I have committed a change to cgd(4) that defers crypto operations to a parallel work queue. This should improve performance on multi-core systems.
On a system with Core i5-2520 (2.5GHz) with 2 Cores and 2 Threads/Core
I now get 150-180MByte/s reading and writing the raw device.
For comparison, 'openssl speed -multi 4 aes-128-cbc' returns an aggregated
rate of ~200MByte/s.
Feedback welcome.
--
--
Michael van Elst
Internet: [email protected]
"A potential Snark may lurk in every tree."
