• Time Sharing: Martin Richards, Ken Thompson and Linus Torvalds

    From Mild Shock@21:1/5 to Mild Shock on Thu Jun 19 14:38:43 2025
    Hi,

    Threads are just a variant of processes that
    share a heap, which can be the bottleneck. Processes
    are good for time sharing systems. Some history:

    How it started, support small CPUs:

    TRIPOS (TRIvial Portable Operating System)
    https://en.wikipedia.org/wiki/TRIPOS

    How it is going, more small CPUs excitement:

    The Origins of Linux—Linus Torvalds https://www.youtube.com/watch?v=WVTWCPoUt8w

    But what about GUI programming, or even web
    servers database access? Threads have massively
    fallen back with other models mainly through

    JavaScript, in the GUI via the browser and
    on the server via node.js. So don't be fooled
    by multi-threading Prolog systems,

    could be a complete nonsense...

    Mild Shock schrieb:

    I wasted a half dozen months with simulated
    “threads” by using CheerpJ. This was a very
    interesting experience, I could run formerly
    Jekejeke Prolog written in Java inside JavaScript.

    Main problem a simulation of “threads” cannot
    use certain GUI programming patterns from
    JavaScript. Because then you have still pre-emptive
    yield, like if you do it with auto-yield.

    But in JavaScript you have critical regions
    for free, just between two awaits, there will
    be no yield. Such properties might impact how
    you model your “concurrent” logic.

    This puts some weight on having a shield/1
    primitive. The CheerpJ threads were not that
    bad, I added auto yield, but shield/1 is a
    later invention from Dogelog Player.

    Mild Shock schrieb:
    Hi,

    When did the quantum leap happen in Dogelog Player?
    Well very early on: The first binary release was here:

    ANN: Dogelog Player 0.9.7
    The new version 0.9.7 of the Dogelog
    player now masters DCG.
    Admin User, created 09/06/2025

    And '$YIELD'/1 stuff came 9 months (sic!) later:

    ANN: Dogelog Player 0.9.8
    The Dogelog Player supports a new interpreter
    that can relinquish control.
    Admin User, created 21/03/2022

    See for your self:

    https://www.xlog.ch/izytab/doclet/home/12_news/package.html

    9 months, like Homo sapiens gestation period. (*)

    LoL

    Bye

    (*) To be frank, it was the perfect pass time during
    the COVID-19 pandemic.

    Mild Shock schrieb:
    Hi,

    Somebody wrote:

    Right, but what I’m saying is that with the
    current implementation as a separate process,
    running independently of the editor and only
    communicating over the channel defined by the
    protocol, the ability to have fine-grained
    asynchronous communication is limited.

    Not my problem not seeking the state of the art. You
    need of course a Prolog system that is all flesh
    and blood cooperative multitasking. If you want

    to do a GUI in Prolog similar like JavaScript
    does it now, or a native GUI adapter. And I don't
    count Tau-Prolog, although they showed amazing

    stuff. But not amazing enough, still based
    on continuations, and not based on promises. Or then
    async/await. I am not sure what SWI WASM now delivers?

    SWI Tinker shows me errors talking about threads when
    I use the query background execution feature and if I
    abort. But people from cooperative multitasking tend

    to not use the terminology "thread", they prefer to
    talk about "tasks". Ciao Playground doesn't count either,
    since their channels to a background worker can even

    not process an interrupt correctly.

    Bye

    Mild Shock schrieb:
    Hi,

    The fukkers stole my name:

    Jan - Local AI Assistant
    https://github.com/menloresearch/jan

    Bye

    P.S.: Meanwhile I am struggling with Quantum Paper
    whille doing a dark/light CSS. Look my light olive turned dark
    olive, that is somewhere near chocolate or is it dog shit?

    P.P.S.: Just figured out yesterday that some browser defaults
    are more Motif/OSF than Material/Google, aka Quantum Paper.



    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Mild Shock@21:1/5 to Mild Shock on Thu Jun 19 14:54:20 2025
    Hi,

    I find fewer and fewer use cases for threads.
    Here some fun with <parallel> Ant Task:

    <target name="dogelog_transpile">
    <parallel>
    <ant dir="../player/cross"
    antfile="run.xml"/>
    <ant dir="../playerpy/cross"
    antfile="run.xml"/>
    <ant dir="../playerj/cross"
    antfile="run.xml"/>
    </parallel>
    </target>

    Transpiling Dogelog Player in parallel. Except
    for a totally broken Ant log file, works
    smooth and shows a solid speed up. Although everything

    is heavily forked and possibly processes. I assume
    Ant offers tries to use threads, back from the time
    Java people were proud of their threads,

    but it nowadays with high speed solid state disks
    and large memory capacity, it doesn't make a dent anymore.

