Pith. sign in

REVIEW 2 major objections 5 minor 73 references

Thermal Modeling and Optimal Allocation of Avionics Safety-critical Tasks on Heterogeneous MPSoCs

T0 review · 2 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Software-only task placement can cut MPSoC temperature by up to 5.5 °C in safety-critical avionics workloads.

desk verdict New problem formulation with real-hardware validation, but the temperature superiority claim needs more statistical support. read the letter →

arxiv 2505.22214 v2 pith:GZ6WC2YV submitted 2025-05-28 cs.SE

Ondřej Benedikt , Michal Sojka , Přemysl Šůcha , Pavel Zaykov , Zdeněk Hanzálek This is my paper
classification cs.SE
keywords thermal-awareschedulingARINC653temporalisolationMPSoCintegerlinearprogrammingpowermodelingheterogeneousmulticoreavionics
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 argues that the way safety-critical avionics tasks are assigned to cores and to ARINC-653 temporal isolation windows can measurably change how hot an MPSoC runs, even when clock frequencies are fixed and dynamic voltage/frequency scaling is forbidden. Its central claim is that an integer-linear-programming optimizer built around an empirical Sum-Max power model, called ILP-SM, finds allocations that reduce steady-state temperature more consistently than six competing methods on three real platforms, with up to 5.5 °C difference against the worst method. The same paper proves the underlying allocation problem is weakly NP-hard and shows that fixing window lengths makes it polynomially solvable through a Minimum Cost Flow formulation. If true, this gives avionics designers a software-only thermal management lever that could reduce reliance on heavy or costly external cooling.

What carries the argument

The load-bearing object is the Sum-Max power model, which predicts a window's average power as the sum over clusters of per-task dynamic coefficients scaled by duty cycle, plus the largest static offset among tasks running in the window, plus idle power. A large-constant linearization turns the max term into linear constraints, so the whole allocation problem becomes an integer linear program whose objective is estimated average power over the major frame. The argument is carried by the combination of this linearized objective with the ARINC-653 window constraints, together with the steady-state single-output thermal relation that links average power to temperature linearly.

What would settle it

Run the same benchmark suite with a 100-second major frame instead of a one-second frame and compare measured steady-state temperatures of the ILP-SM and ILP-IDLE-MAX schedules; if the gap disappears or the ordering reverses, the linear average-power-to-temperature assumption that carries the method is violated in that regime. A more direct check is to compare the schedules' ranking by measured big-cluster temperature against their ranking by average power across many workloads with sub-second frames.

Watch

Extended reading notes

Core claim

The central claim is that minimizing a carefully chosen estimate of average power consumption, specifically the empirical Sum-Max power model, is the right objective for minimizing steady-state on-chip temperature under ARINC-653 temporal isolation. The paper reports that ILP-SM outperformed all other tested methods on all three platforms in nearly every instance, achieving up to 5.5 °C lower steady-state temperature than the worst uninformed idle-time optimizer and saving up to 1.6 °C over a simple greedy heuristic. A secondary finding is that the more accurate linear-regression power model led to worse optimization outcomes than the less accurate but consistently pessimistic Sum-Max model, which the authors attribute to the latter preserving the ordering between schedules even while overestimating power. The paper also establishes the formal hardness of the allocation problem and identifies a polynomial-time subproblem.

Load-bearing premise

The argument depends on a steady-state, single-output thermal model in which average power maps linearly to temperature near the big cluster, so that minimizing estimated average power also minimizes measured temperature; the paper validates this mainly for short major frames around one second, and longer frames show transient heating and cooling that this model ignores.

Editorial extensions

If this is right

  • On the three tested platforms, switching from the worst uninformed schedule to ILP-SM lowers steady-state temperature by up to 5.5 °C, a 22 % reduction relative to ambient temperature.
  • The best-performing methods all use the Sum-Max model even though the linear-regression model has lower prediction error, so model-optimizer harmony matters more than raw model accuracy.
  • A genetic-algorithm variant using the same Sum-Max model is nearly as good as ILP-SM and scales better, and it does not require a commercial solver.
  • Fixing window lengths renders the decision problem polynomial via Minimum Cost Flow, which offers a route to scalable decomposition methods for larger task sets.

Reading between the lines

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

  • Editorial inference: the ordering-preserving property of the Sum-Max model may transfer to other heterogeneous platforms, but only if the per-task coefficients are re-measured; the 5.5 °C gap is platform- and workload-specific.
  • Editorial inference: if major frames grow to tens of seconds, the steady-state assumption weakens and the temperature ranking of schedules may change; the paper's own 100-second experiments suggest such a regime exists.
  • Editorial inference: the minimum-cost-flow subproblem could be embedded in a logic-based Benders decomposition to scale beyond the tested instances, which the authors mention as future work.
  • Editorial inference: the same design pattern, choose an upper-bound power model that preserves schedule ordering, could apply to other safety-critical standards with temporal partitioning, such as automotive or industrial systems.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. The paper studies offline thermal-aware allocation of periodic safety-critical tasks on heterogeneous MPSoCs running at fixed clock frequency under ARINC-653 temporal isolation. It formalizes an ARINC allocation problem, proves weak NP-hardness by reduction from PARTITION, and shows that the fixed-window-length variant is solvable in polynomial time via minimum-cost flow. It then introduces three power models (empirical Sum-Max, linear regression, and an upper-bound variant LR-UB) and integrates them into several optimizers, including ILP, QP, genetic-algorithm black-box search, a greedy heuristic, and two uninformed idle-time ILP formulations. The experimental evaluation on three platforms (i.MX8 MEK, i.MX8 Ixora, TX2) measures both power and steady-state temperature, reporting that the ILP-SM method, which minimizes the Sum-Max power model, achieves the lowest temperature among all compared methods, with up to 5.5 °C difference from the worst method.

