FlyDSL Kernel Authoring Skill

What problem does it solve?

This Skill provides a powerful framework for writing and compiling high-performance GPU kernels, specifically targeting AMD GPUs, by offering fine-grained control over data layout and memory access.

Core Features & Use Cases

• GPU Kernel Development: Write custom kernels for AMD GPUs (MI300X/MI350) using a Python DSL.
• Layout Algebra: Precisely control data movement, tiling, and memory access patterns with an explicit layout system.
• MLIR Compilation: Leverages MLIR for a robust compilation pipeline from Python DSL to HSACO binaries.
• Use Case: Optimize complex matrix multiplication (GEMM) or other compute-intensive operations for AMD hardware by defining custom data layouts and kernel logic.

Quick Start

Use the FlyDSL Kernel Authoring Skill to define a new GPU kernel for matrix multiplication.