@KimmoK

Quote:

Quote: OlafS25 wrote: No surprise they have shifted to ARM and will produce existing PPC chips as long there is enough need for it but not invest in new chip generations. It was easy predictable when you look at the company they now belong to. KimmoK wrote: @OlafS25 Yep. In 2016 they had some "hints" of some new Power Architecture based layerscape chip. But now it would seem there won't be any. To me it seems Qoriq chips are selling ok, but it seems they consider ARM will deliver the same in the long run. Interesting to see if Open Power deliver any new low...mid range chip. Most likely not. PowerPC chips will continue to be available for 10 years or so, but that's it. (plenty to play with, then I hope there will be nice host for AOS4 SW on some other CPU architecture... is it AROS or MorphOS... we'll see.)

NOTHING (for us) interesting has come in the shape of PPC since Apple abandoned it in 2005. Immediately after Apples decision, Freescale and IBM shifted focus and shelved several planned PPC architectures meant for Apple computers that could have been interesting to us, like the e700, follow-ups on 970MP, etc.

About when Genesi left PPC for ARM it was pretty clear that a PPC suitable for genuine Desktops or Laptops would never happen again. This was confirmed by various roadmaps. And it has been reconfirmed time and time again, as new roadmaps appeared over the years, ARM easily replaced PPC in every possible segment.

CPU’s stocked in warehouses will be sold out. Perhaps even some occasional production run, if needed to honor the 10 year longevity programs. But that’s about it.

@KimmoK

Quote:

KimmoK wrote: My crystal ball is fuzzy vs what CPU architecture is the best option in the long run.

ARM has by far the best diversity, ranging from very low power/performance CPU to desktop/laptop class in one (or a few ) architectures. It’s cheap, lean, power efficient, and it comes in many shapes and forms from a big and growing number of manufacturers. Problem with ARM is that the platform is so darn device centered, developments are made for single and specific end-user devices and not the more “general” Desktop or Laptop platform like on x64.

Perhaps this is about to change now, though? Microsoft has committed heavily to ARM now with their full, regular Windows 10 including an x64->ARM JIT compiler. They are targeting an emerging ultra low-cost laptop segment, that actually grows unlike x64 laptop and desktop segments. Maybe Microsofts Windows will function as a common denominator, helping standardize a regular laptop market based on ARM that more and more manufacturers can join in on? This could be helpful! Maybe the same will happen for desktop kind of devices as well?


Quote:

Ideally OS should be fully and easily portable and application binary independent (etc). But I'm afraid it might result in bloated and slow OS, unlike AmigaOS.

I like the idea of a CPU agnostic bytecode. The compilers in our toolchain is made to output that bytecode instead of some other ISA. The entire OS and all applications compiled for this. Then a JIT like some “Trance NG x64” or “Trance NG ARMv8” (or “Trance NG PPC-64” ) would compile CPU native bytecode for the platform it runs on.

For MorphOS, the time to make a change like that is now, since MorphOS 4 won’t be entirely source compatible anyway and will need at least a recompile of all third party apps to work natively anyway. And then the term “morph” would earn yet another meaning!

It doesn’t by any means mean slowness or bloatness per concept. A JIT is really just a compiler, that takes an input stream (be it ASCII stream or binary stream) and outputs it to bytecode of a certain ISA. Just like any other compiler. I remember Itix answered to a question once about “emulated” 68k code in MorphOS, and he said something in the line of (not a quote): “a recompiled 68k binary stream can even become more efficient thus faster on PPC since Trance is aware of and can use more registers and other resources that the old 68k compiler simply had no access to when generating the original bytecode”. And a from-scratch developed intermediary ISA would obviously have flexibility in mind by design, and its bytestreams would be predictable and nice to the various Trance end-compilers. Stuff probably exist already, in various shapes and forms. One should google it...

@TRIPOS

Quote:

NOTHING (for us) interesting has come in the shape of PPC since Apple abandoned it in 2005.

