Quote:

That's if you consider 3.9 to have been worthwhile at all. You could live quite happily without it, if you already had 64 bit aware versions of scsi.device and FastFileSystem. The rest was mainly a hodgepodge of unnecessary reworks of Amiga preferences in ReAction and some glowicons via enabled via slightly-less-hacky way. Only really worth it for the new HDToolbox and IconEdit. (Although the ppc-native picture.datatype did finally allow you to remap and dither pictures within seconds to an 8 bit screen rather than minutes. Pity many people didn't use 8 bit screens any more...)

As Piru said, the context switching rigged demo that came with WOS was outed as a lie as soon as games compiled for WarpUp proved to have awful performance before being optimised to reduce context switches. At least one WOS proponent who I won't name (but whose name rhymes strongly with Steffen Hauesser... oops) was known to completely disable sound in his ports to reduce context switching without having to do any real work.

Only real advantage of the whole affair was that eventually it became possible to use PUP under WOS with an emulation library and so users didn't need to worry about any of it any more.

Quote:
I do not think EHF was such a disaster, it fits quite well within the HUNK format. And H&P proved with (Storm)GCC, that it was possible to use a "standard" compiler to create EHF object files. FWIW, in 2004 I created a cross-compiler with their sources using Frank Willes pasm and vlink as tools. I successfully compiled mpega-libmad with it as CISC can certainly tell you. BTW, are current sources for mpega-libmad available somewhere?

Recently I noticed that H&P added some EHF support to bfd and gas in their binutils. However, that support was not completed.

Quote:

http://morphos-team.net/files/src/3.11/mpega.library.tar.bz2

For some reason they're not linked on the home page in the correct section. I'll make sure to create a ticket to fix that.

Quote:

I remember that it was more complicated with the WarpOS / PowerUP 'war' back then to get the right version of the software to run on your machine for the endusers.

Why H&P invented their own kernel and not licenced PowerUp is strange.

Quote:

Probably due to ambitions to create a range of PPC accelerator boards or even full systems, starting with the Brainstormer G4. They had no intention of allowing Phase5 to control that niche by defining the PPC kernel standards.

Of course, they completely underestimated the difficulty of designing a PPC board for Amiga and all that was left of their plans was division and a WarpOS that turned out to be absolutely useless to the OS4 guys when they picked it up.

Quote:

I believe they have been deluded by Petro, that a pure PPC AmigaOS was "around the corner" and had to do this quick Frankenstein-like card redesign when Petros promises proved to be groundless. There was no time to write MorphOS-like solution (too many features would be missing). The minimal 0.4 version was ready not earlier than 2001 and the market was long gone by then (as well as phase5 itself).

In Poland we have a phrase saying that "A drowning man will grab even a razor". That's basically what the weird two-CPU board design was - a razor for a sinking company the phase5 was back then.

[ Edited by koszer 10.07.2018 - 13:50 ]

> ambitions to create a range of PPC accelerator boards or even full systems,
> starting with the Brainstormer G4

The start was a little less ambitious with the Brainstormer G3 (to be used in DCE's revised Power A5000). And let me name the failed Predator Plus that arose from the failed Brainstormer G3/G4, and the failed Escena AmigaOne that arose from the failed Predator Plus :-)

Quote:

ELF format is a monster, I remember there were mixed 68k/PPC elfs even, like apple mixed PPC/x86 binaries during Intel transition.

And elves are a bit of monsters!

Thanks Piru. What is analogy? pUp guys went MOS, wOS guys went OS4? (I see warp datatypes and few thingies were endeed OS4 ported) >:)

Well, none of pUp/wUp software does work on my x1000, even it is AOS PPC code it has quite lost relevance. If that ReWarp thingie is done, I might go back in time to explore!

Since I live in not so rich area of Europe had just two friends who ever owned them and likewise wasnt much impressed. 060 impressed me more. And MorphOS as well. Too bad Amiga Inc made the wrong choice of picking Friendes brothers.

Hope one day you could add support for the damned Nemo board aka A1-x1000.

Also, any possibility of latest Odyssey getting 68k (Vamp) and OS4 branches? AROS too. Browser is our week knee, next to Office and Printing.

[ Edited by vox 12.07.2018 - 23:06 ]

Quote:
And they all went reimplementing the very same APIs (AmigaOS API + Cybergraphics + AHI + MUI,...), porting the very same apps, etc...

What a waste.

And now they are pretty much both in the same situation: locked on PowerPC hardware :)

Quote:

Yup the circle is complete. "What if" efforts were joined, we ll never know, we can only thing of each sides strong points.

