Skip to main content

Performance Optimisation and Productivity 2

Periodic Reporting for period 1 - POP2 (Performance Optimisation and Productivity 2)

Reporting period: 2018-12-01 to 2020-05-31

High-performance computing is a fundamental tool for the progress of science and engineering and as such for economic competitiveness. The growing complexity of parallel computers leads to a situation where code owners and users are not aware of the detailed issues affecting the performance of their applications. The result is often an inefficient use of infrastructures. Even in the cases where a need to get further performance and efficiency is perceived, code developers may not have enough insight on its detailed causes to properly address the problem. This may lead to blind attempts to restructure codes in a way that may not be the most productive.

POP2 will continue to support researchers and industry in solving grand societal challenges by improving their productivity and the performance of their applications allowing them to reach new scales. Improved performance will ultimately lead to improved competitiveness for European companies and SMEs through access to CoE expertise and services.
The objective of POP2 is to continue and improve on the project activities of the first phase of POP, by operating a CoE on HPC with a special focus on very large scale. POP2 promotes best practices in performance analysis and optimization, helping developers and users of parallel applications to understand the performance of their applications and identify ways to improve them. The primary customers are code developers and owners. POP2 services are also available to code users, research infrastructure and service centres. Services are offered for free to EU institutions helping a wide user community.

POP2 aims to perform 180 services over a 3-year period. Activities are specifically oriented to improve performance tools to reduce the effort and cost of performing the assessments (ease of installation, portability, coverage of different platforms) rather than performing new academic research.
In the first half of the project, the team maintained the procedures required to manage the CoE, established in the first phase.

The business development component of the project centred around identifying and attracting potential POP users with a particular focus SMEs and CoEs. 1000 leads were identified, 89 new POP users were created, including 15 SME users. A new promotional tool was launched called “POPCasts” - short video interviews with users and POP staff, available on the POP blog. Online training material was produced, useful to those interested in learning about POP methodology and tools. Work also commenced to identify strategy approaches to ensure the project´s sustainability.

The project team was invited to speak about the ROI of performance analysis at multiple international large conferences & workshops, highlighting the awareness and interest of the HPC community in this topic, and in the project´s achievements. Training materials and activities were organized to engage interested parties in using offered services. POP hosted one training event and cooperated in 10 additional training events, co-organized and co-sponsored between POP partners with other EU projects or organizations (mostly PRACE (Advanced) Training Centres and VI-HPS). Communication efforts were maintained through publications: POP newsletter, 58 blog articles; four scientific papers.

At the end of the first period, there were 115 performance assessment (PA) services (detailed audits of code performance), with 91 studies originating since the start of POP2. 54 PAs were completed and 34 were in progress. Results of POP PA services are promising providing useful suggestions to clients for performance improvements. A total of 14 Proof-of-Concept (PoC) activities were conducted; 4 were completed by the end of the period, with 10 activities in progress. These include elements of code development and optimisation techniques.

Collaboration with other CoEs is a key aspect of POP2. In terms of large scale executions, an initial campaign audited all 12 codes of the ChEESE CoE. Flagship codes were separately audited for CompBioMed, ESiWACE and other HPC CoEs, multiple codes were analysed at joint workshops with EoCoE and E-CAM CoEs. Nine codes were audited executing on more than 5000 cores, with the largest at 309,696 cores, as part of readiness assessments for forthcoming exascale computer systems.
POP2 continues to analyse service costs and how to improve customer satisfaction. To obtain customer feedback, POP customers are sent questionnaires after receiving the report for the service provided. By the end of the first period, 54 questionnaires were sent on the services and. 39 customer feedbacks were collected. More than 94% of customers are either very satisfied or satisfied with the service. Over half of the customers wished to continue their collaboration with POP by requesting a PoC service.

A co-design repository was created to build a database of performance metrics characterizing applications such as kernels representative of fundamental behaviors identified in real applications. In the first half, 10 kernels were published on the site, with 6 patterns and 6 best-practices gathered from previous POP reports. The site includes descriptions of programming languages, programming models, disciplines, and algorithms.
The POP methodology is being extended to support hybrid scenarios as well as to consider the efficiencies of I/O and vectorization. Tool development efforts included support for POP specific use cases, release of new features in the analysis tools, and improvements towards better usability of tools.
POP2 continues to focus on a transversal (horizontal) service to offer highly specialized expertise and know-how to all other CoEs facilitating cross-fertilisation between them and other sectors. This could lead to wider access to codes, including specific and targeted measures for industry & SMEs. Collaborations with other CoEs and Projects will help to reach potential users for services and promote support to the governance of HPC Infrastructures. Centres could adopt the POP methodology, helping it embed in the wider HPC community, leveraging the use of existing infrastructures.

All activities of the project directly aim at increasing the adoption of best HPC practices in industrial and academic customers in all application domains. By applying these in real use cases and through the project´s targeted training activities, this directly contributes to developing the skills and human capital resources of customers.

Based on the services, application developers will make better-informed decisions to drive their work in a more productive direction. Infrastructure operators will ensure that their systems are provisioned in the most efficient way to maximise their Return on Investment. End-users (both academic and industrial) will use time and resources more efficiently.

POP2 continues to extend the successful POP Webinars series as a mechanism to transfer technologies, methodologies and engage with customers and the broader community.
POP2 Kick off meeting - Barcelona, Dec.2019
POP2 poster presentation at EuroHPC Summit Week 2019 - Poznan, May 2019
“POP analysis methodology”, University of Queensland, Sep 16-18, 2019