Automating Energy-Efficient GPU Kernel Generation: A Fast Search-Based Compilation Approach

Guohao Dai; Ningyi Xu; Shijie Cao; Sicheng Zhang; Weigang Feng; Yijia Zhang; Zhihong Gou

REVIEW 4 major objections 7 minor 36 references

Automating Energy-Efficient GPU Kernel Generation: A Fast Search-Based Compilation Approach

T0 review · 4 major / 7 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read Search finds GPU kernels that use up to 21.69% less energy.

desk verdict A useful and novel combination for energy-aware GPU kernel search, but the evaluation has a physically impossible power reading in Table 3 that the authors must fix. read the letter →

arxiv 2411.18873 v1 pith:XP6XIVSK submitted 2024-11-28 cs.PF cs.LG

Yijia Zhang , Zhihong Gou , Shijie Cao , Weigang Feng , Sicheng Zhang , Guohao Dai , Ningyi Xu This is my paper · ORCID

classification cs.PFcs.LG

keywords GPUkernelgenerationenergyefficiencysearch-basedcompilationgeneticalgorithmcostmodelXGBoostdeeplearningcompilerspowermeasurement

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

This paper claims that the energy a GPU kernel consumes can be cut substantially without slowing it down, if the search that generates kernels treats energy as a goal equal to speed. The authors show that kernels with nearly identical latency can differ a lot in energy, so among the fastest kernels there are lower-power versions worth choosing. To keep the search practical, they use a machine-learning model that predicts a kernel's energy from its loop structure and memory-access pattern, and they refresh that model during the search so that only a few candidate kernels need direct power measurement. Across operators from transformer and convolution models, their approach reduces kernel energy by up to 21.69% (7.47% on average) on an NVIDIA A100 and by more than 50% on some RTX 4090 cases, with latency staying about the same or even improving. The practical payoff is that data centers could lower both compute and cooling energy simply by choosing better kernel implementations.

What carries the argument

The central mechanism is a genetic algorithm with a two-stage fitness evaluation, plus a learned, dynamically updated energy cost model. In each search round, the algorithm measures kernel latency, keeps the $M$ fastest kernels, then uses the cost model to rank those by predicted energy and keeps a fraction of them. The cost model is an XGBoost regressor trained on high-level kernel features (floating-point and integer operation counts, loop depth and structure, vectorization, and cache-access counts) with a weighted squared-error loss that up-weights low-energy kernels. A parameter $k$ controls how many top-ranked kernels are actually measured with NVML power sampling each round: when the model's prediction error is below a threshold, $k$ shrinks and fewer measurements are needed; when the model drifts, $k$ grows. This online updating lets the search keep accuracy while cutting the main time cost, which is physical power measurement.

What would settle it

Measure the same searched kernels from Table 2 with an independent high-bandwidth power trace, such as a current probe on the GPU power rail, and compare the energy ratios to the reported values; if the 21.69% reduction for MM1 does not reproduce under independent measurement, the reported savings are artifacts of NVML's 30-50 Hz sampling.

Watch

Extended reading notes

Core claim

The paper's central claim is that energy efficiency can be inserted into the kernel-generation search itself, rather than treated as a post-hoc hardware or workload-level fix. Building on a genetic search over tensor-program schedules, the authors select, in each generation, the fastest kernels first and then, among those, the ones with the lowest predicted energy. The key empirical discovery is that this two-stage selection finds kernels with clearly lower energy than a latency-only search while keeping latency essentially unchanged; in some cases the energy reduction even comes with slight latency improvement. The authors attribute the effect to structural differences in grid and block size: a smaller grid leaves more streaming multiprocessors idle, cutting static energy, and larger blocks increase data reuse, cutting memory-access energy. They report up to 21.69% energy reduction on an A100 GPU and up to 53.13% on an RTX 4090, against the Ansor baseline, with average latency changing by less than two percent.

Load-bearing premise

All reported energy reductions and the training data for the cost model depend on NVML power sampling at 30-50 Hz, combined with preheating and thousands of repeated kernel executions, being accurate and repeatable enough to capture millisecond-scale kernel energy.

Editorial extensions

If this is right

Energy becomes a first-class objective in tensor program search, so any workload compiled with such a scheduler can inherit energy savings without requiring new hardware or changes to the model.
Because the method keeps latency close to the latency-optimal kernel, it can be dropped into existing serving or training stacks as a drop-in kernel replacement.
The case study shows that static energy depends on how many SMs are active; choosing smaller grids for small batch sizes could become a general compiler heuristic.
The dynamic cost-model updating strategy reduces the number of on-device measurements, making energy-aware compilation practical at the scale of thousands of kernels.

