> Enhancements to the assembler routines on UltraSparc. > Using the FPU gains about 30% in performance with a 1k RSA key.
The submission is being dismissed in favor of newly developed bn_mul_mont codepath, which exhibits uniform 2x(!) performance improvement for all key lenghts. One can argue that it's not exactly fair, because newly uploaded module target 64-bit ABI, while the code submission in question - 32-bit ABI. But the new sparcv9a-mount.pl was designed with 32-bit ABI in mind and will be adapted for 32-bit as time permits. See http://cvs.openssl.org/chngview?cn=14530 for further details. For reference, both above mentioned 2x and 30% improvement coefficients apply to USI&II processor cores. USIII benchmark results are hardly affected by either code [just few percents in various directions], but(!) the new bn_mul_mont code can be optimized further, estimatedly for further ~50%-ish improvement on either core. A. ______________________________________________________________________ OpenSSL Project http://www.openssl.org Development Mailing List [email protected] Automated List Manager [EMAIL PROTECTED]
