Statically Determining Memory Consumption

In real-time and embedded systems, it is often necessary to place conservative upper bounds on the memory required by a program or subprogram. This can be difficult and error-prone process. In this thesis, I have designed and implemented two (related) compile-time analyses to addresses this problem. The first analysis computes a symbolic upper bound on the maximum number of allocations of each object, showing undetermined facts about the program as symbols. The second analysis determines objects in the program that may be allocated statically, without changing the semantics of the program. The symbolic expression is then simplified by removing factors for statically allocated objects. The overall result is a simplified procedure for computing conservative upper bounds on memory. Results on a number of benchmarks are provided. Thesis Supervisor: Martin Rinard Title: Associate Professor