From 4fcab263fefd67d294156895acac6e5d06df7ad9 Mon Sep 17 00:00:00 2001 From: Florian Faucher Date: Wed, 16 Oct 2024 14:51:40 +0200 Subject: [PATCH] update roadmap haven wp3 --- software/hawen/WP1/WP1.tex | 2 +- software/hawen/WP3/WP3.tex | 15 +++++++++++---- 2 files changed, 12 insertions(+), 5 deletions(-) diff --git a/software/hawen/WP1/WP1.tex b/software/hawen/WP1/WP1.tex index 9aa3002..f9db235 100644 --- a/software/hawen/WP1/WP1.tex +++ b/software/hawen/WP1/WP1.tex @@ -270,7 +270,7 @@ \subsection{12-Month Roadmap} \rowcolor{white} B10 - Scientific Productivity & Improve high-order polynomials operations for HDG discretization. \\ \rowcolor{numpexlightergray} B11 - Reproducibility and Replicability of Computation & Each benchmark will have a separate repository with the input files and associated results to reproduce the simulations. \\ -\rowcolor{white} B6 - Data Management & Provide the meshes used for the benchmark and the associated results in standard format such as vtk. \\ +\rowcolor{white} B6 - Data Management & Provide the meshes used for the benchmark and the associated results in standard format such as vtk and hdf5. \\ \rowcolor{numpexlightergray} B7 - Exascale Algorithms & Control the memory footprint. \\ \end{tabular} } diff --git a/software/hawen/WP3/WP3.tex b/software/hawen/WP3/WP3.tex index c430454..48cd2e7 100644 --- a/software/hawen/WP3/WP3.tex +++ b/software/hawen/WP3/WP3.tex @@ -188,10 +188,17 @@ \subsection{12-Month Roadmap} % \item \textbf{Results Retention:} Plans to maintain benchmark results in a publicly accessible repository with appropriate metadata and documentation, ensuring long-term usability. %\end{itemize} -We wish to improve input/output data management by including more options -to save the results of \hawen, for instance this can include the hdf5 format -which is not allowed currently. - +We wish to investigate the efficient of the linear algebra operations +when working with particularly high polynomial orders, as established +in \cref{sec:WP1:Hawen:software}. +We emphasize the cost on the memory footprint for the matrix factorization, +which is the main bottleneck in time-harmonic wave simulations using a +direct solver. +We will rely on the direct solver MUMPS to investigate the cost and benefit +from its latest development to reduce the computational cost. However, one +must find a good compromise between the reduction of the memory footprint +(e.g., with block low-rank compression or mixed precision), and the accuracy +of the solution. In~\cref{tab:WP3:Hawen:bottlenecks}, we briefly discuss the bottleneck roadmap associated to the software and relevant to the work package. \begin{table}[h!]