Significance. If the empirical claims hold, this is a practically valuable result: it shows that software-only task placement under ARINC-653 isolation can reduce MPSoC operating temperature on real hardware, with a sober trade-off analysis showing that a simpler but consistently biased power model (SM) outperforms a more accurate regression model (LR) when integrated into optimization. The complexity results are clean and strengthen the paper's contribution by delineating the NP-hard core from the polynomially solvable fixed-window subproblem. The paper also has notable reproducibility strengths: the authors release their scheduler, measured data, and optimization code, and they report a substantial measurement campaign (three platforms, roughly 100 hours of power measurements and 42 hours of schedule evaluation per platform). The main weakness is that the headline temperature-based superiority claim rests on only six random instances per condition without error bars or significance testing, and the measured temperature differences among the best methods are small relative to the reported spatial and sensor variability.

major comments (2)
  1. [Section 7.5, Figure 13] The central claim that ILP-SM outperforms all other methods in steady-state temperature is based on averages over only six randomly generated instances per workload type per platform, with no error bars, confidence intervals, or significance tests. The reported advantages over the HEUR baseline (1.6 °C, 1.3 °C, and 0.6 °C on the three platforms) are small relative to the per-instance scatter visible in the power-offset data of Table 5 (for example, on TX2 mixed workloads, HEUR's power offset is equal to or lower than ILP-SM's in instances 3 and 5). Provide per-schedule measured temperature data or paired significance tests across the six instances to support the claim that ILP-SM is statistically better in temperature, not merely better on average.
  2. [Section 7.2 and Eq. (3)] The linear relation between average power and steady-state temperature is validated in Figure 10 over a wide power span (roughly 2-12 W), but the differences among optimized schedules are only about 0.2-1.5 W in average power (Table 5). Since Figure 13 reports directly measured temperatures, the evaluation is not circular, but the manuscript does not demonstrate that these small power offsets translate into measured temperature differences that are distinguishable from measurement noise. This is compounded by Section 7.1.2, which reports a ~1 °C spatial temperature difference between stressing the big and little clusters at comparable total power; the single-output model in Eq. (3) ignores this spatial degree of freedom, and 1 °C is the same order as the claimed 0.6-1.6 °C advantages over HEUR. A direct per-schedule power-versus-temperature scatter for the evaluated schedules, or per-schedule temperature tables, would substantiate the temperature-based ranking.
minor comments (5)
  1. [Section 8] The percentage reductions reported in the conclusion ("up to 16%, 14% and 10%") are inconsistent with Section 7.5, which reports 22%, 19.6%, and 14.4% relative to ambient temperature for the same worst-to-best comparison. Clarify the basis for each percentage.
  2. [Section 7.4] The statement that the SM model "steadily provides an upper bound on the measured power consumption" is stronger than the evidence presented: Table 3 reports mean absolute error but not the frequency or magnitude of underestimation, and no formal bound is proven. Either soften the wording or quantify the proportion of instances where SM underestimates the measured power.
  3. [Appendix D] The schedule illustration labels one method as "ILP-SM-I" while the main text and tables use "ILP-SM"; please make the notation consistent.
  4. [Throughout] There are several typos, including "uniformed heuristics" (should be "uninformed heuristics") in Section 1, "major fame" (should be "major frame") in Section 7.1.1, and "I.MX8 MEX" (should be "I.MX8 MEK") in Section 7.4.
  5. [Figure 14] The legend entry "BB-[SM/LR]" is not defined in the caption or text; specify that it denotes both BB-SM and BB-LR.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the optimization comparison is evaluated on measured physical temperature against independently fitted power models.

full rationale

I found no circularity in the claimed derivation chain. The SM and LR power model coefficients are fitted from measured power data in Section 7.3, and the model predictions are then checked against held-out measured power in Section 7.4, not against the training data. The central comparison in Section 7.5 is based on measured steady-state temperature on three physical platforms, so the claim that ILP-SM outperforms other methods in reducing temperature is not an evaluation of the fitted model against its own fit. Equation (3) is a modeling assumption that relates average power to steady-state temperature; it could be under-validated for the small power differences among optimized schedules, but that is an external-validity or correctness concern rather than a circularity. The self-citations to the authors' earlier work [27] for the SM model and to [39] for the testbed are used as prior context, and the SM model is fully re-derived and empirically fitted within this paper, so no load-bearing argument reduces to an unverified self-citation.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The core engineering rests on measured power coefficients and a steady-state single-node thermal approximation; no new physical entities are introduced.

