Pith. sign in

REVIEW 6 minor 84 references

DMRG software packages differ by up to two orders of magnitude in time-to-accuracy, and so do different settings inside one package.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

A cost-and-quality benchmarking framework applied to eight DMRG codes finds up to 100× performance gaps between packages and between parameter settings within a package.

T0 review reviewed 2026-07-31 challenge →

load-bearing objection Solid multi-package DMRG cost-vs-quality benchmarks with real spreads and a usable framework; scope is deliberately narrow but the claims match the data.

arxiv 2607.28369 v1 pith:JCZAGCPI submitted 2026-07-30 physics.comp-ph cs.CEcs.MSphysics.chem-phquant-ph

Performance Benchmarking: Software for the Density Matrix Renormalization Group

Per Sehlstedt , Paolo Bientinesi , Lars Karlsson This is my paper
classification physics.comp-ph cs.CEcs.MSphysics.chem-phquant-ph
keywords DMRGTensor NetworkFrameworkperformance benchmarkingmatrix product statesscientific softwaretime-to-accuracy
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

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 density matrix renormalization group (DMRG) is a workhorse method for quantum many-body problems, yet more than fifty software packages implement it with almost no shared standard for comparing how fast they reach a given accuracy. This paper argues that meaningful comparison requires jointly tracking computational cost and solution quality, using the same models, controlled parameters, and controlled hardware, and measuring end-to-end behavior rather than isolated sweep times. Applying that framework to eight open-source packages on three standard one-dimensional lattice models, the authors find large gaps: packages can differ by factors of four to twenty (and in places nearly two orders of magnitude) in wall time to a target energy error even when settings are aligned, while changing Krylov dimension, floating-point precision schedule, 1-site versus 2-site updates, or subspace expansion inside a single package can produce comparable or larger swings. The practical message is that rigorous, cost-versus-quality benchmarking is itself a research tool: it exposes non-obvious trade-offs, shows that strategy often matters as much as package choice, and gives users and developers a shared baseline for deciding what to run and what to improve next.

Core claim

When computational cost and solution quality are evaluated jointly under controlled models, parameters, and hardware, DMRG implementations differ by up to two orders of magnitude in time-to-accuracy—both across packages with aligned settings and within one package under different parameter configurations. Sweep time alone is an incomplete metric; end-to-end wall time versus relative energy error is required for meaningful ranking.

What carries the argument

A performance-oriented benchmarking framework whose five requirements are joint cost–quality evaluation, end-to-end metrics, common model problems, controlled parameter configurations, and controlled hardware/software environments; operationalized here as wall time versus relative energy residual on L=100 open chains for three lattice models.

Load-bearing premise

That simple fixed, aligned baseline settings on single-core one-dimensional chains of length 100 are enough to support general comparative claims about package performance, even though specialized features and per-package optimal tuning are deliberately left unused.

What would settle it

Re-run the same eight packages on the same three models and hardware, but allow each package its best documented multi-feature schedule (mixed precision, hybrid 1-/2-site, subspace expansion, symmetries) and check whether the order-of-magnitude gaps in time-to-accuracy shrink below a factor of a few or reverse.

Watch this falsifier. Get emailed when new claim-graph text bears on it.

If this is right

  • Users choosing a DMRG package should treat time-to-target-accuracy curves, not advertised sweep speed or feature lists alone, as the primary decision data.
  • Developers gain more by exposing tunable, switchable strategies (precision, update mode, expansion) than by micro-optimizing a single fixed path.
  • Future package papers can adopt the five-requirement checklist so new claims become cross-comparable instead of isolated scaling plots.
  • Parameter-tuning effort can rival or exceed package choice in impact, so automated or default-adaptive schedules become high-value targets.
  • The same cost–quality framing can be reused for other tensor-network families beyond MPS-DMRG.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • If strategy flexibility dominates package identity, community effort on interoperable schedule languages or shared backends may pay off more than proliferating new full stacks.
  • The large within-package swings suggest that published ‘default’ timings systematically understate what experienced users can achieve and overstate differences between codes.
  • Extending the suite to two-dimensional lattices, quantum-chemistry Hamiltonians, or multi-node/GPU runs would test whether the same ranking and sensitivity patterns survive outside the 1D condensed-matter regime used here.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

