Skip to main content

Using GPU Buffers

See also GPU Commands for guidance on when buffer operations should use immediate resource helpers versus explicit command encoding.

Buffer Operations​

The ability to copy memory between CPU, buffers and textures

DimensionWebGPUWebGL2Description
Buffer.write()✅✅Write a buffer synchronously
Buffer.mapAndWriteAsync()✅✅ *Write a buffer synchronously
Buffer.readAsync()✅✅ *Read a buffer asynchronously without copy.
Buffer.mapAndReadAsync()✅✅ *Read a buffer asynchronously
Buffer.readSyncWebGL()❌✅Read a buffer synchronously
copyBufferToBuffer✅✅Copy a buffer to another buffer without CPU roundtrip
copyBufferToTexture✅✅ *Copy a buffer to a texture without CPU roundtrip
copyTextureToBuffer✅✅ *Copy a buffer to a texture without CPU roundtrip

Remarks:

  • The mapAndWriteAsync() API is available on WebGL2, however a temporary buffer is created. For optimal performance, applications may want to use write() on WebGL2.
  • The mapAndReadAsync() API is available on WebGL2, however the data is actually copied. The lifetime callback parameter indicates whether the ArrayBuffer can be retained.
  • Asynchronous buffer reads are emulated on WebGL2. The actual reads are still synchronous under the hood.
  • A WebGL extension does exist that enables asynchronous buffer reads, but it is not implemented on MacOS which is the primary development environment for luma.gl.
  • On WebGPU, buffer-to-texture and texture-to-buffer copies use linear buffer layouts, so bytesPerRow must satisfy WebGPU's row-alignment rules (typically a multiple of 256). Packed CPU-side data should use Texture.writeData() instead of a buffer copy path.
  • On WebGL, copy commands are best-effort compatibility operations. They are portable, but they do not imply WebGPU-style deferred command recording for rendering.