Announcement

Collapse
No announcement yet.

The State Of Debian & Fedora On The RISC-V Architecture

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

  • The State Of Debian & Fedora On The RISC-V Architecture

    Phoronix: The State Of Debian & Fedora On The RISC-V Architecture

    RISC-V remains of a lot of interest to open-source/Linux users for being a royalty-free and completely open CPU architecture. In part due to the lack of affordable RISC-V hardware limiting developers from working more on this architecture, the state of RISC-V support by Linux distributions varies but at least has improved a lot in recent years...

    http://www.phoronix.com/scan.php?pag...ntu-State-2019

  • #2
    Typo:

    Originally posted by phoronix View Post
    about this open procesosr ISA and various software

    Comment


    • #3
      If risc-v ends up implementing support for posit number format in hardware and those linux distributions start supporting it, my bet is that risc-v can become the next big thing. See https://posithub.org/docs/RISC-V/RISC-V-for-posits.pdf for details.

      Comment


      • #4
        phoronix My understanding from the slides is that the number of successfully built fedora packages is much higher than 20%. It looks like 20% of packages were built in a single week, according to page 18 of the PDF, unless I'm reading it wrong?

        Comment


        • #5
          Author of Fedora/RISCV talk here. Hi!

          Correct, our best week was ~4500 successful builds. This was after we started to use GCC Compile Farm. In general we have significant portion of Fedora packages built, but the numbers of packages included in Fedora is constantly changing (new being added and non-maintained ones being removed). Most likely the package you are looking is already available in the repository. We can even run GNOME Desktop w/ Wayland!

          Next major things will be main server(s) migration (Koji infrastructure) to new server(s) and preparation for a new mass rebuild with GCC 9.

          Comment


          • #6
            Originally posted by pegasus View Post
            If risc-v ends up implementing support for posit number format in hardware and those linux distributions start supporting it, my bet is that risc-v can become the next big thing. See https://posithub.org/docs/RISC-V/RISC-V-for-posits.pdf for details.
            Wow. THAT topic was a most interesting rabbit hole to fall down! Much thanks, pegasus !

            Comment


            • #7
              risc-v is nice for been simpler and open..
              But now with Open Mips32.. I don'tknow..

              Because mips owns the best lowpower 32 bits cpu out there, and now its also open..

              Comment


              • #8
                Originally posted by tuxd3v View Post
                Because mips owns the best lowpower 32 bits cpu out there,
                [citation needed]

                Comment


                • #9
                  Originally posted by starshipeleven View Post
                  [citation needed]
                  Some examples:
                  1) Directly from Baikal
                  2) I think more recent benchmarks: https://medium.com/@malafeev/first-i...d-bc7c1db12046
                  3) "Per-clock, the P5600 also scored ahead of the Cortex-A15, which I believe is the closest competitor in the category of the P5600. "
                  ("Those g++7.3 results positioned the P5600 firmly between the AMD A8-7600 and the Intel Core2 Duo P8600 in the clock-normalized Mandelbrot performance charts...")

                  Comment


                  • #10
                    Originally posted by tuxd3v View Post

                    Some examples:
                    1) Directly from Baikal
                    2) I think more recent benchmarks: https://medium.com/@malafeev/first-i...d-bc7c1db12046
                    3) "Per-clock, the P5600 also scored ahead of the Cortex-A15, which I believe is the closest competitor in the category of the P5600. "
                    ("Those g++7.3 results positioned the P5600 firmly between the AMD A8-7600 and the Intel Core2 Duo P8600 in the clock-normalized Mandelbrot performance charts...")
                    Baikal is the main manufacturer so I'm not going to believe them at face value.

                    The second source does still show ARM and quite old x86 hardware (Pentium 4 or Phenom II) beating it.

                    Your third source actually states the same in a part that isn't taken out of context like you did:

                    Per-clock, the P5600 also scored ahead of the Cortex-A15, which I believe is the closest competitor in the category of the P5600. Where the P5600, or perhaps its incarnation in the Baikal T1, fell short, was in absolute performance due to low clocks. Should that core reach clocks closer to 2GHz, we’d be seeing much more interesting absolute-performance results.

                    Using my limited micro-benchmark set as a basis, that uarchitecture does largely deliver on its promises of good gen-purposes IPC and good SIMD throughput per clock, and could be considered a direct competitor to the best of 32-bit ARM Cortex designs. That said, Baikal T1 could use higher clocks, which would position it in absolute-performance terms right in the group of the Core2 lineup by Intel and the Cortex-A12/15/17 lineup by ARM.

                    That's not really a statement of its superiority by any stretch of the imagination.
                    He says that per-clock it is superior but since it is run at low clock speed it's not better.

                    To be superior to ARM it would have to boost its power consumption quite a bit, as that's what happens when you boost clock speed. Then it's no more "low power".

                    I'm not saying it's bad, all things considered it is a very good result for a small company, but it's not really "the best 32bit low power CPU".

                    You probably won't accept this truth because it's russian and russian stuff is always best for russian people, but that's your own problem.

                    Comment

                    Working...
                    X