0 major / 6 minor

Summary. The manuscript proposes a performance-oriented benchmarking framework for DMRG software that jointly evaluates computational cost and solution quality under controlled models, parameters, and hardware, then applies it to eight version-pinned open-source packages on three standard 1D models (critical transverse-field Ising, spin-1 Heisenberg, Fermi–Hubbard at U/t=8). Performance is reported as relative energy error versus wall time at the end of each sweep (Figs. 1–7), with averages over 10 random initializations on a single core. The central empirical claim is that time-to-accuracy can differ by up to two orders of magnitude both across packages under aligned simple settings and within a single package under different Krylov dimensions, floating-point precision, update modes, and subspace-expansion strategies. The authors explicitly caveat that the baseline deliberately under-uses specialized features and that no definitive ranking is claimed.

Significance. The work fills a genuine gap: DMRG has >50 implementations yet almost no cross-package, end-to-end cost–quality comparisons under shared models and controlled environments. The five requirements in §3.2 and the joint metrics in §3.3 are a clear, reusable standard. Strengths include version-pinned packages, analytic/high-bond reference energies, public reproducibility repository [95], and transparent discussion (§5.1, §6) that aligned baselines under-represent specialized features. The demonstrated intra-package spreads (e.g., Krylov dimension, mixed precision, 1-site/2-site/SSE hybrids) are practically useful for users and developers and support the claim that rigorous joint evaluation is informative.

minor comments (6)
  1. [§4.2.4, Figs. 1–3] §4.2.4 / Figs. 1–3: State explicitly in the figure captions or main text which packages enforce which symmetries in each panel (especially the non-abelian cases in Figs. 2–3), so readers need not reverse-engineer support from the curves alone.
  2. [§5.1] §5.1: A short table summarizing default/aligned settings (max bond dimension, eigensolver type and Krylov dimension, truncation criterion, 1- vs 2-site) for each package would make the ‘aligned parameters’ claim easier to audit.
  3. [Fig. 5] Fig. 5 caption: The absolute-difference treatment of the single-precision energy error is important; consider also showing the raw energies or a brief note that the single-precision Hamiltonian itself is slightly different.
  4. [§4.3] §4.3: Clarify whether the reported wall times include only the DMRG sweeps or also any per-sweep orthogonalization/truncation bookkeeping that some packages expose separately.
  5. [§2.2, §4.2] Minor typography: ‘V ariational’, ‘T runcation’, ‘T ransverse-field’, ‘F ermi–Hubbard’ appear with stray spaces in headings (§2.2, §4.2); fix for production.
  6. [References] References: a few DOIs/URLs are duplicated or slightly malformed (e.g., repeated url fields in [23]–[27]); a quick bibliography pass would help.

Circularity Check

0 steps flagged

No significant circularity: empirical wall-time vs. energy-error benchmarks against external/high-D references, not self-defining predictions.

full rationale

This paper is an empirical software-performance study, not a first-principles derivation. Its load-bearing claims are measured wall time versus relative energy error (eq. 11) on fixed 1D models under controlled hardware and version-pinned packages (Figs. 1–7; §4–§5). Reference energies are either analytic (Ising, eq. 7) or high-bond-dimension computations cross-checked against thermodynamic-limit literature values (§4.2.4); using Block2/MPSKit at D=1600 to set E0 does not force the time-to-accuracy rankings of the D=100–800 runs by construction. Self-citations ([3] landscape survey; [95] benchmark repo) supply context and reproducibility, not uniqueness theorems or fitted inputs renamed as predictions. No step reduces a claimed prediction to its own definition or fit. Score 0 is appropriate.

