On Fri, Jun 04, 2010 at 05:30:54PM +0300, Ber Lost wrote:
> Hello,
> I have some questions about the process of computing a chain point according
> to the first format dp15k32.
> 
> 1) In order to find a distinguished point, first we compute the a51 part(64
> steps majority clock) and then 64 steps round function(lfsr 64 bits wide).
> Afterwards the result is xor-ed and if it is a distinguished point the
> result is the first valid chain point and the whole process is repeated
> until we have 32 valid chain points. If it is not a distinguished point then
> which part do we repeat( for 64 times) until we find one? The a51 part or
> the round function part?

The 64 bits of the round function are not computed at each chain link,
because that would be too expensive. instead they are generated by the
cpu during startup and copied to the DRAM of the GPU.

Apart from that, your understanding of the procedure isn't correct:
If the (64bit keystream) XOR (64bit round function) is not a distinguished
point, then the result of the XOR is used as the contents of A5/1 registers
R1,R2,R3 (which happen to be 64 bits in size, so there is no reduction
function) and the the loop is started from the beginning. the loop is the
following:

repeat
        keystream = generate 64 bits of A5/1 output
        result = keystream XOR round function
        if 15 bits of result == 0
                switch to next round function
        R3 = result[0:22]; R2 = result[23:44]; R1 = result[45:63]
until done (which means round 32 reached)

> 
> 2) What is considered distinguished point?The lsb 15 bits or any consecutive
> 15 bits have to be all zero ?

the first 15 bits of keystream are checked.
You could take any string of 15 bits if you wanted to (but only a single
string) and you would get a length of 2^15 on average. Using the first
15 bits is the simplest implementation.
If you allowed the 15 bits to be anywhere, you will get an average
length of (64-15)/2^15 i suppose.

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

Reply via email to