Announcement

Collapse
No announcement yet.

Linux 4.17 Spring Cleaning To Drop Some Old CPU Architectures

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

  • #21
    Originally posted by schmidtbag View Post
    I'd like to think I'm decently aware of various CPU architectures, but aside from OpenRISC, I'm surprised I haven't heard of any of these.
    I've worked with OpenRISC both Verilog (OR1200) and physical implementation. I don't mind dumping OpenRISC.
    It's ISA served no real purpose since all implementation was and always has been utter crap. Nobody with time or money took OpenRISC under it's wings like RISC-V.
    So it's pretty much dead. Not that there is something intrinsically wrong with the ISA though. Most ISA's have issues. OpenRISC is pretty clean.
    I had serveral devboards with Blackfins. They where pretty popular for small scale DSP's some 15 years ago.
    I have also worked with CRIS. It's a 16-bit opspace RISC if I remember correctly. Address is 32-bit. Developed by Axis Communications in Sweden. Obviously used in all their cameras, printservers etc etc. I think Cannon and other printer companies used them too for ethernet addon-cards with postscript capabilities and stuff like that.
    The others I believe were just esoteric oddballs.

    Comment


    • #22
      Originally posted by Azrael5 View Post
      it's time to drop all the cpus based on 32bit architecture.
      Not yet, at least until all the POS hardware based on the Geode GX/LX die out.
      Pure Intel 80386 support was removed a few years ago but 486+ is still supported and should be for some time.
      Also, a lot of new ARM chips are still 32-bit and a lot of applications do not even need that kind of address space.
      Heck, even Z80 CPUs are still manufactured and used today.

      Comment


      • #23
        Originally posted by onokun View Post


        Loongson is so 2008, should drop it too. There is no contemporary architecture for Linux, only obsolete ones.
        So, the one true architecture is the Turing machine with a single peripheral, a tape. Drop all drivers, too, except SCSI/ATAPI streamers. :-P

        Comment


        • #24
          Originally posted by deant View Post
          It is great decision. i have always been fan of cleanng and optimising. I hope we can get smaller footprint and smaller/faster kernel.
          How do you think this will influence your kernel's performance? Do you compile support for drivers and architectures into your kernel that you're never going to have on your x86 machine?

          WTF?
          Stop TCPA, stupid software patents and corrupt politicians!

          Comment


          • #25
            Originally posted by Azrael5 View Post
            it's time to drop all the cpus based on 32bit architecture.
            No. You probably don't have a clue. And you don't have to use things in the kernel that are not available on your machine so there is no need to complain about other things in the kernel source.
            You are NOT forced to enable and compile them. But you folks argue like you were.
            Stop TCPA, stupid software patents and corrupt politicians!

            Comment


            • #26
              Originally posted by monraaf View Post

              The code being removed is specific to the architectures mentioned in the article. It's never compiled into any Ubuntu kernels as Ubuntu doesn't support any of these architectures.
              So? I responded to the reply of a reply that was talking about smaller kernel sizes in general, not just about this code being removed. Also, Ubuntu was just an example.

              Comment


              • #27
                Originally posted by milkylainen View Post

                I've worked with OpenRISC both Verilog (OR1200) and physical implementation. I don't mind dumping OpenRISC.
                It's ISA served no real purpose since all implementation was and always has been utter crap. Nobody with time or money took OpenRISC under it's wings like RISC-V.
                So it's pretty much dead. Not that there is something intrinsically wrong with the ISA though. Most ISA's have issues. OpenRISC is pretty clean.
                I had serveral devboards with Blackfins. They where pretty popular for small scale DSP's some 15 years ago.
                I have also worked with CRIS. It's a 16-bit opspace RISC if I remember correctly. Address is 32-bit. Developed by Axis Communications in Sweden. Obviously used in all their cameras, printservers etc etc. I think Cannon and other printer companies used them too for ethernet addon-cards with postscript capabilities and stuff like that.
                The others I believe were just esoteric oddballs.
                But if Axis is still using CRIS (and Axis is a pretty popular brand for webcams at least), then why is the kernel team dropping support for it? I could understand if Axis was a very small company and/or they were not using CRIS anymore, but they are not small at all and they might still be using CRIS...

                Comment


                • #28
                  Originally posted by Azrael5 View Post
                  it's time to drop all the cpus based on 32bit architecture.
                  Are you trolling, or just clueless?

                  Comment


                  • #29
                    Originally posted by Azrael5 View Post
                    it's time to drop all the cpus based on 32bit architecture.
                    Lol don't post bullshit. The overwhelming majority of embedded devices uses 32-bit processors, and will not really need more than that for a long time (most 64-bit parts are also run in 32-bit mode because they don't have more than 4GB of RAM).

                    Comment


                    • #30
                      Originally posted by deant View Post
                      It is great decision. i have always been fan of cleanng and optimising. I hope we can get smaller footprint and smaller/faster kernel.
                      your kernel does not include support for those arches and isn't running that code. so you will see zero difference. it only makes difference for kernel maintainers

                      Comment

                      Working...
                      X