2) If I've to add crypto accelerator support in openssl for linux then which is better approach
a) I directly write an engine
b) I use engine written for OCF and I just write my module for OCF in
kernel
From my limited experience with OCF I remember a _significant_
performance penalty for each call, assumed because of the
penalty for transferring data to/from kernel-land. The only
time I saw a benefit was for the larger block sizes...
Here is a post I made a while back, it has some performance
statistics for OCF vs software on a Geode...
http://busybox.net/lists/buildroot/2007-August/004810.html
Basically, if you can do it by writing a real OpenSSL engine, my guess
is that it will probably be faster and a better course of action. That
said, I haven't run any tests on hardware supported by both OCF and
directly as an OpenSSL engine to provide any real knowledge there,
perhaps someone else could better answer your question.
-Brad
______________________________________________________________________
OpenSSL Project http://www.openssl.org
Development Mailing List openssl-dev@openssl.org
Automated List Manager [EMAIL PROTECTED]
