Re: Solaris x86 version of Matlab - Solaris

This is a discussion on Re: Solaris x86 version of Matlab - Solaris ; Gulleiot Ulleiot wrote: > I'm interested in a version of Matlab for Solaris x86. I see > a lot of talk about this recently, and would like to add my > voice to those wanting it. > > Wolfram Resarch ...

+ Reply to Thread
Page 1 of 2 1 2 LastLast
Results 1 to 20 of 22

Thread: Re: Solaris x86 version of Matlab

  1. Re: Solaris x86 version of Matlab

    Gulleiot Ulleiot wrote:
    > I'm interested in a version of Matlab for Solaris x86. I see
    > a lot of talk about this recently, and would like to add my
    > voice to those wanting it.
    >
    > Wolfram Resarch have done this for Mathematica, and given
    > Mathworks have versions for both Solaris (SPARC) and Linux,
    > producing an x86 port should be easy. Any Solaris specific
    > code should simply recompile. Any x86 assmebler should move
    > from Linux quite easily. A port should not be too hard.
    >
    >


    You should have perhaps copied that to the Solaris newsgroups - there
    has been some discussion about Matlab on Solaris recently there. I took
    the liberty of sending this to a couple of Solaris newsgroups.

    I suggested to the other person wanting Matlab on Solaris x86 that they
    consider using Mathematica, since that does exist for Solaris x86.

    Actually, Mathematica on Solaris needs an AMD 64-bit CPU. Clearly
    Wolfram Research have decided 32-bit is not worth worrying with.

    Personally I find that requirement for 64-bit AMD CPUs a bit annoying,
    as it means it can't be used on many laptops. As far as I know, all the
    AMD 64-bit CPUs, including the mobile ones, are a lot more power hungry
    than their Intel equivalents. But judging by some benchmarks I have
    seen, Mathematica seems to run very well on Solaris x86 (or is it
    Solaris x64?). But it does need a 64-bit AMD CPU.

    As you say, a port of Matlab should be quite easy, so if Mathworks will
    not do it, consider some similar(ish) software that will.


  2. Re: Solaris x86 version of Matlab

    Andrew Gabriel wrote:

    >> Actually, Mathematica on Solaris needs an AMD 64-bit CPU. Clearly

    >
    > I presume it's OK on Intel 64-bit x86 CPU too, as it was
    > a Sun Ultra 24 Workstation running Solaris 10 x64 which
    > held the Mathematica performance record.
    >


    Em, you have a point there.

    But

    http://www.wolfram.com/products/mathematica/platforms/
    says:

    Sun x86
    Solaris 10 64-bit
    Requires AMD 64-bit machine.



    >> Wolfram Research have decided 32-bit is not worth worrying with.

    >
    >> As you say, a port of Matlab should be quite easy, so if Mathworks will
    >> not do it, consider some similar(ish) software that will.

    >
    > It's surprising if Mathworks won't do it, as they're likely
    > missing out on meeting their top performance results, if
    > Mathematica is anything to go by.
    >



    Yes, very good point.

  3. Re: Solaris x86 version of Matlab

    Dave wrote:
    > Gulleiot Ulleiot wrote:
    >
    >> I'm interested in a version of Matlab for Solaris x86. I see
    >> a lot of talk about this recently, and would like to add my
    >> voice to those wanting it.
    >> Wolfram Resarch have done this for Mathematica, and given
    >> Mathworks have versions for both Solaris (SPARC) and Linux,
    >> producing an x86 port should be easy. Any Solaris specific
    >> code should simply recompile. Any x86 assmebler should move
    >> from Linux quite easily. A port should not be too hard.
    >>
    >>

    >
    > You should have perhaps copied that to the Solaris newsgroups - there
    > has been some discussion about Matlab on Solaris recently there. I took
    > the liberty of sending this to a couple of Solaris newsgroups.
    >
    > I suggested to the other person wanting Matlab on Solaris x86 that they
    > consider using Mathematica, since that does exist for Solaris x86.
    >
    > Actually, Mathematica on Solaris needs an AMD 64-bit CPU. Clearly
    > Wolfram Research have decided 32-bit is not worth worrying with.
    >
    > Personally I find that requirement for 64-bit AMD CPUs a bit annoying,
    > as it means it can't be used on many laptops. As far as I know, all the
    > AMD 64-bit CPUs, including the mobile ones, are a lot more power hungry
    > than their Intel equivalents. But judging by some benchmarks I have
    > seen, Mathematica seems to run very well on Solaris x86 (or is it
    > Solaris x64?). But it does need a 64-bit AMD CPU.
    >
    > As you say, a port of Matlab should be quite easy, so if Mathworks will
    > not do it, consider some similar(ish) software that will.
    >


    ISTR that there is an open source version of Matlab. Many years ago I
    got a copy from an AT&T website. I also recall that it's written in
    Fortran so you'll need a Fortran compiler.