free parameters (2)
  • SM task coefficients a_i,k and o_i,k = Tables B.7/B.8, e.g., canrdr-4M on I.MX8 MEK little cluster: a=0.41, o=1.36
    Fitted from per-kernel power measurements on each platform; these coefficients enter the objective of ILP-SM and BB-SM.
  • LR regression coefficients beta_k,r = Table 2, e.g., I.MX8 MEK: beta_1,1=1.205, beta_2,1=0.270, beta_1,2=0.969, beta_2,2=0.456
    Obtained by linear regression on 1000 random 1-second intervals per platform; used in QP-LR-UB and BB-LR.
assumptions (5)
  • domain assumption Steady-state, single-output thermal model: T = P/B + G*Tamb/B (Eq. 3)
    Assumes workloads have periods much shorter than thermal time constants and that a single node (big cluster zone) represents chip temperature; justified in Section 7.1 for one workload and 1s frames.
  • domain assumption Average power is linearly related to steady-state temperature and leakage power is negligible
    Used to replace temperature with power in optimization (Section 4.1); the paper notes leakage is ignored and Figure 10 shows strong correlation, but no rigorous bound is given.
  • domain assumption Tasks are independent, non-preemptive, periodic, share a common period h, and have deadline h
    Problem statement in Section 3.2; standard in ARINC-653 partition scheduling but limits applicability to more general task models.
  • domain assumption DVFS is forbidden and clock frequencies are fixed
    Motivated by safety certification (Section 1, [18,23]); this removes frequency scaling as a control knob.
  • ad hoc to paper The Sum-Max model's structural form (z*a_i,k + o_i,k + P_idle for parallel instances) is valid
    Proposed by the authors in [27]; the max-over-tasks o_i,k term is a modeling assumption not derived from first principles.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Thermal Modeling and Optimal Allocation of Avionics Safety-critical Tasks on Heterogeneous MPSoCs." pith.science (2026). https://pith.science/paper/GZ6WC2YV

@misc{pith2026250522214,
  author       = {Pith},
  title        = {Pith review of: Thermal Modeling and Optimal Allocation of Avionics Safety-critical Tasks on Heterogeneous MPSoCs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GZ6WC2YV}},
  note         = {Machine review of arXiv:2505.22214}
}
read the original abstract

Multi-Processor Systems-on-Chip (MPSoC) can deliver high performance needed in many industrial domains, including aerospace. However, their high power consumption, combined with avionics safety standards, brings new thermal management challenges. This paper investigates techniques for offline thermal-aware allocation of periodic tasks on heterogeneous MPSoCs running at a fixed clock frequency, as required in avionics. The goal is to find the assignment of tasks to (i) cores and (ii) temporal isolation windows while minimizing the MPSoC temperature. To achieve that, we propose and analyze three power models, and integrate them within several novel optimization approaches based on heuristics, a black-box optimizer, and Integer Linear Programming (ILP). We perform the experimental evaluation on three popular MPSoC platforms (NXP i.MX8QM MEK, NXP i.MX8QM Ixora, NVIDIA TX2) and observe a difference of up to 5.5{\deg}C among the tested methods (corresponding to a 22% reduction w.r.t. the ambient temperature). We also show that our method, integrating the empirical power model with the ILP, outperforms the other methods on all tested platforms.

Figures

Figures reproduced from arXiv: 2505.22214 by the authors.

Figure 1
Figure 1. Three steps (bechmarking, optimization and evaluation) towards thermally efficient scheduling. [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. Schedule of seven tasks on six cores and three windows [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗
Figure 3
Figure 3. Illustration of processing-idling intervals needed for the [PITH_FULL_IMAGE:figures/full_fig_p011_3.png] view at source ↗
Figures from the paper (11 more)
Figure 4
Figure 4. Figure 4: Illustration of a simplified window with a single processing-idling interval only, as used by [PITH_FULL_IMAGE:figures/full_fig_p013_4.png]
Figure 5
Figure 5. Figure 5: Embedded platforms used for the evaluation: (a) I.MX8QuadMax Multisensory Enablement Kit by NXP [PITH_FULL_IMAGE:figures/full_fig_p017_5.png]
Figure 6
Figure 6. Figure 6: Relative speedup on a CPU from the high-performing cluster. [PITH_FULL_IMAGE:figures/full_fig_p019_6.png]
Figure 7
Figure 7. Figure 7: Influence of the on-chip temperature near the high-performing cluster on the major frame length for three instances [PITH_FULL_IMAGE:figures/full_fig_p020_7.png]
Figure 8
Figure 8. Figure 8: Temperatures obtained for pntrch-4M-100s from on-chip sensors near little and big cluster thermal zones. 20 [PITH_FULL_IMAGE:figures/full_fig_p020_8.png]
Figure 9
Figure 9. Figure 9: Spatial on-chip temperature T(x, y) on the left and hot spots h(x, y) on the right of I.MX8 Ixora with little (A53) cluster stressed at the top, and big (A72) cluster stressed at the bottom. platform, we observe that both thermal zones report the same value. This might…
Figure 10
Figure 10. Figure 10: Average power and steady-state temperature (measured, at the thermal zone near the big cluster) of various [PITH_FULL_IMAGE:figures/full_fig_p022_10.png]
Figure 11
Figure 11. Figure 11: Values of task characteristics coefficients [PITH_FULL_IMAGE:figures/full_fig_p023_11.png]
Figure 12
Figure 12. Figure 12: Measured and predicted power consumption of 1000 testing instances (mixed workload windows); instances are [PITH_FULL_IMAGE:figures/full_fig_p025_12.png]
Figure 13
Figure 13. Figure 13: Average difference between the measured steady-state temperature [PITH_FULL_IMAGE:figures/full_fig_p027_13.png]
Figure 14
Figure 14. Figure 14: Average computation time of different methods w.r.t. the instance size [PITH_FULL_IMAGE:figures/full_fig_p028_14.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

