Re: [Iperf-users] When will real UDP, connectionless (sendto(), recvfrom(), broadcast) be iperf'd?

2017-01-13 Thread Bob McMahon
Which version of iperf did you modify?

Bob

On Fri, Jan 13, 2017 at 10:27 AM, Hoshall, Paul <hos...@textronsystems.com>
wrote:

> No versions of iperf I’ve seen use sendto() for UDP…..
>
>
>
> And Yes….I modified iperf and took measurements…..
>
>
>
> After my UDP application could not achieve the iperf UDP throughput….
>
> which subsequently is really only UDP by the protocol ID in the IP header….
>
> All other protocol layers are connection-oriented…..socket
> connect/listen/etc…
>
>
>
> *From:* Bob McMahon [mailto:bob.mcma...@broadcom.com]
> *Sent:* Friday, January 13, 2017 1:15 PM
> *To:* Hoshall, Paul
> *Cc:* iperf-users@lists.sourceforge.net; Gibbs, James
> *Subject:* Re: [Iperf-users] When will real UDP, connectionless
> (sendto(), recvfrom(), broadcast) be iperf'd?
>
>
>
> Which version of iperf are you using?   Do you have any measurements for
> UDP using sendto() vs send() with bind() and connect()?  I haven't seen any
> performance difference on my systems but haven't really spent much time
> profile it either.
>
>
>
> http://man7.org/linux/man-pages/man7/udp.7.html
>
>
>
> Bob
>
>
>
> On Fri, Jan 13, 2017 at 5:52 AM, Hoshall, Paul <hos...@textronsystems.com>
> wrote:
>
> Iperf does a connect with the UDP socket first….then sends() to a specific
> address
>
>
>
> Sendto ( broadcast ) is way slower than send()…..in Linux and Windows
>
>
>
> *From:* Bob McMahon [mailto:bob.mcma...@broadcom.com]
> *Sent:* Thursday, January 12, 2017 5:48 PM
> *To:* Hoshall, Paul
> *Cc:* iperf-users@lists.sourceforge.net; Gibbs, James
> *Subject:* Re: [Iperf-users] When will real UDP, connectionless
> (sendto(), recvfrom(), broadcast) be iperf'd?
>
>
>
> Not sure why this is needed.  I think a UDP socket with bind() using
> send() is equivalent to sendto() with respect to both being connectionless
> at the network layer.  Are you trying to test system or OS calls?  If not,
> I think a bind for UDP should be fine.
>
>
>
> Bob
>
>
>
>
>
>
>
--
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today. http://sdm.link/xeonphi___
Iperf-users mailing list
Iperf-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/iperf-users


Re: [Iperf-users] When will real UDP, connectionless (sendto(), recvfrom(), broadcast) be iperf'd?

2017-01-13 Thread Hoshall, Paul
Iperf does a connect with the UDP socket first….then sends() to a specific 
address

Sendto ( broadcast ) is way slower than send()…..in Linux and Windows

From: Bob McMahon [mailto:bob.mcma...@broadcom.com]
Sent: Thursday, January 12, 2017 5:48 PM
To: Hoshall, Paul
Cc: iperf-users@lists.sourceforge.net; Gibbs, James
Subject: Re: [Iperf-users] When will real UDP, connectionless (sendto(), 
recvfrom(), broadcast) be iperf'd?

Not sure why this is needed.  I think a UDP socket with bind() using send() is 
equivalent to sendto() with respect to both being connectionless at the network 
layer.  Are you trying to test system or OS calls?  If not, I think a bind for 
UDP should be fine.

Bob


--
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today. http://sdm.link/xeonphi___
Iperf-users mailing list
Iperf-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/iperf-users


Re: [Iperf-users] When will real UDP, connectionless (sendto(), recvfrom(), broadcast) be iperf'd?

2017-01-12 Thread Bob McMahon
Not sure why this is needed.  I think a UDP socket with bind() using send()
is equivalent to sendto() with respect to both being connectionless at the
network layer.  Are you trying to test system or OS calls?  If not, I think
a bind for UDP should be fine.

Bob



On Thu, Jan 12, 2017 at 2:06 PM, Hoshall, Paul 
wrote:

>
>
> 
> --
> Developer Access Program for Intel Xeon Phi Processors
> Access to Intel Xeon Phi processor-based developer platforms.
> With one year of Intel Parallel Studio XE.
> Training and support from Colfax.
> Order your platform today. http://sdm.link/xeonphi
> ___
> Iperf-users mailing list
> Iperf-users@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/iperf-users
>
>
--
Developer Access Program for Intel Xeon Phi Processors
Access to Intel Xeon Phi processor-based developer platforms.
With one year of Intel Parallel Studio XE.
Training and support from Colfax.
Order your platform today. http://sdm.link/xeonphi___
Iperf-users mailing list
Iperf-users@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/iperf-users