Periodic Reporting for period 1 - DEEP-SEA (DEEP – SOFTWARE FOR EXASCALE ARCHITECTURES)
Periodo di rendicontazione: 2021-04-01 al 2022-09-30
DEEP-SEA (“DEEP – Software for Exascale Architectures”) will deliver the programming environment for future EU-exascale systems, adapting all levels of the software (SW) stack to support highly heterogeneous compute and memory configurations and to allow code optimization across existing and future architectures and systems.
The society at large will benefit from the increase in the scientific throughput of HPC systems, which enables new discoveries in scientific fields as important as climate research, medicine, or new sources of energy.
The overall objectives of the project are:
PObj-1: Co-design the software- and programming environment of the upcoming European Exascale systems.
PObj-2: Provide tools to map complex applications and non-uniform workflows onto heterogeneous and modular computer architectures.
PObj-3: Enhance the system software, programming paradigms, tools, and runtimes in order to extract the maximum performance from heterogeneous computer platforms and improve performance portability.
PObj-4: Improve the use and management of new memory technologies and the placement of data in compute devices with deep and heterogeneous memory hierarchies.
PObj-5: Release the DEEP-SEA software stack in production-ready quality to enable its use and exploitation in upcoming European Exascale systems.
The society at large will benefit from the increase in the scientific throughput of HPC systems, which enables new discoveries in scientific fields as important as climate research, medicine, or new sources of energy.
The overall objectives of the project are:
PObj-1: Co-design the software- and programming environment of the upcoming European Exascale systems.
PObj-2: Provide tools to map complex applications and non-uniform workflows onto heterogeneous and modular computer architectures.
PObj-3: Enhance the system software, programming paradigms, tools, and runtimes in order to extract the maximum performance from heterogeneous computer platforms and improve performance portability.
PObj-4: Improve the use and management of new memory technologies and the placement of data in compute devices with deep and heterogeneous memory hierarchies.
PObj-5: Release the DEEP-SEA software stack in production-ready quality to enable its use and exploitation in upcoming European Exascale systems.
Intensive co-design interaction has been established between all technical WPs, from WP1 to WP5.
-WP1: has analised the applications participating in the project, identified requirements, and provided them as co-design input to the rest of the project developers. A benchmarking strategy based on JUBE is in place. The DEEP system, used for SW development, has been updated both from hardware (with a uniform interconnect) as from the software perspective (e.g. with a new OS). Deliverables D1.1 D1.2 and D1.3 have been submitted.
- WP2: further develops tools to measure and model the performance of applications, to map them onto the HW resources, and to monitor both system and applications. The basis are well established SW packages, on which enhancements have been done to better support accelerators (e.g. GPGPUs and FPGAs) and heterogeneous memory systems (e.g. DDR, High Bandwidth Memory (HBM), Non-Volatie Memory (NVM)). The first release of all the WP2 tools has been done in M18. Two deliverables have been submitted: D2.1 and D2.2.
- WP3: The first version of the overall DEEP-SEA software stack has been released in M8, as part of D3.1 which describes the software packages developed within the DEEP-SEA project and the Optimisation Cycles (OCs) that combine them in order to achieve specific improvements. Development work has advanced, with particular emphasis on malleability. The infrastructure for Continuous Integration (CI) is in deployed on the DEEP system, based on Gitlab, Jacamar, EasyBuild, and JUBE.
-WP4: Programming model at node level focuses on the support of heterogenous compute and memory devices, on achieving a good load balance, and on malleability. Support different hardware accelerators (GPUs, FPGAs, etc.) has been improved in the latest OmpSs-2 release and the high-level programming models DaCe and NabLab. Mechanisms and policies for dynamic load balancing (DLB) were implemented in OmpSs, OpenMP and the DLB library. Memory management support was improved across the software stack. D4.1 was submitted.
-WP5: Composability and interoperability between programming models has been improved to enable application developers to combine different approaches without having to fully rewrite their codes (D5.1). Support for malleability is implemented in various layers of the DEEP-SEA software stack and a prototype malleability plugin has been developed (D5.2). MSA-awareness for collective operations has been enhanced in Open MPI and ParaStation MPI. For resiliency, the trace-based Simgrid simulation framework was enhanced with a consistent fault model.
-WP6: The project's website and social media channels are in place and kept up to date. These are completed by virtual conferences, events and workshops, for which numerous presentations and publications have been produced. Joint communication with other EuroHPC projects has been enhanced. The team is involved in standardisation bodies in the fields of MPI, OpenMP, and PMIx. Innovation carried in and out of the project are monitored and documented. A Scientific Advisory Board (SAB) has been established. Training activities focus on knowledge transfer within the project and with fellow SEA and EuroHPC-01-2019 projects, with a training and education programme established an a Web repository for training material created. D6.1 D6.2 and D6.3 have been submitted.
-WP7: The Project Management Team (PMT) continuously monitors the status of the work to guarantee that it progresses according to the Description of Action (DoA), and to identify any risk that could endanger its execution. The management bodies are established, as well as the legal frame that regulates the interactions between the Partners: the Consortium Agreement (CA). Communication tools for intra-project communication are in place. For quality control, each deliverable follows an internal review process before it is submitted. For financial control each Partner sends their Quarterly Financial Reports (QFR) to the PMT to detail all expenses. The WP quarterly reports contain a summary of the activities performed with the reported PMs. Collaboration between the 10 complementary grants is ongoing under the Collaboration Agreement signed by all 62 Parties in M12. D7.1 to D7.5 have been submitted.