73 extracted references · 59 canonical work pages

  1. [1]

    C. B. Watkins, R. Walter, Transitioning from federated avionics architectures to Integrated Modular Avionics, in: 2007 IEEE/AIAA 26th Digital Avionics Systems Conference, 2007, pp. 2.A.1–1–2.A.1–10.doi:10.1109/DASC.2007.4391842. URLhttps://ieeexplore.ieee.org/abstract/document/4391842

    work page arXiv 2007

  2. [2]

    Avionics Application Software Standard Interface, Part 1, Required Services, ARINC Specification 653 Part 1 Supplement 4 (653P1-4) (2015)

    work page 2015

  3. [3]

    P. J. Prisaznuk, Arinc 653 role in integrated modular avionics (ima), in: 2008 IEEE/AIAA 27th Digital Avionics Systems Conference, 2008, pp. 1.E.5–1–1.E.5–10.doi:10.1109/DASC.2008.4702770

    work page arXiv 2008

  4. [4]

    A. T. A. Dugo, J.-B. Lefoul, F. G. De Magalhaes, D. Assal, G. Nicolescu, Cache locking content selection algorithms for ARINC-653 compliant RTOS, ACM Trans. Embed. Comput. Syst. 18 (5s).doi:10.1145/3358196. URLhttps://doi.org/10.1145/3358196

    work page doi:10.1145/3358196

  5. [5]

    P. Han, Z. Zhai, B. Nielsen, U. Nyman, Model-based optimization of ARINC-653partition scheduling, International Journal on Software Tools for Technology Transfer 23 (2021) 1–20.doi:10.1007/s10009-020-00597-6

    work page doi:10.1007/s10009-020-00597-6 2021

  6. [6]

    Potteiger, A

    B. Potteiger, A. Dubey, F. Cai, X. Koutsoukos, Z. Zhang, Moving target defense for the security and resilience of mixed time and event triggered cyber–physical systems, Journal of Systems Architecture 125 (2022) 102420,https://www. sciencedirect.com/science/article/pii/S1383762122000212.doi:10.1016/j.sysarc.2022.102420

    arXiv 2022

  7. [7]

    C. E. Salloum, M. Elshuber, O. H¨ oftberger, H. Isakovic, A. Wasicek, The across mpsoc – a new generation of multi-core processors designed for safety-critical embedded systems, in: 2012 15th Euromicro Conference on Digital System Design, 2012, pp. 105–113.doi:10.1109/DSD.2012.126

    work page doi:10.1109/dsd.2012.126 2012

  8. [8]

    Lentaris, K

    G. Lentaris, K. Maragos, I. Stratakos, L. Papadopoulos, O. Papanikolaou, D. Soudris, M. Lourakis, X. Zabulis, D. Gonzalez-Arjona, G. Furano, High-performance embedded computing in space: Evaluation of platforms for vision- based navigation, Journal of Aerospace Information Systems 15 (4) (2018) 178–192.arXiv:https://doi.org/10.2514/1. I010555,doi:10.2514/...

    doi:10.2514/1 2018

Show all 73 references
  1. [9]

    Lakshminarayanan, N

    V. Lakshminarayanan, N. Sriraam, The effect of temperature on the reliability of electronic components, in: 2014 IEEE International Conference on Electronics, Computing and Communication Technologies (CONECCT), 2014, pp. 1–6.doi: 10.1109/CONECCT.2014.6740182

    2014

  2. [11]

    Zhuravlev, J

    S. Zhuravlev, J. C. Saez, S. Blagodurov, A. Fedorova, M. Prieto, Survey of energy-cognizant scheduling techniques, IEEE Transactions on Parallel and Distributed Systems 24 (7) (2013) 1447–1464.doi:10.1109/TPDS.2012.20

    2013 doi

  3. [12]

    K. Pi, S. Ghosh, K. Khan, P. Bikki, A fully differential operational amplifier with slew rate enhancer and adaptive bias for ultra low power, Journal of Low Power Electronics 13 (2017) 67–75.doi:10.1166/jolpe.2017.1467

    2017

  4. [13]