Reading between the lines

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

The two-stage selection (fastest first, then most energy-efficient) could be applied to other search-based compilers beyond Ansor, and the energy cost model could be retrained on any hardware with an energy API, so the approach may generalize to CPUs, FPGAs, or future GPU generations beyond the two tested.
The reported 53.13% energy reduction for the MV operator on RTX 4090 suggests that memory-bound operators, common in large language model inference, might be the biggest beneficiaries; this is not a claim the paper makes explicitly.
The grid-size/block-size mechanism implies that deliberately under-utilizing a GPU (leaving SMs idle) can save energy when the workload is small; a testable extension is to benchmark whether a single search round can also optimize power-capping settings together with kernel structure.
Since the energy cost model is trained on a small set of operators, its predictions could drift on unseen shapes; the paper's dynamic updating strategy mitigates this, but a stronger test would be to evaluate the energy model's ranking accuracy (not just the final kernel) on a held-out operator family.

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

The paper's core idea is sound and worth engaging: add energy as an explicit objective in Ansor-style genetic search, backed by an XGBoost energy cost model and an adaptive measurement budget. That combination is new in the cited literature, and the case study in Section 8 gives a concrete, plausible explanation for why one kernel can use less energy than another at similar latency. The A100 results are internally consistent and show a meaningful 7.5% average energy reduction with negligible latency impact.

The soft spots are real and one is serious. Table 3 reports the MV baseline on RTX 4090 as 6.909 mJ at 0.0118 ms. That implies 585 W average power, which exceeds the RTX 4090's board limit of about 450 W. This is not a rounding issue; the entry is impossible. Since every energy number and the cost-model training data come from the same NVML pipeline, this internal inconsistency undermines the credibility of the measurement methodology, not just a single cell. The paper also gives no error bars, and several Table 2 reductions are only 1.5-3%, which could easily be noise at NVML's 30-50 Hz sampling rate on millisecond kernels even with thousands of repetitions. Add a small operator set, only two GPUs, no code or data release, and comparisons restricted to Ansor and cuBLAS, and you have an evaluation that is too thin for strong claims.

Still, the central mechanism is principled. The paper does not claim to model energy from first principles; it measures what it can and uses the cost model only to accelerate search. The latency-focused selection within each generation is a sensible way to avoid trading away performance. The A100 numbers pass a sanity check (e.g., 239 W and 184 W for the MM1 kernels), so I do not think this is fabricated. The likely story is a misreported unit or a bad measurement on one platform.

This paper deserves a serious referee because the problem is important and the approach is novel. But it needs major revision: correct or explain Table 3, add confidence intervals, report the exact measurement protocol, and ideally release artifacts. As it stands, I would not cite it in my own work, but I would want to see a fixed version.

My recommendation: send it to peer review with a clear request for additional measurements and reproducibility. The idea should not be buried over a fixable data error.

Referee Report

4 major / 7 minor

Summary. The paper proposes an energy-aware search-based kernel generation framework for GPU tensor compilers. The method builds on Ansor's genetic-algorithm search, adding an energy objective: in each generation it first selects the fastest kernels, then ranks them by predicted energy from an XGBoost cost model, and finally measures a dynamic subset using NVML to update the model. The authors report up to 21.69% energy reduction on an NVIDIA A100 (average 7.47%), up to 53.13% on an RTX 4090, latency comparable to Ansor, and further comparison against cuBLAS. A case study attributes the savings to reduced static and memory-access energy due to smaller grid sizes and larger thread blocks.

Significance. If the reported energy reductions are accurate, this is a genuinely useful step toward making energy efficiency a first-class objective in tensor compilation. The paper identifies a real latency-power trade-off in kernel search, provides a fast machine-learning energy cost model, and introduces a dynamic measurement strategy that plausibly reduces search time. The evaluation against Ansor and cuBLAS is appropriate, and the A100 numbers are internally consistent. However, the absence of code, the lack of statistical repeatability, and a physically impossible baseline measurement in the RTX 4090 table currently limit confidence in the headline claims.

major comments (4)

