This presentation addresses the following topic(s):
- What needs to be done in software to reduce the widening gap between theoretical performance and actual performance?
The use of simulation and AI is an alternative to the experiments and decisions that humans have made over time and cost. These technologies are essential for the creation of a sustainable world in the future. These workloads require massive computing power and must be able to handle large computing environments with high efficiency in order to take the world to the next level. However, the use of large-scale computing environments often causes the problem of performance degradation depending on the architecture. Therefore, the modification which matches architecture and network topology must be introduced in the software side. In this talk, we at Fujitsu introduce the problems we have encountered in computing on large scale computers and how to optimize them from an architectural point of view.
