@NutsAboutAmiga

Not just unix. Oh, and IPC happens not so much through pipes and TCP as through socket files and a variety of message buses (akin to arexx), like DCOM, D-BUS etc.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

However you want to spin it, one of exec's strengths is also a weakness. The ability to send zero-copy messages between processes was key in making a fast IPC mechanism. However, it's also part of the reason why we a stuck with a single, unprotected memory model.

_________________
Doing stupid things for fun...

@Hypex

Quote:

Hypex wrote: Not very well back then. It would have been like a sell out to spend resources designing a new Amiga only to run Windows on it! But what's interesting is that what became of the PA-RISC is the Intel Itanium. So had it survived the future Hombre Amiga may have been running on an Intel Itanium. Oh the irony. An Amiga with an Intel Inside!

Amiga, as almost any system, was Intel inside since day zero.

In fact, it was Intel that invented the DRAMs...
Quote:

Hypex wrote: @cdimauro Quote: It's the exact opposite: the Amiga o.s. has only threads (running on a single process). That sounds like old joke about AmigaOS being a process and all running tasks being its threads.

It's not a joke: it's exactly like this.
Quote:

Quote: So, what's missing is the process concept & implementation. Problem is even with the share thread like nature of AmigaOS tasks they aren't that close. A child task or child process is a misnomer. They don't share a process context and are implemented as separate processes.

With the term "process" and "thread" I was referring to the o.ses literature.

The Amiga / post-Amiga land can call them differently, but actually there's only one type of entity running on the Amiga o.s./-like, and this is a thread. Everything on the Amiga o.s./-like is just a thread, which shares ALL resources on the single address space which is available (AKA: the single process).
Quote:

A well known example is bsdsocket.library not giving shareable library bases so main process cannot share base with child processes.

I don't see the problem: the mathieee libraries give back a non-shareable library base to be used specifically on that task, AFAIR.
Quote:

And there are a few threads online about getting in this. It seems to me people must be writing advanced programs if they need child processes, as they complicate it with more to manage, and code that has more strict rules to be thread safe so to speak. All the stuff I write manages it all in one process with signal handling. Except things like libraries that use IPC with a master process. I get the impression that people are being too lazy or just don't understand the thread semantics on AmigaOS. It's as if they program it in a way to act like threads on other systems then wonder why their program breaks on something. It's easy to understand when you know AmigaOS does not offer real threading capability for processes.

Well, it's the exact opposite. As I've said, everything on the Amiga o.s./-like is a thread.

It's "only" processes which are missing.
Quote:

The following post about OS4 having real threading features doesn't help. No info is given about. And questions on specifics are not answered so I take fact as a grain of salt. https://forum.amiga.org/index.php?topic=1217.msg13351#msg13351

If "real threading" is what's is found on the literature, then I fail to see how this is possible with OS4...


@Karlos

Quote:

Karlos wrote: @Hypex Quote: That sounds like old joke about AmigaOS being a process and all running tasks being its threads. That's not really a joke, though, at from the commonly understood perspective. What AmigaDOS calls Processes are just Tasks augmented with IO support. Tasks are even more lightweight than threads in that context. Quote: The following post about OS4 having real threading features doesn't help. No info is given about. One of the more frustrating problems I had back in the day was that I was unable to port some C++ code to OS4 from OS3.x because it used exceptions in a multiple task/process context. For OS3.x, I'd been using StormGCC and it came with threadsafe.lib that, among other things, made sure that an exception handler only executes in the same task that the excpetion was raised in. This all goes to pot in OS4 as the GCC compiler model has no equivalent thread safety and as soon as a task throws an exception, it can get picked up by a different one and the wrong stack gets unwound. Hilarity and frustration ensue. I don't know if this has changed since, but at the time, unhelpful stuff like "just rewrite it in C" was the answer.

That looks like a horror story...


@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote: @Hypex One main differences between AmigaOS and UNIX systems, is that programs are allocated differently memory locations on AmigaOS, while UNIX style, memory locations is same for all exe files, its MMU map that changes, in UNIX world programs can most only talk to each other using pipes or TCP or something like that, of course you use share memory but its also attached to file system, it’s another clumsy thing to use. because address space is same for exe files, it manes also fork creates copy of parent task, and all address point to right places, even if variable content is different. Forks on Amiga will kind of almost be impossible, the only why I think it can be possible, is if we use exception handler, to trick the child task into writing and read into offset variable space. Of course, that be really slow.

Forking is one of the most stupid things which unfortunately Unix has given to the humankind.

