As far as I know, in a parallel proof session, the slaves of the selector copy the objects; and they are merged at the end.
I would like to process the entries in parallel, store part of the processed information in a std::map, and at the end (1 thread) work with the map.
So, I can imagine two ways of doing it:
-Share a unique std::map (protected by a mutex)
-Each slave fill a map, and merged them at the end. This could be time consuming, but the size of the map would be small (around several million of elements).
Question: is any of the options possible, or should I use std::thread?
I would like a container to be filled by different slaves, and at the end have only one container. Is it possible? (I think that a global container will be filled independently by the different slaves).