Since the "attracting outside developers" thread derailed with discussion about modernizing Amiga, I'm starting a new thread. Now I'm going to pose a few questions:

Should Amiga-like systems abandon the Exec.library kernel?

Should Amiga-like systems go bi-endian like AROS? Little endian?

Should we try to maintain cross-compatibility via some bytecode so we don't need to care about the OS at all?

Who will supply drivers?

@Samurai_Crow

From the previous thread:

Quote:

Hosted AROS on Linux is slower than native AROS.

On the same hardware?

Can you cite benchmark results that demonstrate the speed differences?

I can see the outcome going either way. Linux is much larger than AROS, but it also uses more sophisticated algorithms that may be faster in some circumstances. Linux hosted requires more memory, that much is obvious.

Last edited by bison on 27-Sep-2019 at 07:09 PM.

_________________
"Unix is supposed to fix that." -- Jay Miner

@Samurai_Crow

how does "bi-endian" work, have not seen nor noticed thing like that while working with any data files or memory.

PowerPC does support reverse endian instruction does that make PowerPC bi-endian? if so how does that help?

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

@bison

Well it depends I guess, optimized drivers are faster than none optimized drivers, drivers for specific graphics card is faster then the generic graphic driver.

So if you do not have the right driver I guess that a hosted OS can be faster than none hosted OS on hardware it barely support. If however you do have all drivers for all hardware and you do not need to use generic crap drivers, then it might be that a native OS is faster.

Maybe we just stick a Amiga sticker on IPhone and call it Amiga it makes no difference to me what people do.

Last edited by NutsAboutAmiga on 27-Sep-2019 at 07:36 PM.

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

@Samurai_Crow

Are we dreaming, or are we being practical?

Quote:

Should Amiga-like systems abandon the Exec.library kernel?

Dreaming: A modernized Amiga system should abandon any part of exec.library that holds it back from being modern. Practices like blatantly snooping into and even modifying another process's memory have to stop. The lack of resource tracking is likewise a non-starter for serious work. This doesn't mean you have to embrace the inefficiency of monolithic kernels; QNX and others have shown that you can have simplicity, memory safety, and efficiency (QNX can even swap its kernel while running, without going down).

Practical: Probably not, because who will rewrite what's there, and what software will run on it once things like the current message passing system no longer work?

Quote:

Should Amiga-like systems go bi-endian like AROS? Little endian?

Dreaming: Whatever the hardware requires. Apple showed that you can survive an endian switch.

Practical: Big-endian for compatibility.

Quote:

Should we try to maintain cross-compatibility via some bytecode so we don't need to care about the OS at all?

Dreaming: AmigaOS should have modernized and discarded backwards compatibility long ago, with classic software run in a JIT-translated sandbox, the way Apple was doing it 20 years ago (funny I keep coming back to them).

Practical: This would have to be done incrementally, with some though put into the design, and lots of coordination between several teams, and will probably cost money, so... no.

Quote:

Who will supply drivers?

Dreaming: Open-source the operating system and find a way to convert Linux drivers to a new driver subsystem. Even this is problematic, though, because Linux's developers have no interest in backward compatibility, and routinely change interfaces, then change them back (I've seen it happen on a USB wireless antenna driver). Maybe there's a way to get at the driver's core behavior without running this risk, but I don't know enough about that sort of thing.

Practical: No one without $$$ and/or time, so... no one.

Basically, AROS has the right idea, but a lot of developers who like AROS don't have the combination of time and energy necessary to work on it. Except, apparently, the Vampire developers, who are helping to get drivers for their machine.

_________________
I've decided to follow an awful lot of people I respect and leave AmigaWorld. If for some reason you want to talk to me, it shouldn't take much effort to find me.

@NutsAboutAmiga

AROS has support for both endiannesses. ARM is normally little-endian but older ARM CPUs support big-endian mode also. PPC was originally biendian as you said and could support a little-endian version of AROS as well.

@bison

Good question! I've always run hosted on my devices because of poor driver support. I've heard on AROS-Exec that native AROS is more responsive.

Of the factors that slow down Linux are hard drive swapping when memory runs out, memory paging in general costs a little due to the MMU accessing its page tables, and the fact that everything on Linux is memory protected adding to the second factor.