[Section 7.2, Table 3] The RTX 4090 MV baseline is internally inconsistent and physically implausible: 6.909 mJ at 0.0118 ms implies an average power of 585 W, which exceeds the RTX 4090's ~450 W board power limit. The CONV baseline likewise implies 468 W. The proposed MV kernel gives 263 W, which is plausible. Because the 53.13% MV energy reduction and the statement that reductions are 'even higher on the RTX 4090' rest on this measurement, the NVML measurement pipeline described in Section 4.4 is called into question. Please re-measure these baselines and correct the table and derived percentages.
[Section 7, Tables 2 and 3] No error bars, standard deviations, confidence intervals, or repeated runs are reported. Several A100 reductions (for example MM2 4.57%, MM3 3.08%, MV1 2.92%, MV2 1.71%, CONV3 1.54%) are small enough to be comparable to NVML sampling noise at 30–50 Hz for millisecond-scale kernels. The paper should report variance over multiple independent searches and ideally a paired statistical test across operator shapes to establish that the reductions are not measurement artifacts.
[Section 5.4 and Figure 4] The energy cost model is claimed to be accurate, but the only evidence is a scatter plot with no quantitative error metric (R², MAPE, Spearman correlation) and no comparison to a trivial baseline predictor. Additionally, the weighted loss in Eq. (1) is introduced without an ablation showing it outperforms unweighted MSE. Since the dynamic updating strategy in Algorithm 1 relies on prediction-error thresholds, the paper should quantify the model's accuracy and report sensitivity of the final search quality to the threshold μ.
[Section 7.4 and Algorithm 1] The claim that the cost-model-based search is 'nearly twice as fast' as NVML-only search is demonstrated only for a specific adjustment of μ that halves the number of measurements; this is a hyperparameter choice, not a parameter-free property of the method. The paper should provide a sensitivity analysis over μ and k, and should show that the final kernel energy and latency are not degraded relative to the NVML-only search. The prediction error is called 'SNR' in Algorithm 1 but the definition of this signal-to-noise ratio is never given.

minor comments (7)

[Algorithm 1] The word 'vaule' should be 'value', and the undefined term 'SNR' should either be defined formally or replaced with a concrete error metric.
[Figure 2 caption] The caption says 'The kernel generate by our method' and should be 'generated'; also 'Ansor's' would be clearer.
[Section 6.4] The phrase 'An prediction error' should be 'A prediction error'.
[Section 7.3] The description of 'normalized measured energy' and 'normalized predicted energy' is vague; please specify the normalization procedure so the plot can be interpreted.
[Section 7.4] The time costs are given in hours (7.3h, 6.5h, etc.) but the number of search rounds and the number of kernels measured per round are not specified; please state the experimental protocol.
[Section 8] The case study attributes the energy difference to static and memory-access energy, but no quantitative power or counter data is shown to support this decomposition; consider reporting measured power breakdowns or an ablation over grid/block sizes.
[Section 1] The claim of being the 'first search-based energy-aware GPU kernel generation framework' is strong given prior work such as reference [19] on energy-aware tile size selection; please qualify the novelty statement.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: final energy reductions are measured, the cost model is validated on held-out data, and no load-bearing step reduces to its own inputs.

full rationale

The paper's central claim is an empirical system evaluation. Final kernel energies in Tables 2-4 are obtained by NVML measurement of the chosen kernels, not by the energy cost model, so the reported reductions are not predictions of a fitted model. The energy cost model (Section 5.4) is trained on measured energies and evaluated on a held-out 20% test split (Section 7.3), giving it independent content. The dynamic updating strategy (Algorithm 1) uses measured energies to correct model predictions during search, but the final reported energies remain direct measurements. The only tunable element, the mu threshold in Section 7.4, is explicitly adjusted to demonstrate a ~2x speedup; this is an experimental setting, not a fitted parameter renamed as a prediction, and it does not affect the headline energy reductions. There are no author self-citations, uniqueness theorems, or imported ansatze in the derivation chain. A separate correctness concern is that the RTX 4090 MV baseline (6.909 mJ at 0.0118 ms) implies ~585 W average power, which is physically implausible for that GPU; this is a measurement/validity risk, not a circularity, and does not change the circularity verdict.

