X-HEEP: An Open-Source, Configurable and Extendible RISC-V Platform for TinyAI Applications

David Atienza; Giovanni Ansaloni; Miguel Pe\'on-Quir\'os; Pasquale Davide Schiavone; Simone Machetti

REVIEW 3 major objections 6 minor 35 references

X-HEEP: An Open-Source, Configurable and Extendible RISC-V Platform for TinyAI Applications

T0 review · 3 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read X-HEEP: an open-source RISC-V host whose 65 nm implementation costs 0.15 mm² and 29 µW leakage, yet accelerates early-exit TinyAI inference by 7.3× with 3.6× energy savings.

desk verdict A solid open-source RISC-V host platform with a useful early-exit accelerator demo, but the headline area/leakage and speedup numbers are pre-silicon estimates and need more configuration detail to be reproducible. read the letter →

arxiv 2508.16959 v1 pith:JCYA6N3Y submitted 2025-08-23 cs.AR

Simone Machetti , Pasquale Davide Schiavone , Giovanni Ansaloni , Miguel Peón-Quirós , David Atienza This is my paper · ORCID

classification cs.AR

keywords RISC-Vultra-low-poweredgecomputingTinyAIacceleratorinterfacenear-memoryearly-exitneuralnetworksopen-sourcehardwareheterogeneousSoC

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper introduces X-HEEP, an open-source RISC-V microcontroller platform whose central claim is that a highly configurable host for edge-AI accelerators can be almost free in area and power. In a 65 nm CMOS implementation the host occupies 0.15 mm² and leaks 29 µW, with most of that cost sitting in memory banks rather than control logic. Its eXtendible Accelerator InterFace (XAIF) bundles bus connections, DMA, interrupts, and power-control signals so accelerators with different requirements plug in without custom RTL. As a demonstration, pairing X-HEEP with a near-memory accelerator for early-exit neural networks yields up to 7.3× speedup and 3.6× energy savings versus CPU-only execution. A sympathetic reader would care because the platform aims to remove the integration overhead that usually blocks accelerator exploration in ultra-low-power systems.

What carries the argument

The eXtendible Accelerator InterFace (XAIF) is the central mechanism: a standardized set of configurable master/slave bus connections, DMA extensions, interrupt lines, and power-management signals that lets accelerators access shared memory and run their own life cycle through dynamic power control. Around it sits a set of SystemVerilog parameters that generate tailored RTL for core choice, memory banks, bus topology, and peripherals, plus an always-on power manager implementing clock gating, power gating, and memory retention. Together they make accelerator integration a plug-in operation rather than a hardware redesign, which is what gives the low area and leakage claims their force.

What would settle it

Tape out the 65 nm X-HEEP plus near-memory accelerator system, measure die area, leakage at 0.8 V, and run the same transformer and CNN early-exit benchmarks on the chip; the quantitative claims are falsified if measured area, leakage, speedup, or energy differ substantially from 0.15 mm², 29 µW, 7.3×, and 3.6×.

Watch

Extended reading notes

Core claim

The paper's central claim is that X-HEEP is a parameterized RISC-V host in which the core, memory size and banking, bus topology, and peripherals are all configurable at synthesis time, and whose standardized XAIF lets external accelerators attach as tightly coupled devices with access to DMA, interrupts, and dynamic power management. The authors assert this flexibility does not cost area or leakage: a 65 nm CMOS implementation at 300 MHz and 0.8 V occupies 0.15 mm² and burns 29 µW of leakage, which can be cut to 3 µW by power-gating unused blocks. They then show that a heterogeneous system built around X-HEEP plus the NM-Carus near-memory accelerator, running early-exit transformer and CNN models for seizure detection, reaches 5.4× and 7.3× kernel speedups and 3.6× and 3.4× energy improvements over the CPU-only baseline, depending on the model.

Load-bearing premise

The area, leakage, speedup, and energy numbers come from post-synthesis simulation with switching activity rather than measurement of fabricated silicon, so the quantitative case assumes the synthesis library and power-analysis flow predict real-chip behavior accurately.