    If you want the very latest and greatest with support, you'll have to
    pay for it. I used the free version because I was on a zero budget.
    The commercial version is maintained and supported. The free version,
    needless to say, is not!


  4. Re: Solaris x86 version of Matlab

    Richard B. Gilbert wrote:

    > ISTR that there is an open source version of Matlab. Many years ago I
    > got a copy from an AT&T website. I also recall that it's written in
    > Fortran so you'll need a Fortran compiler.
    >
    > If you want the very latest and greatest with support, you'll have to
    > pay for it. I used the free version because I was on a zero budget.
    > The commercial version is maintained and supported. The free version,
    > needless to say, is not!
    >


    There are a few.

    http://www.dspguru.com/sw/opendsp/mathclo2.htm

    Someone recently told me Octave was very good. What with that, and
    seeing someone post a question about a problem building Octave on
    Solaris x86, I decided to give the compile a try on SPARC. But it bombs
    out as soon as I try to run gmake

    I'm expecting to be sent some matlab files soon. I'll probably try to
    convert them to Mathematica.




  5. Re: Solaris x86 version of Matlab

    Andrew Gabriel wrote:
    > In article <473f3c3a@212.67.96.135>,
    > Dave writes:
    >> Andrew Gabriel wrote:
    >>> In article <473efd1f@212.67.96.135>,
    >>> Dave writes:
    >>>> Actually, Mathematica on Solaris needs an AMD 64-bit CPU. Clearly
    >>> I presume it's OK on Intel 64-bit x86 CPU too, as it was
    >>> a Sun Ultra 24 Workstation running Solaris 10 x64 which
    >>> held the Mathematica performance record.

    >> Em, you have a point there.
    >>
    >> But
    >>
    >> http://www.wolfram.com/products/mathematica/platforms/
    >> says:
    >>
    >> Sun x86
    >> Solaris 10 64-bit
    >> Requires AMD 64-bit machine.

    >
    > Intel has been supporting most of the AMD64 instruction set
    > for a good while now, although there are some specialised
    > instructions they have their own versions of -- don't know
    > if Mathematica goes near those, but most likely it just
    > wants the base 64-bit instruction set which Intel copies.


    I did a bit more digging ...

    It looks like that record on the Sun Ultra 45 was set on Suse Linux, not
    Solaris x86.

    http://www.sun.com/desktop/workstati...nchmarks.jsp#1
    http://blogs.sun.com/bmseer/entry/ma..._world_records

    I assume if it was set on Solaris, Sun would be promoting that more than
    setting the record on Suse.

    Anyone ever tried Mathematica 6 on an Intel based Solaris x86 system?

  6. Re: Solaris x86 version of Matlab

    Andrew Gabriel wrote:


    >> I did a bit more digging ...
    >>
    >> It looks like that record on the Sun Ultra 45 was set on Suse Linux, not
    >> Solaris x86.
    >>
    >> http://www.sun.com/desktop/workstati...nchmarks.jsp#1
    >> http://blogs.sun.com/bmseer/entry/ma..._world_records

    >
    > I was thinking of this one, but it's AMD64:
    > http://www.sun.com/desktop/workstati...nchmarks.jsp#2
    > http://www.sun.com/customers/software/wolfram.xml
    >


    Yes, as you say, an AMD chip.

    It would be interesting to know if it does in practice run on Intel
    chips, since you say Intel support most of the AMD instructions. I don't
    have a 64-bit intel chip myself, so are unable to test it. But if anyone
    does, they can get a month demo of Mathematica for any platform,
    including Solaris x86.

    If Mathematica really does need an AMD chip, I wonder how much Wolfram
    gain from using an AMD specific instructions? It would have to be quite
    a bit for them to ignore the Intel users. I've no idea how many Intel
    64-bit Solaris boxes exist compared to AMD 64-bit Solaris x86 ones, but
    given Intel is a lot bigger than AMD, I assume that means there are more
    Intel ones.

