feat: Always flushing

apps/typegpu-docs/src/content/docs/fundamentals/pipelines.mdx

@@ -325,9 +325,6 @@ It accepts the number of vertices and optionally the instance count, first verte
 After calling the method, the shader is set for execution immediately.
 
 Compute pipelines are executed using the `dispatchWorkgroups` method, which accepts the number of workgroups in each dimension.
-Unlike render pipelines, after running this method, the execution is not submitted to the GPU immediately.
-In order to do so, `root['~unstable'].flush()` needs to be run.
-However, that is usually not necessary, as it is done automatically when trying to read the result of computation.
 
 ### Drawing with `drawIndexed`
 
@@ -376,14 +373,14 @@ const mainFragment = tgpu['~unstable'].fragmentFn({
 const indexBuffer = root
   .createBuffer(d.arrayOf(d.u16, 6), [0, 2, 1, 0, 3, 2])
   .$usage('index');
-  
+
 const pipeline = root['~unstable']
   .withVertex(vertex, { color: vertexLayout.attrib })
   .withFragment(mainFragment, { format: presentationFormat })
   .createPipeline()
   .withIndexBuffer(indexBuffer);
 
-  pipeline 
+  pipeline
     .with(vertexLayout, colorBuffer)
     .drawIndexed(6);
 ```
@@ -407,8 +404,6 @@ root['~unstable'].beginRenderPass(
     pass.draw(3);
   },
 );
-
-root['~unstable'].flush();
 ```
 
 It is also possible to access the underlying WebGPU resources for the TypeGPU pipelines, by calling `root.unwrap(pipeline)`.

apps/typegpu-docs/src/examples/image-processing/blur/index.ts

@@ -65,7 +65,7 @@ const ioLayout = tgpu.bindGroupLayout({
   outTexture: { storageTexture: d.textureStorage2d('rgba8unorm') },
 });
 
-const tileData = tgpu['~unstable'].workgroupVar(
+const tileData = tgpu.workgroupVar(
   d.arrayOf(d.arrayOf(d.vec3f, 128), 4),
 );
 
@@ -191,7 +191,7 @@ function render() {
 
   for (const i of indices) {
     computePipeline
-      .with(ioLayout, ioBindGroups[i])
+      .with(ioBindGroups[i])
       .dispatchWorkgroups(
         Math.ceil(srcWidth / settings.blockDim),
         Math.ceil(srcHeight / 4),
@@ -203,7 +203,6 @@ function render() {
     loadOp: 'clear',
     storeOp: 'store',
   }).draw(3);
-  root['~unstable'].flush();
 }
 render();
 

apps/typegpu-docs/src/examples/rendering/3d-fish/index.ts

@@ -306,8 +306,6 @@ function frame(timestamp: DOMHighResTimeStamp) {
     .with(renderFishBindGroups[odd ? 1 : 0])
     .draw(fishModel.polygonCount, p.fishAmount);
 
-  root['~unstable'].flush();
-
   requestAnimationFrame(frame);
 }
 enqueuePresetChanges();

apps/typegpu-docs/src/examples/rendering/cubemap-reflection/icosphere.ts

@@ -310,8 +310,6 @@ export class IcosphereGenerator {
       .with(bindGroup)
       .dispatchWorkgroups(xGroups, yGroups, 1);
 
-    this.root['~unstable'].flush();
-
     return nextBuffer;
   }
 

apps/typegpu-docs/src/examples/rendering/cubemap-reflection/index.ts

@@ -315,8 +315,6 @@ function render() {
     .with(renderBindGroup)
     .with(textureBindGroup)
     .draw(vertexBuffer.dataType.elementCount);
-
-  root['~unstable'].flush();
 }
 
 function loop() {

apps/typegpu-docs/src/examples/rendering/simple-shadow/index.ts

@@ -352,7 +352,6 @@ function render() {
       }
     },
   );
-  root['~unstable'].flush();
 }
 frameId = requestAnimationFrame(render);
 

apps/typegpu-docs/src/examples/rendering/two-boxes/index.ts

@@ -303,7 +303,6 @@ function render() {
   drawObject(cubeBuffer, bindGroup, 36, 'clear');
   drawObject(secondCubeBuffer, secondBindGroup, 36, 'load');
   drawObject(planeBuffer, planeBindGroup, 6, 'load');
-  root['~unstable'].flush();
 }
 
 function frame() {

apps/typegpu-docs/src/examples/simulation/slime-mold-3d/index.ts

@@ -487,14 +487,16 @@ function frame() {
 
   params.writePartial({ deltaTime });
 
