Quote:

[Edit] Come to think of it, even fat binaries wouldn't solve that. Imagine that you own an x86 AROS box, and write a shoot-em-up. You test it thoroughly and release it. Immediately someone complains that it crashes on his SAM, and a few others chime in saying it happens to them too. You try to reproduce the problem, but you can't. That's when you realize that there's some subtle incompatibility between x86 AROS and PPC AROS that you'll need to put in a workaround for. Unless you buy yourself a SAM440 or find a patient volunteer to help you, how are you going to test your fix?

AROS PPC on PearPC-JIT perhaps.

_________________
Amiga 1200 (rev 1D1, KS 3.2, PiStorm32/RPi CM4/Emu68)
Amiga 500 (rev 6A, ECS, KS 3.2, PiStorm/RPi 4B/Emu68)
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB

@CodeSmith

Quote:

Come to think of it, even fat binaries wouldn't solve that. Imagine that you own an x86 AROS box, and write a shoot-em-up. You test it thoroughly and release it. Immediately someone complains that it crashes on his SAM, and a few others chime in saying it happens to them too. You try to reproduce the problem, but you can't. That's when you realize that there's some subtle incompatibility between x86 AROS and PPC AROS that you'll need to put in a workaround for. Unless you buy yourself a SAM440 or find a patient volunteer to help you, how are you going to test your fix?

Come to think of it, even fat binaries wouldn't solve that. Imagine that you own an x86 Linux box, and write a shoot-em-up. You test it thoroughly and release it. Immediately someone complains that it crashes on his Alpha, and a few others chime in saying it happens to them too. You try to reproduce the problem, but you can't. That's when you realize that there's some subtle incompatibility between x86 Linux and Alpha Linux that you'll need to put in a workaround for. Unless you buy yourself an Alpha or find a patient volunteer to help you, how are you going to test your fix?


Write portable code and use emulators if you don't have the real machine or want to test it yourself. Or even better... release the sourcecode and let others fix the bugs and submit you the changes.

Last edited by Crumb on 23-May-2008 at 01:22 PM.
Last edited by Crumb on 23-May-2008 at 01:21 PM.

_________________
The only spanish amiga news web page/club: CUAZ

Quote:

Some users dont like PPC, some users dont like 68k, some users dont like x86. And some users dont like OS4, some users dont like MorphOS, some users dont like AROS. Some users dont like CGX, some users dont like P96, some users dont like MUI, some users dont like Reaction.

That's the biggest problem of this community. Nobody knows what they want. Maybe it's only good we have many different solutions. But that way there will be always too few developers.

_________________
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

@crumb:

You're not serious - are you suggesting that we regress to the way of distributing software that Linux is moving away from?

First of all, source distribution is just not usable for someone who isn't a programmer. It's one of the biggest reasons why people say that Linux isn't "ready for the desktop", why do you think Linux distros keep on improving package managers if all they had to do was distribute tarballs?
Secondly, a lot of people aren't going to want to distribute software in source form, especially if they charge for the program and have some sort of a niche. Imagine you've worked out how to reliably burn blu-ray on the amiga, no-one else has yet, and you want to make a few bucks. Until someone else figures it out too, you've got a nice bit of business going, and if you make the program nice enough to use and keep your customers happy, no-one else is going to try too hard. If you release the source code, aminet will end up with a bunch of clones that have a cut'n'pasted version of your burning routines and some half-assed GUI to make it work. They may not be as nice to use as yours, but they will be free and so people will put up with the crappiness and you won't make a cent. Why do you think OSS businesses give away the software and charge for support?

co-exist or die, is the moto.
I prefer PPC, and X86 is black box, that till this day does not interest me. Maybe in the future.

_________________
Quote:

Easy Pocket Money, Freelancers & Experts Online

MiniMig FPGA, Sam440 Flex 733Mhz PPC, Amiga OS 4.1 Update 2, MorphOS 2.4, Other - AmiKit + Cloanto Amiga Forever 2008 + E-UAE, AmigaSYS

Metalheart : A program that was compiled for x86 must be compiled again for PPC. This is the same behaviour on Linux.

Of course you can find x86 board cheaper than Sam440 but this board seems to be really good and it comes from Amiga companies, so I would be pleased to support it. PowerPC is a really good technology.

AROS is another OS for the Sam440, that's a good point.

Michal, you did a great job ! And quickly !

Quote:

That's the biggest problem of this community. Nobody knows what they want. Maybe it's only good we have many different solutions. But that way there will be always too few developers.

The choices Amiga fans have don't always match what we would like. I'm sure if you gave Amiga fan a low cost yet reasonably powerful device that could run all classic software, AROS, MorphOS and OS4 and wasn't based on some bland everyday hardware (but instead showed respect to the classic Amiga architecture), you'd have a big winner on your hands.

So, best hope Natami gets its (optional) PPC accelerator and its AROS + MorphOS + OS4 ports then!

Incidentally, here's an interesting video I just watched about choice, definitely worth watching:
http://video.google.com/videoplay?docid=6127548813950043200&ei=&hl=en

Quote:

First of all, source distribution is just not usable for someone who isn't a programmer. It's one of the biggest reasons why people say that Linux isn't "ready for the desktop", why do you think Linux distros keep on improving package managers if all they had to do was distribute tarballs?

It's quite possible to make source distribution as easy as installing programs on Windows, Linux has just been doing it wrong. You need to see that it is possible to compile and install a program through a single click.

The way it should be done is that every program that is distributed as uncompiled source code comes with an Install script. You can call it 'Install.exe' if you want to make Windows users more comfortable clicking it. This script brings up the options linked to installing the program. Once you've selected the options, the Install button is pressed. The script then starts your compiler (lets say GCC for arguments sake) compiles the program just for your Linux distribution (so you could have flags in the source code for features only the newest Linux kernel supports for instance), and once the program is compiled puts the compiled files in a folder ready for the program to be run.

