External:Annual Reports

Milestones, Outcomes, and Metrics

Year 1:

1. Design the telescopic approach for anisotropic parallel mesh generation and MRTS system based on current and emerging low-level multithreading systems and communication substrates.

1.A Design Telescopic Approach for Anisotropic Parallel Mesh Generation [Nikos/David_M/David_D]

Action Items:

i) Choose 2 to 3 geometries pertinent to this project and generate surface mesh, normals, flag couple of features that might need further processing [David_D/David_M]

   -- The geometries should expose the benefits of AFLR3 and adaptation through MOESS's 
       error-based metrics vs.  isotropic meshing
      -- Simple aerospace geometry described by surface triangulation, focus on parallel 
          volume mesh generation [David_D, Due by Jan. 2016]
      -- Simple ideal benchmark geometry described by both surface triangulation and analytic 
          means, focus on parallel volume mesh generation and adaptation using error-based 
          metrics [David_D/David_M, Due by Jan. 2016]
      -- NASA/LaRC Test cases from 1 and 2 like  and  
          described by surface triangulation, normals, some features (flaged), focus on 
          overheads for scalable geometry, jump-start parallel volume mesh generation 
          [Nikos/NASA-LaRC, Due by Dec. 2015]

ii) Identify AFLR3 operations for fine-grain concurrency for optimistic layer [David_M/Nikos]

  David_M: AFLR3 is under re-organization. It is expected to be completed by March 2016. 
                  Goal explore concurrency for AFT-based methods at fine-grain level. 
  David_M: Explore the triple (Mesh, Solver, Refinement) paradigm
  Nikos: Can look into this, if I get the AFLR3 src code and pertinent papers to leverage 
              understanding from Delaunay-Based find-grain level.
  Nikos/David_D: consider the evaluation of an  alternative approach based on Delaunay 
               based methods in parametric space

iii) PDR.AFLR3 implementation:

 -- Given fixed geometries (from 1.A.i) and target 32-core nodes  study impact of costs like: 
     pre-process geometry, data-decomposition, roll-backs to meet MOESS's metrics, etc.. 
     [David_M/Nikos]
     Nikos: I can run AFLR3 and perform these studies

 -- Integration of PDR with AFLR3 [David_M/Nikos]

iv) Rough estimates of overheads to jump-start process [Nikos: given he has (i) and (iii)]

1.B Design MRTS for current and emerging parallel architectures [Nikos/Andriy]

i) Install current version of PREMA on ITS machines [Nikos/Andriy]

  -- provide accounts to MIT/MSU

ii) Explore power-aware options [Nikos]

  -- profile existing codes, lessons learned
  -- explore use of OpenMP for PDR.AFLR3 to leverage power-aware RTS (from UK)
  -- explore C++ 11 (and above) concurrency constructs for PO.AFLR3 and their interactions 
      with power-aware RTS (from UK)

Outcome: A research report describing the overall design of a multi-layer framework for extreme-scale parallel mesh generation. Included in the report will be specific test cases and metrics for assessing the weak scaling of the mesh-generation framework on NASA capability-class hardware.

2. Design the PDR layer of the multi-layer framework for parallel mesh generation and investigate AFLR modifications in order to be integrated with the PDR framework and issues related to the scalability of querying the geometry.

2.A Design PDR.AFLR3 layer [Nikos/David_M]

-- Proposed PDR API [Nikos]

  -- PDR needs to pass/receive to/from AFLR3:
              -- a list of pointers of  "bad" elements w.r.t  error-based metrics and 
                  potential obstacles to explore concurrency at the current level of refinement 
              -- number of obstacles

  -- AFLR3 needs to modify:
              -- elements: introduce a new counter
              -- communicate relevant events to PDR. eg. remove "bad" elements. 

-- Prepare an abstract [Nikos/David_M, Due by Feb. 2016]

2.B Investigate issues related to the scalability of querying the geometry for preparing the input for PDR.AFLR3 [David_M/Nikos/David_D]

Outcome: A presentation and paper at an open technological conference or similar forum describing the implementation approach.

3. Design of the MOESS API.

Define API between sequential AFLR3 and MIT's MOESS system [David_D/David_M]

Outcome: A research report describing the MOESS API. Included in the report will be specific test cases and metrics for assessing the weak scaling of the MOESS framework on NASA capability-class hardware.

4. Preliminary investigation of scalable optimization algorithms for solving the MOESS optimization problem. [David_D]

Outcome: A presentation and paper at an open technological conference or similar forum investigating MOESS optimization algorithms for extreme scale computations. The metrics developed in Milestone (3) will be used to assess the performance of the optimization algorithm.

Issues for next (Dec. 4) WebEx meeting:

Agenda:

I. Introduce Dr. A. Kot (NCSA/UIUC)

  -- Multi-layered Runtime System (Brief overview of RTS)
  

II. Input (use .tri format) to expose advantages of proposed approach and calibrate current vision

  -- Address short to medium term needs of NASA/LaRC, identify issues to calibrate current vision presented in the proposal
  -- Expose potential advantages of AFLR3 and MOESS, issues to be addressed in the future
  

III. Parallel Mesh Generation

  -- PDR: targets cores in a single node (Year 2) nodes within a single rack (Year 3)
   -- Issues: 
        Pros:required minimum (but non trivial) interaction between software pieces & groups
                 multiplier for speculative approach for cores within a single chip (initially node). 
        Cons: Scalability (small) due to local synchronization, startup overhead
           -- Use input geometries to identify I/O overheads due to geometry, use this is the mac budget to jump-start PDR
        Programming Paradigm: OpenMP (need modifications)
        Define an API between PDR and AFLR3

-- Optimistic fine-grain parallelization: targets cores within a single chip, shared memory

         A. direct use of a fully functional AFLR3 (new re-organized code)
         B. indirect use of parallel Delaunay-based PODM approach in the parametric space
         C. investigate building from scratch new parallel implementation from a simpler AFT-code and build functionality as we go
         D. evaluate a bottom up approach starting from parallel triangulation, etc..   (Juliette) 
       Programming Paradigm: C++ with concurrency constructs, potential extensions of paradigm used in Jlab's QCD codes (challenging)

-- Constrained Parallel Mesh generation: targets racks or super-nodes

        Issues:
        Cons: asynchronous
        Pros: not straight forward generalization of 2D (split of a face is not unique)
    

Remaining Internal (ODU/MIT/MSU) Issues:

    i)  Interaction between MOESS and AFLR3 i.e., API (sequential code can be used as is at this point)
    ii  Interaction between PDR and AFLR3 (API is the goal, a closer look of the AFLR3 code at ODU 
        will help to   build an optimal API) and study separation of concerns between PDR and AFLR3     
    iii) Study overheads for pre-processing (Geometry, AFLR3 might require).
    vi) Set milestones and deliverables for semi-annual report
    vii) Software License and potential IP issues.