Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-21 Thread Honnappa Nagarahalli
> > > > Hi Jakub, > > I am trying to review this patch. I am having difficulty in > > understanding the implementation for the queue/ring, appreciate if you > > could help me understand the logic. > > 'ring' refers to a ring buffer holding packet descriptors. These descriptors > hold metadat

Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-21 Thread Ferruh Yigit
On 9/21/2020 1:21 PM, Jakub Grajciar -X (jgrajcia - PANTHEON TECH SRO at Cisco) wrote: Using 'rte_mb' to synchronize the shared ring head/tail between producer and consumer will stall the pipeline and damage performance on the weak memory model platforms, such like aarch64. Relax the expensive

Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-21 Thread Jakub Grajciar -X (jgrajcia - PANTHEON TECH SRO at Cisco)
> > > > > > > > Using 'rte_mb' to synchronize the shared ring head/tail between > > > > producer and consumer will stall the pipeline and damage > > > > performance on the weak memory model platforms, such like aarch64. > > > > > > > > Relax the expensive barrier with c11 atomic with explicit memor

Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-21 Thread Phil Yang
Jakub Grajciar -X (jgrajcia - PANTHEON TECH SRO at Cisco) writes: > > Subject: RE: [PATCH] net/memif: relax barrier for zero copy path > > > > Hi Jakub, > > I am trying to review this patch. I am having difficulty in > understanding > > the implementation for the queue/ring, appreciate if yo

Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-21 Thread Jakub Grajciar -X (jgrajcia - PANTHEON TECH SRO at Cisco)
Hi Honnappa, Inline comments... > -Original Message- > From: Honnappa Nagarahalli > Sent: Saturday, September 19, 2020 12:49 AM > To: Phil Yang ; Jakub Grajciar -X (jgrajcia - PANTHEON > TECH SRO at Cisco) ; dev@dpdk.org > Cc: Ruifeng Wang ; nd ; Honnappa > Nagarahalli ; nd > Subject: R

Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-18 Thread Honnappa Nagarahalli
Hi Jakub, I am trying to review this patch. I am having difficulty in understanding the implementation for the queue/ring, appreciate if you could help me understand the logic. 1) The S2M queues - are used to send packets from slave to master. My understanding is that, the slave thread

Re: [dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-18 Thread Ferruh Yigit
On 9/11/2020 6:38 AM, Phil Yang wrote: Using 'rte_mb' to synchronize the shared ring head/tail between producer and consumer will stall the pipeline and damage performance on the weak memory model platforms, such like aarch64. Relax the expensive barrier with c11 atomic with explicit memory orde

[dpdk-dev] [PATCH] net/memif: relax barrier for zero copy path

2020-09-10 Thread Phil Yang
Using 'rte_mb' to synchronize the shared ring head/tail between producer and consumer will stall the pipeline and damage performance on the weak memory model platforms, such like aarch64. Relax the expensive barrier with c11 atomic with explicit memory ordering can improve 3.6% performance on thro