Solaris crypto group decides to donate AMD64 optimizations
For the hardcore nerds:
The Solaris crypto group at Sun has decided to donate optimizations assembly optimizations for the bignum library for the AMD64 instruction set. This code roughly doubles the RSA performance. On a v20z with opteron model 248 CPU at 2.2 GHz, using gcc and -O3 optimizations on Solaris 10, rsaperf goes from 585 ops/s for the pure C implementation, to 1186 ops/s with this code. This is for 1024 bit RSA keys, with rsaperf running single-threaded, ie. it is a per-CPU ops number.
Even more in the bug report