@Fl@sh

Quote:

I wonder if OS3 and OS4 will have an unified API, is this in the plans?

It has the same functions.

But the ABI is different, 680x0 ABI is register based, the AmigaOS4.1 PowerPC ELF ABI is stack based,
there is translation layer built into AmigaOS4.1 that takes care of register to stack. there is no counterpart for 680x0, that translate the other way. and there no PowerPC JIT to 680x0 for AmigaOS3.2, that can execute the code. (You need a amazing fast 680x0 pull that off.)

AmigaOS4.1 can run 680x0 and PowerPC binaries.

AmigaOS3.2 can only run 680x0 binaries,
and it can be somewhat source code compatible.

Just to explain the complexity of this:

AmigaOS4.1 Stubs can be generated to call AmigaOS3.1 libraries from PowerPC programs, this is own file called libname.librariy.1.main while PowerPC native library can include stubs pointed to by jmp table, the jmp table is optional. Native PowerPC libraries might not include backwards compatibility, if not needed. EmuTraps can be setup this will cause PowerPC native routines to called instead of the 680x0 routines as well, this is useful when you have a callback functions that is expected in the 680x0 format, for backwards compatibility. Like the sana2 standard.

Last edited by NutsAboutAmiga on 22-Dec-2021 at 07:59 PM.
Last edited by NutsAboutAmiga on 22-Dec-2021 at 07:42 PM.

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

Chris_Y Quote:

It would be nice if SetPatch would patch the critical ROM bugs, rather than requiring the entire module is replaced - this would stop the annoying double-boot. Obviously on major component changes it wouldn't be possible, but 3.2 ROM with 3.2.1 OS shouldn't have any major changes that require module replacement, just bug fixes. So that critical graphics.library bug would be patched, but maybe the colour match stuff wouldn't. The user could decide if they want full module replacement - the OS should be able to run with a SetPatched Kickstart at least until 3.3.

Mobileconnect Quote:

That's not how rom patches work though.

The LVO jump table and library structure are created in memory based on the library (code and read only data) in ROM. SetPatch uses SetFunction() to replace buggy function LVO jump table entries in memory to point to SetPatch bug fixed functions. It sounds like Chris is asking for the traditional SetPatch support even though most modern Amiga users likely load the ROM into memory for performance anyway. How is this "not how ROM patches work"?

@matthey

thats only option for the lazy that don't brun the new rom.

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

I think it's funny that people will boycot Hyperion or Cloanto for shitty behaviour but probably never question many other products they buy where actual human rights abuses are taking place somewhere in the supply chain.

@BigD

No built-in PDF viewer as far as I know.

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

@Mobileconnect

That's *exactly* how ROM patches work. It's what SetPatch was designed to do!

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

Update ...

Today I registered my copy of AmigaOS 3.2 and downloaded the update.

I did not have an unarc program or lha program on my actual OS3.2 installation, so I had to use OS4 to extract the new update from the Update3.2.1.lha file.

The actual update process went well from OS3.2 ...

Now running AmigaOS 3.2.1 emulated with E-UAE and KickA2000.ROM on OS4.



redfox

Last edited by redfox on 23-Dec-2021 at 01:38 AM.

I'll have to update as well. I bought AmigaOS 3.2 when it came out.

_________________
Sent from my Quantum Computer.

@Rob

Quote:

Rob wrote: I think it's funny that people will boycot Hyperion or Cloanto for shitty behaviour but probably never question many other products they buy where actual human rights abuses are taking place somewhere in the supply chain.

I'll consider boycotting Amazon if they ban Visa Credit transactions. They already milk their staff and tax dodge so I think they can pay the charges to Visa!

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

@NutsAboutAmiga

I think it would be more accurate to say that 3.2 is new development, lots of bug fixes, and it has *some* things back ported or re-implemented from OS4. I would assume this would likely continue with 3.2.2 etc.

_________________
AmigaOS 4.x Beta Tester - Classic Amiga enthusiast - http://www.hd-zone.com is my Amiga Blog, check it out!

Why not 3.3?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Because this is a bugfix release, not a feature release. Do you understand how version numbering usually works?

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

@Chris_Y

I do, that’s why I ask. It doesn’t look like the OS devs care much though, as there are clearly added features in 3.2.1 over 3.2 - not just bug fixes. So by saying next will be 3.2.2, is it then also implied that there lots of known bugs still in 3.2.1?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@NutsAboutAmiga
Quote:

680x0 ABI is register based, the AmigaOS4.1 PowerPC ELF ABI is stack based,

This is incorrect BTW, the fact that you claim the PowerPC SysV ABI used by OS4 is "stack based".

