[viff-devel] What to benchmark

[Martin Geisler]

ivan at daimi.au.dk writes:

> I think there was earlier some version where arithmetic was done by
> calling some external C code. From that I am guessing that it is
> feasible to make a version where all or most of the stuff in 1) is
> done by calling specific functions we can name and track rather than
> using the internal Python arithmetic, for instance. In such a
> version, it should be possible to find out how much time is spent on
> 1). If this gets much slower than the normal version, we are in
> trouble and then I don't know what to do.

Sigurd is actually testing this at this very moment (we talked about
it on IRC) and I hope he will give some benchmark results. This is
about using GMPY for field arithmetic:

  http://tracker.viff.dk/issue10

> 2) I suppose can be measured by hooking into the event loop of
> Twisted

That was what I described in the mail before -- I saw very few calls
to the select() function, which is the one used in the event loop to
sleep while waiting for data from a set of file descriptors.

> Exercise: if you can measure 1) and 2), how do you measure 3)?  :-)

Hehe :-)

-- 
Martin Geisler

VIFF (Virtual Ideal Functionality Framework) brings easy and efficient
SMPC (Secure Multi-Party Computation) to Python. See: http://viff.dk/.