Factors that can speed up AROS hosted are HostGL accessing Linux graphics drivers and multicore support for drivers running in their own tasks.

@Samurai_Crow

Forgive me in advance, I have no intention to be disrespectful.

Watching other wishful thinking threads, I came to realize that before drafting feature lists it would be wise to think of who should do the job.
I mean, you and which (dev) army?

If you could gather a big enough team of people committed to share their free time for an organized development process, then discussing and/or opening feature polls would start to make sense.

@IridiumFX

That's the best point to make. While NextGen Amiga-like operating systems languish and are ready to die on the vine, the classics seem to catch new life.

If there is a third generation of Amiga, and I doubt there will be, it would have to support classic hardware because that's where all Amiga development comes from. That's why Vampire is selling.

Last edited by Samurai_Crow on 27-Sep-2019 at 08:58 PM.

@Samurai_Crow

> Should Amiga-like systems abandon the Exec.library kernel?

At this point, yes, it should be abandoned as the underlying kernel, but you can still provide the library for (partial) compatibility.

> Should Amiga-like systems go bi-endian like AROS? Little endian?

Little endian, absolutely. Let's not be 'religious' about it because little endian makes the most sense for a number of reasons.

> Should we try to maintain cross-compatibility via some bytecode so we don't need to care about the OS at all?

I think it remains important to have alternatives, and there is yet a lot to like about the Amiga OS - in short yes, there needs to be some compatibility, but it can (reasonably) be provided via UAE.

> Who will supply drivers?

If not derived from / integrated with open source, I can see no option.

Last edited by Snorg on 27-Sep-2019 at 09:24 PM.

@Samurai_Crow

It seems like this subject has been discussed before. But not for at least three months now, so here we go...

I think there are only two realistic scenarios for a third-generation Amiga:

1. Something that's an incremental improvement on 3.x and 4.x. This would not be a modern system, but would modestly extend the usefulness of what is currently available. Ideally, it should run on commodity hardware like amd64 and/or RPi. Failing that, a very minimal Linux system should be created to host the system. A system like this would keep Exec.library, remain big endian, etc. If hosted, it could use the drivers on the host system.

2. Create an Amiga-like UI *and* GUI toolkit for existing Linux and BSD systems so that an ecosystem of Amiga-like apps could be created for these systems. Such a system should be able to run the new apps along side existing native apps, and use native drivers. This is similar to what Apple did with OS X.

I'm in in favour of either approach, preferably both.

As far as a brand new from-scratch Amiga system, I think the train left the station on that a long time ago.

Last edited by bison on 27-Sep-2019 at 09:54 PM.

_________________
"Unix is supposed to fix that." -- Jay Miner

@bison

The Aeros distribution of hosted Aros on Linux comes bundled with a command line utility that launches Linux programs from IconX or the Aros shell. I don't know if it's open source though. It originally wasn't.

Also, the AMD64 and ARM EB versions of Aros implement the multicore support in Exec experimentally.

Finally, Michal Schulz wants to make an Aros/Linux hybrid called Arix. It was originally supposed to be a GPL AROS varient that used Linux drivers but all he has revealed recently is that plans have changed.

@bison

Quote:

bison wrote: @Samurai_Crow From the previous thread: Quote: Hosted AROS on Linux is slower than native AROS. On the same hardware?

And how much of that can be due to X11/Xorg?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

And how much of that can be due to X11/Xorg?

The official answer from Xorg: nobody knows.

Quote:

X performance is extremely difficult to quantify, and there is a large amount of work to be done in this space

https://www.x.org/wiki/Development/Documentation/Performance/

_________________
"Unix is supposed to fix that." -- Jay Miner

@Samurai_Crow

Hardware for PC and Linux is not all that good. I have 3 PCs which don't have drivers for one reason or another to work on AROS Linux hosted.

Having TRUSTED hardware matters, having trusted GNU Linux drivers matter, or we could be thinking its an AROS problem when really its a GNU Linux or Virtual machine error.

A-EON dedicated to Linux is good, for compatibility, efficiency, stability, security, for Amiga testing but also Linux community - sanity that the hardware will work with the software.

