Pith. sign in

REVIEW 3 major objections 5 minor 41 references

PowerScope: ML-based Intra-Cycle Power Estimation

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read PowerScope claims to be the first ML framework that estimates post-layout intra-cycle power directly from RTL simulation traces, achieving 5.88% median and 9% mean absolute percentage error at roughly 80x speedup.

desk verdict A genuinely new intra-cycle power estimation approach with a clever STA-guided feature selection, but the headline numbers don't add up and the feature set may miss input-side internal power. read the letter →

arxiv 2608.05339 v2 pith:C6WPKMW6 submitted 2026-08-05 cs.AR

Jayanth Balasubramanian , Sujay Pandit , Radha Vaidya , Anand Raghunathan This is my paper · ORCID
classification cs.AR
keywords PowerestimationIntra-cycleMachinelearningRTLsimulationSTA-guidedfeatureselectionLightGBMTVLASide-channelanalysis
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

PowerScope claims to be the first machine-learning framework that estimates post-layout intra-cycle power—power at multiple time bins within each clock cycle—directly from RTL simulation traces. It argues that the expensive post-layout gate-level simulation and power analysis step only needs to be run once per design, during training, and that inference can then run entirely at the RTL level. Across five designs, PowerScope reports 5.88% median and 9% mean absolute percentage error against a commercial post-layout flow (Synopsys PrimeTime PX), at roughly 80x end-to-end speedup. It also reports that its bin-level power traces support TVLA side-channel leakage assessment with only 11.7% average deviation from the ground-truth t-statistics.

What carries the argument

The load-bearing mechanism is STA-guided, bin-specific feature selection. Static timing analysis assigns each gate a switching window $[t_{\min}(g), t_{\max}(g)]$ from the SDF file; a gate is considered to contribute to a bin's dynamic power only if that window overlaps the bin. Tracing each such gate's fan-in back to registers yields a per-bin register set, and from cycle-level RTL simulation each selected register contributes two binary features: whether it toggles in the current cycle and its state in the previous cycle. A per-bin LightGBM model then maps these features to that bin's power, trained on labels from post-layout PrimeTime PX analysis with quantile-based stratified subsampling to cover rare high-activity cycles in bimodal bins.

What would settle it

Run PowerScope on a set of workloads for a design with known unbalanced path delays (e.g., a ripple-carry adder or a multiplier with a slow carry chain), then compare its per-bin predictions against post-layout PrimeTime PX power at a fine sampling period; if the median absolute error rises substantially above the reported 5.88% and the extra error concentrates in bins where glitches occur or where actual switching falls outside the STA-derived windows, the overlap-based feature assumption is the cause.

Watch

Extended reading notes

Core claim

PowerScope's central claim is that intra-cycle power, sampled into $B$ uniform time bins per clock cycle, can be predicted per bin by a small design-specific set of RTL signals: the toggle activity and previous-cycle state of the registers in the transitive fan-in of gates whose static-timing switching windows overlap that bin. Because those "potential switching subsets" are derived from the netlist and SDF file in a single pass, and because one LightGBM model is trained per bin with stratified subsampling to handle skewed power distributions, inference needs only RTL simulation. The paper reports that this yields per-bin power estimates whose median absolute error is 5.88% and mean absolute error 9% relative to post-layout gate-level power analysis, and that the predicted traces rank power samples well (Spearman $\rho$ up to 0.96 for FMA), which is what TVLA-style side-channel screening needs.

Load-bearing premise

PowerScope assumes that a gate's contribution to a bin's power is determined by whether its static timing window overlaps the bin, and that the bin's dynamic power can be predicted from the toggle and previous-cycle state of the gate's fan-in registers, which ignores glitch energy and input-dependent delay variation.

Editorial extensions

If this is right

  • Pre-silicon power side-channel evaluation can be run directly from RTL simulation, since PowerScope's bin-level traces track PrimeTime PX TVLA t-statistics within 11.7% average deviation.
  • Feature selection cost no longer scales with the number of bins: the STA-guided pass runs once over netlist and SDF, giving an average 12.5x faster proxy selection than MCP-based methods like ML-Power.
  • Finer sampling periods (smaller $T_s$) tend to raise error because bins near the switching edge become bimodal; coarser bins pool transitions and reduce MAPE, so users can trade temporal resolution against accuracy.
  • Stratified subsampling improves accuracy mainly for designs with bimodal per-bin power distributions (e.g., FMA and AES), where random sampling under-covers lower-probability high-activity cycles.

