Browsing by Author "Tallent, Nathan Russell"

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    Binary analysis for attribution and interpretation of performance measurements on fully-optimized code
    (2007) Tallent, Nathan Russell; Mellor-Crummey, John
    Modern scientific codes frequently employ sophisticated object-oriented design. In these codes, deep loop nests are often spread across multiple routines. To achieve high performance, such codes rely on compilers to inline routines and optimize loops. Consequently, to effectively interpret performance, transformed loops must be understood in the calling context of transformed routines. To understand the performance of optimized object-oriented code, we describe how to analyze optimized object code and its debugging sections to recover its program structure and reconstruct a mapping back to its source code. Using this mapping, we combine the recovered static program structure with dynamic call path profiles to expose inlined frames and loop nests. Experiments show that performance visualizations based on this information provide unique insight into the performance of complex object-oriented codes written in C++. This work is implemented in HPC-TOOLKIT, a performance analysis toolkit.
  • Thumbnail Image
    Item
    Performance analysis for parallel programs from multicore to petascale
    (2010) Tallent, Nathan Russell; Mellor-Crummey, John
    Cutting-edge science and engineering applications require petascale computing. Petascale computing platforms are characterized by both extreme parallelism (systems of hundreds of thousands to millions of cores) and hybrid parallelism (nodes with multicore chips). Consequently, to effectively use petascale resources, applications must exploit concurrency at both the node and system level --- a difficult problem. The challenge of developing scalable petascale applications is only partially aided by existing languages and compilers. As a result, manual performance tuning is often necessary to identify and resolve poor parallel and serial efficiency. Our thesis is that it is possible to achieve unique, accurate, and actionable insight into the performance of fully optimized parallel programs by measuring them with asynchronous-sampling-based call path profiles; attributing the resulting binary-level measurements to source code structure; analyzing measurements on-the-fly and postmortem to highlight performance inefficiencies; and presenting the resulting context- sensitive metrics in three complementary views. To support this thesis, we have developed several techniques for identifying performance problems in fully optimized serial, multithreaded and petascale programs. First, we describe how to attribute very precise (instruction-level) measurements to source-level static and dynamic contexts in fully optimized applications --- all for an average run-time overhead of a few percent. We then generalize this work with the development of logical call path profiling and apply it to work-stealing-based applications. Second, we describe techniques for pinpointing and quantifying parallel inefficiencies such as parallel idleness, parallel overhead and lock contention in multithreaded executions. Third, we show how to diagnose scalability bottlenecks in petascale applications by scaling our our measurement, analysis and presentation tools to support large-scale executions. Finally, we provide a coherent framework for these techniques by sketching a unique and comprehensive performance analysis methodology. This work forms the basis of Rice University's HPCTOOLKIT performance tools.