    Huang, S

    W. Huang, S. Ghosh, S. Velusamy, K. Sankaranarayanan, K. Skadron, M. Stan, Hotspot: A compact thermal modeling methodology for early-stage vlsi design, IEEE Transactions on Very Large Scale Integration Systems - VLSI 14 (2006) 501–513.doi:10.1109/TVLSI.2006.876103

    2006

  5. [14]

    C. Yoon, S. Lee, Y. Choi, R. Ha, H. Cha, Accurate power modeling of modern mobile application processors, Journal of Systems Architecture 81 (2017) 17–31.doi:https://doi.org/10.1016/j.sysarc.2017.10.001

    2017 doi

  6. [15]

    J. Zhou, T. Wei, M. Chen, J. Yan, X. S. Hu, Y. Ma, Thermal-Aware Task Scheduling for Energy Minimization in Heterogeneous Real-Time MPSoC Systems, IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems 35 (8) (2016) 1269–1282.doi:10.1109/TCAD.2015.2501286

    2016

  7. [16]

    Chien, R.-G

    T.-H. Chien, R.-G. Chang, A thermal-aware scheduling for multicore architectures, Journal of Systems Architecture 62 (2016) 54–62.doi:https://doi.org/10.1016/j.sysarc.2015.12.003

    2016 doi

  8. [17]

    J. Zhou, J. Yan, K. Cao, Y. Tan, T. Wei, M. Chen, G. Zhang, X. Chen, S. Hu, Thermal-aware correlated two-level scheduling of real-time tasks with reduced processor energy on heterogeneous mpsocs, Journal of Systems Architecture 82 (2018) 1–11.doi:https://doi.org/10.1016/j.sysa...

    2018 doi

  9. [18]

    Z. Wu, L. Han, J. Liu, Y. Robert, F. Vivien, Energy-aware mapping and scheduling strategies for real-time workflows under reliability constraints, Journal of Parallel and Distributed Computing 176 (2023) 1–16.doi:10.1016/j.jpdc.2023.02.004. URLhttps://www.sciencedirect.com/sci...

    2023 doi

  10. [19]

    H. Xu, B. Zhang, C. Pan, K. Li, Energy-efficient triple modular redundancy scheduling on heterogeneous multi-core real- time systems, Journal of Parallel and Distributed Computing 191 (2024) 104915.doi:10.1016/j.jpdc.2024.104915. URLhttps://www.sciencedirect.com/science/articl...

    2024

  11. [20]

    Schieber, B

    B. Schieber, B. Samineni, S. Vahidi, Interweaving Real-Time Jobs with Energy Harvesting to Maximize Throughput, in: C.-C. Lin, B. M. T. Lin, G. Liotta (Eds.), W ALCOM: Algorithms and Computation, Springer Nature Switzerland, Cham, 2023, pp. 305–316.doi:10.1007/978-3-031-27051-2_26

    2023 doi

  12. [21]

    X. Li, Z. Li, W. Zhou, Z. Duan, Accurate on-chip temperature sensing for multicore processors using embedded thermal sensors, IEEE Transactions on Very Large Scale Integration (VLSI) Systems 28 (11) (2020) 2328–2341.doi:10.1109/ TVLSI.2020.3012833

    2020

  13. [22]

    Sadiqbatcha, J

    S. Sadiqbatcha, J. Zhang, H. Amrouch, S. X.-D. Tan, Real-time full-chip thermal tracking: A post-silicon, machine learning perspective, IEEE Transactions on Computers 71 (6) (2022) 1411–1424.doi:10.1109/TC.2021.3086112

    2022

  14. [23]

    Fakih, A

    M. Fakih, A. Lenz, M. Azkarate-Askasua, J. Coronel, A. Crespo, S. Davidmann, J. C. Diaz Garcia, N. G. Romero, K. Gr¨ uttner, S. Schreiner, R. Seyyedi, R. Obermaisser, A. Maleki, J. ¨Oberg, M. T. Mohammadat, J. P´ erez-Cerrolaza, I. Sander, I. S¨ oderquist, SAFEPOWER project: A...

    2017 doi

  15. [24]

    URLhttps://www.nxp.com/design/development-boards/i-mx-evaluation-and-development-boards/ i-mx-8quadmax-multisensory-enablement-kit-mek:MCIMX8QM-CPU

    NXP, i.MX 8QuadMax/QuadPlus Multisensory Enablement Kit (2021). URLhttps://www.nxp.com/design/development-boards/i-mx-evaluation-and-development-boards/ i-mx-8quadmax-multisensory-enablement-kit-mek:MCIMX8QM-CPU

    2021

  16. [25]

    NXP, I.MX 8QuadMax Applications Processor (Sep. 2021). URLhttps://www.nxp.com/webapp/Download?colCode=IMX8QMRM

    2021

  17. [26]

    URLhttps://developer.nvidia.com/embedded/jetson-tx2-developer-kit

    NVIDIA, Harness ai at the edge with the jetson tx2 developer kit (2022). URLhttps://developer.nvidia.com/embedded/jetson-tx2-developer-kit

    2022

  18. [27]

    Benedikt, M

    O. Benedikt, M. Sojka, P. Zaykov, D. Hornof, M. Kafka, P. ˇS ˚ ucha, Z. Hanz´ alek, Thermal-aware scheduling for mpsoc in the avionics domain: Tooling and initial results, in: 2021 IEEE 27th International Conference on Embedded and Real-Time Computing Systems and Applications ...

