Trace-based Performance Analysis Tools on a Heterogeneousulticore Platform

Performance evaluation is key to the optimization of computer applications on multicore systems. While many techniques and profiling tools are available for measuringerformance on homogeneous multicore platforms, most of them depend on the hardware support from the vendors. For developing applications on heterogeneous multicore systems, very few analysis tools exist to help the developers. This thesis describes a pure software tracing toolkit, called ParallelTracer that can be ported to a variety of platforms via code instrumentation at the source level. In this thesis, we use the IBM Cell processor as a case study to demonstrate the capabilities of ParallelTracer. Our results show that ParallelTracer provides useful information forrogrammers to understand program behaviors and identify potential performance bottlenecks via graphical visualization. We also discuss the runtime overhead of ParallelTracer. With proper usage, the performance and code size overhead introduced by our toolkit are about 9.73% and 9KB, for the benchmark programs in the case study.