The PowerPC SysV ABI, also used by OS4, passes integer/address parameters in registers r3-r10 first, before it falls back to stack based parameters, and floating point parameters in registers f1-f8 first, before falling back to stack based argument passing. (Not counting the r500 cores, where the SPE might use a different floating point ABI.) Function return values are in register r3-r4 and f1, respectively. OS4 uses a slight tweak to this in an Amiga .library context, where the library base is also passed as the first argument to every function. (This might be hidden by "interfaces magic" and macros on the C language level, but this is how it _actually_ works underneath.) This register-window size for parameters covers almost the entire AmigaOS API, so I doubt you'll see a lot of stack-based parameter passing in practice, except maybe in some RTG or graphics.library functions. MorphOS and AROS are also SysV ABI based, btw, with their own quirks on top of it (whch quirks are in place largely to assist 68k AmigaOS compatibility, if that applies on the host platform.)

Ironically, the m68k SysV ABI also exists, used by Linux/m68k for example, and it _IS_ stack based, registers are not used for parameter passing, except for return values. Also, on 68k AmigaOS ABI isn't standardized, and it's described on a per-function basis, so there _could_ be some libraries which use the m68k SysV ABI in an Amiga .library context, and pass arguments on stack. AROS' C library arosc.library is one example, that does this.

Source: I worked on a compiler which had to adhere to these ABIs (and then some), plus documentation is out there if you need to find it...

Anyway, regardless, all of this has little to do why it's not possible to run OS4 software on OS3... It's just two wildly different CPU architectures which can't run each-others software. Technically a translation could be done either way, but performance differences between actual processor implementations make this impractical in the PPC-on-68k direction.

_________________
MorphOS, classic Amiga, demoscene, and stuff
"When a bridge is not enough, build a Viaduct!"
"Strip the Amiga community of speculation and we can fit every forum on a 720k floppy" (by resle)

@Chain-Q

Ok, checked my notes, the registers are pushed to stack, so they be restored, after the function call.

so the PPC does pass arguments by registers, as you say.

Anyway, point is there is a lot that has to go on from 68k to ppc, and from ppc to 68k,
beyond JIT that also need and trap code to handle special cases.

Linux / AROS I never talked about, not same operating systems.

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

@kolla

There are a couple of minor new features but no API changes AFAIK. 2.05 had new features over 2.04, so there is precedent.

3.3 would denote a paid release, so would need some significant new features.

_________________
"Miracles we do at once, the impossible takes a little longer" - AJS on Hyperion
Avatar is Tabitha by Eric W Schwartz

@NutsAboutAmiga
Quote:

Ok, checked my notes, the registers are pushed to stack, so they be restored, after the function call.

Which also happens on 68k, BTW. A certain set of registers needs to be preserved through function calls, and these are usually saved on stack, inside the callee function.

Quote:

Linux / AROS I never talked about, not same operating systems.

Indeed, but they're highly relevant and related, when discussing "modern" ABIs, like it or not.

_________________
MorphOS, classic Amiga, demoscene, and stuff
"When a bridge is not enough, build a Viaduct!"
"Strip the Amiga community of speculation and we can fit every forum on a 720k floppy" (by resle)

Quote:

NutsAboutAmiga wrote: As I understand it OS3.2 mostly back port of AmigaOS4.1 sources code, ...

Could be wrong, Hyperion doesn't have the AOS4.x sources, they are owned by the developers.

@OneTimer1

A few years back at Amiwest they talked about what Exec Team / Trever owned, and what Hyperion owned, every AmigaOS4.x release has to be a collaboration. Between the two groups, Hyperion own reaction something recently acquired. (technically Reaction is a set of BOOPSI classes.)

Reaction is now part of AmigaOS3.2

Last edited by NutsAboutAmiga on 24-Dec-2021 at 12:54 PM.
Last edited by NutsAboutAmiga on 24-Dec-2021 at 12:52 PM.

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

@Chain-Q

Programmig some functions in PowerPC assembly I have to confirm your reply.
Parameters are passed by registers, it's also obvious due Lage set of PowerPC registers.
My question was addressed mainly to API implementation and differences between os3 and os4, and if is planned to make them overlap each other.
In this way every projects could be targeted to os3 or os4 with just by a compiler switch.
Apple did the same in xcode during PPC to x86 migration and recently with x86 to arm migration.

_________________
Pegasos II G4@1GHz 2GB Radeon 9250 256MB
AmigaOS4.1 fe - MorphOS - Debian 9 Jessie