    2021

  19. [28]

    Tama¸ s-Selicean, P

    D. Tama¸ s-Selicean, P. Pop, Optimization of time-partitions for mixed-criticality real-time distributed embedded systems, in: 2011 14th IEEE International Symposium on Object/Component/Service-Oriented Real-Time Distributed Computing Workshops, 2011, pp. 1–10.doi:10.1109/ISOR...

    2011 doi

  20. [29]

    T˘ ama¸ s-Selicean, P

    D. T˘ ama¸ s-Selicean, P. Pop, Design optimization of mixed-criticality real-time embedded systems, ACM Trans. Embed. Comput. Syst. 14 (3).doi:10.1145/2700103. URLhttps://doi.org/10.1145/2700103

    doi

  21. [30]

    Waszniowski, J

    L. Waszniowski, J. Kr´ akora, Z. Hanz´ alek, Case study on distributed and fault tolerant system modeling based on timed automata, Journal of Systems and Software 82 (10) (2009) 1678–1694, sI: YAU.doi:https://doi.org/10.1016/j.jss. 2009.04.042. URLhttps://www.sciencedirect.com...

    2009 doi

  22. [31]

    J. Chen, C. Du, P. Han, Scheduling independent partitions in integrated modular avionics systems, PLOS ONE 11 (12) (2016) 1–18.doi:10.1371/journal.pone.0168064. URLhttps://doi.org/10.1371/journal.pone.0168064

    2016 doi

  23. [32]

    Chen, C.-F

    J.-J. Chen, C.-F. Kuo, Energy-Efficient Scheduling for Real-Time Systems on Dynamic Voltage Scaling (DVS) Platforms, in: 13th IEEE International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA 2007), 2007, pp. 28–38.doi:10.1109/RTCSA.2007.37. URL...

    2007

  24. [33]

    Fisher, J.-J

    N. Fisher, J.-J. Chen, S. Wang, L. Thiele, Thermal-aware global real-time scheduling and analysis on multicore systems, Journal of Systems Architecture 57 (5) (2011) 547–560.doi:10.1016/j.sysarc.2010.09.010

    2011 doi

  25. [34]

    M. E. T. Gerards, J. L. Hurink, P. K. F. H¨ olzenspies, A survey of offline algorithms for energy minimization under deadline constraints, Journal of Scheduling 19 (1) (2016) 3–19.doi:10.1007/s10951-015-0463-8

    2016 doi

  26. [35]

    Perez Rodriguez, P

    J. Perez Rodriguez, P. Meumeu Yomsi, Thermal-aware schedulability analysis for fixed-priority non-preemptive real-time systems, in: 2019 IEEE Real-Time Systems Symposium (RTSS), 2019, pp. 154–166.doi:10.1109/RTSS46320.2019.00024

    2019

  27. [36]

    URLhttp://developer.nvidia.com/embedded/dlc/jetson-tx2-series-thermal-design-guide

    NVIDIA, Jetson TX2 Series Thermal Design Guide (TDG-09420-001 v1.0) (2019). URLhttp://developer.nvidia.com/embedded/dlc/jetson-tx2-series-thermal-design-guide

    2019

  28. [37]

    Manna, P

    K. Manna, P. Mukherjee, S. Chattopadhyay, I. Sengupta, Thermal-Aware Application Mapping Strategy for Network-on- Chip Based System Design, IEEE Transactions on Computers 67 (4) (2018) 528–542.doi:10.1109/TC.2017.2770130

    2018

  29. [38]

    Kanduri, M

    A. Kanduri, M. Haghbayan, A. M. Rahmani, M. Shafique, A. Jantsch, P. Liljeberg, adBoost: Thermal Aware Performance Boosting Through Dark Silicon Patterning, IEEE Transactions on Computers 67 (8) (2018) 1062–1077.doi:10.1109/TC. 2018.2805683

    2018

  30. [39]

    Sojka, O

    M. Sojka, O. Benedikt, Z. Hanz´ alek, P. Zaykov, Testbed for thermal and performance analysis in mpsoc systems, in: 2020 15th Conference on Computer Science and Information Systems (FedCSIS), 2020, pp. 683–692.doi:10.15439/2020F174

    2020 doi

  31. [40]

    G. Ara, T. Cucinotta, A. Mascitti, Simulating execution time and power consumption of real-time tasks on embedded platforms, in: Proceedings of the 37th ACM/SIGAPP Symposium on Applied Computing, SAC ’22, Association for Computing Machinery, New York, NY, USA, 2022, pp. 491–50...

