%0 Conference Paper %1 c5c7030455624f8691e05f963bf1b7e1 %A Iskandar, Veronia %A El Ghany, Mohamed A.Abd %A Goehringer, Diana %B Proceedings - 31st IEEE International Symposium on Field-Programmable Custom Computing Machines, FCCM 2023 %C United States of America %D 2023 %I Institute of Electrical and Electronics Engineers Inc. %K FIS_scads nopdf topic_federatedlearn %P 222 %R 10.1109/FCCM57271.2023.00050 %T Compiler-Assisted Kernel Selection for FPGA-based Near-Memory Computing Platforms %U https://www.fccm.org/past/2023/ %X The speed of modern computing systems has improved significantly, thanks to advances in CMOS technology. However, the memory bandwidth of DRAM has not kept pace with these improvements in terms of latency and energy consumption, which is known as the memory wall 1. FPGAs with high-bandwidth memory (HBM) provide significantly improved performance on memory-intensive tasks, such as graph processing and machine learning. By leveraging 3D-stacked DRAM memory on FPGAs, it is possible to realize the Near-Memory Computing (NMC) paradigm, which involves offloading some kernels to be processed close to the memory. While there have been many studies on NMC accelerators, there is no established method for determining which application kernels are suitable for execution near the HBM. To fully realize the potential of FPGA-HBM architectures, it is important to identify offloading candidates without relying on programmers' knowledge. However, this is a non-trivial task due to the complexity of modern applications. To address this issue, we propose a compiler-assisted tool-flow for the automatic selection of kernels to be offloaded.

@inproceedings{c5c7030455624f8691e05f963bf1b7e1, abstract = {The speed of modern computing systems has improved significantly, thanks to advances in CMOS technology. However, the memory bandwidth of DRAM has not kept pace with these improvements in terms of latency and energy consumption, which is known as the memory wall [1]. FPGAs with high-bandwidth memory (HBM) provide significantly improved performance on memory-intensive tasks, such as graph processing and machine learning. By leveraging 3D-stacked DRAM memory on FPGAs, it is possible to realize the Near-Memory Computing (NMC) paradigm, which involves offloading some kernels to be processed close to the memory. While there have been many studies on NMC accelerators, there is no established method for determining which application kernels are suitable for execution near the HBM. To fully realize the potential of FPGA-HBM architectures, it is important to identify offloading candidates without relying on programmers' knowledge. However, this is a non-trivial task due to the complexity of modern applications. To address this issue, we propose a compiler-assisted tool-flow for the automatic selection of kernels to be offloaded.}, added-at = {2024-11-28T16:27:18.000+0100}, address = {United States of America}, author = {Iskandar, Veronia and {El Ghany}, {Mohamed A.Abd} and Goehringer, Diana}, biburl = {https://puma.scadsai.uni-leipzig.de/bibtex/2967f51f482d2f4d7d18d5ba042a8b4a8/scadsfct}, booktitle = {Proceedings - 31st IEEE International Symposium on Field-Programmable Custom Computing Machines, FCCM 2023}, doi = {10.1109/FCCM57271.2023.00050}, interhash = {5460228fc9467b8a1cbeeceb923d8fa8}, intrahash = {967f51f482d2f4d7d18d5ba042a8b4a8}, keywords = {FIS_scads nopdf topic_federatedlearn}, language = {English}, note = {Publisher Copyright: {\textcopyright} 2023 IEEE.; 31st IEEE International Symposium on Field-Programmable Custom Computing Machines, FCCM 2023 ; Conference date: 08-05-2023 Through 11-05-2023}, pages = 222, publisher = {Institute of Electrical and Electronics Engineers Inc.}, series = {Annual IEEE Symposium on Field-Programmable Custom Computing Machines (FCCM)}, timestamp = {2025-03-13T14:28:13.000+0100}, title = {Compiler-Assisted Kernel Selection for FPGA-based Near-Memory Computing Platforms}, url = {https://www.fccm.org/past/2023/}, year = 2023 }