Develop an OpenMP offload enabled BLAS

PI Name Ye Luo
PI Institution Argonne National Laboratory
Collaborating ANL Division Computational Science (CPS)
Project Description

It is a summer student project

Basic Linear Algebra Subprograms (BLAS) is widely used in scientific and machine learning applications. Although the most performant libraries are provided by hardware vendors, those libraries add additional hurdles to open-source compilers and tools. Having a portable OpenMP offload enabled BLAS for accelerators removes such dependency during the application development process and enables cross platform validation. This project works on implementing the most used BLAS APIs in both batch and non-batched fashion with best practices in software development.

Testbed

NVIDIA GPU