Skip to main content

An Exascale Programming, Multi-objective Optimisation and Resilience Management Environment Based on Nested Recursive Parallelism

Deliverables

AllScale Compiler prototype

A prototype compiler that translates the API of Deliverables D3.1 and D3.2 to the interface offered by the runtime system of Deliverable D4.1

Installation, integration, and deployment of the AllScale environment and pilot applications (b)

Defines the guidelines, tools and procedures for integrating the AllScale environment and all pilot applications (outcome of T6.3).

Testing, evaluation and tuning of the AllScale environment with the pilot applications

Describes the different evaluation criteria and the results of the validation activities performed for every pilot application (outcome of Task 6.4) on the AllScale computing infrastructure.

API implementation for recursive parallelism (b)

A C++ implementation of the AllScale API comprising the Core and User API, being compatible to state-of-the-art C++ compilers.

Early AllScale runtime system prototype

A first proto-type of the runtime system implementing the interface defined by Deliverable D4.1 providing a unified scheduling interface and an implementation of it focusing on high resource utilisation within shared memory systems.

Full AllScale Compiler prototype

An extended version of Deliverable D3.3 adding automated porting support targeting (i) distributed memory systems and (ii) accelerators as well as (iii) facilities creating and restoring local checkpoints.

API implementation for recursive parallelism (a)

A C++ implementation of the AllScale API comprising the Core and User API, being compatible to state-of-the-art C++ compilers.

Installation, integration, and deployment of the AllScale environment and pilot applications (a)

Defines the guidelines, tools and procedures for integrating the AllScale environment and all pilot applications (outcome of T6.3).

Implementation and Evaluation of Application Specific Resilience Techniques (a)

An implementation and evaluation of the resilience techniques identified by Deliverable D5.1 utilising the AllScale API in the context of the pilot application.

AMDADOS Deepwater Horizon application preparation and evaluation (b)

Describes the incremental preparation, deployment and validation of the AMDADOS pilot application.

Multi-Objective Dynamic Optimiser (b)

This Deliverable explores and utilises the capabilities of the other deliverables to research strategies to effectively steer applications towards fulfilling customisable trade-offs among (conflicting) objectives.

iPIC3D implicit particle-in-cell code for space weather applications preparation and evaluation (b)

This deliverable describes the incremental preparation, deployment and validation of the iPIC3D pilot application.

AllScale API specification (a)

This deliverable will specify the AllScale Core and User API as well as their intended usage within the pilot applications (compare with Deliverable D3.1 covering its actual implementation)

On-Demand, On-Line Monitoring Infrastructure (a)

An infrastructure for collecting and aggregating monitoring information within the runtime system.

AllScale API specification (b)

This deliverable will specify the AllScale Core and User API as well as their intended usage within the pilot applications (compare with Deliverables D3.1 and 3.2 covering its actual implementation)

Data management plan (a)

This deliverable will contain an analysis of the main elements of the data management policy used by the applicants with regard to all data sets generated by the project. This will include a data set reference, name, description, standards, metadata, and management principles such as sharing, archiving and preservation.

AllScale runtime system monitoring infrastructure

Integration of the IPM tool with the HPX framework to develop an infrastructure for collecting, aggregating and querying (abstract) monitoring information effectively throughout connected runtime system instances.

Resource management support

An extended version of Deliverable D4.2 providing resource management support and an adapted scheduler targeting minimal energy usage.

AllScale computing infrastructure

This deliverable describes the characteristics of the AllScale computing infrastructure for the three pilot applications (outcome of Task T6.1).

AllScale runtime system interface specification

This deliverable will provide a document with the specification of the API to be used to interact with the runtime system as well as an implementation of the API itself.

AllScale system architecture (b)

This deliverable will describe the detailed software architecture of the AllScale environment comprising all components, their interfaces and responsibilities.

Project dissemination and communication strategy and reporting (a)

This deliverable will provide information about the project dissemination plans including the specification of the target dissemination groups, publication policy, event planning and marketing. AllScale will exploit popular web 2.0 channels and social media for further enhancing the project image. It is an ongoing deliverable with subsequent updated versions throughout the project lifetime.

AMDADOS Deepwater Horizon application preparation and evaluation (a)

Describes the incremental preparation, deployment and validation of the AMDADOS pilot application.

Implementation and Evaluation of Application Specific Resilience Techniques (b)

An implementation and evaluation of the resilience techniques identified by Deliverable D5.1 utilising the AllScale API in the context of the pilot application.

On-Demand, On-Line Monitoring Infrastructure (b)

An infrastructure for collecting and aggregating monitoring information within the runtime system.

AllScale system architecture (a)

This deliverable will describe the detailed software architecture of the AllScale environment comprising all components, their interfaces and responsibilities.

Data management support

An extended version of Deliverable D4.2 providing data management support and an adapted scheduler targeting high resource utilisation within distributed memory systems.

iPIC3D implicit particle-in-cell code for space weather applications preparation and evaluation (a)

This deliverable describes the incremental preparation, deployment and validation of the iPIC3D pilot application.

Project dissemination and communication strategy and reporting (b)

This deliverable will provide information about the project dissemination plans including the specification of the target dissemination groups, publication policy, event planning and marketing. AllScale will exploit popular web 2.0 channels and social media for further enhancing the project image. It is an ongoing deliverable with subsequent updated versions throughout the project lifetime.

Resilience Primitives

An API implementation for saving and restoring (local) recovery information and associated cost models and their documentation.

Multi-Objective Dynamic Optimiser (a)

This Deliverable explores and utilises the capabilities of the other deliverables to research strategies to effectively steer applications towards fulfilling customisable trade-offs among (conflicting) objectives.

Dissemination materials and impact reporting strategy

This deliverable will provide information about additional AllScale dissemination materials to be produced in the project, such as project leaflets, posters, newsletters, calendar of events, etc. It will also include other dissemination support materials are required to give higher impact to the project as it progresses.

Requirement specifications and reports on external technological developments (a)

This deliverable will describe requirements of the project with respect to the pilot applications and project objectives and specify benchmarks for performance eval-uations. Furthermore, it will identify external technological developments potentially influencing the project objectives and requirements.

Data management plan (b)

This deliverable will contain an analysis of the main elements of the data management policy used by the applicants with regard to all data sets generated by the project. This will include a data set reference, name, description, standards, metadata, and management principles such as sharing, archiving and preservation.

Final project report

This report will describe the overall achievements of the project.

Resilience Manager

A component identifying failures, estimating failure compensation costs and advising preparation and recovery operations to the runtime system scheduler. The deliverable will outline those techniques and provide an evaluation of those in the context of the AllScale pilot applications.

Requirement specifications and reports on external technological developments (b)

This deliverable will describe requirements of the project with respect to the pilot applications and project objectives and specify benchmarks for performance evaluations. Furthermore, it will identify external technological developments potentially influencing the project objectives and requirements.

Application Specific Resilience Strategies

A comprehensive description of developed application specific resilience strategies to be applied on the AllScale pilot applications as well as an evaluation of their expected capabilities.

Project web site and factsheet

This deliverable will describe the project website comprising the main dissemination and exploitation channels for promoting and communicating the project, relevant activities and achievements. This deliverable also provides information about the AllScale project factsheet. The web site will also have a part acting as an active “blog of research”, where partners can put articles about intermediate results, events, etc. Periodically updates of the website, according to users’ evaluation session planned in dissemination will be carried out.

Searching for OpenAIRE data...