I suggest you to look at Windows ports instead of Unix ports, since Windows has no fork concept, like the Amiga o.s.. So, it's much easier to ports stuff from there.
Quote:

And also, I think only way solve network problem, is if socket was its own process, then it be easy to do IPC communication between tasks, and connection, so it becomes isolated form everything else. So I do not think fork problem is 100% realistic to solve, but at least, spin up process that talk to intermediate, seems like a possible solution. Anyhow it be interesting to play with virtual networks and middle layer, perhaps it be way bypass the lack of IPv6, and other missing features in the network stack.

It's a non-sense on the Amiga o.s., since everything is shared. However sometimes resources requires to be duplicated for the single tasks.


@Karlos

Quote:

Karlos wrote: However you want to spin it, one of exec's strengths is also a weakness. The ability to send zero-copy messages between processes was key in making a fast IPC mechanism. However, it's also part of the reason why we a stuck with a single, unprotected memory model.

*

@Hypex

Quote:

Hypex wrote: @Karlos Quote: Get some new material for pity's sake, you're embarrassing yourself and all of trollkind. And yet, despite low post count and interactions, ppcamiga1 is the catalyst for this whole thread and reason for its existence. He gave you a gift horse. Do you bite the hand that feeds you?

Pigeons don't blame other pigeons for having the cat set amongst them.

Last edited by agami on 13-Dec-2023 at 11:40 PM.
Last edited by agami on 13-Dec-2023 at 11:40 PM.

_________________
All the way, with 68k

@agami

Wait, am I a cat or a pigeon? Do I need to cancel my Dutch citizenship application now?

_________________
Doing stupid things for fun...

@Karlos

Quote:

Karlos wrote: @agami Wait, am I a cat or a pigeon? Do I need to cancel my Dutch citizenship application now?

Only if NGtherlands doesn't allow Dutch-Pigeon dual citizenship.

_________________
All the way, with 68k

@agami

Dunno about pigeons, but cat is probably a protected gender identity, so that'll be ok...

_________________
Doing stupid things for fun...

@ppcamiga1

Quote:

ppcamiga1 wrote: one and only thing from orginal hardware that classic amiga still have to have is cpu working in 32 bit big endian mode. which means no x86, no arm, no risc-v. chipset is not important. In 256 or thousands colors on amiga from commodore graphics is made as on pc from that time everything by cpu. chipset has no support to 256 or thousands colors modes. It only display them and work as dumb framebuffer. 68k may be changed to anything. software for amiga is made mostly in c/c++

1. Emu68 operates big-endian mode.
2. AGA has 256 color display modes without any tricks.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Karlos

Quote:

Wait, did you just imply I'm a troll?

More like implying I'm a stirrer.

Quote:

That's not really a joke, though, at from the commonly understood perspective. What AmigaDOS calls Processes are just Tasks augmented with IO support. Tasks are even more lightweight than threads in that context.

As I would say, every Process is a Task, but not every Task is a Process.

DOS support in this case, or DOS process. The problem is it doesn't work as commonly understood so people can get stuck.

Quote:

One of the more frustrating problems I had back in the day was that I was unable to port some C++ code to OS4 from OS3.x because it used exceptions in a multiple task/process context. For OS3.x, I'd been using StormGCC and it came with threadsafe.lib that, among other things, made sure that an exception handler only executes in the same task that the excpetion was raised in.

It must have had good support as usually porting C++ code to OS3 wasn't easy. Even now it looks like GeekGadgets is still relevant. Where as OS4 compilers managed to include most of these functions built in without needing extra libs.

Quote:

This all goes to pot in OS4 as the GCC compiler model has no equivalent thread safety and as soon as a task throws an exception, it can get picked up by a different one and the wrong stack gets unwound. Hilarity and frustration ensue.

I can imagine. A few years back I was trying to fix up FreeCiv on OS4. I managed to get it compiled but the code didn't work correctly. It was designed to run as a client/server model and spawned both off. But when it loaded on OS4 it broke. It sat there with broken communication. I suspected it was a threading issue but never found out the issue.

Quote:

I don't know if this has changed since, but at the time, unhelpful stuff like "just rewrite it in C" was the answer.

It wouldn't help since C++ is supposed to have better support on OS4.

Last edited by Hypex on 30-Dec-2023 at 07:58 AM.

@Hypex

I appreciate I come across as a troll when I am venting over PPC, but this is the end result of watching a seemingly non-stop traincrash for 20+ years.

At first, PPC was reasonable. When it first landed on the platform, everyone knew the 68K was not going to be developed further, 060 was your lot. Nobody then imagined that we'd have FPGA implementations of the ISA that are faster. UAE was still no faster than a high end 68K: I remember having an AMD Thunderbird that could reach the same speed as my 040 for compute bound code, though somewhat faster for memory bandwidth limited code (of course), since it was all still running interpretively.

For a brief time, all the talk around phase5's ABox and proposed Caipirinha custom chips sounded great, though in hindsight it was obvious vapourware.

