ARM ISA

Article: Why Apple Won’t ARM the MacBook
By: David Kanter (dkanter.delete@this.realworldtech.com), May 10, 2011 9:12 am
Room: Moderated Discussions
Wilco (Wilco.Dijkstra@ntlworld.com) on 5/10/11 wrote:
---------------------------
>David Kanter (dkanter@realworldtech.com) on 5/10/11 wrote:
>---------------------------
>>none (none@none.com) on 5/10/11 wrote:
>>---------------------------
>>>There are some mistakes about ARM:
>>>
>>>1. except for NEON, decoding is easy, even for detecting
>>>2-byte vs 4-byte instructions (which are the only sizes
>>>available in Thumb mode)
>
>Yes, there are a few misconceptions in the article. There is effectively one ARM
>ISA which has 2 encodings: 32-bit ARM and 16/32-bit Thumb-2 (the 16-bit encodings
>are Thumb-1). VFP and Neon are in both of those, and use >almost identical encodings (so they can share a decoder).
>
>Neon is a big and complex instruction set, but decoding is not as difficult as
>it may seem at first - it is true the encoding bits are spread around but to hardware
>that doesn't matter at all. Like all ARM instructions, the encoding allows you to
>quickly determine whether an instruction is a load or store and which registers
>are accessed. And that is what matters most.
>
>>>2. Corte-A15 is Eagle and vice versa :-)
>>
>>Thanks for that correction.
>>
>>>3. the implicit barrel shifter indeed is an oddity, but it is
>>>no real issue: either split the instruction or add a pipe
>>>stage
>>
>>Still ugly.
>
>Not nearly as ugly has having instructions which access >memory, do a computation and then access memory again...
>
>It is definitely not "very expensive" hardware as ALUs form a tiny part of a core,
>and ALUs have a barrel shifter anyway. It is obviously reused for plain shifts,
>bitfield operations, sign extension and some SIMD operations. Many other architectures
>support at least shift(1-3)+add, which is useful for address computations and small multiplies.
>
>ARM has always been able to do more with fewer executed instructions, not just
>compared to its purist RISC siblings but even compared to x86. The 3 operands, conditional
>execuction, shifted immediates and shifts on ALU operations means you can pack a
>lot of computation into a single instruction, allowing a >relatively narrow design
>to be very efficient.

I don't disagree that it's useful. But it's also something I would not have put in an ISA if I knew I had to make an OOOE version. It's clever, but it's annoying : )

>How many ISAs can emulate an IEEE floating point multiply >in 25 instructions?

Well most ISAs do it in a single instruction : P

>>>4. in ARM mode (as opposed to Thumb mode) setting condition
>>>flags is an option
>>
>>Yes, I got that.
>
>It is also optional in 32-bit Thumb-2 instructions - only >the 16-bit Thumb-1 instructions
>always set the flags. Either option is not detrimental to >performance, as Power and x86 prove.

Yes, I agree. I think my whole point was that ARM is cleaner than x86, but is no beauty queen. There are things there that complicate high performance design.

>>>5. Thumb initially dropped the shifter and "predication" (I
>>>certainly don't like that term, it has become tainted, I
>>>prefer "conditional execution" :-) to fit in 16-bit opcodes,
>>>(32-bit Thumb came later) not to ease micro-architecture
>>>design.
>>
>>Predication/conditional execution. It's all the same. You can argue predication
>>is more for SIMD architectures, but I'm not sure it matters what you call it. The
>>point is that you now have at least 3 inputs for most operations...
>
>ARM can have max 4 register inputs, 2 outputs plus 1 in/out for the condition bits.
>The 3 and 4 register cases are rare (more so in Thumb-2) and could be split early
>on. Conditional execution does indeed pose a challenge to OoOE but not much more
>than branches do (one possible implementation is to treat a conditional instruction
>as a merged conditional branch + an unconditional >instruction).


>Anyway, I liked your article, and agree there is no chance Apple will replace their
>x86 laptops completely with ARM in the next few years. However I think it is extremely
>likely Apple will introduce laptops based on ARM - even if it is just to give people
>an iPad+keyboard option. There is no need for SB or >Bulldozer-like performance for
>this to make sense - after all we have Atom and Bobcat >laptops.

Yeah, I think a keyboard+tablet might work. Although that really does seem antithetical to Steve Jobs' aesthetic tendencies.

David
< Previous Post in ThreadNext Post in Thread >
TopicPosted ByDate
New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/10 12:19 AM
  New Article: Why Apple Won’t ARM the MacBookMark Christiansen2011/05/09 11:24 PM
  New Article: Why Apple Won’t ARM the MacBookLouis Gerbarg2011/05/09 11:54 PM
  New Article: Why Apple Won’t ARM the MacBooknone2011/05/10 12:57 AM
    New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/10 01:18 AM
      New Article: Why Apple Won’t ARM the MacBooknone2011/05/10 02:14 AM
        New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/10 02:19 AM
          New Article: Why Apple Won’t ARM the MacBookGoose2011/05/10 02:34 AM
          New Article: Why Apple Won’t ARM the MacBookRatherNotSay2011/05/10 06:25 AM
          ARM predication not so horriblePaul A. Clayton2011/05/10 08:05 AM
            ARM predication not so horriblenone2011/05/10 08:12 AM
              Partial update seems kind of oddPaul A. Clayton2011/05/10 08:29 AM
                Partial update seems kind of oddExophase2011/05/10 08:51 AM
                  Partial update seems kind of oddMichael S2011/05/10 12:56 PM
                    Partial update seems kind of oddMichael S2011/05/10 01:25 PM
                      Partial update seems kind of oddWilco2011/05/10 02:39 PM
                        Partial update seems kind of oddanonymous422011/05/10 07:09 PM
                          Partial update seems kind of oddrwessel2011/05/10 09:57 PM
                  Condition register usesPaul A. Clayton2011/05/10 05:56 PM
                  Partial update seems kind of oddMegol2011/05/11 01:06 AM
          New Article: Why Apple Won’t ARM the MacBookKevin G2011/05/11 01:21 AM
      New Article: Why Apple Won’t ARM the MacBookWilco2011/05/10 03:47 AM
        New Article: Why Apple Won’t ARM the MacBookMichael S2011/05/10 04:06 AM
          New Article: Why Apple Won’t ARM the MacBookWilco2011/05/10 04:25 AM
            New Article: Why Apple Won’t ARM the MacBookMichael S2011/05/10 05:13 AM
        Predication and OoOEExophase2011/05/10 08:34 AM
          Predication and OoOEWilco2011/05/10 03:46 PM
            Predication and OoOEExophase2011/05/11 07:51 AM
        ARM ISADavid Kanter2011/05/10 09:12 AM
          ARM ISAUngo2011/05/10 12:30 PM
  New Article: Why Apple Won’t ARM the MacBookThibs2011/05/10 02:45 AM
  New Article: Why Apple Won’t ARM the MacBookLlew2011/05/10 03:33 AM
    New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/10 09:34 AM
  New Article: Why Apple Won’t ARM the MacBookKevin G2011/05/10 03:55 AM
    New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/10 10:07 AM
      New Article: Why Apple Won’t ARM the MacBookKevin G2011/05/11 01:33 AM
        New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/11 02:00 AM
  New Article: Why Apple Won’t ARM the MacBook ... The FACTS say otherwise !fabrice2011/05/10 04:07 AM
    New Article: Why Apple Won’t ARM the MacBook ... The FACTS say otherwise !Rakesh Malik2011/05/11 03:03 PM
      New Article: Why Apple Won’t ARM the MacBook ... The FACTS say otherwise !Richard Cownie2011/05/12 08:50 AM
        New Article: Why Apple Won’t ARM the MacBook ... The FACTS say otherwise !Rakesh Malik2011/05/12 09:10 AM
    Hrmmmm...David Kanter2011/05/14 12:19 AM
  iPad-writer?Jouni Osmala2011/05/10 04:35 AM
