archived-dots

[rustic rain]

spawn random trees

like paint tool

over terrain

then you need editor code, that yes spawns Game Objects

with all authoring components

basically designer tools

have nothing to do with ECS yet

[ocean portal]

The V Rising team does something close to it 🤷‍♀️ https://www.twitch.tv/videos/1517796421 at 52:00

They did mention they had to modify the package to get Entities running in the editor properly, but didn't elaborate what exactly

[rustic rain]

well yeah

there are 2 ways to achieve it

the one I mentioned above

or through procedural generation during conversion

so either your tools add game objects into subscene during editing mode

either you add game objects that represent instruction to conversion how it will be converted into whatever you want

for example you put cube with special tag and it gets converted to random bush

[ocean portal]

Yeah at least there's a way, even if not as simple as I first thought it would be. I'll have to think about the best approach, thanks for all the input!

[errant hawk]

@ocean portal Alright, after a bit of digging the best way to create a custom procedural level generator editor tool is to do something similar to this:

• Create level generator scene (NOT subscene)
• Add a GameObject at the root of the scene for you generator script to be attached to
• Create another scene with the name like "Level 1" (again, not a subscene)
• In your generator script, get a reference to that scene and load it as additive.
• procedrually generate stuff using GameObjects like you would for a non-ECS project and move them to the "Level 1" scene.
• mark scene as dirty, and save when necessary.
• Since it's a scene, you can loop over all the GameObjects in that scene pretty easily and set authoring component fields, add/remove components, etc
• Profit $$$

also, here's an example script for adding a GameObject when you reset the script:

    public class TestAddGameObjectScript : MonoBehaviour {

        private void Reset () {
            Scene scene = EditorSceneManager.OpenScene("Assets/Scenes/SomeScene.unity", OpenSceneMode.Additive);
            EditorSceneManager.MoveGameObjectToScene(new GameObject(), scene);
            EditorSceneManager.MarkSceneDirty(scene);
        }
    }

[ocean portal]

Ohh and it could be split in to subscenes later

[errant hawk]

yep

subscenes are literally normal scenes

[ocean portal]

Yeah, there's literally 0 difference right?

[errant hawk]

just loaded by a script called "Sub Scene"

yes, they're normal scenes

[ocean portal]

You just get an entity representation which you can instantiate

[errant hawk]

What?

[ocean portal]

When you do this sceneSystem.LoadSceneAsync(sceneGuid, loadParameters); it returns an entity representation of the subscene, which I believe you can even instantiate multiple times with an EntityManager

[errant hawk]

I assume so, yes

[rotund token]

at runtime subscenes are just a giant chunk of pre-converted entity data

the scene no longer exists

[ocean portal]

Right. I kind of wish the conversion system didn't exist and we could just deal with pure entities :P

[errant hawk]

yes but this is an editor tool

[errant hawk]

What I complained about when I first started using DOTS lol

[rotund token]

what everyone complains about for whatever reason

then i link them the gdc talk

and we go on the same little journey

[ocean portal]

Maybe one day when everything is dots and go's don't exist

I mean in a few decades maybe :P

[errant hawk]

But here's the thing, an Entity is not a "singular object" like GameObjects.

it has components that can be scattered across many points of memory

GameObjects being managed doesnt have that issue

granted Entities are chunked, but that's beyond the point im making

[ocean portal]

Yeah they're a bit different, sure, but I don't think it would stop "prefab entities" existing for example

[errant hawk]

That would be just normal prefabs since they're serialized the same way as scenes

you can get a prefab reference and have it be used during creation of entities

[ocean portal]

I guess, still the authoring workflow irks be a bit

[errant hawk]

Yeah not disagreeing

idc if under the hood it's GameObjects, just make the workflow a bit more intuitive

From what I've been told it's to not make all the editor tools useless

Oh, also I think something entities-only is planned because their roadmap has a DOTS runtime as a potential feature in the future

[ocean portal]

For sure they have good reasons to keep it like this, at least for the forseeable future, just daydreaming 😅

[rustic rain]

no reason to tbh

game objects are easier to understand for humans

entities are easier to work for CPU

having both is good

[ocean portal]

Maybe, can't see in to the future, but I wouldn't be surprised if everything was entities in the backend at some point. But even so the difference between a go and entity is pretty small, so this gets easily in to semantics

[rustic rain]

difference is huge

entity is just data

game object is self registered object with tons of allocations

[ocean portal]

This is the semantics part I meant 😅 Cause I could argue that data = allocations

I know there's a huge difference in the backend, but we were talking about the authoring side mostly

[rustic rain]

key point is self registered here

[ocean portal]

I've come to like the entity way more, it feels easier with all the help you get from unity, so I could also argue go's aren't necessarily easier for all humans ^^

[rustic rain]

GO's support inheritance 😅

can't do the same with entities

[ocean portal]

What if I don't design with inheritance in mind? :P

[rustic rain]

it's not always for game logic though

sometimes it's for other stuff

like UI

[ocean portal]

But you can do inheritance without GO's too

[rustic rain]

yeah and here you go recreating everything GO's already have yourself xD

[ocean portal]

I forgot what the argument was, I can only say I haven't missed the managed components

[rustic rain]

the need for GO's in the future 😅

[ocean portal]

Hmm yeah I can see how they might be around, but I usually steer away from the GO stuff even in GO projects and just use custom classes which store references to components, so I personally wouldn't mind if GO's didn't exist :P But yeah ppl are different, it was just a far off throw thinking about the future in a few decades or even more, tried to defend it, failed :D

[viral sonnet]

of all the things to complain about people chose the authoring workflow most often 😄 so weird

[ocean portal]

The hardest part in the ECS learning process for me has been the hybrid approach, it took months to understand how to interact with both sides from a system for example. Maybe I'm just slow idk

[viral sonnet]

i mostly blame that examples and samples are hard to find. once you get the grasp it's very straight forward. just a lot of bad terminology like hybrid itself but then you want to add a MB component to an entity and you have to call AddComponentObject. Then the whole thing about (the near obsolete) CompanionLink that i've given up now because it's mostly useless by design.

[ocean portal]

Pretty much this^ Once I saw stuff like go transforms in components it became a lot easier, but there was a lot of stuff to shift through to find the good stuff like that

Same goes for CopyTransformToGameObject and the like

[errant hawk]

because it's shit. Entities are a seperate workflow. forcing entities and GOs to work together isn't the best idea

[pliant pike]

I don't suppose anyone knows what happened to this method or what it doesCollectionHelper.AssumePositive()

my guess is it works like math.abs()

[errant hawk]

IIRC it's now an optimizer for the burst compiler

[dense crypt]

Unity.Collections.CollectionHelper.cs:

/// <summary>
/// Tell Burst that an integer can be assumed to map to an always positive value.
/// </summary>
/// <param name="value">The integer that is always positive.</param>
/// <returns>Returns `x`, but allows the compiler to assume it is always positive.</returns>
[return: AssumeRange(0, int.MaxValue)]
internal static int AssumePositive(int value)
{
    return value;
}```

Where Unity.Burst.CompilerServices.AssumePositive:

/// <summary>
/// Can be used to specify that a parameter or return has a range assumption.
/// Assumptions feed directly into the optimizer and allow better codegen.
///
/// Only usable on values of type scalar integer.
///
/// The range is a closed interval [min..max] - EG. the attributed value
/// is greater-than-or-equal-to min and less-than-or-equal-to max.
/// </summary>
[AttributeUsage(AttributeTargets.ReturnValue | AttributeTargets.Parameter)]
public class AssumeRangeAttribute : Attribute
{
    /// <summary>
    /// Assume that an integer is in the signed closed interval [min..max].
    /// </summary>
    /// <param name="min">The inclusive minimum value.</param>
    /// <param name="max">The inclusive maximum value.</param>
    public AssumeRangeAttribute(long min, long max) { }

    /// <summary>
    /// Assume that an integer is in the unsigned closed interval [min..max].
    /// </summary>
    /// <param name="min">The inclusive minimum value.</param>
    /// <param name="max">The inclusive maximum value.</param>
    public AssumeRangeAttribute(ulong min, ulong max) { }
}```

[pliant pike]

how do you even find that

[dense crypt]

Rider Decompiler 😄
You can find it if you go through the package source as well:

[errant hawk]

Visual studio can decompile source code

And yeah, packages are all uncompiled so just a simple search will work

[pliant pike]

well that's great thanks a lot

[viral sonnet]

are you talking about authoring or hybrid because those are 2 different things

[errant hawk]

they're both equally shit

[viral sonnet]

lol, i somewhat agree on hybrid but authoring is the best unity ever decided. to split runtime and design data

[errant hawk]

How does adding entities have anything to do with "splitting runtime and design data" that you couldn't do already?

Point is, Entity workflow shouldn't use GameObjects purely because the workflow is completely different

Again, I don't care if they Have to use GameObjects in the background, there just need to be an obvious distinction between what qualifies a GameObject to be a GameObject, and an Entity to be an Entity

You have to keep track of what is going to be converted, and what is not because both normal Mono runtime and Entity conversion use Monoscripts

Having "Authoring" at the end doesn't really make it qualify as an Entity explicitly

[viral sonnet]

it's pretty simple though, GOs in subscenes are converted to entities and serialized as binary subscene data. no GO is left and only lives in the editor

[errant hawk]

Yes, I know how it works

that's off topic

I'm specifically pointing out the conflict of interest between Mono workflow and Entity workflow

I'm honestly tempted on making a completely standalone editor tool that doesn't use a single GameObject purely for entities

[viral sonnet]

there is no pure entity workflow. that's what unity decided on and it makes sense

[errant hawk]

No it doesnt

[viral sonnet]

unless you code your own system, archetypes and loading/saving data. and then you'd end up with something like the authoring workflow

[errant hawk]

I don't see how that is relevant

my entire point is to strip GameObjects out of the equation

[viral sonnet]

but there are no GOs when using subscenes. what's with the arbitrary hate on GOs?