Publications

A taxonomy of task-based parallel programming technologies for high-performance computing

Author(s): Peter Thoman, Kiril Dichev, Thomas Heller, Roman Iakymchuk, Xavier Aguilar, Khalid Hasanov, Philipp Gschwandtner, Pierre Lemarinier, Stefano Markidis, Herbert Jordan, Thomas Fahringer, Kostas Katrinis, Erwin Laure, Dimitrios S. Nikolopoulos
Published in: The Journal of Supercomputing, Issue 74/4, 2018, Page(s) 1422-1434, ISSN 0920-8542
DOI: 10.1007/s11227-018-2238-4

Static Compiler Analyses for Application-specific Optimization of Task-Parallel Runtime Systems

Author(s): Peter Thoman, Peter Zangerl, Thomas Fahringer
Published in: Journal of Signal Processing Systems, 2018, ISSN 1939-8018
DOI: 10.1007/s11265-018-1356-9

SCALO

Author(s): Giorgis Georgakoudis, Hans Vandierendonck, Peter Thoman, Bronis R. De Supinski, Thomas Fahringer, Dimitrios S. Nikolopoulos
Published in: ACM Transactions on Architecture and Code Optimization, Issue 14/4, 2017, Page(s) 1-25, ISSN 1544-3566
DOI: 10.1145/3158643

Multi-Objective region-Aware optimization of parallel programs

Author(s): Juan J. Durillo, Philipp Gschwandtner, Klaus Kofler, Thomas Fahringer
Published in: Parallel Computing, 2018, ISSN 0167-8191
DOI: 10.1016/j.parco.2018.03.010

Localised sequential state estimation for advection dominated flows with non-Gaussian uncertainty description

Author(s): Emanuele Ragnoli, Mykhaylo Zayats, Fearghal O'Donncha, Sergiy Zhuk
Published in: Journal of Computational Physics, Issue bimonthly, 2018, Page(s) In Press, ISSN 0021-9991
DOI: 10.5281/zenodo.1462715

Expediting assessments of database performance for streams of respiratory parameters

Author(s): Gillan, Charles; Novakovic, Aleksandar; Shyamsundar, Murali; Marshall, Adele H; Nikolopoulos, Dimitrios
Published in: Journal of Computers in Biology and Medicine, Issue 1, 2018, ISSN 0010-4825
DOI: 10.5281/zenodo.1244669

A Task-Based Particle-in-Cell Method with Automatic Load-Balancing using the AllScale Environment

Author(s): Roman Iakymchuk; Herbert Jordan; Philipp Gschwandtner; Thomas Heller; Peter Thoman; Xavier Aguilar; Thomas Fahringer; Erwin Laure; Stefano Markidis
Published in: Exascale Applications and Software Conference (EASC2018), 2018
DOI: 10.5281/zenodo.1119103

A Taxonomy of Task-Based Technologies for High-Performance Computing

Author(s): Peter Thoman; Khalid Hasanov; Kiril Dichev; Roman Iakymchuk; Xavier Aguilar; Philipp Gschwandtner; Pierre Lemarinier; Stefano Markidis; Herbert Jordan; Erwin Laure; Kostas Katrinis; Dimitrios S. Nikolopoulos; Thomas Fahringer
Published in: In: Wyrzykowski R., Dongarra J., Deelman E., Karczewski K. (eds) Parallel Processing and Applied Mathematics. PPAM 2017. Lecture Notes in Computer Science, Issue 10778, 2017
DOI: 10.5281/zenodo.1162306

Performance and Behavior Characterization of Amazon EC2 Spot Instances

Author(s): Pham, Thanh-Phuong; Ristov, Sasko; Fahringer, Thomas
Published in: Issue 4, 2018
DOI: 10.5281/zenodo.1310921

Exploring the Semantic Gap in Compiling Embedded DSLs

Author(s): Zangerl, Peter; Jordan, Herbert; Thoman, Peter; Gschwandtner, Philipp; Fahringer, Thomas
Published in: Issue 2, 2018
DOI: 10.5281/zenodo.1309474

The AllScale Runtime Application Model (incl. Appendix)

Author(s): Jordan, Herbert; Heller, Thomas; Gschwandtner, Philipp; Zangerl, Peter; Thoman, Peter; Fey, Dietmar; Fahringer, Thomas
Published in: Issue 2, 2018
DOI: 10.5281/zenodo.1322420

Characterizing Performance and Cache Impacts of Code Multi-versioning on Multicore Architectures

