[dpdk-dev] [PATCH] compressdev: add feature flag to specify where processing is done

[Trahe, Fiona]

Hi Shally,

> -----Original Message-----
> From: Verma, Shally [mailto:Shally.Verma at cavium.com]
> Sent: Monday, November 19, 2018 10:17 PM
> To: Trahe, Fiona <fiona.trahe at intel.com>; dev at dpdk.org
> Cc: akhil.goyal at nxp.com; Jozwiak, TomaszX <tomaszx.jozwiak at intel.com>; Gupta, Ashish
> <Ashish.Gupta at cavium.com>; Daly, Lee <lee.daly at intel.com>
> Subject: RE: [PATCH] compressdev: add feature flag to specify where processing is done
> 
> HI Fiona
> 
> >-----Original Message-----
> >From: Fiona Trahe <fiona.trahe at intel.com>
> >Sent: 20 November 2018 07:10
> >To: dev at dpdk.org
> >Cc: akhil.goyal at nxp.com; tomaszx.jozwiak at intel.com; Verma, Shally <Shally.Verma at cavium.com>;
> Gupta, Ashish
> ><Ashish.Gupta at cavium.com>; lee.daly at intel.com; fiona.trahe at intel.com
> >Subject: [PATCH] compressdev: add feature flag to specify where processing is done
> >
> >External Email
> >
> >A new device feature flag, RTE_COMPDEV_FF_SW_OP_DONE_IN_DEQUEUE
> >is added. A PMD which processes operations using a software
> >acceleration engine should set this if the bulk of the
> >processing is done during the dequeue. It should leave it
> >cleared if the bulk of the processing is done during the
> >enqueue (default).
> >An application may find this useful for tuning.
> 
> Help me understand a bit , so how does it work when:
> 1. PMD do ops processing only at enqueue , will not set this flag?
> 2. PMD do ops processing only at dequeue, then?
> 3. PMD can do op processing both at enqueue and dequeue and in such case,  can set this flag to indicate
> this support? So that app can choose?
[Fiona] Both sw PMDs already upstreamed do the bulk of the work on the enqueue.
So no need to set the flag.
It doesn't seem likely that PMDs will split the workload between enqueue and dequeue, if they do for some reason, thy can set whichever makes most sense.
It's like a hint, similar to the HW flag - to help an application to tune - applications don't have to use it if they don't want to.
I'm planning to send same patch for cryptodev - there two PMDs do all the work on dequeue, the rest do it on enqueue. Our Storage team asked for this as they changed some polling timings once they understood
where the work was done.

 
> if supported then app set this feature flag on PMD and PMD take it as a hint to know application
> preference of processing point or PMD always do bulk processing at chosen point?
[Fiona] No, it's the opposite - the PMD reports the capabilities to the appl.
> 
> And why specific to SW PMD, I see HW PMD can do use this feature flag?!
[Fiona] HW is not using CPU cycle as the work is offloaded, so I don't see how it can apply there.



> 
> Thanks
> Shally