Most of it is done in hardware, yeah. As for times, I haven't measured again. I have yet to fix some issues (like the random-ass freeze- I guess the GPU doesn't like gratuitous configuration changes so I'll try to avoid them, and there may also be something with sprites).
Might check for other zero-sized array cases, the GPU doesn't like that at all (renders garbage). Although I think I have fixed them all.
I observed that DKC seemed to take some frameskip. But there's a reason for that. The game enables all layers on both mainscreen and subscreen, but disables subscreen layers via $2130. This part isn't implemented so I end up drawing more than needed.
Same shit happened with the software renderer. Game was slow as shit. Then I implemented that bit from $2130, and it became much faster.
DiscostewSM the idea I have in mind is exactly that. But right now, I'd rather get shit working