Author(s): Peter Zangerl, Peter Thoman, Thomas Fahringer
Published in: 2017 25th Euromicro International Conference on Parallel, Distributed and Network-based Processing (PDP), 2017, Page(s) 209-213
DOI: 10.1109/PDP.2017.77

Task-parallel Runtime System Optimization Using Static Compiler Analysis

Author(s): Peter Thoman, Peter Zangerl, Thomas Fahringer
Published in: Proceedings of the Computing Frontiers Conference on ZZZ - CF'17, 2017, Page(s) 201-210
DOI: 10.1145/3075564.3075574

A Region-Aware Multi-Objective Auto-Tuner for Parallel Programs

Author(s): Klaus Kofler, Juan J. Durillo, Philipp Gschwandtner, Thomas Fahringer
Published in: 2017 46th International Conference on Parallel Processing Workshops (ICPPW), 2017, Page(s) 190-199
DOI: 10.1109/ICPPW.2017.37

A localised data assimilation framework within the 'AllScale' parallel development environment

Author(s): Akhriev, Albert; Gschwandtner, Philipp; Jordan, Herbert; O'Donncha, Fearghal
Published in: MTS/IEEE Oceans 2018, Issue 1, 2018, Page(s) 1-7
DOI: 10.5281/zenodo.1346169

Energy-efficient localised rollback via data flow analysis and frequency scaling

Author(s): Kiril Dichev, Kirk Cameron, Dimitrios S. Nikolopoulos
Published in: Proceedings of the 25th European MPI Users' Group Meeting on - EuroMPI'18, 2018, Page(s) 1-11
DOI: 10.1145/3236367.3236379

Characterizing Performance and Cache Impacts of Code Multi-Versioning on Multicore Architectures

Author(s): Peter Thoman; Peter Zangerl; Thomas Fahringer
Published in: 25th Euromicro International Conference on Parallel, Distributed and Network-based Processing, Issue 1, 2017
DOI: 10.5281/zenodo.375519

On fast large-scale program analysis in Datalog

Author(s): Bernhard Scholz, Herbert Jordan, Pavle Subotić, Till Westmann
Published in: Proceedings of the 25th International Conference on Compiler Construction - CC 2016, 2016, Page(s) 196-206
DOI: 10.1145/2892208.2892226

A Context-aware Primitive for Nested Recursive Parallelism

Author(s): Thomas Fahringer; Thomas Heller; Peter Thoman; Herbert Jordan; Peter Zangerl
Published in: 5th International Workshop on Multicore Software Engineering (IWMSE 2016), Issue 2, 2016
DOI: 10.5281/zenodo.345835

SOUFFLÉ: On Synthesis of Program Analyzers

Author(s): Jordan, Herbert; Scholz, Bernhard; Subotić, Pavle
Published in: International Conference on Computer Aided Verification, CAV 2016, Issue 2, 2016
DOI: 10.1007/978-3-319-41540-6_23

Deploying and optimizing performance of a 3D hydrodynamic model on cloud

Author(s): Fearghal O'Donncha, Srikumar Venugopal, Scott C. James, Emanuele Ragnoli
Published in: OCEANS 2016 MTS/IEEE Monterey, Issue 1, 2016, Page(s) 1-7
DOI: 10.1109/OCEANS.2016.7761131

A Particle-in-Cell Method for Automatic Load-Balancing with the AllScale Environment

Author(s): Iakymchuk, Roman; Jordan, Herbert; Bo Peng, Ivy; Markidis, Stefano; Laure, Erwin
Published in: Exascale Applications and Software Conference (EASC2016), 2016

TwinPCG: Dual Thread Redundancy with Forward Recovery for Preconditioned Conjugate Gradient Methods

Author(s): Kiril Dichev, Dimitrios S. Nikolopoulos
Published in: 2016 IEEE International Conference on Cluster Computing (CLUSTER), 2016, Page(s) 506-514
DOI: 10.1109/CLUSTER.2016.99