Data centric techniques for mapping performance data to program variables

Rutar, Nick; Hollingsworth, Jeffrey K.

doi:10.1016/j.parco.2011.10.006

Cited by 3 publications

(2 citation statements)

References 8 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…PAPI's interface allows memory events to be counted and sampled, but with less precision than modern instruction sampling. For architectures without instruction sampling, Rutar and Hollingsworth [17], [18] have developed techniques to mitigate instruction skid so that data-centric attribution can be used with traditional PMUs and PAPI. These techniques would allow our measurement techniques to be applied with some precision loss on architectures that do not provide PEBS or IBS.…”

Section: A Memory Instruction Measurementmentioning

confidence: 99%

“…Many tools have been developed to highlight cache locality problems [19]- [21], to highlight NUMA issues in code [22], [23], and to map these issues to particular variables through static analysis [17], [21], [24]. These techniques differ from ours in that they examine a much more limited set of correlations, and do not provide intuitive linked views of the different domains for analysis.…”

Section: B Memory Profiling Toolsmentioning

confidence: 99%

See 1 more Smart Citation

Dissecting On-Node Memory Access Performance: A Semantic Approach

Giménez

Gamblin

Rountree

et al. 2014

SC14: International Conference for High Performance Computing, Networking, Storage and Analysis

View full text Add to dashboard Cite

Abstract-Optimizing memory access is critical for performance and power efficiency. CPU manufacturers have developed sampling-based performance measurement units (PMUs) that report precise costs of memory accesses at specific addresses. However, this data is too low-level to be meaningfully interpreted and contains an excessive amount of irrelevant or uninteresting information.We have developed a method to gather fine-grained memory access performance data for specific data objects and regions of code with low overhead and attribute semantic information to the sampled memory accesses. This information provides the context necessary to more effectively interpret the data. We have developed a tool that performs this sampling and attribution and used the tool to discover and diagnose performance problems in real-world applications. Our techniques provide useful insight into the memory behavior of applications and allow programmers to understand the performance ramifications of key design decisions: domain decomposition, multi-threading, and data motion within distributed memory systems.

show abstract

Section: A Memory Instruction Measurementmentioning

confidence: 99%