True, but it was surprising to see even even a similar decline for embedded CPUs, I don't think we will ever see a RaspberryPI like system (price and performace wise) with a PPC.

@BigD

Quote:

Hmmm.. I guess everyone needs a hobby or is it just better than counting sheep?

Yes, it's always better than counting sheep.

@KimmoK
Quote:

if in future we have only one or two available CPU ISA

Nowadays some people are capable of making their own CPU's and ISA's on FPGA's. Of course, the FPGA is not as fast as a real CPU but that doesn't stop people making them. It's not the same as big companies competing with each other obviously.

@OneTimer1
Quote:

I don't think we will ever see a RaspberryPI like system (price and performace wise) with a PPC.

If you add addons to RPi to make it a proper computer then the price goes up to same with the cheapest low end sale priced laptops/netbooks which have x86 CPU's.

_________________
Rock lobster bit me. My Workbench has always preferences. X1000 + AmigaOS4.1 FE
"Anyone can build a fast CPU. The trick is to build a fast system." -Seymour Cray

@Tomppeli

Quote:

Nowadays some people are capable of making their own CPU's and ISA's on FPGA's.

Absolutely, since the Color Computer 3 fpga project started block copy, dma, and a higher res 256 color mode have been added and I'd like to see some cpu enhancements adopted (there are the instructions added by the Hitachi 6309, as well as other possible additions).

I don't think cpus or ISA will be limited at all, no more than OS' were limited to two choices (MacOS and Windows).

We have the option of creating not just our own operating systems, but now we have the option of creating the hardware they run on.

Personally, right now I'm curious what could be done with Minix 3 as a base.
It needs a better user interface, but its micro kernel nature is attractive to me (OS9, which I've used on the 6809 and the 68K, and MorphOS are both micro kernel operating systems).

Maybe I'll join that community and try to contribute to the port to the 'dead' PPC ISA.

Long live it all, be it dead or alive.

Last edited by Beans on 19-Oct-2017 at 03:34 PM.
Last edited by Beans on 19-Oct-2017 at 02:27 PM.
Last edited by Beans on 19-Oct-2017 at 02:27 PM.

_________________
Amiga! "Our appeal has become more selective"

@Beans

Funnily enough, MINIX 3 is one of the most used OS' in use today.
http://blog.ptsecurity.com/2017/04/intel-me-way-of-static-analysis.html

@WolfpackN64

Quote:

Funnily enough, MINIX 3 is one of the most used OS' in use today.

As well it ought to be.
Its even embedded in Intel hardware in their latest chipsets.

I was a big advocate of micro kernel OS' in the '80s and '90s when I was selling 68K based systems running Microware's OS9 operating system (which in itself is somewhat UNIX-like).

A BSD based OS with a micro kernel?
Not exactly a new idea, or didn't anyone pay attention when Jobs created the NextStep then when he moved back to Apples he tanked OS9 and in favor of the Mach kernel based OSX (which is really just a retooled NextStep)?

With all regards to Torvalds, he was wrong, micro kernels rule!

_________________
Amiga! "Our appeal has become more selective"

@Beans

It's just to bad Intel uses MINIX 3 as a giant backdoor scheme on a separate processor that's embedded in their CPU's.

OSX isn't really that great of an example. It's most of the FreeBSD kernel on top of a Mach Microkernel. Not too dissimilar of the the old Mac OS userspace on top of the Mac OS 9 Nanokernel.

Microkernels are awesome. But Linux still works great despite or thanks to it's kernel.

@WolfpackN64

Quote:

But Linux still works great despite or thanks to it's kernel.

Actually, Linux IS the kernel.
Nothing prevents BSD or Minix3 from being the base for a Linux compatible OS, and in fact there are BSD variants that will run Linux binaries.

I'd like to see an AROS distro set up to run on top of Minix3, and just to increase the overall perversity of the idea, quite possibly a PPC variant.

