Modernizing air quality forecasting tools
PREV’AIR is the French national system for air quality forecasting managed by INERIS (French public research body under the aegis of the ministry of environment). This is the result of the work led by a consortium integrating also Météo France, CNRS (French National Center for Scientific Research), and LCSQA (reference laboratory for air quality monitoring). This platform relies on the results of numerical simulations and near real time pollutant observations to predict and map concentrations of regulated air pollutants.
After nearly 15 years of production, this platform has crossed in 2017 a major technological threshold that reinforces its quality and reliability. Atos’ Center for Excellence (CoE) supported INERIS for this move. The project stretched over a year and was led by Frédérik Meleux (INERIS) in collaboration with Météo-France. “The CoE, by its support in carrying and optimizing our workflow on Météo-France’s supercomputer, contributed to set up our new high resolution production. The work undertaken led to new perspectives like air quality forecasts at higher resolution and more service offers to our users” said Frédérik Meleux.
The CoE migrated INERIS’ workflow from an old generation cluster to the new Météo-France cluster. The old workflow was mainly based on several hundred scripts launched at fixed times. The CoE builts a new workflow integrating a full chain of dependent scripts launched at once. Though the new resolution is higher, this evolution reduced overall time to solution.
Code modernization with the French HPC community
GENCI is the French entity in charge of developing numerical simulation usage so as to improve French research and industry competitiveness. GENCI initiated in 2016 a technology watch project for assisting French HPC developers to move to new architectures by evaluating the productivity gain and the portability effort in many domains (weather forecast, CFD, high energy physics, chemistry…).
For this purpose, Atos delivered one of the very first Bull Sequana X1000 systems at the end of 2016 in Montpellier (CINES). The system is composed of 48 compute nodes equipped with Intel Knight Landing processors (68 cores).
Atos’ Center for Excellence was involved in the profiling and optimizations for enabling each application to the Intel KNL architecture. The obtained performance was published in many workshops and conferences, including the Intel Xeon Phi User Group (IXPUG).
Setting up Ivory Coast’s first national HPC center
Ivory Coast is the first country in West Africa to set up a national supercomputing centre with the purpose of sharing computing systems between universities and industrials. Atos’ Center for Excellence supports the Ivory Coast Ministry in their ambitious project with two concurrent actions.
- By delivering organizational recommendations for setting up an efficient operational structure and a balanced decisional system for sharing computing resources between national research communities.
- The second mission targets application knowledge transfer to the national research ecosystem, so as to push national excellence to a level recognized by the HPC community. Many actions are scheduled such as regular trainings, workshops, and creating HPC academic training.
To achieve this ambitious goal, the first CoE satellite in Africa was created within the national center building, so as to give the highest customized support to the Ivory Coast community.
Code optimisation for SKA-France
The SKA (Square Kilometer Array) will be the largest radio telescope ever built and will produce science that changes our understanding of the universe. The SKA will be collocated in Australia and in Africa. The project involves 100 organisations across about 20 countries. SKA-France is a national coordination of industrial, technical and scientific activities preparatory to the SKA project in France. SKA-France’s coordination work to optimise new algorithms for radioastronomy through collaboration between researchers and HPC companies is producing interesting results. For the first experiments, Atos’s CoE worked on the calibration and imaging code “DDFacet” by Cyril Tasse (OBSPM) with two different compiler suites: GNU and Intel. The aim was to dive into the DDFacet software stack in order to understand the different processing phases and extract their respective part from the total execution time. These experiments have highlighted potential improvements, which will be investigated through a tighter collaboration with developers.
Porting SEISCOPE to Intel Knights Landing
SEISCOPE is a consortium managed by the 3 French public laboratories LJK, Geoazur and ISTERRE, and sponsored by 9 Oil & Gas companies working together in quantitative seismic imaging, with the aim to reap the outcome of their common R&D endeavor to enhance their operations.
As part of SEISCOPE, an efficient 3D finite-difference time-domain modelling and frequency domain inversion code of Full Waveform Inversion called GeoInv3D is developed. GeoInv3D is memory bounded and is expected to benefit from Intel High Memory Bandwidth MCDRAM that equips the last generation of Intel® Xeon Phi™ Knights Landing (KNL) processors.
Atos’s Center for Excellence presents results obtained on an Intel® Xeon Phi™ processor (7210).
GPU accelerated implementation of NCI calculations using promolecular density
A scientific paper produced by Atos’s CoE and the University of Reims Champagne-Ardenne, a long-time partner and customer of Atos in HPC.
The NCI approach is a modern tool to reveal chemical noncovalent interactions. It is particularly attractive to describe ligand–protein binding. A custom implementation for NCI using promolecular density is presented. It is designed to leverage the computational power of NVIDIA graphics processing unit (GPU) accelerators through the CUDA programming model. The code performances of three versions are examined on a test set of 144 systems. NCI calculations are particularly well suited to the GPU architecture, which reduces drastically the computational time. On a single compute node, the dual-GPU version leads to a 39-fold improvement for the biggest instance compared to the optimal OpenMP parallel run (C code, icc compiler) with 16 CPU cores. Energy consumption measurements carried out on both CPU and GPU NCI tests show that the GPU approach provides substantial energy savings.