Axiom & Free-Parameter Ledger

5 free parameters · 5 axioms · 1 invented entities

The central claims rest on standard DMRG/tensor-network practice, three conventional lattice Hamiltonians, chosen numerical protocols (random MPS init, fixed max bond dimensions, single-core runs), and the authors’ five benchmark requirements. No new physical entities are postulated. Free choices are experimental controls that define the measured regime rather than fitted constants in a theory.

free parameters (5)
  • Chain length L=100, open boundaries = L=100
    Fixed by hand for all experiments; sets finite-size regime and cost scale of every timing curve.
  • Maximum bond dimensions (100 / 400 / 800) = Ising 100; Heisenberg 400; Hubbard 800
    Per-model caps chosen for the package comparison; strongly affect both runtime and reachable accuracy.
  • Model couplings (h/J=1, J=-1, U/t=8 half-filling) = as stated in §4.2.4
    Standard but selective points (critical Ising, AF Heisenberg, intermediate Hubbard); rankings could shift in other regimes.
  • Number of averaged random initializations = 10
    N=10 runs chosen to suppress init noise; not derived.
  • Krylov subspace dimensions tested (3,6,16,30,60) = set {3,6,16,30,60}
    Discrete sweep of solver depth in the MPSKit study; illustrates sensitivity rather than a universal optimum.
axioms (5)
  • domain assumption Variational DMRG energy is an upper bound; lower computed energy (closer to reference E0) indicates better solution quality for the ground state.
    Used throughout §3.3 and eq. (11) as the quality metric.
  • ad hoc to paper Meaningful package comparison requires jointly reporting cost and quality, end-to-end metrics, shared models, controlled parameters, and controlled environments.
    The five requirements in §3.2 are the authors’ normative framework, not a theorem.
  • domain assumption Reference energies (analytic or high-bond SU(2)/U(1)×SU(2) runs) are accurate enough that relative error εr ranks implementations fairly.
    §4.2.4 references underpin all vertical axes in Figs. 1–7.
  • ad hoc to paper Excluding Hamiltonian build and MPS init, and warming up Julia JIT, isolates ‘DMRG call’ performance.
    Measurement protocol in §4.3; affects absolute times and language comparisons.
  • standard math Standard linear-algebra and iterative eigensolver behavior underpins interpretation of sweep cost vs. Krylov dimension and precision.
    Background for §2 and parameter studies in §5.2.
invented entities (1)
  • Performance-oriented DMRG benchmarking framework (five requirements + cost/quality metrics) independent evidence
    purpose: Define what counts as a meaningful cross-implementation comparison and structure the experiments.
    Introduced in §3.2–3.3 as the paper’s methodological object; it is a protocol, not a physical entity.

reviewed 2026-07-31 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Performance Benchmarking: Software for the Density Matrix Renormalization Group." pith.science (2026). https://pith.science/paper/JCZAGCPI

@misc{pith2026260728369,
  author       = {Pith},
  title        = {Pith review of: Performance Benchmarking: Software for the Density Matrix Renormalization Group},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/JCZAGCPI}},
  note         = {Machine review of arXiv:2607.28369}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

The performance of scientific software often determines the scale of problems that can be solved in practice. As multiple implementations of the same algorithm emerge, systematic evaluation is needed to compare their strengths and limitations. The density matrix renormalization group (DMRG) algorithm, widely used to study quantum systems, has over 50 software implementations. These implementations vary in multiple aspects that can strongly affect performance. However, despite the need, performance evaluations of these implementations are scarce and lack a consistent standard; many existing evaluations are either too incomplete to enable meaningful comparisons or focus on objectives other than direct performance comparisons, thereby limiting understanding of how the implementations compare. Here, we present a performance-oriented benchmarking framework to facilitate meaningful comparisons of DMRG implementations, and we apply it to quantify the performance of eight implementations, highlighting similarities and differences among them. Furthermore, we examine multiple parameter settings, optimization strategies, and implementation-specific features to demonstrate how parameter configuration can affect performance and how systematic evaluation can reveal non-obvious trade-offs. The results show significant performance differences, up to two orders of magnitude in some cases, not only between different implementations when aligning parameters, but also within the same implementation when comparing different parameter configurations. Hence, our results demonstrate the significant value and insight that can be gained from conducting rigorous performance evaluations. Using our results and framework as a starting point, more rigorous benchmarking will ultimately help users and developers make informed decisions and support future development efforts to build better, more efficient software.

Figures

Figures reproduced from arXiv: 2607.28369 by Lars Karlsson, Paolo Bientinesi, Per Sehlstedt.