Reading between the lines

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

  • The STA-window overlap criterion is a static proxy; input-dependent path delays and glitch energy can spill switching activity outside the computed windows or across bins. A stress test on a heavily glitching combinational design (e.g., a multiplier or carry-save adder) would reveal whether the reported 9% MAPE holds when the window assumption is violated.
  • Because the features are only binary register toggles and states, PowerScope's accuracy likely degrades for designs where combinational depth is large and intra-cycle power depends on the exact arrival time distribution rather than just which registers fired; extending features to cycle-relative timing or glitch counts might extend the method to finer $T_s$.
  • The downstream TVLA validation suggests a practical rule of thumb: ML-based intra-cycle power is good enough for rank-based side-channel screening (TVLA), but may not yet be sufficient for absolute-value tasks like IR-droop signoff, where the paper itself notes errors propagate into voltage droop and timing analysis.
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.

Referee Report

3 major / 5 minor

Summary. PowerScope proposes a per-design machine-learning flow for intra-cycle (bin-level) power estimation from RTL register traces. For each of B temporal bins, static timing analysis selects gates whose output switching windows overlap the bin, traces their fan-in registers, and uses per-cycle toggle and previous-state features to train a LightGBM model against PrimeTime PX post-layout power labels. Inference requires only RTL simulation, bypassing post-layout gate-level simulation and power analysis. The paper reports 5.88% median and 9.05% mean absolute percentage error relative to PrimeTime PX across five designs, a roughly 80x speedup, and a TVLA side-channel case study on an OpenTitan AES core.

Significance. If the reported accuracy holds, PowerScope is a useful step: it extends ML-based power estimation from average or per-cycle granularity to intra-cycle granularity, which matters for PDN design, droop analysis, and pre-silicon side-channel evaluation. The method is not circular: labels come from a commercial signoff tool, features come from RTL simulation, and the STA-based feature selection is independent of the power labels. The experimental protocol is clearly described and compares against a commercial post-layout flow, and the downstream TVLA validation is a constructive addition. However, the evidence base is narrow (five designs, no confidence intervals, no released artifacts), the headline accuracy numbers do not follow from the table as printed, and there is a conceptual gap in the STA-guided feature selection that needs quantitative justification. These issues are fixable, but they affect the strength of the central claim.

major comments (3)
  1. [Section 4.1.2 / Algorithm 1] The bin-membership criterion in Algorithm 1 assigns a gate to a bin based on the gate's output switching window, but standard-cell power models dissipate input-pin internal power at the input arrival time, which can precede the output transition by the gate delay. A gate whose output window lies in a later bin, or whose output never toggles because of a controlling input, can therefore dissipate internal power in an earlier bin while being absent from that bin's potential switching subset. The model also lacks the states of the gate's other inputs, so it cannot distinguish cycles that differ only in those co-input states. This is a completeness gap in the feature construction, not a model-capacity issue, and it affects exactly the early, high-activity bins. Please quantify the fraction of per-bin energy attributable to such excluded gates, or extend the PSS to include gates whose input windows overlap each bin, or otherwise show empirically that the omitted component is negligible.
  2. [Section 5.2 / Table 1 / abstract] The headline claim of 'average MedAPE of 5.88% and average MAPE of 9.05%' is not reproduced by an unweighted average of the five per-design averages in Table 1, which gives 7.37% MedAPE and 10.98% MAPE. If a weighted average or a pooled-per-bin calculation is intended, the weighting must be stated explicitly; otherwise the numbers should be corrected. Similarly, the '~80x' speedup is consistent with the geometric mean of the five speedups in Figure 7 (the arithmetic mean is 120.7x), so the averaging convention should be stated wherever an aggregate speedup is reported.
  3. [Section 5.1 / Section 5.2] The empirical evaluation is under-specified for the strength of the central claim. The manuscript does not state which specific workloads were used for testing, how many cycles each test workload contains, how the test workloads differ from the training micro-benchmarks, or any per-design or per-bin variance, confidence intervals, or repeated-run statistics. With only five designs and no released code, models, or data, the 5.88%/9.05% claim is not reproducible as presented. Please report the test workload details, per-design error bars, and consider releasing artifacts to support the quantitative claims.
minor comments (5)
  1. [Figure 1 / Figure 7] The caption of Figure 1 states '~80x avg. speedup' without defining the average; since the per-design speedups range from 15.2x to 248.4x, please state explicitly whether the aggregate is an arithmetic or geometric mean.
  2. [Section 4.1.1] The description of stratified subsampling says that when a stratum has fewer cycles than requested, 'the shortfall is redistributed proportionally to denser strata'; the exact rounding and redistribution rule should be specified for reproducibility.
  3. [Section 5.2.4] The runtime comparison should state whether the reported PowerScope inference time includes RTL simulation plus model inference for the same 100k-cycle workload, and what parallelism assumptions are used, given that the server in Section 5.1 has 512 cores.
  4. [Table 2] Since the features are register toggles and states, Table 2 should also report the total number of registers in each design, not only the total number of RTL signals, to make the 'selected proxies' fractions more interpretable.
  5. [Section 6 / Figure 8] The statement that the PowerScope-based TVLA t-statistic deviates by 'only 11.7% on average' needs a precise definition of the relative error; on a log-scale plot, small t-statistics can dominate such an average, so a per-bin table or an alternative error measure would be more informative.