Editorial extensions

If this is right

A designer can assemble an edge SoC with any of several RISC-V cores, a chosen memory hierarchy, bus topology, and only the needed peripherals, all from parameters rather than manual RTL edits.
Since memory banks account for 44% of area and 84% of leakage, scaling or power-gating memory is the dominant lever for meeting an edge power budget.
Accelerator teams that attach through XAIF inherit DMA, interrupts, and power control, eliminating glue logic and custom bus bridges.
Combining early-exit termination with near-memory acceleration can cut both latency and energy on this host, with up to 7.3× speedup and 3.6× energy improvement over CPU-only execution.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

If the post-synthesis numbers survive silicon, the host becomes nearly negligible in an edge SoC: its entire area is less than that of an additional memory bank, so design effort can concentrate on memory and accelerator optimization.
The same standardized XAIF could serve as a neutral mounting point for apples-to-apples accelerator comparisons under identical host, toolchain, and power-management conditions.
The 3 µW power-gated leakage points toward deeply duty-cycled sensing nodes that sleep almost indefinitely and wake only for short inference bursts, a regime the paper itself does not quantify.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Colleague,

I read the X-HEEP arXiv paper. The platform itself is a genuine engineering contribution: a configurable RISC-V host with a standardized accelerator interface (XAIF), multiple core options, memory/bus/peripheral configurability, and a power manager. The architecture is an evolution of the earlier X-HEEP papers [6,7], so the genuinely new piece here is the early-exit NM-Carus case study, which demonstrates the interface with a near-memory accelerator and reports speedup/energy gains on two adaptive network benchmarks.

The related-work section is fair and useful, covering PULPissimo, Cheshire, BlackParrot, OpenTitan, Chipyard, LiteX, and ESP with clear distinctions. The area and leakage breakdown showing memory dominance is informative and plausible, as is the claim that the host logic itself is low-overhead.

The soft spots are real but fixable. The headline numbers (0.15 mm², 29 µW, 7.3× speedup, 3.6× energy) come from post-synthesis simulation with switching-activity power analysis—no tape-out is reported. The stress-test note is on the mark: the paper omits the exact memory configuration (size, banks) and doesn't say whether the 0.15 mm² includes the pad ring, which matters because memory is 44% of area and 84% of leakage. A different SRAM config moves those numbers substantially. The speedup/energy figures are also best-case selections from a parameter sweep, with early-exit rates tuned to 73% and 82% at a real cost in F1. That's a co-design result, not a generic property of the host, and the paper should say so more prominently.

The self-citations to prior X-HEEP and NM-Carus work are appropriate, not a red flag. This is an incremental platform paper, and it does not overclaim in the main text beyond the abstract's 'implemented' phrasing.

Bottom line: this deserves a serious referee. The engineering value for the low-power RISC-V/TinyAI community is real, especially as an open-source reference. But I would push for revisions before acceptance: specify the configuration, give a commit hash or artifact link, and clearly label all quantitative claims as pre-silicon estimates. With those changes, it becomes a dependable building block for others.

Referee Report

3 major / 6 minor

Summary. The paper presents X-HEEP, an open-source RISC-V microcontroller platform with configurable core, memory, bus, peripherals, and an accelerator interface (XAIF), together with FPGA, ASIC, and mixed SystemC-RTL development flows. It reports a TSMC 65 nm synthesis result of 0.15 mm^2 area and 29 µW leakage, and demonstrates integration with the NM-Carus near-memory accelerator for early-exit transformer and CNN seizure-detection networks, claiming up to 7.3× performance speedup and 3.6× energy improvement over CPU-only execution.

Significance. If the quantitative claims hold, X-HEEP is a useful open infrastructure for TinyAI accelerator research: it is built on standard open IPs, is released as open source, supports FPGA and ASIC flows, and provides a structured accelerator interface used by several examples. The area and leakage breakdown is informative, and the early-exit demo is a plausible use case. However, the significance of the numerical claims is currently limited because they are pre-silicon estimates with unspecified configuration and methodology details, and the performance numbers come from tuned best-case operating points.

major comments (3)

