> The "vectorizing" and "pipelining" here seem to work when describing changes/deployments made to the system That's not what I mean. But you're correct, what

> The "vectorizing" and "pipelining" here seem to work when describing changes/deployments made to the system That's not what I mean. But you're correct, what you want to do with the data informs your decision of what should be separate processes, and you once you know, you might decide to split (modularize) the processing code accordingly. Doing that other way around (i.e. to design the modules before you understand the data flow) is just going to cause more trouble.