A comparison of the CM with the DAP for lattice gauge theory (Q1262690)

Many problems requiring large-scale computations need a code language, for an advanced architecture of programs. One of these is the lattice gauge theory which, using massively parallel computers such as the connection machine (CM) and distributed array processor (DAP), became one of the most challenging large-scale scientific computations. The authors show that they have successfully implemented an example of gauge theory, the quantum electro-dynamics (QED), that is to say the theory which describes the interactions between electrons and photons, on the CM. The performance obtained on a 64K CM-2, with hardware support for floating-point operations is of order of a million site updates per second, for QED. They compare the CM with the DAP for lattice gauge theory, too and they find that for QED the CM calculate roughly three times faster, but communicate six times slower than DAP.