In this paper, the performance improvements of an inverse compressor cascade design are investigated. A Navier-Stokes code is used for flow analysis. A least-square optimization method is used to minimize pressure discrepancies between initial and target cascades. In optimization process, most of the CPU time is associated with the sensitivity analyses. Sensitivities are evaluated on parallel computers by using a finite difference approach. Results show that, analyzing sensitivities on parallel computers reduces the execution time of design significantly. © 2002 The American Institute of Aeronautics and Astronautics Inc. All rights reserved.