-  blurPipeline.with(computeLayout, bindGroups[currentTexture])
+  blurPipeline
+    .with(bindGroups[currentTexture])
     .dispatchWorkgroups(
       Math.ceil(resolution.x / BLUR_WORKGROUP_SIZE[0]),
       Math.ceil(resolution.y / BLUR_WORKGROUP_SIZE[1]),
       Math.ceil(resolution.z / BLUR_WORKGROUP_SIZE[2]),
     );
 
-  computePipeline.with(computeLayout, bindGroups[currentTexture])
+  computePipeline
+    .with(bindGroups[currentTexture])
     .dispatchWorkgroups(
       Math.ceil(NUM_AGENTS / AGENT_WORKGROUP_SIZE),
     );
@@ -505,12 +507,8 @@ function frame() {
       loadOp: 'clear',
       storeOp: 'store',
     })
-    .with(
-      renderLayout,
-      renderBindGroups[1 - currentTexture],
-    ).draw(3);
-
-  root['~unstable'].flush();
+    .with(renderBindGroups[1 - currentTexture])
+    .draw(3);
 
   currentTexture = 1 - currentTexture;
 

apps/typegpu-docs/src/examples/tests/texture-test/index.ts

@@ -80,14 +80,14 @@ const pipeline = root['~unstable']
 
 function render() {
   pipeline
-    .with(layout, bindGroup)
+    .with(bindGroup)
     .withColorAttachment({
       view: context.getCurrentTexture().createView(),
       loadOp: 'clear',
       storeOp: 'store',
     })
     .draw(3);
-  root['~unstable'].flush();
+
   requestAnimationFrame(render);
 }
 requestAnimationFrame(render);

packages/typegpu/src/core/buffer/buffer.ts

@@ -176,6 +176,8 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
   public readonly resourceType = 'buffer';
   public flags: GPUBufferUsageFlags = GPUBufferUsage.COPY_DST |
     GPUBufferUsage.COPY_SRC;
+
+  readonly #device: GPUDevice;
   private _buffer: GPUBuffer | null = null;
   private _ownBuffer: boolean;
   private _destroyed = false;
@@ -189,12 +191,13 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
   usableAsIndex = false;
 
   constructor(
-    private readonly _group: ExperimentalTgpuRoot,
+    root: ExperimentalTgpuRoot,
     public readonly dataType: TData,
     public readonly initialOrBuffer?: Infer<TData> | GPUBuffer | undefined,
     private readonly _disallowedUsages?:
       ('uniform' | 'storage' | 'vertex' | 'index')[],
   ) {
+    this.#device = root.device;
     if (isGPUBuffer(initialOrBuffer)) {
       this._ownBuffer = false;
       this._buffer = initialOrBuffer;
@@ -205,14 +208,12 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
   }
 
   get buffer() {
-    const device = this._group.device;
-
     if (this._destroyed) {
       throw new Error('This buffer has been destroyed');
     }
 
     if (!this._buffer) {
-      this._buffer = device.createBuffer({
+      this._buffer = this.#device.createBuffer({
         size: sizeOf(this.dataType),
         usage: this.flags,
         mappedAtCreation: !!this.initial,
@@ -317,7 +318,6 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
 
   write(data: Infer<TData>): void {
     const gpuBuffer = this.buffer;
-    const device = this._group.device;
 
     if (gpuBuffer.mapState === 'mapped') {
       const mapped = gpuBuffer.getMappedRange();
@@ -330,16 +330,12 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
       this._hostBuffer = new ArrayBuffer(size);
     }
 
