1. Forum
  2. USENET
  3. comp.lang.prolog

  • =?UTF-8?Q?ANN:_Dogelog_Player_2.2.5_=28=cf=80-WAM_Channels=29?=

    From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Wed Aug 12 02:18:11 2026
    From Newsgroup: comp.lang.prolog

    Dear All,

    We are happy to announce a new edition of
    the Dogelog Player:

    - GPU Backend:
    The two pilars of -C-calculus are processes and
    channels. While processes in the CPU Backend
    mapped straight forward to some platform threads
    and a new predicate execute/[1,2]. The predicate
    introduced here is expedite/[1,2] and it does
    the same for your GPU via compute shaders. We
    observed good results with iGPUs from
    budget AI laptops.

    - CPU Channels:
    We mentioned already that the two pilars of
    -C-calculus are processes and channels. In
    search of easy channel objects, we arrived at
    a kind of ADA RendezVous without ACK or NACK.
    This is basically a 1-element mpmc queue, now
    available via the predicate chan/1 for
    the CPU backend.

    - GPU Channels:
    Channel objects are now also available for the
    GPU backend via the predicate flit/1. The approach
    is again some busy-wait, so dont use CPU or GPU
    channels for isolated long waits. Because of a
    batch oriented GPU command API, we could not
    yet provide online communication
    between CPU and GPU .

    Have Fun!

    Jan Burse, August 12, 2026, https://www.herbrand.ai/
    --- Synchronet 3.22a-Linux NewsLink 1.2
  • From Mild Shock@janburse@fastmail.fm to comp.lang.prolog on Fri Aug 14 22:03:50 2026
    From Newsgroup: comp.lang.prolog

    Hi,

    Years ago Sam Altman said to have no idea how to
    generate revenue, but when the generally intelligent
    system is in place, he might ask it. We take Sudoku
    as the drosophila how different schools deal with
    rCLgeneralityrCY, in this particular instance attacking
    NP problems with constraint solvers.

    Every turing machine T has a turing machine TrCO that
    can do multiple steps at once. In propagation methods
    repeated goal reconstructions happens, while static
    methods construct goals once. The later is applied in
    Dogelog Player and looking at AI Escargot it can beat
    EyeProlog in the browser.

    Bye

    See also:

    BlumrCOs Speed-Up Theorem and Sudoku
    https://medium.com/2989/bb8acd0c3284

    Mild Shock schrieb:
    Dear All,

    We are happy to announce a new edition of
    the Dogelog Player:

    - GPU Backend:
    The two pilars of -C-calculus are processes and
    channels. While processes in the CPU Backend
    mapped straight forward to some platform threads
    and a new predicate execute/[1,2]. The predicate
    introduced here is expedite/[1,2] and it does
    the same for your GPU via compute shaders. We
    observed good results with iGPUs from
    budget AI laptops.

    - CPU Channels:
    We mentioned already that the two pilars of
    -C-calculus are processes and channels. In
    search of easy channel objects, we arrived at
    a kind of ADA RendezVous without ACK or NACK.
    This is basically a 1-element mpmc queue, now
    available via the predicate chan/1 for
    the CPU backend.

    - GPU Channels:
    Channel objects are now also available for the
    GPU backend via the predicate flit/1. The approach
    is again some busy-wait, so dont use CPU or GPU
    channels for isolated long waits. Because of a
    batch oriented GPU command API, we could not
    yet provide online communication
    between CPU and GPU .

    Have Fun!

    Jan Burse, August 12, 2026, https://www.herbrand.ai/

    --- Synchronet 3.22a-Linux NewsLink 1.2