(RP25) Development of Performance Assessment Method Based on Aspen DSL and Micro-Kernels Benchmarking
Performance Analysis and Optimization
TimeTuesday, June 18th8:30am - 10am CEST
LocationSubstanz 1, 2
DescriptionAccurate assessment and predicting the applications performance are essential for the effective usage of modern multi-core computers. Performance models can allow to describe the dynamical behavior of applications on different computing platforms so they can be useful for the design of future supercomputers.
In this research, we propose the application performance assessment method based on Aspen DSL modeling complimented with direct performance measurements and experimental data analyzing.
On the poster an analytical performance model for multispectral images processing application is presented as an example of more general approach. The model construction steps include application profiling, experimental direct measurements of actual FLOPs, processor cycles and different memory usage data.