1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

IDF Prague: Terascale Computing

Discussion in 'Reviews' started by W1zzard, Oct 8, 2006.

  1. W1zzard

    W1zzard Administrator Staff Member

    May 14, 2004
    14,859 (3.93/day)
    Thanks Received:
    Last edited: Oct 10, 2006
  2. Lozza New Member

    Oct 19, 2004
    44 (0.01/day)
    Thanks Received:
    Great Wyrley-UK
    My personal thoughts are:

    This is just a general summary of what everyone's been thinking lately. Or me at least.

    Adding more cores is useless unless people learn how to code for them, but it is a method of increasing computing power cheaply.

    How long will it be before the common processor will be measured in "amount of cores" instead on "Mhz"?

    Memory and programming are the biggest blockades.

    As far as I know, loads of people have known this sort of stuff for a while now.

    Thanks for the article, it's a good read.
    Last edited: Oct 11, 2006
  3. PVTCaboose1337

    PVTCaboose1337 Graphical Hacker

    Feb 1, 2006
    9,513 (3.02/day)
    Thanks Received:
    San Antonio, Texas
    Were getting ahead in one field and staying behind in others. While processors are "quad multi, multi, super octal decahexal..." Other components are bottlenecking.
  4. Alec§taar New Member

    May 15, 2006
    4,677 (1.53/day)
    Thanks Received:
    Someone who's going to find NewTekie1 and teach hi
    It's NOT that difficult, especially when implementing what I call "implicitly ready smp"!

    Done MAINLY via the use of Win32 API via "CreateThread" in say, Delphi!

    (Delphi makes it simpler imo, to do them, than any other language for Win32 'standalone' non-interpreted apps (like VB & VC++ are, interpreted languages via runtime libs).

    (Moreso VB, it is TOTALLY so, & VC++ can do a statically linked stand-alone, afaik too, but iirc, its runtime largely only runs the interface code - could be wrong here, I don't use it a lot anymore, & have not for years).

    If you want to get MORE complex, & do "explicitly ready smp" code (as I call it), you need to do much more, like Processor Detections (how many) & SetThreadAffinityMask API calls etc.

    NOW, the TRULY hard part?

    Identifying WHAT functions and data can do this, w/ out hitting walls like "race conditions" (2 or more threads chasing time for working on the SAME data - not a GOOD idea, period, imo @ least).

    W1zz's document states INTEL's working on that being more "automagic" in their compilers, & this? I do NOT doubt... INTEL has some of the BEST compilers & libs for them out there, @ least for C++...

    VB.NET though, from my using it this year quite a lot, does make doing threadwork pretty much a 'snap' & maybe even simpler than Delphi, but I did hit snags w/ something called using "delegates" but this may just be my inexperience & incomplete mastery of VB.NET &/or ASP.NET @ this point is all!


    P.S.=> W1zz: Nice article, VERY enlightening man... thanks! I think I will have to 'browse' thru this section more tonite - if the reviews in here are of this level? Will be fun, & yes, enlightening... apk
    Last edited: Oct 11, 2006
  5. spectre440 New Member

    Jul 18, 2005
    948 (0.28/day)
    Thanks Received:
    great article.
    very interesting read.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page