WP7: Showroom, Benchmarking, and Co-Design Coordination

WP7 integrates, tests, and benchmarks methods and software from all WPs, coordinates co-design activities with ExaDIP, and delivers the showroom, CI/CD infrastructure, and training materials.

Recent Highlights (2024-2025)

CI/CD Automation

CI/CD pipelines validating new methods and software deliveries

24h Turnaround

Automated code → EuroHPC runs in < 24h

Showroom Updated

Showroom content updated with meshing, solver, and ML workflows

Containers Published

Containers and packages published (Spack, Guix, Docker, Apptainer)

Training Available

Training modules on CI/CD and benchmarking best practices

1. Objectives

WP7 provides the infrastructure and coordination for:

  • Software testing from simple to advanced benchmarking

  • Verification of exascale capabilities and handling of identified challenges (B1-B13)

  • Delivery of software packages with CI/CD framework

  • Coordination of co-design activities within Exa-MA and with ExaDIP

  • Showroom of Exa-MA results and training material

2. Integration Framework

Diagram

3. Key Tasks

T7.1: Testing and Benchmarking Environment

  • Identify each relevant demonstrator to define a validation laboratory

  • Three types of demonstrators:

    • Level 1: Covers one to two WPs (e.g., AMR techniques)

    • Level 2: Covers three to four WPs

    • Level 3: Potentially covers all WPs

  • Some demonstrators retained by PC5 benefit from "mini apps" development and support

T7.2: Co-design Activities Coordination

  • Integration process based on all demonstrators

  • Broken down into non-regression, verification and validation processes

  • Each demonstrator must deliver current test process to guarantee integrity

  • Guarantee non-regression while evaluating new features

  • Record cases as new non-regression, verification or validation tests

  • Add new tests based solely on functionality tested

  • Enrich original base if necessary to better evaluate contribution of new methods

T7.3: Showroom Coordination

  • Describe obtained results in unique format

  • Present in dedicated web page

  • Compare results with initial objectives in terms of performance

  • Illustrate with figure of merit and raw data (clock time, resources used, computer)

  • Systematically compare with initial performance of the demonstrator

T7.4: Training

  • Produce training material on exascale toolboxes and mini-apps

  • Document best practices for exascale computing

  • Create reproducible examples and tutorials

4. Leads & Partners

Lead Institution

CEA (initially UNISTRA)

Co-Leaders

UNISTRA, Inria, École Polytechnique, Sorbonne Université

Duration

Months 1-60+

5. Addressed Exascale Bottlenecks

WP7 addresses ALL bottlenecks (B1-B13) through:

  • Benchmarking to verify exascale capabilities

  • Handling of all identified challenges

  • Non-regression, verification and validation testing

  • CI/CD integration with ExaDIP

Bottlenecks: B1 (Energy efficiency), B2 (Interconnect), B3 (Memory), B4 (System software), B5 (Programming systems), B6 (Data Management), B7 (Exascale Algorithms), B8 (Discovery/design/decision), B9 (Resilience/robustness/accuracy), B10 (Scientific productivity), B11 (Reproducibility/replicability), B12 (Pre/Post processing), B13 (Integrate uncertainties)

6. Deliverables

ID Title Due Dates

D7.1-B

Benchmarking analysis report (bottlenecks and breakthroughs)

M12, M24, M36, M48, M60

D7.2-MR

Co-design report

M48, M60

D7.3-TR

Training material

M12, M24, M36, M48, M60

D7.4-MR

Activity reports (included in annual report D0.2-TR)

M12, M24, M36, M48, M60

7. Collaborations

  • All WPs (WP1-WP6): Benchmarking and testing of all developments

  • WP0: Follows project management plan

  • NumPEx Projects: ExaSoft, ExaDoST, ExaAToW, ExaDIP (CI/CD and co-design)