[diroussel]

Yeah. Thinking of them as a starting point is a good way to think of the approach.

I had some code that was slower than I intuitively expected it to be. I looked around for approaches and came across cache oblivious algorithms. In my case I was thinking of RAM as a cache and some of the inputs would fit in RAM and some were larger than RAM. All inputs and outputs were in network drives. The final output was 4.5 GB. Using cache obvious algorithms as a inspiration I managed to get my runtime down from 5 mins to 20 seconds (numbers are approximate as this was nearly ten years ago now).

I was quite pleased.