Home

JavaToGpu Wiki (WIP)

JavaToGpu is a source-first Java-to-OpenCL pipeline with an additional structured ASM frontend for advanced integrations.

Pages

• Getting Started
• Kernel Programming Model
• Helpers, Pointers, Vectors, Structs
• OpenCL Features
• Runtime Dispatch and ABI
• ASM Frontend
• Troubleshooting
• Examples and References
• Roadmap and Limitations

Short Summary

• write kernels in Java with @GPU
• reuse GPU helpers with @CCode
• use pointer wrappers, vectors, and @GPUStruct
• map kernel parameters to OpenCL address spaces
• compile through source or structured ASM frontends

How It Works

When you mark a method with @GPU, JavaToGpu:

1. parses the Java source,
2. validates it against the supported GPU subset,
3. lowers it to an internal IR,
4. emits OpenCL C,
5. generates a launcher,
6. rewrites direct calls so runtime execution goes through GpuRuntime.

Main Features

• source-first Java kernel authoring
• reusable GPU helper methods
• OpenCL math, indexing, atomics, fences, and barriers
• pointer wrappers for mutable helper-style code
• vector wrappers for OpenCL vector types
• @GPUStruct for user-defined data layouts
• structured ASM frontend for compiler and generator integrations

Suggested Reading Order

If you want to write kernels by hand:

1. Getting Started
2. Kernel Programming Model
3. Helpers, Pointers, Vectors, Structs
4. OpenCL Features
5. Runtime Dispatch and ABI

If you want to integrate another compiler:

1. ASM Frontend
2. Kernel Programming Model
3. Examples and References