From: Honnappa Nagarahalli <Honnappa.Nagarahalli@arm.com>
To: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Cc: "bruce.richardson@intel.com" <bruce.richardson@intel.com>,
"pablo.de.lara.guarch@intel.com" <pablo.de.lara.guarch@intel.com>,
"dev@dpdk.org" <dev@dpdk.org>,
"yipeng1.wang@intel.com" <yipeng1.wang@intel.com>,
Dharmik Thakkar <Dharmik.Thakkar@arm.com>,
"Gavin Hu (Arm Technology China)" <Gavin.Hu@arm.com>,
nd <nd@arm.com>, "thomas@monjalon.net" <thomas@monjalon.net>,
"ferruh.yigit@intel.com" <ferruh.yigit@intel.com>,
"hemant.agrawal@nxp.com" <hemant.agrawal@nxp.com>,
"chaozhu@linux.vnet.ibm.com" <chaozhu@linux.vnet.ibm.com>,
"Kapoor, Prasun" <Prasun.Kapoor@cavium.com>, nd <nd@arm.com>
Subject: Re: [dpdk-dev] [PATCH v7 4/5] hash: add lock-free read-write concurrency
Date: Fri, 9 Nov 2018 15:37:48 +0000 [thread overview]
Message-ID: <AM6PR08MB3672EDE2EA5DFF096816DF7E98C60@AM6PR08MB3672.eurprd08.prod.outlook.com> (raw)
In-Reply-To: <20181109092821.GB4934@jerin>
> > > > Agree. There are multiple micro-architectures in Arm eco-system.
> > > > We
> > > should establish few simple rules to make sure algorithms perform
> > > well on all the available platforms. I established few rules in VPP
> > > and they are working fine so far.
> > >
> > > Can you share that rules for everyone's benefit?
> > >
> > These are just few simple rules anyone can think of, but avoid the surprises.
> > We identified a owner for each platform (we have this already in DPDK,
> > even across platforms) Each patch submitted for Arm platforms is
> > marked with -2 (VPP uses Gerrit) Every platform owner tests on her/his
> platform. -2 will be removed only if it does not cause regression on any
> platforms. Platform owner helps out with optimization where required as they
> understand their micro-architecture best. I guess this is what is supposed to
> happen through the review process in DPDK. But making sure everyone tests it
> before it gets merged avoids the surprises.
>
> I think, The very same philosophy can be implemented with exiting mailing list
> method, if
>
> 1) Author Cc all the architecture maintainers and platform owners for any fast
> path change where it can introduce performance regression.
> 2) Author can CC the same list to request for performance check along with
> test command if area of performance regression known before.
>
> I agree with last minute surprises are bad for both Author and platform owner.
> I think, it can fixed by above scheme.
>
Makes sense to me.
> >
> > > > >
> > > > > IMO, This scheme won't work. I think, we are introducing such
> > > > > performance critical feature, we need to put under function
> > > > > pointer scheme so that if an application does not need such
> > > > > feature it can use
> > > plain loads.
> > > > >
> > > > IMO, we should do some more debugging before going into exploring
> > > > other
> > > options.
> > >
> > > Yes. But, how do we align with v18.11 release?
> > >
> > I think I have spent enough time optimizing the code. Please provide the
> feedback and I will work on completing the fix.
> >
> > However, if the new patch is not satisfactory enough, we need another plan.
>
> Based on the public release meeting held yesterday, RC3 date is on next
> Monday.
>
> I would suggest:
> - Send your exiting tested patch in mailing list for review. In my
> setup, The regression reduced to 5.7% from 24%
> - Extend the patch for hash bulk case as well and check
> NO_HASH_MULTI_LOOKUP
> as zero with EM_HASH_LOOKUP_COUNT value as 16 or 8 for arm64.
>
Thank you Jerin for testing. Unfortunately, when I looked at making the corresponding changes for the hash_add API, they look more involved. They may not be acceptable for RC3. I also need more time to spend on bulk case. For RC3, we will split the lookup path for RW-lock and Lock-free. I will add the rest in a follow up patch.
> >
> > You had mentioned about using function pointers. I suggest, we use the
> function pointer only for lookup function. Otherwise, it will be too much of
> code duplication.
> > When lock-free is not used, the function with no memory-orderings will be
> called. However, I am not sure about the function pointer overhead. But this
> will be a simple change.
>
> It may not be very simple change as we need to take care secondary process
> case as well, see struct rte_mempool::ops_index scheme.
>
Yes, I realized it while making changes. I have used a if statement with an existing lock-free flag. Will send out the patch soon.
> Since rte_hash_lookup() already NOT a inline function, so making it as inline
> and calling a function pointer inside may not attract much overhead. But we
> can tell only after testing(Which may be not possible for
> RC3)
>
> I think, in future it make sense to have function pointer scheme to avoid new
> APIs for different hash library and we can plugin other proven hash library like
> urcu based one etc to DPDK.
>
> https://lwn.net/Articles/573431/
next prev parent reply other threads:[~2018-11-09 15:37 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-10-26 5:37 [dpdk-dev] [PATCH v7 0/5] Address reader-writer concurrency in rte_hash Honnappa Nagarahalli
2018-10-26 5:37 ` [dpdk-dev] [PATCH v7 1/5] hash: separate multi-writer from rw-concurrency Honnappa Nagarahalli
2018-10-26 5:37 ` [dpdk-dev] [PATCH v7 2/5] hash: support do not free on delete Honnappa Nagarahalli
2018-10-26 5:37 ` [dpdk-dev] [PATCH v7 3/5] hash: fix key store element alignment Honnappa Nagarahalli
2018-10-26 5:37 ` [dpdk-dev] [PATCH v7 4/5] hash: add lock-free read-write concurrency Honnappa Nagarahalli
2018-11-03 11:52 ` Jerin Jacob
2018-11-03 15:40 ` Jerin Jacob
2018-11-06 6:07 ` Honnappa Nagarahalli
2018-11-06 9:10 ` Jerin Jacob
2018-11-06 9:13 ` Thomas Monjalon
2018-11-06 9:47 ` Jerin Jacob
2018-11-09 1:34 ` Honnappa Nagarahalli
2018-11-09 2:20 ` Honnappa Nagarahalli
2018-11-09 9:28 ` Jerin Jacob
2018-11-09 15:37 ` Honnappa Nagarahalli [this message]
2018-11-07 2:15 ` Wang, Yipeng1
2018-11-09 0:47 ` Honnappa Nagarahalli
2018-10-26 5:37 ` [dpdk-dev] [PATCH v7 5/5] test/hash: read-write lock-free concurrency test Honnappa Nagarahalli
2018-10-26 10:52 ` [dpdk-dev] [PATCH v7 0/5] Address reader-writer concurrency in rte_hash Thomas Monjalon
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=AM6PR08MB3672EDE2EA5DFF096816DF7E98C60@AM6PR08MB3672.eurprd08.prod.outlook.com \
--to=honnappa.nagarahalli@arm.com \
--cc=Dharmik.Thakkar@arm.com \
--cc=Gavin.Hu@arm.com \
--cc=Prasun.Kapoor@cavium.com \
--cc=bruce.richardson@intel.com \
--cc=chaozhu@linux.vnet.ibm.com \
--cc=dev@dpdk.org \
--cc=ferruh.yigit@intel.com \
--cc=hemant.agrawal@nxp.com \
--cc=jerin.jacob@caviumnetworks.com \
--cc=nd@arm.com \
--cc=pablo.de.lara.guarch@intel.com \
--cc=thomas@monjalon.net \
--cc=yipeng1.wang@intel.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).