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

No new physical or conceptual entities are introduced; the energy cost model is a fitted ML model, not a postulated entity. The central claim rests on measurement accuracy, feature sufficiency, and a heuristic search design.

free parameters (5)

mu (SNR threshold in Algorithm 1) = Not reported; adjusted in experiments
Controls when the energy cost model is trusted; in Section 7.4 the authors tune mu to nearly halve NVML measurements, so the speedup result depends on this hand-picked value.
k (measurement fraction) and update step 0.2 = Initial 1.0, step 0.2
Algorithm 1 starts k at 1.0 and adjusts by 0.2 based on prediction error; this schedule is chosen by the authors, not derived.
M (number of top-latency kernels per generation) = Not specified
The genetic search keeps M fastest kernels each round; M is a free search hyperparameter not reported.
XGBoost energy cost model hyperparameters and feature set = Not reported
The model is trained on measured kernel data with a chosen feature set and XGBoost settings; its predictions are fitted values, and the exact features are only qualitatively described.
Measurement warm-up duration and repetition count = 'Several seconds' and 'thousands of iterations'
These experimental constants are chosen by hand to stabilize NVML readings; exact values are not given, so measurements may not be exactly replicable.

assumptions (5)

domain assumption GPU energy can be decomposed into constant, static, and dynamic power, with dynamic power driven by compute and memory access.
Used to justify high-level features for the energy cost model; cited to AccelWattch [21] in Sections 2.3 and 5.3.
domain assumption High-level kernel features (loop depth, op counts, vectorization, cache access) are sufficient to predict relative kernel energy.
Central to the energy cost model in Section 5.4; no evidence is given that these features capture all energy-relevant behavior.
domain assumption NVML power sampling at 30-50 Hz, with preheating and repeated execution, gives accurate average power for millisecond-scale kernels.
Underlies all measured energy values and cost model training data; Section 4.4.
ad hoc to paper Selecting low-latency kernels first and then energy-efficient ones preserves latency while reducing energy.
This is the search design choice in Section 4.4 and Algorithm 1; it is a heuristic with no formal guarantee.
ad hoc to paper Weighted squared error loss (Ep-Em)^2/Em improves prediction quality for low-energy kernels.
Chosen in Eq. (1) without theoretical or empirical comparison to unweighted loss.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Automating Energy-Efficient GPU Kernel Generation: A Fast Search-Based Compilation Approach." pith.science (2026). https://pith.science/paper/XP6XIVSK

@misc{pith2026241118873,
  author       = {Pith},
  title        = {Pith review of: Automating Energy-Efficient GPU Kernel Generation: A Fast Search-Based Compilation Approach},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/XP6XIVSK}},
  note         = {Machine review of arXiv:2411.18873}
}

read the original abstract

Deep Neural Networks (DNNs) have revolutionized various fields, but their deployment on GPUs often leads to significant energy consumption. Unlike existing methods for reducing GPU energy consumption, which are either hardware-inflexible or limited by workload constraints, this paper addresses the problem at the GPU kernel level. We propose a novel search-based compilation method to generate energy-efficient GPU kernels by incorporating energy efficiency into the search process. To accelerate the energy evaluation process, we develop an accurate energy cost model based on high-level kernel features. Furthermore, we introduce a dynamic updating strategy for the energy cost model, reducing the need for on-device energy measurements and accelerating the search process. Our evaluation demonstrates that the proposed approach can generate GPU kernels with up to 21.69% reduced energy consumption while maintaining low latency.

Figures

Figures reproduced from arXiv: 2411.18873 by the authors.

**Figure 1.** Our search-based compilation for energy [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗

**Figure 3.** The inverse correlation between latency and [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗

Figures from the paper (2 more)

**Figure 4.** Figure 4: The normalized predicted energy v.s. the normalized measured energy. [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]

**Figure 5.** Figure 5: The time cost of NVML-only and costmodel [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 24 canonical work pages

[1]

[n. d.]. Energy Crisis. https://www.trendforce.com/news/2024/03/28/ news-ai-confronts-an-energy-crisis-electricity-may-have-emerged- as-a-challenge/

work page 2024
[2]

[n. d.]. Meta LLaMA 3. https://ai.meta.com/blog/meta-llama-3/

work page
[3]

[n. d.]. NVIDIA Accelerates LLaMA 3. https://blogs.nvidia.com/blog/ meta-llama3-inference-acceleration/

work page
[4]

[n. d.]. Nvidia Power Capping. https://docs.nvidia.com/networking/ display/bluefieldbmcv23107/power+capping

work page
[5]

[n. d.]. nvmlApi. https://developer.nvidia.com/management-library- nvml

work page
[6]

[n. d.]. Sam Altman’s talk at World Economic Forum. https://www. youtube.com/watch?v=xUoAhu2hlWo

work page
[7]

