Reductionism and simplicity everywhere

I thought I’d carry on talking the _ism_s that feature heavily in my thinking, after the last musing on pragmaticism. It’s a further generalisation of the basic idea I talked about in surface area.

Mostly borne out of my performance analysis work over the years, where I need to extract meaningful data from a complex, interconnected system (the GPU) where the system often displays unexplainable emergent properties, I’m a big fan of reductionist thinking and simple explanation.

Any keen problem solver will have the same tools in their arsenal. It’s usually manifest in a situation where you have to chop a problem, in my case the understanding of the GPU’s performance, into smaller, discrete atomic parts and solve those first, before reassembling the newly understood pieces to get the bigger picture.

In many cases with complex systems it’s the only way to explain something. There can be far too much happening to hold in your head at once and understand, so you need to reduce the system into simpler parts and figure out how they fit together first.

In the PowerVR graphics architecture there’s a notion that the hardware is split into two parts when 3D rendering: everything before pixel shading (TA phase) and everything including it and after it (3D phase). To say that a frame of rendering is therefore “TA limited” with a straight face is to understand a complex pipeline of atomic operations inside of a much more complex enclosing system first. The system has to be reduced and analysed first before you can honestly understand any truth behind the statement.

That desire to mentally break things down into simpler to understand parts can be applied to large swathes of life. There’s no shame to saying, “I don’t understand”, no matter what the subject matter, be it how to put a piece of furniture together or how does this entire GPU work.

By keeping my view of the world simple and reduced, before building up any complex ad-hoc understanding that I need, things seem to mentally work a lot better across the board.