faatal

Merging optimizations to a19.

Very playable. I played it briefly last week to test feral sense. All the normal stuff works fine, just some features like the block shape UI update are in flux.

Stealth still works. Feral sense is basically zombie hearing and seeing x2.5, which is mod-able in xml and may be tweaked before release.

Low. I just read a Unity is getting DLSS article and it seems to just be the hi def render pipeline getting it, which we are not planning on updating to, since it would be months of rewriting gfx code and shaders. From what I've seen. A lot different. Much less human.

We are planning to redo/improve/change/something with the UI someday, so it might actually happen, but to be safe we can just say there is a 1000000% chance it will never happen. Is that good?

Settings are nice and all, but they also clog up the UI and can be confusing to a lot of players, so there is a line. I already had to move 3 settings from tab 2 to 1 to even make room for this one setting, since we don't really have room for more tabs. Maybe someday we will overhaul the UI and get more room in general.

There are productivity benefits for us like I can profile in the Unity Editor and get release mode timings similar to making a build. Players will probably see a small performance improvement, but that depends on your particular hardware. No. Blood moon code specifically targets players, so hear/see does not matter. The next alpha will be out .1% sooner.

No special situations other than it being on or off. Like it is night, so those zeds can detect better. If they would hear you banging from 20m now they can at 50m or they would see you at 30m, but now 75m (made up numbers). Code dealing with what is indoors or out has not changed.

There is no indicator other than zombies coming to kill you like they do any other time you are detected. Dynamic music would react when they target you like it does now. The system simply gives zombie hearing and seeing distances a big bonus (defined in xml), so they hear and see you more easily. The Feral Sense bonus is either on or off depending on if set to day, night or both.

Added Feral Sense option (None, Day, Night, All). I'm still tweaking it, but I expect to be playing with it set to Night and actually being afraid of the dark.

Most of the optimizations have to do with making the game smoother, not higher peak FPS, so less frequent and milder FPS drops. For example, I can run at 60 FPS, but it would often drop to 30, 20, 15 for multiple frames, which feels bad. Now I rarely drop below 30 and it just feels much smoother. I have yet to try A20 optimizations on lower end computers, so results there many be more, less or the same as a higher end computer.

I tried Vulkan last month on Windows and the Unity editor crashed in Vulkan mode while making a build. I have a ticket open with Unity and they have reproduced the crash. I stopped testing at that point and did not try a normal build and switching it to Vulkan. I will probably try again once Unity fixes the editor crash.

I actually tested DX12 a few weeks back (you just add to the project as a gfx choice), but it caused a crash when the occlusion manager was on. Unity replicated and thinks it was nVidia's drivers, so I retested yesterday with the latest drivers and did not crash, but I do get a DX12 error (x2) on world exit, which I told Unity about. We do want to support it, but FPS seems the same. I still have DX12 enabled in the Unity editor, which also has that error (x2), but they don't seem to break anything other than making the console open.

It is on the back burner. Optimizations basically got bumped to higher priority. Just plans. We may not do the Unity upgrade for A19, since there is more risk, but many of the optimizations can be moved over without updating Unity.

I've used that slider in my own games. Only regret not putting more choices for values. I'd prefer 5 or 10 % increments for many of our options. Someday.

This is very subjective. I would fully expect some people to hate it, since you can always find someone who hates anything. Looks cool to me, but I have not seen one in person, so can't say 100% that it is great.

Got the baking thread working today and with full baking quality instead of simple/fast. Full was required with how Unity implemented it. Hopefully I won't find any issues in final testing tomorrow and that will get committed.

Recent a19 changes have to do with Twitch. Map restart should not be required for the optimizations I'm doing. Data formats are not changing.

Many of the a20 optimizations will move to a19 easy enough. I'm even thinking we upgrade a19 to Unity 2020.3, since it was not hard and some optimizations, like mesh baking, need newer Unity versions.

We just finished our project review last week with Unity. Made a few performance changes as we did that. We have tickets open with them on Vulkan and DX12, so might see some stability/crash improvements in that area. Improved our texture atlas tool, since art complained about how slow it was and caused errors using during game play. Now I'm working on collision mesh baking off the main thread, since that was being a pain to frame split from the vis mesh and Unity pointed out they added BakeMesh a while ago (2019.3). Trees are certainly not done and are on my list of FPS hogs. They are not consistent due to changes over time, like old speed tree vs new speed tree and we may be changing how their shadow casters work and the excessive choices we use for far LODs.

Hopefully not, since it is a good feature to keep. Punted in this case means not A20. Feral Sense has to do with the level that zombies can see and hear you, not blood moons or pathing. Feral sense at night for instance, would make them find you much easier, so if you go out at night, biome zeds will find you.

I'd give it a 95% a20 will be on U2020 LTS.

I would think not for horde zombies. If you ran into a POI on horde night, then POI zeds would follow sense rules? Be a while before this gets worked on. We pool a lot of stuff, but you often get spikes anyway. Even changing parent from the pool to another owner can cause spikes. Instantiating after the first time is typically cheaper, since Unity keeps resources around, so instantiate is often not that bad. Unity 2020.3 LTS released today and in testing so far has been working great and in general, due to optimizing, the game is running quite well now.

It is not about distance or rendering or scripts. All those GameObjects and their components are expensive to create, which means lag spikes. It need further analysis to find all the slow parts and find solutions for each.

As Blake_ had said, there are many issues with that. The main overhead is instantiating the entity, not AI or animation. There are ways to speed up spawning and they are not trivial, but it will probably be improved at some point.