In recent years the Itanium support in the Linux kernel has went downhill with not many users left testing new kernels on aging Itanium servers. There also hasn’t been any major active contributors to the Itanium code for keeping it maintained and making any serious improvements to the architecture code. On and off for months there’s been talk of retiring Itanium from the Linux kernel and now it’s finally happened.
With Linux 6.6 expected to be this year’s Long-Term Support (LTS) kernel version, there was the proposal recently to drop Itanium in Linux 6.7 and indeed it’s successfully happened.
This is a complete outrage, and a sign Torvalds has completely lost the plot. Itanium is the future, and dropping it from the Linux kernel will be its death knell.
I’m going back to DOS.
Oh no! Anyway….
In all seriousness, this is one of those things that seems indifferent now, but we’ll look back on it in 10 years and say “well that was a shame”
Itanium was an interesting idea. Totally misguided, and reppeated a lot of mistakes. But interesting. In this ARM/x86 dominated world, it’s always sad to see another ISA fall by the wayside.
Itanium probably deserved it, but it’s sad all the same.
The123king,
I agree. It’s the pragmatic thing to do, but a lot of people cherish old obscure tech as a hobby. Thom has spouted sarcasm here, but I think it’s fair to note that even he has the “old tech bug”.
The problem with Itanium is that the tasks that were supposed to be decided at compile-time in order to make PA-RISC work performantly could not be decided at compile-time (not in the real world anyway). Nobody was into Itanium for the performance of its implementations (there wasn’t much), they were into Itanium because they were locked into HP-UX.
If I had to lament an architecture, it would be MIPS (aka the architecture of the SGI IRIX workstations).
kurkosdr,
This could have worked even with real applications, except compilers were not designed to do this. So it never took off. As a result it became known for notoriously bad performance running typical software that wasn’t hand tuned for itanium. For a long time, the same could be said of auto-vectorized SSE on x86 too. Real performance gains were only achievable with hand tuned algorithms and typical software didn’t really benefit until compilers slowly caught up. Obviously a big difference is that x86 was already dominant and not going anywhere. x86 had no trouble getting more investment “we’re already using x86 everywhere, let us make full use of it”, whereas investing in itanium was a huge risk for such a niche market.
Yeah, I think the alpha had merit over itanium too. Intel and HP were much stronger companies though and for better or worse that can matter more than merit.
Agreed though I am not going to ask others to put in the work for such limited nostalgia.
The good news is that we are seeing architectures increase again. You had to say ARM / x86 as ARM is a more successful architecture than anything that was challenging x86 when Itanium was introduced. And new RISC-V is coming on very strong.
If we can keep on ISA from completely dominating, it will be easier for smaller ones to exist. With 3 major ones, niche ones will feel less pressure to throw in the towel.
“x86 as ARM”?
Also, I personally don’t like the explosion of different architectures. I don’t want to give up my library of software for marginal performance-per-watt benefits. You see. people have libraries of commercial software they don’t want to lose or have to run in a crummy emulator. As a result, each OS has a “preferred” ISA and everything else is a distraction at best and a headache at worst. Even on Desktop Linux, whatever commercial software exists on it is on x86. Android also failed to become multi-architecture (despite originally being designed for it) because some Native Development Kit apps are ARM-only. Windows of course prefers x86.
Only MacOS has done multiple switches ISA because users weren’t really given a choice. For example, Apple could have moved to AMD instead of forcing an ISA switch to ARM, but they want to own the stack, so ARM with custom cores it was. It doesn’t have anything to do with the benefits to the user.
This is also why I won’t hold my breath over that new Windows ARM SoC from Qualcomm. We are talking about an industry that has trouble switching from Intel to AMD (despite Intel’s fab issues), and it’s supposed to go to ARM? Which brings me to my other point: Most of the flack x86 gets is because of Intel making CPUs on ancient processes. Sure, x86 introduces some performance penalty, but it’s small. After all, x86 was able to beat all RISC workstation CPUs in the 2000s.
kurkosdr,
The software I use is already open source, which makes architectural transitions a breeze. The first time I booted up a linux desktop on an ARM SBC, I was thinking “wow, everything is exactly the same”. If it weren’t for the much lower performance of my SBC I wouldn’t even have noticed it was a new architecture.
I understand your predicament though. With windows software especially, you’re stuck with x86 executable and that’s that. Emulation can be ok for a temporary software transition period, but for long term, why even bother because the proposition value for ARM disappears.
However if ARM became more popular among windows users, it would actually result in more multi-platform windows software, which wouldn’t necessarily be a bad thing for you. It’s just a chicken and egg problem.
I agree completely.
Yea, microsoft has commitment issues and it shows. When they talk about new platforms, be it hardware or software, nobody believes them. Even departments within microsoft are like “eh, let’s wait and see where it goes”.
intel stumbled badly, though x86 itself still managed fairly well thanks to AMD. Now intel have mostly recovered. The area where ARM has a strong lead is energy efficiency.
Are Itanium chips even made anymore? The last iteration was “Kittson” from November 2017, aka 6 years ago. At this point, we could be talking about retro hardware.
Which version of DOS tho’?
Is there even a version of DOS that runs on the Itanium? I would not think so.