High-Level Synthesis Techniques for FPGA Design

High-Level Synthesis (HLS) techniques have transformed the development of field-programmable gate arrays (FPGAs) by raising the abstraction level from hardware description languages to widely used programming languages such as C, C++ and Python. This shift enables designers to express algorithmic intent without dealing directly with low-level register transfer logic, while automated tooling handles scheduling, resource allocation and optimisation. Key HLS strategies include loop transformations (unrolling, pipelining and tiling), resource-sharing directives and memory partitioning to balance throughput, latency and area constraints. Modern toolchains integrate performance modelling and design space exploration to guide trade-off decisions, often incorporating roofline-style analyses to identify compute- or memory-bound kernels. Emerging trends further exploit domain-specific languages, algorithmic skeletons and machine-learning-driven directive selection to reduce manual intervention. Collectively, these advances have broadened FPGA applicability across high-performance computing, image and signal processing, machine learning inference and embedded systems, delivering customisable accelerators that combine high computational density with energy efficiency.

Research from Nature Portfolio

No recent Nature Portfolio content available.

High-Level Synthesis Techniques for FPGA Design publication trend

The graph below shows the total number of articles in high-level synthesis techniques for fpga design across all publications each year (not limited to Nature Index journals).

Technical terms

High-Level Synthesis (HLS): A methodology that converts algorithmic code in high-level languages into hardware descriptions synthesised for FPGA architectures.

Field-Programmable Gate Array (FPGA): A reconfigurable integrated circuit comprising an array of logic blocks and interconnects, programmable to implement custom digital circuits.

Design Space Exploration (DSE): The systematic evaluation of multiple hardware implementations generated by varying synthesis directives to optimise performance, area and power trade-offs.

Directives (Pragmas): Annotations inserted into high-level source code that guide the synthesis tool on loop unrolling, pipelining and resource bindings.

Dataflow Model: A representation that captures the movement of data between computational stages, used to schedule operations and manage on-chip buffering in HLS flows.

References

1. Towards Automatic High-Level Code Deployment on Reconfigurable Platforms: A Survey of High-Level Synthesis Tools and Toolchains. IEEE Access (2020).
2. Performance Modeling for FPGAs: Extending the Roofline Model with High‐Level Synthesis Tools. International Journal of Reconfigurable Computing (2013).
3. PyLog: An Algorithm-Centric Python-Based FPGA Programming and Synthesis Flow. IEEE Transactions on Computers (2021).
4. FPGA design space exploration for scientific HPC applications using a fast and accurate cost model based on roofline analysis. Journal of Parallel and Distributed Computing (2019).