SparCity: An Optimization and Co-design Framework for Sparse Computation

Risultati finali

Public release of database of execution efficiency in terms of performance, power, and energy usage.

Publication and release of performance and energy models.

Profiling and instrumentation tools will be developed by INESC ID and shared with other partners and public.

KU will develop core set of sparse matrix features for ML learning techniques and make it available to the general public as a publication or report.

ML-based prediction model

Machine learning based prediction model and its source code for sparse computation will be made public.

Reordering Library

Publication and source release of the reordering library by Simula.

SuperTwin and API

Publication on the SuperTwin tool that describes its API and use-cases

Mixed precision support in SparCity Framework

Publication and source release of mixed precision support of SparCity framework

Visualization tools

Release of first version of the visualization tools for the current performance, power and energy usage, and communication and data movement patterns.

Topology-aware heterogeneous multi-level partitioner

Source code release and a publication on topology-aware heterogeneous multi-level partitioner

Fully functioning code generator

SU will release the code generator source code to public along with its documentation.

Microbenchmarking tools

Publications and release of microbenchmarking tools

Mid-term usability report of the SparCity framework

Simula will collect the results of testing the SparCity Framework on the sample use-cases and share the scientific findings with the public.

Scientific evaluation report of the SparCity framework

Simula will release an evaluation report on the SparCity framework based on the partners' scientific findings and the use-cases of SparCity.

2nd Workshop about the SparCity framework and the modeling tools

LMU will organize the second workshop that demonstrates the SparCity tools and provides tutorials on how to use them.

Prototype of performance and energy models

Demonstration of performance and energy models

Prototype for the dynamic topology information system

This deliverable will release a prototype of a dynamic topology information system that integrates existing data sources such as the hwloc and vendor-specific tools to capture the complex setup of modern HPC systems.

Prototype of kernel template generator and kernel analyzer

We will release the prototype of a source-to-source compiler that generates a kernel template auto-tuned for the target device.

1st Workshop about the SparCity framework and the modeling tools

LMU will organize a workshop to demonstrate the Sparcity framework and its tools.

Pubblicazioni

Modelling Data Locality of Sparse Matrix-Vector Multiplication on the A64FX.

Autori: Sergej Breiter, James D. Trotter, and Karl Fürlinger
Pubblicato in: Association for Computing Machinery, 2023
Editore: ACM
DOI: 10.1145/3624062.3624198

Bringing Order to Sparsity: A Sparse Matrix Reordering Study on Multicore CPUs.

Autori: James D. Trotter, Sinan Ekmekçibaşı, Johannes Langguth, Tugba Torun, Emre Düzakın, Aleksandar Ilic, and Didem Unat
Pubblicato in: 2023
Editore: ACM
DOI: 10.1145/3581784.3607046

iPUG for Multiple Graphcore IPUs: Optimizing Performance and Scalability of Parallel Breadth-First Search

Autori: Luk Burchard, Xing Cai, Johannes Langguth
Pubblicato in: 2021 IEEE 28th International Conference on High Performance Computing, Data, and Analytics (HiPC), 2021, Pagina/e 162-171
Editore: IEEE
DOI: 10.1109/hipc53243.2021.00030

Performance Modelling-driven Optimization of RISC-V Hardware for Efficient SpMV. Proceedings of the International workshop on RISC-V for HPC.

Autori: Alexandre Rodrigues, Leonel Sousa and Alexandar Ilic
Pubblicato in: High Performance Computing. ISC High Performance 2023. Lecture Notes in Computer Science, 2023
Editore: Springer, Cham.
DOI: 10.1007/978-3-031-40843-4_36

Precise event sampling-based data locality tools for AMD multicore architectures

Autori: Muhammad Aditya Sasongko, Milind Chabbi, Paul H. J. Kelly, Didem Unat
Pubblicato in: Concurrency Computat Pract Exper., 2023, ISSN 1532-0626
Editore: John Wiley and Sons Ltd.
DOI: 10.1002/cpe.7707

Mixed and Multi-Precision SpMV for GPUs with Row-wise Precision Selection

Autori: Erhan Tezcan, Tugba Torun, Fahrican Koşar, Kamer Kaya, and Didem Unat
Pubblicato in: IEEE 34th International Symposium on Computer Architecture and High Performance Computing, 2022
Editore: IEEE

Explaining the Performance of Supervised and Semi-Supervised Methods for Automated Sparse Matrix Format Selection

Autori: Sunidhi Dhandhania, Akshay Deodhar, Konstantin Pogorelov, Swarnendu Biswas, Johannes Langguth
Pubblicato in: ICPP Workshops '21: 50th International Conference on Parallel Processing Workshop, 2021
Editore: ACM
DOI: 10.1145/3458744.3474049

Using GNNs for Misinformation Spreader Detection via Assortativity-Aware Node Label Classification in Twitter Networks

Autori: Asep Maulana, Johannes Langguth
Pubblicato in: 2023 Tenth International Conference on Social Networks Analysis, Management and Security (SNAMS), 2023
Editore: IEEE
DOI: 10.1109/snams60348.2023.10375407

Degree-Aware Kernels for Computing Jaccard Weights on GPUs

Autori: Amro Alabsi Aljundi, Taha Atahan Akyildiz, Kamer Kaya
Pubblicato in: 2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS), 2022, ISBN 978-1-6654-8106-9
Editore: IEEE
DOI: 10.1109/ipdps53621.2022.00092

Tensor-Accelerated Fourth-Order Epistasis Detection on GPUs

Autori: Ricardo Nobre, Aleksandar Ilic, Sergio Santander-Jiménez, and Leonel Sousa
Pubblicato in: 51st International Conference on Parallel Processing (ICPP ’22), 2022
Editore: ACM

Unlocking Personalized Healthcare on Modern CPUs/GPUs: Three-way Gene Interaction Study.

