From: "Morten Brørup" <mb@smartsharesystems.com>
To: "Ananyev, Konstantin" <konstantin.ananyev@intel.com>,
"Van Haaren, Harry" <harry.van.haaren@intel.com>,
"Stephen Hemminger" <stephen@networkplumber.org>
Cc: <olivier.matz@6wind.com>, <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH] mbuf: add bulk free function
Date: Sun, 15 Sep 2019 11:07:08 +0200 [thread overview]
Message-ID: <98CBD80474FA8B44BF855DF32C47DC35B42A9C@smartserver.smartshare.dk> (raw)
In-Reply-To: <2601191342CEEE43887BDE71AB9772580191962C86@irsmsx105.ger.corp.intel.com>
> -----Original Message-----
> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Ananyev, Konstantin
> Sent: Wednesday, September 11, 2019 1:42 PM
> To: Van Haaren, Harry; Stephen Hemminger; Morten Brørup
> Cc: olivier.matz@6wind.com; dev@dpdk.org
> Subject: Re: [dpdk-dev] [PATCH] mbuf: add bulk free function
>
>
>
> > -----Original Message-----
> > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Van Haaren, Harry
> > Sent: Wednesday, September 11, 2019 12:30 PM
> > To: Stephen Hemminger <stephen@networkplumber.org>; Morten Brørup
> <mb@smartsharesystems.com>
> > Cc: olivier.matz@6wind.com; dev@dpdk.org
> > Subject: Re: [dpdk-dev] [PATCH] mbuf: add bulk free function
> >
> > > -----Original Message-----
> > > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Stephen Hemminger
> > > Sent: Wednesday, September 11, 2019 12:19 PM
> > > To: Morten Brørup <mb@smartsharesystems.com>
> > > Cc: olivier.matz@6wind.com; dev@dpdk.org
> > > Subject: Re: [dpdk-dev] [PATCH] mbuf: add bulk free function
> > >
> > > On Wed, 11 Sep 2019 09:19:08 +0000
> > > Morten Brørup <mb@smartsharesystems.com> wrote:
> > >
> > > > Add function for freeing a bulk of mbufs.
> > > >
> > > > Signed-off-by: Morten Brørup <mb@smartsharesystems.com>
> > > > ---
> > > > lib/librte_mbuf/rte_mbuf.h | 17 +++++++++++++++++
> > > > 1 file changed, 17 insertions(+)
> > > >
> > > > diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
> > > > index 98225ec80..f2e174da1 100644
> > > > --- a/lib/librte_mbuf/rte_mbuf.h
> > > > +++ b/lib/librte_mbuf/rte_mbuf.h
> > > > @@ -1907,6 +1907,23 @@ static inline void rte_pktmbuf_free(struct
> rte_mbuf
> > > *m)
> > > > }
> > > > }
> > > >
> > > > +/**
> > > > + * Free a bulk of mbufs back into their original mempool.
> > > > + *
> > > > + * @param mbufs
> > > > + * Array of pointers to mbufs
> > > > + * @param count
> > > > + * Array size
> > > > + */
> > > > +static inline void
> > > > +rte_pktmbuf_free_bulk(struct rte_mbuf **mbufs, unsigned count)
> > > > +{
> > > > + unsigned idx = 0;
> > > > +
> > > > + for (idx = 0; idx < count; idx++)
> > > > + rte_pktmbuf_free(mbufs[idx]);
> > > > +}
> > > > +
> > >
> > > You can optimize this to use mempool bulk put operation.
> >
> > I believe there's a nuance here - not all mbufs may come from the same
> mempool.
> > The for() approach will free each to its "home" mempool.
> > The bulk() approach may return mbufs to pools they didn't originate from.
> >
> > For performance reasons it would be nice if they did, but we (in the DPDK
> library)
> > should not blindly assume that.
>
> I suppose Stephen is aware of that and suggests something similar to
> What many PMDs are already doing. Let say in ixgbe:
> static __rte_always_inline int
> ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
> {
> ....
> for (i = 0; i < txq->tx_rs_thresh; ++i, ++txep) {
> /* free buffers one at a time */
> m = rte_pktmbuf_prefree_seg(txep->mbuf);
> txep->mbuf = NULL;
>
> if (unlikely(m == NULL))
> continue;
>
> if (nb_free >= RTE_IXGBE_TX_MAX_FREE_BUF_SZ ||
> (nb_free > 0 && m->pool != free[0]->pool)) {
> rte_mempool_put_bulk(free[0]->pool,
> (void **)free, nb_free);
> nb_free = 0;
> }
>
> free[nb_free++] = m;
> }
> }
>
> Of course generic function will also need to go through all segments in
> each packet.
>
Thank you for the clarifying example!
It looks like this optimization behaves as if RTE_LIBRTE_MBUF_DEBUG is not set.
In that case, the existing rte_pktmbuf_free() could be optimized similarly to free multiple segments in bulk. (Except that the use of likely/unlikely in the mbuf library heavily favors single-segment mbufs, so such an optimization would go against this favoritism.)
And yes, I see how rte_pktmbuf_free_bulk() could be optimized similarly. But I don't think it's appropriate for mbuf library functions to assume that RTE_LIBRTE_MBUF_DEBUG is not set. Although it could be implemented both ways, controlled by #ifdef RTE_LIBRTE_MBUF_DEBUG / #else / #endif.
> > We could consider adding a 2nd functions,
> rte_pktmbuf_free_bulk_to_single_mempool()
> > or some better descriptive name.
>
> Probably a good idea too.
> Konstantin
>
Feature creep. I prefer the generic function only.
Med venlig hilsen / kind regards
- Morten Brørup
next prev parent reply other threads:[~2019-09-15 9:07 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-09-11 9:19 Morten Brørup
2019-09-11 11:18 ` Stephen Hemminger
2019-09-11 11:29 ` Van Haaren, Harry
2019-09-11 11:41 ` Ananyev, Konstantin
2019-09-15 9:07 ` Morten Brørup [this message]
2019-09-15 9:24 ` Morten Brørup
2019-09-11 11:33 ` Olivier Matz
2019-09-11 11:39 ` Stephen Hemminger
2019-09-11 12:14 ` Morten Brørup
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=98CBD80474FA8B44BF855DF32C47DC35B42A9C@smartserver.smartshare.dk \
--to=mb@smartsharesystems.com \
--cc=dev@dpdk.org \
--cc=harry.van.haaren@intel.com \
--cc=konstantin.ananyev@intel.com \
--cc=olivier.matz@6wind.com \
--cc=stephen@networkplumber.org \
/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).