Announcement

Collapse
No announcement yet.

Loongson Continues Working On LoongArch For Linux, But It's Mostly Copying MIPS Code

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

  • #11
    Originally posted by torsionbar28 View Post
    Who cares about Loongson though, it's effectively a China-domestic product [...] Not sure why Linux should bother mainlining anything like this, has no practical value to anyone outside of China..
    This is kind of what open-source, GNU and Linux is all about.

    If Linux can support the old N64 hardware, why not an internal Chinese processor which has many more users? Also, why should Apple have their shite internal processor supported but not one from a Chinese company?

    Yes, useless for many (i.e getting hold of a Loongson laptop was extremely difficult for me) but Linux really is a big bargain bucket of hardware support. It seems a bit messy but ultimately is one of its major strengths.

    Plus looking at some reviews such as this: https://www.techspot.com/news/90544-...rst-ryzen.html
    It is certainly looking "good enough" as a viable day to day processor.
    Last edited by kpedersen; 25 August 2021, 11:08 AM.

    Comment


    • #12
      Originally posted by s_j_newbury View Post
      but Loongson calling it MIPS obviously has legal issues.
      Does it? And to what extent? Simply implementing it as a processor in the kernel's MIPS family isn't the same as commercially advertising it as a MIPS processor.

      Also potentially relevant:

      In December 2018, Wave Computing, the new owner of the MIPS architecture, announced that MIPS ISA would be open-sourced in a program dubbed the MIPS Open initiative. The program was intended to open up access to the most recent versions of both the 32-bit and 64-bit designs making them available without any licensing or royalty fees as well as granting participants licenses to existing MIPS patents.

      Source: https://en.wikipedia.org/wiki/MIPS_a...re#cite_ref-18
      Last edited by coder; 25 August 2021, 11:09 AM.

      Comment


      • #13
        Originally posted by torsionbar28 View Post
        Who cares about Loongson though, it's effectively a China-domestic product, and it's slow and buggy. Not sure why Linux should bother mainlining anything like this, has no practical value to anyone outside of China. It is all a political move from the CCP, part of their "Made in China 2025" initiative,
        I think such political considerations are outside the Linux charter. If there's a platform that meets all the technical qualifications, and there's someone to maintain it, then it should be on equal grounds for inclusion as anything else.

        Comment


        • #14
          Originally posted by ddriver View Post
          Really? You find it hard to figure out a set of instructions that may be required for a computer to run?
          Yes, actually.

          You clearly haven't written a compiler or done any substantial degree of kernel development. There's a lot more that goes into an ISA than deciding what to name the instructions, how many operands they should have, and choosing a binary instruction format.

          Comment


          • #15
            Originally posted by coder View Post
            Yes, actually.

            You clearly haven't written a compiler or done any substantial degree of kernel development. There's a lot more that goes into an ISA than deciding what to name the instructions, how many operands they should have, and choosing a binary instruction format.
            You clearly have no idea of having an idea who has done what. You are confusing ISA design with hardware and software implementation of a compliant system. The ISA is clearly defined as a specification that is not bound to any particular implementation. And both hardware and software implementations are entirely different fields that have absolutely nothing to do with ISA design.

            Also, I currently happen to work on a tool that allows to define and simulate different hardware platforms and generate native assemblies out of the definitions and IR code format that I have personally designed. So I think I can do one better than "write a compiler" - writing a compilers compiler.

            A design that can conform abstract instructions to ANY ISA, a wee bit more complicated than designing any ISA. Also, requires a fair amount of existing ISA knowledge. You may not be able to appreciate the gravity here.
            Last edited by ddriver; 25 August 2021, 11:37 AM.

            Comment


            • #16
              Originally posted by ddriver View Post
              You are confusing ISA design with hardware and software implementation of a compliant system.
              No, I'm not. The ISA dictates a lot about how the kernel, compilers, and libraries interact with the CPUs implementing it. TLBs, interrupt-handling, memory-consistency, cache control, atomics, paging, context switching, etc. all exist as part of the ISA definition.

