Dynamic loading and linking, overlays, swapping

Dynamic loading loads a routine into memory only when it is first called, so unused routines never occupy memory. Static linking copies library code into every executable; dynamic linking postpones linking to run time, using a stub that locates the routine, and lets many processes share a single copy of a library.

Overlays let a program larger than its allotted memory run by keeping only the currently needed section resident and loading other sections over it; the programmer designs the overlays and an overlay driver loads them. Example: a two-pass assembler with pass 1 = 70 KB, pass 2 = 80 KB, symbol table 30 KB, common routines 20 KB and driver 10 KB needs 30 + 20 + 10 + 80 = 140 KB with overlays.

Swapping temporarily moves a whole process between main memory and a backing store. Swap time is dominated by transfer time plus latency: a 10 MB process at 40 MB/s with 8 ms latency takes 250 + 8 = 258 ms each way, so 516 ms for out and in. A process with pending I/O into its own memory must not be swapped out carelessly, because the device would write into memory owned by someone else.

Exam trap: swapping moves entire processes, whereas paging moves fixed-size pages.