GPU Tables

From v9.4Experimental API

@luma.gl/experimental/gpu-tables provides private batch-preserving table helpers for rendering, transforms, and compute. It builds on the primitive objects in @luma.gl/gpgpu/gpu-data.

Start Here

• Read the GPU Tables guide for end-to-end attribute and storage workflows.
• Read GPU Table Structure for the logical-column and physical-batch object model.
• Use GPUTable for data ownership and GPUTableShaderBindings for backend-specific shader resources.
• Use @luma.gl/arrow adapters when source data is Apache Arrow.

API Reference

• GPU Table Structure
• GPU Table Lifecycle
• GPUTable
• GPURecordBatch
• GPUSchema
• GPUInputSchema
• GPUTable Shader Bindings
• GPUTableBufferPlanner
• GPGPU Data Primitives
• Supported Arrow Types

The subpath owns GPURecordBatch, GPUTable, GPUSchema, GPUField, GPUTypeMap, table bindings, table computations, and generic table planners. Models can publish GPUInputSchema declarations for the prepared table inputs they accept. GPUTableShaderBindings resolves those declarations into owned, batch-preserving attribute and storage resources for a shader layout. Table-oriented execution helpers include TableTransform, GPUTableComputation, GPUTableBufferPlanner, and generated-buffer batching utilities. GPUTableModel renders preserved table batches through one model pipeline, while GPUTableGeometry exposes a packed static table as ordinary GPU geometry. Specialized path and polygon models live separately in @luma.gl/experimental/models.

Arrow-specific construction and analysis helpers live in @luma.gl/arrow. Applications that ingest Apache Arrow data should use those adapters to build the generic GPU table objects exposed here.

Installing

This subpath is experimental in 9.4 and is installed with its containing package. Install GPGPU as well when constructing primitive data objects directly:

npm install @luma.gl/gpgpu @luma.gl/experimental