  7. Re: Solaris x86 version of Matlab

    * Dave:

    > Personally I find that requirement for 64-bit AMD CPUs a bit annoying,
    > as it means it can't be used on many laptops.


    It can. Mathematica doesn't require an "AMD 64bit CPU" but an "AMD64
    CPU". The difference is that the first one means the CPU must be made by
    AMD, the second one just refers to AMD64-compatible CPUs (aka "x64" aka
    "EM64T" aka "intel64").

    Mathematica runs on all CPUs that use the AMD64/x64/EM64T/intel64
    extensions, that means besides AMD Athlon/Opteron/Phenom it also runs on
    intel P4 (with EM64T)/XEON (with EM64T/intel64)/Core2.

    > As far as I know, all the
    > AMD 64-bit CPUs, including the mobile ones, are a lot more power hungry
    > than their Intel equivalents.


    Well, the difference between AMD Turion and intel Core2 Mobile is quite
    small and the power consumption around(!) the same.

    Benjamin

  8. Re: Solaris x86 version of Matlab

    Benjamin Gawert wrote:

    > Mathematica runs on all CPUs that use the AMD64/x64/EM64T/intel64
    > extensions, that means besides AMD Athlon/Opteron/Phenom it also runs on
    > intel P4 (with EM64T)/XEON (with EM64T/intel64)/Core2.


    It needs the AMD_3DNOW extension, which my laptop (Intel Core 2 Duo
    T7200 2.0 GHz) does not support.

  9. Re: Solaris x86 version of Matlab

    Dave wrote:
    > Andrew Gabriel wrote:
    >
    >
    >>> I did a bit more digging ...
    >>>
    >>> It looks like that record on the Sun Ultra 45 was set on Suse Linux,
    >>> not Solaris x86.
    >>>
    >>> http://www.sun.com/desktop/workstati...nchmarks.jsp#1
    >>> http://blogs.sun.com/bmseer/entry/ma..._world_records

    >>
    >> I was thinking of this one, but it's AMD64:
    >> http://www.sun.com/desktop/workstati...nchmarks.jsp#2
    >> http://www.sun.com/customers/software/wolfram.xml
    >>

    >
    > Yes, as you say, an AMD chip.
    >
    > It would be interesting to know if it does in practice run on Intel
    > chips, since you say Intel support most of the AMD instructions. I don't
    > have a 64-bit intel chip myself, so are unable to test it. But if anyone
    > does, they can get a month demo of Mathematica for any platform,
    > including Solaris x86.
    >
    > If Mathematica really does need an AMD chip, I wonder how much Wolfram
    > gain from using an AMD specific instructions? It would have to be quite
    > a bit for them to ignore the Intel users. I've no idea how many Intel
    > 64-bit Solaris boxes exist compared to AMD 64-bit Solaris x86 ones, but
    > given Intel is a lot bigger than AMD, I assume that means there are more
    > Intel ones.


    Well, I was bored, you see, and I have a S10 system on an Intel P4, and,
    well, I was intrigued...

    The install fails:
    ld.so.1: MathKernel: fatal:
    /opt/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1:
    hardware capability unsupported: 0x100 [ AMD_3DNow ]

    Cheers,
    Gary B-)

    --
    __________________________________________________ ____________________________
    Armful of chairs: Something some people would not know
    whether you were up them with or not
    - Barry Humphries

  10. Re: Solaris x86 version of Matlab

    Gary R. Schmidt wrote:

    > Well, I was bored, you see, and I have a S10 system on an Intel P4, and,
    > well, I was intrigued...
    >
    > The install fails:
    > ld.so.1: MathKernel: fatal:
    > /opt/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1:
    > hardware capability unsupported: 0x100 [ AMD_3DNow ]
    >
    > Cheers,
    > Gary B-)
    >


    Yes, I got exactly the same after installing Solaris 10 on my laptop -
    which I only done last night. I hate to think why the Mathematica kernel
    needs 3D instructions.



  11. Re: Solaris x86 version of Matlab

    "Gary R. Schmidt" writes:

    >Well, I was bored, you see, and I have a S10 system on an Intel P4, and,
    >well, I was intrigued...


