New algorithms for applications on European exascale supercomputers -

Expected Outcome

At the end of the action, the following will be accomplished:

• Contribution to the realisation of the EuroHPC overall and specific objectives
• Novel and outstanding ideas to address the exascale challenge in HPC applications will be identified and the potential impact on HPC applications substantiated by a proof-of-concept implementation
• Wider participation of researchers in the exascale challenge from different domains with so far little exposure to exascale technology in the European HPC ecosystem
• Prepared underrepresented research communities to participate in upcoming calls in the field of HPC algorithms and applications

Each third-party which received financial support in this action will deliver by the end of the project at least:

• A proof-of-concept implementation running on real hardware including meaningful benchmark data that demonstrates significant supremacy compared to existing solutions
• With respect to the base line provided in the application for financial support: a revised estimate of the reduction for time-to-solution of a list of applications and use cases, based on the best available benchmark data
• An impact asessment on the reduction of HPC resource consumption taking into account the present usage pattern and potential use cases of the algorithms
• Documentation of the new algorithms and their technical implementation

Objective

The action will provide financial support through an open and competitive call to small teams of researchers to develop novel algorithms for the upcoming European exascale supercomputers. Targeting the highest possible impact on HPC applications, the developed proofs of concepts will demonstrate great potential to solve currently untractable computational challenges or to substantially improve time-to-solution and energy-to-solution for important use cases.

Scope

Proposals are expected to define an outreach approach for identifying and attracting innovative ideas to efficiently exploit current and future European exascale HPC systems. A mechanism involving financial support to third parties should be defined, which will adequately stimulate such innovation potential of researchers and software engineers participating in the action. The focus of the action should aim at solutions to computational problems in an exascale context which are common in European flagship HPC applications, offering outstanding potential to improve their efficiency on European HPC systems and reduce time-to-solution for common use cases, possibly across scientific domains.

The proposal should define the process of selecting novel algorithms to be developed from TRL 0-1 to TRL 3-4 (proof-of-concept), for which financial support will be granted. The central selection criterion for granting financial support to third parties should be based on the expected impact for the imminent exascale era, measured in recovered compute time or reduced time-to-solution taking into account the expected usage pattern on European exascale systems, but also consider novel solutions to computational problems with high impact. Considering the complexity of developing, implementing and validating algorithms for new European hardware and exascale supercomputer architectures, typically the financial support will be in the order of EUR 50 000 – 200 000 per third-party project and should primarily cover personnel costs. At least 80% of the total grant should be allocated to financial support for these third parties.

Calls for proposals should be widely disseminated among the EU and Participating States of the EuroHPC JU. The proposal should clearly describe how relevant target groups are addressed in the different countries. In particular, scientific communities in relevant domains such as applied mathematics, theoretical physics and computer science should be specifically addressed and encouraged to apply.

Additionally, the consortium should

• provide relevant information to applicants for the preparation of their proposals, e. g. on system architecture, software and technical specifications of European exascale technology. Ideally, the consortium should also provide access to exascale technology for specific development purposes to selected projects of third parties as appropriate.
• perform a final review of the funded projects carried out by third parties, including a technical inspection and approval of the developed software, and provide an assessment on each algorithm and implementation. The final reports should include clear and comparable indicators for all projects on the progress made, taking into account the baseline, and the maturity of the implementation at the end of the project.
• collaborate with the national HPC Competence Centres, European Centres of Excellence for HPC Applications and the related coordination actions to organise additional support for third-party projects.
• provide support for coordination with other European projects such as the Advanced Pilots towards the European supercomputers, if relevant for a third-party project.

The JU considers that proposals requesting a contribution from the JU of up to EUR 5 million and a duration of 18 months would allow this specific challenge to be addressed appropriately. Nonetheless, this does not preclude submission and selection of proposals with another duration or requesting other amounts. Only one proposal will be selected.

Selection of third parties for financial support: Award criteria and evaluation procedure

The consortium is expected to describe the procedure and the criteria used to ensure appropriate organisations and projects will be selected, in agreement with the provisions of the Digital Europe programme.

When launching calls for proposals for the award of financial support, the consortium members may use their own procedure provided this procedure complies with the principles of proportionality, sound financial management, equal treatment and nondiscrimination.

The process should also aim at reducing administrative burden for third parties, while at the same time ensuring sound financial management. In particular, the application forms and procedures should be tailor-made and appropriate to the technical and managerial capacities of the targeted applicants and scope of the projects. Calls for proposals must remain open for at least two months.

Beneficiaries must ensure transparency with adequate publication of calls for proposals and prevent conflicts of interest throughout the entire award procedure. They will have to clearly demonstrate this in their application and report on it throughout the project.

The consortium must publish the outcome of the call(s) on their website, including a description of the selected projects, award dates, project durations, final recipients’ legal names and countries of establishment. The indicative timeline for this publication is within two months after the submission deadline of the call the proposal or equivalent date for open-end calls.

Financial support to third parties: Budget, duration and location

• The indicative amount of financial support per third party should be the equivalent of 2 FTEs for 12 months, based on the usual remuneration of the institution for staff in a relevant role. Financial support to one application must not exceed a total amount of EUR 200000.
• Applicants for financial support to third parties should be able to use simplified cost options and in particular lump sums.
• Activities carried out by third parties must take place in EU Member States or associated countries that are members of the EuroHPC JU.
• Only cost incurred during the grant implementation period can be considered eligible.

Types of organisations/third parties that may receive financial support:

In order to be eligible for financial support, the third parties must neither be affiliated entity(ies) of the beneficiaries nor associates nor contractors. Natural persons are not eligible.

Third-party projects receiving financial support should be subject to at least the following requirements:

• Identify and propose novel, forward-looking and potentially disruptive approaches to the solution of complex mathematical, numerical or data processing problems on current and future European exascale supercomputers.
• Explain the concept and design a fundamentally new and innovative algorithm - porting, reimplementation, incremental improvements or parallelization of an existing implementation are not sufficient.
• Present a sound theoretical concept, substantiated e.g. by peer-reviewed publications, with a credible and convincing plan to achieve a first proof-of-concept implementation which clearly demonstrates significantly superior performance compared to existing solutions and exploits the specific capabilities of exascale supercomputers.
• Solutions must demonstrate the potential to be integrated into important applications, addressing relevant use cases with a broad user base.
• Clearly describe the state of the art (baseline) of the concept and implementation which should be in line with the scope of the call regarding TRL.
• Provide a list of applications frequently used on HPC systems with typical use cases which could substantially benefit from the proposed solution including an estimate of the reduction of time-to-solution for the use cases.
• Algorithms using emerging technology such as quantum computers are only eligible if linked to HPC, for example, by exploiting hybrid quantum-classical exascale architectures.

Activities are expected start at TRL 0-1 and achieve TRL 3-4 by the end of the project – see General Annex B.

News flashes