Figure 1
Figure 1. Figure 1: Performance of DMRG implementations on the transverse-field Ising [PITH_FULL_IMAGE:figures/full_fig_p014_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Performance of DMRG implementations on the isotropic spin-1 [PITH_FULL_IMAGE:figures/full_fig_p015_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Performance of DMRG implementations on the Fermi–Hubbard model [PITH_FULL_IMAGE:figures/full_fig_p016_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Performance of MPSKit on the transverse-field Ising model using different Krylov subspace dimensions. 5.2.2 Mixed-precision In [PITH_FULL_IMAGE:figures/full_fig_p018_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Performance of ITensorMPS on the Fermi–Hubbard model using three different floating-point precision strategies: single-precision only, double￾precision only, and starting with single- and switching to double-precision after four sweeps. Because the single-precision Hamiltonian yields lower energy due to numerical imprecision, the single-precision result error is calculated as an absolute difference in eq. … view at source ↗
Figure 6
Figure 6. Figure 6: Performance of YASTN on the Fermi-Hubbard model using three update mode strategies: 1-site only, 2-site only, and alternating 1-site and 2- site. dimension because it is more computationally cost-effective than the 2-site method. Still, it becomes stuck in a suboptimal configuration, as it tends to do when symmetries are enforced, and it reaches only an accuracy of 10−5 . More￾over, since the traditional 1… view at source ↗
Figure 7
Figure 7. Figure 7: Performance of TeNPy on the Fermi-Hubbard model using 2-site DMRG and 1-site DMRG with and without subspace expansion. To start, we again observe the typical drawbacks of the traditional 1-site method, i.e., the need to start at the target bond dimension and the tendency to get stuck in local minima when enforcing symmetries, in this case, resulting in only 10−2 accuracy after approximately 200 s. In contr… view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

84 extracted references · 23 canonical work pages · 1 internal anchor

  1. [1]

    Density matrix formulation for quantum renormaliza- tion groups

    Steven R. White. “Density matrix formulation for quantum renormaliza- tion groups”. In:Physical Review Letters69.19 (Nov. 1992), pp. 2863– 2866.issn: 0031-9007.doi:10.1103/physrevlett.69.2863

    doi:10.1103/physrevlett.69.2863 1992

  2. [2]

    Density-matrix algorithms for quantum renormaliza- tion groups

    Steven R. White. “Density-matrix algorithms for quantum renormaliza- tion groups”. In:Physical Review B48.14 (Oct. 1993), pp. 10345–10356. issn: 1095-3795.doi:10.1103/physrevb.48.10345

    doi:10.1103/physrevb.48.10345 1993

  3. [3]

    The software landscape for the density matrix renor- malization group

    Per Sehlstedt et al. “The software landscape for the density matrix renor- malization group”. In:Computer Physics Communications324 (July 2026), p. 110136.issn: 0010-4655.doi:10.1016/j.cpc.2026.110136

    arXiv 2026

  4. [4]

    The density-matrix renormalization group in the age of matrix product states

    Ulrich Schollw¨ ock. “The density-matrix renormalization group in the age of matrix product states”. In:Annals of Physics326.1 (Jan. 2011), pp. 96– 192.issn: 0003-4916.doi:10.1016/j.aop.2010.09.012

    doi:10.1016/j.aop.2010.09.012 2011

  5. [5]

    A practical introduction to tensor networks: Matrix prod- uct states and projected entangled pair states

    Rom´ an Or´ us. “A practical introduction to tensor networks: Matrix prod- uct states and projected entangled pair states”. In:Annals of Physics349 (Oct. 2014), pp. 117–158.issn: 0003-4916.doi:10.1016/j.aop.2014. 06.013

    doi:10.1016/j.aop.2014 2014

  6. [6]

    Renormalization and tensor prod- uct states in spin chains and lattices

    J Ignacio Cirac and Frank Verstraete. “Renormalization and tensor prod- uct states in spin chains and lattices”. In:Journal of Physics A: Math- ematical and Theoretical42.50 (Dec. 2009), p. 504004.issn: 1751-8121. doi:10.1088/1751-8113/42/50/504004

    work page doi:10.1088/1751-8113/42/50/504004 2009

  7. [7]

