Skip to content

Milestones

List view

  • Finish first iteration of MLO

    The current system developed during the 3rd year group project still lacks certain features... Need to implement them to make it fully operational. 1.Need to match previous performance 2.Need to test all the requested features.

    Overdue by 13 year(s)
    •
    Due by January 25, 2013
    •9/11 issues closed
    81% complete2 open9 closed

  • Asynchronous MLO

    With a grid of machines available for benchmark evaluation distribution of workload is a clear way of optimization acceleration. Unfortunately numerous issues arise when extending MLO with this feature. When sampling the design space the active learning scheme has to take into account that there are design being evaluated - otherwise valuable resources could be wasted on redundant evaluations. Integration with non-uniform design spaces poses further problems..

    No due date
    0% complete0 open0 closed

  • Adapting MLO to non-uniform cost design spaces.

    Utilize the fact that only certain parameter changes induce extra bit-stream generation (Non-uniform change induced cost). This phenomena exists due to caching, non-linear bit-stream and benchmark evaluation time. For example map and route problem grows exponentially with chip utilization rate. Work should conclude in a paper containing the following contributions: 1.Mathematical formulation of the new problem 2.New MLO flavour which takes into account non-uniform cost fitness functions. 3.Evaluation of the new flavour using 2-3 examples. Preferably Quadrature-method based and one or two news ones.

    Overdue by 13 year(s)
    •
    Due by March 1, 2013
    0% complete0 open0 closed

  • Multi-objective Optimization

    The idea is, to see how we should modify MLO to automatically optimize reconfigurable computing designs in terms of multiple objectives like power, area or speed. Ideally a number of different flavours of the algorithm are going to be developed and evaluated. The work should be summarized in a paper by March 2013. The following contributions are suggested: 1. Mathematical description of multi-objective optimization in reconfigurable applications 2.A new multi-objective flavour of MLO algorithm. 3.We evaluate the new multi-objective MLO using two reconfigurable computing applications. First one is the Quadrature-based application with extra power metric added.

    Overdue by 13 year(s)
    •
    Due by March 1, 2013
    0% complete0 open0 closed