Any progress on MATLAB port to Solaris x86? - Solaris

This is a discussion on Any progress on MATLAB port to Solaris x86? - Solaris ; Given MATLAB runs on Solaris (SPARC), it should be quite easy for Mathworks to do a port to Solaris x86. Unless there is assembly code written, it is likely to need no more than a recompile. If there is aseembly ...

+ Reply to Thread
Results 1 to 3 of 3

Thread: Any progress on MATLAB port to Solaris x86?

  1. Any progress on MATLAB port to Solaris x86?

    Given MATLAB runs on Solaris (SPARC), it should be quite easy for
    Mathworks to do a port to Solaris x86. Unless there is assembly code
    written, it is likely to need no more than a recompile. If there is
    aseembly code in the source, then using the Linux assembly code as a
    base, it should be easy to generate Solaris assembler.

    Hence technically, a port should be quite easy.

    Is there any work taking place on a port? There have been rumors
    circulating for some time a port would be done, but I've yet to see any
    official announcement.

    Mathematica runs ok on Solaris 86, so why not MATLAB?

  2. Re: Any progress on MATLAB port to Solaris x86?

    In alt.solaris.x86 Dave wrote:
    > Given MATLAB runs on Solaris (SPARC), it should be quite easy for
    > Mathworks to do a port to Solaris x86. Unless there is assembly code
    > written, it is likely to need no more than a recompile. If there is
    > aseembly code in the source, then using the Linux assembly code as a
    > base, it should be easy to generate Solaris assembler.
    >
    > Hence technically, a port should be quite easy.


    I don't think it is a technical issue so much as a resource issue. Adding
    support for a new platform (which looks and behaves very much like the
    existing SPARC platform) requires a bunch of effort at all stages of the
    release process. It isn't just a matter of compiling it up one afternoon
    and then sending it out the door. You can be sure that it will have its
    own set of bugs associated with it, and they will require extra resources,
    which might be better spent supporting other platforms.

    If they receive enough customer feedback asking for this new platform, then
    it might look like a good idea to them. I have been asking this question
    since they announced that support for SGI Irix was to be stoppped.

    Have you asked The MathWorks directly for support for Solaris x86, or just
    posted questions on Usenet groups? I suggest you do both.

    >
    > Is there any work taking place on a port? There have been rumors
    > circulating for some time a port would be done, but I've yet to see any
    > official announcement.


    That would be a question to ask your favourite MathWorks representative.

    In the mean time, you can run MATLAB in an lx-zone. There are some
    restrictions, but it does work, and quite well. You can only use MATLAB
    R2007a or earlier (glibc requirements) and only 32 bit versions (brandz
    only supports 32 bit applications). Both of these restrictions have been
    overcome in latest opensolaris implementations of brandz, but I don't
    believe they have made their way into the official Solaris release yet.

    After reading the brandz documentation, it took me about 30 minutes to set
    up a Linux zone and get MATLAB to run on it. I posted some information in
    comp.soft-sys.matlab back in January. Give it a try.


    --
    Dr Tristram J. Scott
    Energy Consultant

  3. Re: Any progress on MATLAB port to Solaris x86?

    Tristram Scott wrote:

    >> Hence technically, a port should be quite easy.

    >
    > I don't think it is a technical issue so much as a resource issue. Adding
    > support for a new platform (which looks and behaves very much like the
    > existing SPARC platform) requires a bunch of effort at all stages of the
    > release process.


    I understand that, but I doubt a huge effort and so I doubt it would
    take a lot of resources.

    > It isn't just a matter of compiling it up one afternoon
    > and then sending it out the door. You can be sure that it will have its
    > own set of bugs associated with it, and they will require extra resources,
    > which might be better spent supporting other platforms.


    My experience has been that when software is ported to another platform,
    you often notice new bugs, but really they are not new, but just had not
    been triggered before. So testing on multiple platforms tends to improve
    the code.

    > Have you asked The MathWorks directly for support for Solaris x86, or just
    > posted questions on Usenet groups? I suggest you do both.


    I've asked before and was told there were no plans. But I know others
    who have asked before that, and were told there were plans. I guess it
    depends who you talk to.

    > In the mean time, you can run MATLAB in an lx-zone. There are some
    > restrictions, but it does work, and quite well. You can only use MATLAB
    > R2007a or earlier (glibc requirements) and only 32 bit versions (brandz
    > only supports 32 bit applications). Both of these restrictions have been
    > overcome in latest opensolaris implementations of brandz, but I don't
    > believe they have made their way into the official Solaris release yet.
    >
    > After reading the brandz documentation, it took me about 30 minutes to set
    > up a Linux zone and get MATLAB to run on it. I posted some information in
    > comp.soft-sys.matlab back in January. Give it a try.


    Thanks for that. In the above you say R2007A or earlier, yet when I
    found the post you made in Jan of this year

    http://groups.google.com/group/comp....c3d86937a37b1b

    it says "MATLAB versions up to R2006a (7.2)"

    Mathematica once held a speed record on Solaris x86, with it faster on
    that platform than any other. From what results I have seen, Mathematica
    works a LOT faster on Solaris x86 than on any SPARC processor, so I
    guess MATLAB would see a big jump in performance compared to SPARC.


+ Reply to Thread