1. Forum
  2. Usenet
  3. COMP.LANG.FORTRAN

  • OT: A tensor language prototype for high-performance computers

    From Gary Scott@21:1/5 to All on Wed Feb 16 09:52:56 2022
    Oh boy, another new language!

    A tensor language prototype for high-performance computers: https://techxplore.com/news/2022-02-tensor-language-prototype-high-performance.html

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Jos Bergervoet@21:1/5 to Gary Scott on Sun Feb 20 12:42:40 2022
    On 22/02/16 4:52 PM, Gary Scott wrote:
    Oh boy, another new language!

    A tensor language prototype for high-performance computers: https://techxplore.com/news/2022-02-tensor-language-prototype-high-performance.html


    Well.. it seems to have advantages over Java, the article says!

    --
    Jos

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Thomas Koenig@21:1/5 to Jos Bergervoet on Sun Feb 20 12:21:14 2022
    Jos Bergervoet <jos.bergervoet@xs4all.nl> schrieb:
    On 22/02/16 4:52 PM, Gary Scott wrote:
    Oh boy, another new language!

    A tensor language prototype for high-performance computers:
    https://techxplore.com/news/2022-02-tensor-language-prototype-high-performance.html


    Well.. it seems to have advantages over Java, the article says!

    It is not 100% clear, but it seems they solved the halting problem...

    # Coq had another intrinsic feature that made it attractive to the
    # MIT-based group: programs written in it, or adaptations of it,
    # always terminate and cannot run forever on endless loops (as can
    # happen with programs written in Java, for example).

    ... by making their language Turing-incomplete, they must effectively
    have banned loops.

    Just think of it - no Newton iterations. You cannot even implement
    an iterative sparse solver. Great!!!11!1!1!

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From fiziqs@gmail.com@21:1/5 to Thomas Koenig on Sun Feb 20 10:40:27 2022
    On Sunday, February 20, 2022 at 1:21:17 PM UTC+1, Thomas Koenig wrote:
    Jos Bergervoet <jos.ber...@xs4all.nl> schrieb:
    On 22/02/16 4:52 PM, Gary Scott wrote:
    Oh boy, another new language!

    A tensor language prototype for high-performance computers:
    https://techxplore.com/news/2022-02-tensor-language-prototype-high-performance.html


    Well.. it seems to have advantages over Java, the article says!
    It is not 100% clear, but it seems they solved the halting problem...

    # Coq had another intrinsic feature that made it attractive to the
    # MIT-based group: programs written in it, or adaptations of it,
    # always terminate and cannot run forever on endless loops (as can
    # happen with programs written in Java, for example).

    ... by making their language Turing-incomplete, they must effectively
    have banned loops.

    Just think of it - no Newton iterations. You cannot even implement
    an iterative sparse solver. Great!!!11!1!1!

    Not necessarily. Just your Newton-iteration must have a number of iterations, after which it does give up if it has not yet converged. I usually program it like that in Fortran or C++ too.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From gah4@21:1/5 to fiz...@gmail.com on Mon Feb 21 07:16:40 2022
    On Sunday, February 20, 2022 at 10:40:32 AM UTC-8, fiz...@gmail.com wrote:

    (snip)

    Not necessarily. Just your Newton-iteration must have a number of iterations, after which it does give up if it has not yet converged. I usually program it like that in Fortran or C++ too.

    The square root Newton-Raphson implementations I know do it in a fixed number, and are not written as loops. If you have a good initial guess, it is two for single
    and four for double precision. The last one is slightly tricky in other than binary floating point.

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)