data view versus data set

I mostly use data views as I deal with very large sets of information. My question is on using a view versus the ordinary data step when the data is split and merged back together after operations are performed. For example:

data a/view=a;
set library.stored_data;
operations on the data;
run;

data split1/view=split1;
set a(where some condition splits the data);
operations on the data;
run;

data split2/view=split2;
set a(where some condition splits the data in some other way);
operations on the data;
run;

data remerge/view=remerge;
merge split1 split2;
by key_variable;
run;

proc means data=remerge;
run;

Now my question is library.stored_data read from the hard drive once for the initial view or twice, once for each partition? If so, for more complex partitioning of data, could it be advantageous to save the data to an ordinary data step at any point? Processing time efficiency and storage space trump elegance or simplicity for the work I am doing. In practice my work involves over a hundred data steps and procedures, but a good portion is dividing data and merging it back together after having been transformed.