[errant hawk]

ok

I don't hate GameObjects

The entire workflow behind data oriented design is to be completely linear compared to OOP which is GameObjects

the two dont mix well, that's why the Job system is a PITA to work with because you need to squeeze in these little changes to work with OOP and DOD (really depends on what you're doing)

What I personally want along with a lot of other people want is a new/similar workflow that doesn't involve GameObjects

needing to convert your MonoBehaviour script into an entity component type is a bit absurd at times, especially with "simple" components that have multiple aspects to them

Need an array? Too bad, gotta use a dynamic buffer

[viral sonnet]

you mix a lot of topics together. i get the impression you have not fully understood why this workflow system exists. it's not made for the solo dev but big teams with designers who work daily with the comps the programmers made. you say a lot of other people want this but this channel is full of people who came and went with the same prejudice and then realised that their understanding of the workflow was just lacking and they changed their mind. can it be improved upon? sure, that's what 1.0 is about with the new baker feature. but nobody who worked longer with it is really talking about a huge paradigm shift in the workflow or find the need to voice that they don't want any GOs involved. GOs in the workflow are just data containers anyway but for some reason you ignore this.

[rustic rain]

mono workflow is easier

you have inheritance here and all sorts of stuff

easier to mod and extend

if you only had tools to construct entities...

ooof

runtime data of entities can be very awful to work with

so you end with same approach

of having authoring components

[safe lintel]

i think regarding "new/similar workflow that doesn't involve GameObjects" i sort of get it and agree, that in an ideal world maybe something newer for an authoring workflow that didnt have some of the cruft that years of ancient and quasi obsolete monobehaviour functions would be nice.
sadly in reality that would extend the production of dots unreasonably so unity is making do with what they have and its versatile enough for the authoring conversion workflow

[errant hawk]

Unity is considering making a DOTS-only runtime that wont have anything to do with GameObjects, so it's not unreasonable at all

[safe lintel]

they canned tiny and that dots only runtime was once part of the active roadmap, only to be shifted to under consideration, so time considerations apply again.

[errant hawk]

This has nothing to do explicitly with the mono workflow, but the combination for mono and ecs

[rustic rain]

by mono workflow I mean transform hieararchies and managed components (monob)

which is far more simpler to work with rather than runtime entity data

[errant hawk]

yes, but that's not what I'm talking about at all

[errant hawk]

unmanaged data types are also far easier to work with for a lot of reasons. ECS/DOTS/unmanged forces you to actually write clean code, have an understanding of the logic flow, and how to manage data correctly. The only difficult aspect regarding unmanaged types is pointers and managing array-like types (NativeArrays for example.) While it's not "difficult" it's something to be cautious about.

[rustic rain]

I can already see setting up rotation for objects through quaternions 😅

authoring components is a really cool concept

[viral sonnet]

good example. now that would be fun 🙂 the authoring concept of quaternions is angles. easy to understand and work with

[rustic rain]

also colliders, hehe

allthough maybe setting up all that authoring data is possible through pure entities

but I find it quite painful, considering every component change equals = other archetype

[viral sonnet]

there's a plethora of examples. like an ai authoring MB that chains into many ecs comps and/or buffers. maybe I'm missing something here but I'd see no improvement when these have to be setup in the editor with pure entities

miss one comp, the whole thing doesn't work. we had this for a long time with standard MBs too when devs didn't know what they were doing

same goes for any type of string selection in an array that gets converted into an integer index

[twin raven]

I need a second world with the systems in FixedSimulationSystemGroup. Can this be done automatically instead of always remembering to add each new system to the list for the separate world?

[rotund token]

not without using reflection unless you want to include all your systems in the world

[late mural]

how would i burst a for loop, according to the docs it should happen automatically, is this true? Does this also happen in mono behaviours or only systems?

[rustic rain]

you can only burst functions or jobs

[BurstCompile] attribute

[late mural]

https://docs.unity3d.com/Packages/com.unity.burst@1.6/manual/docs/OptimizationGuidelines.html seems to disagree

[rustic rain]

well first of all, that will be burst compiled only if it was called from burst called

then you have

[MethodImpl(MethodImplOptions.NoInlining)]

[late mural]

im afraid im a bit of a noob, so you are gonna have to try explain simply lol

[rustic rain]

Start with this
https://docs.unity3d.com/Packages/com.unity.burst@1.6/manual/docs/QuickStart.html

[late mural]

ah, i shall, thank you

[rustic rain]

https://docs.unity3d.com/Packages/com.unity.burst@1.6/manual/docs/AdvancedUsages.html#function-pointers

here more specific on compiling specific methods with burst

[late mural]

ok thanks!

ok i think i get it, so all for loops within a IJob thingy should burst compile?

[rustic rain]

all code within burst compiled job is bursted

[late mural]

ok fascinating!

having an unsafe IJob is ok right? (im using c style arrays)

[late mural]

oh i see having a managed array inside a job is not allowed, that is unfortunate

[rustic rain]

And this is another part of manual 😅
https://docs.unity3d.com/Packages/com.unity.burst@1.6/manual/docs/CSharpLanguageSupport_Types.html

[late mural]

ah thanks!

[narrow scaffold]

Hello everyone,
happy to have found this discord.
Quick question:
Would you recomend stickinging with HDRP for a new DOTS project at the moment or is URP sufficently supported?

[late mural]

hey, wondering how you debug errors in jobs, as im getting this error, and i dont even know where to start looking lol

[balmy thistle]

What's the callstack?

[late mural]

What's a callstack?

[balmy thistle]

That's kind of like asking how long a man's legs should be, it depends.

[balmy thistle]

Click on that message, it should include more text that describes the sequence of function calls that lead to the issuing of that error message

[late mural]

oh ok sure!

i feel like an idiot now, i only just realised you can scroll in them, and now i know atleast where the error is coming from, thanks so much, ill come back if i cant work out why my spaghetti code is not working!

[balmy thistle]

sweet

[late mural]

aha turns out i was accidentally running it on every physic frame and every regular frame, and also every time the particle system wanted an update

oh also it really doesnt seem to like converting from ParticleSystemNativeArray3 to Vector3

now that im thinking about it, how on earth do i convert ParticleSystemNativeArray3 to Vector3? the more i think about it, the less it seems possible

they appear to be 3 dimensional (perhaps 4?) so im not quite certain how to interact with them in any way, and i cant ignore them either sadly, how unusual and terrifying

[balmy thistle]

by components would be my guess

[robust scaffold]

https://github.com/Unity-Technologies/DOTS-training-samples/blob/dots-training-2022-09-na-group1/DotsVersion/ZombieMaze/Packages/manifest.json
DOTS has landed 1.0 at Unity. Exp.2 so very early stage but it's probably in feature freeze rn.

https://github.com/Unity-Technologies/DOTS-training-samples/blob/dots-training-2022-09-na-group1/DotsVersion/ZombieMaze/Assets/Scripts/Systems/RotationSystem.cs
Unity intends for all stages of a system to be bursted (create, destroy, and update seem to have burst compile tag in the prefab)

https://github.com/Unity-Technologies/DOTS-training-samples/blob/96420b673580e3dad381c06bbeb602dd84cb3abc/DotsVersion/ZombieMaze/Assets/Scripts/Systems/TileCreationSystem.cs#L177
There's now a quick access singleton buffer component function, which is nice.

No appearance of a scheduled / threaded integrated aspect - foreach loop. Only mainthread so I dont think they've managed to make that work. So far.

Foreach loop as in the C# foreach () {}, not the foreach jobs.

[viral sonnet]

great find!

[robust scaffold]

I wake up every day and check the github for new training branches.

[errant hawk]

I imagine we're getting really close to 1.0!

So excited lmao

[narrow scaffold]

I'll rephrase. I'd like to use URP, but I seem to remember that some features of URP are not supported with DOTS and HDRP had more supported features. That might be outdated information though. So is there any reason not to use URP with DOTS?

Another question:
0.51 does not seem to work with the animation package, so right now I am proxying animated/skinned characters as gameobjects to be able to use the non-dots animation system. Does anyone know of a better solution or will I have to wait until animation is implemented?
Even just getting gpu-skinning to work would be great so I can to dots based procedural animations.

[rustic rain]

no animation as of now

you have to use hybrid approach

for animated characters

[viral sonnet]

look on the forum for kinemation or dmotion. it requires compute shaders though https://github.com/Dreaming381/Kinemation-Skinning-Prototype

other than that you are out of luck and need to use GOs

[ocean portal]

Thanks for the links! Do we know if aspects can be used to create entities? 👀

[robust scaffold]

No clue. No example has been done using it in entity archetypes thus far.

[viral sonnet]

fingers crossed we get something on wednesday or next week 🤞

[errant hawk]

Pretty much everything normal GOs can do entities can (eventually)

[unborn totem]

hello all

does anyone here know if com.unity.netcode requires entites 0.51?