Its a no brainer. The package managers have certain advantages when it comes to dependencies but otherwise the way I described is much more user friendly and would be a more logical way of doing things.

Quote:

Secondly, a lot of people aren't going to want to distribute software in source form, especially if they charge for the program and have some sort of a niche. Imagine you've worked out how to reliably burn blu-ray on the amiga, no-one else has yet, and you want to make a few bucks.

Then offer it as a closed source pre-compiled binary. You do know closed source programs exist for Linux, right? Nobody is directly forcing developers to do the right thing (IMHO) and go with the open source distribution model.

Last edited by HenryCase on 23-May-2008 at 07:40 PM.

@HenryCase :

Please follow my sig then...

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

This screenshot sums up my feelings:

_________________
A1200, 68060/64MB/1.2GB/WiFi/AGAtoCRT/OS3.9 Pegasos I, G3 600Mhz/512/9200SE/80GB WinUAE, Ryzen 5 2400G/Vega11, 8GB DDR4, 256GB SSD,Win 10 Pro x64 Amiga Forever !

@pixie
Quote:

Please follow my sig then...

What sig?

@DBAlex
Nice screenshot!

sig doesn't appear but in forum posts, my bad, I take it as a given... check my profile then...

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

@HenryCase:

Quote:

It's quite possible to make source distribution as easy as installing programs on Windows, Linux has just been doing it wrong. You need to see that it is possible to compile and install a program through a single click.

That's not the problem, in Linux instead of clicking a button you run "make config & make & make install" - hardly rocket science. The problem is that, depending on what version of the compiler you have installed, you may get build warnings or even build errors that require you to make small tweaks to the source code, and some versions of gcc produce code that is incompatible with code build with earlier versions (this happens when they tweak the ABI - doesn't happen often, but it does happen). And then of course there's dependency hell...

Quote:

Then offer it as a closed source pre-compiled binary. You do know closed source programs exist for Linux, right?

Sure, will that be a precompiled binary for x86, amd64, PPC, M68K or ARM?

Last edited by CodeSmith on 24-May-2008 at 04:51 AM.

This is good news, but when the Efika will have available this OS, and the Holywood programmers we will do programs for Efika and Sam?

Quote:

But seriously, could you imagine if we all just focused on one option. I mean I think a lot of people were hoping on AROS. Still, AROS, but even that is starting to fragment.

Now please visit the www.aros.org site, especially the AROS introduction. Please show me which out of five mentioned AROS goals implies that we are bound to the x86 platform? AROS is an operating system which
Quote:

2. Can be ported to different kinds of hardware architectures and processors, such as x86, PowerPC, Alpha, Sparc, HPPA and other.

Sam440, m68k (in form of AfAOS) and pure 64-bit version on x86_64 prove that this goal is valid.

@michalsc

I still think that there's such a thing as too much choice, especially if it means that you need to have different binaries. Someone mentioned using the M68K as the "canonical" CPU for AmigaOS binaries, I think that's a great idea. It makes sense because of the Amiga's history, and also the M68K is the weakest of the possible CPUs so it makes more sense to run 68K code emulated in an x86 than vice versa. Also, 68K emulation is very mature because of UAE's long history, whereas eg PPC and x86 emulation are still relatively new.

I wonder how difficult it would be to use 68K code as a universal binary format, where one takes a 68K binary, runs it through another program, and produces an equivalent x86 (or PPC, etc) program. A bit like JIT, only all done ahead of time.

@CodeSmith
That's basically emulation. Emulation of an outdated CPU as the standard for the next-gen Amiga...great. OS4 and MorphOS developpers working to make a PPC-AmigaOS : useless. 68K is the way to go! WinUAE is t3h n3w Amiga!

Sorry for being sarcastic but I really don't get it.

Last edited by Lynx on 26-May-2008 at 04:08 PM.

@CodeSmith

What have you been smoking?

Emulation is slow, end of story!

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

Correct, you don't get it

It's not emulation at all. Say that there's a 68K AROS program on aminet (let's call it X), I run AROS on PPC, and you run it on x86. We both download program X. I run it through a program (let's call it PPCGen) that converts X into a PPC native program, and you run it through another (let's call it X86Gen) that converts X into an x86 native program. When we both run X, we are running program X "on the metal" as a native program, no emulation anywhere.

How do PPCGen and X86Gen work? they take the 68K binary, treat it as if it were source code and compile it into a PPC or x86 binary. You can do this because even though the targets are completely different, the API is the same.

There's a program that's part of the .net SDK called ngen that does exactly this, it converts .net bytecodes into either x86 or amd64 native binaries. I wouldn't be surprised if the same thing existed for java bytecodes.

If AROS is to exist for multiple platforms and not have the problem that program X exists only for PPC and not x86 or vice versa, something like this will have to be done. I don't believe that source distribution is the correct answer (see my previous posts), so unless someone comes up with something better, a portable binary format is probably the best compromise.

Last edited by CodeSmith on 27-May-2008 at 12:13 AM.
Last edited by CodeSmith on 27-May-2008 at 12:08 AM.

@CodeSmith

This is an old idea that was talked about extensively when we was still hoping for ColdFire. What you describe would be possible but not with current Amiga 68k executable files. The main problem being that there is no way to distinguish data from code.

That's what make it different when emulating (instead of translating before running as you suggest). When the code runs, JIT can translate only the parts that are to be executed leaving data alone.

Ah, you're talking about self-modifying code aren't you. Darn, I forgot about that old dinosaur. Still, if we're talking about new code going forward then SMC could be simply declared verboten - it's not like you really need to squeeze every last cycle out of a 2GHz Core 2...