DPDK patches and discussions
 help / color / mirror / Atom feed
From: Neil Horman <nhorman@tuxdriver.com>
To: "De Lara Guarch, Pablo" <pablo.de.lara.guarch@intel.com>
Cc: "dev@dpdk.org" <dev@dpdk.org>
Subject: Re: [dpdk-dev] [PATCH 0/3] New Thread Safe Hash Library
Date: Thu, 18 Sep 2014 12:09:20 -0400
Message-ID: <20140918160920.GK20389@hmsreliant.think-freely.org> (raw)
In-Reply-To: <E115CCD9D858EF4F90C690B0DCB4D89722616F62@IRSMSX108.ger.corp.intel.com>

On Thu, Sep 18, 2014 at 03:31:34PM +0000, De Lara Guarch, Pablo wrote:
> Hi Neil,
> 
> > -----Original Message-----
> > From: Neil Horman [mailto:nhorman@tuxdriver.com]
> > Sent: Thursday, September 18, 2014 1:21 PM
> > To: De Lara Guarch, Pablo
> > Cc: dev@dpdk.org
> > Subject: Re: [dpdk-dev] [PATCH 0/3] New Thread Safe Hash Library
> > 
> > On Thu, Sep 18, 2014 at 11:34:28AM +0100, Pablo de Lara wrote:
> > > This is an alternative hash implementation to the existing hash library.
> > > This patch set provides a thread safe hash implementation, it allows users
> > > to use multiple readers/writers working on a same hash table.
> > > Main differences between the previous and the new implementation are:
> > >
> > > - Multiple readers/writers can work on the same hash table,
> > >   whereas in the previous implementation writers could not work
> > >   on the table at the same time readers do.
> > > - Previous implementation returned an index to a table after a lookup.
> > >   This implementation returns 8-byte integers or pointers to external data.
> > > - Maximum entries to be looked up in bursts is 64, instead of 16.
> > > - Maximum key length has being increased to 128, instead of a maximum of
> > 64.
> > >
> > > Basic implementation:
> > >
> > > - A sparse table containing buckets (64-byte long) with hashes,
> > >   most of which are empty, and indexes to the second table.
> > > - A compact table containing keys for final matching,
> > >   plus data associated to them.
> > >
> > Thread safe hash tables seem to me like a configuration option rather than a
> > new
> > library.  Instead of creating a whole new library (with a new API and ABI to
> > maintain, why not just add thread safety as a configurable option to the
> > existing hash library.  That saves code space in the DPDK, and reduces
> > application complexity (as the same api is useable for thread safe and unsafe
> > hash tables)
> 
> Makes sense, but implementation has changed so much to add it directly into the existing library.
> At first, this was designed to be a replacement of the existing library, 
> but since API is a bit different from the old one, it was thought to leave it as an alternative,
>  so users are not forced to have to change their applications if they don't want to use thread safe hash tables.

What are you talking about?  The API calls between rte_hash and the new
rte_tshash are identical.  The only thing that differs are the names slightly
(rte_hash vs rte_tshash), and some of the elements of the internal data
structure, which really shouldn't be accessed by the application anyway (though
that does play into some of the ABI work we've started looking at).  It should
be pretty easy to modify the rte_hash library to optionally include thread
safety.  A flag in the config structure, a spinlock in the internal
representation, and you're home free.

Neil

  parent reply	other threads:[~2014-09-18 16:03 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2014-09-18 10:34 Pablo de Lara
2014-09-18 10:34 ` [dpdk-dev] [PATCH 1/3] eal: add const in prefetch functions Pablo de Lara
2014-09-18 10:34 ` [dpdk-dev] [PATCH 2/3] lib/librte_tshash: New Thread Safe Hash library for DPDK Pablo de Lara
2014-09-18 10:34 ` [dpdk-dev] [PATCH 3/3] app/test: Added unit tests for Thread Safe Hash library Pablo de Lara
2014-09-18 12:21 ` [dpdk-dev] [PATCH 0/3] New Thread Safe Hash Library Neil Horman
2014-09-18 15:31   ` De Lara Guarch, Pablo
2014-09-18 15:45     ` Thomas Monjalon
2014-09-18 16:09     ` Neil Horman [this message]
2014-09-18 22:30 ` Stephen Hemminger

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=20140918160920.GK20389@hmsreliant.think-freely.org \
    --to=nhorman@tuxdriver.com \
    --cc=dev@dpdk.org \
    --cc=pablo.de.lara.guarch@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

DPDK patches and discussions

This inbox may be cloned and mirrored by anyone:

	git clone --mirror https://inbox.dpdk.org/dev/0 dev/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dev dev/ https://inbox.dpdk.org/dev \
		dev@dpdk.org
	public-inbox-index dev

Example config snippet for mirrors.
Newsgroup available over NNTP:
	nntp://inbox.dpdk.org/inbox.dpdk.dev


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git