Assigning Blame: Mapping Performance to High Level Parallel Programming Abstractions

Abstract

Abstract. Parallel programs are increasingly being written using programming frameworks and other environments that allow parallel constructs to be programmed with greater ease. The data structures used allow the modeling of complex mathematical structures like linear systems and partial differential equations using high-level programming abstractions. While this allows programmers to model complex systems in a more intuitive way, it also makes the debugging and profiling of these systems more difficult due to the complexity of mapping these high level abstractions down to the low level parallel programming constructs. This work discusses mapping mechanisms, called variable blame, for creating these mappings and using them to assist in the profiling and debugging of programs created using advanced parallel programming techniques. We also include an example of a prototype implementation of the system profiling three programs.

Decision support via automated metric comparison for the palladio-based performance blame analysis

“…Another approach that deals with performance blame analysis is exhibited by Rutar and Hollingsworth [15]. The approach computes the blame for variables and data structures on the basis of data flow.…”

“…In case they do support component performance contracts, these contracts are not parameterized by the component's context (e.g. [15,17]). The context of a component is defined as the component's usage, its connected external services, and its allocation on execution environments.…”

PBlaman: performance blame analysis based on Palladio contracts

“…Another performance blame analysis approach is exhibited by Rutar and Hollingsworth [6]. The approach computes blame for variables and data structures on the basis of data flow.…”

“…The process of deciding who is able to and responsible for fixing a performance error in a component-based system is called performance blame analysis. Performance blame analysis is especially important in componentbased and service-oriented software engineering where development processes are decoupled and entail strictly separated developer roles that interact through a component or service market.Existing performance blame analysis approaches [6,7] do not consider the expected performance of components at all. As a consequence, they are restricted to providing aggregated performance results, which a system architect must filter and interpret manually.…”

“…Existing performance blame analysis approaches [6,7] do not consider the expected performance of components at all. As a consequence, they are restricted to providing aggregated performance results, which a system architect must filter and interpret manually.…”