This part is very short. Only 400 words or so. So it’s quite challenging to fit in everything I want:
- Benchmarking! This is critical for evaluating systems
- Pet project of using MLIR
- Proof that it’s worth improving PostgreSQL by benchmarking how much time is spent inside the CPU
I am a bit tempted to just double this page and say it’s 4, or even 6, pages. It is only worth 20% of the literature review though, and probably the biggest time sink. So get the benchmarking and proof for postgresql done, but only start the MLIR pet project if everything else is already done.
I will also need to account for all the garbage pages that the LaTeX format demands. It’s going to add a lot of space for references and so on.