I was playing surf (made by @Quaternions) with @Zomebody and noticed that I was lagging in certain areas where he wasn’t. Now the obvious question is ‘how good is your pc?’ but this is not the case. I was maintaining a full 60fps but in this certain area in the map I dropped to just a few frames.
There are quite a few people with good computers that have complained to me about lag on my game, however this does just look like it’s specifically from the portals. Is there some kind of magical profiling tool I can use to see how many times per second functions are being called or loops are running?
To investigate this you’ll have to get to the area of the map where it’s very slow and record & upload the microprofile capture here (Ctrl+F6 -> Dump -> 32 frames; this’ll save an HTML file to C:\Users\Younite or whatever your Windows account name is).
It’d also help to do the same in the area of the map where it isn’t this slow.
Uh oh, at least that helped us find an issue with on-screen profiler rendering on AMD cards :-/ (with a fix shipping next week). When you do get an “insane lag” case though please make sure to post a capture.
Sorry for bothering you, but while we are on the topic… since this weekend I have started to notice some lag spikes in a game where me and few others are working on. The most part that was changed in the past few days is basically geometry (buildings).
I am unsure what is causing this, the microprofiler indicates “Physics” as the main “cause” of the spikes. The main thing that is being affected is CPU and Physics in the performance stats, GPU stays stable.
I can also see LuaGC, Server HB etc in the dev console taking a dip but I think this could be due to the client not getting this data on time.
As far as I am aware, we didn’t add non anchored objects to the place so I am unsure what is causing this. The spikes also appear on a +/- frequent interval.
Here is a log to a dump of the microprofiler (the same basically happens around the whole map and practically for everyone who we tested with so far).
This looks like in some cases there’s a LOT of narrowphase work occurring intermittently, which may mean that some non-anchored objects compute their collisions with some anchored objects, or something along the lines.
Can you create a separate thread for this, put the same link to the microprofile dump, and additionally specify a place link and where in the map you should be for this to happen? I can then tag our physics engineers and they’ll take a look.
