Improvement Patch?http://csv.swordofmoonlight.net/SomEx.dll/1.2.3.4.zipThis patch switches over to a fixed frame rate. I don't know if it's best for everyone, but it certainly helps my system. I think Windows is just an unstable environment for games. But it could depend on hardware. I really don't think there's anything SOM can do to help this.
What I was using until now (since recently really) is average time over 20 or 30 frames. Not doing this caused smoothness to tank. Lately I've been having bad feedback in the control system intermittently and I took it to be frame rate irregularities most likely.
Edited: Note, my main motivation was to determine if the new much improved gait system had math issues or if the troubles with it lied elsewhere. So far it seems the frame rate mainly was at issue, which at least for the gait system is good news.
What I tried today was to fix the frame rate to the closest standard frame rate. That seemed to remove it, although it changes the experience a little bit. Ultimately I decided to defer to the display adapter's reported frame rate where available.
I also added a [Window] extension called "seconds_per_frame" (not fps since this doesn't change the frame rate) that sets the length of the fixed time step. It's not really meant to be used normally, but it can speed up and slow down the game effectively by covering more time. It can be used to speed up the game like people usually play KF2 on emulator.
But what I wanted to see was if the control system held up at different time steps. It's a little hard to judge because it's unnatural to play in slow-motion or sped up speed, but it doesn't appear to have any issues. It seems that the problems arise due to irregular timing. In that case they're probably purely down to human perception and not really errors.
I think this will be the best experience when hiccups occur, but more importantly it seems to eliminate turbulence from the control system. It seems to make jumping last longer or be more full. I might even reduce the jump height. I'm also thinking of reducing the do_lap extension when turning to see if it can make it a little less blurry. (In theory giving more weight to the more recent frame would give the blur some directionality, more like afterimages probably are IRL. Unfortunately this would degrade the do_aa effect. I wonder if the new contrast detection system can help to reason about if things are moving, then give a slight weight to the current frame if so. In that case the do_aa effect can't do its magic anyway, but I don't know if every pixel having a different weight would be odd or if small weights would be noticeable. It's possible even a slight weight could aid the brain to isolate the real image. Edited: in the debug mode build I can hit the grave/tilde key to turn off the effects pass. In that case the blur is not majorly reduced, so I think the improvement will be nice but not night and day.)