I was an early adopter and got my BlizzardPPC. Which ended up having to be sent back to get reworked due to having all sorts of early revision issues. Came back months later in a fit state and all was well for a time.

Then there was the whole Blizzard G4 thing, sounded great, never happened. Followed by AmiJoe G3. Sounded an OK commiseration prize, but never happened either.

Meanwhile there was the original AmigaOne, which also sounded great at first, expanding the A1200 directly but that was eventally reduced to just a dodgy PPC board full of bugs. Same with the Peg1.

Despite all the warning signs then of increasing cost, decreasing compatibility and a list of issues as long as your arm, I still kept up and remained enthusiastic.

Until Amithlon / AmigaOS XL. Suddenly there was a 68K solution that put a fat question mark over the PPC. So what if it didn't have native chipset support? Neither did the A1 or Peg.

So I worked away with both. It wasn't until the A1 physically died from the exhaustion of basically *just existing* that I stepped back. By then I had other responsibilities too. Since then all I have seen is more and more farce:

For OS4:

* Less frequent releases of ever more expensive vanity hardware
* Hardware that can't be used properly by the OS it is allegedly designed for
* Hardware decades in the pipeline that ends up an *incompatible* FPU.

Alternatively, for MOS:

* QBox all but forgotten, MorphOS *is* the ABox today.
* Climbed up a melting iceberg of old PPC abandoned HW.
* Only legitimate claim to fame is that it's better than OS4 on equivalent hardware and has a functional browser.

If you peel away the superficial improvements to UI and graphics, neither OS has actually improved meaningfully over 3.x on 68K for the hardware they run on:

* No support for memory sizes > 2GB
* No support for SMP
* No support for Memory Protection. There's an MVP notion of privately allocated memory in OS4.

I appreciate those are hard problems and likely won't be solved without breaking compatibility. And maybe it doesn't even matter, it just makes a bit of a mockery of the term "next generation". It's really just "the same generation" on slightly less ancient hardware. It's been decades now. Where's the progress?

That's an easy one. There aren't enough devs of the right calibre, right? Or is it that nobody is actually interested in working on the most beige cpu architecture that history will record as an "also ran".

The PPC: One of the architectures of all time.

Last edited by Karlos on 29-Dec-2023 at 05:03 PM.
Last edited by Karlos on 29-Dec-2023 at 05:00 PM.

_________________
Doing stupid things for fun...

@Karlos

I think our fellow NG die-hard Amigans arguing for PPC is a mute point by this stage. We have so many FPGA and PiStorm physical hardware solutions on 68k hardware/emulation options both Arm and x86-64. PPC is not really on my radar other than a nice A1222+ box that popped up on the forums/Discord this week! The AAA Early Adopter box looked nice too!

I guess the PiStorm, THEA500 Mini and all the SOFTWARE DEVELOMENT that has been rejuvenated due to the 68k hardware is exciting! In person Amiga meets/shows and Discord channels are great fun! Moaning about PPC isn't fun!

The PS3 Cell was interesting and I liked that it made The Last of Us possible on 2006 hardware! Those times are gone.

Last edited by BigD on 29-Dec-2023 at 06:42 PM.
Last edited by BigD on 29-Dec-2023 at 06:41 PM.
Last edited by BigD on 29-Dec-2023 at 06:40 PM.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@Karlos


All reasonable technical observations.

I guess I would suggest meaningful (to me) improvements as:
- Good dual monitor support.
- Good interaction with the OS including SMB.
- Good browser.
- Good email.
- Good crm.
- Good dtp.

All of the above have significantly changed over the last few years. I view meaningful as tangible things I can get to use. To me this is very significant over 4.1 at this point, as pointed out it is a dead OS. Some users still support it and develop for it, but it is a dead end because it was destroyed in court. Even if a miracle happened and Hyperion and AEON allow people to make money and build trust in them again, the fruit of the poisoned tree chased likely too many away to ever make a real difference.

I think the most realistic future is for a large bounty and some meeting of the minds with the developers that could actually port MorphOS somewhere else. They have the right idea of creating/improving native apps and continue moving the OS along the path.

Until then, I'm 100% content on my G5 as almost every other day it seems I have another update or enhancement to get stuff done...

G5's are very cheap to pick up and if you get a 2.5GHz PCIe or a 2.7GHz PCIX you will have very modernish speeds for what you use a computer for. Nothing else is really close in Amiga Land.

Last edited by Matt3k on 29-Dec-2023 at 06:43 PM.
Last edited by Matt3k on 29-Dec-2023 at 06:40 PM.

@Matt3k

Other than the dual monitor support and maybe the samba, depending on how integrated it is, which of those are OS improvements? The browser and other clients are applications.