    Laptops finally moving to touchscreen GUI?mpx2011/05/10 06:03 AM
  What about MS?Joe Kerr2011/05/10 06:26 AM
  New Article: Why Apple Won’t ARM the MacBookRichard Cownie2011/05/10 06:41 AM
    Dual CPU makes more sense than dropping x86Mark Roulo2011/05/10 07:54 AM
      Dual CPU makes more sense than dropping x86Richard Cownie2011/05/10 09:02 AM
        Dual CPU makes more sense than dropping x86Mark Roulo2011/05/10 09:28 AM
          Dual CPU makes more sense than dropping x86Richard Cownie2011/05/10 09:55 AM
            Dual CPU makes more sense than dropping x86Wes Felter2011/05/10 11:20 AM
      Dual CPU makes more sense than dropping x86rwessel2011/05/10 11:01 AM
        Dual CPU makes more sense than dropping x86fabrice2011/05/10 12:02 PM
    New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/10 10:18 AM
      New Article: Why Apple Won’t ARM the MacBookRichard Cownie2011/05/10 12:10 PM
        Form factors and performanceDavid Kanter2011/05/10 12:39 PM
          Form factors and performanceRichard Cownie2011/05/10 05:25 PM
            Form factors and performanceLinus Torvalds2011/05/10 06:24 PM
              Form factors and performanceRichard cownie2011/05/11 02:12 AM
                Form factors and performanceRichard Cownie2011/05/11 03:44 AM
              Form factors and performanceGabriele Svelto2011/05/11 09:03 AM
                Form factors and performanceLinus Torvalds2011/05/11 10:16 AM
                  Form factors and performanceMark Christiansen2011/05/11 11:06 AM
                  Form factors and performanceRichard Cownie2011/05/11 12:36 PM
              Form factors and performanceDoug Siebert2011/05/11 10:19 AM
                Form factors and performanceRakesh Malik2011/05/11 03:36 PM
            Form factors and performanceFoo_2011/05/11 04:33 AM
              Form factors and performanceRichard Cownie2011/05/11 05:35 AM
              Form factors and performanceJoe Kerr2011/05/11 09:05 PM
    New Article: Why Apple Won’t ARM the MacBookUngo2011/05/10 01:33 PM
      New Article: Why Apple Won’t ARM the MacBookRichard Cownie2011/05/10 04:54 PM
  Maybe after Cortex A15 ?Nicolas2011/05/10 07:07 AM
    Maybe after Cortex A15 ?anonymous422011/05/10 11:01 AM
      Maybe after Cortex A15 ? Nicolas2011/05/11 01:34 PM
  One way for ARM to beat Inteletienne2011/05/10 04:38 PM
  Perhaps ARMed apple netbooks?John H2011/05/16 08:14 PM
    Perhaps ARMed apple netbooks?blah2011/05/17 08:02 PM
      Perhaps ARMed apple netbooks?Brett2011/05/18 02:50 AM
        Perhaps ARMed apple netbooks?Llew2011/05/18 03:43 AM
        Perhaps ARMed apple netbooks?David Kanter2011/05/18 09:46 AM
          Perhaps ARMed apple netbooks?Brett2011/05/18 11:43 AM
            Perhaps ARMed apple netbooks?blah2011/05/18 08:01 PM
            Perhaps ARMed apple netbooks?David Kanter2011/05/18 11:47 PM
              Perhaps ARMed apple netbooks?Rob Thorpe2011/05/19 01:07 AM
                Perhaps ARMed apple netbooks?blah2011/05/19 11:01 AM
                  Perhaps ARMed apple netbooks?Richard Cownie2011/05/19 11:48 AM
              iOS changes from OSXBrett2011/05/21 12:18 PM
                iOS changes from OSXBrett Davis2011/08/26 04:46 PM
              iOS changes from OSX Brett2011/06/03 10:50 AM
  New Article: Why Apple Won’t ARM the MacBookRobi52011/05/20 11:55 PM
    New Article: Why Apple Won’t ARM the MacBookDavid Kanter2011/05/21 01:36 PM
Reply to this Topic
Name:
Email:
Topic:
Body: No Text
How do you spell avocado?