Really Complicated Custom Maps Are Laggy

I noticed something about really complicated custom maps in 2.1.X. For some reason, since 2.1.X I have issues with the framerate on really complicated maps. It happens on both SDL and DD. Unfortunately, I can't show gifs of the lag since it happens really badly.

The question is, what custom maps are you talking about specifically?

Duon the Adventurer said:

The question is, what custom maps are you talking about specifically?

Click to expand...

It is definitely a problem on CD City. On CD City, my framerate is around 1 or 2 fps. It also jacks up my ping on netgames. I think the issue is caused by lua, but I'm not sure. It is also a small problem on SF94-City. I noticed custom maps that lag really bad on me have lua scripts. It doesn't happen with all maps that have lua scripts though.

From my own experience, not only lua causes lots of lag depending on what's being activated, but sometimes too many sectors and FOFs can cause lag a while section of a level. Best example is the 2.1 version of the Riders Mod, many levels, especially in OGL will induce an unhealthy amount of lag.

Duon the Adventurer said:

From my own experience, not only lua causes lots of lag depending on what's being activated, but sometimes too many sectors and FOFs can cause lag a while section of a level. Best example is the 2.1 version of the Riders Mod, many levels, especially in OGL will induce an unhealthy amount of lag.

Click to expand...

I think it is more a limitation with the engine itself, because Doom is really old.

How does 2.1.15's performance compare with 2.1.14's on the maps in question? Unless the performance issues are from something introduced with the patch, we're gonna be hard-pressed to do anything about it.

I know that after the most recent update for SC_Heroes in 2.1.14 I was no longer getting lag around the first pillar platforming segment and now I am with 2.1.15.

Well, SRB2 SDL is very unoptimized still. Huge maps, unoptimized SDL, and big lua scripts equal a lots of lag in 2.1.X. I would avoid big lua scripts in netgames since high ping is common when SRB2 is lagging on online multiplayer.

Duon the Adventurer said:

many levels, especially in OGL will induce an unhealthy amount of lag.

Click to expand...

Lag in GL is pretty much entirely dependent on your GPU. Not sure why, but it seems GL only performs well with absurdly tough GPUs. I'm using a GeForce GTX 970 and get full frames in all levels consistently.

Wolfy said:

Lag in GL is pretty much entirely dependent on your GPU. Not sure why, but it seems GL only performs well with absurdly tough GPUs. I'm using a GeForce GTX 970 and get full frames in all levels consistently.

Click to expand...

I used to have a computer that had issues in OpenGL mode. We need to start reporting bugs for SRB2 SDL.

Well, I got some good news about this issue. I made a really big and empty field and the map didn't lag at all. I used ZDBSP to build this map instead of ZenNode. It seems that the lag is an issue with ZenNode itself. All of the SRB2 maps I've downloaded used ZenNode, because ZDBSP is known to cause issues with the netcode. I have not tested ZenNode without the reject itself though. I only tested ZenNode with reject and ZDBSP.

Your testing is not helpful if you admit that there are circumstances under which you have not tested. Please test under those as many people do not have the same lag you do. Also, please test using the exe available here as it solves many common lag issues with 2.1.15's main release. Just so you know - we do not support executables you have compiled yourself, as there may be external factors making them laggier. With the exception of the primary release of 2.1.15, SRB2 executables provided via srb2.org are significantly less likely to give you anything but the best performance experience.

We know why ZDBSP causes problems with the netcode; it's because the game doesn't check whether the length of the REJECT lump is long enough for the number of sectors in the map. The game then reads random bits of memory beyond the end of the REJECT matrix in memory as an extension, and these random bits of memory are definitely not network safe (enemies which do not see you locally may see you server side, or vicea versa). This issue is fixed in the next version of SRB2, whatever version number that may be.

The reason ZDBSP causes this problem with every map using it is because it creates a REJECT lump of the shortest possible length (an empty header lump). It IS possible to recreate the issue if you prevent ZenNode from saving the REJECT properly, but it's relatively unlikely.