[Abstract, §V, §VI-A] All headline quantitative results (0.15 mm^2, 29 µW leakage, 7.3× speedup, 3.6× energy gain) come from post-synthesis simulation with switching-activity power analysis, yet the abstract calls the platform 'implemented' and §V calls the simulated energy 'measured'. No tape-out or silicon measurement is reported, and the paper does not provide the synthesis library corner, SRAM compiler settings, supply/corner conditions, or the simulation flow (input stimuli, number of cycles, switching-activity generation) used to obtain these figures. Since the claims are only as trustworthy as the EDA estimates, the wording should be corrected and the missing methodology details reported.
[§VI-A, §III-C] The area and leakage figures are not reproducible as stated because the memory configuration is unspecified: the paper does not give the total memory size, number of banks, word width, or memory wrapper, and it does not state whether the 0.15 mm^2 footprint includes the pad ring and pad controller described in §III-C. Memory accounts for 44% of area and 84% of leakage, so these omissions can move the headline numbers substantially; at minimum the exact configuration and pad-ring treatment must be reported before the low-overhead claim can be evaluated.
[§V, §VI-B] The reported 7.3× and 3.6× improvements are selected from a parameter sweep over early-exit loss weight, entropy threshold, and exit placement, with the chosen operating points reducing F1 from 0.6223 to 0.53 (transformer) and from 0.57 to 0.49 (CNN). As presented, these are best-case tuned results, and the paper does not report the behavior across the swept configurations or any measure of variability (e.g., range, median, or sensitivity), nor does it compare the same workload on another host platform; accordingly the 'up to' claims overstate the typical benefit and should be accompanied by the full sweep or a sensitivity analysis.

minor comments (6)

[Title and §III-A] The title uses 'Extendible' while §III-A uses 'eXtendible'; choose one spelling consistently.
[§IV-B, reference [29]] Reference [29] (Terzano et al., 'Just TestIt! An SBST Approach To Automate System-Integration Testing') is cited for the Im2Col accelerator, but the title suggests a testing paper; please verify that this citation matches the described work.
[Figure 3] Figure 3 contains a stray duplicated 'CPU CPU' line in the caption area; remove it.
[§V] In §V, 'The measured energy consumption' should read 'the estimated energy consumption' because the values come from post-synthesis simulation, not from chip measurement.
[§VI-A] In §VI-A, 'the integration overhead introduced by X-HEEP is low' is unclear: state the reference point for 'overhead' (e.g., area/leakage of the platform excluding memory, or the accelerator-only system).
[§V] In §V, describe the input stimuli and cycle counts used for switching-activity extraction; without this, the power estimates cannot be reproduced.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the area/leakage and speedup/energy results are synthesis/simulation measurements using cited components, not derivations that assume their own conclusions.

full rationale

The central quantitative claims are (i) X-HEEP synthesized in TSMC 65nm occupies 0.15 mm2 and leaks 29 uW, and (ii) integrating the NM-Carus accelerator yields up to 7.3x speedup and 3.6x energy improvement. Both are empirical results from post-synthesis simulation and power analysis, as stated in Section V: 'Post-synthesis simulations are used to assess performance and extract switching activity, which is then fed into power analysis tools for accurate energy estimation.' No equation or derivation reduces the reported numbers to the paper's own assumptions. The NM-Carus accelerator is prior cited work [4], not re-derived here, and the dynamic-network models and datasets are external references ([33]–[35]). The early-exit thresholds are chosen through a parameter sweep to maximize early-exit rate while limiting F1 degradation; this is a hyperparameter selection procedure, not a fitted parameter renamed as a prediction. The platform's area and leakage breakdown in Section VI-A is independent of the accelerator integration and follows from the synthesized netlist and power analysis. The remaining concerns—pre-silicon simulation rather than measured silicon, unspecified memory configuration and pad-ring inclusion, and best-case tuned thresholds—are verification and reproducibility limitations, not circularity. Self-citations such as refs. [6], [7], [21], [23], and [28] describe the open-source platform and related accelerators, but they are not load-bearing evidence for the reported measurements and do not replace a derivation. No circular step can be exhibited from the paper's text.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The paper introduces no new physical or mathematical entities. The XAIF interface is a design feature, not a postulated entity. The quantitative results depend on several tuned hyperparameters (early-exit weights, thresholds, exit locations) and on standard hardware-simulation assumptions about synthesis and power analysis fidelity.

free parameters (5)

Early-exit loss weight (transformer) = 0.1
Selected through a parameter sweep to minimize energy while preserving accuracy. This tuning affects the 73% early-exit rate and the reported speedup.
Early-exit loss weight (CNN) = 0.01
Selected via the same sweep, yields 82% early-exit rate. The final speedup figures depend on this choice.
Entropy threshold (transformer) = 0.45
Part of the sweep that determines when early exit triggers. Higher threshold lowers accuracy but increases energy savings.
Entropy threshold (CNN) = 0.35
Chosen to balance F1 score and early-exit rate; directly influences the reported energy improvement.
Exit point location = After first encoder layer / first convolutional block
Hand-selected per model to minimize energy; not justified by a generalizable method.

assumptions (4)

domain assumption TSMC 65 nm synthesis library models are accurate for area, leakage, and timing
The paper reports 0.15 mm2 area and 29 uW leakage from synthesis, not from measured silicon. Accuracy of the foundry library models is assumed.
domain assumption Post-synthesis simulation with switching activity correctly estimates dynamic energy
The power analysis flow uses switching activity from simulations; the paper provides no validation against measured chip power.
domain assumption The seizure detection benchmark is representative of TinyAI workloads
The paper claims general applicability to TinyAI but evaluates only two models in one medical domain. The early-exit benefits may not transfer to other workloads.
domain assumption The NM-Carus accelerator's integer arithmetic is a fair comparison to the CPU baseline
The paper states NM-Carus targets integer arithmetic but does not specify whether the CPU baseline used integer or floating-point operations, which affects the speedup comparison.

how reviews work

0 comments

Cite this review

Pith. "Pith review of X-HEEP: An Open-Source, Configurable and Extendible RISC-V Platform for TinyAI Applications." pith.science (2026). https://pith.science/paper/JCYA6N3Y

@misc{pith2026250816959,
  author       = {Pith},
  title        = {Pith review of: X-HEEP: An Open-Source, Configurable and Extendible RISC-V Platform for TinyAI Applications},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JCYA6N3Y}},
  note         = {Machine review of arXiv:2508.16959}
}

read the original abstract

In this work, we present X-HEEP, an open-source, configurable, and extendible RISC-V platform for ultra-low-power edge applications (TinyAI). X-HEEP features the eXtendible Accelerator InterFace (XAIF), which enables seamless integration of accelerators with varying requirements along with an extensive internal configuration of cores, memory, bus, and peripherals. Moreover, it supports various development flows, including FPGA prototyping, ASIC implementation, and mixed SystemC-RTL modeling, enabling efficient exploration and optimization. Implemented in TSMC's 65 nm CMOS technology (300 MHz, 0.8 V), X-HEEP achieves a minimal footprint of only 0.15 mm2 and consumes just 29 uW of leakage power. As a demonstrator of the configurability and low overhead of X-HEEP as a host platform, we present a study integrating it with near-memory accelerators targeting early-exit dynamic network applications, achieving up to 7.3 x performance speedup and 3.6 x energy improvement on the resulting heterogeneous system compared to CPU-only execution.

Figures

Figures reproduced from arXiv: 2508.16959 by the authors.

**Figure 1.** X-HEEP architecture with integrated accelerators and co-processor. reducing power when idle. This feature supports energyefficient designs without compromising memory availability. Communication across the platform is managed by buses based on the open bus interface (OBI) [19]. The topology can be customized: A one-at-a-time configuration minimizes resource usage for simpler systems, whereas a fully connected cros… view at source ↗

**Figure 2.** Area and leakage distributions among the internal components of the X-HEEP platform. from 0.6223 to 0.53. The CNN model, configured with a weight of 0.01 and a threshold of 0.35, achieves an even higher early-exit rate of 82 %, with the F1 score decreasing from 0.57 to 0.49. To evaluate the benefits of the software strategies analyzed, we integrate the near-memory accelerator NM-Carus [4] with X-HEEP. The resulting … view at source ↗

