Hi there,
while working on a quite low-power embedded System (a 68k derivative
clocked at 166 MHz), I realised that libtommath was simply too slow.
Just connecting to an ssh-server using dbclient took about 30-40 sec.
Therefore I adapted the whole dropbear code and libtomcrypt (which
already comes with an appropriate interface) to use libtomsfastmath,
which heavily uses fix-precision numbers and assembly.
The patch is nearing its completion, for now a connection using
dbclient takes about 11 sec, from local commandline to logoff after a
remote "echo hello". Unfortunately the dropbear server seems to have
broken while doing m68k assembly speedups, but this should be
fixable. On an X86 this patch likewise greatly increases the
performance of all public-key algorithms.
It would be interesting how close Marcos and my approaches regarding
speedups are under real word conditions, say slow/underpowered/
embedded computers like routers, ,,netbooks'', dboxen,etc.
If there is any demand for a pre-alpha release maybe I would put it
on my webpage for you to download, just let me know.
Best regards,
Peter
On Dec 17, 2008, at 11:20 PM, [email protected] wrote:
Dear dropbear developers,
I was experimenting with the Toom-Cook multiplication algorithm,
for large
numbers, and I run across LibTomMath. Unfortunately the main
project seems
dead and forgotten, so I searched for an active project using its
code and
I've found Dropbear!
I patched the 3-way Toom from libtommath with some simple
optimizations
( http://ln.bodrato.it/FasterToomConvolution_pdf ), now it is
significantly faster: on my laptop it "obsoletes" the Karatsuba code,
since it is as fast a Karatsuba for small operands, but faster for
bigger
ones.
Now the question is: are you interested in the patch?
(maybe you prefer not to modify an "external and imported" library...)
In the case you are interested, how should I submit the new files?
Thank you for any answer,
Marco
--
http://bodrato.it/
