FPGA-based accelerators demonstrated high energy efficiency compared to GPUs and CPUs. However, single FPGA designs may not achieve sufficient task parallelism. In this work, we optimize the mapping of high-performance multi-kernel applications, like Convolutional Neural Networks, to multi-FPGA platforms. First, we formulate the system level optimization problem, choosing within a huge design space the parallelism and number of compute units for each kernel in the pipeline. Then we solve it using a combination of Geometric Programming, producing the optimum performance solution given resource and DRAM bandwidth constraints, and a heuristic allocator of the compute units on the FPGA cluster.

Exact and Heuristic Allocation of Multi-kernel Applications to Multi-FPGA Platforms / Shan, Junnan; Casu, Mario R.; Cortadella, Jordi; Lavagno, Luciano; Lazarescu, Mihai T.. - ELETTRONICO. - (2019), pp. 1-6. (Intervento presentato al convegno Design Automation Conference 2019 tenutosi a Las Vegas, NV (USA) nel 2-6 giugno 2019) [10.1145/3316781.3317821].

Exact and Heuristic Allocation of Multi-kernel Applications to Multi-FPGA Platforms

Shan, Junnan;Casu, Mario R.;Lavagno, Luciano;Lazarescu, Mihai T.
2019

Abstract

FPGA-based accelerators demonstrated high energy efficiency compared to GPUs and CPUs. However, single FPGA designs may not achieve sufficient task parallelism. In this work, we optimize the mapping of high-performance multi-kernel applications, like Convolutional Neural Networks, to multi-FPGA platforms. First, we formulate the system level optimization problem, choosing within a huge design space the parallelism and number of compute units for each kernel in the pipeline. Then we solve it using a combination of Geometric Programming, producing the optimum performance solution given resource and DRAM bandwidth constraints, and a heuristic allocator of the compute units on the FPGA cluster.
2019
9781450367257
File in questo prodotto:
File Dimensione Formato  
dac19.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 820.97 kB
Formato Adobe PDF
820.97 kB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2740832
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo