Announcement

Collapse
No announcement yet.

Intel Itanium IA-64 Support To Be Deprecated By GCC 10, Planned Removal In GCC 11

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

  • Hugh
    replied
    Technically VLIW was interesting when HP proposed the project. It had been preceded by Fisher's "Bulldog" compiler (PhD research), commercialized by Multiflow. It looked performant for scientific applications. The common wisdom is that commercial workloads are too irregular and branchy (at a small scale) for VLIW to get good traction.

    Commercially, it looked to me as if Itanium succeeded in blowing most RISC architectures out of the market. PA gone. Alpha gone a little later. MIPS retreated to embedded space. SPARC and Power lost mindshare. That damage scorched the earth, making room for x86 when it got good enough. So failure of Itanium was almost irrelevant to Intel's success moving up the box hierarchy.

    The Itanium systems seemed to be designed from the start for server systems: no stupid BIOS history, HA features, ECC, not tied to having a keyboard and VGA display. These looked like the style of engineering around boxes from Sun, IBM, HP, and SGI. PC hardware still hasn't gotten there completely. I think EFI came out of Itanium. Surely there was some second system syndrome too.

    It appeared as if Intel x86 was in mortal combat with AMD and Itanium got few engineering cycles. I seem to remember it was always a generation behind in semiconductor process. No wonder it looked bad.

    Today's x86 processors seem really profligate with transistors. But that seems to be OK when there are so many transistors on a chip. In retrospect, transistor budgets really determined optimal architecture style. For a while it was a no-brainer that RISC was best. Now it seems almost irrelevant for desktops and above.

    Leave a comment:


  • jacob
    replied
    Originally posted by phuclv View Post

    Botched implementation? Maybe. But VLIW is definitely not the worst idea in CPU design. The idea is good at the time where superscalar wasn't mature. Have you ever heard of the Transmeta Crusoe? An "x86" CPU that's smaller, cheaper, faster and much less power hungry than Intel CPUs in its era
    Maybe you mean dynamic OoO execution and register renaming? Superscalar was very mature at the time. Intel had been using it since the first Pentium, several years before the Itanium, and the Pentium itself was already late to the game.

    Of course I've heard of the Crusoe. It has an internal VLIW core, just like most modern CPUs including virtually all current Intels and AMDs. But that's the whole point: it uses VLIW the right way, e.g. dynamically, where the execution slots can be fed from a dispatch queue based on the current context and thus maximise the core utilization. The Itanium idea was to get rid of all the dynamic translation/dispatch "overhead" and expose the naked core, with the assumption that optimisation would instead be done statically ahead of time by the compiler. I don't know if it's just me, but I really don't understand how could someone believe that it was somehow not equivalent to the Turing machine's halting problem? I practice, I've read somewhere that in real world applications (as opposed to benchmarks) the instruction slot utilisation was around 60% at the best of times. So the whole Itanic saga ultimately amounted to offering extremely expensive processors with subpar performance to begin with, and then running them at around 60% of their capabilities. AMD's proposition, on the other hand, was much cheaper processors with excellent performance and backward compatibility on top of it. Put like that it's obvious who won and why.

    Leave a comment:


  • phuclv
    replied
    Originally posted by jacob View Post
    The Itanic will be remembered as one of the worst ideas in CPU design history, with a botched implementation to match.
    Botched implementation? Maybe. But VLIW is definitely not the worst idea in CPU design. The idea is good at the time where superscalar wasn't mature. Have you ever heard of the Transmeta Crusoe? An "x86" CPU that's smaller, cheaper, faster and much less power hungry than Intel CPUs in its era

    Leave a comment:


  • DanL
    replied
    Originally posted by torsionbar28 View Post
    If you recall, Netburst P4's were so terrible, they regularly overheated, couldn't compete with Athlon on performance, and only scaled to 3.8 Ghz max.
    Not exactly. The P4 Northwood core was faster than Athlon XP (T-bred), and it didn't have the overheating problems of its succesor, (Prescott), unless you pumped in too much Voltage when trying to OC.
    Netburst couldn't compete with Athlon64, even when Intel got the TDP under control with the move to 65nm (Cedar Mill/Presler).

