Running a large number of diffs

[nadim khemir]

I have a list of files which are potentially different:

p/f1, p2/f1
p/p/f2, p2/p/f2
... ten thousands more entries

is it possible to pass the list to diff without having to start 10 000
processes?

I know I could pass the directories but the entries are not always
under the same root.

with -q I found diff blazing fast to find which files differ but I'd
like to run diff in parallel, having the list of files I can do that
but wonder if there's already way to do that.