This is a talk that was delivered at the AWS/Arm Cloud Hackathon in July 2021.

John Linford is Arm’s director of HPC and in this talk he covers why you should evolve your profiling beyond using print(), and how much time and effort this can save you. He also dives into some reasons to avoid premature optimization.

The Summer Hackathon ran for a week from July 12-16 in 2021. It’s purpose was to assemble the HPC community around a common goal of beginning the porting/testing/tuning process for dozens of codes to use Arm-based processors, in this case, AWS’s Graviton2’s.

During the week we had a series of talks from global leaders and experts in various fields from compilers to networking, profiling debugging and tooling. We’ve published all the talks here in the HPC Tech Shorts channel, in the hope that everyone gets some value from them.

If you want to join in our collective effort to boost the performance of HPC workloads on Arm, DM us at Twitter (@TechHpc) or head to a-hug.org/ and join the community.