From: Jerin Jacob <jerin.jacob@caviumnetworks.com>
To: Bruce Richardson <bruce.richardson@intel.com>
Cc: Thomas Monjalon <thomas.monjalon@6wind.com>, <dev@dpdk.org>,
<harry.van.haaren@intel.com>, <hemant.agrawal@nxp.com>,
<gage.eads@intel.com>
Subject: Re: [dpdk-dev] [PATCH 1/4] eventdev: introduce event driven programming model
Date: Tue, 29 Nov 2016 09:31:42 +0530 [thread overview]
Message-ID: <20161129040141.GA11674@svelivela-lt.caveonetworks.com> (raw)
In-Reply-To: <20161128091610.GB168972@bricha3-MOBL3.ger.corp.intel.com>
On Mon, Nov 28, 2016 at 09:16:10AM +0000, Bruce Richardson wrote:
> On Sat, Nov 26, 2016 at 08:24:55AM +0530, Jerin Jacob wrote:
> > On Fri, Nov 25, 2016 at 11:00:53AM +0000, Bruce Richardson wrote:
> > > On Fri, Nov 25, 2016 at 05:53:34AM +0530, Jerin Jacob wrote:
> > > > On Thu, Nov 24, 2016 at 04:35:56PM +0100, Thomas Monjalon wrote:
> > > > > 2016-11-24 07:29, Jerin Jacob:
> > > > > > On Wed, Nov 23, 2016 at 07:39:09PM +0100, Thomas Monjalon wrote:
> > > > > > > 2016-11-18 11:14, Jerin Jacob:
> > > > > > > > +Eventdev API - EXPERIMENTAL
> > > > > > > > +M: Jerin Jacob <jerin.jacob@caviumnetworks.com>
> > > > > > > > +F: lib/librte_eventdev/
> > > > > > >
> > > >
> > > > I don't think there is any portability issue here, I can explain.
> > > >
> > > > The application level, we have two more use case to deal with non burst
> > > > variant
> > > >
> > > > - latency critical work
> > > > - on dequeue, if application wants to deal with only one flow(i.e to
> > > > avoid processing two different application flows to avoid cache trashing)
> > > >
> > > > Selection of the burst variants will be based on
> > > > rte_event_dev_info_get() and rte_event_dev_configure()(see, max_event_port_dequeue_depth,
> > > > max_event_port_enqueue_depth, nb_event_port_dequeue_depth, nb_event_port_enqueue_depth )
> > > > So I don't think their is portability issue here and I don't want to waste my
> > > > CPU cycles on the for loop if application known to be working with non
> > > > bursts variant like below
> > > >
> > >
> > > If the application is known to be working on non-burst varients, then
> > > they always request a burst-size of 1, and skip the loop completely.
> > > There is no extra performance hit in that case in either the app or the
> > > driver (since the non-burst driver always returns 1, irrespective of the
> > > number requested).
> >
> > Hmm. I am afraid, There is.
> > On the app side, the const "1" can not be optimized by the compiler as
> > on downside it is function pointer based driver interface
> > On the driver side, the implementation would be for loop based instead
> > of plain access.
> > (compiler never can see the const "1" in driver interface)
> >
> > We are planning to implement burst mode as kind of emulation mode and
> > have a different scheme for burst and nonburst. The similar approach we have
> > taken in introducing rte_event_schedule() and split the responsibility so
> > that SW driver can work without additional performance overhead and neat
> > driver interface.
> >
> > If you are concerned about the usability part and regression on the SW
> > driver, then it's not the case, application will use nonburst variant only if
> > dequeue_depth == 1 and/or explicit case where latency matters.
> >
> > On the portability side, we support both case and application if written based
> > on dequeue_depth it will perform well in both implementations.IMO, There is
> > no another shortcut for performance optimized application running on different
> > set of model.I think it is not an issue as, in event model as each cores
> > identical and main loop can be changed based on dequeue_depth
> > if needs performance(anyway mainloop will be function pointer based).
> >
>
> Ok, I think I see your point now. Here is an alternative suggestion.
>
> 1. Keep the single user API.
