Announcement

Collapse
No announcement yet.

Intel Sapphire Rapids Target Added To LLVM/Clang 12.0

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Intel Sapphire Rapids Target Added To LLVM/Clang 12.0

    Phoronix: Intel Sapphire Rapids Target Added To LLVM/Clang 12.0

    Intel developers engaging with upstream LLVM have been adding AMX support and other new features for next year's Xeon "Sapphire Rapids" while as of a few days ago in LLVM 12 Git is the actual enabling of -march=sapphirerapids support...

    http://www.phoronix.com/scan.php?pag...Rapids-LLVM-12

  • #2
    AMX extensions....LOL. Isn't that cute. Instruction extensions to vector SIMD registers that have been around since...what... the MMX days? And we won't even get these "kludges" duct taped onto an over priced Xeon until MAYBE the end of 2021 ?? Which really means mass production and in use by the end of 2022.

    And 10nm ?? HAHAHAHAHAHA !!!

    But today you can buy a $399 Apple iPhone SE with a 7nm Apple Silicon A13 SoC that has an ACTUAL hardware Advanced Matrix math processor seperate from its own ARM based Vector SIMD unit.

    And in October the 5nm Apple Silicon A14 comes out for iPhones, iPads and an Apple Macbook probably in December that will have an improved Advanced Matrix math processor among other things AND will be competitive wirh Intel Ice Lake and AMD Zen 2.

    x86 is tired and played out. It's easier to add powerful features to an already efficient and power sipping design than to try to take an old CISC hased architecture and make it efficient and power sipping AND STILL retain high performance.
    Last edited by Jumbotron; 09-01-2020, 01:51 PM.

    Comment


    • #3
      Originally posted by Jumbotron View Post
      But today you can buy a $399 Apple iPhone SE with a 7nm Apple Silicon A13 SoC that has an ACTUAL hardware Advanced Matrix math processor seperate from its own ARM based Vector SIMD unit.

      And in October the 5nm Apple Silicon A14 comes out for iPhones, iPads and an Apple Macbook probably in December that will have an improved Advanced Matrix math processor among other things AND will be competitive wirh Intel Ice Lake and AMD Zen 2.
      As long as you stick to using iOS or macOS for the rest of your life, pay $99 per year if using iOS, stick to either Objective-C or Swift, stick to Xcode and accept Apple's severe lock-in measures.




      ...


      You could just buy an RTX 2060 for less than $399 and have Tensor cores for some degree of matrix processing (of course, as long as you are OK with their proprietary drivers).

      Comment


      • #4
        Yeah...you could go the above route. And then compare your power bill at the end of the year. LOL!

        Oh..just to let you know my uniformed friend. Apple will be introducing ARM based Apple Silicon across their entire hardware stack including Mac Pro's. Every piece of hardware will be specced out with a separate Neural Net Processor, a separate Matrix Math processor, a separate Vector unit, and DSP. And a full suite of sensors. At 5nm and less while AMD STILL can't get their Sensor Hub shit straight AND STILL are not feature or performance full concerning Intel AVX or AVX2 instructions as opposed to Intel. And Intel can't even get their shit straight with 10nm. Also Microsoft is going ARM as well...just a little behind Apple. As they always do they just to take a look at what Apple is doing and then tell their "engineers"..."Uhhh, can you guys just copy THAT for us?" LOL !!

        Tick Tock Linux and Linux software makers. Better start porting over AND OPTIMIZING for ARM ASAP !

        Comment


        • #5
          Originally posted by Jumbotron View Post
          Yeah...you could go the above route. And then compare your power bill at the end of the year. LOL!

          Oh..just to let you know my uniformed friend. Apple will be introducing ARM based Apple Silicon across their entire hardware stack including Mac Pro's. Every piece of hardware will be specced out with a separate Neural Net Processor, a separate Matrix Math processor, a separate Vector unit, and DSP. And a full suite of sensors. At 5nm and less while AMD STILL can't get their Sensor Hub shit straight AND STILL are not feature or performance full concerning Intel AVX or AVX2 instructions as opposed to Intel. And Intel can't even get their shit straight with 10nm. Also Microsoft is going ARM as well...just a little behind Apple. As they always do they just to take a look at what Apple is doing and then tell their "engineers"..."Uhhh, can you guys just copy THAT for us?" LOL !!

          Tick Tock Linux and Linux software makers. Better start porting over AND OPTIMIZING for ARM ASAP !
          Take a USB stick and install RHEL on your precious little iPhone and then do a few trillion inferences faster than Sapphire Rapids. Then stop shilling for a company that makes Microsoft look like a champion of open source and open software systems. I don't come here to hear d-bag losers like you dump on the #1 contributor to the Linux kernel while actively shilling for a company that literally enslaves both its Chinese throw-aways and its own "customers".

          Comment


          • #6
            "on top of Ice Lake Server..." ??
            Ice Lake Server has no amx.

            Comment


            • #7
              Originally posted by jayN View Post
              "on top of Ice Lake Server..." ??
              Ice Lake Server has no amx.
              Referring to 'in addition' to what Ice Lake server exposes.
              Michael Larabel
              http://www.michaellarabel.com/

              Comment

              Working...
              X