A visual performance debugger for Concordia Parallel Programming Environment

This thesis presents the design and implementation of the Visual Performance Debugger for the Concordia Parallel Programming Environment (CPPE), a simulator for parallel programming environment. The purpose of this visual performance debugger is to provide a rich set of correctness and performance debugging tools for CPPE to make CPPE a real flexible and efficient system for parallel program development. The challenge in the design of the performance debugger is the comprehensive recognition of the interaction of performance affecting factors and their impact on parallel applications running on a parallel system. We intend to provide the user with the tool set that can gather performance data regarding different performance affecting factors at different level of details to facilitate the performance analysis and fine-tuning. To this end, a comprehensive survey of performance affecting factors for parallel programming is conducted. Based on this recognition, we have designed and implemented a rich set of correctness and performance debugging tools that can significantly help the user in their parallel programming exercise for correctness debugging and performance tune. We provide a visual environment for this performance debugger in order to achieve these goals more easily