slackware performance- processor type relation - Slackware

This is a discussion on slackware performance- processor type relation - Slackware ; I'm using slackware 10.2 .According to /boot/config the kernel is compiled for x486. My SBC board has a Pentium M processor .Should I expect a big performance increase ,when I compile the kernel for Pentium M? Does anybody have benchmark ...

+ Reply to Thread
Results 1 to 10 of 10

Thread: slackware performance- processor type relation

  1. slackware performance- processor type relation

    I'm using slackware 10.2 .According to /boot/config the kernel is
    compiled for x486.
    My SBC board has a Pentium M processor .Should I expect a big
    performance increase ,when I compile the kernel for Pentium M?

    Does anybody have benchmark results for x486 ,x586,x686 for
    comparison?

    Thanks

    yekta ayduk


  2. Re: slackware performance- processor type relation

    yekta wrote:
    > I'm using slackware 10.2 .According to /boot/config the kernel is
    > compiled for x486.
    > My SBC board has a Pentium M processor .Should I expect a big
    > performance increase ,when I compile the kernel for Pentium M?
    >
    > Does anybody have benchmark results for x486 ,x586,x686 for
    > comparison?


    I do not know exactly but the kernel is only a component (corresponding
    to the %sy value that you see in top. I believe a benchmark will not be
    affected since the cpu is occupied mostly by the benchmark not the
    kernel (it is the benchmark that would have to be compiled for P-M). To
    have better performance you should have to recompile all applications
    (or at least the ones that use a lot of CPU cycles).

    I don't really understand this policy of compiling for such old
    hardware. Lots of packages would not be compatible anyway (KDE is a good
    example) because the performances of a 486 are not sufficient.

    Olive

  3. Re: slackware performance- processor type relation

    On 11/07/07 08:06, yekta wrote:
    > I'm using slackware 10.2 .According to /boot/config the kernel is
    > compiled for x486.
    > My SBC board has a Pentium M processor .Should I expect a big
    > performance increase ,when I compile the kernel for Pentium M?


    Not in all areas. Do not expect a big increase in speed, but a 486
    may not make some features available, thought I think that the kernel
    tests at boot the type of CPU it is running.

    > Does anybody have benchmark results for x486 ,x586,x686 for
    > comparison?


    Ciao
    Giovanni
    --
    A computer is like an air conditioner,
    it stops working when you open Windows.
    Registered Linux user #337974 < http://giovanni.homelinux.net/ >

  4. Re: slackware performance- processor type relation

    Olive wrote:

    > I don't really understand this policy of compiling for such old
    > hardware. Lots of packages would not be compatible anyway (KDE is a
    > good example) because the performances of a 486 are not sufficient.


    What does "performance" have to do with "compatible". The 486 is
    most certainly "compatible" with KDE and similarly large applications.
    That one would have insufficient performance means simply that it would
    be unusable for us mere humans, but the application (or applications, in
    KDE's case) would certainly run (given sufficient virtual-memory space).

    --
    ----------------------------------------------------------------------
    Sylvain Robitaille syl@alcor.concordia.ca

    Systems and Network analyst Concordia University
    Instructional & Information Technology Montreal, Quebec, Canada
    ----------------------------------------------------------------------

  5. Re: slackware performance- processor type relation

    Are all the linux distributions behave like this?Actually the
    distribution should determine the CPU type of the hardware and should
    compile the kernel and applications for best performance .What about
    Windows OS.?


  6. Re: slackware performance- processor type relation

    On Wed, 07 Nov 2007 12:41:03 +0000, Giovanni wrote:

    > On 11/07/07 08:06, yekta wrote:
    >> I'm using slackware 10.2 .According to /boot/config the kernel is
    >> compiled for x486.
    >> My SBC board has a Pentium M processor .Should I expect a big
    >> performance increase ,when I compile the kernel for Pentium M?

    >
    > Not in all areas. Do not expect a big increase in speed, but a 486
    > may not make some features available, thought I think that the kernel
    > tests at boot the type of CPU it is running.
    >
    >> Does anybody have benchmark results for x486 ,x586,x686 for
    >> comparison?

    >
    > Ciao
    > Giovanni
    >

    Different processors have different feature sets. AFAIK, the most
    compatible kernel selection for Intel 386 derivatives is "386" with
    math coprocessor emulation in software. Slackware's default kernel
    selects 486 compatibility (and with math coprocessor emulation still
    present.) I could be wrong here, but my guess is that the kernel will only
    know about the feature set as selected. It won't attempt to use any
    assembly instructions that are not present in the actual 486 CPU and
    its assocatied 487 math coprocessor. I think you are correct that this is
    not ideal for Pentium M. For one thing, Pentium M includes the advanced
    SSEx assembly instructions and these will perform better than other
    work-alike instructions. A kernel which includes support for it will
    have a better performance on that hardware.