    LoL

    Bye

    P.S.: No GitHub Actions nonsense allowed here. Although
    matrix build looks interesting. Its possibly to uniform
    paralleliz. Unfortunately I have cases where I mix <parallel>

    with <sequential> in ant task. And for the lazy ones there
    is <macrodef> . What I didn't try yet is a better logger,
    it seems Gradel would provide that? One could have a DOM

    as Log datastructure, before the log gets committed.
    Simulating the mailboxes of SWI-Prolog message system.
    But the idea of "mailboxes" already indicate messages would

    leave process borders. Thank god I have delegated everything
    to Ant, don't need to make my hands dirty here.


    Mild Shock schrieb:
    Hi,

    Threads are just a variant of processes that
    share a heap, which can be the bottleneck. Processes
    are good for time sharing systems. Some history:

    How it started, support small CPUs:

    TRIPOS (TRIvial Portable Operating System) https://en.wikipedia.org/wiki/TRIPOS

    How it is going, more small CPUs excitement:

    The Origins of Linux—Linus Torvalds https://www.youtube.com/watch?v=WVTWCPoUt8w

    But what about GUI programming, or even web
    servers database access? Threads have massively
    fallen back with other models mainly through

    JavaScript, in the GUI via the browser and
    on the server via node.js. So don't be fooled
    by multi-threading Prolog systems,

    could be a complete nonsense...

    Mild Shock schrieb:

    I wasted a half dozen months with simulated
    “threads” by using CheerpJ. This was a very
    interesting experience, I could run formerly
    Jekejeke Prolog written in Java inside JavaScript.

    Main problem a simulation of “threads” cannot
    use certain GUI programming patterns from
    JavaScript. Because then you have still pre-emptive
    yield, like if you do it with auto-yield.

    But in JavaScript you have critical regions
    for free, just between two awaits, there will
    be no yield. Such properties might impact how
    you model your “concurrent” logic.

    This puts some weight on having a shield/1
    primitive. The CheerpJ threads were not that
    bad, I added auto yield, but shield/1 is a
    later invention from Dogelog Player.

    Mild Shock schrieb:
    Hi,

    When did the quantum leap happen in Dogelog Player?
    Well very early on: The first binary release was here:

    ANN: Dogelog Player 0.9.7
    The new version 0.9.7 of the Dogelog
    player now masters DCG.
    Admin User, created 09/06/2025

    And '$YIELD'/1 stuff came 9 months (sic!) later:

    ANN: Dogelog Player 0.9.8
    The Dogelog Player supports a new interpreter
    that can relinquish control.
    Admin User, created 21/03/2022

    See for your self:

    https://www.xlog.ch/izytab/doclet/home/12_news/package.html

    9 months, like Homo sapiens gestation period. (*)

    LoL

    Bye

    (*) To be frank, it was the perfect pass time during
    the COVID-19 pandemic.

    Mild Shock schrieb:
    Hi,

    Somebody wrote:

    Right, but what I’m saying is that with the
    current implementation as a separate process,
    running independently of the editor and only
    communicating over the channel defined by the
    protocol, the ability to have fine-grained
    asynchronous communication is limited.

    Not my problem not seeking the state of the art. You
    need of course a Prolog system that is all flesh
    and blood cooperative multitasking. If you want

    to do a GUI in Prolog similar like JavaScript
    does it now, or a native GUI adapter. And I don't
    count Tau-Prolog, although they showed amazing

    stuff. But not amazing enough, still based
    on continuations, and not based on promises. Or then
    async/await. I am not sure what SWI WASM now delivers?

    SWI Tinker shows me errors talking about threads when
    I use the query background execution feature and if I
    abort. But people from cooperative multitasking tend

    to not use the terminology "thread", they prefer to
    talk about "tasks". Ciao Playground doesn't count either,
    since their channels to a background worker can even

    not process an interrupt correctly.

    Bye

    Mild Shock schrieb:
    Hi,

    The fukkers stole my name:

    Jan - Local AI Assistant
    https://github.com/menloresearch/jan

    Bye

    P.S.: Meanwhile I am struggling with Quantum Paper
    whille doing a dark/light CSS. Look my light olive turned dark
    olive, that is somewhere near chocolate or is it dog shit?

    P.P.S.: Just figured out yesterday that some browser defaults
    are more Motif/OSF than Material/Google, aka Quantum Paper.




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