Towards a high-performance AI compiler with upstream MLIR

arXiv:2404.15204v1 Announce Type: cross
Abstract: This work proposes a compilation flow using open-source compiler passes to build a framework to achieve ninja performance from a generic linear algebra high-level abstraction. We demonstrate this flow with a proof-of-concept MLIR project that uses input IR in Linalg-on-Tensor from TensorFlow and PyTorch, performs cache-level optimizations and lowering to micro-kernels for efficient vectorization, achieving over 90% of the performance of ninja-written equivalent programs. The contributions of this work include: (1) Packing primitives on the …

abstract abstraction algebra arxiv build cache compilation compiler concept cs.ai cs.ar cs.dc cs.lg cs.pl flow framework linear linear algebra mlir performance project proof-of-concept pytorch tensor tensorflow type work