udf-optimize-cudf

What problem does it solve?

This Skill helps improve the GPU performance of cuDF RapidsUDF implementations without sacrificing correctness, replacing guesswork with a disciplined optimization workflow.

Core Features & Use Cases

• Profiling-Driven Optimization: Uses nsys profiling and kernel statistics to identify bottlenecks before making changes.
• Iterative Benchmarking Loop: Applies one targeted change at a time, then validates it with unit tests and microbenchmarks.
• Safety and Regression Control: Keeps backup checkpoints, restores failed attempts, and maintains a log of what helped or hurt performance.
• Use Case: A Spark engineer can use this Skill to tune a slow custom UDF, compare benchmark results across iterations, and converge on a faster GPU implementation.

Quick Start

Ask the assistant to optimize a specific RapidsUDF class by profiling it, testing it, benchmarking it, and keeping only changes that improve GPU performance.