You don’t really need a serious amount of computation even for tokamak feedback systems. The computers of the 90s are adequate. Most other confinement machines (magnetic or otherwise) don’t even need feedback. It’s just bog standard industrial automation/control.

We’ve been on the ball for optics for a long time now. Where computation really comes in handy is in simulation and optimization. Plasma physics is still a very unfinished field and computation is a very valuable tool in speeding up the theory <-> experiment loop.