Optimizing DDA Code on a POWER5 Processor

In this paper we take an existing scientific computation code, DDA, and optimize it to run on an IBM Power5 processor. The DDA code, originally developed by a Ph.D. candidate in physics, suffers from excessive execution time caused by a high number of cache accesses and a low rate of instructions per cycle. Our goal is to improve the code’s performance by making a series of optimizations in a step-by-step manner. The first and second stages of optimizations were done by selecting specific optimization parameters available from IBM’s compiler, xlC. Our next step was to perform handmade optimizations to the code, concentrating mainly on loop fusion techniques. Our last stage of optimization was to incorporate OpenMP into the code in order to take advantage of the dual-cores available on the Power5 system. By using the IBM High Performance Toolkit, we were able to record the change in number of L1 data cache misses and references, IPC, and execution time after each phase of optimization. Using the original source code with no optimizations as the base for our experiments, we were able to obtain a speedup of 12x for “compiler only” optimizations, and an overall speedup of 42x after all modifications were made.