Re: FC4 on Dual Pentium III

From: Nico Kadel-Garcia (nkadel_at_comcast.net)
Date: 11/27/05

  • Next message: Nico Kadel-Garcia: "Re: Suse 9.0 slooowness"
    Date: Sun, 27 Nov 2005 08:13:20 -0500
    
    

    "Enrique Perez-Terron" <enrio@online.no> wrote in message
    news:op.s0u8pwgznxtvbs@apeiron.home.lan...
    > On Sat, 26 Nov 2005 20:47:41 +0100, Aaron Gray <ang.usenet@gmail.com>
    > wrote:
    >
    >> Hi,
    >>
    >> I am going to install Fedora Core 4 on a Dual Pentium III server.
    >>
    >> Do I have to rebuild the kernel for it or does the install do Dual CPU
    >> machines ?
    >
    > Go ahead and try; if it doesn't, by all means tell me.
    >
    > FC4 does have smp kernels in their installation media. It would be very
    > surprising (to me, at least) if the install still installs a uniprocessor
    > kernel.
    >
    > (I only have uniprocessors, so I have not tried)
    >
    > (smp = "symmetric multi-processor". "Symmetric" because all cpus can run
    > kernel code.)

    >From experience, it generally installs *BOTH*. The information about which
    should be the default kernel is stored in /etfc/sysconfig/kernel, and is
    used to set the default kernel when you do kernel updates. I'm personally a
    big believer in ripping out kernels you don't use or plan to use, such as
    ripping out the single-processor kernel when you have an SMP machine, just
    to lighten the load on your disk and your RPM database and help avoid
    confusion when booting.

    Do note: if you're running a single CPU kernel, there is *no way* short of
    installing a LinuxBIOS (which I don't recommend for amateurs) to detect that
    you actually have another CPU or hyper-threading available, short of
    rebooting with an SMP kernel and looking around in /proc/cpuinfo or using a
    tool that looks in /proc/cpuinfo.. There may be some deep kernel probing
    tools out there in user-land, but the only tools I've seen that might have
    been useful some day were never complete enough to even be compilable in
    most circumstances.

    It's a shame, really. There was some toolset from IBM I saw years ago, but
    the consultant I met working on it got dropped as part of the dot-bomb
    layoff years ago.


  • Next message: Nico Kadel-Garcia: "Re: Suse 9.0 slooowness"