Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Ananyev, Konstantin
> -Original Message- > From: Thomas Monjalon > Sent: Wednesday, November 6, 2019 12:19 PM > To: Ananyev, Konstantin > Cc: techbo...@dpdk.org; Honnappa Nagarahalli ; > dev@dpdk.org; Zhang, Roy Fan > ; Doherty, Declan ; > akhil.go...@nxp.com; nd > Subject: Re: [dpdk-techboard] [RFC 0/

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Hemant Agrawal
> 06/11/2019 12:33, Ananyev, Konstantin: > > > > > > > > > Originally both SW and HW crypto PMDs use rte_crypot_op > > > > > > > based API to process the crypto workload asynchronously. > > > > > > > This way provides uniformity to both PMD types, but also > > > > > > > introduce unnecessary perf

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Thomas Monjalon
06/11/2019 12:33, Ananyev, Konstantin: > > > > > > > Originally both SW and HW crypto PMDs use rte_crypot_op based API to > > > > > > process the crypto workload asynchronously. This way provides > > > > > > uniformity to > > > > > > both PMD types, but also introduce unnecessary performance pena

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Ananyev, Konstantin
> > > > > Originally both SW and HW crypto PMDs use rte_crypot_op based API to > > > > > process the crypto workload asynchronously. This way provides > > > > > uniformity to > > > > > both PMD types, but also introduce unnecessary performance penalty to > > > > > SW > > > > > PMDs that have to

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Ananyev, Konstantin
> > > > Originally both SW and HW crypto PMDs use rte_crypot_op based API to > > > > process the crypto workload asynchronously. This way provides > > > > uniformity to > > > > both PMD types, but also introduce unnecessary performance penalty to SW > > > > PMDs that have to "simulate" HW async

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Thomas Monjalon
06/11/2019 10:35, Thomas Monjalon: > 06/11/2019 05:54, Honnappa Nagarahalli: > > > > > > > Originally both SW and HW crypto PMDs use rte_crypot_op based API to > > > process the crypto workload asynchronously. This way provides uniformity > > > to > > > both PMD types, but also introduce unneces

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-06 Thread Thomas Monjalon
06/11/2019 05:54, Honnappa Nagarahalli: > > > > Originally both SW and HW crypto PMDs use rte_crypot_op based API to > > process the crypto workload asynchronously. This way provides uniformity to > > both PMD types, but also introduce unnecessary performance penalty to SW > > PMDs that have to "

Re: [dpdk-dev] [dpdk-techboard] [RFC 0/4] cpu-crypto API choices

2019-11-05 Thread Honnappa Nagarahalli
> > Originally both SW and HW crypto PMDs use rte_crypot_op based API to > process the crypto workload asynchronously. This way provides uniformity to > both PMD types, but also introduce unnecessary performance penalty to SW > PMDs that have to "simulate" HW async behavior (crypto-ops > enqueue