Circularity Check

0 steps flagged · score 0.0 of 10

PowerScope is a supervised ML approximation trained on independent commercial-tool labels and tested on held-out workloads; no circular reduction found.

full rationale

PowerScope's derivation chain is a standard supervised approximation: per-bin LightGBM models map RTL register toggle/state features (Algorithm 1) to intra-cycle power labels produced by PrimeTime PX from post-layout GLS of micro-benchmarks. The STA-guided feature selection in Sec. 4.1.2 uses only the gate-level netlist and SDF timing windows to identify per-bin registers; it does not use the power labels, so the feature set is not defined in terms of the prediction target. Evaluation is on workloads that are, in the paper's words, 'completely unseen during model training' (Sec. 5.2), and the reference is an external commercial tool, PrimeTime PX, so no fitted parameter is smuggled into the test labels. The TVLA downstream comparison uses the same ground-truth tool as an evaluation metric, not as part of the derivation of the predictions. Self-citations such as ML-Power [22] as a baseline and earlier Raghunathan et al. references are not load-bearing premises; they are related-work comparisons and do not force the central result. A separate arithmetic inconsistency (the abstract's 5.88% MedAPE and 9% MAPE do not match a simple average of Table 1's per-design values) is a reporting/correctness concern, but it is not circularity. No step in the paper reduces by construction to its own inputs, and no load-bearing claim depends on an unverified self-citation.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

PowerScope rests on supervised learning from one design's commercial flow to its own future workloads. The axioms are the domain assumptions that the commercial labels are ground truth, that STA windows localize gate contributions to bins, and that cycle-granular register features carry enough information to predict bin power. Free parameters include the temporal binning, stratified sampling configuration, and per-bin tree hyperparameters, none of which are provided as fixed values. No new physical entities are introduced.

free parameters (3)
  • Number of intra-cycle bins B (or sampling period T_s) = B = 3, 5, 8, 10 across designs (Table 1)
    The central accuracy numbers depend on this user-chosen temporal resolution; Figure 6 shows error changes by several percentage points as T_s changes.
  • Stratified subsampling parameters K (number of strata) and N (target training set size) = Not reported
    Chosen by hand in Section 4.1.1; the paper does not give the values, though the ablation shows stratified sampling changes MAPE by up to 3 to 9 percentage points on bimodal designs.
  • Per-bin LightGBM hyperparameters (leaf count, depth, regularization) = Grid-searched per bin
    Selected to fit each bin's distribution in Section 4.2; the selected configurations are not enumerated, and the reported accuracy is a function of this per-bin tuning.
assumptions (4)
  • domain assumption PrimeTime PX post-layout power analysis with back-annotated parasitics provides accurate ground-truth intra-cycle power.
    Section 4.1.1 uses it to generate labels and Section 5 compares against it; silicon measurements are not presented.
  • domain assumption A gate's switching window, derived from SDF min/max arrival times, determines which bins the gate's dynamic power falls into.
    This is the basis of STA-guided feature selection in Algorithm 1; glitch and path-delay variation can violate the static window model.
  • domain assumption Per-bin power is a function of the toggle and previous state of the fan-in registers at cycle granularity.
    The only inference features in Section 4.1.2 are register toggles and states; sub-cycle timing of toggles is not available from RTL simulation.
  • domain assumption Models trained on micro-benchmarks generalize to unseen application workloads on the same design.
    Section 5.2 evaluates on workloads completely unseen during model training but assumes the learned mapping is workload invariant.

