The user now launched the "Convert all" function, it should run a seperate process for each of them items.
What's the recommeded way, convert them one after the other one by one, or launch a seperate conversion thread for each of the items letting all the processes work simultaneously?
My question is, maybe since there are too many progresses at a time, all of them work slower which might then be better converting them one by one? What is the most efficient approach?
The answer to this is highly specific to the nature of the processing (it's highly dependent on what the bottlenecks are, what the shared resources are, etc.). Only you can answer the question by writing it both ways and using a profiler to measure the performance.