Closed dankcushions closed 8 years ago
I do think the people complaining about input lag are a relatively small number - and in most cases is due to the screen (ie not in the right mode, or a bad screen). I personally do not notice any lag with fceumm, however some people might be more sensitive to it.
We definitely need more testing, but I am reluctant to switch based on 1 frame of extra lag - it's also easy enough for people to switch. As you know, changing defaults often causes problems :-) I would need further convincing - also in terms of compatibility. Some rpi0/1 tests would be useful though.
Have you read the link? I trust the assessment given the detail they've gone into, so I think it's for sure better in these terms, but agreed - I'm personally going to change but we need to get rpi0/1 confirmation first.
I read it yes. Hence specifically mentioning the 1 additional frame delay
1 frame might not be much, but considering that RetroPie with fceumm and snes9x-next (before my fix that was just committed) has 7-8 frames (117-133 ms) of lag even before adding the display's input lag, I'd say that we need to do what we can to reduce it. Adding a display with 3-4 frames of input lag makes the setup as a whole approach 200 ms. The reason that I started looking into this from the beginning was the fact that those 200 ms (yeah, my TV is slow) made Super Mario World really hard to play.
However, I do see where you're coming from. Although I personally prefer Nestopia and have been using it for a long time, both on my Pi 2 and Pi 3, I definitely see how switching could cause issues. Unless someone can provide compelling info regarding compatibility and RPi 0/1 tests, it might be better to simply wait for a fix in fceumm.
nestopia has better accuracy and less input lag yet fceumm it's faster and has better compatibility with non-intro, unlicensed and specially hacked roms. Personally nestopia as standard core might lag on Rpi1 plus some users might complaint about some games being unable to run...
the scripts can handle one default for one platform, one for another. that was my assumption - nestopia for 2/3, fceumm for 0/1. perhaps anyone who cares enough about the input lag will read the wiki and do the change manually anyway...
I've found a fix for the issue and created a pull request here: https://github.com/libretro/libretro-fceumm/pull/46
Talk about small fix (moving one line of code one line up...).
even better! cheers @Brunnis :) I recommend we update the binaries once that one gets pulled. Closing.
I have updated the binaries.
Please see http://libretro.com/forums/showthread.php?t=5428&p=41746&viewfull=1#post41746 - until https://github.com/libretro/libretro-fceumm/issues/45 is resolved (and I'm not sure it will be given that lr-nestopia exists), lr-nestopia has measurably better input lag. For this reason I think we should adopt it as the 'main' NES emulator, with lr-fceumm as the 'opt' (currently it is the other way around). I am sure I've seen some users complain about input lag in the NES emulator, so this should help!
One caveat here - I understand that lr-fceumm runs faster than lr-nestopia, but is this measurable on any Pi systems? I think there was a reason we have lr-fceumm as the current default and it might be this. I have a Pi 3 and I get 60FPS on both cores. Any Pi 2/1/0 users who can check? If lr-fceumm runs better, we could put a platform check in the script.