Dario Amodei and Danny Hernandez. 2018. AI and compute. https: //openai.com/research/ai-and-compute

work page 2018
[8]

Srikant Bharadwaj, Shomit Das, Yasuko Eckert, Mark Oskin, and Tushar Krishna. 2021. DUB: Dynamic underclocking and bypassing in nocs for heterogeneous GPU workloads. In Proceedings of the 15th IEEE/ACM International Symposium on Networks-on-Chip . 49–54

work page 2021

Show all 36 references

[9]

Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al . 2020. Language models are few-shot learners. Advances in neural information processing systems 33 (2020), 1877–1901

2020
[10]

Tianqi Chen and Carlos Guestrin. 2016. Xgboost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining . 785–794

2016
[11]

Tianqi Chen, Thierry Moreau, Ziheng Jiang, Lianmin Zheng, Eddie Yan, Haichen Shen, Meghan Cowan, Leyuan Wang, Yuwei Hu, Luis Ceze, et al. 2018. {TVM}: An automated{End-to-End} optimizing compiler for deep learning. In 13th USENIX Symposium on Operating Systems Design and Imple...

2018
[12]

Tianqi Chen, Lianmin Zheng, Eddie Yan, Ziheng Jiang, Thierry Moreau, Luis Ceze, Carlos Guestrin, and Arvind Krishnamurthy. 2018. Learning to optimize tensor programs.Advances in Neural Information Processing Systems 31 (2018)

2018
[13]

Sharan Chetlur, Cliff Woolley, Philippe Vandermersch, Jonathan Co- hen, John Tran, Bryan Catanzaro, and Evan Shelhamer. 2014. cudnn: Efficient primitives for deep learning. arXiv preprint arXiv:1410.0759 (2014)

2014 arXiv
[14]

Miyuru Dayarathna, Yonggang Wen, and Rui Fan. 2015. Data cen- ter energy consumption modeling: A survey. IEEE Communications surveys & tutorials 18, 1 (2015), 732–794

2015
[15]

Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei
[16]

Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova
[17]

Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. 2016. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition . 770–778

2016
[18]

Sunpyo Hong and Hyesoon Kim. 2010. An integrated GPU power and performance model. In Proceedings of the 37th annual international symposium on Computer architecture . 280–289

2010
[19]

Malith Jayaweera, Martin Kong, Yanzhi Wang, and David Kaeli. 2024. Energy-Aware Tile Size Selection for Affine Programs on GPUs. In 2024 IEEE/ACM International Symposium on Code Generation and Opti- mization (CGO). IEEE, 13–27

2024
[20]

Yunfan Jiang, Agrim Gupta, Zichen Zhang, Guanzhi Wang, Yongqiang Dou, Yanjun Chen, Li Fei-Fei, Anima Anandkumar, Yuke Zhu, and Linxi Fan. 2022. Vima: General robot manipulation with multimodal prompts. In NeurIPS 2022 Foundation Models for Decision Making Workshop

2022
[21]

Vijay Kandiah, Scott Peverelle, Mahmoud Khairy, Junrui Pan, Amogh Manjunath, Timothy G Rogers, Tor M Aamodt, and Nikos Hardavel- las. 2021. AccelWattch: A power modeling framework for modern GPUs. In MICRO-54: 54th Annual IEEE/ACM International Symposium on Microarchitecture. 738–753

2021
[22]

Jonathan Lew, Deval A Shah, Suchita Pati, Shaylin Cattell, Mengchi Zhang, Amruth Sandhupatla, Christopher Ng, Negar Goli, Matthew D Sinclair, Timothy G Rogers, et al. 2019. Analyzing machine learning workloads using a detailed GPU simulator. In 2019 IEEE international symposiu...

2019
[23]

Lingxiao Ma, Zhiqiang Xie, Zhi Yang, Jilong Xue, Youshan Miao, Wei Cui, Wenxiang Hu, Fan Yang, Lintao Zhang, and Lidong Zhou. 2020. Rammer: Enabling holistic deep learning compiler optimizations with {rTasks}. In 14th USENIX Symposium on Operating Systems Design and Implementa...

2020
[24]

David Patterson, Joseph Gonzalez, Quoc Le, Chen Liang, Lluis-Miquel Munguia, Daniel Rothchild, David So, Maud Texier, and Jeff Dean. 2021. Carbon emissions and large neural network training. arXiv preprint arXiv:2104.10350 (2021)

2021 arXiv
[25]

Jonathan Ragan-Kelley, Connelly Barnes, Andrew Adams, Sylvain Paris, Frédo Durand, and Saman Amarasinghe. 2013. Halide: a language , , Yijia Zhang 1, Zhihong Gou1, Shijie Cao2, Weigang Feng1, Sicheng Zhang1, and Guohao Dai1, Ningyi Xu1 and compiler for optimizing parallelism, ...

