JT -- On Fri, Jul 26, 2013 at 11:25 PM, JT Olds <he...@jtolds.com> wrote:
> So, yeah, openssl speed -evp aes-128-cbc definitely shows a huge > improvement with cryptodev. It's huge orders of magnitude faster with > cryptodev.ko inserted into the kernel than without, and it's actually > about twice as fast as AF_ALG (which is actually what we've been using > for the last few months, since I couldn't get cryptodev to work > without this and similar issues). > > I haven't ever run the speed command without -evp though. Should I? Probably not. At the time, I was still trying to figure out what was what. Nikos has indicated that omitting the "-evp" parameter will force the fallback to software implementation, so it's almost certainly the wrong answer here. What you *can* play with, though, is the "-elapsed" switch to the "openssl speed" command: that changes the calculation from CPU time to wall-clock times, and gives you the true throughput of your device. E.g., where CPU clock would imply that my device can do 670MB/s, "-elapsed" gave the true value of about 33MB/s. (Which is still a very nice 5x speedup over the 6MB/s provided by software impl on same device, but not the 100x improvement the cpu-time would imply.) > Interesting. Yeah, I'm a little confused because it seems like my > platform (Kirkwood) is pretty heavily used with cryptodev, so I'm sure > I'm probably doing something wrong. Good luck. :-/ Best regards, Anthony Foiani _______________________________________________ Cryptodev-linux-devel mailing list Cryptodev-linux-devel@gna.org https://mail.gna.org/listinfo/cryptodev-linux-devel
