You should play around with the RetoArch run-ahead option to help with input lag
https://www.libretro.com/index.php/r...nahead-method/
Using that on my Nvidia Shield and a bluetooth 8bitdo, I get less input lag than I do on real hardware. It takes some trial and error to figure out how many frames to set it to, it's helpful to bind hotkeys to pause and frame advance to do this, so you can measure exactly how many frames it takes between pressing a button and showing an action on-screen. My personal setup is NES with Nestopia core with 1 frame of run-ahead, and SNES with SNES9x with 2 frames of run-ahead. I can play both Punch-Out and Super Punch-Out with no issues using those settings, which is basically the gold standard for input lag mitigation