              Originally posted by ddriver View Post
              I currently happen to work on a tool that allows to define and simulate different hardware platforms and generate native assemblies out of the definitions and IR code format that I have personally designed. So I think I can do one better than "write a compiler" - writing a compilers compiler.
              And does anyone actually use your toy, at least for anything beyond microcontrollers?

              Comment


              • #17
                Originally posted by Pheckphul View Post
                Of course it's a MIPS ripoff. China's tech industry still hasn't risen to the level of inventing from whole cloth anything that can compete with x86/POWER/ARM. They haven't funneled enough stolen IP back to the fatherland to accomplish that yet. 🤡
                I don't think you can call it a ripoff. Loongson Technology Corp. Ltd. have licensed the MIPS32 and MIPS64 architectures from MIPS Technologies, Inc. fair and square. I don't know why they don't want to call it MIPS, maybe because they've been evolving it.

                Comment


                • #18
                  Originally posted by phoronix View Post
                  Phoronix: Loongson Continues Working On LoongArch For Linux, But It's Mostly Copying MIPS Code
                  In danger of stating the obvious, the Chinese are stereotypically known for copying all the things.
                  But maybe there is something a bit more than just the base MIPS ISA? AMD processors of today can still run i386 code, but somehow the world stopped calling it i386 in 2003 because they added long mode and a bunch of instructions. Ccome to speak of it, the (somewhat short-lived) k1om/l1om architectures was basically Pentium plus a bit on top too.

                  Comment


                  • #19
                    Originally posted by ddriver View Post

                    Hello, it is 2021 calling...

                    Today it is fairly easy to roll out a toolchain. Heck, it is getting "pedestrian" level to design semiconductors, with modern design tools, you can simply import and interconnect IP blocks and have that manufactured if you can afford to.

                    Today it would be stupid to resort to anything legacy, much less bother to reverse-engineer a complex design, not for the sake of manufacturing and selling products anyway. It is so much cleaner and easier to start from scratch.
                    Ignoring that all of that only looks easy if you've never done it yourself, it also costs an enormous amount of man power just to shuffle the paper around. This is why the Linux kernel devs should be resistant to merging another architecture which is just a duplication of the MIPS code.
                    Last edited by Developer12; 25 August 2021, 08:49 PM.

                    Comment


                    • #20
                      Originally posted by coder View Post
                      No, I'm not. The ISA dictates a lot about how the kernel, compilers, and libraries interact with the CPUs implementing it. TLBs, interrupt-handling, memory-consistency, cache control, atomics, paging, context switching, etc. all exist as part of the ISA definition.

                      And does anyone actually use your toy, at least for anything beyond microcontrollers?
                      What you are overlooking is the component of legacy. Computers have changed quite a lot since the olden days, what people use computers for has changed even more so. Your basic grasp on ISA design appears to be vaguely based on all the mud legacy ISAs had to be dragged through to accommodate contemporary computing requirements, rather than on what a contemporary ISA needs to be.

                      Today you don't design an ISA in the dark about what the future may bring, at this point computing is a mature subject, it has peaked, and it is more available than ever, why we get over a dozen new hardware architectures every year. You can scrap all the legacy garbage, which was introduced for reasons that are no longer relevant. This eliminates a lot of complexity and overheads. What might look over-complicated from the perspective of a basic "coder" is more akin muscle memory for a proficient system designer.

                      Kinda like doing practically anything - those that have never done it struggle incapable to do it for a long, long time, whereas to those that have done it enough to attain proficiency it is a triviality. It would take a very modest skill set, in terms of both diversity and level of proficiency, to somehow remain ignorant of that little fact, as you evidently have. You are but an infant struggling to walk more than a couple of steps before tumbling down, thus the concept of running seems like something extremely advanced, for reasons unknown to an infant.

                      As for my "toy" - it can actually model and simulate an arbitrary level of features, complexity and optimizations for every platform. It will either find a way to accomplish an abstract task given the target platform capabilities or request you to provide a custom implementation. And again, do not expect a self-proclaimed "coder" to have the level of knowledge and experience to fully appreciate that, as you clearly don't.

                      So I am kind of curious as to what your work involves, in order to put you on a standing where you get to evaluate my work as a toy? Care to disclose your expertise? Unless of course, you were just being condescending because you don't have anything better.

                      Comment

                      Working...
                      X