Featured is a method for reducing overhead associated with system input output (I/O) operations in a computer system having a plurality of processors and a physical memory accessed and used by the plurality of processors. The physical memory being accessed can be a global physical memory such as that used with SMP types of architectures or distributed physical memory such as that used with CCNUMA types of architectures. Such a method includes creating a pinned virtual memory range database in which is stored virtual memory address information corresponding to pinned physical memory for each applications program being run on the computer system. Also featured is an operating system for execution with a multiprocessor computer system and a multiprocessor computer including such an operating system for execution therein.