Locality of Data

I’m reading about caching and the difference between temporal and spatial locality.

Spatial locality: is the likelihood of accessing memory addresses near recently accessed locations.

Example: MP3 data has strong spatial locality since you know you’re going to need t = 1s and then t = t + 1s (sequential traversal) and so forth as you listen to a track
Code: related functions living close in memory
Data: contiguous data structures (ie. arrays)

Temporal locality is the likelihood of reusing recently accessed memory locations.

Example: simple loops executing over the same data structure repeatedly (ie. games, photoshop filter, etc.)
Code: hot paths being executed frequently (ie. game loop, event handlers)
Data: frequently accessed state (ie. LRU cache)

Using a media player as a case study:

It has good spatial locality as we’re sequentially reading the audio/video frames
It has mixed temporal locality:
- High code (decoder runs continuously)
- Low data (frames usually processed once)

Keeping the above in mind could help make better decisions while: