The Physics of Computational Physics

Vincent Reverdy

Over the last decades, numerical simulations have become an essential element of the physicist toolbox. However, even if the physics is well known, as well as algorithms and numerical methods, implementing a physics code still often requires years of development. Coding is even becoming the main activity of a growing number of physicists. The common story is that this is due to the fact that physics is a complex field and that the nature of research requires constant changes. The very same story also tells us that computational power is the main limitation of modern numerical simulations. But what if one of the main challenge of computational physics was its lack of theorization ? What if codes were everything but technical details ? In this talk, built as a discussion, I will explore this strange question : if the objects of computational physics, namely simulation codes, were physical objects then what the physics of computational physics would look like ? What would be the benefits of the search for symmetries, invariance and of the application of Occam’s razor ? Mathematicians have made proofs mathematical objects a while ago, giving birth to proof theory. It may be time to explore a similar path in computational physics.