Speaking of dual monitor, admittedly not quite the same, but I used to use RTG and AGA simultaneously back in the day. It worked surprisingly well. Well, at least until I figured 57kHz audio was more useful to me than that.

_________________
Doing stupid things for fun...

If we want to truly talk about "next generation", that addresses the core OS features regarding SMP, memory protection, address space, source but not necessarily binary compatible and ease of running software from "other" platforms, then ironically, Ax Runtime has already been there and done it - because it's Linux. Which is why it'll never satisfy NG crowd, despite doing what their OS of choice has not done, regardless of promises, for decades.

_________________
Doing stupid things for fun...

@Karlos

Well dual monitor worked with any Amiga that had either 2 GFX cards or 1 plus the chipset.

Dual monitor out of 1 card was the reason why had that G550 in my Amithlon system and not some GeForce2.

Was a pain in the #### to use till I did SpringMaus, which isn't needed on MorphOS anymore.

So........

Stuff that was and is complicated on 68k real or fake just works with MorphOS just like you can try to port the "applications" you mentioned over to any version of AOS and see how far you get....

68k should, could and should could.
MorphOS does.

Last edited by Kronos on 29-Dec-2023 at 09:17 PM.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Matt3k

Quote:

I guess I would suggest meaningful (to me) improvements as: - Good dual monitor support. - Good interaction with the OS including SMB. - Good browser. - Good email. - Good crm. - Good dtp.

I wonder if Pistorm and an hypervisor like solution wouldn't be enought to 'eat the cake and have it too'..
On x86 it seems the best solution for using Linux alongside Windows, without having to dual boot. I will always have those windows apps I use, but it would be far easier to try Linux apps also, and with Pistorm it might be the same, use both Amiga side and Linux as well.

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

@pixie

I'd say at that point just run UAE on the rPi and stop fooling yourself.....

Last edited by Kronos on 29-Dec-2023 at 09:08 PM.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada

@Kronos

You overestimate the achievement of NG. Sure, you have evolved APIs and a lot of peripheral functionality that yawningly standard on pretty much every mainstream (and many esoteric) operating systems, but there's not a single thing it does worth a damn that's actually dependent on it being a PPC exclusive. That's the issue here and the relevance to this thread.

You want an example killer app for an Uber 68K Amiga? How about the MorphOS ABox? Given how utterly coupled it is to the Amiga 68K legacy (31 bit addressing, single address space, big endian, no memory protection, etc) anyway, what single meaningful technical* obstacle is there to porting it?

So many people voiced the opinion that MorphOS was the true AmigaOS successor and here we are at a point in time that it could (in theory) run well on a bonafide Amiga for the first time since the early PowerUp version, but it won't happen, because people equate the 68K ISA with the 68K CPU and think of it as a retrograde step rather than the binary virtualisation opportunity it represents.

*As opposed to the core team having no interest or someone making it actually depedent on Altivec.

_________________
Doing stupid things for fun...

@Karlos

I think we fundamentally agree on most items, the semantics just get in the way :)

Well we can select any point in time to measure that. So if we say we are going to say MorphOS 3.13 will be the starting point. Then we can look over the 5 version updates to the specific improvements from that point to 3.18... Since the team tracks so much, it is relatively easy to look at each specific improvement and bug fix for any point of time even back to 2.x or maybe even earlier...

Sure, my point has been the apps on MorphOS are very well supported and updated. Between PolyOrga, Iris, Wayfarer, and other programs you can really see the progression and maturity.

Yes, I also run dual monitor on my 3000D with 2 rtg cards and your right it isn't that good in comparison. I did get both screen to use two different screen savers though!

MorphOS uses dual head or one video card for dual monitor, has a MUI button you can have in your window to just click and it automatically goes to the other display, and the screen savers already understand dual monitor and just work on both. The OS when booted already sets up the other display when it sees it. So it is much more advanced and simple to use than other Amiga like systems, even with subtle differences sometimes, they add up to a great experience.

Don't get me wrong, I would love for it to get ported someday to something else, I'm just content with the software progression that takes my focus off the hardware. 4.1 focuses on the hardware by releasing bespoke hardware and with the state of 4.1 you don't get much os/software. PowerMac stuff is cheap and performs better than any other PPC Amiga hardware by a decent margin, so I just don't care as much right now.

I don't have the time or understanding to spearhead such a major endeavor that may not even be of interest to the developers right now. If someone had time and understanding to get this serious, I would certainly donate to the bounty.

@Karlos

Quote:

Karlos wrote: (in theory)

I could be the next queen of England.



I don't care what could have been if we all followed the teaching of St Charles_LXVIII.


What is is what is and thats what it is.

_________________
- We don't need good ideas, we haven't run out on bad ones yet
- blame Canada