Memory-Aware Functional IR for Higher-Level Synthesis of Accelerators

Autor: Christof Schlaak, Tzung-Han Juang, Christophe Dubach

Publikováno v: Schlaak, C, Juang, T-H & Dubach, C 2022, ' Memory-Aware Functional IR for Higher-Level Synthesis of Accelerators ', ACM Transactions on Architecture and Code Optimization, vol. 19, no. 2, 16 . https://doi.org/10.1145/3501768

Specialized accelerators deliver orders of a magnitude of higher performance than general-purpose processors. The ever-changing nature of modern workloads is pushing the adoption of Field Programmable Gate Arrays (FPGAs) as the substrate of choice. H

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::c2f01eee8e6f6bb54a69cab6ac83b1b2
https://doi.org/10.1145/3501768

Mapping parallelism in a functional IR through constraint satisfaction: a case study on convolution for mobile GPUs

Autor: Naums Mogers, Lu Li, Valentin Radu, Christophe Dubach

Publikováno v: Mogers, N, Li, L, Radu, V & Dubach, C 2022, Mapping Parallelism in a Functional IR through Constraint Satisfaction : A Case Study on Convolution for Mobile GPUs . in B Egger & A Smith (eds), Proceedings of the 31st ACM SIGPLAN International Conference on Compiler Construction . pp. 218-230, ACM SIGPLAN 2022 International Conference on Compiler Construction, 2/04/22 . https://doi.org/10.1145/3497776.3517777

Graphics Processing Units (GPUs) are notoriously hard to optimize for manually. What is needed are good automatic code generators and optimizers. Accelerate, Futhark and Lift demonstrated that a functional approach is well suited for this challenge.

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::95b434b8daf6f79c260909cafa94018b
https://doi.org/10.1145/3497776.3517777

SparseAdapt: Runtime Control for Sparse Linear Algebra on a Reconfigurable Accelerator

Autor: Ronald G. Dreslinski, Siying Feng, Christophe Dubach, Michael O'Boyle, Subhankar Pal, Aporva Amarnath

Publikováno v: MICRO
Pal, S, Amarnath, A, Feng, S, O'Boyle, M, Dreslinski, R & Dubach, C 2021, SparseAdapt: Runtime Control for Sparse Linear Algebra on a Reconfigurable Accelerator . in The 54th Annual IEEE/ACM International Symposium on Microarchitecture Proceedings . pp. 1005-1021, 54th IEEE/ACM International Symposium on Microarchitecture, Athens, Greece, 18/10/21 . https://doi.org/10.1145/3466752.3480134

Dynamic adaptation is a post-silicon optimization technique that adapts the hardware to workload phases. However, current adaptive approaches are oblivious to implicit phases that arise from operating on irregular data, such as sparse linear algebra

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::50269edb5ff2ce0191b07c3ac145b65f
https://doi.org/10.1145/3466752.3480134

Tiling Optimizations for Stencil Computations Using Rewrite Rules in L ift

Autor: Sergei Gorlatch, Christophe Dubach, Larisa Stoltzfus, Bastian Hagedorn, Michel Steuwer

Publikováno v: Stoltzfus, L, Hagedorn, B, Steuwer, M, Gorlatch, S & Dubach, C 2020, ' Tiling Optimizations for Stencil Computations Using Rewrite Rules in Lift ', ACM Transactions on Architecture and Code Optimization, vol. 16, no. 4, 52, pp. 52:2-52:25 . https://doi.org/10.1145/3368858

Stencil computations are a widely used type of algorithm, found in applications from physical simulations to machine learning. Stencils are embarrassingly parallel, therefore fit on modern hardware such as Graphic Processing Units perfectly. Although

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_dedup___::e4a4257ec1670b91a932219b4bd02af2
https://doi.org/10.1145/3368858

Fast Optimisation of Convolutional Neural Network Inference using System Performance Models

Autor: Christophe Dubach, Valentin Radu, Rik Mulder

Publikováno v: EuroMLSys@EuroSys

The choice of convolutional routines (or primitives) for implementing the operations in a Convolutional Neural Network (CNN) has a tremendous impact over the inference time. To optimise the execution latency for a target system, a lengthy profiling s

Externí odkaz: https://explore.openaire.eu/search/publication?articleId=doi_________::b9947d8b697864c62e32ae66adcab1a0
https://doi.org/10.1145/3437984.3458840