Self-decompressing FPGA Bitstreams

SRAM based FPGAs (field programmable gate arrays) are volatile devices, and need to reload its configuration (bitstream) every time after power up. Bitstream compression is one of the major method to reduce the cost of storing the configuration storage and speed up configuration. However, existing c...

Full description

Saved in:
Bibliographic Details
Main Authors: Ma, Shenghou, Ampadu, Paul
Format: Conference Proceeding
Language:English
Subjects:
Clocks
configuration compression
configuration datastreams
decompression hardware
Engines
Fabrics
Field programmable gate arrays
Loading
Logic gates
reconfigurable logic
Routing
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:SRAM based FPGAs (field programmable gate arrays) are volatile devices, and need to reload its configuration (bitstream) every time after power up. Bitstream compression is one of the major method to reduce the cost of storing the configuration storage and speed up configuration. However, existing complete bitstream (as compared to partial bitstream) compression methods require an external decompresser implemented either in another FPGA or CPU. This drawback prevents deployment of bitstream compression to already in-field systems. This paper drew inspiration from self-extracting archive, and utilizes the self partial reconfiguration capabilities of modern FPGAs to create a self-decompressing FPGA bitstreams so that the decompression engine is located inside the compressed bitstream itself. Analysis showed that this method will never increase the size of bitstream, and can achieve configuration time reduction whenever the compression ratio is smaller than 7/9 and the configuration reduction scales almost linearly with the compression ratio. Furthermore, the paper shows that by using bitstream assembly techniques, it is possible to choose the compression algorithm after building the target bitstream, therefore, the best compression ratio can be achieved given a pool of available decompression engines.
ISSN:1558-3899
DOI:10.1109/MWSCAS.2019.8885346