Its good business.

Using Open Source drivers that are the most data optimized to chip as possible.

This system should be also Hackintosh compatible, AROS (hardware) compatible, Virtual Machine (hardware) compatible with full hardware acceleration 2D, 3D, Networking, etc.

Amiga systems need to start using proper benchmark algorithms, not tweaked for the test !

For a start : Lets see some Amiga OS3, OS4, Linux whets numbers :
http://www.roylongbottom.org.uk/whets.c

_________________
Download 499.26 Mbps, 659.94 Mbps Upload :)

@asymetrix

There is room for improvement even above Amiga. AmigaDE was able to inject inline code from the driver into the executables and run an update when the driver changed. Maybe someday we can get that back.

@Samurai_Crow

If have rearranged your questions, because the answer to earlier questions are important for a better understanding of the next answer.

Quote:

Should Amiga-like systems go bi-endian like AROS? Little endian?

I see Amiga as an OS that should run on broadly available and affordable hardware.

Currently you will only find AMD64 or ARM compatible systems on the market.

Both systems are supporting little-endian format, outside this Amigaoid filter bubble people people declare the Motorola format to be dead:

https://www.reddit.com/r/linux/comments/3467gq/bigendian_is_effectively_dead/

rant: If I could get an Euro, for every time someone misunderstood data because of this stupid format, I would be rich.

Quote:

Should Amiga-like systems abandon the Exec.library kernel?

Exec is big-endian
Exec could not support multiprocessing
Exec could not support user right management
Exec could not support decent resource tracking
Exec could not support memory virtualisation

Maybe a modern AmigaOS should abandon more than just Exec.


Quote:

Should we try to maintain cross-compatibility via some bytecode so we don't need to care about the OS at all?

Impossible, at least to impossible when trying to have a compatibility to AOS3.x, because of the endianess problems.

Quote:

Who will supply drivers?

Depends on the system that would replace Exec, if you switch to Linux or BSD as kernal you will get your drivers from there.

Last edited by OneTimer1 on 28-Sep-2019 at 11:27 PM.
Last edited by OneTimer1 on 28-Sep-2019 at 11:27 PM.

@OneTimer1

Good logical answers, they make sense (hopefully to everyone here, but just wait for the crazy opposition, and fantasy stuff to show up now).

I wonder how badly the performance of AmigaOS4 would suffer, if it were run hosted on top of Linux, and the "Rabbit Hole" feature were implemented, to allow running Linux programs seamlessly from the AmigaOS4 desktop? Using standard, top of the line x64 hardware of course, to minimize the performance problems. Surely the top of the line x64 hardware won't cost more than an X5000 system, and the under lying Linux OS would run any Linux native software fast, through the "Rabbit Hole" feature. Of course we are not likely to ever see this happen, as it would kill A-Eon's hardware sales, and they are the only company who has the funds to make this kind of change to AmigaOS4 happen.

It seems more evident to me every day that running existing Amiga software should be done only through UAE on modern hardware, unless you are only "playing around" and are okay with your old original Amiga hardware, or the more powerful FPGA variants. The amount of software written for any PPC variant of the Amiga, is just too little, and too late, and not likely to change EVER!

That sentiment is not going to make many of my friends happy to read, but it is of course just my personal opinion. Vampire is a fun toy for running Amiga software at a faster speed, and it is an interesting "Retro" project, but will never turn into anything serious, or for daily computer use. AROS could turn into something better than what it is today, but it has never really caught on with most of the community, and probably never will. MorphOS was good for a while, but is starting to show it's age, and has probably run it's course, unless the MorphOS Dev. Team can create a miracle with their switch to the x64 hardware platform, and there is no telling how long that will take, what software it will be able to run when, or IF it is ever released, and how many users will adopt it from the existing ever shrinking user community.

Maybe a 3rd generation is possible, but where will it come from, the MorphOS Dev. Team, or the AROS group (or my preference, both groups combined)? I doubt that AmigaOS4 can afford, or survive a jump to anything like a 3rd generation, even with Trevor's seemingly bottomless pockets funding it.

