inducer/loopy

Loopy applies symbolic transformation passes—loop tiling, unrolling, vectorization, and data layout restructuring—to high-level loop specifications, enabling fine-grained control over parallelization, memory access patterns, and instruction-level optimizations. It integrates with PyOpenCL and generates OpenCL/CUDA code targeting both GPUs and multi-core CPUs, while remaining domain-specific to dense array computations like linear algebra, convolutions, and PDE solvers.

626 stars and 6,604 monthly downloads. Used by 1 other package. Actively maintained with 13 commits in the last 30 days. Available on PyPI.