> 2. Have both single and burst function pointers in the driver
> 3. Call appropriately in the eventdev layer based on parameters. For
> example:
>
> rte_event_dequeue_burst(..., int num)
> {
> if (num == 1 && single_dequeue_fn != NULL)
> return single_dequeue_fn(...);
> return burst_dequeue_fn(...);
> }
>
> This way drivers can optionally special-case the single dequeue case -
> the function pointer check will definitely be predictable in HW making
> that a near-zero-cost check - while not forcing all drivers to do so.
> It also reduces the public API surface, and gives us a single enqueue
> and dequeue function.
The alternative suggestion looks good to me. Yes, it makes sense to reduces the
public API interface if possible.
Regarding the implementation, I thought to have a bit approach like below
to reduce the cost of additional AND operation.(with const "1", compiler
can choose with correct one with out any overhead)
rte_event_dequeue_burst(..., int num)
{
if (num == 1)
return single_dequeue_fn(...);
return burst_dequeue_fn(...);
}
"single_dequeue_fn" populated from the driver layer.
In the absence of populating the "single_dequeue_fn" from the driver layer,
The common code can create the single_dequeue_fn using driver
provided "burst_dequeue_fn"
something like
generic_single_dequeue_fn(dev){
{
dev->burst_dequeue_fn(..,1);
}
Any concerns?
>
> /Bruce
>
next prev parent reply other threads:[~2016-11-29 4:03 UTC|newest]
Thread overview: 109+ messages / expand[flat|nested] mbox.gz Atom feed top
2016-11-18 5:44 [dpdk-dev] [PATCH 0/4] libeventdev API and northbound implementation Jerin Jacob
2016-11-18 5:44 ` [dpdk-dev] [PATCH 1/4] eventdev: introduce event driven programming model Jerin Jacob
2016-11-23 18:39 ` Thomas Monjalon
2016-11-24 1:59 ` Jerin Jacob
2016-11-24 12:26 ` Bruce Richardson
2016-11-24 15:35 ` Thomas Monjalon
2016-11-25 0:23 ` Jerin Jacob
2016-11-25 11:00 ` Bruce Richardson
2016-11-25 13:09 ` Thomas Monjalon
2016-11-26 0:57 ` Jerin Jacob
2016-11-28 9:10 ` Bruce Richardson
2016-11-26 2:54 ` Jerin Jacob
2016-11-28 9:16 ` Bruce Richardson
2016-11-28 11:30 ` Thomas Monjalon
2016-11-29 4:01 ` Jerin Jacob [this message]
2016-11-29 10:00 ` Bruce Richardson
2016-11-25 11:59 ` Van Haaren, Harry
2016-11-25 12:09 ` Richardson, Bruce
2016-11-24 16:24 ` Bruce Richardson
2016-11-24 19:30 ` Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 0/6] libeventdev API and northbound implementation Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 1/6] eventdev: introduce event driven programming model Jerin Jacob
2016-12-06 16:51 ` Bruce Richardson
2016-12-07 18:53 ` Jerin Jacob
2016-12-08 9:30 ` Bruce Richardson
2016-12-08 20:41 ` Jerin Jacob
2016-12-09 15:11 ` Bruce Richardson
2016-12-14 6:55 ` Jerin Jacob
2016-12-07 10:57 ` Van Haaren, Harry
2016-12-08 1:24 ` Jerin Jacob
2016-12-08 11:02 ` Van Haaren, Harry
2016-12-14 13:13 ` Jerin Jacob
2016-12-14 15:15 ` Bruce Richardson
2016-12-15 16:54 ` Van Haaren, Harry
2016-12-07 11:12 ` Bruce Richardson
2016-12-08 1:48 ` Jerin Jacob
2016-12-08 9:57 ` Bruce Richardson
2016-12-14 6:40 ` Jerin Jacob
2016-12-14 15:19 ` Bruce Richardson
2016-12-15 13:39 ` Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 2/6] eventdev: define southbound driver interface Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 3/6] eventdev: implement the northbound APIs Jerin Jacob
2016-12-06 17:17 ` Bruce Richardson
2016-12-07 17:02 ` Jerin Jacob
2016-12-08 9:59 ` Bruce Richardson
