Announcement

Collapse
No announcement yet.

RISC-V With Linux 6.8 Restores XIP Kernel Support

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

  • RISC-V With Linux 6.8 Restores XIP Kernel Support

    Phoronix: RISC-V With Linux 6.8 Restores XIP Kernel Support

    With Linus Torvalds back to work, merged to mainline on Wednesday were the RISC-V architecture updates for the in-development Linux 6.8 kernel cycle...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    It might be a silly question - but what is the problem here? The NOR Flash shows somewhere in the physical address room, the CPU starts execution from there. If there is no hard coded addresses (aka: PIE) in the kernel this should work just fine.
    That can be even done on normal PCs - ok, one should try a bit older one (P2/P3 could be feasible): a stub ROM (instead of the BIOS) at 0xC0000 from which the CPU starts, and then a jump to another ROM in the adress space (e.g. a big Boot ROM in a PCI-Slot) which is big enough to hold the whole kernel (and perhaps a R/O file system). Ok, the stub ROM should at least init the RAM, but on these older machines this is quite simple.

    Comment


    • #3
      Intel please make a RISC-V board with Intel Xe graphics and Intel Ethernet, with audio and Wi-Fi by Intel.

      Comment


      • #4
        Why should intel give up voluntarily their patented x86-64 cash cow that secured their overpriced products and duopoly-like position for decades? I don't see that happening as even now they are still trying to beat arm in the small form factor sector despite having the more power hungry x86-64 chips. In 2021 intel tried to buy SiFive for 2 billion probably to make them disappear as a driving force behind RISC-V just as they did with transmeta. And as for horse creek i doubt intel's motivations for making a serious and competitive product. After less than half a year intel quietly killed the RISC-V pathfinder dev environment. No, i think the revolution may come from completely different companies who are hungry for real innovation in a more fair market.

        Comment


        • #5
          Originally posted by uid313 View Post
          Intel Ethernet
          Hell no.

          Comment


          • #6
            Originally posted by M.Bahr View Post
            Why should intel give up voluntarily their patented x86-64 cash cow that secured their overpriced products and duopoly-like position for decades? I don't see that happening as even now they are still trying to beat arm in the small form factor sector despite having the more power hungry x86-64 chips. In 2021 intel tried to buy SiFive for 2 billion probably to make them disappear as a driving force behind RISC-V just as they did with transmeta. And as for horse creek i doubt intel's motivations for making a serious and competitive product. After less than half a year intel quietly killed the RISC-V pathfinder dev environment. No, i think the revolution may come from completely different companies who are hungry for real innovation in a more fair market.
            Because if Intel sells a RISC-V board then I am buying it!

            If not, well then they risk being left behind with unattractive products and having their marketshare taken by the competition. I don't find x86 products interesting because they are inefficient so I am looking to buy ARM or RISC-V products instead.

            Consumers see how great the Apple Macbooks powered by Apple M1, M2 and M3 are. Microsoft is noticing it too and optimizing Windows for ARM and partnering with Qualcomm. I guess HP, Dell, and Lenovo will notice it too.

            Comment


            • #7
              Originally posted by M.Bahr View Post
              Why should intel give up voluntarily their patented x86-64 cash cow that secured their overpriced products and duopoly-like position for decades?
              You bring up a fair question. Let me ask you equally fair ones:
              • Why should Nokia give up voluntarily their patented 3320 phone cash cow that secured their overpriced products and duopoly-like position for decades?
              • Why should Kodak give up voluntarily their patented film-based camera cash cow that secured their overpriced products and duopoly-like position for decades?
              • Why should Ford give up voluntarily their patented petrol engine cash cow that secured their overpriced products and duopoly-like position for decades?
              • Why should Blockbuster give up voluntarily their patented rental store cash cow that secured their overpriced products and duopoly-like position for decades?
              And so on, and so forth. It's called Disruption.

              Comment


              • #8
                I am still not convinced that intel would give up their ISA cash cow and follow RISC-V as a "Disruptive innovation". And one of the many strong proofs for that are new patented extensions again. Those are added to the x86-64 basis to keep the dependency on it alive. This prevents the competition to copy the ISA fully but only older patents which are running out and become obsolete.

                Instead of breaking that circle i sadly see even linux distros following the example of windows by lifting the ISA requirements for the OS. The plans are to compile distro packages from x86-64-2v up as some here may know. This is not only counter-productive to the spirit of FLOSS but i even dare to say it is subversive. There is neglible to almost no performance benefit in doing so. It cuts off a big user base with older hardware. This decision mainly assists intel and amd to keep customers in their ISA vendor-lock-in for probably more decades to come.

                New ISAs don't have automatically disruptive potential by simply being better. The adoption rate is very slow and the ISA needs to be backwards compatible. We saw this on many examples in history like with the Z80, transmeta chips and even newer products from amd and intel themselves. The users wouldn't have moved to amd64 (x86-64) if there was no backwards compatibility to x86. Intel failed completely in establishing itanium as a successor. Even apple despite being able to simply dictate an ISA change by their m chips offer backwards compatibility in form of rosetta 2.

                All in all i see many obstacles for a real disruptive innovation but foremost a big chicken-or-the-egg problem protected by two gigantic gate keepers. As i said in my initial comment i rather think the revolution may come from different companies. Intel and amd may follow a global RISC-V success in the end but they wouldn't do it voluntarily.

                Comment


                • #9
                  Originally posted by uid313 View Post

                  Because if Intel sells a RISC-V board then I am buying it!

                  If not, well then they risk being left behind with unattractive products and having their marketshare taken by the competition. I don't find x86 products interesting because they are inefficient so I am looking to buy ARM or RISC-V products instead.

                  Consumers see how great the Apple Macbooks powered by Apple M1, M2 and M3 are. Microsoft is noticing it too and optimizing Windows for ARM and partnering with Qualcomm. I guess HP, Dell, and Lenovo will notice it too.
                  Apple Silicon is not more efficientl because of Arm. It is because they are always a node ahead of the rest, and because they use on-chip memory, multiple accelerators for compute heavy stuff, etc.

                  The accelerators route is a really good one, however, by having on-chip RAM, these processors are only useful for a select group of computers. You cannot scale M3 to be in Ryzen APUs, Ryzen desktop, ThreadRipper and EPYC. You'd need to design every variation of the CPU, and the cost would be absurd. In contrast AMD's are a bit less efficient, while being easy to scale by using chiplets with a separate IO die (except for mobile, shere they are competitive with Apple's). Each comes with its own tradeoffs.

                  Comment

                  Working...
                  X