Shop OBEX P1 Docs P2 Docs Learn Events
Multicore is hype! Why not the propeller? — Parallax Forums

Multicore is hype! Why not the propeller?

ErNaErNa Posts: 1,749
edited 2008-06-09 05:53 in Propeller 1
East to west and north to south everybody is talking about multi cores.
TI does multicore? Really?

Join TI’s Debbie Shemony, Bill Mills and Alan Gatherer in a 4-part series tracking the latest advances in multicore technology.


www.ti.com/corp/docs/landing/multicore/videos.htm

It is a pity that the propeller is not promoted in such a way. cry.gif

Comments

  • AleAle Posts: 2,363
    edited 2008-06-06 07:23
    And everyone else is avoiding to talk about the *lack* of software wink.gif
  • Graham StablerGraham Stabler Posts: 2,507
    edited 2008-06-06 08:27
    Really, that was terrible!
  • Graham StablerGraham Stabler Posts: 2,507
    edited 2008-06-06 10:46
    Interestingly in there written presentation they say that making multicore invisible to the programmer is the holy grail, I can see why this might be the case, if you are just using multiple cores to make faster processor.

    To me it is the difference between being the manager of a team or the manager of the team manager, I love the control over the players but if you just want the job done then the other tack might be better.
  • bambinobambino Posts: 789
    edited 2008-06-06 16:38
    It's just plan cute though that now they want to advertise!
    And do it in a ficious way in an attempt to say "I can't beleive you didn't know!"
    "Gee, we must be really good at keeping secrets"
  • Paul BakerPaul Baker Posts: 6,351
    edited 2008-06-06 17:59
    It may be the industry's goal to hide a multicore from the programmer, but it will never be our goal. It seems everyone is wedded to this notion of using complex task management and scheduling so that it appears as a homogeneous single system, but you give up so much to achieve that goal that we feel the benefit doesn't measure up to what has to be sacrificed.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Paul Baker
    Propeller Applications Engineer

    Parallax, Inc.
  • AleAle Posts: 2,363
    edited 2008-06-06 19:48
    Paul:

    You are right. But the "problem" is that when you do not rewrite most of the application(s) and just port it... if someone else made that "abstraction" it may be enough.

    The advantage in that regard to the propeller is that such lack of compatibility means that applications have to be rewritten and using more cores is just natural, unavoidable and in some cases the only way to go because there are no interrupts for instance or dedicated communications hw. I see it as an advantage, some other more structured people/way of doing things may fear it somehow. The good thing is that COGs do not compete for program or data memory making them predictable and avoiding race conditions and so on.
  • Oldbitcollector (Jeff)Oldbitcollector (Jeff) Posts: 8,091
    edited 2008-06-07 02:23
    Parallax:

    Thank you for not spending time to advertise in this manner..

    If I used TI multicore products this drivel would insult my intelligence.
    (Canned laughter didn't help) I felt like I was watching something
    produced by Steve Ballmer..

    Sorry...

    OBC

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    New to the Propeller?

    Getting started with the Protoboard? - Propeller Cookbook 1.4
    Updates to the Cookbook are now posted to: Propeller.warrantyvoid.us
    Got an SD card? - PropDOS
    Need a part? Got spare electronics? - The Electronics Exchange
  • jazzedjazzed Posts: 11,803
    edited 2008-06-07 05:07
    Looked like a poorly done SNL "Weekend Update" spoof to me. Marketing folks are just freaky sometimes.

    Linux SMP handles multi-core very nicely and transparently with the exception of device driver resource spin-locks which seem unavoidable·for SMP. Linux SMP of course requires a fairly substantial computing platform and a C99+ GNU toolset .... all outside the Propeller box. Given the target market, Propeller is a good tool except for memory and IO·constraints.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔


    Post Edited (jazzed) : 6/7/2008 7:16:06 AM GMT
  • LeonLeon Posts: 7,620
    edited 2008-06-07 09:17
    ADI makes the ADSP-BF561 Blackfin embedded symmetric multiprocessor with two 600 MIPS cores. Each core has two 16-bit MACs, two 40-bit ALUs, four 8-bit video ALUs,
    and a 40-bit shifter.

    Leon

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Amateur radio callsign: G1HSM
    Suzuki SV1000S motorcycle
  • J. A. StreichJ. A. Streich Posts: 158
    edited 2008-06-07 14:46
    The other thing is that most multi-core chips are being used for more "regular computers," where it might make sense to have the multi-core nature abstracted so that the same code can run on 1 core or 12 cores without recompiling. That said, in an embedded system with already specialized cores, and real-time demands, it makes sense to dedicate cores to specific tasks, and abstracting that away would be an uphill battle to embedded system designers. "Is there really going to be core looking at this pin always, or will there be times when that thread isn't executing so the scheduler can squeeze out a few more 'flops'?"

    I think Parallax has it right for propeller, and I think Intel/AMD have it right for PC. That said, I like to see Intel and AMD begin to think differently and creatively like Parallax has been doing for a long time.
  • ErNaErNa Posts: 1,749
    edited 2008-06-09 05:53
    The good old news is: Parallax's business is independent from the mainstream. Multi-core technology allows for increased computing power, if there is inherent parallelism in the program. That is not alway the case. Anyway, the multiple cores are hidden to the user. The propeller allows for defining separates tasks and run them in parallel by using a communication mechanism. Only sometimes it is necessary to serialise some task to place them into one cog for lack of cogs (and the prop II)
    So we will go on being the spearhead of innovation. wink.gif
Sign In or Register to comment.