[balmy thistle]

Depends on the version

[unborn totem]

their documentation is a little confusing. Most versions say this

    2020.1.2 and later (recommended)

This package uses Unity’s Entity Component System (ECS) as a foundation. As such, you must know how to use ECS to use this package.```

, but then once netcode reaches version 0.51, it starts to say

```This version of Unity NetCode is compatible with the following versions of the Unity Editor:

    2020.2.4f1-dots-5 and later (recommended) Get From Here

This package uses Unity’s Entity Component System (ECS) as a foundation. As such, you must know how to use ECS to use this package.```

i was hoping there would be a list of explicit dependencies in the documentation

[balmy thistle]

All the 0.51 packages were released together

[unborn totem]

ok, so Netcode 0.5, the version just before 0.51 could be for Entities 0.17 then

[balmy thistle]

0.50, and no because all the 0.50 packages were also released together

before then it was the wild west

[unborn totem]

oh 😛

ok then

well I can always manually install it and look at the package-lock file I guess

[balmy thistle]

always an option

[unborn totem]

thanks

[unborn totem]

looks like Netcode 0.6.0 is the last version that works with Entities 0.17 before moving to 0.50 or 0.51

though netcode documentation shows there was a 0.7.0 and 0.8.0, but they don't appear anywhere in the package manager

also the dates in the netcode changelist don't match the dates in the package manager for when they were released

maybe because the preview.# versions aren't documented in the changelog

[elfin spire]

did you guys know that handle dependencies is much slower than schedule then complete? it was a surprise

[whole gyro]

They sometimes have internal versions that were never made publicly available. Same thing happened with the entities package for version 0.18 and 0.19

[rotund token]

a few weeks ago I actually benchmarked our project to show this. while schedule costs aren't insignificant, dependency costs are quite a bit higher

179/frame - Median 1.23ms, Mean 1.24ms, Min 1.02ms, Max 1.41ms

m_DependencyManager->AddDependency
171/frame - Median 0.20ms, Mean 0.24ms, Min 0.17ms, Max 0.48ms

JobHandle.ScheduleBatchedJobs
171/frame - Median 0.62ms, Mean 0.61ms, Min 0.47ms, Max 0.69ms```

[viral sonnet]

unity is aware and some time ago a dev made a forum post that the job system is completely rewritten and (i think) scheduled to arrive in unity 2022 or was it 2023. i can't remember

[elfin spire]

That matches our findings and emphasizes once more the importance of benchmarking everything instead of buying into the Unity best practice

[elfin spire]

Good to hear, so maybe next year when 2023 is rid of all the many regressions we can use job best practice

[raw mica]

@rotund token which version of Unity are those timings from?

[viral sonnet]

how can i read these timings and what could we do about it? so GetDependency is really expensive for some reason

[raw mica]

GetDependency does a lot of work compared to Scheduling a job. It has to collect all JobHandles mapped to each component being read/written to and then it combines the dependencies which is the same thing as scheduling an empty job with all dependencies.

[rotund token]

This is work project so still 2020.3.30f1

[rotund token]

Get dependency is what is called at start of system

AddDependency is called at end of system

Then scheduled batch jobs is called right after it

Which is what actually schedules all the jobs you just created in the system

[rotund token]

Should be in 2022.2

It might actually already be public - I haven't bothered to confirm though due to effort of getting entities to run there

[raw mica]

Yes, the job system improvements came in 2022.2.0b2. From the release notes:

Core: Improved the Job System to better scale as core counts increase and reduced the cost of scheduling jobs and combining dependencies.

[rotund token]

As for what you can do about it, use ISystem

Also use less systems

About all you can do until 1.0

[devout prairie]

Just looking at this code in the unity physics samples:

any clues why they've memcopy'd the indexs but not the verts?

it's actually just in the unit test code

i'm guessing that basically the vert positions aren't important

but obviously the indexs are

[viral sonnet]

seems like a mistake to me. the verts are still used but empty

[rotund token]

depends what it's testing

[viral sonnet]

true, still an odd setup

[rotund token]

agreed

[devout prairie]

i'm guessing it's simply testing the creation of the collider/blob

it basically just creates one of each type

i'm trying to create a collider that matches the camera's frustrum to run a collider cast with

so this seems like a nice setup to quickly build one, i just feed in the points i've calculated for the corners of the frustrum

[nocturne dust]

Quick question: Unity NetCode for dots, does it have platform limitations? aka can it run on WebGL with websockets?

[rotund token]

probably not

not out of box anyway

might need to setup your own transport

https://github.com/Unity-Technologies/multiplayer-community-contributions#transports

these are transports people have written for netcode fo gameobjects (in particular websocket/photon for webgl)

you will probably need something similar for entities netcode (though the transport layer should be somewhat decoupled so hopefully not that hard to port)

[devout prairie]

So uhh.. after getting this frustrum collider setup and running it's clear that the cast only returns hits on the surface of the mesh collider..

Which renders the whole idea pointless 😐

Any ideas on how to return collisions within the volume of the collider?

I should have realised this would be the case tbh, it just never occurred to me

obviously the alternative would be just create a spacial hashmap of all units, rebuild every frame and poll that

but i figured leveraging the physics bvh would be a neat and tidy way to do the same thing, and the data's already there

One workaround i think might work is cast the collider from the end point back towards the camera position

I'm only collecting when the user makes a click-drag selection rectangle so i don't think performance of that would be a huge issue

[viral sonnet]

how can i imagine this? an rts click & drag selection?

[devout prairie]

yeah pretty much

sweeping the frustrum-shaped collider back towards the camera works

[viral sonnet]

huh, i'd just use the collisionWorld and a overlapBox cast. you probably know the starting and endpoint

[devout prairie]

yeah the problem with that is, as the box gets further from the camera, it gets smaller due to perspective..

so if you're drawing a rect in screen space.. objects that are say 100 units from the camera near the edge of that rect might be missed by the box overlap/cast

[viral sonnet]

yeah i see the problem 🙂

[rotund token]

oh that is an interesting problem

[viral sonnet]

hm, maybe you need to rethink that completely. the entity picker approach renders the entity index to a texture. you could do something similar. it turns into a 2d problem then

[rotund token]

i'd probably still do some type of overlap and then just cull individually

[viral sonnet]

optimized you don't even need to render a fullscreen texture but just the part that was selected. for a prototype fullscreen is much easier though

[devout prairie]

well if you think about it.. a box that is worldspace the width of the screen at a distance of say 1 from the camera, will be tiny by the time it's cast out to say 100 units

if you do the opposite ( make it large enough at 100 units so that it covers the width of the screen ) - it would be enormous at camera z

[rotund token]

🤷‍♂️

you can crunch a lot of frustrums pretty quickly

[devout prairie]

there is one disadvantage with this method i'm relying on Camera.WorldToScreenPoint unless i can find the math that goes on behind that

for each hit, i use that to check if the hit is within the screen area of the drawn rect

i guess i'd have to use WorldToScreenPoint and reference the camera whichever method i used to collect the objects within view

[viral sonnet]

how many units are we talking here?

my suggestion has the obvios flaw that bunched up, non visible units wouldnt be rendered. though i think this case could be rare or non existent

depends on the camera angle

because with units <10k i would not overthink this. just brute force it. burst and entities is fast enough

[devout prairie]

that's true also yep

[devout prairie]

Just to point out as it dawned on me, WorldToScreenPoint requires a reference to the camera and isn't burstable, and you're having to check every entity world position against it

[viral sonnet]

yeah that sucks. it's not that easy, what you need is the inverse matrix to cast from world to screen position. some shader tutorials should get you up to speed how to do this

or its just multiplication with the perspective matrix. it's been many years I've done this the last time

i think ive some code around that does this

[devout prairie]

Yeah i mean for what it's worth, creating a custom mesh collider that is the shape of the camera field of view and required depth and then casting that from camPos+(camFwd*depth)/towards camPos and collecting the results works and does the job quickly enough..

I might try and dig out the code behind Camera.WorldToScreenPoint though to avoid needing that managed reference

[devout prairie]

I just looked at this again and got what you actually meant 😛 Yeah i'm going to open the unity core assembly in dnspy and check if the code is visible, barring that i'll use what i've got just now ( which collects all entities/positions from the collider cast inside a bursted job and then iterates the results with Camera.WorldToScreenPoint in a non bursted job.

[devout prairie]

Found a method of WorldToScreenPoint online and converted into burstable code:

[MethodImpl(MethodImplOptions.AggressiveInlining)]
private static float3 WorldToScreenPoint(float3 wp, float4x4 camProjectionMatrix, float4x4 camWorldToCameraMatrix, float camPixelWidth, float camPixelHeight)
{
    // calculate view-projection matrix
    float4x4 mat = math.mul(camProjectionMatrix, camWorldToCameraMatrix);

    // multiply world point by VP matrix
    float4 temp = math.mul(mat, new float4(wp.x, wp.y, wp.z, 1f));

    if (temp.w == 0f)
    {
        // point is exactly on camera focus point, screen point is undefined
        // unity handles this by returning 0,0,0
        return float3.zero;
    }
    else
    {
        // convert x and y from clip space to window coordinates
        temp.x = (temp.x / temp.w + 1f) * .5f * camPixelWidth;
        temp.y = (temp.y / temp.w + 1f) * .5f * camPixelHeight;
        return new float3(temp.x, temp.y, wp.z);
    }
}

[rustic rain]

can swap if with math.select

should be a tiny bit faster and potentially vectorizable

[devout prairie]

ah okay - thanks for that!

so burst doesn't pickup on that and compile it in the same anyway?

[rustic rain]

I have no idea how exactly it works

but

KornFlaks showed comparison

and normal if was different to math.select

which was quite surprising

