On Wed, Dec 14, 2011, Marius Peschke wrote: > Marius Peschke > > Sadly I had to experience that my MD5-speed dropped by roughly 20-25% again > without using assembler optimization.(Measurements made by "openssl speed > md5") >
Looking at how we measure speeds in apps/speed.c the results may not be realistic: it would depend on how the application uses digest contexts. In speed.c we call EVP_Digest() on each interation. This will initialise a EVP_MD_CTX, use it, then clean it. That adds an extra initialisation and clean up overhead to each operation: which will be more significant for smaller blocks. A more efficient way is to instead reuse the same EVP_MD_CTX for all operations. That way some initialisation and cleanup overheads only occur on the initial operation and the final cleanup respectively. I can't recall if we added some additional operations to initialisation and cleanup which would account for the difference in 1.0.0x. Steve. -- Dr Stephen N. Henson. OpenSSL project core developer. Commercial tech support now available see: http://www.openssl.org ______________________________________________________________________ OpenSSL Project http://www.openssl.org User Support Mailing List openssl-users@openssl.org Automated List Manager majord...@openssl.org