Autori: Diogo Marques, Rafael Campos, Sergio Santander-Jiménez, Zakhar Matveev, Leonel Sousa, and Aleksandar Ilic
Pubblicato in: IEEE International Parallel and Distributed Processing Symposium (IPDPS), 2022
Editore: IEEE
DOI: 10.1109/ipdps53621.2022.00023

A Profiling-based Approach to Cache Partitioning of Program Data

Autori: Sergej Breiter, Josef Weidendorfer, Minh Thanh Chung and Karl Fürlinger
Pubblicato in: Proceedings of the 23rd International Conference on Parallel and Distributed Computing, Applications and Technologies (PDCAT’22), 2022
Editore: Springer Lecture Notes in Computer Science (LNCS)

Fast and High-Quality Influence Maximization on Multiple GPUs

Autori: Gökhan Göktürk, Kamer Kaya
Pubblicato in: 2022 IEEE International Parallel and Distributed Processing Symposium (IPDPS), 2022, ISBN 978-1-6654-8106-9
Editore: IEEE
DOI: 10.1109/ipdps53621.2022.00093

Interpreting High Order Epistasis Using Sparse Transformers.

Autori: Miguel Graça, Diogo Marques, Sergio Santander-Jiménez, Leonel Sousa, and Aleksandar Ilic
Pubblicato in: IEEE/ACM Conference on Connected Health: Applications, Systems and Engineering Technologies (CHASE), 2023
Editore: IEEE
DOI: 10.1145/3580252.3586982

Fourth-Order Exhaustive Epistasis Detection for the xPU Era

Autori: Ricardo Nobre, Aleksandar Ilic, Sergio Santander-Jiménez, Leonel Sousa
Pubblicato in: 50th International Conference on Parallel Processing, 2021, Pagina/e 1-10, ISBN 9781450390682
Editore: ACM
DOI: 10.1145/3472456.3472509

Boosting Graph Embedding on a Single GPU

Autori: Amro Alabsi Aljundi, Taha Atahan Akyildiz, Kamer Kaya
Pubblicato in: IEEE Transactions on Parallel and Distributed Systems, Numero 10459219, 2022, Pagina/e 3092 - 3105, ISSN 1045-9219
Editore: Institute of Electrical and Electronics Engineers
DOI: 10.1109/tpds.2021.3129617

Fast and error-adaptive influence maximization based on Count-Distinct sketches

Autori: Gökhan Göktürk, Kamer Kaya
Pubblicato in: Information Sciences, Numero Volume 655, January 2024, 119875, 2024, ISSN 1872-6291
Editore: Elsevier
DOI: 10.1016/j.ins.2023.119875

Mansard Roofline Model: Reinforcing the Accuracy of the Roofs

Autori: Diogo Marques, Aleksandar Ilic, Leonel Sousa
Pubblicato in: ACM Transactions on Modeling and Performance Evaluation of Computing Systems, Numero 6/2, 2021, Pagina/e 1-23, ISSN 2376-3639
Editore: Association for Computing Machinery
DOI: 10.1145/3475866

PhyloMissForest: a random forest framework to construct phylogenetic trees with missing data

Autori: Diogo Pinheiro; Sergio Santander-Jimenéz; Aleksandar Ilic
Pubblicato in: BMC Genomics, Numero 14712164, 2022, Pagina/e 1-21, ISSN 1471-2164
Editore: BioMed Central
DOI: 10.1186/s12864-022-08540-6

Targeting performance and user-friendliness: GPU-accelerated finite element computation with automated code generation in FEniCS.

Autori: James D. Trotter, Johannes Langguth and Xing Cai
Pubblicato in: Parallel Computing, 2023, ISSN 0167-8191
Editore: Elsevier BV
DOI: 10.1016/j.parco.2023.103051

Detailed Modeling of Heterogeneous and Contention-Constrained Point-to-Point MPI Communication

Autori: Andreas Thune; Sven-Arne Reinemo; Tor Skeie; Xing Cai
Pubblicato in: IEEE Transactions on Parallel and Distributed Systems, 2023, ISSN 1045-9219
Editore: Institute of Electrical and Electronics Engineers
DOI: 10.1109/tpds.2023.3253881

Enabling Unstructured-Mesh Computation on Massively Tiled AI-Processors: An Example of Accelerating In-Silico Cardiac Simulation.

Autori: Luk Burchard, Kristian Gregorius Hustad, Johannes Langguth and Xing Cai
Pubblicato in: Front. Phys. Sec. Statistical and Computational Physics, 2023, ISSN 2296-424X
Editore: Frontiers Media
DOI: 10.3389/fphy.2023.979699

iPUG: Accelerating Breadth-First Graph Traversals Using Manycore Graphcore IPUs

Autori: Luk Burchard, Johannes Moe, Daniel Thilo Schroeder, Konstantin Pogorelov, Johannes Langguth
Pubblicato in: High Performance Computing - 36th International Conference, ISC High Performance 2021, Virtual Event, June 24 – July 2, 2021, Proceedings, Numero 12728, 2021, Pagina/e 291-309, ISBN 978-3-030-78712-7
Editore: Springer International Publishing
DOI: 10.1007/978-3-030-78713-4_16

È in corso la ricerca di dati su OpenAIRE...

Risultati finali

Pubblicazioni

Condividi questa pagina

Scarica