**Figure 3.** Performance and energy improvements obtained by running the adaptive network benchmark with early exit and on the NM-Carus accelerator compared to the baseline CPU-only execution. demonstrating how our configurable and low-overhead host platform can support substantial gains in real-world edge scenarios. VIII. ACKNOWLEDGEMENTS The authors thank the entire X-HEEP team for their significant contributions to creating … view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 25 canonical work pages

[1]

A Hardware/Software Co-Design Vision for Deep Learning at the Edge

Flavio Ponzina et al. “A Hardware/Software Co-Design Vision for Deep Learning at the Edge”. In: IEEE Micro 42.6 (2022), pp. 48–54

2022
[2]

Mr. Wolf: An energy-precision scalable parallel ultra low power SoC for IoT edge processing

Antonio Pullini et al. “Mr. Wolf: An energy-precision scalable parallel ultra low power SoC for IoT edge processing”. In: IEEE Journal of Solid-State Circuits 54.7 (2019), pp. 1970–1981

2019
[3]

A multi-core reconfigurable architecture for ultra- low power bio-signal analysis

Loris Duch et al. “A multi-core reconfigurable architecture for ultra- low power bio-signal analysis”. In: IEEE BioCAS. 2016, pp. 416–419

work page 2016
[4]

Scalable and RISC-V Programmable Near- Memory Computing Architectures for Edge Nodes

Michele Caon et al. “Scalable and RISC-V Programmable Near- Memory Computing Architectures for Edge Nodes”. In: IEEE TETC (2025), pp. 1–15

work page 2025
[5]

ASIP design for motion estimation in video compression algorithms

Simone Machetti. “ASIP design for motion estimation in video compression algorithms”. In: Master Thesis, Polito (2018)

work page 2018
[6]

X-HEEP: An Open-Source, Configurable and Extendible RISC-V Microcontroller for the Exploration of Ultra-Low- Power Edge Accelerators

Simone Machetti et al. X-HEEP: An Open-Source, Configurable and Extendible RISC-V Microcontroller for the Exploration of Ultra-Low- Power Edge Accelerators. 2024. arXiv: 2401.05548

arXiv 2024
[7]

X-HEEP: An Open-Source, Config- urable and Extendible RISC-V Microcontroller

Pasquale Davide Schiavone et al. “X-HEEP: An Open-Source, Config- urable and Extendible RISC-V Microcontroller”. In: ACM CF. 2023, pp. 379–380

2023
[8]

Quentin: an Ultra-Low-Power PULPissimo SoC in 22nm FDX

Pasquale Davide Schiavone et al. “Quentin: an Ultra-Low-Power PULPissimo SoC in 22nm FDX”. In: (2018), pp. 1–3

work page 2018

Show all 35 references

[9]

Cheshire: A Lightweight, Linux-Capable RISC-V Host Platform for Domain-Specific Accelerator Plug-In

Alessandro Ottaviano et al. “Cheshire: A Lightweight, Linux-Capable RISC-V Host Platform for Domain-Specific Accelerator Plug-In”. In: IEEE TCAS (2023), pp. 1–1

2023
[10]

BlackParrot: An Agile Open-Source RISC-V Multicore for Accelerator SoCs

Daniel Petrisko et al. “BlackParrot: An Agile Open-Source RISC-V Multicore for Accelerator SoCs”. In: IEEE Micro 40.4 (2020), pp. 93– 102

2020
[11]

OpenTitan

LowRISC. OpenTitan. URL: https://github.com/lowRISC/opentitan
[12]

Chipyard: Integrated Design, Simulation, and Implementation Framework for Custom SoCs

Alon Amid et al. “Chipyard: Integrated Design, Simulation, and Implementation Framework for Custom SoCs”. In: IEEE Micro 40.4 (2020), pp. 10–21

2020
[13]

LiteX: an open-source SoC builder and library based on Migen Python DSL

Florent Kermarrec et al. LiteX: an open-source SoC builder and library based on Migen Python DSL . 2020. arXiv: 2005.02506

2020 arXiv
[14]

Agile SoC Development with Open ESP : Invited Paper

Paolo Mantovani et al. “Agile SoC Development with Open ESP : Invited Paper”. In: IEEE ICCAD. 2020, pp. 1–9

