Skip to main content

Sustainable Performance for High-Performance Embedded Computing Systems

Periodic Reporting for period 1 - SuPerCom (Sustainable Performance for High-Performance Embedded Computing Systems)

Reporting period: 2018-06-01 to 2019-11-30

"SuPerCom addresses the huge challenge of providing ""high and sustainable performance"" (hsperf) covering the highest-ever computation performance needs of critical software with strong guarantees on sustainability for safe operation. This will be tackled for future mixed-criticality embedded computers comprising high-performance hardware with unprecedented complexity levels. To reach its goals, SuPerCom proposes a radical new approach by combining performance analysis, hardware design, and statistical and machine learning analysis. SuPerCom solutions can become integral part in the ecosystem of next-generation embedded computing systems by allowing them to use increasingly-complex high-performance hardware features on which strong guarantees of sustainable performance can be placed soundly a priori. This will allow developers to use computer performance-demanding functionalities (with sound guarantees) such as complex algorithms to control CO2 emissions or to provide driving assistance in cars; or devices with increasing performance requirements in the medical market such as pacemakers or infusion pumps. Hence, the SuPerCom breakthrough can have significant economic and societal impact."
From the beginning of the project, several main activities have be performed as follows:
1) Simulations tools and case studies. This captures a set of requirements emanating from (a) the parallel hardware architecture and (b) performance analysis techniques that influence hardware designs and high and sustainable performance models respectively. It also consolidates the baseline tools to use and work in their inter-operability. We have a solid set of requirements, and an integrated version of the evaluation toolchain to start the development of SuPerCom innovative solutions.

2) Requirements, tools and evaluations. SuPerCom researchers have been jointly working with the PI to define the main project requirements. The engineers have developed and prepared a sophisticated and comprehensive baseline simulations infrastructure featuring state-of-the-art architectural support and industry-level accuracy. We are actively working on keeping this tools updated and adding features which are required.
SuPerCom will propose innovative solutions that push the limits of current approaches for sustainable performance: (1) SuPerCom will propose novel hardware support for performance prediction that captures the sustainable performance requirements emanating from applications, with minimum or no impact on average performance. (2) For hard-to-predict resources, unlike current approaches that propose invasive methods to increase predictability, SuPerCom eases performance modelling via novel hardware sensors that expose key performance indicators to the performance analysis tools; the latter (3) will be extended with statistical and machine learning analysis to automatically manage the high amounts of information provided by hardware sensors, both during system design and operation. This is coupled with (4) advanced software techniques that generate adjustable load on resources to obtain high and sustainable performance under different usage scenarios.
Summary of the publications