2. There is a simple relation between coverage (C), window length (WL), read length (RL) and average number of reads per bin (m): C=m*RL/WL. cn.MOPS has a good performance if the average number of reads in a window is 50-100. Assuming your reads have read length of 100 and a coverage of 50X, this suggests a WL of 100bp-200bp.
3. cn.mops does not provide a function for calculating the coverage but it is easy to determine if you have already calculated the read counts:
You could - for example - do an initial run with a large WL, then calculate the read counts and coverage. Based on these coverage values you can determine a more suitable WL and recalculate the read counts. Probably there is a function to calculate the coverage in Rsamtools.
4. Yes, of course. Almost all functions have an argument "parallel" that you can set to the desired number of processing cores:
cn.mops(..., parallel=24) would run cn.mops on 24 cores.
Thanks for your questions!