Last edited by Beans on 19-Oct-2017 at 03:59 PM.

_________________
Amiga! "Our appeal has become more selective"

@Beans

I know Linux is the kernel and GNU is most of the userspace. I used Linux as a term for the whole since it's easier. Running Linux API's on FreeBSD and derivatives is mostly possible due to them sharing a lot of similarities (partly thanks to them both being UNIX systems). But It's not "easy" to plop Linux API's onto another OS and frankly, MINIX 3 doesn't have enough development or support to realistically directly support Linux syscalls.

AROS on top of MINIX would be kind of a freak, you'd basically run a microkernel on a microkernel.

@WolfpackN64

Quote:

AROS on top of MINIX would be kind of a freak, you'd basically run a microkernel on a microkernel.

I rather LIKE that idea.
I think you may have helped talk me into it.
Thanks.

_________________
Amiga! "Our appeal has become more selective"

@olegil

NXP, since it absorbed Freescale has:

Coldfire
PPC
And ARM.

All are still listed on the website.

Coldfire is not actively developed, but the cores are licensable.
PPCs are still produced, and while the Qorlq line is transitioning to ARM, many of the existing PPC SOCs will be produced for years to come.
And, as has been pointed out, ARM cores aren't even designed by the vendors, just the elements added to the Socs.

Frankly I find them a tad boring. At least X64 designers create their own designs.

_________________
Amiga! "Our appeal has become more selective"

@KimmoK

Just how good are you at setting up Linux?
Because the Linux installations on my AMD hardware work flawlessly.
And AMD video drivers have performed better for me than Nvidia's.

_________________
Amiga! "Our appeal has become more selective"

@Beans

With all due respect, there are two levels of ARM licenses. ISA and Architecture. There’s two dozens of architecture licensees and these produce custom implementations. famously qualcomm snapdragon/kryo core and Apple A, with 20 smaller players that simply don’t get the deserved visibility (Google being one of these)

As for AMD Radeons, while the situation is getting somewhat better, they notoriously underperform Nvidias under Linux. Checking on phoronix you may easily notice how even the Radeon Vega (including the mighty frontier edition) tend to perform slightly better than a 1 generation old Nvidia. The guy running phoronix sits on the x.org board, so I tend to believe he knows a thing or two about x.org performance testing. On the other hand they are fully documented to the ISA Level, thus i’d prefer them too

@Tellurium

I don't know which benchmarks you saw, but the Vega is hitting between the GTX 1070 and in some cases even beating the GTX 1080Ti. AMD's Open Source driver have achieved parity with their Windows driver.

@TRIPOS

Quote:

I like the idea of a CPU agnostic bytecode. The compilers in our toolchain is made to output that bytecode instead of some other ISA. The entire OS and all applications compiled for this.

This is pretty much the AmigaDE idea. In fact this is Java in a nutshell. Except Java has endian problems, which is strange because IIRC, the Java bytecode is big endian encoded.

If Java was integrated into AmigaOS4 it might make a transition to a new ISA a whole lot easier. But I suppose being CPU agnostic wasn't a specified feature. And AmigaDE never made it anywhere into anything Amiga.

@Hypex

Quote:

Hypex wrote: @TRIPOS This is pretty much the AmigaDE idea. In fact this is Java in a nutshell.

Not at all. I wasn’t talking about new languages, new toolchains, virtual machines or whatever. I was talking about what we have today, like a Trance JIT compiler, but instead of compiling native binaries out of 68k binaries it would compile it out of ”Morph” binaries. That’s all. So applications would run completely down to the metal native, using the MorphOS API, just like they do today. No layers in between, no boxes, no virtual machines or whatever.

@TRIPOS

So you can run MorphOS ELF on AmigaOS4.x, that should like bad idea,
not optimized, and so on.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@NutsAboutAmiga

Eh? What?!

@TRIPOS

Anyway its lot of work, way not make Hollywood programs, they run on anything.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS