Parallel Programming - Painful or Productive?
Programming Models & Languages
Scientific Software Development
TimeTuesday, June 18th12pm - 12:30pm
DescriptionIn pursuit of exascale computing, the complexity of parallel hardware architectures will increase and, hence, will the complexity of software development of scientific applications. Developers will spend painfully more effort to parallelize applications or port them to new hardware architectures. Thus, besides expenses for hardware and power consumption, we need to consider 'brainware' costs when moving to the exascale era.
In this talk, we look at parallel programming models in terms of software development effort and productivity. Based on human-subject studies, the effort-performance tradeoff in parallel programming can be evaluated. Scaling such results to an HPC center's perspective, we can get insights to the center's cost effectiveness.
Deputy Group Head HPC