Institutional Repository
Technical University of Crete
Technical University of Crete
EN
|
EL
| URI: | http://purl.tuc.gr/dl/dias/05A79EB2-CC89-4B96-9B64-6B75447380DF | ||
| Year | 2024 | ||
| Type of Item | Diploma Work | ||
| License |
|
||
| Bibliographic Citation | Aikaterini Tsimpirdoni, "Remote execution of an FPGA-based cellular automata accelerator on the Amazon F1 cloud", Diploma Work, School of Electrical and Computer Engineering, Technical University of Crete, Chania, Greece, 2024 https://doi.org/10.26233/heallink.tuc.101224 | ||
| Appears in Collections |
Cellular automata, introduced by John von Neumann and Stanislaw Ulamin the 1940s, are discrete mathematical models used to simulate complex systemsthrough simple rules. They are widely applied in various scientificfields to study dynamic systems. In this thesis, a reprogrammable FPGAbasedframework was developed to efficiently simulate cellular automata(CA) models on the AWS F1 platform. The design builds upon the architectureinitially developed by Nikolaos Kyparissas and later extended by EmmanouilMilonakis, with additional improvements introduced in this work.The AWS FPGA Developer AMI was employed, offering a pre-configuredenvironment with tools like Xilinx Vivado, accessed remotely via NICE DCV,eliminating the need for a physical FPGA board.The framework was createdby generating an Amazon FPGA Image (AFI) bitstream, compatible with theAWS F1 instance. After implementing logic changes in the VHDL code, thebitstream was synthesized and deployed. Optimizations included upgradingthe memory system from a 128-bit DDR2 to a 512-bit DDR4 configuration,enhancing data handling and increasing burst size. Data transfer between thehost and FPGA was managed via PCIe using Direct Memory Access (DMA)by configuring the PCIe to AXI bridge for efficient communication.The FPGAexecuted the CA model, achieving a 21.2x performance improvement overtraditional software methods, particularly when processing a 1920x1080 CAgrid.
Copyright © DIAS 2013