2020
[15]

X-HEEP: An Open-Source, Configurable and Extendible RISC-V Microcontroller for the Exploration of Ultra-Low- Power Edge Accelerators

Simone Machetti et al. “X-HEEP: An Open-Source, Configurable and Extendible RISC-V Microcontroller for the Exploration of Ultra-Low- Power Edge Accelerators”. In: EcoCloud Event 2023 (2023)

2023
[16]

Slow and steady wins the race? A comparison of ultra-low-power RISC-V cores for Internet-of-Things applications

Pasquale Davide Schiavone et al. “Slow and steady wins the race? A comparison of ultra-low-power RISC-V cores for Internet-of-Things applications”. In: IEEE PATMOS. 2017, pp. 1–8

2017
[17]

Near-threshold RISC-V core with DSP extensions for scalable IoT endpoint devices

Michael Gautschi et al. “Near-threshold RISC-V core with DSP extensions for scalable IoT endpoint devices”. In: IEEE VLSI 25.10 (2017), pp. 2700–2713

2017
[18]

URL: github.com/openhwgroup/core-v-xif

CORE-V-XIF. URL: github.com/openhwgroup/core-v-xif
[19]

URL: https://github.com/openhwgroup/ obi

Open Bus Interface Protocol . URL: https://github.com/openhwgroup/ obi
[20]

URL: https://github.com/olofk/fusesoc

FuseSoC. URL: https://github.com/olofk/fusesoc
[21]

FEMU: An Open-Source RISC-V Emulation Platform for the Exploration of Accelerator-based Edge Applications

Simone Machetti et al. “FEMU: An Open-Source RISC-V Emulation Platform for the Exploration of Accelerator-based Edge Applications”. In: DATE (2024)

2024
[22]

BLADE: An in-cache computing architecture for edge devices

William Andrew Simon et al. “BLADE: An in-cache computing architecture for edge devices”. In: IEEE Transactions on Computers 69.9 (2020), pp. 1349–1363

2020
[23]

HEEPocrates: An ultra-low-power RISC- V microcontroller for edge-computing healthcare applications

Simone Machetti et al. “HEEPocrates: An ultra-low-power RISC- V microcontroller for edge-computing healthcare applications”. In: Europractice Activity Report (2024)

2024
[24]

ARCANE: Adaptive RISC-V Cache Architecture for Near-memory Extensions

Vincenzo Petrolo et al. “ARCANE: Adaptive RISC-V Cache Architecture for Near-memory Extensions”. In: arXiv preprint arXiv:2504.02533 (2025)

2025 arXiv
[25]

HEEPstor: an Open-Hardware Co- design Framework for Quantized Machine Learning at the Edge

Pedro Palacios Almendros et al. “HEEPstor: an Open-Hardware Co- design Framework for Quantized Machine Learning at the Edge”. In: 2025

2025
[26]

Modular Design and Optimization of Biomedical Applications for Ultralow Power Heterogeneous Plat- forms

Elisabetta De Giovanni et al. “Modular Design and Optimization of Biomedical Applications for Ultralow Power Heterogeneous Plat- forms”. In: IEEE TCAD 39.11 (2020), pp. 3821–3832

2020
[27]

STRELA: STReaming ELAstic CGRA Acceler- ator for Embedded Systems

Daniel Vazquez et al. STRELA: STReaming ELAstic CGRA Acceler- ator for Embedded Systems . 2024. arXiv: 2404.12503

2024 arXiv
[28]

e-GPU: An Open-Source and Configurable RISC-V Graphic Processing Unit for TinyAI Applications

Simone Machetti et al. “e-GPU: An Open-Source and Configurable RISC-V Graphic Processing Unit for TinyAI Applications”. In: arXiv preprint arXiv:2505.08421 (2025)

2025
[29]

Just TestIt! An SBST Approach To Automate System-Integration Testing

Tommaso Terzano et al. Just TestIt! An SBST Approach To Automate System-Integration Testing. 2025. arXiv: 2504.07555

2025 arXiv
[30]

Quadrilatero: A RISC-V programmable matrix coprocessor for low-power edge applications