I mean in assembly code

maybe Burst has it's own conditional code

which compiles certain functions with special constraints

[devout prairie]

yeah that's interesting

i assume it depends on the nested code, how burst handles it and if it gets vectorized

ie if it's vectorizable, it'll get the benefit of math.select

if not, it's maybe marginal

so it's probably better to use it everywhere possible than not

[rotund token]

i think you'd probably better be off just doing if (Hint.Unlikely(temp.w == 0f)) unless you're actually getting proper simd on this

since it's such a rare case

math select isn't always faster, it requires execution of both paths for the sake of removing branching

and in this case using math.select will cause nans in one of the paths

does that matter? hmm maybe not

[devout prairie]

interesting, thanks

[snow iron]

just raising awareness of this tweetstory that I posted last week in case it's helpful: https://twitter.com/MrAndyPuppy/status/1568716388499619840

Andrew Parsons (@MrAndyPuppy)

I know people are increasingly asking about DOTS this year, so here is my current top ten DOTS links:

1. DOTS Beta Forums - Dev Status and Milestones - https://t.co/1rhV0ZglXf

Likes

113

Twitter · Sep 10, 2022, 9:41 PM

[karmic basin]

Great ! We have some in the pinned messages, but always nice to refresh with more recent resources 🙂

[safe lintel]

pinned message here should probably be updated 🙂

[devout prairie]

@solar spire

[silk nimbus]

Hello, I am facing two issues in my project which uses DOTS. Can anyone please help me with this?

1. I am controlling a sphere object through below movement script. It works fine but when I reload the scene, this is no longer working. I am not able to control the sphere.

2. I am increasing the composite scale of a plane at runtime but mesh collider is not increasing automatically with it. So objects which are on the plane are falling through the plane.

[viral sonnet]

are you reloading the domain? if you don't, statics will have old values

[safe lintel]

for 2. physics only uses the translation and rotation for non static rigidbodies(and i dont think statics use more than ltw?)

you will have to edit the collider blob data

[rotund token]

class System : ISystem
i need unity to add a static analyzer for this to save me from myself

[balmy thistle]

Great idea

[dense crypt]

How can I get ComponentData if all I know is the TypeIndex?

[rotund token]

so you don't know what component it is?

you can get ComponentType using ComponentType.FromTypeIndex()

from there you can get dynamic handles

[dense crypt]

I have an attribute I mark on my Components, so I know what component it is, but I can only find <T> ways of getting ComponentData

Closest I found on DynamicHandles is GetDynamicComponentDataArrayReinterpret<T>

which still requires a Type, I suppose I could maybe just pass IComponentData?

What I'm trying to do atm is to just dump the data of all components marked with my attribute, if I have to crawl raw pointers that's fine too. But I think I could probably get the info from reflection as long as I can get the structs somehow

[rotund token]

you can just pass it byte

and use info from typemanager about elementsize etc

to do with it as you will

var components = chunk.GetDynamicComponentDataArrayReinterpret<byte>(this.ComponentType, this.ElementSize);

[dense crypt]

Hmm interesting, I didn't think that was a thing

Is there then a way to reinterp a byte array to the actual component given its System.Type? Assuming I've already split my array up in to the byte representation of the component

[rotund token]

not really

[dense crypt]

I suppose I could use a binary formatter or similar, but was hoping there's a native solution

[rotund token]

c# is for the most part statically typed however you could maybe use 'dynamic' to get this to work. maybe...

{
    public void Execute(in ArchetypeChunk chunk, int unfilteredChunkIndex, bool useEnabledMask, in v128 chunkEnabledMask)```

i find it strange Unity has updated IJobChunk for the new enable features

when IJobChunk was meant to be deprecated for IJobEntityBatch

        while (chunkEntityEnumerator.NextEntityIndex(out var i))
        {
            var tmComp = transforms[i];```

i am somewhat concerned by this api though

though I assume if I am not use enable masks I can just iterate the array normally

for the sake of simd

[rustic rain]

but what exactly are you trying to achieve here?

what do you reinterpret it as?

and how do you plan to use it?

[dense crypt]

The basic idea is to supply a UI system with any Component's value from DOTS. So in essence someone could say MyUI.Width = "Namespace.HealthComponent.Value". Note that our UI isn't made in Unity, so we'll resolve the string to the actual Type. Then my idea was to use this Type to query for data in ECS Worlds, then use reflection to grab the property (Value) and pass it back to the UI.

Still early prototyping so looking if there's a nice way to get entity data from a System.Type

[rustic rain]

I had idea about it

my solution was to create personal type manager that solves componentdata into string

to do it untyped

it'll require to implement either extended type with method that does conversion

or conversion from byte array

let's say component velocity

and you want to show on UI it's value as

"Velocity: {value}"

you reinterpret byte array as value type

viola

and it can be made native

but it'll be very low level copy of Unity ISystem implementation xD

you'll have to store managed and unmanaged function pointers basically

somewhere

which will be accessed by your converter's type index

[tropic venture]

just tested my subscenes / dots / ecs scene for the first time on android, skybox renders nicely, other than that - nothing is showing. everything seemed to show fine in editor player game mode, but nothing showing on the actual mobile device. gonna start trying to investigate.. meanwhile if there are any obvious caveats / should-knows / usual suspects etc. please advise

[rustic rain]

why not start by attaching logcat?

[tropic venture]

is that in the profiler?

[rustic rain]

nah, it's package

or you do it manually through adb

should just look up Unity tutorial on it

[tropic venture]

hope this is the right one https://docs.unity3d.com/Packages/com.unity.mobile.android-logcat@0.1/manual/index.html

i vaguely remember previous times Unity came shipped with logcat... maybe when i was playing with unity 2022.*

ok i got profiler and logcat, anything specific i'm looking for?

[rustic rain]

errors

what if it's screaming errors on mobile

xD

[tropic venture]

hm what if 🤔

i thought i already made the shaders SRP-batcher compatible....

[misty wedge]

Is it somehow possible to schedule a deferred parallel job based on the length of a specific dynamic buffer?

[dense crypt]

If it's a IJobParallelFor you can specify the batch count, if that's what you mean

[misty wedge]

But I can't access the length of the buffer without calling Complete on the systems modifying it

I don't think it's safe to hold a copy to a dynamic buffer while it's being change though and then scheduling a deferred job on it though since the memory may be moved by ECS

[tropic venture]

why would the URP/* shaders have compatibility issues with hybrid renderer?

[tropic venture]

ok so Vulkan works...

but i was hoping it will work on OpenGL ES 3 maybe even 2...

does using dots/ecs require higher-end mobile devices? are signifigant swaths of older devices like Android 6.0 etc. going to not be supported?

[misty wedge]

OpenGL and GLES are not supported in the current version of hybrid renderer, but it is planned for the future

It only supports Vulkan on android and linux atm

[misty wedge]

DOTS / ECS does not have any specific hardware requirements itself afaik

[tropic venture]

ok so how do i troubleshoot why standard shipped shaders like URP/Lit or URP/SimpleLit don't work with OpenGL ES in dots/ecs

[misty wedge]

Are you using hybrid renderer?

[tropic venture]

yes

[misty wedge]

^

[tropic venture]

wha 😱

is this documented somewhere?

[misty wedge]

Of course

[tropic venture]

please do share link

[misty wedge]

https://docs.unity3d.com/Packages/com.unity.rendering.hybrid@0.51/manual/requirements-and-compatibility.html

[tropic venture]

awww maannnnn

is it planned for ECS 1.0 ?

[misty wedge]

I'm not sure if they've announced anything, I don't follow ECS development nearly as closely as some of the others here, they'd probably know more

I don't use hybrid renderer for my own project

[tropic venture]

this does imply some degree of hardware requirements kinda sorta

[misty wedge]

Well yes, for the hybrid renderer, but not for DOTS / ECS as a whole

They are entirely separate

[tropic venture]

right

[tropic venture]

so you probably use dots/ecs mostly for speeding up logic processing, like moving lots of code which would have been in Update() methods into ECS systems. But for my prototype scene where i'm mostly just testing pushing graphic limits of how much static environment i can render (trees, vegetation, grass, rocks, props etc) on mobile, no significant loads of logics to process ATM, i would specifically benefit most from the hybrid renderer and NOT just plain putting all the game objects into ECS world?

so if i specifically want openGL support, then i have nothing to gain from ecs/dots at all for my particular use case?

[misty wedge]

If you have no significant gameplay code your game will (probably) not run faster by moving stuff over to ECS

[tropic venture]

..unless i benefit from the hybrid renderer?

[misty wedge]

Yes, hybrid renderer should run faster than using game objects for rendering

[tropic venture]

putting it a bit differently; the initial flashy marketing about dots/ecs was that MegaCity demo, millions of of static environment objects rendering blazing fast. am i wrong to desire to achieve a similar feat by using dots/ecs for a dense scene of lots of static environment?

if someone want to achieve dots/ecs awesomeness like megacity nowadays, is the answer basically = hybrid renderer?

[misty wedge]

I mean it's certainly possible, as they have shown. In the current iteration you would have to resort to vulkan. I'm not sure how well the megacity demo ran on mobile devices

[tropic venture]

technically the did show it off on mobile as well

[misty wedge]

Then you have your answer

[tropic venture]

so back in 2018 they had some sort of hybrid renderer which also might have only worked with vulkan?

[misty wedge]

I'm guessing it's still not easy and requires "manual" optimization

[misty wedge]

I'd guess so

[tropic venture]

i see

[misty wedge]

There are limits you will run into regardless, especially on mobile devices

[balmy thistle]

Megacity ran on mobile

[misty wedge]

hybrid renderer isn't magic and will let you render infinite objects at no cost

[tropic venture]

funny you should mention that,

[tropic venture]

i was asking about "manual" optimizing in dots/ecs here https://forum.unity.com/threads/how-to-implement-megacity-lods-and-culling-system.1334756/ they were all saying - no need, it's all automagic, just put your scene in ECS land (subscenes or whatever) and you'll enjoy all the benefits & optimizations similar to / better than megacity