    >The install fails:
    >ld.so.1: MathKernel: fatal:
    >/opt/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1:
    >hardware capability unsupported: 0x100 [ AMD_3DNow ]


    I just had a look at the libsunperf.so that comes with Sun Studio 11 and 12:

    Studio 11:
    amd64/libsunperf.so.1: ELF 64-bit LSB dynamic lib AMD64 Version 1 [SSE2 SSE AMD_3DNow CMOV FPU], dynamically linked, not stripped, no debugging information available
    sse2/libsunperf.so.2: ELF 32-bit LSB dynamic lib 80386 Version 1 [SSE2 SSE MMX CMOV FPU], dynamically linked, not stripped, no debugging information available
    Studio 12:
    amd64/libsunperf.so.2: ELF 64-bit LSB dynamic lib AMD64 Version 1 [SSE2 SSE CMOV FPU], dynamically linked, not stripped, no debugging information available
    sse2/libsunperf.so.3: ELF 32-bit LSB dynamic lib 80386 Version 1 [SSE2 SSE MMX CMOV FPU], dynamically linked, not stripped, no debugging information available

    So it looks like the newer versions of libsunperf.so for amd64 are no longer compiled
    with the AMD_3DNow option. So it might really be possible to just replace the one
    supplied with Mathematica with a newer one.

    Regards,

    Matthias

    --
    +----------------------------------------+-----------------------------------+
    | Matthias Ernst | Phone: +41-1-632-4366 |
    | ETH Zürich, HCI D 227 | Fax: +41-1-632-1621 |
    | Laboratorium für Physikalische Chemie | |
    | ETH-Hönggerberg - HCI D 227 | Email: maer@nmr.phys.chem.ethz.ch |
    | CH-8093 Zürich, Switzerland | maer@scientist.com |
    +----------------------------------------+-----------------------------------+

  12. Re: Solaris x86 version of Matlab

    Matthias Ernst wrote:
    > "Gary R. Schmidt" writes:
    >
    >> Well, I was bored, you see, and I have a S10 system on an Intel P4, and,
    >> well, I was intrigued...

    >
    >> The install fails:
    >> ld.so.1: MathKernel: fatal:
    >> /opt/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1:
    >> hardware capability unsupported: 0x100 [ AMD_3DNow ]

    >
    > I just had a look at the libsunperf.so that comes with Sun Studio 11 and 12:
    >
    > Studio 11:
    > amd64/libsunperf.so.1: ELF 64-bit LSB dynamic lib AMD64 Version 1 [SSE2 SSE AMD_3DNow CMOV FPU], dynamically linked, not stripped, no debugging information available
    > sse2/libsunperf.so.2: ELF 32-bit LSB dynamic lib 80386 Version 1 [SSE2 SSE MMX CMOV FPU], dynamically linked, not stripped, no debugging information available
    > Studio 12:
    > amd64/libsunperf.so.2: ELF 64-bit LSB dynamic lib AMD64 Version 1 [SSE2 SSE CMOV FPU], dynamically linked, not stripped, no debugging information available
    > sse2/libsunperf.so.3: ELF 32-bit LSB dynamic lib 80386 Version 1 [SSE2 SSE MMX CMOV FPU], dynamically linked, not stripped, no debugging information available
    >
    > So it looks like the newer versions of libsunperf.so for amd64 are no longer compiled
    > with the AMD_3DNow option. So it might really be possible to just replace the one
    > supplied with Mathematica with a newer one.
    >
    > Regards,
    >
    > Matthias
    >



    I tried, but it does not work, as then a file (libfsu.so.l) appears to
    be missing. Although I have not tried much - it is possible that file is
    somewhere on the disk.


    I can't possibly see any justificatiofor requiring the kernel (which can
    be run without any user interface) to use 3D instructions. People run
    multiple kernels and one front end, and use the kernels on machines with
    no display connected. It seems strange that 3D instructions would be
    needed.


