Announcement

Collapse
No announcement yet.

Linux Looks To Retire Itanium/IA64 Support

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

  • #41
    Originally posted by D-Nis View Post
    and they're in some kind of partnership with SiFive.
    They did try to acquire SiFive back in the day, but due to financial trouble Intel has stopped RISC-V work.

    Comment


    • #42
      Originally posted by coder View Post
      IA64 could employ modern techniques like branch-prediction, speculative execution, and limited out-of-order execution, but these avenues were never explored. Nor did they ever augment the instruction set with vector extensions, like SSE or AVX.
      One could do that, but at that point why bother with all that VLIW-inspired jazz of instruction bundles, extensive predication etc. that tended to massively bloat Itanium code? Just do a relatively straightforward RISC like aarch64 or RISC-V.

      Comment


      • #43
        Originally posted by stormcrow View Post
        There is one thing going for OpenVMS, though. Almost no one is going to know how to write malware for it! I'm positive it can be done, just there's not enough people out there versed in VMS internals to make it a likely target. There is at least some minor comfort in using an obscure OS from a security standpoint. Script kiddies will never be after you. Only the very highly skilled and resourced.
        All that means is that you're not a low-hanging fruit for ransomware gangs. However, people running it might be sufficiently juicy targets that it's worth attackers' time to put in the work. Government agencies, likely on both sides.

        Comment


        • #44
          Originally posted by Mark Rose View Post
          due to financial trouble Intel has stopped RISC-V work.
          That's an overstatement/misreading of what actually happened.

          Intel's Pathfinder program was basically like a SDK that helped people who wanted to build a RISC-V SoC prototype their software and select compatible hardware IP. It was probably a big help to the RISC-V market, when they launched it, but Intel claims there are now solutions from other industry players that cover a lot of those bases.

          Intel did not cancel Horse Creek, which is far more consequential. From what's externally visible, Intel seems no less committed to RISC-V, overall, than before. FWIW, I'm not sure how much committed that actually was/is.

          Comment


          • #45
            Originally posted by jabl View Post
            One could do that, but at that point why bother with all that VLIW-inspired jazz of instruction bundles, extensive predication etc. that tended to massively bloat Itanium code? Just do a relatively straightforward RISC like aarch64 or RISC-V.
            The benefit would be to make data-dependencies cheaper to detect. That said, I'm not sure how much die area or energy are currently spent on detection of data dependencies, so maybe that's why it hasn't been copied by AArch64, for instance.

            FWIW, it does seem as if there's a spiritual relative (some say cousin, not descendant) in ELBRUS. If anyone can share some insight here, please do.

            Tachyum had originally announced plans to use VLIW, but it seems they've gone a more conventional route:
            Last edited by coder; 15 February 2023, 02:41 PM.

            Comment


            • #46
              Originally posted by coder View Post
              You mean legacy 32-bit x86? There are still some currently-manufactured microcontrollers that run it.

              Given that probably millions of time as many 32-bit x86 CPUs shipped as IA64, I'd say 32-bit x86 support should probably hang on a little bit longer.
              I don't understand why these microcontrollers would try to run a 6.x kernel. They should use ancient and stable 3.x/4.x LTS kernels

              Comment


              • #47
                Originally posted by edxposed View Post
                I don't understand why these microcontrollers would try to run a 6.x kernel. They should use ancient and stable 3.x/4.x LTS kernels
                Uh, security, features, software compatibility? I dunno. I'm not even saying people run Linux on them, as I don't really know. Just mentioning they exist.

                These claim to support Linux, but I haven't investigated to see which kernels/distros/etc.:

                Comment


                • #48
                  Originally posted by coder View Post

                  i860 was promising but I heard Intel cancelled its successor after they realized x86 was far more profitable for them.
                  More because other RISC designs by competitors were in better market and performance positions and Intel's resources were spread out too much. Intel transitioned to ARM based XScale instead and canceled several alternative processor tech products. XScale would eventually be sold off to Marvel in the mid 00s as neither a major success nor a complete flop. The irony is that i960 (only related to the i860 in that both were RISC based) would be very successful in comparison and eventually be discontinued soon after XScale was sold.

                  Comment


                  • #49
                    Originally posted by coder View Post
                    Probably the worst legacy of IA64 is the lesson Intel seems to have taken to stay away from anything non-x86. I suspect that influenced their decision to base Larrabee on x86, rather than their iGPU ISA. Had they made a different choice, the past decade of GPU development might've looked very different.

                    The curse of IA64 might still haunt Intel. I predict they'll keep all their CPU eggs in the x86 basket for too long.
                    I would argue that it's probably deeper than that, because at the time there seemed to be a huge assumption that Intel could do everything on its own.

                    It took a lot of soul searching but it finally seems to have sunk in to Intel that they can't. I honestly think that at the engineering table that bit of hubris compounded with the fact that Intel used x86 for everything, meanwhile the world was evolving all around it in the semiconductor space.
                    It's now at a point where several producers of lead silicon do not need Intel for anything.

                    Comment


                    • #50
                      Originally posted by jabl View Post

                      HP-UX is an old-school unix. I think originally it ran on M68k(?) before switching over to HP's own RISC architecture, PA-RISC.
                      I can confirm HPUX 8.X ran on M68000 based HP hardware. I had such a system attached to an old mass spec in the mid '90s. From memory it was BSD based compared with HPUX 9.X running on the Apollo Domain PARISC 1.X (700 series) workstations which was more System V.3 and more so with HPUX 10.X. The M68000 might have been HP1000 hardware?

                      All a bit vague now - only remember it because I rolled up with a HPUX 10 install CD to upgrade the OS

                      Comment

                      Working...
                      X