Vulkan Columnar

Contents

• GPU Programming with Vulkan Columnar
  • 1. Why a GPU for Columnar Data?
  • 2. The Vulkan Compute Pipeline (What You Actually Need)
  • 3. The GPU Memory Model
  • 4. Shader Design Patterns
  • 5. Parallel Algorithms in the Library
  • 6. Cross-Vendor Architecture
  • 7. Library Architecture
  • 8. Optimization Principles
  • 9. Testing and Benchmarking
  • 10. Contributing a New Operation
• Getting Started
  • Prerequisites
  • Building
  • Running Tests
  • Hello Compute
  • Library Usage (C API)
• API Reference
  • Core Types
  • Context Lifecycle
  • Column Management
  • Null Support (Arrow Validity Bitmaps)
  • Arrow C Data Interface
  • Operations
  • Chained Execution
  • Utilities
  • Gather (Indirect Indexing)
  • Element-wise Arithmetic
  • Element-wise Comparison
  • Null Operations
  • Unique / Distinct
  • Boolean Mask Operations
• Contributing
• Performance Optimizations
  • Release Benchmarks (RX 9070, RDNA 4, GCC 13.2 -O3)
  • Planned Optimizations

GPU columnar compute library via Vulkan — filter, cast, and scan operations on typed column arrays, exposed through a C ABI for binding into Polars, DataFusion, Spark, and Python.

Indices and tables

• Index

• Module Index

• Search Page