    The file Mathematica wants is

    /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1

    On my system, I found /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2 and
    running the file command does not mention 3D_NOW:

    # file /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2: ELF 64-bit LSB dynamic
    lib AMD64 Version 1 [SSE2 SSE CMOV FPU], dynamically linked, not
    stripped, no debugging information available

    So I created a link (after renaming the file supplied with Mathematica)

    # ln -s /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1

    Then when I try to run the program:

    # /usr/local/bin/math
    ld.so.1: MathKernel: fatal: libfsu.so.1: version `SUNWprivate_1.1' not
    found (required by file
    /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1)
    ld.so.1: MathKernel: fatal: libfsu.so.1: open failed: No such file or
    directory
    Killed


    Does anyone know the ratio of users of Solaris x86 on Intel to those on
    AMD CPUs? I can't believe Intel is in the minority, so I'm baffled why
    there is no support for Intel CPUs on Solaris x86, but there is AMD CPUs


  13. Re: Solaris x86 version of Matlab

    Dave wrote:
    [SNIP]
    >
    > I tried, but it does not work, as then a file (libfsu.so.l) appears to
    > be missing. Although I have not tried much - it is possible that file is
    > somewhere on the disk.
    >
    >
    > I can't possibly see any justificatiofor requiring the kernel (which can
    > be run without any user interface) to use 3D instructions. People run
    > multiple kernels and one front end, and use the kernels on machines with
    > no display connected. It seems strange that 3D instructions would be
    > needed.
    >
    >
    > The file Mathematica wants is
    >
    > /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1
    >
    >
    > On my system, I found /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2 and
    > running the file command does not mention 3D_NOW:
    >
    > # file /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    > /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2: ELF 64-bit LSB dynamic
    > lib AMD64 Version 1 [SSE2 SSE CMOV FPU], dynamically linked, not
    > stripped, no debugging information available
    >
    > So I created a link (after renaming the file supplied with Mathematica)
    >
    > # ln -s /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    > /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1
    >
    >
    > Then when I try to run the program:
    >
    > # /usr/local/bin/math
    > ld.so.1: MathKernel: fatal: libfsu.so.1: version `SUNWprivate_1.1' not
    > found (required by file
    > /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1)
    >
    > ld.so.1: MathKernel: fatal: libfsu.so.1: open failed: No such file or
    > directory
    > Killed
    >

    Hmmm, I have Studio 12 on my system, and libfsu is in various places
    under /opt/SUNWspro.

    Presumably the folk at Wolfram are *so* smart that hey have to control
    every little bit of what goes on. If they follow up the trial request,
    I may have a bit of a dig at them.

    > Does anyone know the ratio of users of Solaris x86 on Intel to those on
    > AMD CPUs? I can't believe Intel is in the minority, so I'm baffled why
    > there is no support for Intel CPUs on Solaris x86, but there is AMD CPUs
    >

    It seems a bit limiting to not have as generic a product as possiible.
    I can understand being 64-bit only, but this is just silly.

    Perhaps all they have to do is re-compile under Studio 12?

    Oh, and *is* libsunperf re-distributable? I don't recall reading
    anything about it in Studio 11, and I don't want to assume that it is
    unchanged in Studio 12.

    Cheers,
    Gary B-)

  14. Re: Solaris x86 version of Matlab

    Dave writes:

    >The file Mathematica wants is


    >/usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1


    >On my system, I found /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2 and
    >running the file command does not mention 3D_NOW:


    ># file /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    >/opt/SUNWspro/prod/lib/amd64/libsunperf.so.2: ELF 64-bit LSB dynamic
    >lib AMD64 Version 1 [SSE2 SSE CMOV FPU], dynamically linked, not
    >stripped, no debugging information available


    >So I created a link (after renaming the file supplied with Mathematica)


    ># ln -s /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    >/usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1


    >Then when I try to run the program:


    ># /usr/local/bin/math
    >ld.so.1: MathKernel: fatal: libfsu.so.1: version `SUNWprivate_1.1' not
    >found (required by file
    >/usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64/libsunperf.so.1)
    >ld.so.1: MathKernel: fatal: libfsu.so.1: open failed: No such file or
    >directory
    >Killed


    I just tried the following: From Studio 12 I copied the following libraries
    into the Mathematica library directory after moving the ones supplied by
    Mathematica to a save place:
    libsunperf.so.2
    libsunmath.so.1
    libfsu.so.1
    libfui.so.2
    I then renamed libsunperf.so.2 to libsunperf.so.1 und Mathematica starts up
    on an AMD machine without any libraries that have the 3DNow option set. I
    don't have an Intel machine yet, so I cannot test whether it runs on Intel
    or not but guessing from the libraries it should.

