The last nail in OS/2's coffin

[Lars]

Sorry, it's in German (unfortunately it's also not free):
https://www.heise.de/select/ct/2023/26/2317912025061981023

I suppose that this has also been.picked up by other IT press articles about the new X86-S architecture.
What it basically states is that Intel will now finally remove all 16-bit support (true for real mode, protected mode and VM86 mode), all segmentation, Ring 1 and 2 of the protection rings.
That means: NO OS/2 device driver will work, NO  OS/2 1.x application will work, no application with an IOPL segment will work, no VDM will work, some parts of the kernel will cease to work.

OS/2 is going to be put to it's well deserved, permanent rest.
The current x86 will likely live on for a couple of years but there clearly is no future for OS/2, not even in a virtualization environment like Virtualbox.

[Digi]

Time to pay attention to the project http://osfree.org/ 

[Martin Iturbide]

Hello

This is part of our 64bit dilemma.

I guess, reading some similar news articles, that Intel is trying to save costs and complexity on their processors by cutting the "legacy" of them. Intel calls it the "64-Bit Mode-Only Architecture". Yes, that will be the final nail in the coffin for running OS/2 on real hardware. I guess some emulators like 86Box can evolve to run ArcaOS at that time, but real hardware will be gone for us.

It's good to have the heads up about this, but the question is what we should do, as a community and as the little market that Arca Noae has.  Is the answer to this question binary?

Option A: When the moment comes, let OS/2 and ArcaOS rest, switch to a different OS, and talk about the good old OS/2 times with our friends.
or
Option B: Try to do something and work in a different strategy for our community and software.

Option B will be the "Long Term Strategy" that I insisted on the past that we should have. The short term necessities are wifi drivers, more applications, translations, etc. But the long term strategy should be to finally "free / release" OS/2 from the IBM owned binaries.

My "Dreaming In Technicolor" idea will be to first (first step of several) having an approach to run OS/2 binaries over a 64bits kernel. Create a layer that can interpret OS/2 to keep running all the OS/2 experience like native applications, the PM and WPS (GUI), etc.

Ryan C Gordon, tried it with a different approach, try to interpret an OS/2 application over Linux. But it was too much work to clone all the OS/2 APIs in one shot.

The crazy/drunk idea will be to grab a 64bits open source multi-kernel and make OS/2 run interpreted (not emulated) from it. Like grabbing the Zircon kernel (from Fuchsia OS) and developing some layers of OS/2 interpretation over it for binaries, Kernel instructions and basic $Drivers. CPI, PM, WPS and native OS/2 apps think they are running over OS/2. (of course all OS/2 native drivers will be useless at that moment).

This will require a lot of knowledge on the OS/2 kernel and the Zircon kernel and a lot of effort too.


Reference:
- May 21, 2023. https://www.pcgamer.com/intel-proposes-x86s-a-64-bit-cpu-microarchitecture-that-does-away-with-legacy-16-bit-and-32-bit-support/
- 25 May 2023. https://www.theregister.com/2023/05/25/intel_proposes_dropping_16_bit_mode/
- Specification Proposal x86S - https://www.intel.com/content/www/us/en/developer/articles/technical/envisioning-future-simplified-architecture.html

[Mentore]

Sorry, it's in German (unfortunately it's also not free):
https://www.heise.de/select/ct/2023/26/2317912025061981023

I suppose that this has also been.picked up by other IT press articles about the new X86-S architecture.
What it basically states is that Intel will now finally remove all 16-bit support (true for real mode, protected mode and VM86 mode), all segmentation, Ring 1 and 2 of the protection rings.
That means: NO OS/2 device driver will work, NO  OS/2 1.x application will work, no application with an IOPL segment will work, no VDM will work, some parts of the kernel will cease to work.

OS/2 is going to be put to it's well deserved, permanent rest.
The current x86 will likely live on for a couple of years but there clearly is no future for OS/2, not even in a virtualization environment like Virtualbox.

Dunno.
Basically this means we will build new machines on AMD CPUs.

(Jus' kidding, but not so much...)

Mentore

[Lars]

Taken the fact that AMD was the driving force to intoduce 64-bit mode and Intel just following, I very much doubt that AMD would put the burden onto itself to support legacy HW concepts that no current OS needs. That would throw them behind Intel with no benefit at all. You need the die for modern stuff (AI in HW or some such, larger on-chip cache, larger SIMD registers etc.) and not for stuff that is not needed at all any more.

Intel taking that step is effectively Intel's confession that 16-bit and segmentation have no future at all.
As to the ring protection, it turned out that differentiating between "user" (Ring 3) and "system" (Ring 0) is already sufficient. For paging, these are already the only 2 "Ring Protection" levels (and they are consequently called "user" and "system").

Unfortunately, OS/2 offers use of Ring 2 (in conjunction with the IOPL flag in the EFLAGS register) to simplify I/O port access and therefore there are some applications that make use of that and would need to be rewritten to use a device driver instead. But see my statement about OS/2 device drivers ...