    However, the default kernel is better from a compatibility point of view.
    This "486" level kernel will still work fine on Pentium M. This is
    because the Pentium M's instruction set is backward compatible with the
    i486. So, it depends on what you want or need: compatibility or maximum
    performance for your specific hardware.

    The new instructions for Pentium M make compiling a tailored kernel very
    desirable, IMO. The extra features built on that processor will then
    be availble for use- SSEx, speedstep centrino processor scaling, etc.

    --
    Douglas Mayne

  7. Re: slackware performance- processor type relation

    openwaterfrostbites@gmail.com wrote:

    >>> I don't really understand this policy of compiling for such old
    >>> hardware. Lots of packages would not be compatible anyway (KDE is a
    >>> good example) because the performances of a 486 are not sufficient.


    >> What does "performance" have to do with "compatible". *The 486 is
    >> most certainly "compatible" with KDE and similarly large applications.
    >> That one would have insufficient performance means simply that it would
    >> be unusable for us mere humans, but the application (or applications, in
    >> KDE's case) would certainly run (given sufficient virtual-memory space).

    > Are all the linux distributions behave like this?Actually the
    > distribution should determine the CPU type of the hardware and should
    > compile the kernel and applications for best performance .What about
    > Windows OS.?


    No normally each distribution uses some sort of baseline line i486 686 or
    something. Some distro's like Gentoo and Sourcemage compile all the
    packages to the hardware set by the user. Windows is like most programs in
    that it is compiled with a fall back option. If the CPU does not support
    the hardware feature the software falls back to a specific set of code that
    does it all in hardware. This sort of effect can be seen by comparing the
    printing speed of Win98 and Win95 on a non-MMX machine. Win95 will print
    faster on such hardware than win98. Install on a machine with MMX support
    and suddenly Win98 goes a lot faster. This also is the difference between
    the -march and -mtune settings in GCC. Even though a package may be
    compiled for support for better processors with -mtune they are set to run
    on at least the -march processor settings.

    Note these settings are irrelevant in most cases. The cases where they do
    make a difference are in programs that use advanced features like MMX and
    SSE. Programs such as media decoders or encoders often use these settings.
    Some games and graphics programs can benefit from use SSE/MMX but only if
    the author used MMX/SSE code in their program. On some architectures I
    think that recompiling can make a significant difference, Itanium for
    example runs native i386 code much slower than Itanium code.

    Unless you know that a program will actually run faster recompiled you are
    wasting your time trying. I will take you more time to recompile a package
    than the few minutes over a month you will gain from recompiling. Of course
    if you use a program a lot you might need to reconsider that estimate.

    Richard James

  8. Re: slackware performance- processor type relation

    -----BEGIN PGP SIGNED MESSAGE-----
    Hash: SHA1

    On 2007-11-10, Richard James wrote:
    > On some architectures I
    > think that recompiling can make a significant difference, Itanium for
    > example runs native i386 code much slower than Itanium code.


    That's not true at all. Anytime I run native i386 code on an Itanium
    it completes almost immediately.

    alan@itanium~# some_i386_app
    Segmentation fault. Core dumped.

    (Hint: Itanium can't run i386 code at all.)

    - --
    It is better to hear the rebuke of the wise,
    Than for a man to hear the song of fools.
    Ecclesiastes 7:5
    -----BEGIN PGP SIGNATURE-----
    Version: GnuPG v1.4.7 (GNU/Linux)

    iD8DBQFHOOzjrZS6hX/gvjoRAtc5AKCYDREhI1pSn9godqNF8RIP1heyvgCglGn5
    8+LJ4qgT+CecUf/26UhdoTs=
    =1UP1
    -----END PGP SIGNATURE-----

  9. Re: slackware performance- processor type relation

    +Alan Hicks+ wrote:
    > Anytime I run native i386 code on an Itanium
    > it completes almost immediately.
    >
    > alan@itanium~# some_i386_app
    > Segmentation fault. Core dumped.


    well, you know, alan, perhaps you should take that as a hint not to run
    i386 apps on your itanium boxes...

    (just a thought.)

    ;-P


    --
    Joost Kremers joostkremers@yahoo.com
    Selbst in die Unterwelt dringt durch Spalten Licht
    EN:SiS(9)

  10. Re: slackware performance- processor type relation

    +Alan Hicks+ wrote:

    > -----BEGIN PGP SIGNED MESSAGE-----
    > Hash: SHA1
    >
    > On 2007-11-10, Richard James wrote:
    >> On some architectures I
    >> think that recompiling can make a significant difference, Itanium for
    >> example runs native i386 code much slower than Itanium code.

    >
    > That's not true at all. Anytime I run native i386 code on an Itanium
    > it completes almost immediately.
    >
    > alan@itanium~# some_i386_app
    > Segmentation fault. Core dumped.
    >
    > (Hint: Itanium can't run i386 code at all.)


    That's what I get for typing and relying on my memory instead of checking
    first.

    Richard James

+ Reply to Thread