 
them, VIFF doesn't multiply locally, but executes the multiplication for 
shares of higher threshold. You have to do the following to achieve 
local multiplication:

a = open(share_a)
b = open(share_b)
c = gather_shares([a, b])
c.addCallback(lambda (a, b): a * b)

Therefore, I think that VIFF computes on field elements as well as on 
shares almost equally.

> Computation on field elements or rather the representation of a Share 
> can be useful as an optimization, however this optimization should be 
> confined within applications or runtimes, and should not progress over 
> interface boundaries as I fear you are suggesting.

I believe that if you would implement an OrlandiTuple class and overload 
operators there with the current _plus, _minus, and _const_mul 
functions, the OrlandiRuntime could reuse the add, sub, and lin_comb 
functions from PassiveRuntime and maybe even the mul function from 
BasicActiveRuntime. The OrlandiTuple would then be something equivalent 
to a FieldElement in the older runtimes.

Best regards,
Marcel