-WP1: has analised the applications participating in the project, identified requirements, and provided them as co-design input to the rest of the project developers. A benchmarking strategy based on JUBE is in place. The DEEP system, used for SW development, has been updated both from hardware (with a uniform interconnect) as from the software perspective (e.g. with a new OS). Deliverables D1.1 D1.2 and D1.3 have been submitted.
- WP2: further develops tools to measure and model the performance of applications, to map them onto the HW resources, and to monitor both system and applications. The basis are well established SW packages, on which enhancements have been done to better support accelerators (e.g. GPGPUs and FPGAs) and heterogeneous memory systems (e.g. DDR, High Bandwidth Memory (HBM), Non-Volatie Memory (NVM)). The first release of all the WP2 tools has been done in M18. Two deliverables have been submitted: D2.1 and D2.2.
- WP3: The first version of the overall DEEP-SEA software stack has been released in M8, as part of D3.1 which describes the software packages developed within the DEEP-SEA project and the Optimisation Cycles (OCs) that combine them in order to achieve specific improvements. Development work has advanced, with particular emphasis on malleability. The infrastructure for Continuous Integration (CI) is in deployed on the DEEP system, based on Gitlab, Jacamar, EasyBuild, and JUBE.
-WP4: Programming model at node level focuses on the support of heterogenous compute and memory devices, on achieving a good load balance, and on malleability. Support different hardware accelerators (GPUs, FPGAs, etc.) has been improved in the latest OmpSs-2 release and the high-level programming models DaCe and NabLab. Mechanisms and policies for dynamic load balancing (DLB) were implemented in OmpSs, OpenMP and the DLB library. Memory management support was improved across the software stack. D4.1 was submitted.
-WP5: Composability and interoperability between programming models has been improved to enable application developers to combine different approaches without having to fully rewrite their codes (D5.1). Support for malleability is implemented in various layers of the DEEP-SEA software stack and a prototype malleability plugin has been developed (D5.2). MSA-awareness for collective operations has been enhanced in Open MPI and ParaStation MPI. For resiliency, the trace-based Simgrid simulation framework was enhanced with a consistent fault model.
-WP6: The project's website and social media channels are in place and kept up to date. These are completed by virtual conferences, events and workshops, for which numerous presentations and publications have been produced. Joint communication with other EuroHPC projects has been enhanced. The team is involved in standardisation bodies in the fields of MPI, OpenMP, and PMIx. Innovation carried in and out of the project are monitored and documented. A Scientific Advisory Board (SAB) has been established. Training activities focus on knowledge transfer within the project and with fellow SEA and EuroHPC-01-2019 projects, with a training and education programme established an a Web repository for training material created. D6.1 D6.2 and D6.3 have been submitted.
-WP7: The Project Management Team (PMT) continuously monitors the status of the work to guarantee that it progresses according to the Description of Action (DoA), and to identify any risk that could endanger its execution. The management bodies are established, as well as the legal frame that regulates the interactions between the Partners: the Consortium Agreement (CA). Communication tools for intra-project communication are in place. For quality control, each deliverable follows an internal review process before it is submitted. For financial control each Partner sends their Quarterly Financial Reports (QFR) to the PMT to detail all expenses. The WP quarterly reports contain a summary of the activities performed with the reported PMs. Collaboration between the 10 complementary grants is ongoing under the Collaboration Agreement signed by all 62 Parties in M12. D7.1 to D7.5 have been submitted.
DEEP-SEA contributes to the EuroHPC objectives by developing, enhancing and optimising of key system software components, programming environment, tools, and applications, and by making these more mature. The project thus creates a software stack for future European Exascale systems with full support for heterogeneous compute and memory system configurations, and for the Modular Supercomputing Architecture (MSA) in particular. The project’s developments contribute to and are well aligned with the ETP4HPC Strategic Research Agenda (SRA).
Since the beginning of the project progress was made towards the major project objectives. The groundwork to create a software stack that matches the innate requirements of the co-design applications is therefore completed, timely reaching the milestone MS3 in M8. Implementation work on the individual components is progressing, exploiting synergies between the various elements by combining their functionalities whenever doing it provides added value. The first roll-out of the DEEP-SEA software on the DEEP system has been performed in M12 (reaching milestone MS4), and a first release of the WP2 tools was ready by M18. The CI-infrastructure is now in operation, and it was an important factor to reach this goal.
The project will deliver key elements of the software environment for future Exascale systems and provide vital contributions to get Europe ready for Exascale. DEEP-SEA, together with the IO-SEA and RED-SEA projects, is centrally positioned in the EuroHPC Exascale roadmap: its hardware-agnostic developments will benefit the European Processor Initiative (EPI), as well as the upcoming pilot projects EUPEX and EUPilot. Together with all these and further EuroHPC R&D projects, DEEP-SEA is paving the way towards the first Exascale systems in Europe.
Since the beginning of the project progress was made towards the major project objectives. The groundwork to create a software stack that matches the innate requirements of the co-design applications is therefore completed, timely reaching the milestone MS3 in M8. Implementation work on the individual components is progressing, exploiting synergies between the various elements by combining their functionalities whenever doing it provides added value. The first roll-out of the DEEP-SEA software on the DEEP system has been performed in M12 (reaching milestone MS4), and a first release of the WP2 tools was ready by M18. The CI-infrastructure is now in operation, and it was an important factor to reach this goal.
The project will deliver key elements of the software environment for future Exascale systems and provide vital contributions to get Europe ready for Exascale. DEEP-SEA, together with the IO-SEA and RED-SEA projects, is centrally positioned in the EuroHPC Exascale roadmap: its hardware-agnostic developments will benefit the European Processor Initiative (EPI), as well as the upcoming pilot projects EUPEX and EUPilot. Together with all these and further EuroHPC R&D projects, DEEP-SEA is paving the way towards the first Exascale systems in Europe.