Design pattern and/or data structure for cache friendly memory allocation?

I am a bit new to this site. Feel free to edit or move me to where I or question would fit better.

I am a applied mathematician / computational engineer searching for design patterns and data structures exist for cache-friendly memory allocation (for fast numerical floating point algorithms)?

For example I have N activities for which I want to group data to be local in memory when allocated as I know that when activity # k is executed it will greatly benefit if all of the data it uses is within one cache window.