Re: symbol catenation and montgomery

[phr-2000]

        What does "openssl speed" use for the exponent?  With 65537, GMP goes
        down to 0.33ms per modexp.  With 3, we're down to 0.11ms.

     65537 I'm pretty sure.

   In that case, the current GMP sources are 7 times faster than openssl
   on Athlon.  That sounds like an unlikely large edge for GMP.  I tried
   on a Alpha 21164 and got even greater difference.  I really think I am
   misreading the openssl output.  Here are the test runs again:...

                     sign    verify    sign/s verify/s
   rsa 1024 bits   0.0067s   0.0004s    148.9   2797.4

If it's doing 2797 encryptions/sec that's about 0.35 ms per modexp, or
about the same as GMP.  An RSA encryption is a full 1024-bit modexp,
though with a small exponent.  You can't use the CRT because you don't
know the secret factors.  The CRT is only for decryption.