[dpdk-dev] [PATCH 1/4] eventdev: introduce event driven programming model

Bruce Richardson bruce.richardson at intel.com
Thu Nov 24 17:24:11 CET 2016

On Fri, Nov 18, 2016 at 11:14:59AM +0530, Jerin Jacob wrote:
> In a polling model, lcores poll ethdev ports and associated
> rx queues directly to look for packet. In an event driven model,
> by contrast, lcores call the scheduler that selects packets for
> them based on programmer-specified criteria. Eventdev library
> adds support for event driven programming model, which offer
> applications automatic multicore scaling, dynamic load balancing,
> pipelining, packet ingress order maintenance and
> synchronization services to simplify application packet processing.
> By introducing event driven programming model, DPDK can support
> both polling and event driven programming models for packet processing,
> and applications are free to choose whatever model
> (or combination of the two) that best suits their needs.
> Signed-off-by: Jerin Jacob <jerin.jacob at caviumnetworks.com>
> ---

Hi Jerin,

Thanks for the patchset. A few minor comments in general on the API that
we found from working with it (thus far - more may follow :-) ).

1. Priorities: priorities are used in a number of places in the API, but
   all are uint8_t types and have their own MAX/NORMAL/MIN values. I think
   it would be simpler for the user just to have one priority type in the
   library, and use that everywhere. I suggest using RTE_EVENT_PRIORITY_*
   and drop the separate defines for SERVICE_PRIORITY, and QUEUE_PRIORITY
   etc. Ideally, I'd see things like this converted to enums too, rather
   than defines, but I'm not sure it's possible in this case.

2. Functions for config and setup can have their structure parameter
   types as const as they don't/shouldn't change the values internally.
   So add "const" to parameters to:

3. in event schedule() function, the dev->schedule() function needs the
   dev instance pointer passed in as parameter.

4. The event op values and the event type values would be better as
   enums rather than as a set of #defines.


More information about the dev mailing list