-    // Flushing any commands yet to be encoded.
-    this._group.flush();
-
     this._writeToTarget(this._hostBuffer, data);
-    device.queue.writeBuffer(gpuBuffer, 0, this._hostBuffer, 0, size);
+    this.#device.queue.writeBuffer(gpuBuffer, 0, this._hostBuffer, 0, size);
   }
 
   public writePartial(data: InferPartial<TData>): void {
     const gpuBuffer = this.buffer;
-    const device = this._group.device;
 
     const instructions = getWriteInstructions(this.dataType, data);
 
@@ -352,7 +348,7 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
       }
     } else {
       for (const instruction of instructions) {
-        device.queue.writeBuffer(
+        this.#device.queue.writeBuffer(
           gpuBuffer,
           instruction.data.byteOffset,
           instruction.data,
@@ -365,19 +361,15 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
 
   public clear(): void {
     const gpuBuffer = this.buffer;
-    const device = this._group.device;
 
     if (gpuBuffer.mapState === 'mapped') {
       new Uint8Array(gpuBuffer.getMappedRange()).fill(0);
       return;
     }
 
-    // Flushing any commands yet to be encoded.
-    this._group.flush();
-
-    const encoder = device.createCommandEncoder();
+    const encoder = this.#device.createCommandEncoder();
     encoder.clearBuffer(gpuBuffer);
-    device.queue.submit([encoder.finish()]);
+    this.#device.queue.submit([encoder.finish()]);
   }
 
   copyFrom(srcBuffer: TgpuBuffer<MemIdentity<TData>>): void {
@@ -386,16 +378,13 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
     }
 
     const size = sizeOf(this.dataType);
-    const encoder = this._group.commandEncoder;
+    const encoder = this.#device.createCommandEncoder();
     encoder.copyBufferToBuffer(srcBuffer.buffer, 0, this.buffer, 0, size);
+    this.#device.queue.submit([encoder.finish()]);
   }
 
   async read(): Promise<Infer<TData>> {
-    // Flushing any commands yet to be encoded.
-    this._group.flush();
-
     const gpuBuffer = this.buffer;
-    const device = this._group.device;
 
     if (gpuBuffer.mapState === 'mapped') {
       const mapped = gpuBuffer.getMappedRange();
@@ -410,12 +399,12 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
       return res;
     }
 
-    const stagingBuffer = device.createBuffer({
+    const stagingBuffer = this.#device.createBuffer({
       size: sizeOf(this.dataType),
       usage: GPUBufferUsage.COPY_DST | GPUBufferUsage.MAP_READ,
     });
 
-    const commandEncoder = device.createCommandEncoder();
+    const commandEncoder = this.#device.createCommandEncoder();
     commandEncoder.copyBufferToBuffer(
       gpuBuffer,
       0,
@@ -424,8 +413,8 @@ class TgpuBufferImpl<TData extends AnyData> implements TgpuBuffer<TData> {
       sizeOf(this.dataType),
     );
 
-    device.queue.submit([commandEncoder.finish()]);
-    await device.queue.onSubmittedWorkDone();
+    this.#device.queue.submit([commandEncoder.finish()]);
+    await this.#device.queue.onSubmittedWorkDone();
     await stagingBuffer.mapAsync(GPUMapMode.READ, 0, sizeOf(this.dataType));
 
     const res = readData(

packages/typegpu/src/core/pipeline/computePipeline.ts

@@ -190,7 +190,8 @@ class TgpuComputePipelineImpl implements TgpuComputePipeline {
       ...setupTimestampWrites(this._priors, branch),
     };
 
-    const pass = branch.commandEncoder.beginComputePass(passDescriptor);
+    const commandEncoder = branch.device.createCommandEncoder();
+    const pass = commandEncoder.beginComputePass(passDescriptor);
 
     pass.setPipeline(memo.pipeline);
 
@@ -216,6 +217,7 @@ class TgpuComputePipelineImpl implements TgpuComputePipeline {
 
     pass.dispatchWorkgroups(x, y, z);
     pass.end();
+    branch.device.queue.submit([commandEncoder.finish()]);
 
     if (memo.logResources) {
       logDataFromGPU(memo.logResources);