Any idea when we'll have a semi-working GPC simulation?
Depends on which simulation model you mean. The AP-101S emulation is currently mostly for the fun of it, the SimpleGPC stuff can be made working in a rather short time, maybe three weeks.
The SimpleGPC stuff is purely C++ and native code, the AP-101S emulation would need also a large set of tools and source codes to be working. But I can already see the SimpleGPC concept getting to its limits soon, since we also need to structure the code for fitting into Orbiters timesteps, what would be no longer needed for the AP-101S emulation. And the emulation could also simpler be moved to multicore processors.
Coding also for different software versions and software distributions would also be simpler for the emulation.
So, the SimpleGPC should be primary as long as the emulation has no complete tool chain ready. Once it is possible to actually write complete software for the emulation, or convince NASA to give us the original software, the SimpleGPC could be made option and the emulation primary, since the emulation offers a lot of improvements for getting the exact behavior of the GPCs right.
Since both models would share the same I/O implementation (= connection of subsystems to the GPCs), this is currently my focus there... Poscik had stated interest to help there.