2013
[26]

Farui Wang, Weizhe Zhang, Shichao Lai, Meng Hao, and Zheng Wang
[27]

Jian Weng, Animesh Jain, Jie Wang, Leyuan Wang, Yida Wang, and Tony Nowatzki. 2021. UNIT: Unifying tensorized instruction compila- tion. In 2021 IEEE/ACM International Symposium on Code Generation and Optimization (CGO). IEEE, 77–89

2021
[28]

Jie You, Jae-Won Chung, and Mosharaf Chowdhury. 2023. Zeus: Un- derstanding and optimizing{GPU} energy consumption of{DNN} training. In 20th USENIX Symposium on Networked Systems Design and Implementation (NSDI 23). 119–139

2023
[29]

Gyeong-In Yu, Joo Seong Jeong, Geon-Woo Kim, Soojeong Kim, and Byung-Gon Chun. 2022. Orca: A distributed serving system for {Transformer-Based} generative models. In 16th USENIX Symposium on Operating Systems Design and Implementation (OSDI 22) . 521–538

2022
[30]

Lianmin Zheng, Chengfan Jia, Minmin Sun, Zhao Wu, Cody Hao Yu, Ameer Haj-Ali, Yida Wang, Jun Yang, Danyang Zhuo, Koushik Sen, et al. 2020. Ansor: Generating{High-Performance} tensor programs for deep learning. In 14th USENIX symposium on operating systems design and implementa...

2020
[31]

Size Zheng, Yun Liang, Shuo Wang, Renze Chen, and Kaiwen Sheng
[32]

Pengfei Zou, Ang Li, Kevin Barker, and Rong Ge. 2020. Indicator- directed dynamic power management for iterative workloads on GPU- accelerated systems. In 2020 20th IEEE/ACM International Symposium on Cluster, Cloud and Internet Computing (CCGRID) . IEEE, 559–568. Guohao Dai1,...

2020
[2009]

In2009 IEEE conference on computer vision and pattern recognition

Imagenet: A large-scale hierarchical image database. In2009 IEEE conference on computer vision and pattern recognition . Ieee, 248–255
[2019]

In Proceedings of NAACL-HLT

BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of NAACL-HLT. 4171–4186
[2020]

In Pro- ceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems

Flextensor: An automatic schedule exploration and optimization framework for tensor computation on heterogeneous system. In Pro- ceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems . 859–873
[2021]

IEEE Transactions on Parallel and Distributed Systems 33, 11 (2021), 2943–2954

Dynamic GPU energy optimization for machine learning training workloads. IEEE Transactions on Parallel and Distributed Systems 33, 11 (2021), 2943–2954

2021

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [1]

[n. d.]. Energy Crisis. https://www.trendforce.com/news/2024/03/28/ news-ai-confronts-an-energy-crisis-electricity-may-have-emerged- as-a-challenge/

work page 2024

[2] [2]

[n. d.]. Meta LLaMA 3. https://ai.meta.com/blog/meta-llama-3/

work page

[3] [3]

[n. d.]. NVIDIA Accelerates LLaMA 3. https://blogs.nvidia.com/blog/ meta-llama3-inference-acceleration/

work page

[4] [4]

[n. d.]. Nvidia Power Capping. https://docs.nvidia.com/networking/ display/bluefieldbmcv23107/power+capping

work page

[5] [5]

[n. d.]. nvmlApi. https://developer.nvidia.com/management-library- nvml

work page

[6] [6]

[n. d.]. Sam Altman’s talk at World Economic Forum. https://www. youtube.com/watch?v=xUoAhu2hlWo

work page

[7] [7]

Dario Amodei and Danny Hernandez. 2018. AI and compute. https: //openai.com/research/ai-and-compute

work page 2018

[8] [8]

Srikant Bharadwaj, Shomit Das, Yasuko Eckert, Mark Oskin, and Tushar Krishna. 2021. DUB: Dynamic underclocking and bypassing in nocs for heterogeneous GPU workloads. In Proceedings of the 15th IEEE/ACM International Symposium on Networks-on-Chip . 49–54

work page 2021