Using cache optimizing compiler for managing software cache on distributed shared memory system

On distributed shared memory (DSM) system, the optimization of memory access is very important for achieving good performance. In this paper we propose an optimizing compiler which controls a software cache system implemented on DSM. The software cache consists of static part related with compiler and dynamic part related with cache-managing runtime routines. Our compiler controls static part of the software cache by using information from static analysis. For applications whose behavior can only be dynamically determined, the compiler uses dynamic part of the software cache. We also propose application of RISC-oriented optimization techniques to parallel applications on our software cache system. We evaluate the efficiency of our compiler and RISC-oriented optimization techniques on distributed parallel machine, CM-5. Our results show that the compiler and the optimizations considerably improve the performance of basic linear algebra routines: matrix multiply, Cholesky decomposition and Gaussian elimination.