[misty wedge]

I'm not sure what kind of optimizations they did for megacity besides not using game objects, I was referring more to things like optimizing shaders for mobile, which is something you would still need to do. Again, hybrid renderer itself isn't magic, eventually the GPU will have to handle draw calls

(hopefully significantly less than if a different approach was used)

[tropic venture]

ah right well ye shaders prolly still need to be optimized as needed, i guess i was referring more to like scene optimization, LODs, culling, scene graph etc. concepts like what was discussed in that ECS talk https://www.youtube.com/watch?v=k_ORJXmPu9M&ab_channel=Unity

and apparently that stuff is supposed to be automagic in hybrid renderer?..

anyway i guess for now the main lesson is; dots/ecs 0.51 + hybrid renderer = only vulkan supported

[misty wedge]

I can't really comment as I've never used hybrid renderer in a significant capacity myself. If I were you I'd look at the unity samples and see how they did things

[tropic venture]

ok

[misty wedge]

Hybrid renderer uses the BatchRendererGroup api to create draw calls, and that only supports Vulkan on android / linux

[tropic venture]

i see

[tropic venture]

what's the plan for openGL ES then, they wont use BatchRendererGroup for that?

[misty wedge]

I'm guessing they will add support for it in the future, at least the hybrid renderer docs state as much

[tropic venture]

hopefully nearer future than later... maybe they'll just decide meh forget OpenGL/ES... only future devices supported kthxbai...

[misty wedge]

I'm happy I have very low rendering requirements so I can just use normal URP rendering 😅

[rustic rain]

why is it even a problem?

[tropic venture]

lemme know when you're making a game trying to look anything vaguely close to this 😄

[misty wedge]

You'd probably be able to do something like this in "normal" unity with lots of performance tricks

even on mobile

[tropic venture]

i'm trying/hoping to support a few levels of older devices too... if at all possible. ex. i'm testing with an LG G4 having Android 6, where vulkan support is 1.0 and famously android devices have flaky early vulkan support and will never be corrected (as the hybrid renderer docs say)

[rustic rain]

how old do you want ,kek

Vulkan is supported on almost anything past 2016, no?

[dull scroll]

Hello where can I learn about best practices to reload a scene with DOTS or load another scene? Whatever I tried I get problems.

[tropic venture]

i guess i'd be satisfied with this LG G4

[tropic venture]

but flaky?...

TBH i dont have that LG G4 with me right now so can't check, will do so in a few hrs

[tropic venture]

i guess i thought the easiest "performance tricks" would be to throw it into ecs/dots/hybrid renderer, and why not - if there's a motherload of performance awesome to not lose out on right there. performance by default. then you can even optimize further with even more optimization tricks...

having said that - i'd love to achieve great performance in "normal" unity even on mobile... what tricks do you have in mind? frustum culling? occlusion culling? LODs? anything else?

[misty wedge]

I mean it depends a lot on the scene, but I'd start with draw calls first

[tropic venture]

hey if i could interest you to join in this little prototype project - it would be awesome to learn good tricks from a seasoned veteran 🤩

[misty wedge]

Make sure everything that can be is instanced

[misty wedge]

I have very little graphics programming experience and no time sadly 😅

[tropic venture]

instancing itself requires opengl es 3.1 i think, no? not even sure if that's available on this LG G4.... maybe tho

[misty wedge]

There's a reason why my project has low rendering requirements

[tropic venture]

heh

2D? 😛

[misty wedge]

of course

[tropic venture]

hah of course

[misty wedge]

There's lots of stuff that developers have used over the years, replacing distant trees with billboards, instancing, pooling, etc

[tropic venture]

right

[misty wedge]

Depends a lot on what's bottlenecking you

[tropic venture]

last i profiled it in regular game-objects on mobile i think the main load was Draw Opaque and Draw Transparent calls

[misty wedge]

How many draw calls did you have?

Also why transparent calls? The scene above doesn't look like it has anything transparent except maybe the water

[tropic venture]

tree leaves use transparency cut-outs

alpha clipping

otherwise this

[misty wedge]

But why not make it opaque and use alpha clipping

The actual leaves aren't transparent

[tropic venture]

right

i thought that's what Draw Transparent calls were about, the alpha clipping

so you could see other graphics through the transparency of the leaves texture

i might be horribly wrong about that tho

i also thought "but wait it's a PNG with transparency - it should automagically already be alpha cut out!" but no apparently a shader needs to specifically perform that

which i'm guessing is what causes some Draw Transparent calls or whatever

[misty wedge]

What surface type are you rendering it as? Transparent or opaque?

[tropic venture]

opaque

i figured that one out a little while ago

not sure if it matters but i'm also seeting render face = both

[misty wedge]

I'd check the frame debugger and see how exactly the frame is rendered

[misty wedge]

No, the frame debugger

[tropic venture]

ah

[misty wedge]

Also 2k draw calls is a lot for such a small amount of geometry

especially on mobile

[tropic venture]

i think i didnt turn on instancing

but again - not sure instancing is available for the LG G4 i'm hoping to target

[misty wedge]

I'd check this part of the docs, it explains a lot of basic performance optimization techniques as they pertain to Unity

https://docs.unity3d.com/Manual/graphics-performance-profiling.html

[dull scroll]

can anyone helpme reloading a scene in a system?

I want to implement a simple method to reload the scene once the game is over

               var sceneSystem = World.GetExistingSystem<SceneSystem>();
                var guid = sceneSystem.GetSceneGUID("Assets/Scenes/GameScene.unity");
                var sceneEntity = sceneSystem.LoadSceneAsync(guid);```

using this

how can I load the scene as single?

[misty wedge]

What do you mean?

[dull scroll]

I mean I just want to reset the world and reload the scene

[misty wedge]

Different ways to do it, ideally you'd just unload the old scene

If you want a "true" reset (including system state), you can dispose the world and create a new one

[dull scroll]

from a monobehaviour?

[misty wedge]

From wherever you like

(as long as it's on the main thread)

[tropic venture]

what would you like to see?

[dull scroll]

I see

[misty wedge]

I'd check where your 2k draw calls are coming from

[tropic venture]

i'd have to test on the LG G4 in a few hours, the LG V60 ThinQ i am using now definitely has proper GPU and instancing and good performance, 322 draw calls, 24 batches

168.9 k tris

hm now i see indeed most vegetation even those transparent leaves are part of Draw Opaque. only my "water" plane is Draw Transparent

[tropic venture]

IIRC there were about 2K objects in the scene when i screenshotted that

very cool tool this frame debugger 👍

[viral sonnet]

are you using subscenes? that way you can control it much easier as you'll have the entities coupled in it

handling entities just with scenes needs quite a lot of manual work to get right

[dull scroll]

I will, I am learning about them now

[viral sonnet]

a subscene is a special component in a scene that automatically converts your gameobjects to entities.

[misty wedge]

Any sense in pooling entities for performance gains?

I guess it would depend on the amount and how long it takes to iterate over the "disabled" ones, maybe easier with the new enable/disable functionality in 1.0...

[viral sonnet]

depends in what capacity they change and how long they live

[misty wedge]

a few seconds maybe

[viral sonnet]

if you can batch create the entities i found pooling not necessary.

[misty wedge]

Basically I have a lot of data that belongs to map chunks, and currently I am storing that data in native collections. That makes it a lot more difficult to schedule jobs per map chunk though, which is trivial if you have an entity for it

[viral sonnet]

where's the difference in scheduling?

[misty wedge]

Each map chunk belongs to a separate "world" (I talked about this recently, not sure if you remember), not an ECS world, but a place where an Entity can be. As each world needs to store data separately, this already makes it kind of awkward, as I need to use a Dictionary<Entity, NativeCollection> or resort to unsafe collections, which I'd rather not. Then I need to schedule each "world"s update separately since I cannot pass that dictionary to a job, and there is no entity for a map chunk to grab data from using something like a CDFE

It's hard to explain, but I feel like by trying not to use entities for this, it kind of removes this part from the normal (simple) entity dependency management, and requires a lot of manual scheduling and passing job handles between systems. I feel like I could make it a lot simpler if I moved more data to entities

For example, I'm currently playing around with using the manual navmesh API instead of my own written version for pathfinding locally. My idea was to use a job that gathers all collidable primitive shapes (NavMeshBuildSource) for a map chunk, and then in the next frame the previous dependency handle is completed, and the static NavMesh and NavMeshBuilder APIs are called (since I'm guessing these are main thread only).

The question is what kind of data structure to use to store the primitive shapes. If each map chunk is an entity, that would be easy, a DynamicBuffer. But since it isn't, I'd need to use something like Dictionary<Entity, NativeMultiHashMap<int2, NavMeshBuildSource>>, which seems very akward...

(in the entity version map chunks are mapped to worlds using a sharedcomponentdata that contains a single Entity field)

[karmic basin]

On the topic of navmesh, are you aware they have jobified the API now ?

Anyone got good info on joint motors support ? Just had a quick look at the physics forum. It is... depressing 😞

[misty wedge]

By jobified do you mean the navmeshquery?

[tropic venture]

testing now

😭

[misty wedge]

Is there a way to do the opposite of a raycast in the dots physics API? basically stop once the ray doesn't hit anything?

[rotund token]

!

[safe lintel]

@karmic basin dont think they exist yet

at this rate im not sure if they ever will 🥲

[rustic rain]

you can test on virtual android

on PC

[tropic venture]

but i was hoping to target medium-low end devices such as this LG G4

if it doesnt support vulkan, then i might try pulling off this project in regular game objects after all...

at least until they support opengl maybe

[rotund token]

they do specifically warn you about this in the manual

