Re: DEVICE_POLLING with SMP

From: Vlad Galu (dudu_at_diaspar.rdsnet.ro)
Date: 01/29/04

  • Next message: Kevin Day: "Re: DEVICE_POLLING with SMP" 
    Date: Thu, 29 Jan 2004 09:04:13 +0200
To: freebsd-net@freebsd.org

    veedee@c7.campus.utcluj.ro writes:

    |On Wed, Jan 28, 2004 at 01:25:13PM -0600, Kevin Day wrote:
    |>
    |> Has anyone made any headway with getting polling(4) to work with a
    |SMP > kernel? Last May this was discussed on here briefly with me,
    |Luigi and > Don Bowman, which seemed to indicate that the majority of
    |what needed > to be fixed to make this work would be some kind of
    |locking in > idle_loop to make sure only one CPU gets into the polling
    |code at once.>
    |> I've got a case where enabling SMP gives me a 20-30% performance
    |boost, > or if I turn polling on I get another 10-15% boost. I'd love
    |to see > what the two together could do.
    |
    |I was just about to ask the same question. We've just received 2 Athlon
    |MPs 2400+ and we're about to buy a gigabit Intel PRO/1000MT Dual for
    |one of our servers.
    |
    |It would be great if SMP could be combined with polling.

            I see no reason for it. Having to switch between multiple kernel
    threads to handle polling may bring too much overhead.

    |--
    || Radu Bogdan 'veedee' Rusu
    || NetSysAdm at campus dot utcluj dot ro
    || Personal gallery at http://rbrusu.com
    || ...mirroring FreeBSD and coffee
    |_______________________________________________
    |freebsd-net@freebsd.org mailing list
    |http://lists.freebsd.org/mailman/listinfo/freebsd-net
    |To unsubscribe, send any mail to "freebsd-net-unsubscribe@freebsd.org"
    | 

    ----
If it's there, and you can see it, it's real.
If it's not there, and you can see it, it's virtual.
If it's there, and you can't see it, it's transparent.
If it's not there, and you can't see it, you erased it.

  • Next message: Kevin Day: "Re: DEVICE_POLLING with SMP"

    Relevant Pages

    • Re: panic: lock (sleep mutex) inp not locked
      ... > got this panic today on an SMP system. ... > Crashdump and debug kernel are available for further investigation. ... I successfully used polling on my SMP machine a while back, ... #ifdef causing polling build to fail on SMP is removed. ...
      (freebsd-stable)
    • Re: decreasing interrupt CPU load
      ... > question suggests why it is disabled for SMP. ... It seems that polling ... > particular cpu? ...
      (freebsd-performance)
    • Re: Poor NFS server performance in 6.0 with SMP and mpsafenet=1
      ... >> I'm seeing incredibly poor performance when serving files from an SMP ... >> experimenting and have discovered that either removing SMP from the ... I'll have a play with both ULE and POLLING to see what ...
      (freebsd-current)
    • Re: DEVICE_POLLING with SMP
      ... > Has anyone made any headway with getting pollingto work with a SMP ... > idle_loop to make sure only one CPU gets into the polling code at once. ... > I've got a case where enabling SMP gives me a 20-30% performance boost, ... | NetSysAdm at campus dot utcluj dot ro ...
      (freebsd-net)