Image
Over the past two years, MaX has made significant strides in developing and optimizing simulation software able to run on exascale supercomputing environments. These advancements are particularly beneficial for both academic and industrial users seeking to enhance materials simulations. By optimizing its software for emerging architectures and aligning with European HPC strategies, MaX can strengthen its leadership in computational materials science and ensure long-term relevance and applicability of its mission within the EuroHPC ecosystem.
Quantum ESPRESSO - Version 7.4.1 is available for download and integrates essential bug fixes and enhancements. The openMP GPU offloaded version of the Quantum ESPRESSO release 7.4.1 (qe-7.4.1omp) is also available. It has been tested with cce15 Cray compiler on the LUMI supercomputer equipped with AMD MI250X GPUs. Most of the main features of the PWscf code have been ported, while a few of them are still in the process to (eigensolver, forces, stress).
YAMBO - Yambo version 5.3 is now ready for production. This release brings significant advancements, particularly in the GPU support. Below, you will find a comprehensive overview of the updates and improvements introduced in this version, particularly in GPU porting, BSE, non-linear optics, and other compilation processes and upgrades. Read more.
FLEUR - The team is working to progressively increase the parallel efficiency of FLEUR. The effort concerns accelerating more kernels with OpenACC, improving the usage of parallel linear algebra for GPUs. FLEUR team is working closely with the FZJ HPC experts to optimize FLEUR performance on the oncoming JUPYTER supercomputer. The team has optimized the CMake build system and integrated it in a Spack recipe.
SIESTA - The team released SIESTA 5.2.2 incorporating a number of new functionalities. Among them, we highlight: a fix for an obnoxious memory leak due to the new mesh algorithm, some improvements on the basis set generation algorithm, and a few new features for the ELSI interface.
BigDFT - The code functionalities have been extended to platforms with SYCL programming environment. Production results can now be extracted on single nodes with systems of more than 10k atoms, employing the linear-scaling algorithm. The OpenCL acceleration of the BigDFT convolution has proven to be effective both on NVIDIA and AMD GPU. Fully-fledged containered versions of the application (both the runtime and a Software Development Kit) are now available in the intel oneapi25 environment.