how reviews work

0 comments
Cite this review

Pith. "Pith review of PowerScope: ML-based Intra-Cycle Power Estimation." pith.science (2026). https://pith.science/paper/C6WPKMW6

@misc{pith2026260805339,
  author       = {Pith},
  title        = {Pith review of: PowerScope: ML-based Intra-Cycle Power Estimation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/C6WPKMW6}},
  note         = {Machine review of arXiv:2608.05339}
}
read the original abstract

Power estimation at sub-clock-cycle temporal resolutions is critical for tasks such as power delivery network (PDN) design, dynamic voltage droop analysis, and pre-silicon power side-channel security evaluation. Designers commonly rely on commercial post-layout gate-level power analysis tools for these tasks, but these flows are computationally expensive and scale poorly with design size and workload length. Machine learning (ML)-based power estimation frameworks have shown promise in accelerating power estimation, but prior efforts only address average power or per-cycle power estimation. We propose PowerScope, the first ML-based intra-cycle power estimation framework. PowerScope operates purely on RTL simulation traces at inference time, eliminating the need for post-layout gate-level simulation and power analysis per workload. Across a diverse benchmark suite, PowerScope achieves 5.88% median and 9% mean absolute percentage error compared to commercial post-layout gate-level power estimates while running ~80x faster. We further demonstrate that PowerScope's predictions can be reliably used for the downstream task of pre-silicon power side-channel leakage assessment.

Figures

Figures reproduced from arXiv: 2608.05339 by the authors.

Figure 1
Figure 1. (a) PowerScope Inference vs. the Traditional flow for intra-cycle power generation (b) Per-bin power comparison of PowerScope prediction and Traditional flow, on the 64-bit Integer Radix-4 Divider [10] over 25 cycles. Speedup achieved is ∼80× end-to-end measured across designs. (excludes the one-time logic synthesis and layout cost) simulation (GLS), while layout-level analysis additionally requires place-and-route … view at source ↗
Figure 2
Figure 2. PowerScope Training and Inference flow and ML-Power [22] make use of MCP (Minimax Concave Penalty) regularization [35] during regression, which automatically selects power proxies while building power models through regression. A second class of methods, including GRANNITE [37] and PRI￾MAL [39], uses neural network based models to learn mappings between switching activity and power consumption. These methods achieve… view at source ↗
Figure 3
Figure 3. (a) Illustration of temporal bins 𝑏0 through𝑏9 within a single clock cycle. Histogram of per-bin power across 60K simulated cycles for two temporal bins of an OpenTitan AES design [19]. (b) Bin 𝑏7 is dominated by a leakage mode at low power, with a smaller activity mode at higher power, pro￾ducing a bimodal distribution (c) Bin 𝑏3 exhibits a unimodal leakage-dominated distribution 4.1 Training Dataset Generation 4.1… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Accuracy and feature-selection speedup of [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: PowerScope’s stratified subsampling (green) vs. random sampling (orange) compare it with random sampling [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]
Figure 6
Figure 6. Figure 6: Effect of sampling period 𝑇𝑠 on prediction accuracy 15.2x 49.2x 83.3x 248.4x 207.6x [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: Runtime breakdown and speedup of PowerScope inference (left bar per design) versus the traditional flow (right bar per design) for intra-cycle power across workload of 100k cycles. Speedup (annotated in blue) ranges from 15.2× on the Fixed-point adder design to 248.4× …
Figure 8
Figure 8. Figure 8: TVLA Welch’s |𝑇 | statistic per intra-cycle bin for the AES design (𝑇𝑠 = 500 ps, 𝐵 = 10), comparing PrimeTime PX ground truth (green) against PowerScope predictions (orange) 7 Conclusion We presented PowerScope, the first ML-based framework to esti￾mate intra-cycle pow…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

41 extracted references · 1 linked inside Pith

  1. [1]

    Luca Amarù, Pierre-Emmanuel Gaillardon, and Giovanni De Micheli. 2015. The EPFL Combinational Benchmark Suite. InProceedings of the 24th International Workshop on Logic & Synthesis (IWLS). Mountain View, CA, USA

    2015

  2. [2]

    Alessandro Bogliolo, Luca Benini, and Giovanni De Micheli. 2000. Regression- based RTL power modeling.ACM Trans. Des. Autom. Electron. Syst.5, 3 (July 2000), 337–372. doi:10.1145/348019.348081

    arXiv 2000

  3. [3]

    Eric Brier, Christophe Clavier, and Francis Olivier. 2004. Correlation Power Analysis with a Leakage Model. InCryptographic Hardware and Embedded Systems – CHES 2004 (Lecture Notes in Computer Science, Vol. 3156), Marc Joye and Jean-Jacques Quisquater (Eds.). Springer, Berlin, Heidelberg, 16–29. doi:10.1007/978-3-540-28632-5_2

    doi:10.1007/978-3-540-28632-5_2 2004

  4. [4]

    Cadence Design Systems, Inc. 2026. Innovus Implementation Sys- tem. https://www.cadence.com/en_US/home/tools/digital-design-and-signoff/ soc-implementation-and-floorplanning/innovus-implementation-system.html

    2026

  5. [5]

    Contreras and M

    G. Contreras and M. Martonosi. 2005. Power prediction for intel XScale® processors using performance monitoring unit events.Proceedings of the In- ternational Symposium on Low Power Electronics and Design(2005), 221–226. doi:10.1145/1077603.1077657 2005 International Symposium on Low Power Electronics and Design ; Conference date: 08-08-2005 Through 10-08-2005

    arXiv 2005

  6. [6]

    Costa, J

    J. Costa, J. Monteiro, L.M. Silveira, and S. Devadas. 1999. A probabilistic approach for RT-level power modeling. InICECS’99. Proceedings of ICECS ’99. 6th IEEE International Conference on Electronics, Circuits and Systems (Cat. No.99EX357), Vol. 2. 911–914 vol.2. doi:10.1109/ICECS.1999.813262

    arXiv 1999

  7. [7]

    José Costa, Luís Silveira, Srinivas Devadas, and José Monteiro. 2004. Power Estimation Using Probability Polynomials.Design Automation for Embedded Systems9 (2004), 19–52. doi:10.1007/s10617-005-5344-5

    doi:10.1007/s10617-005-5344-5 2004

  8. [8]

    Electronic Design. n.d.. What’s the Difference Between SPICE and FastSPICE Circuit Simulators. https://www.electronicdesign.com/technologies/test- measurement/article/21801299/whats-the-difference-between-spice-and- fastspice-circuit-simulators Accessed: 2026-05-27

    arXiv 2026

Show all 41 references
  1. [9]

    Wenji Fang, Yao Lu, Shang Liu, Qijun Zhang, Ceyu Xu, Lisa Wu Wills, Hongce Zhang, and Zhiyao Xie. 2023. MasterRTL: A Pre-Synthesis PPA Estimation Frame- work for Any RTL Design. InProceedings of the 2023 IEEE/ACM International Conference on Computer-Aided Design (ICCAD). 1–9. ...

    2023

  2. [10]

    Neel Gala, Arjun Menon, Rahul Bodduna, G. S. Madhusudan, and V. Kamakoti

  3. [11]

    GlobalFoundries. 2026. 22FDX Platform. https://www.cmc.ca/globalfoundries- 22fdx-fdsoi-22-nm/

    2026

  4. [12]

    Gilbert Goodwill, Benjamin Jun, Josh Jaffe, and Pankaj Rohatgi. 2011. A Test- ing Methodology for Side-Channel Resistance Validation. InNIST Non-Invasive Attack Testing Workshop. Cryptography Research Inc

    2011

  5. [13]

    Guolin Ke, Qi Meng, Thomas Finley, Taifeng Wang, Wei Chen, Weidong Ma, Qiwei Ye, and Tie-Yan Liu. 2017. LightGBM: A Highly Efficient Gradient Boosting Decision Tree. InAdvances in Neural Information Processing Systems 30 (NeurIPS). Curran Associates, Inc., 3149–3157

    2017

  6. [14]

    Pantea Kiaei, Zhenyuan Liu, Ramazan Kaan Eren, Yuan Yao, and Patrick Schau- mont. 2021. Saidoyoki: Evaluating side-channel leakage in pre- and post-silicon setting. Cryptology ePrint Archive, Paper 2021/1235. https://eprint.iacr.org/ 2021/1235

    2021

  7. [15]

    Donggyu Kim, Jerry Zhao, Jonathan Bachrach, and Krste Asanović. 2019. Sim- mani: Runtime Power Modeling for Arbitrary RTL with Automatic Signal Selec- tion. InProceedings of the 52nd Annual IEEE/ACM International Symposium on Microarchitecture (MICRO). ACM, 1050–1062. doi:10.1...

    2019

  8. [16]

    Paul Kocher, Joshua Jaffe, and Benjamin Jun. 1999. Differential Power Analysis. InAdvances in Cryptology — CRYPTO’ 99, Michael Wiener (Ed.). Springer Berlin Heidelberg, Berlin, Heidelberg, 388–397. doi:10.1007/3-540-48405-1_25

    1999 doi

  9. [17]

    Michael LeBeane, Jee Ho Ryoo, Reena Panda, and Lizy Kurian John. 2015. Watt Watcher: Fine-Grained Power Estimation for Emerging Workloads. In2015 27th International Symposium on Computer Architecture and High Performance Com- puting (SBAC-PAD). 106–113. doi:10.1109/SBAC-PAD.2015.26

    2015 doi

  10. [18]

    Strong, Jay B

    Sheng Li, Jung Ho Ahn, Richard D. Strong, Jay B. Brockman, Dean M. Tullsen, and Norman P. Jouppi. 2009. McPAT: an integrated power, area, and timing modeling framework for multicore and manycore architectures. InProceedings of the 42nd Annual IEEE/ACM International Symposium o...

    2009

  11. [19]

    lowRISC contributors. 2019. OpenTitan: Open Source Silicon Root of Trust. https://opentitan.org. Accessed: 2026-05-11

    2019

  12. [20]

    Yao Lu, Qijun Zhang, and Zhiyao Xie. 2024. Unleashing Flexibility of ML-based Power Estimators Through Efficient Development Strategies. InProceedings of the 29th ACM/IEEE International Symposium on Low Power Electronics and Design (Newport Beach, CA, USA)(ISLPED ’24). Associa...

    2024

  13. [21]

    F.N. Najm. 1995. Power estimation techniques for integrated circuits. InPro- ceedings of IEEE International Conference on Computer Aided Design (ICCAD). 492–499. doi:10.1109/ICCAD.1995.480162

    1995

  14. [22]

    Sujay Pandit, Sujit Dey, and Anand Raghunathan. 2025. ML-Power: Machine Learning based Power Estimation for SoCs. In2025 IEEE/ACM International Symposium on Low Power Electronics and Design (ISLPED). IEEE, 1–7. doi:10.1109/ ISLPED65674.2025.11261702

    2025

  15. [23]

    Raghunathan, S

    A. Raghunathan, S. Dey, and N.K. Jha. 1996. Register-transfer level estimation techniques for switching activity and power consumption. InProceedings of International Conference on Computer Aided Design. 158–165. doi:10.1109/ICCAD. 1996.569539

    1996

  16. [24]

    S. Ravi, A. Raghunathan, and S. Chakradhar. 2003. Efficient RTL power esti- mation for large designs. In16th International Conference on VLSI Design, 2003. Proceedings.431–439. doi:10.1109/ICVD.2003.1183173

    2003 arXiv

  17. [25]

    Michał Rewiénski. 2011. A Perspective on Fast-SPICE Simulation Technol- ogy. InSimulation and Verification of Electronic and Biological Systems, Peng Li, Luis Miguel Silveira, and Peter Feldmann (Eds.). Springer, 23–42. doi:10.1007/978- 94-007-0149-6_2

    2011 doi

  18. [26]

    Synopsys, Inc. 2026. Design Compiler: RTL Synthesis. https://www.synopsys. com/implementation-and-signoff/rtl-synthesis-test/design-compiler.html

    2026

  19. [27]

    Synopsys, Inc. 2026. PrimeTime Suite for Static Timing Analysis. https://www. synopsys.com/implementation-and-signoff/signoff/primetime.html

    2026

  20. [28]

    Synopsys, Inc. 2026. VCS Functional Verification Solution. https://www. synopsys.com/verification/simulation/vcs.html

