>>106783249
This thing is going to run like ass if they’re trying to emulate the hardware. On modern consoles you just shim all the APIs and JIT the non-x86 assembly into equivalent x86. Reimplementing the hardware level shit is something that stopped being sensible after the SNES.

>>106783249
>Switch 2 emulator
>On github
Some people just don't learn, do they?

>>106783300
>On modern consoles you just shim all the APIs and JIT the non-x86 assembly into equivalent x86
I always wondered how they achieved the performance but that would explain it. But wouldn't it make sense to aot compile it once and translate it that way to x86 or is that too much effort?

>>106783374
Takes too long, easier to JIT and cache it, and the results are typically the same. It’s also not easy to know ahead of time what will actually be executed, and you’ll have to patch and intercept various calls to APIs anyway. You don’t have all the information the original compiler would’ve had, and thus can’t make sensible optimisations that aren’t already there. Most people also aren’t willing to wait for 30 minutes for their game to start, so the only stuff that does get AOT treatment is typically shaders, because they cause stuttering if they’re handled on the fly

>>106783249
>doesn't work and isn't useful
>designed such that it never will be
who cares

>>106783374
>wouldn't it make sense to aot compile it once
No. Why would it? Same performance but you have to wait half an hour before playing the game?

>>106783510
Yeah fair
>>106783545
>Why would it?
Thought maybe compile/translation times would be faster by some degree when done in one pass and upfront instead of doing them all the time while the game is also running