On Mon, Aug 20, 2012 at 1:28 PM, Stefan Fritsch <s...@sfritsch.de> wrote: > On Fri, 17 Aug 2012, Mike Belopuhov wrote: >> >> On Fri, Aug 17, 2012 at 1:17 PM, Stefan Fritsch <s...@sfritsch.de> wrote: >>> >>> On Thursday 16 August 2012, Mike Belopuhov wrote: >>>> >>>> could you please tell me if you're using tx interrupt or not? >>>> if yes, why do you need to have a watchdog code implement >>>> a txeof path and not the actual reset? >>> >>> >>> In normal operation, tx interrupts are not used and the sent >>> descriptors are cleaned up either by rx interrupts or by the watchdog. >>> The sole exception is when the tx queue is full. >> >> >> exactly, so please don't do it. use a tx interrupt for txeof and use a >> watchdog to reset a "controller". > > > Always using the tx interrupt decreases performance significantly. On my > test system (in MBytes/s): > > Sending UDP from 75 to 55 > Sending TCP from 34 to 25 >
Why? > I dont' think we want to do this. Perhaps you got your OACTIVE handling wrong and don't call a start routing from the appropriate places. Have you considered that?