Danilo Cammarata et al. Quadrilatero: A RISC-V programmable matrix coprocessor for low-power edge applications . 2025. arXiv: 2504.07565

2025 arXiv
[31]

Increasing the Energy-Efficiency of Wearables Using Low-Precision Posit Arithmetic with PHEE

David Mallas ´en et al. Increasing the Energy-Efficiency of Wearables Using Low-Precision Posit Arithmetic with PHEE. 2025. arXiv: 2501. 18253

2025
[32]

ATHOS: A Hybrid Accelerator for PQC CRYSTALS-Algorithms Exploiting New CV-X-IF Interface

Alessandra Dolmeta, Maurizio Martina, and Guido Masera. “ATHOS: A Hybrid Accelerator for PQC CRYSTALS-Algorithms Exploiting New CV-X-IF Interface”. In: IEEE Access 12 (2024), pp. 182340– 182352

2024
[33]

MetaWearS: A Shortcut in Wearable Systems Lifecycle with Only a Few Shots

Alireza Amirshahi et al. MetaWearS: A Shortcut in Wearable Systems Lifecycle with Only a Few Shots . 2024. arXiv: 2408.01988

2024 arXiv
[34]

BiomedBench: A benchmark suite of TinyML biomedical applications for low-power wearables

Dimitrios Samakovlis et al. “BiomedBench: A benchmark suite of TinyML biomedical applications for low-power wearables”. In: IEEE Design and Test (2024), pp. 1–1

2024
[35]

ACE: Automated Optimization Towards Iterative Classification in Edge Health Monitors

Yuxuan Wang et al. “ACE: Automated Optimization Towards Iterative Classification in Edge Health Monitors”. In: TBioCAS 19.1 (2025), pp. 82–92

2025

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [1]

A Hardware/Software Co-Design Vision for Deep Learning at the Edge

Flavio Ponzina et al. “A Hardware/Software Co-Design Vision for Deep Learning at the Edge”. In: IEEE Micro 42.6 (2022), pp. 48–54

2022

[2] [2]

Mr. Wolf: An energy-precision scalable parallel ultra low power SoC for IoT edge processing

Antonio Pullini et al. “Mr. Wolf: An energy-precision scalable parallel ultra low power SoC for IoT edge processing”. In: IEEE Journal of Solid-State Circuits 54.7 (2019), pp. 1970–1981

2019

[3] [3]

A multi-core reconfigurable architecture for ultra- low power bio-signal analysis

Loris Duch et al. “A multi-core reconfigurable architecture for ultra- low power bio-signal analysis”. In: IEEE BioCAS. 2016, pp. 416–419

work page 2016

[4] [4]

Scalable and RISC-V Programmable Near- Memory Computing Architectures for Edge Nodes

Michele Caon et al. “Scalable and RISC-V Programmable Near- Memory Computing Architectures for Edge Nodes”. In: IEEE TETC (2025), pp. 1–15

work page 2025

[5] [5]

ASIP design for motion estimation in video compression algorithms

Simone Machetti. “ASIP design for motion estimation in video compression algorithms”. In: Master Thesis, Polito (2018)

work page 2018

[6] [6]

X-HEEP: An Open-Source, Configurable and Extendible RISC-V Microcontroller for the Exploration of Ultra-Low- Power Edge Accelerators

Simone Machetti et al. X-HEEP: An Open-Source, Configurable and Extendible RISC-V Microcontroller for the Exploration of Ultra-Low- Power Edge Accelerators. 2024. arXiv: 2401.05548

arXiv 2024

[7] [7]

X-HEEP: An Open-Source, Config- urable and Extendible RISC-V Microcontroller

Pasquale Davide Schiavone et al. “X-HEEP: An Open-Source, Config- urable and Extendible RISC-V Microcontroller”. In: ACM CF. 2023, pp. 379–380

2023

[8] [8]

Quentin: an Ultra-Low-Power PULPissimo SoC in 22nm FDX

Pasquale Davide Schiavone et al. “Quentin: an Ultra-Low-Power PULPissimo SoC in 22nm FDX”. In: (2018), pp. 1–3

work page 2018