Advertisement
If you have a new account but are having problems posting or verifying your account, please email us on hello@boards.ie for help. Thanks :)
Hello all! Please ensure that you are posting a new thread or question in the appropriate forum. The Feedback forum is overwhelmed with questions that are having to be moved elsewhere. If you need help to verify your account contact hello@boards.ie
Hi there,
There is an issue with role permissions that is being worked on at the moment.
If you are having trouble with access or permissions on regional forums please post here to get access: https://www.boards.ie/discussion/2058365403/you-do-not-have-permission-for-that#latest

Series Processing?

  • 25-11-2009 3:48am
    #1
    Overheal
    Registered Users, Registered Users 2 Posts: 86,729 ✭✭✭✭


    A thought just occurred to me, and google seems stumped by the query,

    If todays multicores run Parallel processing (multi-threading) cant we technically get multiple cores to run in a series to have one Really blazing fast Single Thread?

    Kinda like running batteries in parallel (gameboy) or running them in a series (forklift).


Comments

  • #2
    leninbenjamin
    Closed Accounts Posts: 8,983 ✭✭✭leninbenjamin


    Overheal wrote: »
    A thought just occurred to me, and google seems stumped by the query,

    If todays multicores run Parallel processing (multi-threading) cant we technically get multiple cores to run in a series to have one Really blazing fast Single Thread?

    Kinda like running batteries in parallel (gameboy) or running them in a series (forklift).

    No, because you'd have massive amounts of latency transferring data and instructions between cores, and even if you could miraculous transfer the data from core to core instantaneously, I don't think you'd have a scheduling algorithm accurate enough to handle it.


  • #3
    L31mr0d
    Registered Users, Registered Users 2 Posts: 8,067 ✭✭✭L31mr0d


    Overheal wrote: »
    If todays multicores run Parallel processing (multi-threading) cant we technically get multiple cores to run in a series to have one Really blazing fast Single Thread?

    Well... "technically" speaking, we should never of moved away from a single core, but Moores law could only go on for so long I guess. Once multiple cores where even entertained they admitted defeat to being able to increase transistor density.

    Parallel is nearly always better. Plus, in your scheme that places the cores in a series, there would still be a need for threading between the cores, only in your case they would all be sharing the same bus.

    Plus the way CPU's work currently they aren't 100% parallel or in series, they can share properties of both depending on the application.


  • #4
    leninbenjamin
    Closed Accounts Posts: 8,983 ✭✭✭leninbenjamin


    L31mr0d wrote: »
    Well... "technically" speaking, we should never of moved away from a single core

    This is the average lazy coder's response tbh. It really depends on the algorithm, but there are plenty of applications out there where a multi-core implementation scales better than a single core. Indeed most home applications (music, video, general browsing, games) could all be fairly easily parallelised too if lazy devs. got round to it (like google chrome actually). There are very very few truly serial applications tbh.

    Remember a single core would only have on L1, so what happens if you were to operate on a large data set? Memory has to be dumped back into main memory, and the processor sits idle. And it's not an easy task just to increase cache sizes, the larger the cache the slower the look up performance. That's why Intel and co started introducing multiple cache at different levels along the memory pipeline.

    L31mr0d wrote: »
    Plus the way CPU's work currently they aren't 100% parallel or in series, they can share properties of both depending on the application.

    I disagree, each processor is pretty much independent. They share a L2 or L3 cache and of course main memory, but you have to design the application itself in such a way as to explicitly eliminate the latency of going up the memory hierarchy. That's where the tedium kicks in. Certain processors (like the one in the Xbox) can dump results straight onto the main bus but that's about it. All it means is you can get data faster to each independent unit. I don't think it's correct to think of this as 'sharing properties'.


Advertisement