Currently, Hybrid Renderer does not support desktop OpenGL or GLES. For Android and Linux, you should use Vulkan. However, be aware that the Vulkan drivers on many older Android devices are in a bad shape and will never be upgraded. This limits the platform coverage Hybrid Renderer can currently offer on Android devices. OpenGL and GLES3.1 support are planned for a future version of Hybrid Renderer.

[tropic venture]

wish it was a bit more louder & clearer on the tin before getting too much into the weeds...

it's ok tho - i'll still work on a dots/hybrid renderer version for higher end devices

bit clunky to support multiple versions like that... but i'll try and see how it goes

[karmic basin]

Honestly, I would expect anybody to have read the manual of the packages at least once before writing the first line of code 😉

Always study feasibility before jumping right into it with your two feet 😄

[rotund token]

haha reading manuals

[karmic basin]

I don't have hope neither :/ How would you approach it ? Apply linear/torque/velocity impulse, managed by a p(i)d controller ?

[rotund token]

start writing code, then complain about a bunch of design decisions

this is the way

[karmic basin]

🤓 haha yeah where's the fun in books !

[safe lintel]

@tropic venture you did have your early warning 🙂 #archived-dots message

anyway Ive never used a joint motor in physx, I just want other features they talked about ages ago. they sound kinda important 😦

[tropic venture]

wow impressive necro 😄 i probably didnt fully understand what i was talking about then, was just getting started, but yes apparently you're right i was told about it...

[safe lintel]

only a month ago. just wait until you've been waiting for features and fixes for years 😅

[tropic venture]

still hardly know what's what... dots, ecs, hybrid renderer, subscenes or no subscenes, convert to ecs or dont convert, manual optimizations or automagic optimizations... fun times

[rotund token]

or you had a feature and it was ripped away from you 😄

[safe lintel]

animation 😭

[karmic basin]

alright, thanks for feedback 👌

[karmic basin]

Yeah that one hit hard

[viral sonnet]

UIToolkit has so many controls and just a fraction is available in the UI builder. so weird

i only realized with intellisense that there's ui:Image and i don't need to use ui:IMGUIContainer

which stands not for image UI container ... lol

[rotund token]

lol

generally you can just use visual element + background image

unless you need some very specific image control

you'll also probably end up just writing a bunch of your own elements

[viral sonnet]

that's what these controls do anyway. write out a background-image

but ui:Image at least has a sprite property where I can set it directly

haven't found the binding path. seems missing 🤷

who needs images in games. such a niche thing

not in a world with no animations

[rotund token]

BindingPath is part of BindableElement

[viral sonnet]

yeah i saw that. image is just derived from visualElement. i'm just looking for an image something that can bind a sprite

UIToolkit feels very code centric. I don't like querying all these elements. Such a step back from UnityUI. (and yeah, i know, great rendering and fast) UnityUI could also been fast without this shitty workflow

[rotund token]

UnityUI
i'm not sure that's true

[viral sonnet]

probably not for the current version but the workflow is excellent imo

[misty wedge]

The idea is to go from something nobody knows (unity UI) to something that actually kind of works in a team (html and css)

Also it was originally editor only, which means code only

[rotund token]

layout groups are a huge problem in ugui

they may be friendly to setup

but they destroy performance

the best practice is to use layout groups to setup all your stuff

then delete them all so they never run

it's pretty horrific

[viral sonnet]

maybe i'm just pessimistic but we have very little numbers for UIToolkit - at least i don't know any - so who's to tell that they solve all complex UI stuff.

if it was just editor stuff. no editor GUI is utterly complex in layout

[misty wedge]

They just released a large ui toolkit specific sample

I haven't looked at it though

[rotund token]

it's actually on my todo list for my day off today

just curious how it compares to what i've been doing

[viral sonnet]

same, i'll take a look at it

i mean, not that we can really compare as there's no unityUI sample of the same thing

[rotund token]

outside of assigning others to do UI work, I haven't used unity UI in 2 years

so i don't really care about a comparison

i just want to see if I've overlook some useful functionality

[viral sonnet]

when the major downside of UnityUI is performance it's the most important metric

[rotund token]

well personally

i think ui toolkit workflow is much better

[misty wedge]

UnityUI performance is terrible

[viral sonnet]

i don't see it yet. unityUI has great WYSIWYG. binding alone pisses me off and can lead to weird errors in teams. it's very rigid.

[misty wedge]

You mean binding as in callbacks on classes / IDs?

[viral sonnet]

yeah i've read that a bunch of times by now and i can never figure out what kind of UIs one need to build to have terrible performance. i'm looking at all those UnityUI centric games and they barely hit any cpu.

[rotund token]

you know you don't have to use bindings right

[rotund token]

we have 25-40ms spikes on consoles from ugui

and up to 10ms on 3xxx ryzen

[misty wedge]

Someone I know worked on a large scale project and they optimized forever to get the UI not to spike FPS

[viral sonnet]

what's happening in those spikes?

[rotund token]

you update 1 element

every canvas that touches that in any z direction

needs to be rebuilt

which triggers every layout group

basically unless you're very very careful

any update to any element ends up triggering a full UI rebuild

you have to follow very specific and pretty gross practices

to avoid this

https://forum.unity.com/threads/unity-ui-performance-tips-sharing-my-findings.524916/

this guy summarized it pretty well

A lot of the optimization tips tells you to not do what the tutorials tell you to do (e.g. avoid Layouts etc).

[viral sonnet]

how much UI panels are we talking about here that this gets into such a spike?

[rotund token]

the TLDR

• don't use any layout groups, manually layout everything
• have a completely flat hierarchy, no nesting
• don't stack any UI

this matches our experience

[viral sonnet]

so the main problem seems to be the overly aggressive updates of layouts that aren't even needed

[rotund token]

if ugui is your jam, it's unlikely to be removed anytime soon

hell we still have OnGUI support

[viral sonnet]

i don't give a fk about unityUI 😄 i'll take the next best thing

uiToolkit sample in profiler

hits the 144fps cap

looks pretty cool, very smooth

[viral sonnet]

you should've heard what i said about unityUI when i first started it. now it's just something i got used to and all it's quirks. don't think for a second that i like the vertical/horizontal layout groups or layout element that only support percentages in a kind of way. but to its credit. the implementation and interface for it is nice. i like the stretch and anchoring stuff. i can't even figure out what half of those icons in UI builder are. i have to look at the code to understand most of them.

oh boy

still off with match game view.

they certainly use VisualElement for images and set gearElement.style.backgroundImage = new StyleBackground(m_EmptyGearSlotSprite);

[rustic rain]

I use static resolution in builder

and special setting

that scales texts

makes workflow pretty fluent

[viral sonnet]

match game view sets the resolution to the game view. what's your special setting?

[rustic rain]

ah, I don't remember

but you choose asset for those

in UI Document

component

match game view is awful, because in Editor resolution can be very small

and there's no autosize for text, without such setting

[viral sonnet]

well that explains a bit. is there no way to get that in ui builder?

[rustic rain]

in builder

you simply set that "default" resolution

you picked in ui asset

and don't change it

everything else will be scaled

in runtime

[viral sonnet]

btw @rotund token they bunch all screens in the MainMenu.uxml. haven't found a dynamic load yet of a panel

[rustic rain]

which you can test right in editor

[rotund token]

doesn't surprise me

[rotund token]

yeah i only use visual element for images

[viral sonnet]

didn't know about StyleBackground. seems easy enough

[rotund token]

well redid my whole asset management, thanks eizenhorn...

[safe lintel]

any particular reason why? also was this something eizenhorn posted on the forums about?

[rotund token]

just a random discussion that was on in here a couple of days ago

IDs has been something that has bothered me for ages

i like the general concept he was using and it inspired me to properly solve this

i even solved the issue of 2 people creating new objects and being assigned same ID in different branches

[safe lintel]

can you link me part of the convo if it was here? getting a case of fomo 🙂

[rotund token]

#archived-dots message

here onwards

[safe lintel]

ty

[rotund token]

(you can see the previous convo that he replied to)

[safe lintel]

my asset management is so haphazard, one scheme here, another scheme there, then just folders where I dump things willy nilly

[rotund token]

yeah i unified everything

everything uses the same definition

ties 1:1 to an entity

data driven categories that a design can create

and would be usable in AI / etc

as well as guaranteed unique ID even with branch merges

[safe lintel]

on first glance maybe I dont need this complexity

[rustic rain]

Is there a way to check whether current execution is in burst?

[rotund token]

yes

ecb does that

[rustic rain]

Hmm

I just got a fun idea

Burstable delegates

[viral sonnet]

hm, i use the guid from the AssetDatabase for SOs

[rotund token]

that was basically the system i replaced

[rotund token]

        private static bool IsBursted()
        {
            return false;
        }

        public static void Method()
        {
            var isBursted = true;
            isBursted = IsBursted();

            if (isBursted)
            {
            }
            else
            {
            }
        }```

something like this

[rustic rain]

Which will be generic struct with pointers to managed and unmanaged function (actually just index that points to shared static array of pointers)

The idea behind it - create a way to queue callbacks

[rotund token]

i was using a WeakAssetReference which was just the GUID of the unity object

but my new implementation has quite a few nice benefits

• 4x smaller component size
• i can replace all references in project by changing the prefab in the definition instead of having to find all references in project and replace them
• i can convert my database of assets from a hashmap to an array

[safe lintel]

hmm that second part sounds really useful

[rustic rain]

Hmmm

That gave me an idea

For entity reference

What if you create special native container

That stores pointer to entity struct

Similar to blob

On the other hand nvm

Approach doesn't help with reference preservation after all

[viral sonnet]