    Tensor Network States and Geometry

    G. Evenbly and G. Vidal. “Tensor Network States and Geometry”. In: Journal of Statistical Physics145.4 (June 2011), pp. 891–918.issn: 1572- 9613.doi:10.1007/s10955-011-0237-4

    doi:10.1007/s10955-011-0237-4 2011

  8. [8]

    Tensor networks for complex quantum systems

    Rom´ an Or´ us. “Tensor networks for complex quantum systems”. In:Nature Reviews Physics1.9 (Aug. 2019), pp. 538–550.issn: 2522-5820.doi:10. 1038/s42254-019-0086-7

    2019

  9. [9]

    Classical simulation of quantum many-body systems with a tree tensor network

    Y.-Y. Shi, L.-M. Duan, and G. Vidal. “Classical simulation of quantum many-body systems with a tree tensor network”. In:Physical Review A 74.2 (Aug. 2006).issn: 1094-1622.doi:10.1103/physreva.74.022320

    doi:10.1103/physreva.74.022320 2006

  10. [10]

    Efficient tree tensor network states (TTNS) for quantum chemistry: Generalizations of the density matrix renormalization group algorithm

    Naoki Nakatani and Garnet Kin-Lic Chan. “Efficient tree tensor network states (TTNS) for quantum chemistry: Generalizations of the density matrix renormalization group algorithm”. In:The Journal of Chemical Physics138.13 (Apr. 2013).issn: 1089-7690.doi:10.1063/1.4798639

    doi:10.1063/1.4798639 2013

  11. [11]

    Tensor-Train Decomposition

    I. V. Oseledets. “Tensor-Train Decomposition”. In:SIAM Journal on Sci- entific Computing33.5 (Jan. 2011), pp. 2295–2317.issn: 1095-7197.doi: 10.1137/090752286. 24

    doi:10.1137/090752286 2011

  12. [12]

    The non-Abelian density matrix renor- malization group algorithm

    I. P McCulloch and M Gul´ acsi. “The non-Abelian density matrix renor- malization group algorithm”. In:Europhysics Letters (EPL)57.6 (Mar. 2002), pp. 852–858.issn: 1286-4854.doi:10.1209/epl/i2002-00393-0

    doi:10.1209/epl/i2002-00393-0 2002

  13. [13]

    Non-abelian symmetries in tensor networks: A quantum symmetry space approach

    Andreas Weichselbaum. “Non-abelian symmetries in tensor networks: A quantum symmetry space approach”. In:Annals of Physics327.12 (Dec. 2012), pp. 2972–3047.issn: 0003-4916.doi:10.1016/j.aop.2012.07. 009

    work page doi:10.1016/j.aop.2012.07 2012

  14. [14]

    Studying Two-Dimensional Sys- tems with the Density Matrix Renormalization Group

    E.M. Stoudenmire and Steven R. White. “Studying Two-Dimensional Sys- tems with the Density Matrix Renormalization Group”. In:Annual Re- view of Condensed Matter Physics3.1 (Mar. 2012), pp. 111–128.issn: 1947-5462.doi:10.1146/annurev-conmatphys-020911-125018

    doi:10.1146/annurev-conmatphys-020911-125018 2012

  15. [15]

    Density matrix renormalization group algorithms with a single center site

    Steven R. White. “Density matrix renormalization group algorithms with a single center site”. In:Physical Review B72.18 (Nov. 2005).issn: 1550- 235X.doi:10.1103/physrevb.72.180403

    work page doi:10.1103/physrevb.72.180403 2005

  16. [16]

    Strictly single-site DMRG algorithm with subspace ex- pansion

    C. Hubig et al. “Strictly single-site DMRG algorithm with subspace ex- pansion”. In:Physical Review B91.15 (Apr. 2015).issn: 1550-235X.doi: 10.1103/physrevb.91.155115

    doi:10.1103/physrevb.91.155115 2015

  17. [17]

    Controlled Bond Expan- sion for Density Matrix Renormalization Group Ground State Search at Single-Site Costs