    2022

  32. [41]

    G. Chen, K. Huang, A. Knoll, Energy optimization for real-time multiprocessor system-on-chip with optimal DVFS and DPM combination, ACM Transactions on Embedded Computing Systems 13 (3s) (2014) 111:1–111:21.doi:10.1145/ 2567935

    2014

  33. [42]

    J. Zhou, J. Sun, P. Cong, Z. Liu, X. Zhou, T. Wei, S. Hu, Security-Critical Energy-Aware Task Scheduling for Hetero- geneous Real-Time MPSoCs in IoT, IEEE Transactions on Services Computing 13 (4) (2020) 745–758, conference Name: IEEE Transactions on Services Computing.doi:10....

    2020

  34. [43]

    Salami, H

    B. Salami, H. Noori, F. Mehdipour, M. Baharani, Physical-aware predictive dynamic thermal management of multi-core processors, Journal of Parallel and Distributed Computing 95 (2016) 42–56.doi:10.1016/j.jpdc.2016.03.008

    2016 doi

  35. [44]

    Balsini, L

    A. Balsini, L. Pannocchi, T. Cucinotta, Modeling and simulation of power consumption and execution times for real- time tasks on embedded heterogeneous architectures, ACM SIGBED Review 16 (3) (2019) 51–56.doi:10.1145/3373400. 3373408

    2019 doi

  36. [45]

    Shahid, M

    A. Shahid, M. Fahad, R. R. Manumachu, A. Lastovetsky, Improving the accuracy of energy predictive models for multicore CPUs by combining utilization and performance events model variables, Journal of Parallel and Distributed Computing 151 (2021) 38–51.doi:10.1016/j.jpdc.2021.01.007

    2021 doi

  37. [46]

    Zhang, A

    K. Zhang, A. Guliani, S. Ogrenci-Memik, G. Memik, K. Yoshii, R. Sankaran, P. Beckman, Machine Learning-Based Temperature Prediction for Runtime Thermal Management Across System Components, IEEE Transactions on Parallel and Distributed Systems 29 (2) (2018) 405–419.doi:10.1109/...

    2018

  38. [47]

    D. D. Cofer, M. Rangarajan, Formal Modeling and Analysis of Advanced Scheduling Features in an Avionics RTOS, in: Proceedings of the Second International Conference on Embedded Software, EMSOFT ’02, Springer-Verlag, Berlin, 35 Heidelberg, 2002, pp. 138–152

    2002

  39. [48]

    T. Li, G. Yu, J. Song, Minimizing energy by thermal-aware task assignment and speed scaling in heterogeneous MPSoC systems, Journal of Systems Architecture 89 (2018) 118–130.doi:10.1016/j.sysarc.2018.08.003

    2018 doi

  40. [49]

    D. Zhu, L. Chen, T. M. Pinkston, M. Pedram, TAPP: Temperature-aware application mapping for NoC-based many-core processors, in: 2015 Design, Automation & Test in Europe Conference & Exhibition (DATE), 2015, pp. 1241–1244

    2015

  41. [50]

    Abdollahi, Y

    M. Abdollahi, Y. Firouzabadi, F. Dehghani, S. Mohammadi, THAMON: Thermal-aware High-performance Application Mapping onto Opto-electrical network-on-chip, Journal of Systems Architecture 121 (2021) 102315.doi:10.1016/j. sysarc.2021.102315

    2021

  42. [51]

    Chantem, X

    T. Chantem, X. S. Hu, R. P. Dick, Temperature-Aware Scheduling and Assignment for Hard Real-Time Applications on MPSoCs, IEEE Transactions on Very Large Scale Integration (VLSI) Systems 19 (10) (2011) 1884–1897.doi:10.1109/ TVLSI.2010.2058873

    2011

  43. [52]

    Jiang, K

    X. Jiang, K. Huang, X. Zhang, R. Yan, K. Wang, D. Xiong, X. Yan, Energy-Efficient Scheduling of Periodic Applications on Safety-Critical Time-Triggered Multiprocessor Systems, Electronics 7 (6) (2018) 98.doi:10.3390/electronics7060098

    2018 doi

  44. [53]

    T. Li, T. Zhang, G. Yu, J. Song, J. Fan, Minimizing temperature and energy of real-time applications with precedence constraints on heterogeneous MPSoC systems, Journal of Systems Architecture 98 (2019) 79–91.doi:10.1016/j.sysarc. 2019.07.001

    2019 doi

  45. [54]

    Kuo, Y.-F

    C.-F. Kuo, Y.-F. Lu, Task assignment with energy efficiency considerations for non-dvs heterogeneous multiprocessor systems, SIGAPP Appl. Comput. Rev. 14 (4) (2015) 8–18.doi:10.1145/2724928.2724929

    2015

  46. [55]

    K. Cao, J. Zhou, T. Wei, M. Chen, S. Hu, K. Li, A survey of optimization techniques for thermal-aware 3D processors, Journal of Systems Architecture 97 (2019) 397–415.doi:10.1016/j.sysarc.2019.01.003

    2019 doi

  47. [56]

    H. Liu, B. Liu, L. T. Yang, M. Lin, Y. Deng, K. Bilal, S. U. Khan, Thermal-Aware and DVFS-Enabled Big Data Task Scheduling for Data Centers, IEEE Transactions on Big Data 4 (2) (2018) 177–190.doi:10.1109/TBDATA.2017.2763612