2016-12-14 6:28 ` Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 4/6] eventdev: implement PMD registration functions Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 5/6] event/skeleton: add skeleton eventdev driver Jerin Jacob
2016-12-06 3:52 ` [dpdk-dev] [PATCH v2 6/6] app/test: unit test case for eventdev APIs Jerin Jacob
2016-12-06 16:46 ` [dpdk-dev] [PATCH v2 0/6] libeventdev API and northbound implementation Bruce Richardson
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 " Jerin Jacob
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 1/6] eventdev: introduce event driven programming model Jerin Jacob
2017-01-25 16:32 ` Eads, Gage
2017-01-25 16:36 ` Richardson, Bruce
2017-01-25 16:53 ` Eads, Gage
2017-01-25 22:36 ` Eads, Gage
2017-01-26 9:39 ` Jerin Jacob
2017-01-26 20:39 ` Eads, Gage
2017-01-27 10:03 ` Bruce Richardson
2017-01-30 10:42 ` Jerin Jacob
2017-02-02 11:18 ` Nipun Gupta
2017-02-02 14:09 ` Jerin Jacob
2017-02-03 6:38 ` Nipun Gupta
2017-02-03 10:58 ` Hemant Agrawal
2017-02-07 4:59 ` Jerin Jacob
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 2/6] eventdev: define southbound driver interface Jerin Jacob
2017-02-02 11:19 ` Nipun Gupta
2017-02-02 11:34 ` Bruce Richardson
2017-02-02 12:53 ` Nipun Gupta
2017-02-02 13:58 ` Bruce Richardson
2017-02-03 5:59 ` Nipun Gupta
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 3/6] eventdev: implement the northbound APIs Jerin Jacob
2017-02-02 11:19 ` Nipun Gupta
2017-02-02 14:32 ` Jerin Jacob
2017-02-03 6:59 ` Nipun Gupta
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 4/6] eventdev: implement PMD registration functions Jerin Jacob
2017-02-02 11:20 ` Nipun Gupta
2017-02-05 13:04 ` Jerin Jacob
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 5/6] event/skeleton: add skeleton eventdev driver Jerin Jacob
2016-12-21 9:25 ` [dpdk-dev] [PATCH v4 6/6] app/test: unit test case for eventdev APIs Jerin Jacob
2016-11-18 5:45 ` [dpdk-dev] [PATCH 2/4] eventdev: implement the northbound APIs Jerin Jacob
2016-11-21 17:45 ` Eads, Gage
2016-11-21 19:13 ` Jerin Jacob
2016-11-21 19:31 ` Jerin Jacob
2016-11-22 15:15 ` Eads, Gage
2016-11-22 18:19 ` Jerin Jacob
2016-11-22 19:43 ` Eads, Gage
2016-11-22 20:00 ` Jerin Jacob
2016-11-22 22:48 ` Eads, Gage
2016-11-22 23:43 ` Jerin Jacob
2016-11-28 15:53 ` Eads, Gage
2016-11-29 2:01 ` Jerin Jacob
2016-11-29 3:43 ` Jerin Jacob
2016-11-29 5:46 ` Eads, Gage
2016-11-23 9:57 ` Bruce Richardson
2016-11-23 19:18 ` Thomas Monjalon
2016-11-25 4:17 ` Jerin Jacob
2016-11-25 9:55 ` Richardson, Bruce
2016-11-25 23:08 ` Jerin Jacob
2016-11-18 5:45 ` [dpdk-dev] [PATCH 3/4] event/skeleton: add skeleton eventdev driver Jerin Jacob
2016-11-18 5:45 ` [dpdk-dev] [PATCH 4/4] app/test: unit test case for eventdev APIs Jerin Jacob
2016-11-18 15:25 ` [dpdk-dev] [PATCH 0/4] libeventdev API and northbound implementation Bruce Richardson
2016-11-18 16:04 ` Bruce Richardson
2016-11-18 19:27 ` Jerin Jacob
2016-11-21 9:40 ` Thomas Monjalon
2016-11-21 9:57 ` Bruce Richardson
2016-11-22 0:11 ` Thomas Monjalon
2016-11-22 2:00 ` Yuanhan Liu
2016-11-22 9:05 ` Shreyansh Jain
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20161129040141.GA11674@svelivela-lt.caveonetworks.com \
--to=jerin.jacob@caviumnetworks.com \
--cc=bruce.richardson@intel.com \
--cc=dev@dpdk.org \
--cc=gage.eads@intel.com \
--cc=harry.van.haaren@intel.com \
--cc=hemant.agrawal@nxp.com \
--cc=thomas.monjalon@6wind.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).