Something that builds on top of the strengths of the Linux & BSD community seems to be the only way forward, if we want a bigger selection of drivers, and any chance of staying current with the constantly changing computer world. Using commodity x64 & ARM hardware is the only choice that makes any sense price wise. To have enough manpower to make any kind of change feasible, such a change would need to attract users and programmers from all corners of the Amiga/MorphOS/AROS communities. One thing is pretty clear, anything with higher hopes than pure Retro, needs to make a clean break from the limitations of the Amiga's past. We have several versions of UAE to run the old stuff, and it can do so very well, and can be disguised to run it transparently, so give up any attempt at backward compatibility, and move forward with a clean slate. The retro segment of our community will continue on as it always has, and won't be affected much by anything any NG efforts do, or don't do.

It would be nice to be able to accomplish 90% to 100% of all our computer tasks on a 3rd generation Amiga-Like system, instead of relying on Windows, MacOSX, or Linux for so much of what we need and want to do on a computer.

Edit: I haven't looked at QNX in a long time, but I was really disappointed when the plans to use it as the base kernel for an updated AmigaOS was scrapped. I loved what it could do when they had it all compressed onto a single floppy disk, that could auto-boot on most x86 hardware of it's day, and install and run a web browser, as well as a few other things. I'm not aware of what QNX is doing these days, but I sort of doubt that it could compete with the Linux & BSD communities for driver development and maintenance, unless our hosted 3rd gen. Amiga was limited to a much smaller sub-set of hardware choices, which most users would be okay with, if the choices had good price to performance ratios, and availability was long lived.

Last edited by amigadave on 29-Sep-2019 at 12:45 AM.

_________________
Amiga! The computer that inspired so many, to accomplish so much, but has ended up in the hands of . . . . . . . . . .

Re:POSIX OS

If the choice came between Linux or something else POSIX, I'd choose Haiku. It's as POSIX compliant as BeOS (the 32 bit version is backward compatible to BeOS x86). The 64-bit version is better for its lack of cruft.

It runs on a system with 1 gig of RAM comfortably and is multicore from the ground up. It's only current downfall is it's lack of graphics acceleration in its Mesa port. It is license-incompatible to the GPL but is compatible to BSD.

Last edited by Samurai_Crow on 29-Sep-2019 at 03:48 AM.

@amigadave

Quote:

I'm not aware of what QNX is doing these days...

There's a high probability it's running in your car, at least if their website is to be believed.

Some years back, QNX's founders decided to retire, and sold the company to Blackberry, who tried to release a tablet with an OS based on the QNX kernel, but just like everything else Blackberry has done for the last 10-15 years in the consumer market, they botched the release badly.

(At least they didn't botch things as badly as Palm, who shot themselves in the foot so many times that they no longer exist.)

Quote:

...but I sort of doubt that it could compete with the Linux & BSD communities for driver development and maintenance

Alas.

_________________
I've decided to follow an awful lot of people I respect and leave AmigaWorld. If for some reason you want to talk to me, it shouldn't take much effort to find me.

@amigadave

Quote:

I wonder how badly the performance of AmigaOS4 would suffer, if it were run hosted on top of Linux, and the "Rabbit Hole" feature were implemented, to allow running Linux programs seamlessly from the AmigaOS4 desktop?

AROS runs well hosted on Linux, so I expect AmigaOS would as well. In fact, AROS runs well on a nested X server running on Weston (the Wayland reference compositor). I'm running on a i5-7600, and none of this is taxing my CPU or using much RAM. Emulation, on the other hand, can use a lot of CPU, especially with FS-UAE.

Update:

Some statistics. AROS hosted on Linux (Linux Mint 19.1 Xfce) is using less than 2% CPU and maybe 1 MB RAM. I can't really tell on the RAM, since it falls within the size of a rounding error.

FS-UAE emulating an A4000 and booting into Workbench 3.1 on the same system is using 100% of a single core on my quad-core CPU, and about 130 MB RAM. The fan noise due to high CPU usage is a nuisance.

Last edited by bison on 29-Sep-2019 at 05:28 AM.
Last edited by bison on 29-Sep-2019 at 05:27 AM.

_________________
"Unix is supposed to fix that." -- Jay Miner