On the first day of the conference, Wednesday, March 25, Gaétan Hains (University Paris-Est Créteil, France) will deliver a keynote address entitled “From algorithm to code to hardware to execution: why isn’t computing performance an experimental science?”, scheduled from 9:30 a.m. to 10:30 a.m. The presentation will take place in the Aula Magna.
Abstract. In the age of pervasive and compute-intensive AI it has never been more important to minimize time and energy consumption. The proliferation of models, optimizations and novel architectures is a natural effect of the race for more advanced functionality. But how does one minimize resources that are not scientifically predictable? We should should be able to approximate and extrapolate real-life execution resources (time, energy) based on source code and architecture parameters. I review the history of computational complexity and real-life performance to suggest that it is time we turn our knowledge into an experimental science.