    2018

  48. [57]

    Lucas, B

    J. Lucas, B. Juurlink, MEMPower: Data-Aware GPU Memory Power Model, in: M. Schoeberl, C. Hochberger, S. Uhrig, J. Brehm, T. Pionteck (Eds.), Architecture of Computing Systems – ARCS 2019, Lecture Notes in Computer Science, Springer International Publishing, 2019, pp. 195–207

    2019

  49. [58]

    Mantovani, E

    F. Mantovani, E. Calore, Performance and Power Analysis of HPC Workloads on Heterogeneous Multi-Node Clusters, Journal of Low Power Electronics and Applications 8 (2) (2018) 13.doi:10.3390/jlpea8020013

    2018 doi

  50. [59]

    Y. Lee, K. G. Shin, H. S. Chwa, Thermal-aware scheduling for integrated CPUs–GPU platforms, ACM Trans. Embed. Comput. Syst. 18 (5s).doi:10.1145/3358235

    doi

  51. [60]

    Pallister, S

    J. Pallister, S. J. Hollis, J. Bennett, Identifying Compiler Options to Minimize Energy Consumption for Embedded Platforms, The Computer Journal 58 (1) (2015) 95–109.doi:10.1093/comjnl/bxt129

    2015 doi

  52. [61]

    M. R. Garey, D. S. Johnson, Computers and Intractability; A Guide to the Theory of NP-Completeness, W. H. Freeman & Co., USA, 1990

    1990

  53. [62]

    Hanen, A

    C. Hanen, A. M. Kordon, Fixed-parameter tractability of scheduling dependent typed tasks subject to release times and deadlines, J. Sched. 27 (2) (2024) 119–133.doi:10.1007/S10951-023-00788-4. URLhttps://doi.org/10.1007/s10951-023-00788-4

    2024 doi

  54. [63]

    Pagani, H

    S. Pagani, H. Khdr, W. Munawar, J.-J. Chen, M. Shafique, M. Li, J. Henkel, TSP: Thermal safe power: Efficient power budgeting for many-core systems in dark silicon, in: Proceedings of the 2014 International Conference on Hard- ware/Software Codesign and System Synthesis, CODES...

    2014

  55. [64]

    X. Guo, H. Wang, C. Zhang, H. Tang, Y. Yuan, Leakage-aware thermal management for multi-core systems using piecewise linear model based predictive control, 2019, pp. 64–69.doi:10.1145/3287624.3287665. URLhttps://wanghaiuestc.github.io/papers/ASPDAC19_pwl.pdf

    2019

  56. [65]

    Shahid, M

    A. Shahid, M. Fahad, R. R. Manumachu, A. Lastovetsky, A Comparative Study of Techniques for Energy Predictive Modeling Using Performance Monitoring Counters on Modern Multicore CPUs, IEEE Access 8 (2020) 143306–143332. doi:10.1109/ACCESS.2020.3013812

    2020

  57. [66]

    URLhttps://www.gurobi.com

    Gurobi Optimization, LLC, Gurobi Optimizer Reference Manual (2024). URLhttps://www.gurobi.com

    2024

  58. [67]

    (08 2022)

    Wildart, Evolutionary.jl, online, julia package hosted on Github. (08 2022). URLhttps://wildart.github.io/Evolutionary.jl/stable/

    2022

  59. [68]

    NVIDIA, Tegra X2 Series SoC, Technical Reference Manual Parker TRM DP07821001p (Jun. 2017). URLhttps://developer.nvidia.com/embedded/downloads#?search=Tegra%20X2

    2017

  60. [69]

    URLhttps://www.toradex.com/computer-on-modules/apalis-arm-family/nxp-imx-8

    Toradex, NXP i.MX 8 Computer on Module (2021). URLhttps://www.toradex.com/computer-on-modules/apalis-arm-family/nxp-imx-8

    2021

  61. [70]

    URLhttps://workswell-thermal-camera.com/workswell-infrared-camera-wic/#specifications

    Workswell, Workswell infrared camera (2022). URLhttps://workswell-thermal-camera.com/workswell-infrared-camera-wic/#specifications

    2022

  62. [71]

    URLhttps://www.eembc.org/autobench/

    EEMBC, Autobench 2.0 (2022). URLhttps://www.eembc.org/autobench/

    2022

  63. [72]

    J. A. Poovey, T. M. Conte, M. Levy, S. Gal-On, A benchmark characterization of the eembc benchmark suite, IEEE Micro 29 (5) (2009) 18–29.doi:10.1109/MM.2009.74

    2009 doi

  64. [73]

    Zhang, S

    J. Zhang, S. Sadiqbatcha, W. Jin, S. X.-D. Tan, Accurate power density map estimation for commercial multi-core microprocessors, in: 2020 Design, Automation Test in Europe Conference Exhibition (DATE), 2020, pp. 1085–1090. doi:10.23919/DATE48585.2020.9116545

    2020

  65. [74]

    J. N. Hooker, Logic-based benders decomposition for large-scale optimization, Large scale optimization in supply chains and smart manufacturing: Theory and applications (2019) 1–26. 36

    2019

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.