did i get this right. you now have a SO database which relates guid to integer id?

[rotund token]

kind of?

my definitions are just basic SO

(there is also a hidden unique ID field here)

that reference an entity prefab

authoring references these SO and store the ID on a component (int ID is wrapped in a struct so I could change type in future if I need though will probably not happen)
then can just get the prefab from a buffer to instantiate

i use asset importer to guarantee unique IDs

so if 2 different people merge 2 new definitions with the same ID

on the import it'll detect this and update one of the IDs to a new unique one

but existing references won't break because they reference the SO

the only case I can see this breaking is if someone branches from master right before you publicly release, then merge back in after you release

and updates a released ID

(because it might break save data)

[viral sonnet]

isn't that kind of impossible with guids?

[rotund token]

im not using GUID though

[viral sonnet]

why not, every SO has one

[rotund token]

because i dont want to store GUID on components

and i can't just do
Instantiate(NativeArray[index])

[viral sonnet]

i think eizenhorn does it like this. at least the 2nd screenshot he posted looked like this.

guid#name

[rotund token]

he specifically has sequential IDs

and they're solution to merge issues is just, push the new ID straight away quickly

#archived-dots message

[viral sonnet]

yeah, with that kind of referencing you can generate a sequential unique id without worries

[rotund token]

which has worked fine for them

and honestly, solid rules in a team is usually good enough

i just thought of an easy way I can guarantee this

[viral sonnet]

the only more safe way is to have some kind of database in between

[rotund token]

i don't need that database

that's the point

i can guarantee these IDs without losing references by fixing them on import if there's a merge conflict

and the code is quite simple

another situation this solves is duplicating a definition

which is usually how a designer creates these things

traditionally we'd use OnValidate() for this type of thing

but it's pretty problematic

because it runs on every SO when you build addressables in editor

another nice perk is I can just create the list of all prefabs in a single buffer during authoring without having to actually create a list

[viral sonnet]

some really nice benefits

[rotund token]

so a designer doesn't have to remember to add their new definition to some arbitrary list on a SO

[viral sonnet]

yeah that sucks so hard

[rotund token]

i wasn't actually sure this would work

but it turns out it just does

tldr: var guids = AssetDatabase.FindAssets($"t:{nameof(ObjectDefinition)}");

i just find them all using AssetDatabase

during conversion

so the process for a designer to make a new definition

is to hit Ctrl+D on an existing one

and rename it

bam done

[viral sonnet]

i'm doing something similar. that was the reason i didn't want any GO for the CreateAdditonalEntity if you remember. i even wanted to skip the part to setup a ObjectDefinitionRegistryAuthoring (what you call) in a subscene 😄

[rotund token]

i'm not really sure why you'd need CreateAdditonalEntity

for what i'm doing

[viral sonnet]

you pack it all into one. i seperated them on types

[rotund token]

i don't really get it

why do you need new entities though

[viral sonnet]

oh wait, that was for the SOs to blob converter. slightly different but samey 😄

[rotund token]

the first experimental version coming Pretty Soon Now (tm)
😉

[viral sonnet]

WHO SAYS THAT?

[rotund token]

https://forum.unity.com/threads/subscenes-dlc.1339274/

[viral sonnet]

noice

[rotund token]

https://forum.unity.com/threads/dots-development-status-and-next-milestones-june-2022.1291766/page-2#post-8459483

[viral sonnet]

after that long post, i still don't see much changes

[honest plinth]

https://forum.unity.com/threads/new-batchrenderergroup-api-for-2022-1.1230669/page-2#post-8122013
We have landed GLES 3.0 support for BatchRendererGroup, and newer versions of Hybrid Renderer will support GLES 3.1 (due to some compute requirements)
Don't expect miraculous perf on GLES phones though 😉

[devout prairie]

Hmm..

I dumped this code into a job:

what am i missing here

[rotund token]

are you using other temp memory

inside that loop

[devout prairie]

yes in the job but not in the loop

[rotund token]

hithovered entities

nativehashset

temp memory

?

[devout prairie]

ahh

ok so what's the issue with that?

[rotund token]

you're running into this issue: https://docs.unity3d.com/Packages/com.unity.collections@1.4/manual/issues.html

get rid of your foreach basically

the 'AsArray' is getting invalidated by the other temp allocator

[devout prairie]

ah shitsticks

ok thanks

so just use an index to iterate should be ok??

[rotund token]

yeah

[devout prairie]

yeah that works, thanks again

it makes some kind of sense i guess, and is no big hassle

be nice if they fix it tho

[rotund token]

it's a huge problem imo

like, if I (or someone else with a lot of experience) isn't around to tell you what's up

it could take ages to debug this

this is the most basic case as well

[devout prairie]

it could indeed

yap

[rotund token]

there are much more complex cases where it pops up that are much less clear

i'm hoping this is being addressed with new allocation stuff but i'm not sure

it's been marked as a known issue for 2+ years

[devout prairie]

ahhh

juggling safety handles i guess internally, without causing problems

i'd like them to fix the using 'this' error

if you use a non-local inside a job

it's really unspecific, and would be a lot quicker to debug if the warning would just mention the name of the offending variable

[rotund token]

oh this. flagging annoys me

on the WithStoreEntityQueryInField

it should not flag that as an error

but i'm at the point where I only ever use EFE if i'm quickly throwing something together to test

so whatever s🤷‍♂️

[devout prairie]

it's always the stupid stuff when you're trying to rapidly test 😛

i had easily half a dozen foreachs in that job, and same amount of temp lists 😐

my previous selection system was thrown together using entitymanager for quickness/prototyping

inevitably it gets fairly complex so i wanted to start flexible

but it was as slow as hell and caused lag

now it doesn't impact fps at all

so the frustrum collider idea works well,and shoving it all into jobs and ecb goes without saying, magic

[devout prairie]

hey @rotund token was just reading an old thread - you found that moving entities offscreen was quicker than say disable/enable?

i couldn't gather whether you'd used worldrenderbounds to trick it off screen, or simply moving the translation

[rotund token]

ive said a lot of things

and i dont recall this at all

i'll need a reference to myself 😄

[devout prairie]

haha 2019 i think

https://forum.unity.com/threads/how-to-enable-disable-renderer-of-entity.684025/#post-4590745

[rotund token]

oh this would have been hybrid renderer v1

[devout prairie]

ahh yeah

[rotund token]

oh wait

this was completely different

this was for like invisible units you wanted to exist but not render

but there's a component for that now

DisableRendering

[devout prairie]

currently adding and removing Disabled to my selection indicators and it's not really ideal, bit hacky i think

ah oh DisableRendering

that's probably be better

[rotund token]

i was actually going to pull this library out soon

and update it

been 3 years

and i'm finally at a point where i want to use it

[devout prairie]

when adding and removing Disabled i'm noticing this 1 frame glitch where the objects render out of place for some reason

[rotund token]

i'm going to hate the code

[devout prairie]

hehe no doubt

[rotund token]

DisableRendering shouldn't have this issue

as i believe the components should still update

[devout prairie]

Still getting a slight issue with that, but it's less obvious now..
https://gyazo.com/9c34f50bde836565de3ee0ff9eeebc0d

Can't really see any issue with the code so wondering if it's when and where i'm doing this

[rotund token]

what's the issue for unobservant people

[devout prairie]

when i move the mouse, so the rect selects/deselects units at the edges, you can see the green cubes pop in/out out of place

kinda hard to spot

but it's there

[rotund token]

isn't that the intended behaviour 🤔

what buffer you using?

[devout prairie]

end init

and during it during init

[devout prairie]

you can see sortof like 'fireflies' where the green cubes appear out of place

[rotund token]

i know your stuff isn't really moving atm

but aren't you basically using last frame position then

probably doesn't matter

(i find peoples usages of init system group interesting)

(people use it for all sorts of different things)

[devout prairie]

yeah i just dumped the system there and it works so haven't changed it if i'm honest

just tried simulation group but using the begin init buffer

and it's the same

🤔

[rotund token]

why wouldn't use presentation buffer if you were doing it in simulation

[devout prairie]

almost seems as if the renderer is rendering them some internal initialization pos/scale for one frame because they appear out of place and smaller

[devout prairie]

ahh maybe

[rotund token]

(bit of a tangent but personally a system like this should exist in presentation system group because if you were running a headless test client this system shouldn't be updating since it requires rendering/camera/input)

[devout prairie]

ok no i appreciate that

i'm a little unsure on the exact order and usages of those things still if i'm honest

[rotund token]

init | simulation | presentation

[devout prairie]

and the renderer actually renders between presentation back to init?

using begin presentation with the system in simulation actually solves that issue

[rotund token]

hybrid renderer updates in presentation

the actual render stuff happens after that

init called start* of frame

[devout prairie]

yeap ok, that clears it up a bit in my mind

i think it's when i've been doing it that's the issue there

[misty wedge]

Is there a more efficient way to spawn many unique objects using an ECB besides instantiating a prefab and setting values?

The prefab is always the same but the values are mostly unique

[pliant pike]

maybe don't use the ECB?

[celest jay]

Are the values predefined?

[misty wedge]

The majority are (e.g. position isn't)

[celest jay]

You could define the variations in one or more ScriptableObject(s) and just pass that into your instantiated GameObject

[misty wedge]

I don't follow, I'm not using any gameobjects

[celest jay]

you are spawning prefabs, what do you thing that does?

[misty wedge]

Create an ecs entity

[celest jay]

Sorry, misread the question

[misty wedge]

😅

[misty wedge]

Is it significantly faster to spawn using the EntityManager if you have many unique entities?

[pliant pike]

well using the ecb is like two steps EntityManager is one and you can batch instantiate with the entitymanager, I guess it depends on the use case, but I'm guessing it will be faster, but you can always test

