I have started computing tables with the following parameters:
DP=12bits , rounds=8, extra clockings=100

These tables will be calculated in 2 stages, I am currently doing the
first stage only, which is to a height of 5.4e9 entries. This gives us
2^32 unique endpoints after removing merges.

My computer uses 2 x ATI Radeon 5850 cards, and with some hard core
optimization done by Sascha, I can output > 34000 chains/second. Tables
can then be completed to stage 1 in less than 48hours.

Writing to disk has been modified, so that it has a pre-sorting step,
and writes to 256 "bucket files" - The final sorting can then be
completed in < 40 minutes.

After sorting, the table can be written directly to a block device. ( I
use a 80GB Intel X-25M SSD drive), and an index file is generated.

Finally lookups can now be performed, (this code is in its infancy, and
much speed improvement will come to fully utilize the speed of the SSD)

In a recent test run, I did have some measure of success when I was able
to find the key to frame 257 in the challenges set (SVN:challenge.bin):

[Lookup from table with ID:100]
#### Found potential key (bits: 37)
#### ae5b16cb41710ba0 
#### Stepping back to mix 
#### ae5b16cb41710ba0 -> ae5b16cb41710ba0 
Candidate: f5212b11b0342617 
Candidate: f5212b88d81c2617 
### Frame is 257 ### 

I have previously been able to find keys that produce challenge output,
but this is the first key that passes the backwards clocking test. The
candidates are equivalent, and will both produce the same GSM frame.

Cheers,

  Frank


_______________________________________________
A51 mailing list
[email protected]
http://lists.lists.reflextor.com/cgi-bin/mailman/listinfo/a51

Reply via email to