I dont mean to sound negative, but what exactly would its use be, other than having 1 iso that runs on multiple archs (though admitidly through emulation..)?

I mean, the parts of aros that this would be usefull to, are the parts that need developed in the first place imo. which makes it use-less.

All the code is available for the rest (read - apps etc), so once youve got the building blocks in place (ie aros running on PPC in the first place) the rest should just compile and run.

Thats my opinion anyhow ... sounds like effort would be better spent just getting the PPC port to function.

True.


I don't believe you could ever have 100% compatibility, but maybe we could have a mechanism that could run at least some specific code. There are a few obvious candidates for usefulness (like datatypes and xpk libs perhaps). But even if that won't work, something useful might still come out of it.

Or so I hope.


m

Your interpretation is mostly proper. But please note, that there is limited amount of syscals, and parameters for every single syscall are endian-safe - even if you need to convert them, you know exact types.

Also, linux kernel doesn't provide much global variables and user-accessible data (like sysbase and all libraries devices and such in AmigaOS) - you are forced to do everything through the system. So, there is only one place to apply endian conversions. Sadly, AmigaOS (AROS) is different here :(

Well, no.

qemu is, primarily, an "exotic-Linux"-on-x86-Linux emulator. Therefore the author has made provisions for running Linux binaries intended for different CPUs.

So in the non-x86 CPU scenario, we have:

x86 Linux program -> x86 glibc -> qemu -> non x86-Linux kernel

The paragraph on "you need to download .." is just to say that you need to get the x86 glibc separately.

The scenario above could just as easily be:

x86 Linux program -> x86 glibc -> qemu -> Linux emulation -> real OS

where "real OS" could be just about anything that can emulate Linux system calls. I know NetBSD can, FreeBSD most likely can too.

Or, even:

x86 Linux program that doesn't use libc -> qemu -> Linux kernel (or equivalent)

..in which case there's no need for an x86 glibc since our program is not going to call it anyway.


Now, if we don't intend to run Linux apps and use the special provisions intended for that (syscall "forwarding"), there's no need for an x86 glibc. We'd probably need a lot of hacking on qemu itself though to get our emulated program to communicate with the outside world in a meaningful way. But, in principle, you should be able to run a say, PPC program under qemu and have qemu running on x86 - even under AROS.

That's my interpretation anyway.


m

well, i thought it was interesting (mad dreams of a 68k datatype..lol)...so now it will be logged! =o)

http://wuarchive.wustl.edu/~aminet/util/shell/list68k.readme

michal, thanks! =o)

so, that qemu doesnt help at all! =o\

just try to make this topic of something usefull (as the first post is not) here is another link. (i know the ardi thing, but i think most arosian know, so i will try to make my best)

http://www.vx68k.org/libvm68k/

http://www.lancs.ac.uk/ug/chart/fyp/

Cite from qemu docs:

"On non x86 CPUs, you need first to download at least an x86 glibc"...

In case of AROS that would mean you would need not only eg. m68k emulation, but also the m68k exec and other libs used by the binary. Basically the AROS ROM for UAE emulator :)

http://fabrice.bellard.free.fr/qemu/qemu-doc.html#SEC33

qemu allow to run elf bin on other archs. any comment?
(course when i saw it i remeber our beloved/hated m68k-emul topic)

well, just hope it help, or at least its one more general info! =o)

regards,