decrease memory usage by storing chunk colours as 8bit (at the cost of slower mesh generation)

Sources/Voxelotl/Chunk.swift

@@ -123,7 +123,7 @@ public struct Chunk: Hashable {
 
 public enum BlockType: Hashable {
   case air
-  case solid(_ color: Color<Float>)
+  case solid(_ color: Color<UInt8>)
 }
 
 public struct Block: Hashable {

Sources/Voxelotl/ChunkMeshBuilder.swift

@@ -13,7 +13,7 @@ struct ChunkMeshBuilder {
               .init(
                 position: SIMD3(position) + $0.position,
                 normal: $0.normal,
-                color: SIMD4(color),
+                color: SIMD4(Color<Float>(color).linear),
                 texCoord: $0.texCoord)
             })
             indices.append(contentsOf: sideIndices.map { orig + $0 })

Sources/Voxelotl/Generator/StandardWorldGenerator.swift

@@ -32,9 +32,9 @@ struct StandardWorldGenerator: WorldGenerator {
       if self.ravineMask.get(position * SIMD3(1, 0.441, 1)) >= 0.8 &&
         abs(self.ravineNoise.get(position * SIMD3(1, 0.6, 1))) <= 0.1 { return .air }
 #endif
-      return .solid(.init(
+      return .solid(Color<UInt8>(.init(
         hue: Float(180 + self.colorNoise.get(position) * 180),
-        saturation: 0.47, value: 0.9).linear)
+        saturation: 0.47, value: 0.9)))
     }
 
     let chunkOrigin = chunkID.getPosition()

Sources/Voxelotl/Generator/TerrorTowerGenerator.swift

@@ -19,11 +19,11 @@ struct TerrorTowerGenerator: WorldGenerator {
         let gradient = simd_length(fpos.xz) / 14.0
         let value = self.noise1.get(fpos) - 0.25
       return if gradient + value < threshold {
-        .solid(.init(
+        .solid(Color<UInt8>(.init(
           hue: ((fpos.x * 0.5 + fpos.y) / 30.0) * 360.0,
           saturation: 0.2 + noise2.get(fpos) * 0.2,
           value: 0.75 + noise2.get(SIMD4(fpos, 1) * 0.25)
-        ).linear)
+        )))
       } else {
         .air
       }