Re: DES Key Decryption Time

From: Valdis.Kletnieks@vt.edu
Date: 04/09/03

  • Next message: Slawek: "Re: DES Key Decryption Time" 
    To: Ali Saifullah Khan <saifullah@gamebox.net>
From: Valdis.Kletnieks@vt.edu
Date: Wed, 09 Apr 2003 14:09:42 -0400

    On Wed, 09 Apr 2003 22:17:42 +0500, Ali Saifullah Khan <saifullah@gamebox.net> said:

    > What is the approx. number of clock cycles required to decrypt a 56-bit DES
    > key / signature ?

    In general, difficult to predict.

    Assuming a really smart hardware implementation, you can probably get it
    down to around 32 cycles (assuming a PAL to do the S-box and other stuff),
    doing each of 16 rounds in 2 cycles. A really good software implementation
    will need more, and a stupid implementation will require even more. (I'm
    handwaving about key setup here...)

    In addition, there's key initialization issues - very rarely is DES used to
    encrypt only one block. So you usually take a big hit in setting up the
    keys on the first block, and then subsequent blocks take a much smaller hit.

    Also, note that "clock cycles" is *always* a total crock when trying to
    compare across different architectures - the *SAME* instruction can take
    a different number of cycles on a Pentium-4, a Pentium-3, a Crusoe chip - and
    those are all the same software architecture. Comparing it to cycle count
    on a MIPS or Power3 chipset would be a total loss. Then add in issues like
    instruction parallelism and cache structure.....

  • Next message: Slawek: "Re: DES Key Decryption Time"