Home›Industry News›HLS Goes Straight to FPGA Hard Blocks: Paper Proposes a Native Implementation Path
← Back to Industry News
Research

HLS Goes Straight to FPGA Hard Blocks: Paper Proposes a Native Implementation Path

ResearchHLS Goes Straight toFPGA Hard Blocks:Paper Proposes a…fpga.love
Illustration

High-level synthesis has long been criticized for being able to synthesize logic but struggling to touch hard blocks. An arXiv paper takes aim at this gap: Mohamed Bouaziz and four co-authors propose a path that generates native FPGA hard-block implementations directly from HLS, rather than first mapping to soft logic and then replacing it by hand. For engineers working on high-speed interfaces and boards, this reaches the layer the toolchain has been least willing to touch.

The Gap the Paper Targets

The paper is titled "Native FPGA Hard-Block Implementation from High-Level Synthesis," submitted on November 18, 2026, to arXiv's hardware architecture category, with Mohamed Bouaziz and four other authors. The full text is 88 KB and includes PDF, experimental HTML, and TeX source. The material does not specify the hard-block types, target devices, or process nodes. The real issue lies in the topic itself: HLS abstraction has always stopped at soft logic, and hard-block invocation usually requires going back to RTL or manual instantiation. The paper attempts to bring this step into the HLS flow.

The Engineering Cost of Hard-Block Access

The material does not disclose the methods, evaluation platforms, or result data used in the paper, so it is impossible to judge which hard blocks it covers or how far the support goes. The only verifiable facts are the submission date, the number of authors, and the direction indicated by the paper's title. This means what can be confirmed now is the research intent, not usability. For board design, hard blocks often determine interface rates and pin constraints. If HLS cannot natively express these constraints, the flow must still be closed out at the RTL layer. Whether the paper changes this can only be judged after the methods and experiments are made public.

QuestionMethodFinding
Illustration
Three takeaways
  • The paper argues for generating FPGA hard-block implementations directly from HLS
  • The material does not disclose hard-block types, target devices, or evaluation data
  • In the short term, it does not affect the existing hybrid flow of HLS plus RTL

What it means for us

For teams in China working on high-speed interfaces and boards, this paper is not yet a reason to switch solutions: there is no device support list, no hard-block types, and no reproducible experimental data, so toolchain selection cannot be discussed. The indirect impact is on the observation front: if HLS can truly invoke hard blocks natively, the workload for IP instantiation and constraint closure would shift forward from the RTL side, and the HLS tool gap among domestic FPGA vendors would also be compared on the same yardstick. For now, the more practical approach is to continue maintaining the division of labor in which HLS handles data paths and RTL closes hard blocks, and to evaluate once the paper's methods and open-source implementation land.

HLSFPGA hard blocksEDA toolchainarXiv
fpga.love

Subscribe to fpga.love

Daily Briefing · Weekly Digest · Monthly Deep Dive