As HPC platforms grow in scale and complexity, understanding the performance problems of applications becomes essential to run efficiently on these machines. Many bottlenecks however, only manifest un...

48th VI-HPS Tuning Workshop (BSC, Barcelona, Spain)

Modern high-performance computing is increasingly driven by heterogeneous systems that combine traditional CPUs with accelerators, particularly GPUs. Among the accelerator programming models, HIP (Het...

On modern processors, compiling is very difficult, as it has to deal with large Instruction Level Parallelism and various vector lengths/instruction sets available, so often compilers generate subopti...

The number of scientific applications using GPUs for accelerated compute is steadily growing, which adds new programming models and practices into play. Moreover, ML applications present new paralleli...

Python is the most popular programming language in use today, and it looks like Python usage will keep growing into the foreseeable future. Old-time HPC programmers (like me) need to get past our atti...

Traces of parallel programs contain an overwhelming amount of data, often making it challenging to identify the root causes of performance issues. This talk presents a methodology that synthesizes tra...

High Performance Computing (HPC) systems are large, heterogeneous, sophisticated – and are therefore so complicated that they are difficult to use efficiently. HPC users are allocated finite compute t...