    2026

  21. [29]

    Chi-ying Tsui, José Monteiro, Massoud Pedram, Srinivas Devadas, Alvin Despain, and Bill Lin. 1995. Power Estimation Methods for Sequential Logic Circuits.Very Large Scale Integration (VLSI) Systems, IEEE Transactions on3 (10 1995), 404 – 416. doi:10.1109/92.406998

    1995 doi

  22. [30]

    Ashka Vakil, Houman Homayoun, and Avesta Sasan. 2019. IR-ATA: IR Annotated Timing Analysis, A Flow for Closing the Loop Between PDN Design, IR Analysis & Timing Closure.2019 24th Asia and South Pacific Design Automation Conference (ASP-DAC)(2019), 1–8. https://api.semanticscho...

    2019

  23. [31]

    Zhiyao Xie, Shiyu Li, Mingyuan Ma, Chen-Chia Chang, Jingyu Pan, Yiran Chen, and Jiang Hu. 2022. DEEP: Developing Extremely Efficient Runtime On-Chip Power Meters. InProceedings of the 41st IEEE/ACM International Conference on Computer-Aided Design (ICCAD). doi:10.1145/3508352.3549427

    2022

  24. [32]

    Zhiyao Xie, Haoxing Ren, Brucek Khailany, Ye Sheng, Santosh Santosh, Jiang Hu, and Yiran Chen. 2020. PowerNet: Transferable Dynamic IR Drop Estimation via Maximum Convolutional Neural Network.2020 25th Asia and South Pacific Design Automation Conference (ASP-DAC)(2020), 13–18....

    2020

  25. [33]

    Zhiyao Xie, Xiaoqing Xu, Matt Walker, Joshua Knebel, Kumaraguru Palaniswamy, Nicolas Hebert, Jiang Hu, Huanrui Yang, Yiran Chen, and Shidhartha Das. 2021. APOLLO: An Automated Power Modeling Framework for Runtime Power In- trospection in High-Volume Commercial Microprocessors....

    2021

  26. [34]

    Jianwang Zhai, Chen Bai, Binwu Zhu, Yici Cai, Qiang Zhou, and Bei Yu. 2021. McPAT-Calib: A Microarchitecture Power Modeling Framework for Modern CPUs. In2021 IEEE/ACM International Conference On Computer Aided Design (ICCAD). 1–9. doi:10.1109/ICCAD51958.2021.9643508

    2021

  27. [35]

    Cun-Hui Zhang. 2010. Nearly unbiased variable selection under minimax concave penalty.The Annals of Statistics38 (02 2010). doi:10.1214/09-AOS729

    2010 doi

  28. [36]

    Qijun Zhang, Shiyu Li, Guanglei Zhou, Jingyu Pan, Chen-Chia Chang, Yi- ran Chen, and Zhiyao Xie. 2023. PANDA: Architecture-Level Power Evalu- ation by Unifying Analytical and Machine Learning Solutions.2023 IEEE/ACM International Conference on Computer Aided Design (ICCAD)(202...

    2023

  29. [37]

    Yanqing Zhang, Haoxing Ren, and Brucek Khailany. 2020. GRANNITE: Graph Neural Network Inference for Transferable Power Estimation. InProceedings of the 57th ACM/IEEE Design Automation Conference (DAC). 1–6. doi:10.1109/ DAC18072.2020.9218643

    2020

  30. [38]

    Yanqing Zhang, Haoxing Ren, Akshay Sridharan, and Brucek Khailany. 2022. GATSPI: GPU accelerated gate-level simulation for power improvement. 1231–

    2022

  31. [39]

    Yuan Zhou, Haoxing Ren, Yanqing Zhang, Ben Keller, Brucek Khailany, and Zhiru Zhang. 2019. PRIMAL: Power Inference using Machine Learning. InProceedings of the 56th Annual Design Automation Conference (DAC). Article 24, 6 pages. doi:10.1145/3316781.3317884 7

    2019

  32. [1236]

    doi:10.1145/3489517.3530601

  33. [2016]

    In29th Interna- tional Conference on VLSI Design (VLSID)

    SHAKTI Processors: An Open-Source Hardware Initiative. In29th Interna- tional Conference on VLSI Design (VLSID). IEEE, 7–8. doi:10.1109/VLSID.2016.130

    2016 doi

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.