toyc: Data-Parallel Arithmetic Evaluation on Vulkan Compute for Expression Compilation
Year: 2026
Type: preprint
Abstract: toyc is a compiler for arithmetic expressions that targets Vulkan compute shaders, enabling data-parallel evaluation on the GPU. The compiler implements a full pipeline: a lexer and recursive descent parser produce an AST, which is then flattened and compiled to SPIR-V via a custom Vulkan backend. A CPU fallback backend is also included for comparison. The primary goal is to evaluate whether GPU execution via Vulkan compute offers a practical performance advantage over CPU evaluation for arithmetic expression workloads. Benchmark results and analysis are included. This is a preprint. Source code is available at https://github.com/spy1345a/toyc-repo Benchmark code is available at https://github.com/spy1345a/Toyc-benchmark
Author: M B, Ajesh
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225050394