a lot of useless optimizations

zardoy · Dec 13, 2024 · 71d8e73 · 71d8e73
1 parent 9000445
commit 71d8e73
Show file tree

Hide file tree

Showing 4 changed files with 47 additions and 40 deletions.
diff --git a/prismarine-viewer/examples/Cube.comp.wgsl b/prismarine-viewer/examples/Cube.comp.wgsl
@@ -70,8 +70,8 @@ fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
       clipX = clamp(clipX, -1, 1);
     }
     var pos : vec2u = vec2u(u32((clipX * 0.5 + 0.5) * f32(textureSize.x)),u32((clipY * 0.5 + 0.5) * f32(textureSize.y)));
-
-    if (linearize_depth_ndc(clipDepth, 0.05, 10000) - 20 > linearize_depth_ndc(textureLoad(depthTexture, vec2u(pos.x, textureSize.y - pos.y), 0), 0.05, 10000)  && !nearby) {
+    let k = linearize_depth_ndc(clipDepth, 0.05, 10000) ;
+    if (k- 20 > linearize_depth_ndc(textureLoad(depthTexture, vec2u(pos.x, textureSize.y - pos.y), 0), 0.05, 10000)) {
       return;
     }
     if (nearby) {
@@ -84,7 +84,7 @@ fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
       pos.y = index % textureSize.y;
       }
     }
-    let depth = u32(clipDepth * 10000);
+    let depth = u32(k + 1.5);
     var depthPrev = atomicMin(&depthAtomic.locks[pos.x][pos.y], depth);
     //depthPrev = atomicLoad(&depthAtomic.locks[pos.x][pos.y]);
     if (depth < depthPrev) {

diff --git a/prismarine-viewer/examples/Cube.vert.wgsl b/prismarine-viewer/examples/Cube.vert.wgsl
@@ -29,28 +29,7 @@ struct VertexOutput {
 @group(0) @binding(3) var<storage, read> models: array<CubeModel>;
 @group(1) @binding(1) var<storage, read> visibleCubes: array<CubePointer>;
 @group(1) @binding(2) var<storage, read> chunks : array<Chunk>;
-
-fn rotationX(angle: f32) -> mat4x4<f32> {
-    let c = cos(angle);
-    let s = sin(angle);
-    return mat4x4<f32>(
-        vec4<f32>(1.0, 0.0, 0.0, 0.0),
-        vec4<f32>(0.0, c, -s, 0.0),
-        vec4<f32>(0.0, s, c, 0.0),
-        vec4<f32>(0.0, 0.0, 0.0, 1.0),
-    );
-}
-
-fn rotationY(angle: f32) -> mat4x4<f32> {
-    let c = cos(angle);
-    let s = sin(angle);
-    return mat4x4<f32>(
-        vec4<f32>(c, 0.0, s, 0.0),
-        vec4<f32>(0.0, 1.0, 0.0, 0.0),
-        vec4<f32>(-s, 0.0, c, 0.0),
-        vec4<f32>(0.0, 0.0, 0.0, 1.0),
-    );
-}
+@group(0) @binding(4) var<uniform> rotatations: array<mat4x4<f32>, 6>;
 
 @vertex
 fn main(
@@ -82,37 +61,32 @@ fn main(
 
   var normal : mat4x4<f32>;
   var Uv = vec2(uv.x, (1.0 - uv.y));
+  normal = rotatations[normalIndex];
   switch (normalIndex) {
     case 0:
     {
        Uv = vec2((1.0f-uv.x), (1.0 - uv.y));
-       normal = rotationX(radians(-90f));
        textureIndex = models[modelIndex].textureIndex123 & 1023;
     }
     case 1:
     {
-      normal = rotationX(radians(90f));
       textureIndex = (models[modelIndex].textureIndex123 >> 10) & 1023;
     }
     case 2:
     {
-      normal = rotationX(radians(0f));
       textureIndex = (models[modelIndex].textureIndex123 >> 20) & 1023;
     }
     case 3:
     {
       Uv = uv;
-      normal = rotationX(radians(180f));
       textureIndex = models[modelIndex].textureIndex456 & 1023;
     }
     case 4:
     {
-      normal = rotationY(radians(90f));
       textureIndex = (models[modelIndex].textureIndex456 >> 10) & 1023;
     }
     case 5, default:
     {
-     normal = rotationY(radians(-90f));
      textureIndex = (models[modelIndex].textureIndex456 >> 20) & 1023;
     }
   }

diff --git a/prismarine-viewer/examples/webgpuRenderer.ts b/prismarine-viewer/examples/webgpuRenderer.ts
@@ -93,6 +93,7 @@ export class WebgpuRenderer {
   depthTextureAnother: GPUTexture
   volumetricRenderPassDescriptor: GPURenderPassDescriptor
   tempTexture: GPUTexture
+  rotationsUniform: GPUBuffer
 
 
   // eslint-disable-next-line max-params
@@ -311,6 +312,26 @@ export class WebgpuRenderer {
       usage: GPUBufferUsage.UNIFORM | GPUBufferUsage.COPY_DST,
     })
 
