MIT Students Outperform State-of-the-Art HPC Libraries with Hundreds of Lines of Code

Researchers at MIT's CSAIL have developed Exo 2, a new programming language that allows programmers to write 'schedules' explicitly controlling how the compiler generates code, leading to significantly improved performance. Unlike existing User-Schedulable Languages (USLs), Exo 2 lets users define new scheduling operations externally to the compiler, creating reusable scheduling libraries. This enables engineers to achieve performance comparable to, or better than, state-of-the-art HPC libraries with drastically reduced code, revolutionizing efficiency in AI and machine learning applications.