Extending a Light-weight Runtime System by Dynamic Instrumentation for Performance Evaluation
Mario Kicherer, Fabian Nowak, Rainer Buchty, Wolfgang Karl
Abstract
Mario Kicherer, Fabian Nowak, Rainer Buchty, Wolfgang Karl
Abstract
Increasing complexity of current and future systems poses a new challenge for software engineers. In a previous work we presented a light-weight runtime system for abstraction of heterogeneous parallel systems. This runtime system adapts an application to the current system state in order to improve the utilization of the available resources. For online evaluation of such adaptions, we propose a versatile instrumentation mechanism that can be used by the runtime system. We show that this technique introduces only minor overhead and we compare it to the Dyninst mechanism.
A significance statement is not available in the OpenAlex record.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
Increasing complexity of current and future systems poses a new challenge for software engineers. In a previous work we presented a light-weight runtime system for abstraction of heterogeneous parallel systems. This runtime system adapts an application to the current system state in order to improve the utilization of the available resources. For online evaluation of such adaptions, we propose a versatile instrumentation mechanism that can be used by the runtime system. We show that this technique introduces only minor overhead and we compare it to the Dyninst mechanism.
Key concepts: Computer science, Instrumentation (computer programming), Abstraction, Overhead (engineering), Distributed computing, Software, State (computer science), Embedded system