[Iccrg] About the state of MulTFRC

[Emmanuel Lochin]

Hi Michael,

I read some comments concerning the complexity of the computation. For
the sake of decreasing TFRC computational complexity, note that
Guillaume Jourjon simply proposed a Newton algorithm to initialize the
loss history structure. Compared to the standard implementation you
could find in FreeBSD, the proposed method halves the # of iterations
in the worst case scenario : see http://oatao.univ-toulouse.fr/211/
for details. This does not solve everything but at least this is an
improvement :)

Cheers,

Emmanuel

On 28 July 2012 05:30, Michael Welzl <michawe at ifi.uio.no> wrote:
> Hi,
>
> It's ages since the last time I posted something about MulTFRC (
> http://heim.ifi.uio.no/~michawe/research/projects/multfrc/index.html ) here.
>
> So I thought I'd send this message, to tell you all that we're still on the
> case. The situation is - we got reviews, addressed all comments, EXCEPT for
> the request to come up with a good fixed point version of our algorithm.
> Note that the current draft can be implemented and used whenever floats are
> available (normally in user-space).
>
> Developing a fixed point version has turned out to be muuuuuuuuuuuuuuuuch
> harder than we thought. We're still working on it (right now I have a master
> student who is looking into this), but it just takes it time.
>
> Cheers,
> Michael
>
>
> _______________________________________________
> Iccrg mailing list
> Iccrg at cs.ucl.ac.uk
> http://oakham.cs.ucl.ac.uk/mailman/listinfo/iccrg



-- 
"This email and any attachments are confidential. They may contain legally
privileged information or copyright material. You should not read, copy, use
or disclose them without authorisation. If you are not an intended
recipient, please contact us at once by return email and then delete both
messages. We do not accept liability in connection with computer virus,
data corruption, delay, interruption, unauthorised access or unauthorised
amendment. This notice should not be removed"