OPTIMIZATION OF RECIPROCALS AND SQUARE ROOTS ON THE I860 MICROPROCESSOR

Reciprocal and reciprocal square root operations are partially support ed by the i860 floating point unit, whereas square roots are not. We p oint out the reasons for this, and its consequences for the optimizati on of code involving many reciprocal square roots, such as many-body s imulations involving Coulomb-like potentials. We conclude that code wh ich can be optimized to explicitly combine reciprocals and square root s ill the form of reciprocal square roots can attain significantly hig her performance, and that assembly language coding of such operations can make the greatest use of the hardware by calculating only to the a ccuracy required, which may be less than single precision.