[misty wedge]

Why two steps?

[pliant pike]

I only really use the ECB if I specifically need to schedule a structural change

that's what the ecb does it playbacks the commands later on the main thread

[misty wedge]

Well yeah but it just calls EntityManager functions, how is that two steps

[pliant pike]

well the job runs then it puts that all that into the command buffer, then maybe other jobs run in the meantime then when its time the command buffer is played back on the main thread

[misty wedge]

Ok, that's what you meant. But that's not really different from using my own collections / buffers to write the data in a job and then spawn on the main thread (basically what the ECB does)

[pliant pike]

I mean yeah it depends on how you are doing it but it could be faster to just instantiate with EntityManager then set all those values in a job

[misty wedge]

That's not really possible if calculating the values for those entities is expensive (as is the case in my setup)

Where's the best place to call this main thread entity spawning? Schedule a system to run right after something like EndSimulationCommandBufferSystem?

I'd guess close to any sync point would be best

[pliant pike]

I'm not an expert but I'd just do it when you need it

[misty wedge]

Are there any rules against creating "prefabs" at runtime in ECS?

(specifically using the Prefab component)

[pliant pike]

not as far as I know

ECS is kind of new and everyones figuring it out as they go to be honest

[misty wedge]

I wouldn't really call it new at this point tbh, It's been in development for so long already 😅

[pliant pike]

true but they keep changing stuff so your constantly have to learn and experiment

[rustic rain]

if it's meant to be runtime, why not

smth like

saving blueprint

for some building

[misty wedge]

My use case is that I currently have a single ghost prefab for netcode that is basically an "object". There are lots of different ones that can be created though, and the data for these is stored in blobs.

My idea for a performance improvement was to read all blobs and then create spawnable netcode prefabs on the client and server instead of writing all unique values for that object in an ECB each time it is spawned.

I'm not 100% sure how netcode initializes their prefabs though. They do runtime stripping for components in the ghost collection system, but I'm not sure if there is any other step or if they just use the prefab hash to check what kind of prefab was spawned on the client.
For my runtime prefab, I guess I would just generate a unique hash based on the blob that is creating the prefab (which would be identical on the client and server)

[tropic venture]

great news! any chance i can already start playing / testing with this on one of my lower-end mobile devices which dont have Vulkan?
i currently have 2021.3.9f1 and ECS 0.51.1-preview.21
that forum post mentions 2022.2.0a13 - must I have that version?

[rustic rain]

I don't think it's compatible with ECS

not sure at least

[tropic venture]

ah

[devout prairie]

Any particular reason why NonUniformScale may be ignored with an instantiated prefab?

[rustic rain]

could you rephrase?

[devout prairie]

the entity has the component, isn't a physicsbody or anything, but the visual scale of the object and renderbounds don't reflect the NonUniformScale values

So you see the non uniform scale values

Seems to be ignoring those

Ah

It's..

drumroll

... my fault.

[rustic rain]

🤔

[devout prairie]

I'm setting LTW

[rustic rain]

🕯️

[devout prairie]

But not building scale into the 4x4

Doh

[rustic rain]

why do you even set LTW

[devout prairie]

just simpler for the particular use case

[gloomy venture]

Hi all, I'm trying to import some older assets and use them with DOTS. I have selected the Convert to Entity checkbox, and am getting several undesired behaviours. One of the prefabs is breaking up, one is lying on it's side, one is pushing into the floor and one is floating in the air. I assume this is all symptoms of the same issue. Can anyone give me some hints where to start looking to fix this? In all instances the idle animation also stops playing. The screenshots show without ConvertToEntity ticked, and with it ticked.

[rustic rain]

have you worked with Unity ECS before?

[gloomy venture]

relatively new to it. I've fiddled a bit, but this is the first real attempt to build something substantial.

[rustic rain]

I'd suggest to go over whole manual first then
https://docs.unity3d.com/Packages/com.unity.entities@0.51/manual/ecs_core.html

especially important concepts and conversion workflow

[gloomy venture]

so the first thing I'm seeing is that I shouldn't use the ConvertToEntity as the script is going to be deprecated. Should I place the objects that I want to convert into a subscene and write a custom converter?

[rustic rain]

yes, that's basically what you should do

but just keep going

conversion in Unity ECS is pretty important

[gloomy venture]

Ok, I vaguely remember this, I actually did have some of this written previously, so I'm going to give my brain a moment to absorb and remember, and then I'll take another crack at it. Thanks for pointing me in the right direction 👍

[modern glacier]

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

[whole gyro]

I think you are missing some components. I just tried converted a default cube and I see WorldRenderBounds, HybridChunkInfo, ChunkWorldRenderBounds and built-in material properties. Not sure which components you need and which ones are added by the hybrid renderer systems.

I would look into using the helpers in RenderMeshUtility.cs. They make it easier to set up mesh entities

https://docs.unity3d.com/Packages/com.unity.rendering.hybrid@0.51/api/Unity.Rendering.RenderMeshUtility.html

[errant hawk]

I'm having massive issues with any form of physics using raycasts that are a decent amount away from the local origin point of the parent (root) entity.

// This is how im getting the velocity at each point:
float3 velocity = pv.GetLinearVelocity(m, t, r, contact.hit.Position);

//...

// How I'm applying force:
PhysicsComponentExtensions.GetImpulseFromForce(fd.mass, worldForce, ForceMode.Force, deltaTime, out float3 imp, out PhysicsMass pm);
pv.ApplyImpulse(pm, t, r, imp, contact.hit.Position);

It will bounce 1-2 times (as expected) but then it will start spinning out of control, only if the distance from the center is farther than 2-3 meters locally

[modern glacier]

One message removed from a suspended account.

[whole gyro]

I think the problem is with how you create your RenderMesh. I assume the layerMask field is getting a default value of 0, which means it won't render.

[modern glacier]

One message removed from a suspended account.

One message removed from a suspended account.

One message removed from a suspended account.

[whole gyro]

The first error is probably due to the second error, but sometimes shows up harmlessly as well. For the second error, it looks like you aren't using a compatible material. Need to stick with URP, HDRP, or ShaderGraph shaders.

It is possible to get custom shaders to work with hybrid, but you have to add a bunch of boilerplate to the shader and its not really documented anywhere

[modern glacier]

One message removed from a suspended account.

[gloomy venture]

OK, I have now created a Conversion process just using the IConvertGameObjectToEntity, IDeclareReferencedPrefabs approach, but the results are still the same. The mesh is on its side, and broken into the 3 pieces that make it up. Also, the animations don't seem to be running, but I guess one problem at a time. 😮

[whole gyro]

Animation is not currently supported in ECS

If you need full skeletal animation, best option is a hybrid approach where you have game objects with animators that are synced with corresponding entities in the ECS world.

[gloomy venture]

Ok, I'm trying to get a handle on the options here, so can you perhaps confirm whether I can do something like this?

• New Input for Input with ScriptableObjects as a bridge (similar to the Open Project)
• Can I use new Physics with GOs?
• Can I use DOTS for particles, projectiles and other "dumb" assets, and then GOs for the animated characters?
  Or should I just KISS and wait a little longer for DOTS to mature?

[rustic rain]

• Can I use new Physics with GOs?
  No
• Can I use DOTS for particles, projectiles and other "dumb" assets, and then GOs for the animated characters?
  Yes

In fact, you can everything in Unity ECS, but animations with hybrid approach

[whole gyro]

Not familiar with the Open Project so not sure what you are referring to with that. But you can use the new input system with ECS.

The DOTS physics only works with entities, not game objects. However, you can sync game objects to invisible entities which would work with DOTS physics. But it does make things complicated and may not be worth it.

DOTS doesn't have a particle system so it depends what you mean. For things like bullets or other "dumb" assets, DOTS would work great. But if you are trying to more visually complicated particle effects, I would stick with VFX graph.

[rustic rain]

but you'll have to reimplement many things

as it's pretty raw

currently

[gloomy venture]

https://github.com/UnityTechnologies/open-project-1

[whole gyro]

Just to give an example of what's possible, I'll explain what I'm doing in my game.

Most of the objects in the world do not need skeletal animation, so those are all pure entities. I have a main character that is animated, so I use a game object and animator to do that. The character game object follows an invisible entity around which allows it to collide with the entity world. I use the new input system for everything. All input callbacks are handled by an ECS system and written to a singleton entity. Other systems react to changes to that singleton. For particle effects like smoke and lasers, I use VFX graph.

[whole gyro]

Thanks. I know what the Open Project is, but haven't dug into it enough to know what you are talking about with the scriptable object bridge for input.

[gloomy venture]

ah sorry, checkout this devlog video - https://www.youtube.com/watch?v=u1Zel20rwOk

[gentle gyro]

Hi guys,

It seems you can't use Dots with the Quest 2. Can anyone confirm or deny this?

[gloomy venture]

when you say "follow an entity" are they linked in the scene somehow, or do you move the one and then the other?

[whole gyro]

Hard to recommend the best path forward without knowing more about your game. In general, if you needs lots of animated characters or want to rely on lots of 3rd part assets, you may want to stick with game objects. You can still leverage DOTS jobs to speed up parts of your game.

[gloomy venture]

yeah, it's a prototype, so I'm just using store assets wherever I can, I'm certainly no artist. I'm working on isometric top-down shooter, so 3rd person with mostly animated characters.

[rustic rain]

there's built in companion link

which basically copies entity transform to GO transform

[whole gyro]

I wrote my own syncing thing, but works the same way I think. Every frame, I move the game object to match the entity's position.

[rustic rain]

unless you do it from transform job