    Regards,

    Matthias

    --
    +----------------------------------------+-----------------------------------+
    | Matthias Ernst | Phone: +41-1-632-4366 |
    | ETH Zürich, HCI D 227 | Fax: +41-1-632-1621 |
    | Laboratorium für Physikalische Chemie | |
    | ETH-Hönggerberg - HCI D 227 | Email: maer@nmr.phys.chem.ethz.ch |
    | CH-8093 Zürich, Switzerland | maer@scientist.com |
    +----------------------------------------+-----------------------------------+

  15. Re: Solaris x86 version of Matlab

    On 2007-11-21 20:44:45 +0000, Dave said:

    > I can't possibly see any justificatiofor requiring the kernel (which
    > can be run without any user interface) to use 3D instructions. People
    > run multiple kernels and one front end, and use the kernels on machines
    > with no display connected. It seems strange that 3D instructions would
    > be needed.


    They're not. The AMD "3DNow!" instructions are just SIMD extensions
    with a stupid marketing name. Sort of like MMX and SSE.



    > Does anyone know the ratio of users of Solaris x86 on Intel to those on
    > AMD CPUs? I can't believe Intel is in the minority, so I'm baffled why
    > there is no support for Intel CPUs on Solaris x86, but there is AMD CPUs


    That's an interesting question. During the Dell announcement Schwartz
    said something about a large proportion of Solaris 10 x86 installs
    being on Dell boxes, and since Dells are all Intel...

    Cheers,

    Chris


  16. Re: Solaris x86 version of Matlab

    * Dave:

    > It needs the AMD_3DNOW extension, which my laptop (Intel Core 2 Duo
    > T7200 2.0 GHz) does not support.


    I remember to read somewhere that it also runs without 3DNow, but I'm
    not sure when and where it was.

    I sent an inquiry to Wolfram, let's see what they say.

    Benjamin

  17. Re: Solaris x86 version of Matlab

    Benjamin Gawert wrote:
    > * Dave:
    >
    >> It needs the AMD_3DNOW extension, which my laptop (Intel Core 2 Duo
    >> T7200 2.0 GHz) does not support.

    >
    > I remember to read somewhere that it also runs without 3DNow, but I'm
    > not sure when and where it was.
    >
    > I sent an inquiry to Wolfram, let's see what they say.
    >
    > Benjamin



    I've got the kernel running, by changing a couple of library files. I
    only seem to need to


    1) Move libsunperf.so.1 to libsunperf.so.1

    2) Set up a couple of links in the directory:
    /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64

    lrwxrwxrwx 1 root root 40 Nov 22 19:29 libfsu.so.1 ->
    /opt/SUNWspro/prod/lib/amd64/libfsu.so.1
    lrwxrwxrwx 1 root root 44 Nov 21 20:32 libsunperf.so.1 ->
    /opt/SUNWspro/prod/lib/amd64/libsunperf.so.2
    -bash-3.00$ pwd
    /usr/local/Wolfram/Mathematica/6.0/SystemFiles/Libraries/Solaris-x86-64

    But the front end is not being quite so nice. It crashes, but no obvious
    reason why.

    I'm not sure why Matthias changed other files - they don't seem to have
    this AMD_3DNOW issue, but perhaps he knows something I don't, as the
    front end it not working.

    FWIW, Mathematica 6.0.0 scores 1.392 on the built-in benchmark, but
    6.0.1 scores 2.00 under Vista, so is much quicker under Vista, but I'd
    rather use it under Solaris! (Either that or 6.0.1 is much quicker than
    6, which I doubt).

    # /usr/local/bin/math
    Mathematica 6.0 for Sun Solaris x86 (64-bit)
    Copyright 1988-2007 Wolfram Research, Inc.

    In[1]:= Needs["Benchmarking`"];

    In[2]:= Benchmark[]
    Test 1 of 15: Data Fitting ...
    Test 2 of 15: Digits of Pi ...
    Test 3 of 15: Discrete Fourier Transform ...
    Test 4 of 15: Eigenvalues of a Matrix ...
    Test 5 of 15: Elementary Functions ...
    Test 6 of 15: Gamma Function ...
    Test 7 of 15: Large Integer Multiplication ...
    Test 8 of 15: Matrix Arithmetic ...
    Test 9 of 15: Matrix Multiplication ...
    Test 10 of 15: Matrix Transpose ...
    Test 11 of 15: Numerical Integration ...
    Test 12 of 15: Polynomial Expansion ...
    Test 13 of 15: Random Number Sort ...
    Test 14 of 15: Singular Value Decomposition ...
    Test 15 of 15: Solving a Linear System ...

