Objetivo
The multicore revolution (parallel computing on a chip) is predicted by many to be a major disruptive change for the IT industry. The biggest challenges lie in the productive yet efficient usage of parallel computing resources for the large class of programmers who are not experts in parallel hard- and software architecture. Continuing industry trends leading to increased heterogeneity and asymmetry, together with execution environments that demand dynamic adaptability, will further exacerbate the programming challenges.
The MADAME proposal is designed to help programmers develop efficient parallel programs by providing two complementary components. The first component is an efficient monitoring toolset for manycore computing that extends previous work in the area of OpenMP profiling. The second component is a modeling approach that lets users understand performance vs. productivity tradeoffs when developing their code and allows the detection of scalability issues by estimating run-times on platforms that need not be available in real hardware yet. The models are also designed to be used at runtime by upcoming manycore operating systems as a means to communicate quality of service (QoS) requirements and tradeoffs.
Ámbito científico
Convocatoria de propuestas
FP7-PEOPLE-2010-RG
Consulte otros proyectos de esta convocatoria
Régimen de financiación
MC-IRG - International Re-integration Grants (IRG)Coordinador
80539 MUNCHEN
Alemania