AROS Adding 64Bit And SMP Support

[AMIGASYSTEM]

After many years of active development, AROS finally seems to be able to 'evolve' the now 30+ years old architecture of the Amiga API. The original Amiga computers from Commodore brought to home users and professionals the first pre-emptive, window based operating system at affordable prices, although its kernel was tailored to the single Motorola 68000 CPU mounted on the machines. After Commodore's demise in 1994, a long debate started about the evolution of the Amiga platform and, although many announcements were made, current AmigaOS 4.1 is still a 32bit-based, single-core oriented operating system, and the same is true for Amiga-like alternatives MorphOS and AROS.

Things, however, are changing. In his weekly survey about AROS progress on AROS-EXEC.org and Amigaworld.net, Krzysztof Smiechowicz talked about "Work on handling additional CPU cores in x86_64 AROS kernel", adding "Initial version of SMP scheduler has been introduced in AROS i386/x86_64 kernel" just a week later. In the following weeks, a screenshot from coder Nick Andrews and a video on Youtube showed a 64-bit version of AROS, runnning on multicore AMD and Intel processors, handling 4 and 8 cores correctly.

SMP is being added to AROS by experienced coders Nick Andrews and Michal Schulz, and while it is not available in public nightly builds just yet, there is finally the chance to see an Amiga-like operating system handling modern CPUs properly News Source:

Sysmon Update AROS progress
[ Show youtube player ]

[matthey]

After "pre-emptive" you probably want to add "multitasking".

Is it wise to announce SMP support this early? How much testing has been done?

[Locutus]

Why not? the code is there, you can review, build, verify as you want.

Release early, release often, this isn't 1985 anymore :-)

[matthey]

Originally Posted by Locutus

Why not? the code is there, you can review, build, verify as you want.

There could be some instability issue which makes all AROS x86_64 software incompatible for example.

Originally Posted by Locutus

Release early, release often, this isn't 1985 anymore :-)

Right. AROS can always have a 3rd ABI with every developer recompiling for every ABI and target CPU combination .

[wawa]

@matthey:

There could be some instability issue which makes all AROS x86_64 software incompatible for example.

this is work in progress, there is only some test programs, fractyls running on multiple cores. the backwards compatibility is maintained, but one needs to keep an eye on it. lately snoopdos stopped workinng, i suspect some semaphore change in exec, fortunatelly there is aros native snoopy. but im of course concerned about such issues.

Right. AROS can always have a 3rd ABI with every developer recompiling for every ABI and target CPU combination .

given the code already compile for one platform of aros it will usually compile for another, 32 or 64 bit, smp or not.

of course things like pointers need to be defined to a proper type (IPTR if i recall right) which defaults to ULONG on 32bit archs. and on 68k you need to remember to use parametrizing macros and proper syntax for hooks, which will otherwise work on x86, even if wrong, but fail on 68k.

so 68k is actually least forgiving and proper reference platform.

[Locutus]

Originally Posted by matthey

There could be some instability issue which makes all AROS x86_64 software incompatible for example.

Such is the way of a development process, i really dont see how that would keep someone from doing a early release. This is exactly what branches are for after all.

Right. AROS can always have a 3rd ABI with every developer recompiling for every ABI and target CPU combination .

does that really matter? most x86 'users' are on the v0 ABI anyway aren't they?

[matthey]

Originally Posted by wawa

given the code already compile for one platform of aros it will usually compile for another, 32 or 64 bit, smp or not.

Is SMP being added to ABIv0 or ABIv1? Do AROS programmers and the AROS group/team have the organization to give "official" news statements on behalf of AROS?

Originally Posted by Locutus

does that really matter? most x86 'users' are on the v0 ABI anyway aren't they?

My understanding was that x86 AROS remained on ABIv0 for compatibility with already compiled programs like on Aminet. Eventually it was to be merged with the more advanced and AmigaOS compatible ABIv1 (which I believe the 68k is using). I guess the plan was to eventually break x86 binary compatibility anyway?

I would have preferred the lead AROS programmer(s) to give the ok on the SMP announcement when/if they decided stability was adequate (less stress for them) and perhaps give an updated plan for future ABIs but I guess anyone can create an "unofficial" news announcement. I can understand some people's excitement even if it may be better to wait for an "official" announcement.

[wawa]

Originally Posted by matthey

Is SMP being added to ABIv0 or ABIv1? Do AROS programmers and the AROS group/team have the organization to give "official" news statements on behalf of AROS?

abi_v1, which is the same branch, amiga-m68k target exists in. abiv_0 is being kept for compatibility, with as much as possible fixes backported from abi_v1 for x86 users.

My understanding was that x86 AROS remained on ABIv0 for compatibility with already compiled programs like on Aminet..

correct. for the still undfined time being till v1 is stable and the users ready for transition.

Eventually it was to be merged with the more advanced and AmigaOS compatible ABIv1 (which I believe the 68k is using). I guess the plan was to eventually break x86 binary compatibility anyway?

yes.

I would have preferred the lead AROS programmer(s) to give the ok on the SMP announcement when/if they decided stability was adequate (less stress for them) and perhaps give an updated plan for future ABIs but I guess anyone can create an "unofficial" news announcement. I can understand some people's excitement even if it may be better to wait for an "official" announcement.

i think there will be some sort of information available on aros site as soon as its advanced enough. usually deadwood writes some summary, except regularly posting week progress reports on various sites.

here you have posts by michalsc, one of both main devs mostly concerned with the current smp development:
http://amigaworld.net/modules/newbb/...e=flat&order=0

the orther one, kalamatee posts only now any then on aros exec, he has initiated that work, and he commits the most, so i wouldnt expect that much statements from him.

[AMIGASYSTEM]

Video test with a fractal generator, we are still only beginning and already notice the difference in the calculation, 10 seconds rendering time on a 4 cores compared to 30 seconds of a single core.

AROS SMP Test#4, Fractal Generator
[ Show youtube player ]

[AMIGASYSTEM]

AROS running on a 12-core AMD Ryzen R5-1600X based virtual machine:

[ Show youtube player ]

[wawa]

looks like moving the window interrupts computation on all cores, s long as its is not released.