Professional Web Applications Themes

iBook vs PowerBook - Mac Portable

Steven Fisher <sdfisherspamcop.net> wrote: > > I've heard the UNIX kernel described as little more than 'bcopy()'. > > In other words, it just moves lots of data around. > > If the Altivec instructions can help in that regard, then maybe he's right. > > Exactly. There's a lot of copying going on, although I think at least > some of this was removed with Quartz Extreme. It may not have as big an > impact on recent G3s as on older models. QE is a supplementary technique as far as I know. Primarily it works on the videocard, ...

  1. #21

    Default Re: iBook vs PowerBook (Altivec instructions)

    Steven Fisher <sdfisherspamcop.net> wrote:
    > > I've heard the UNIX kernel described as little more than 'bcopy()'.
    > > In other words, it just moves lots of data around.
    > > If the Altivec instructions can help in that regard, then maybe he's right.
    >
    > Exactly. There's a lot of copying going on, although I think at least
    > some of this was removed with Quartz Extreme. It may not have as big an
    > impact on recent G3s as on older models.
    QE is a supplementary technique as far as I know. Primarily it works on
    the videocard, doesn't it? Giving the videocard the possibility to
    hardware accelerate the drawing of the different layers that make up the
    composite image on screen.

    So for boxes without QEsupport, a G4 may certainly help in making those
    computations and may do that lot faster than a G3 and probably also help
    less so for those computations in a QE supported system as the videocard
    then is already doing the GUI computations. Possibly though, the G4 may
    be able to feed the videocard faster, so it may still be significant
    even in a QE supported system, but still to a lesser degree.

    I agree that every application that can make use of altivec and have
    been programmed to do so, will run faster with it, but you said earlier:
    > Except that the Altivec instruction set is used in the core of Mac OS X,
    > so it affects every application. My Powerbook 500 MHz G4 runs rings
    > around a 700 MHz iMac.
    AFAIK, Quartz or any part of the GUI is *not* part of the core, as I
    mean that the core is the core in the GUI less Darwinlayer. Maybe you
    meant it more like "the core of OS X"? Anyway, I understood it as you
    were saying that a G4 is essential for having *any* applications run
    fast and as both my personal understanding of how OS X actually works
    and my real life experience does not support that, I expressed
    disagreement and asked you for some clarification, which you have not
    provided nor have you offered the basis for your viewpoint – such as
    real life experience with certain apps or an Apple OS X reference or a
    published speed tests (there should be a few pointing out the
    difference) that supports your notions – so that we could discuss things
    as they are. If we coudl do that we could all learn.

    I am not saying that you're not right that a G4 can run circles around a
    G3, it's just that you sounded very exaggerated in your ysis and
    that you gave the G3 a real bad rep as an OS X processor. I think it
    works fine, though oif course I want a G4. But do we all really need
    one?

    Personally I'd find it suffice to say that say a G4 will help a lot in
    making music, graphical processing (though just *cutting* video in FCP
    may work fine on a G3) and anything else that asks for advanced
    mathematical operations.

    If there's definitely a benifit in the core of OS X for altivec, such as
    all operations involving the core may run faster because of it, then I'd
    love to hear about it in detail. Including online references and tests.
    From anyone. I'm not laying the burden on you to 'prove' it, I just want
    to know. OK?
    Mikael Bystroem Guest

  2. #22

    Default Re: iBook vs PowerBook (Altivec instructions)

    In article <1fz44yh.1pve2hz1krm5xqN%mikael.lastnamemac.com >,
    [email]mikael.lastnamemac.com[/email] (Mikael Bystroem) wrote:

    | Mike <mghallenteract.com> wrote:
    |
    | > If I had any detailed knowledge of the
    | > Altivec instruction set, as you suggest you do,
    | I have not suggested that I do, but I had the impression altivec
    | instructions helps speedwise only for certain mathematical operations.
    | Moving data is not one of those AFAIK.

    Altivec has 128-bit wide registers, and a memory copy routine can
    theoretically be four times as fast using Altivec as one written using
    word (32-bit) loads and stores. However, the cache/memory hierarchy
    cannot keep up with the Altivec unit, so copies which involve L1/L2
    cache misses will be limited by memory speed.

    -- Tim Olson
    Tim Olson Guest

  3. #23

    Default Re: iBook vs PowerBook (Altivec instructions)

    Tim Olson <0jvo5ty02sneakemail.com> wrote:
    > Altivec has 128-bit wide registers, and a memory copy routine can
    > theoretically be four times as fast using Altivec as one written using
    > word (32-bit) loads and stores. However, the cache/memory hierarchy
    > cannot keep up with the Altivec unit, so copies which involve L1/L2
    > cache misses will be limited by memory speed.
    Thanks for the reminder, Tim. Very useful. However, the question still
    remains whether the core of OS X makes use of *such* copy operations in
    any significant amount affecting every application running. I think not.
    At least not in Jaguar.

