Announcement

Collapse
No announcement yet.

AMD Ryzen 5000 Series (Zen 3) Linux Expectations - Should Be Good But No "Znver3" Compiler Yet

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

  • #11
    Originally posted by oleid View Post

    Probably znver2 will do just fine. I'm wondering, though, what -march=native would do.
    If I understand compiler's optimization flags correctly, -march=native would use all instructions supported by the current processor, and -mtune=... will use micro-arch specific instruction scheduler.
    -march doesn't need any change, but -mtune=, maybe?

    Comment


    • #12
      Who came up with names like znver2, znver3 instead of using zen2, zen3 etc.?

      Comment


      • #13
        Originally posted by zxy_thf View Post
        If I understand compiler's optimization flags correctly, -march=native would use all instructions supported by the current processor, and -mtune=... will use micro-arch specific instruction scheduler.
        -march doesn't need any change, but -mtune=, maybe?
        iirc -mtune is implied by -march, which makes sense.

        Comment


        • #14
          Originally posted by AsuMagic View Post

          iirc -mtune is implied by -march, which makes sense.
          Yeah you're right. I checked the manual and it seems there is no easy way to enable all extensions without micro-arch specific scheduler. (One can surely pass tons of -mxxxx, but I won't call that "easy way")

          Comment


          • #15
            Originally posted by zxy_thf View Post
            If I understand compiler's optimization flags correctly, -march=native would use all instructions supported by the current processor, and -mtune=... will use micro-arch specific instruction scheduler.
            -march doesn't need any change, but -mtune=, maybe?
            -march=XYZ will imply -mtune=XYZ

            Comment


            • #16
              Originally posted by oleid View Post

              -march=XYZ will imply -mtune=XYZ
              Unless you're using -march=native on an unknown architecture where it'll enable what optimizations it can detect, set -mtune=generic, and leave you with binary that should run everywhere and use optimizations if they're available.

              Comment


              • #17
                Can't znver3 be added to a minor release of GCC if it's not too much of a change? That way it shouldn't take that long to go.

                Comment


                • #18
                  Originally posted by geearf View Post
                  Can't znver3 be added to a minor release of GCC if it's not too much of a change? That way it shouldn't take that long to go.
                  For us Slavics, it would be better if they called it "zver3", where "zver" means a Beast in Slavic languages.

                  Comment


                  • #19
                    Judging by past experience of the way AMD devs (remember opensource ?!), and Linux devs incorporate (good drivers, features, ...) for Linux into this AMD Ryzen 5000 series cpu's, the "average" user is looking at maybe November 2021 or later for Linux to incorporate full performance advantage of these 5000 series ryzen cpu's, ..., whereas MS Windows users will have all these issues sorted out 8 months earlier.

                    But, of course by then, Zen 4 will be upon us, ...., and Linux will be adivised to rinse n repeat n wait, again.
                    But that's ok, 'cause when it happens again, and again, and ..., Linux will always blame the private-corporate, patent-pending, copyright, ..., world for letting Linux users down, blah blah ..., and they'll BOTH be correct.
                    Last edited by scjet; 12 October 2020, 08:13 AM.

                    Comment

                    Working...
                    X