[ivan]

This sounds very much like desperation on the part of Intel, they are seeing AMD taking over the desktop processor market with far better processors so they need something, anything, to get in front again.

[Lars]

If you want to delve into the details:

https://www.intel.com/content/www/us/en/developer/articles/technical/envisioning-future-simplified-architecture.html

Currently, this is only a proposal. But it gives a very good idea of what Intel thinks is no longer needed.

[Neil Waldhauer]

Intel proposed Itanium 64 bit in 1989. It takes some time for these plans to come about. Even articles about the new 64 bit architecture from Intel mention cores that still provide X86 mixed with 64-bit only cores.

https://en.wikipedia.org/wiki/Itanium

Even if nothing happens for a while, it's worth thinking about how OS/2 runs on CPU features that matter very little to the CPU market.

The other thing that threatens OS/2 is the aging of its users. I predict that will be the actual cause of death.

[Eugene Tucker]

Good point Neil couple that  with a limited kernal and vanishing developers.

[Lars]

Intel proposed Itanium 64 bit in 1989. It takes some time for these plans to come about. Even articles about the new 64 bit architecture from Intel mention cores that still provide X86 mixed with 64-bit only cores.

https://en.wikipedia.org/wiki/Itanium

Even if nothing happens for a while, it's worth thinking about how OS/2 runs on CPU features that matter very little to the CPU market.

The other thing that threatens OS/2 is the aging of its users. I predict that will be the actual cause of death.

With Itanium 64 bit, while moving to 64-bit, Intel wanted to push through a completely new CPU architecture with a completely different instruction set. That failed because nobody wanted to immediately rebuy all their software.

But with x86-S, Intel wants to make room on their die, kicking out all the stuff nobody needs any more (freeing space for valuable things like for example an increased cache size) and also, because the twisted x86 architecture seems to introduce more security issues then would be necessary because the x86 architecture has become too complex.
That is a completely different motivation. It is in Intel's best interest to push this through as fast as possible. And it has become fairly easy because all major OSes and their applications are 64-bit already and also already fullfil all prerequisites to make them run on x86-S.

[Dave Yeo]

In theory, you could have a 286 emulator in ring -1 along with a program to take control when there was an illegal instruction trap and emulate the 16 bit instructions. Ring -1 is for hypervisors and such. The missing ring 2 and IOPL flag would be harder if even possible.
Not going to happen with the shortage of developers. Besides, in this 64bit world, hardware manufacturers no longer care about the lower 4GB's so what is happening, fractured memory leaving too little visible to 32 bit applications and the killer, the framebuffer being mapped above 4GB.
One thing is hardware lasts now, I just retired a 13 year old machine that was still running fine, just slow IO and a crappy EUFI. So there's another decade where OS/2 will run on older hardware. I doubt that I'll be around that long.

[JTA]

There are two ways to run OS/2, better than it runs on hardware (which does have its limitations) ... the first is virtualization (Virtualbox), and the second is emulation (QEMU).

WRT virtualization (and OS/2), it allows things to be used long after most folks can't figure out how to keep using them on existing hardware ... their hardware dies, and they just stop using the old stuff. By virtualizing, the old stuff pretty much runs forever ... implement virtualization, perhaps by the AToF methods in the virtualization sub-forum, and OS/2 will run many, many years past the point where folks have stopped using OS/2 on "old" hardware.

WRT QEMU, it emulates the processor(s) that OS/2 ran on, and given a reasonable platform, you probably couldn't tell that OS/2 was being emulated.

I'd believe that both of these methods, and others that I haven't explored yet, will keep OS/2 running long past the point of current "old" hardware methods.

Given the popularity of all things "vintage", I'd also believe that OS/2 will be running such that the baton will be handed off to many more "generations" of OS/2 users.

I'm not sure we've seen all that virtualization & emulation has to give us ...

[Martin Iturbide]

The other thing that threatens OS/2 is the aging of its users. I predict that will be the actual cause of death.

I'm scared to interpret this like my age group (40's) is going to be last one using OS/2   
Let's try to get people on his 30's to the community !!!   

Regards

[Digi]

The crazy/drunk idea will be to grab a 64bits open source multi-kernel and make OS/2 run interpreted (not emulated) from it.

http://osfree.org/ - The beginning has already been made. As I understand it, most of the Dos* functions have already been implemented.

[Roderick Klein]

If you want to delve into the details:

https://www.intel.com/content/www/us/en/developer/articles/technical/envisioning-future-simplified-architecture.html

Currently, this is only a proposal. But it gives a very good idea of what Intel thinks is no longer needed.

Its expected this will indeed be long term thing to happen. I understand the legacy bit takes up 12% of the space in CPU.
I read an article (I do not the have link handy) of one of the main AMD tech engineers and he said the proposal was every interesting.
However the engineer indicated the changes proposed are "very, very, very, very complex to make".

The other thing that this would imply is (the way I understand it) is that NO current VM would work anymore
to run legacy OS. Its the CPU that provides the virtualization support to hypervisor.

Roderick