[pve-devel] cpu type ppc/s390

[Thomas Lamprecht]

On 10/7/19 7:20 PM, Tomas Dalebjörk wrote:
> thanks for feedback 
> 
> you know that power is an open architecture that are growing on the market as it performs better than x
> 

I never saw or used one, just heard some non-ideal stories about
vendors[0]. So still waiting before calling this the next bigger
thing one need to be able to buy HW to a price that gets you at
least somewhat the same as amd64 HW does. We had the same wishes
for arm64 over 3 years ago, a real "killer" platform is still not
available - either old stuff, which while many cores it has so bad
interconnects that you can not do anything serious with all of them;
or newer stuff which you cannot get if you are not M$ or the like so
not much of use for our community. I'd rather see RISC-V processors
to gain traction, but also there we'll only be able to do anything once
the platform are ripe _and_ publicly available to a sane price.

cheers,
Thomas

[0]: https://drewdevault.com/2019/09/23/RaptorCS-Blackbird-a-horror-story.html

> Skickat från min iPhone
> 
>> 7 okt. 2019 kl. 17:17 skrev Thomas Lamprecht <t.lamprecht at proxmox.com>:
>>
>> On 10/7/19 4:54 PM, Tomas Dalebjörk wrote:
>>> ok
>>>
>>> that is sad
>>> as qemu support other architectures
>>> so it cant be that difficult 
>>>
>>
>> Yes, qemu itself supports them, that's not the issue.
>> But one needs to do the whole management around the CPU and it's platforms
>> hardware, that's what Proxmox VE is actually there.
>> And that can be huge work depending on platform differences, testing, adapations
>> to future changes specific to that platform and all the other little things add
>> up quick.
>>
>> The current implementation[0][1][2][3] can handle amd64 (duh) and somewhat
>> experimental arm64, while that part is rather limited and assumes mostly amd64
>> like HW for the rest of the system. If you check out the implementation
>> you'll see that it's not all that easy, I mean maybe to bring some proof of
>> concept up, yeah, but to actually have it fully integrated, tested with
>> different possible OS and so on it becomes a non-trivial amount of work.
>>
>> Is it dooable, yeah, sure. But we have no need for it, and so do well over
>> 90% of our users (ballmarked). So why invest the time there instead of doing
>> other feature/bug fixes which help much more users?
>>
>> Note that you can always do yourself an implementation, we're opensource and
>> contribute after all. Maybe it's much easier to do than I currently belief (have
>> not looked deeply in PPC, but having the arm64 part in mind I'd guess that I'm
>> not completely off) and if it's somewhat nicely integrated into the current
>> stack, not affecting the current amd64 parts in any negative way, we can take it
>> in as experimental in the project too.
>>
>> cheers,
>> Thomas
>>
>> [0]: https://git.proxmox.com/?p=qemu-server.git;a=blob;f=PVE/QemuServer.pm;h=83762607dcde15acca06bbe877157cbbcf993f1f;hb=HEAD
>> [1]: https://git.proxmox.com/?p=qemu-server.git;a=blob;f=PVE/QemuServer/Memory.pm;h=b579955037f198eb0163eae7a3d97e81a5cc4e52;hb=HEAD
>> [2]: https://git.proxmox.com/?p=qemu-server.git;a=blob;f=PVE/QemuServer/PCI.pm;h=e0bc1a24f104aa3beef08aa909cf4ca7c08a8aa2;hb=HEAD
>> [3]: https://git.proxmox.com/?p=qemu-server.git;a=blob;f=PVE/QemuServer/USB.pm;h=d328148539fc9245a2ac58fb442e5e307578b7bd;hb=HEAD
>>
>>> Skickat från min iPhone
>>>
>>>>> 7 okt. 2019 kl. 12:32 skrev Dietmar Maurer <dietmar at proxmox.com>:
>>>>
>>>> 
>>>>>
>>>>> I would like ask if it is possible to add other CPU types to the proxmox admin interface 
>>>>> Such as PowerPC (ppc64le/be) and s390x
>>>>
>>>> There are no plans to add ppc or s390.
>>>>