+    this.rotationsUniform = device.createBuffer({
+      size: Mat4x4BufferSize * 6,
+      usage: GPUBufferUsage.UNIFORM | GPUBufferUsage.COPY_DST,
+    })
+
+    const matrixData = new Float32Array([
+      ...new THREE.Matrix4().makeRotationX(THREE.MathUtils.degToRad(90)).toArray(),
+      ...new THREE.Matrix4().makeRotationX(THREE.MathUtils.degToRad(-90)).toArray(),
+      ...new THREE.Matrix4().makeRotationY(THREE.MathUtils.degToRad(0)).toArray(),
+      ...new THREE.Matrix4().makeRotationY(THREE.MathUtils.degToRad(180)).toArray(),
+      ...new THREE.Matrix4().makeRotationY(THREE.MathUtils.degToRad(-90)).toArray(),
+      ...new THREE.Matrix4().makeRotationY(THREE.MathUtils.degToRad(90)).toArray(),
+    ])
+
+    device.queue.writeBuffer(
+      this.rotationsUniform,
+      0,
+      matrixData
+    )
+
     this.cameraComputeUniform = device.createBuffer({
       size: Mat4x4BufferSize,
       usage: GPUBufferUsage.UNIFORM | GPUBufferUsage.COPY_DST,
@@ -588,6 +609,12 @@ export class WebgpuRenderer {
           resource: {
             buffer: this.modelsBuffer
           },
+        },
+        {
+          binding: 4,
+          resource: {
+            buffer: this.rotationsUniform
+          }
         }
       ],
     })
@@ -634,6 +661,12 @@ export class WebgpuRenderer {
           resource: {
             buffer: this.modelsBuffer
           },
+        },
+        {
+          binding: 4,
+          resource: {
+            buffer: this.rotationsUniform
+          }
         }
       ],
     })
@@ -962,7 +995,7 @@ export class WebgpuRenderer {
     //   this.multisampleTexture = multisampleTexture
     // }
 
-    
+
 
     // TODO!
     if (this.rendererParams.godRays) {
@@ -987,9 +1020,9 @@ export class WebgpuRenderer {
 
     this.commandEncoder = device.createCommandEncoder()
     //this.commandEncoder.clearBuffer(this.occlusionTexture)
-    
+
     //this.commandEncoder.clearBuffer(this.DepthTextureBuffer);
-    if (this.rendererParams.occlusion) {
+    if (this.rendererParams.occlusionActive) {
       this.commandEncoder.clearBuffer(this.occlusionTexture)
       this.commandEncoder.clearBuffer(this.visibleCubesBuffer)
       device.queue.writeBuffer(this.indirectDrawBuffer, 0, this.indirectDrawParams)
@@ -999,7 +1032,7 @@ export class WebgpuRenderer {
     device.queue.writeBuffer(this.textureSizeBuffer, 0, textureSize)
 
     if (this.realNumberOfCubes) {
-      if (this.rendererParams.occlusion) {
+      if (this.rendererParams.occlusionActive) {
         {
           const computePass = this.commandEncoder.beginComputePass()
           computePass.label = 'Frustrum/Occluision Culling'
@@ -1021,6 +1054,9 @@ export class WebgpuRenderer {
           computePass.setBindGroup(2, this.textureSizeBindGroup)
           computePass.dispatchWorkgroups(Math.ceil(this.canvas.width / 16), Math.ceil(this.canvas.height / 16))
           computePass.end()
+          if (!this.indirectDrawBufferMapBeingUsed) {
+            this.commandEncoder.copyBufferToBuffer(this.indirectDrawBuffer, 0, this.indirectDrawBufferMap, 0, 16)
+          }
           device.queue.submit([this.commandEncoder.finish()])
         }
       }
@@ -1041,9 +1077,6 @@ export class WebgpuRenderer {
         }
         renderPass.end()
 
-        if (!this.indirectDrawBufferMapBeingUsed) {
-          this.commandEncoder.copyBufferToBuffer(this.indirectDrawBuffer, 0, this.indirectDrawBufferMap, 0, 16)
-        }
 
         device.queue.submit([this.commandEncoder.finish()])
       }

diff --git a/prismarine-viewer/examples/webgpuRendererShared.ts b/prismarine-viewer/examples/webgpuRendererShared.ts
@@ -7,7 +7,7 @@ export const defaultWebgpuRendererParams = {
   cameraOffset: [0, 0, 0] as [number, number, number],
   webgpuWorker: workerParam ? workerParam === 'true' : !isSafari,
   godRays: false,
-  occlusion: true
+  occlusionActive: true,
 }
 
 export const rendererParamsGui = {
@@ -17,7 +17,7 @@ export const rendererParamsGui = {
     qsReload: true
   },
   godRays: true,
-  occlusion: true
+  occlusionActive: true,
 }
 
 export type RendererInitParams = GPURequestAdapterOptions & {}