    Leave a comment:


  • torsionbar28
    replied
    Originally posted by Bsdisbetter View Post
    Workstations != servers
    Who said anything about workstations? I'm talking about rack mount servers. Both HP and Dell supported Windows and Linux on Itanium rack mount servers. Fact.

    Edit: see for yourself here. Dell supported OS's on the Poweredge 7150 Itanium server include Red Hat Linux:
    https://www.dell.com/support/home/us...e-7150/drivers
    The newer Poweredge 3250 and 7250 with Itanium2 processors supported RHEL v2.1 and v3 as well.

    I'm not going to dig up an equivalent HP link because their site is horrible how after the HP/HPE split, but I can assure you it's the same situation with HP.
    Last edited by torsionbar28; 14 June 2019, 10:03 PM.

    Leave a comment:


  • Bsdisbetter
    replied
    Originally posted by torsionbar28 View Post
    Complete nonsense, where do you get this stuff?? Plenty of Itanium hardware was sold with vendor installed and supported Linux distros. I worked for HP during this time, and I can assure you, RHEL was an officially supported OS on Itanium. Same goes for the Itanium servers from Dell, which were sold with your choice of Linux (RHEL or SLES) or Windows.
    Workstations != servers

    Leave a comment:


  • torsionbar28
    replied
    Originally posted by Bsdisbetter View Post
    It seems there's enterprises and there's enterprises. The sort of server hardware Itanium is in is totally locked down by the manufacturers.
    The only time you would consider using them with *Linux derivatives is when they are sold off and go into private hands. And then you would use NetBSD for that.
    Itanium based servers are a package of software/hardware; you cannot separate the two. You have a choice: black or black.
    Complete nonsense, where do you get this bad info?? Plenty of Itanium hardware was sold with vendor installed and supported Linux distros. I worked for HP during this time, and I can assure you, RHEL was an officially supported OS on Itanium. Same goes for the Itanium servers from Dell, which were sold with your choice of Linux (RHEL or SLES) or Windows.
    Last edited by torsionbar28; 14 June 2019, 09:53 PM.

    Leave a comment:


  • Bsdisbetter
    replied
    Originally posted by jpg44 View Post

    Linux is used in enterprises, but Itanium is sort of a dead platform, so no one is spending time porting new Linux distros to it, just keeping their legacy applications running and migrating off the platform.
    It seems there's enterprises and there's enterprises. The sort of server hardware Itanium is in is totally locked down by the manufacturers.
    The only time you would consider using them with *Linux derivatives is when they are sold off and go into private hands. And then you would use NetBSD for that.
    Itanium based servers are a package of software/hardware; you cannot separate the two. You have a choice: black or black.

    Leave a comment:


  • jpg44
    replied
    Originally posted by torsionbar28 View Post
    "Considering the GCC compiler is used to compile the Linux kernel and IA-64 doesn't enjoy coverage from other compilers like Clang able to build the Linux kernel, it will effectively mean the end of the road for new Linux support moving forward."

    Pretty sure when RHEL and SLES stopped supporting IA-64 a while back, that marked the end of the road for Linux on IA-64. This is enterprise hardware, nobody is running Gentoo on these things lol.
    Linux is used in enterprises, but Itanium is sort of a dead platform, so no one is spending time porting new Linux distros to it, just keeping their legacy applications running and migrating off the platform.

    Leave a comment:


  • jpg44
    replied
    Originally posted by stormcrow View Post

    In ten years you'll still have those containers running on whatever hardware they moved it to, probably with only difference is whatever they've replaced failed hardware in that intervening period. It's also just as likely those old Itanium servers will still be in use because no one in management wants to either spend money on migration, or none of the IT support staff want to wake the sleeping giant of problems trying to replace legacy hardware can cause. Like the IRS is still using decades old IBM mainframes, and companies still using old punch card calculator systems. If it's not broke, don't run the risk of "fixing it" and having a technical catastrophe (failed/inept migration) and/or political nightmare on your head (raising taxes to pay for it and/or major budget increases to management-voters-shareholders).

    https://www.nextgov.com/cio-briefing...rnment/128599/
    https://www.pcworld.com/article/2499...use-today.html

    If these aren't public facing systems, there's usually not any real risk to leaving them alone other than hardware failure.
    IRS does not use decades old mainframe. IBM has a migration path, they migrated to more recent IBM hardware thats backward compatable

    Leave a comment:

Working...
X