Windows NT PPC on Macs

[Rairii]

For NT on Alpha there was a dynamic recompilation thing from x86 - no idea how well it worked though or whether it made it to the PPC build.

In the NT4 times there was wx86 for Alpha, MIPS and PowerPC.

For PowerPC specifically, there's a version of wx86 that has the emulator replaced with a port of Insignia SoftWindows: https://archive.org/details/softwindows-32-powerpc-v1.01

I've had that running fine under my dingus tree but haven't checked it on real hardware yet.

 

[Rairii]

I’d love to see this evolve further with custom drivers, stability improvements, more supported models, etc. Most of these projects don’t go beyond proof of concept though, just gotta hope this goes further.

I plan to. I have an iBook G3 sitting next to me that runs NT in the wrong endianness (it boots, plenty of weird issues that would be HELL to debug), but thanks to the port to Gossamer-derivatives I learned what MSR_LE actually does, so eventually I'll do a bare metal ARC firmware little endian port to them (yes, I poked around uni-north registers hoping there was a PCI endianness switch register bit like on Bandit, but no luck).

The main things that need to be dealt with for a decent set of drivers are HAL related. There's no DMA support at all yet, mainly because even though the thing's documented due to its usage in Hydra, I have no idea how Apple's overblown DMA works. Also I'm not even sure if PCI interrupts work properly yet, most of the existing implementations just assume firmware sets up the PCI config registers appropriately, and the interrupts are completely different here (for example, I'm not sure how exactly the PCI interrupts would work for devices behind a bridge).

But given that I started porting NT to Wii to learn more about NT internals that I didn't know much about, and then later to Uni-North because I ran into hardware issues and an iBook G3 was cheaper than a USB Gecko, and then later to Gossamer-derivatives because the MPC106 actually has a little endian mode and there were systems that officially ran PowerPC NT that used it, I guess "learning things" is the entire point.

Although bring-up has been getting quicker I guess. I only bought my Lombard two months ago!

 

[obsolete]

Welcome to the forum, @Rairii! You'll fit right in here

 

[MOS8_030]

There is an old Vogons thread talking about PPC NT stuff but sadly the archive.org link doesn’t seem to work anymore:

Windows NT on Motorola Powerstack II \ VOGONS

www.vogons.org

Way back around 1997-8 in the the heady days of the Apple Clone Wars when I was at Motorola we had one of those PowerStack systems running NT in our office. Our admin had it set up as a server since everyone else in the office was still using Macs.
I don't know what became of that PowerStack system. There weren't many of them made to begin with and Microsoft dropped all their work on PPC NT immediately.
I got one of only a couple StarMax systems that made it in to our group before Apple killed licensing.
I ran NT for a while on my 200mhz Starmax via Virtual PC. It was slow but it worked.

 

[Nixontheknight]

Screen!

This just seems... wrong:
View attachment 75689
View attachment 75690
View attachment 75691
Here are the system specific strings from the System Control Panel:

View attachment 75692

View attachment 75693
Plays Pinball:
View attachment 75688

Its surprisingly stable for a hack. It only crashed twice for me.

if my DV iMac wasn't in pieces and in need of a recap, I'd do this

 

[croissantking]

if my DV iMac wasn't in pieces and in need of a recap, I'd do this

But the GitHub readme says it’s only for tray loaders.

 

[Daniël]

But the GitHub readme says it’s only for tray loaders.

On top of that, there's no possibility of adding ADB to a Slotloader. The Trayloaders have a bunch of left over functional legacy connectors on the logic board, including ADB, making it possible to actually use NT4 at this point, given the current lack of (working) USB HID drivers.

 

[croissantking]

On top of that, there's no possibility of adding ADB to a Slotloader. The Trayloaders have a bunch of left over functional legacy connectors on the logic board, including ADB, making it possible to actually use NT4 at this point, given the current lack of (working) USB HID drivers

Ummm. No USB? They say the G4 Yikes is compatible.

 

[Daniël]

Ummm. No USB? They say the G4 Yikes is compatible.

It's a Gossamer Mac, so yes, it theoretically is compatible, but there's no functional input on that Mac either for now, unless you solder the ADB connector and supporting components back, like on the Trayloader.

 

[croissantking]

It's a Gossamer Mac, so yes, it theoretically is compatible, but there's no functional input on that Mac either for now, unless you solder the ADB connector and supporting components back, like on the Trayloader.

I mean, that’s a great idea, assuming I can work out what supporting components it needs.

It’s kind of a catch 22 - machine must be new world but USB is broken. The Lombard is the only machine that works without any hardware modification, since its keyboard and trackpad are connected by ADB.

Edit: Oh, and the G3 Yosemite, of course!

 

[Daniël]

I mean, that’s a great idea, assuming I can work out what supporting components it needs.

There's an article that describes it here: https://ameblo.jp/keroxiee1016/entry-12663017372.html

For the Trayloader, it's just the inductors (or resistors as used for the larger inductors in this article), and the fuse, those capacitors aren't used on the board. That said, the location of the MiniDIN-4 connector on the Trayloader board is right behind a screw post for the side panel, making it unusable in that position.

It very much seems like it was there for testing purposes during the prototype phase, that wasn't removed before production. But that just adds to my theory that the Trayloader was rushed, and the Slotloader being the iMac G3 Apple would have shipped right out of the gate if they had the means and time to do so.

 

[croissantking]

the Trayloader was rushed

Less rushed than the TAM though, right?

 

[croissantking]

There's an article that describes it here:

Whoa, awesome. This is great!

 

[Nixontheknight]

But the GitHub readme says it’s only for tray loaders.

that's what I get for reading too fast

 

[joevt]

I learned what MSR_LE actually does

Does the NT on PowerPC stay in little endian mode all the time, or does it switch to big endian sometimes? In either case, I am wondering if the LE check in dingusppc could be done outside the CPU instruction execution loop so that when it happens, the emulator just goes into a different CPU instruction execution loop.

 

[wottle]

This makes me want to run it on my Black G3 All-In-One that I upgraded with a Yikes CPU. I'm assuming that won't work until they get it working with the Beige G3 (same logic board as the AIO). Would be fun to see NT running on a G4 AIO...

 

[Rairii]

Does the NT on PowerPC stay in little endian mode all the time, or does it switch to big endian sometimes? In either case, I am wondering if the LE check in dingusppc could be done outside the CPU instruction execution loop so that when it happens, the emulator just goes into a different CPU instruction execution loop.

It stays in little endian mode all the time.

 

[Snial]

OK, so based on the current conversation, NT for PPC could run on a 603e-based Mac as it too supports Little-endian mode, if it also supported built-in USB?

 

[finkmac]

to my knowledge, all powerpc cpus do little endian. the usb support for this isn't work yet either. neither of those are requirements for ntppc.

Spoiler: citation needed

virtualpc even uses undocumented nanokernel calls to switch the cpu into little endian

 

[croissantking]

OK, so based on the current conversation, NT for PPC could run on a 603e-based Mac as it too supports Little-endian mode, if it also supported built-in USB?

View attachment 75838

It only supports Gosammer boards currently, not sure what it would take to get it to work on anything older.