(Bounty) Qemu port

[anothername99]

This bounty is for porting Qemu to libretro. To complete the bounty, we need these features:

• Straight-forward to use. A machine config should be loadable through XMB's Load Content, and it should contain all essential information about the emulated machine.

• Support relative paths in machine config

• Only the recompiler is needed. KVM can come later (or not at all).

Bountysource link: https://www.bountysource.com/issues/61468869-bounty-qemu-port

[ghost]

What kind of machine config? Is there an existing format or would something new have to be created?

CD/DVD image swapping through XMB

I think this would likely alienate too many people from even attempting to port the core as not many people understand how this works in RA nor want to... IMO it would be best to leave that part up to the existing RA team if anything is needed there... cores need to be able to work properly in other non-RA frontends also.

Only the recompiler is needed. KVM can come later (or not at all).

Is there an expectation of certain guest architectures to be supported? Like x86, x64, ppc, sparc etc. I would assume it would have to be a separate core for each arch due to how qemu works.

[anothername99]

A plain text file with the typical commands should be enough. Something like:

qemu-libretro -cpu pentium -boot order=d -drive file=disk1.cow -m 1024 -soundhw es1370

If there is a better format, we can go with that too, as long as we don't put these things under Core Settings because that would break any emulated machine too easily. Relative paths are also important because the user might move the emulated machine to another harddrive for instance.

cores need to be able to work properly in other non-RA frontends also.

Are there frontends that don't support swapping disc images? I imagine this to be a big problem for cores like Beetle PSX too. I'm ok with dropping the requirement for this first bounty though.

Is there an expectation of certain guest architectures to be supported?

I kinda assumed that qemu is one big thing. Would there be additional work necessary to support all guest architectures? In that case, the bounty hunter can pick one guest architecture. I'm mainly interested in x86, arm and ppc.

[ghost]

qemu itself builds completely separate programs for each arch, and all the machine types share common function names, so yea you can't easily build all the arches into one binary, it was never designed to work that way.

[anothername99]

We have some cores that are split into different binaries (e.g. Reicast vs Reicast OIT, Beetle PSX vs Beetle PSX HW). That shouldn't be an issue in my opinion. The question is whether the different arches require more porting work. If not, then I want all arches (in different binaries if needed).

[Kreijstal]

Ah wow this is very interesting

[vadosnaprimer]

Does this thing do anything useful? https://gitlab.com/oxr463/qemu-libretro