The public HPC4E deliverables are available in this section.
D2.1 Report of the Implementation of computing Kernels in accelerator based architectures
The selected computing kernels from WP3, WP4, WP5 and WP6 will be optimized for architectures based on accelerators. The goal will be to optimize the performance but keeping a high degree of portability. The main target architecture platforms are those based on Xeon Phi and NVIDIA GPUs, but other platforms based on embedded processors will be also analysed during the second year.
D2.2 Report of the Implementation of computing Kernels in symmetric multicore-based architectures
The selected computing kernels from WP3, WP4, WP5 and WP6 will be ported to architectures based on symmetric multicore processors with NUMA memory. The goal will be to optimize the performance. The main target architectures will be Intel, AMD and SGI, but also new platforms based on ARM processors will be analysed. In this task the key point will be the load balancing and data placement. We will analyse new scheduling algorithms able to improve locality.
D3.2 Scalable Sparse Linear Solver
This report summarizes the advances made in the application of scalable linear equation solvers to the problems of interest of the present project. We have investigated hierarchical preconditioned Krylov solvers, sparse multifrontal strategies and hybrid strategies. Performance measurements were made on current high performance computers available in Brazil, Europe and USA, demonstrating the collaborative aspect of the present project.
D4.1 Performance of ABL-CFD models under HPC
This deliverable (report) deals with microscale models, typically implemented on computational fluid dynamic (CFD) solvers. In particular it focus on the CFD Alya code, that will be used in WP4 for downscaling strategies (Tasks 4.1 and 4.2) and forecast (Task 4.3). In order to define dynamical downscaling (D4.2), statistical downscaling (D4.3) and forecast (D4.6) strategies it is necessary to assess first the model performance on HPC and the numerical error due to the sensitivity of the result to the model set-up. This is particularly important in complex terrain where grid dependencies can be significant.
D5.1 Development of detailed, reduced and tabulated chemical kinetic schemes for biomass combustion
This is the WP5 Deliverable D5.1 associated to the validation and assessment of chemical kinetic mechanisms for biomass-derived gaseous fuels. The open-source code Cantera has been tested and validated against well-established solvers (PREMIX and COSILAB), and then used to evaluate the effects of fuel composition on the chemical structure of laminar flames. The next step was to generate thermochemical databases using the laminar premixed flames already obtained and integrate the source terms by a beta-PDF to take into account the effects of turbulence.
D6.3 Website deploying a suite of geophysical tests for wave propagation problems on extreme scale machines
We have developed the "HPC4E Seismic Test Suite", a collection of four 3D models and sixteen associated tests that can be downloaded freely at the project's website. The models include simple cases that can be used in the development stage of any geophysical imaging practitioner (developer, tester...) as well as extremely large cases that can only be solved in a reasonable time using ExaFLOPS supercomputers. The models are generated to the required size be means of a Matlab/Octave script and hence can be used by users of any OS or computing platform. The tests can be used to benchmark and compare the capabilities of different and innovative seimic modelling approaches, hence simplifying the task of assessing the algorithmic and computational advantages that they pose.
D7.1 Dissemination Plan
This document defines the dissemination objectives for the HPC4E project, as well as the different targets for all its activities, the dissemination tools, the interaction with similar projects, its activities to be done during the project, and the policy used to disseminate the results.