Compute Shader variations

Something.hlsl

//named with '.hlsl' extension to get highlighting, it's really a '.compute'-file
#pragma kernel CSMain

RWTexture3D<float4> Result;

// if Shader Model doesn't support kernel threadgroups in the Z axis
// organize the threads differently
#if SHADER_TARGET < 40
  [numthreads(16, 16, 1)]
#else
  [numthreads(8, 8, 8)]
#endif
void CSMain(uint3 id : SV_DispatchThreadID)
{
  Result[id] = Compute();
}

SomethingCompute.cs

using UnityEngine;
using System;

class SomethingCompute : MonoBehaviour
{
  [SerializeField] private Vector3Int resolution;
  [SerializeField] private ComputeShader shader;
  
  private uint
    numThreadsX,
    numThreadsY,
    numThreadsZ;
  
  private int kernel;
  
  void Awake()
  {
    kernel = shader.FindKernel("CSMain");
    shader.GetKernelThreadGroupSizes(kernel, out numThreadsX, out numThreadsY, out numThreadsZ);
  }
  
  void Dispatch() 
  {
    Debug.LogFormat("Dispatching kernel with groups of size ({0}, {1}, {2})", numThreadsX, numThreadsY, numThreadsZ);
    shader.Dispatch(kernel, resolution.x / (int)numThreadsX, resolution.y / (int)numThreadsY, resolution.z / (int)numThreadsZ);
  }

}

NOTE - resolution (xyz) must be a multiple of numThreads (xyz)