The RECIPE approach to challenges in deeply heterogeneous high performance systems

Abstract

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

TEXTAROSSA: Towards EXtreme scale Technologies and Accelerators for euROhpc hw/Sw Supercomputing Applications for exascale

“…Energy efficiency can then be guaranteed only if, on the SW side, we can rely on a suitable (hierarchical) resource management framework. Although the state-of-theart already includes some solutions, recent projects, like MANGO [28] and RECIPE [29], show that optimized solutions need to take into account the platform-specific characteristics and control knobs to profile the applications at design-time and monitor them at runtime [30], [31], enabling more accurate resource mappings [32], [33]. Furthermore, an integration of the resource manager with the programming model allows dynamically tuning the numerical accuracy (precision) of the tasks, with respect to the actual application requirements and power/energy constraints.…”

The Italian research on HPC key technologies across EuroHPC

“…The cluster prototype is currently maintained for demonstration and will be instrumental to the activities of the CINI Lab. As a further development, the RECIPE project [2] is extending the architecturerelated investigation of MANGO by developing a multi-FPGA node with additional features of relevance for HPC, particularly checkpoint/restart functions and remote direct accelerator memory access, to be used for example in MPI one-sided communication primitives with heterogeneous workloads.…”

“…While several parallel programming models like OmpSs, Legion, Dharma are being mapped on OCR, DF-Threads offers an alternative implementation both in software and hardware [19] to achieve better energy efficiency by reducing unnecessary synchronization and data movement and to achieve resiliency for HPC applications due to the idempotency of DF-threads computations [34]. 2 https://github.com/alpha-unito/capio TAFFO and libVC. TAFFO [12] is a set of plugins for the LLVM Compiler Framework that enables automated precision tuning.…”

Optimizing Energy in Non-Preemptive Mixed-Criticality Scheduling by Exploiting Probabilistic Information

“…The probabilistic information is used only for the energy minimization and not for the schedulability test. Our schedulability test 2 Although the calculation under such a purpose would technically lead to execution thresholds that have nothing to do with 'worst-case', we nevertheless still follow the traditional MC work in the real-time and embedded systems community by calling them WCET under the lo-criticality mode.…”

“…This work is supported by NSF grant CNS-1850851, a startup grant from the University of Central Florida, the EU H2020 project RECIPE [1], [2], [23] grant 801137, and a HiPEAC collaboration grant 2019.…”