Title: To a Compiler-Only Code Generation Path for Matrix Multiplication
Abstract:
To support both Artificial Intelligence and High-Performance Computing
workloads, new processors have introduced hardware acceleration for
matrix multiplication. Examples include the Matrix Multiply Assist
(MMA) in the IBM POWER10 and the Advanced Matrix Extensions (AMX) in
the Intel Sapphire Rapids microarchitecture for Xeon servers. This
talk describes how, in collaboration between the University of
Alberta, the University of Campinas, and IBM, we developed compiler
technology to support such accelerators. An initial solution
delivered a robust pattern matcher for General Matrix Multiplication
(GEMM) computation operating at the LLVM intermediate representation
that allows the replacement of the computation with an invocation of
a high-performance library. A later solution develivered a compiler-only
path for code generation by adapting the layered approach used in
numerical libraries to the compiler code-generation process.
Bio: J. Nelson Amaral, a Computing Science professor at the
University of Alberta with a Ph.D. from The University of Texas at
Austin, has published in optimizing compilers and high-performance
computing. Scientific community service includes general chair for
the 23rd International Conference on Parallel Architectures and
Compilation Techniques in 2014, for the International Conference
on Performance Engineering in 2020, and for the International
Conference on Parallel Processing in 2020. Accolades include ACM
Distinguished Engineer, IBM Faculty Fellow, IBM Faculty Awards,
IBM CAS "Team of the Year", awards for excellence in teaching,
the University of Alberta Graduate-Student Association Award for
Excellence in Graduate Student Supervision, an University of
Alberta Award for Outstanding Mentorship in Undergraduate Research
& Creative Activities, and a recent University of Alberta 2020
COVID-19 Remote Teaching Award. [Web page]