CubeCL
Multi-platform high-performance compute language extension for Rust
...CubeCL focuses on delivering predictable performance and composability by exposing explicit control over memory layouts, parallelism, and execution patterns while still maintaining a developer-friendly syntax. The framework is built to integrate tightly with modern ML stacks, enabling efficient tensor operations and custom kernel development that can outperform generic libraries in specialized workloads. By combining compiler optimizations with a domain-specific language, CubeCL allows developers to generate highly optimized code for different hardware backends while maintaining a single source of truth.