@vgpu/render

0.0.7 — early preview

@vgpu/render is the small rendering layer on top of @vgpu/core. It focuses on explicit WebGPU-style control: create pipelines, encode standalone render passes, or build one frame command encoder with multiple user-ordered passes.

Install

pnpm add @vgpu/render

Exports

Runtime

createRenderPipeline / createRenderPipelineAsync
RenderPass
beginFrame / Frame
createRenderBundle / RenderBundleRecorder
RapidRenderer

Types

RenderPipelineOptions
ColorAttachment
DepthStencilAttachment
RenderPassOptions
RenderPassDrawOptions
RenderPassDynamicOffsets
FrameOptions
RenderBundleOptions

Standalone pass

import { App } from "@vgpu/core";
import { createMockAdapter } from "@vgpu/adapter-mock";
import { createRenderPipeline, RenderPass } from "@vgpu/render";

const { device } = await App.create({ adapter: createMockAdapter() });
const shader = device.createShader(`
@vertex fn vs_main(@builtin(vertex_index) i: u32) -> @builtin(position) vec4f {
  var p = array<vec2f, 3>(vec2f(0.0, 0.5), vec2f(-0.5, -0.5), vec2f(0.5, -0.5));
  return vec4f(p[i], 0.0, 1.0);
}
@fragment fn fs_main() -> @location(0) vec4f { return vec4f(0.2, 0.6, 1.0, 1.0); }
`);
const target = device.createTexture({ size: [64, 64], format: "rgba8unorm", usage: ["render_attachment", "copy_src"] });
const pipeline = createRenderPipeline(device, {
  shader,
  vertex: { entry: "vs_main" },
  fragment: { entry: "fs_main", targets: [{ format: "rgba8unorm" }] },
});
const pass = new RenderPass(device, {
  colorAttachments: [{ view: target, loadOp: "clear", storeOp: "store", clearValue: [0, 0, 0, 1] }],
});
pass.setPipeline(pipeline);
pass.draw(3);
pass.end(); // finishes and submits this one-shot pass

Rapid renderer

RapidRenderer is available for examples and simple one-draw submissions:

import { RapidRenderer } from "@vgpu/render";

const renderer = new RapidRenderer(device);
material.writeUniforms({ viewProjection, model, cameraPosition, light });
await renderer.draw({ material, mesh, target, depthTarget });

For homepage-grade hot paths, do not resolve shaders or rebuild pipelines inside the animation-frame path. If shader source truly changes dynamically, resolve and create the replacement pipeline outside the frame loop, then stage or double-buffer the swap so a completed pipeline is installed at a frame boundary.

Explicit multipass frame

Render bundles are setup-time reusable draw packets, not hidden passes or a render graph:

import { beginFrame, createRenderBundle } from "@vgpu/render";

const bundle = createRenderBundle(device, {
  label: "hero.light-sources.bundle",
  colorFormats: ["rgba8unorm"],
  depthStencilFormat: "depth24plus",
  sampleCount: 1,
  record(bundle) {
    bundle.setPipeline(lightPipeline);
    bundle.setBindGroup(0, lightBindGroup);
    bundle.draw(lightVertexCount);
  },
});

Native WebGPU multipass code uses one command encoder, explicit pass begin/end calls, then one finish/submit:

const encoder = device.gpu.createCommandEncoder({ label: "hero.frame" });

const lightPass = encoder.beginRenderPass(lightPassDescriptor);
lightPass.executeBundles([bundle]);
lightPass.end();

encoder.writeTimestamp(querySet, 0);
encoder.copyBufferToBuffer(srcBuffer, 0, dstBuffer, 0, byteLength);

const compositePass = encoder.beginRenderPass(compositePassDescriptor);
compositePass.setPipeline(compositePipeline);
compositePass.setBindGroup(0, compositeBindGroup);
compositePass.draw(3);
compositePass.end();

device.queue.gpu.submit([encoder.finish()]);

With VGPU Frame, the lifecycle is the same but pass end and final submission are harder to get wrong:

const frame = beginFrame(device, { label: "hero.frame" });
frame.renderPass(lightPassDescriptor, (pass) => {
  pass.executeBundles([bundle]);
});
frame.gpu.writeTimestamp(querySet, 0);
frame.copyBufferToBuffer(srcBuffer, dstBuffer, byteLength);
frame.renderPass(compositePassDescriptor, (pass) => {
  pass.setPipeline(compositePipeline);
  pass.setBindGroup(0, compositeBindGroup);
  pass.draw(3);
});
frame.submit(); // finishes once and submits once

Frame preserves authored ordering and exposes its raw GPUCommandEncoder as frame.gpu for advanced commands. Direct raw encoder calls follow WebGPU behavior; VGPU helper methods guard use after submit() with VGPU-FRAME-SUBMITTED.

For homepage-grade hot paths, create pipelines (sync or async), buffers, bind groups, and render bundles during setup/warmup or resize, then keep per-frame code to command encoding and one queue submit. createRenderPipelineAsync() defaults to fallback: "sync" with a once-only diagnostic when native async creation is unavailable; use fallback: "throw" if warmup must fail rather than block. Avoid runtime resolveShader()/shader creation in frame loops; dynamically changing shader source should be resolved and pipelined off the frame path, with staged or double-buffered swaps at frame boundaries.

Raw .gpu properties such as device.gpu, queue.gpu, buffer.gpu, texture.gpu, shader.gpu, frame.gpu, and RenderBundleRecorder.gpu are intentional advanced escape hatches to native WebGPU objects and are treated as semver-protected public API. Native WebGPU validation and lifecycle rules still apply when using them directly.

License

MIT.