Where to parallelize

[steob92]

When generating backgrounds we need to think of where and how to parallelize the code. It's not clear to me what option is better.

Say we have a runlists of N runs each of which require X hours of backgrounds. To find these backgrounds we need to look at M background runs.

Should we:

• Set each run of interest to be a parallel job. This means means M background runs are looked at in sequence.
• Run each N runs in sequence and look at the M background runs in parallel.

While N (O(10)) << M (O(100)) the answer is clear but as N -> M we might want to look at better optimizations.