    Out[2]//InputForm=
    {"MachineName" -> "kingfisher", "System" -> "Sun Solaris x86 (64-bit)",
    "BenchmarkName" -> "MathematicaMark6", "FullVersionNumber" -> "6.0.0",
    "Date" -> "November 22, 2007", "BenchmarkResult" -> 1.392,
    "TotalTime" -> 61.9, "Results" -> {{"Data Fitting", 2.509},
    {"Digits of Pi", 0.878}, {"Discrete Fourier Transform", 2.319},
    {"Eigenvalues of a Matrix", 10.737}, {"Elementary Functions", 11.44},
    {"Gamma Function", 0.704}, {"Large Integer Multiplication", 1.406},
    {"Matrix Arithmetic", 7.259}, {"Matrix Multiplication", 4.139},
    {"Matrix Transpose", 2.078}, {"Numerical Integration", 3.553},
    {"Polynomial Expansion", 1.968}, {"Random Number Sort", 4.62},
    {"Singular Value Decomposition", 4.148}, {"Solving a Linear System",
    4.142}}}


  18. Re: Solaris x86 version of Matlab

    Dave wrote:

    >
    > I've got the kernel running, by changing a couple of library files. I
    > only seem to need to
    >
    >
    > 1) Move libsunperf.so.1 to libsunperf.so.1



    Oops, move it to libsunperf.so.1.orig (i.e take a backup)

  19. Re: Solaris x86 version of Matlab

    Matthias Ernst wrote:
    > Killed
    >
    > I just tried the following: From Studio 12 I copied the following libraries
    > into the Mathematica library directory after moving the ones supplied by
    > Mathematica to a save place:
    > libsunperf.so.2
    > libsunmath.so.1
    > libfsu.so.1
    > libfui.so.2
    > I then renamed libsunperf.so.2 to libsunperf.so.1 und Mathematica starts up
    > on an AMD machine without any libraries that have the 3DNow option set. I
    > don't have an Intel machine yet, so I cannot test whether it runs on Intel
    > or not but guessing from the libraries it should.
    >
    > Regards,
    >
    > Matthias
    >



    What made you choose those files? All I can see is that the original
    libsunperf.so.1 had the AMD_3DNow issue. I don't see it with
    libsunmath.so.1 for example.

    -bash-3.00$ file * | grep 3D
    libsunperf.so.1.orig: ELF 64-bit LSB dynamic lib AMD64 Version 1 [SSE2
    SSE AMD_3DNow CMOV FPU], dynamically linked, not stripped, no debugging
    information available


    But the front end it not working, but the kernel is.

    -bash-3.00$ /usr/local/bin/math
    Mathematica 6.0 for Sun Solaris x86 (64-bit)
    Copyright 1988-2007 Wolfram Research, Inc.

    In[1]:= 100!

    Out[1]=
    93326215443944152681699238856266700490715968264381 6214685929638952175\

    >

    99993229915608941463976156518286253697920827223758 2511852109168640000000\

    > 00000000000000000




  20. Re: Solaris x86 version of Matlab

    Dave wrote:
    > Benjamin Gawert wrote:
    >> * Dave:
    >>
    >>> It needs the AMD_3DNOW extension, which my laptop (Intel Core 2 Duo
    >>> T7200 2.0 GHz) does not support.

    >>
    >> I remember to read somewhere that it also runs without 3DNow, but I'm
    >> not sure when and where it was.
    >>
    >> I sent an inquiry to Wolfram, let's see what they say.
    >>
    >> Benjamin

    >
    >
    > I've got the kernel running, by changing a couple of library files. I
    > only seem to need to


    I've now got it working completely. I decided to start this on another
    thread, and copy to sci.math.symbolic, but leave off the Matlab one.
    There's probably some point on it going on
    comp.soft-sys.math.mathematica too, but I get fed up with the moderation
    on that


    See the new thread: "A hack to run Mathematica 6 on Solaris x86 with an
    Intel CPU"


+ Reply to Thread
Page 1 of 2 1 2 LastLast