Periodic Reporting for period 1 - HYSCALA (Hybrid SCAlable sparse matrix Linear Algebra for industrial applications)
Période du rapport: 2018-05-01 au 2019-10-31
HYSCALA is a commercialization project for the linear solver library GaspiLS.
GaspiLS efficiently solves sparse linear systems of equations. Sparse linear systems typically appear after the discretization of partial differential equations (PDEs). That’s the quintessence of many modelling and/or simulation problems in engineering and scientific computing. Examples are fluid flow for aircraft, ship and car design.
GaspiLS relies on the programming model GASPI / GPI-2 for distributed memory applications. GASPI /GPI-2 shows a superior performance and scaling behavior compared to other state-of-the-art programming models. GaspiLS makes the benefits of the underlying programming model for parallel programming GASPI / GPI-2 accessible to a broad range of applications on a high abstraction level, which is easily adaptable also by legacy applications.
The uptake of GaspiLS in industry applications will improve their competitiveness due to a better scalability, performance and elasticity with respect to developments in current and future compute architectures. Thereby advancing European industries relying on HPC services.
Our objectives in the scope of HYSCALA are to understand the customers needs and build up a concrete value proposition. Our ultimate goal is to establish a market segment for GaspiLS.
GaspiLS efficiently solves sparse linear systems of equations. Sparse linear systems typically appear after the discretization of partial differential equations (PDEs). That’s the quintessence of many modelling and/or simulation problems in engineering and scientific computing. Examples are fluid flow for aircraft, ship and car design.
GaspiLS relies on the programming model GASPI / GPI-2 for distributed memory applications. GASPI /GPI-2 shows a superior performance and scaling behavior compared to other state-of-the-art programming models. GaspiLS makes the benefits of the underlying programming model for parallel programming GASPI / GPI-2 accessible to a broad range of applications on a high abstraction level, which is easily adaptable also by legacy applications.
The uptake of GaspiLS in industry applications will improve their competitiveness due to a better scalability, performance and elasticity with respect to developments in current and future compute architectures. Thereby advancing European industries relying on HPC services.
Our objectives in the scope of HYSCALA are to understand the customers needs and build up a concrete value proposition. Our ultimate goal is to establish a market segment for GaspiLS.
HYSCALA has explored HPC market segments and developed a value proposition based on GaspiLS for industrial applications. To reach the final goal, to sell commercial licenses for GaspiLS therein, we developed a deep understanding of the requirements on features and functionalities of diverse industrial clients in several market segments. In the scope of the project we have reached out to industrial customers with our web page, during trade shows, conferences and tutorials and established sustainable business contacts.
GaspiLS achieves its superior scalability and performance by striving for 100 percent resource utilization. GaspiLS delivers superior performance and scalability, which means:
• Time to solution can be practically reduced as much as required by adding extra resources.
• There is practically no limit on the modelling complexity and its enormous memory footprint, as memory is simply accumulated across nodes.
• There is no need for expensive resources. Instead of high-end fat nodes, use a bunch of cheap commodity hardware nodes.
• Optimal energy efficiency because hardware is used to capacity and never sits around being idle.
The purpose of this project has been to commercialize GaspiLS and to show that our parallel linear solver is available to industry, fostering European companies relying on HPC services.
• Time to solution can be practically reduced as much as required by adding extra resources.
• There is practically no limit on the modelling complexity and its enormous memory footprint, as memory is simply accumulated across nodes.
• There is no need for expensive resources. Instead of high-end fat nodes, use a bunch of cheap commodity hardware nodes.
• Optimal energy efficiency because hardware is used to capacity and never sits around being idle.
The purpose of this project has been to commercialize GaspiLS and to show that our parallel linear solver is available to industry, fostering European companies relying on HPC services.