However its good some things have become cross platform standards like MUI5,AHI, Odyssey ...

As well as we got shared hardware at least in form of Peg2,SAM460ex and x5000. Sadly could be more if there was a better cooperation.

Quote:
Technical considerations did not play a role in their choice. They were offered to get a PowerPC-native port of AmigaOS within months for a measly 25.000 EUR. Only to then get sued by their contractor and ultimately lose the intellectual property rights altogether.

This gave a totally new meaning to "You get what you pay for."

Quote:
Just buy a PowerPC Mac already. They cost a tiny fraction of an X1000.

Quote:
Technically, AROS (on x86) uses a more recent Webkit version, which is an open source browser engine and can be ported by everyone who wants to. You are barking up the wrong tree, I am afraid.

> I remember there were mixed 68k/PPC elfs even

No, there were not.

"Problem with ELF binary format was that [...] mixing PPC/68k code was not possible. Extended Hunk Format (EHF), developed by Haage & Partner, allowed mixing PPC and 68k code to single executable without modifying the existing system if PowerPC accelerator was not installed."
https://en.wikipedia.org/wiki/Amiga_Hunk#Extended_Hunk_format

> I see warp datatypes [...] were endeed OS4 ported

...as well as to OS3/m68k and MorphOS.

Quote:

That is less of a problem. I can get bare mini G4 for 40e, plus added value of Linux working out of box and macOS X. Problem is I do have x1000.

Quote:

I dont see that one included in AROS 68k, or am I wrong? (not all AROS is equal). On OS4 situation is severe.


Quote:

As usual your memory and tech knowledge is right. If I missremembered it was ELF, I remember someone even did the mixed mode.



[ Edited by vox 13.07.2018 - 00:16 ]

>> pUp guys went MOS, wOS guys went OS4?

> they all went reimplementing the very same APIs
> (AmigaOS API + Cybergraphics + AHI + MUI,...)

Has AHI ever been reimplemented? I've been thinking the original implementation is used by both.

> porting the very same apps

Thanks to using and/or reimplementing the same APIs, porting between the two has been mostly a piece of cake, relatively speaking (if no proprietary APIs are used in addition) :-)

> They were offered to get a PowerPC-native port of AmigaOS within months
> for a measly 25.000 EUR. Only to then get sued by their contractor

Let's not forget that they sued their contractor first :-)

>> any possibility of latest Odyssey getting 68k (Vamp) and OS4 branches?

> Webkit [...] is an open source browser engine and can be ported by
> everyone who wants to. You are barking up the wrong tree, I am afraid.

I guess he refers to Odyssey v1.25 that comes with MorphOS 3.11 specifically. Is its source available to "everyone who wants to"?

> I remember someone even did the mixed mode.

And I even quoted who :-)

Quote:
This is blatantly wrong! Check mpega_libmad for powerUP which is a shared library with the PPC part being an ELF file included _in_ the HUNK structure. This required an additional program (eg. Data2Object). This technique can be used for "normal" programs as well.
With EHF you would prelink all PPC code with -r into a single object using a suitable linker (eg. vlink) which would then be used in the final link.

Quote:
Sometimes information on wikipedia is indeed wrong.

>>> I remember there were mixed 68k/PPC elfs even

>> No, there were not. [...]
>> https://en.wikipedia.org/wiki/Amiga_Hunk#Extended_Hunk_format

> This is blatantly wrong! Check mpega_libmad for powerUP which is a
> shared library with the PPC part being an ELF file included _in_ the
> HUNK structure. This required an additional program (eg. Data2Object).
> This technique can be used for "normal" programs as well.
> Sometimes information on wikipedia is indeed wrong.

Thanks for correction. I should have known better than to think that claims on Wikipedia going unchallenged for 7 years means they're probably true ;-)

Quote:
Thank you for providing access to the sources!

Is the MorphOs team interested in changes for the amiga parts?

Quote:
I did not check when this misinformation was entered in wikipedia, but apparently the impossible was done years before.

Furthermore, there seems to be a very well hidden "secret": The last SAS/C patchset for PPC available on AmiNet added support for mixing m68k and PPC code in one executable for powerUP. This is a custom extension of the SAS/C compiler and had to be activated when compiling with scppc with the HUNKOBJ keyword (without this argument scppc produces normal ELF objects). There is little documentation for this feature thus using it is challenging, especially for a shared library, but nevertheless it can be done.

I never checked the structure of the HUNKOBJ objects. It is probably possible to add some basic support for these objects to the amigaos binutils port.