Skip to main content

Lund University Publications

LUND UNIVERSITY LIBRARIES

Parametric actors networks to efficient hardware : CORDIC-based QR decomposition

Callanan, Gareth LU orcid and Gruian, Flavius LU orcid (2025) p.135-142
Abstract
In this work, we present an actors-to-hardware design flow where applications are described as networks of actors and compiled to hardware. Specifically, we use the CAL actor language to describe these applications. CAL has previously been used to specify applications such as the MPEG-RVC decoder and is well suited for compilation to hardware. A key limitation of CAL is that the applications described have a fixed structure, limiting its applicability. In this work, we show how recent additions to the CAL language and upgrades to the StreamBlocks CAL compiler have overcome these challenges, enabling parametric networks to be specified and compiled to hardware. To illustrate our approach, we implement a CORDIC-based QR Decomposition... (More)
In this work, we present an actors-to-hardware design flow where applications are described as networks of actors and compiled to hardware. Specifically, we use the CAL actor language to describe these applications. CAL has previously been used to specify applications such as the MPEG-RVC decoder and is well suited for compilation to hardware. A key limitation of CAL is that the applications described have a fixed structure, limiting its applicability. In this work, we show how recent additions to the CAL language and upgrades to the StreamBlocks CAL compiler have overcome these challenges, enabling parametric networks to be specified and compiled to hardware. To illustrate our approach, we implement a CORDIC-based QR Decomposition application as a systolic array in CAL, targeting Massive-MIMO systems used in telecommunications. We describe how this toolchain converts the CAL description into a combination of Verilog and C++ HLS code and demonstrate how this format allows for performance tuning. Our approach expands the capabilities of CAL and CAL toolchains allowing it to be used to compile a broader range of applications to hardware. (Less)
Please use this url to cite or link to this publication:
author
and
organization
publishing date
type
Chapter in Book/Report/Conference proceeding
publication status
published
subject
host publication
27th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC)
pages
135 - 142
publisher
IEEE - Institute of Electrical and Electronics Engineers Inc.
external identifiers
  • scopus:105038327426
DOI
10.1109/SYNASC69064.2025.00027
language
English
LU publication?
yes
id
335c0cbf-d105-4791-811c-9984d74bfe61
date added to LUP
2026-04-24 10:31:28
date last changed
2026-06-24 04:00:55
@inproceedings{335c0cbf-d105-4791-811c-9984d74bfe61,
  abstract     = {{In this work, we present an actors-to-hardware design flow where applications are described as networks of actors and compiled to hardware. Specifically, we use the CAL actor language to describe these applications. CAL has previously been used to specify applications such as the MPEG-RVC decoder and is well suited for compilation to hardware. A key limitation of CAL is that the applications described have a fixed structure, limiting its applicability. In this work, we show how recent additions to the CAL language and upgrades to the StreamBlocks CAL compiler have overcome these challenges, enabling parametric networks to be specified and compiled to hardware. To illustrate our approach, we implement a CORDIC-based QR Decomposition application as a systolic array in CAL, targeting Massive-MIMO systems used in telecommunications. We describe how this toolchain converts the CAL description into a combination of Verilog and C++ HLS code and demonstrate how this format allows for performance tuning. Our approach expands the capabilities of CAL and CAL toolchains allowing it to be used to compile a broader range of applications to hardware.}},
  author       = {{Callanan, Gareth and Gruian, Flavius}},
  booktitle    = {{27th International Symposium on Symbolic and Numeric Algorithms for Scientific Computing (SYNASC)}},
  language     = {{eng}},
  pages        = {{135--142}},
  publisher    = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}},
  title        = {{Parametric actors networks to efficient hardware : CORDIC-based QR decomposition}},
  url          = {{http://dx.doi.org/10.1109/SYNASC69064.2025.00027}},
  doi          = {{10.1109/SYNASC69064.2025.00027}},
  year         = {{2025}},
}