Hi,
Device specific drivers operate without interrupts, so they are expected to have
a smaller jitter than generic driver. In practice there is no performance
difference when the control loop runs below 1 KHz.
regards
---
sebastien
On 01/06/2016 11:26 PM, John Hubbard wrote:
Hello all.
I would like to know what kinds of performance changes I might expect to see if
I switched from the generic NIC driver to a device specific driver (e1000e in my
case). With the generic NIC driver I am currently seeing ecrt_master_receive
calls averaging about 10us and ecrt_master_send times that are a few micro
seconds longer. I don't need exact numbers just a rough idea. Would using the
custom e1000e driver reduce times by an order of magnitude, 50%, 5%?
I would also like to know how sensitive the times are to PDO size? I've tried
registering fewer PDOs (i.e. ecrt_domain_reg_pdo_entry_list) and the difference
in transmit/receive times seems to be nearly unaffected. In addition I tried
reducing the number of devices attached to my EK-1100 coupler and that also had
little to no change in run times. Are those results as expected?
Thanks
--
-john
_______________________________________________
etherlab-users mailing list
[email protected]
http://lists.etherlab.org/mailman/listinfo/etherlab-users
