|
Dear Readers,
Welcome to our MaX April 2020 Newsletter. Our warmest wishes for these challenging times! In the following we update you on some of our ongoing activities, and start by offering some stimulating #StayHome opportunities available through the MaX website.
Our novel, capacity building materials include video lectures from the recent Yambo and Fleur schools as well as lectures recorded in the occasion of the school on High-Performance & High-Throughput Materials Simulations using Quantum ESPRESSO and AiiDA and the training material of the 3 weeks electronic structure course held by S. de Gironcoli: from blackboard to source code. Hands-on Tutorials on the MaX flagship codes can be followed online and you may also wish to visit the online school on Wannier90 that features the MaX codes.
For an introduction to first-principles electronic structure calculations, please see the online course by S. Cottenier. An introduction to Density Functional Theory will also be offered by Nicola Marzari with some live online seminars, based on MaX open-source codes such as Quantum Espresso, and on our Materials Cloud platform and tools.
A series of webinars for more advanced users and developers of our codes will be organized in the coming weeks: see the first dates in the news below. In these webinars we will highlight how the codes are prepared to run on the new HPC architectures, which are often accelerated.
Unfortunately, we had to postpone some events and schools due to the COVID-19 outbreak, including the Advanced School on Quantum Transport, the Digital Learning for Electronic Structure events and the school on First Principle Simulation with SIESTA. We will inform you about new plans as soon as possible.
Last but not least: if you need any support on MaX flagship code usage, the fora and mailing lists of the MaX flagship codes and the MaX help-desk are always active.
Please enjoy our insights below, do stay tuned, and most importantly, stay safe!
|
|
|
|
|
|
|
|
Save the dates for the first three MaX upcoming webinars on its codes in the next months of 2020: Quantum ESPRESSO on May 13, AiiDA on May 27, and Yambo on June 17.
|
|
DFT and GW towards the exascale: porting of the MaX codes on GPUs
MaX developers are working to support the European HPC community in view of the forthcoming supercomputing architectures. As of today, a number of MaX codes, as well as core libraries have been released production-ready with GPU-support on heterogeneous architectures. This result, obtained through a deep refactoring of the main codes, is key to better exploit the potential offered by the future pre-exascale and exascale EuroHPC machines. The porting on GPUs has been done by adopting different GPU-aware programming models and libraries: CUDA-Fortran for Quantum ESPRESSO, Yambo and FLEUR, CUDA and OpenCL for BigDFT, CUDA and ROCm for SIRIUS and the DBCSR, COSMA and SpFFT libs. Noticeably, we have addressed the possibility of delivering GPU acceleration as a library feature, like for the case of PSolver and the forthcoming libconv (BigDFT) and DBCSR-SpFFT (CP2K, CSCS).
|
|
|
Co-design in Materials Science Towards Exascale
Within the MaX co-design activity, Quantum  ESPRESSO, one of the MaX flagship codes, has been used as a demonstrator for cutting edge new HPC technology (ARM + NVIDIA GPUs). In fact, at SC19 in Denver, NVIDIA showcased Quantum ESPRESSO running "live" on an ARM TX2+V100 system, comparing "live" performance with the ARM TX2 host alone. The exhibition was supported by a stand alone pod (in the NVIDIA booth) with an expert available all the time to run the demo with the many visitors.
read more >>
|
|
|
Machine Learning: Improved predictions for time-to-solution of material science simulations
The accurate prediction of the time-to-solution required by massively parallel scientific codes is a key goal in HPC. It allows scientists to better program and allocate their computational tasks, and benefits environmental sustainability since energy waste due to sub-optimal execution parameters could be easily detected.
However, this predictions are especially challenging for hybrid and complex architectures. Using machine learning techniques for DFT-based material science codes marks an important step in reducing such complexities. The results by Pittino et al, presented at PASC 19, show how accurate predictions obtained with machine learning approaches can outperform parametrized analytical performance models made by domain experts.
read more >>
|
|
|
|
Computational School on Electronic Excitations in Novel Materials using Yambo Code
|
|
Hackathon on Domain-Specific Libraries for Materials Modelling
|
|
|
|
|
|
|
