Abstract
Collective operations such as scatter, gather, reduce, etc are utilized broadly to implement distributed HPC applications and are the target of extensive optimization in all MPI implementations as well as dedicated collective libraries by accelerator vendors (e.g. NCCL and RCCL by NVidia and AMD respectively). We present ACCL, an open-source FPGAaccelerated collectives library designed to serve applications running primarily in Xilinx FPGAs. Compared to previous collective communication solutions for FPGA, ACCL is flexible and extensible, easily portable, and fast. We evaluate ACCL up to 8 nodes and demonstrate that ACCL outperforms OpenMPI over 100 Gbps TCP-IP for large messages. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000510849Publication status
publishedExternal links
Book title
2021 IEEE/ACM International Workshop on Heterogeneous High-performance Reconfigurable Computing (H2RC)Pages / Article No.
Publisher
IEEEEvent
Subject
FPGA; collectives; MPIOrganisational unit
03506 - Alonso, Gustavo / Alonso, Gustavo
03506 - Alonso, Gustavo / Alonso, Gustavo
Notes
Conference lecture held on November 15, 2021More
Show all metadata
ETH Bibliography
yes
Altmetrics