How to run query over same component in parallel?

Hi,

I have a flag component (seasoncolor) and depending on this flag and ANOTHER component i want to do different things.

Flag + ComponentA = DoXXX();

Flag + ComponentB = DoYYY();

Flag + ComponentC = DoZZZ();

Code (CSharp):

1. var handle1 = Entities
2.                 .WithBurst()
3.                 .WithNone<Plane>()
4.                 .ForEach((ref SeasonColor seasonColor, in LerpData lerpData) =>
5.                 {
6.                         seasonColor.Value = colortA;
7.                 })
8.                 .Schedule(inputDeps);
9.  
10.             var handle2 = Entities
11.                 .WithBurst()
12.                 .WithAll<Plane>()
13.                 .ForEach((ref SeasonColor seasonColor) =>
14.                 {
15.                     seasonColor.Value = colorB;
16.                 })
17.                 .Schedule(inputDeps);

Here you can see that i want to assign colorB to all entities with a plane component and colorA to all entities with lerpdata attached ..

The error is telling me that i write to SeasonColor two times ...

Is there a way to prevent this error from happening? Thanks

 

handle2 must have handle1 as a dependency, since it write to the same type of component

 

I know that this is a workaround, but this shouldnt be needed as a Plane cannot have lerpdata

 

For what purposes or reasons would I use IJobChunk over entities.foreach?

I plan on creating my own ECS physics character controller implementation based on the one of the sample repository which uses archetypechunk syntax. However, I thought of using entities.foreach since it is more concise.

 

If your logic is totaly different then you should porbably do several systems.(or entiteis.foreach)
If most of your logic is shared, then using the IJobChunk like sugested by unity team is a good fit.
You'll write the sahred logic just once and the specifics can be handles within the same job, meaning you get the thread safety.

For exemple (probably a bad one), you may want your character controller to move the entity through physics if it has physicsvelocity component and use through translation otherwise. that is the spécific part but the fact that your character has a speed, a direction, maybe a ground check, that is common. so instead of writing these in 2 systems, you may do it just once, inssuring consistency and maintainability of your code without losing performance.

 

For example, you can sum up total taken damage of each entity, per each chunk. Store summed damage per chunk, it in Native Array, which is size of number of chunks.

Then in second chained job, singlethreaded, you just iterate through that native array of taken damage sums, to calculate final total sum of taken damage.

You have typically much fewer chunks to work on, than number of entities. So running secondary job is fast.

 

What I found interesting that the character controller code from the unity.physics sample uses chunk api while the quite similar dots sample controller uses foreach which makes me wonder if it just boils down to preference in many cases instead of performance considerations.

I haven't gotten that far yet where I need a complex system like unity.transform. I focus for now on the study of the physics character controller which seems to be geared for serial work units and only the last one writes to the physics velocity component. I worry about npc movement which may or may not be suitable to be covered by the same system later (probably depending on the output of the to developed decision making and pathfinding systems).

 

Is this kind of granularity on the roadmap?