The scientific simulation often requires very long computational time. Several hours for a simulation on a desktop-type workstation is really a daily thing for a computing physicist. And even such cheap and small workstations can perform ten thousand million operations in an hour. But who writes a program which consists of ten thousand million lines? No one writes such a long program and even if you succeeded in making such program, there may be no room in the disk-system for it and it takes much more time for disk-access than for the simulation. The trick is very simple. In scientific simulation, almost all time is usually consumed in DO-loop blocks. Vector computers accelerate the execution of such loops tremendously if the loop satisfies some conditions. The arithmetic performance of one non-vector processor is at most 100 MFLOPS1 now. But if you use a vector computer appropriately, several GFLOPS can be achieved. The purpose of this chapter is to give an elementary introduction to the vector computer and vector processing. In Tab.1, some typical vector computers are listed.
KeywordsAssembly Line Arithmetic Operation Main Memory Float Point Number Arithmetic Performance
Unable to display preview. Download preview PDF.