[Question] Should incremental checkpoints be faster than full checkpoints?

[lloydbrownjr]

Description

I'm trying to understand the expected difference of time taken to dump when doing an incremental checkpoint relative to a regular checkpoint.

Let's say I've taken a program and instrumented it such that at each run I reliably perform checkpoints at the exact same points during the program's execution.

Say I have three different runs:
A - For each dump I use --leave-running
B - For each dump I use --leave-running and --track-mem
C - For each dump I use --leave-running and --track-mem and I pass in the last checkpoint with --prev-images-dir

Should I expect meaningful differences between the time it takes to dump in cases A, B and C assuming that there are a nontrivial amount of pages that are not touched between any two set of checkpoints?