    My best indicator of this not being so, at least not on every G4
    system, is that few applications, including the finder, seem to
    generally run noticeably faster under 10.2.x compared to a G3 system
    with the same CPU speed. Those that do run faster typically makes use of
    altivec natively in its own code, so faster execution should come as no
    surprise there. This is also the typical example for when G4 systems can
    run in circles around a G3 system. But I find it unlikely this is always
    so, due to the OS X core, even though I can not rule it out.

    I still wish that Steven Fisher would mention what core operations or
    systemwide affecting operations in his experience will run faster on the
    G4 with comparable G4 and G3 systems. Also, some URL's indicating real
    life working the way he describes would be nice.

    For the moment, I think Stevens postulate is based in sound theory,
    however not in the actual reality of the moment, meaning it could
    possibly work the way he describes, but does it? I'm getting both a G4
    and a G5 later this year, so of course I'd prefer him to be right on
    this issue.....
    Mikael Bystroem Guest

  4. #24

    Default Re: iBook vs PowerBook (Altivec instructions)

    In article <1fz5m0z.1ldcrw76nfde6N%mikael.lastnamemac.com> ,
    [email]mikael.lastnamemac.com[/email] (Mikael Bystroem) wrote:

    | Thanks for the reminder, Tim. Very useful. However, the question still
    | remains whether the core of OS X makes use of *such* copy operations in
    | any significant amount affecting every application running. I think not.
    | At least not in Jaguar.

    Well, the Altivec unit *is* used in general copy operations. Have a
    look at this disassembly of a portion of the memmove() library routine:

    0x90074460 <memmove+1012>: dcbt r10,r4
    0x90074464 <memmove+1016>: dcbt r11,r4
    0x90074468 <memmove+1020>: lvx v1,r0,r4
    0x9007446c <memmove+1024>: lvx v2,r6,r4
    0x90074470 <memmove+1028>: lvx v3,r7,r4
    0x90074474 <memmove+1032>: lvx v4,r8,r4
    0x90074478 <memmove+1036>: addi r4,r4,64
    0x9007447c <memmove+1040>: dcba r0,r12
    0x90074480 <memmove+1044>: stvx v1,r0,r12
    0x90074484 <memmove+1048>: stvx v2,r6,r12
    0x90074488 <memmove+1052>: dcba r7,r12
    0x9007448c <memmove+1056>: stvx v3,r7,r12
    0x90074490 <memmove+1060>: stvx v4,r8,r12
    0x90074494 <memmove+1064>: addi r12,r12,64
    0x90074498 <memmove+1068>: bdnz+ 0x90074460 <memmove+1012>

    That's actually a very efficient copy routine; it loads and stores four
    Altivec vector registers in the main loop, making appropriate use of the
    data cache block touch and data cache block allocate instructions.

    Now whether it actually makes a measureable difference in general system
    use is another matter.

    -- Tim Olson
    Tim Olson Guest

  5. #25

    Default Re: iBook vs PowerBook (Altivec instructions)

    Tim Olson <0jvo5ty02sneakemail.com> wrote:
    > Well, the Altivec unit *is* used in general copy operations. Have a
    > look at this disassembly of a portion of the memmove() library routine:
    > <snip>
    > That's actually a very efficient copy routine; it loads and stores four
    > Altivec vector registers in the main loop, making appropriate use of the
    > data cache block touch and data cache block allocate instructions.
    Looks interesting. Thanks.
    > Now whether it actually makes a measureable difference in general system
    > use is another matter.
    Exactly. I'm glad if OS X takes advantage of the possibilities in the
    hardware. I can't wait for Panther.....
    Mikael Bystroem Guest

Page 2 of 2 FirstFirst 12

Similar Threads

  1. 12 inch PowerBook vs iBook
    By JeffS87 in forum Mac Portable
    Replies: 7
    Last Post: September 15th, 07:37 AM
  2. iBook or PowerBook
    By Nikola in forum Mac Portable
    Replies: 1
    Last Post: August 22nd, 02:19 AM
  3. iBook vs Powerbook from a durability standpoint
    By Ryan Quick in forum Mac Portable
    Replies: 5
    Last Post: August 13th, 10:58 PM
  4. iBook USB 1.0 or 2.0?
    By AJ in forum Mac Portable
    Replies: 6
    Last Post: July 1st, 12:55 AM
  5. iChat Status: How does it work with iBook/PowerBook?
    By F. Todd Wilson in forum Mac Portable
    Replies: 4
    Last Post: June 29th, 10:05 AM

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •  

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139