    Andreas Gleis, Jheng-Wei Li, and Jan von Delft. “Controlled Bond Expan- sion for Density Matrix Renormalization Group Ground State Search at Single-Site Costs”. In:Physical Review Letters130.24 (June 2023).issn: 1079-7114.doi:10.1103/physrevlett.130.246402

    doi:10.1103/physrevlett.130.246402 2023

  18. [18]

    Mixed-Precision Implementation of the Density Matrix Renormalization Group

    Yingqi Tian et al. “Mixed-Precision Implementation of the Density Matrix Renormalization Group”. In:Journal of Chemical Theory and Computa- tion18.12 (Nov. 2022), pp. 7260–7271.issn: 1549-9626.doi:10.1021/ acs.jctc.2c00632

    2022

  19. [19]

    Mixed-PrecisionAb InitioTensor Network State Meth- ods Adapted for NVIDIA Blackwell Technology via Emulated FP64 Arith- metic

    Cole Brower et al. “Mixed-PrecisionAb InitioTensor Network State Meth- ods Adapted for NVIDIA Blackwell Technology via Emulated FP64 Arith- metic”. In:Journal of Chemical Theory and Computation(Apr. 2026). issn: 1549-9626.doi:10.1021/acs.jctc.6c00203

    doi:10.1021/acs.jctc.6c00203 2026

  20. [20]

    The ab-initio density matrix renormaliza- tion group in practice

    Roberto Olivares-Amaya et al. “The ab-initio density matrix renormaliza- tion group in practice”. In:The Journal of Chemical Physics142.3 (Jan. 2015).issn: 1089-7690.doi:10.1063/1.4905329

    doi:10.1063/1.4905329 2015

  21. [21]

    Low communication high per- formance ab initio density matrix renormalization group algorithms

    Huanchen Zhai and Garnet Kin-Lic Chan. “Low communication high per- formance ab initio density matrix renormalization group algorithms”. In: The Journal of Chemical Physics154.22 (June 2021).issn: 1089-7690. doi:10.1063/5.0050902

    doi:10.1063/5.0050902 2021

  22. [22]

    High-Performance Computing for Density Matrix Renormalization Group

    Yingqi Tian and Haibo Ma. “High-Performance Computing for Density Matrix Renormalization Group”. In:Current Chinese Science3.3 (June 2023), pp. 178–186.issn: 2210-2981.doi:10.2174/2210298103666221125162959

    work page doi:10.2174/2210298103666221125162959 2023

  23. [23]

    The LINPACK Benchmark: past, present and future

    Jack J. Dongarra, Piotr Luszczek, and Antoine Petitet. “The LINPACK Benchmark: past, present and future”. In:Concurrency and Computation: Practice and Experience15.9 (July 2003), pp. 803–820.issn: 1532-0634. doi:10.1002/cpe.728.url:http://dx.doi.org/10.1002/cpe.728. 25

    work page doi:10.1002/cpe.728.url:http://dx.doi.org/10.1002/cpe.728 2003

  24. [24]

    An Iterative Solver Benchmark

    Jack Dongarra, Victor Eijkhout, and Henk van der Vorst. “An Iterative Solver Benchmark”. In:Scientific Programming9.4 (Jan. 2001), pp. 223– 231.issn: 1875-919X.doi:10.1155/2001/527931.url:http://dx.doi. org/10.1155/2001/527931

    work page doi:10.1155/2001/527931.url:http://dx.doi 2001

  25. [25]

    Introducing the graph 500

    Richard C Murphy et al. “Introducing the graph 500”. In:Cray Users Group (CUG)19.45-74 (2010), p. 22

    2010

  26. [26]

    High-performance conjugate-gradient benchmark: A new metric for ranking high-performance computing systems

    Jack Dongarra, Michael A Heroux, and Piotr Luszczek. “High-performance conjugate-gradient benchmark: A new metric for ranking high-performance computing systems”. In:The International Journal of High Performance Computing Applications30.1 (Aug. 2015), pp. 3–10.issn: 1741-2846.doi: 10 . 1177 / 1094342015593158.url:http : / / dx . doi . org / 10 . 1177 / 10...

    2015

  27. [27]

    MLPerf: An Industry Standard Benchmark Suite for Machine Learning Performance

    Peter Mattson et al. “MLPerf: An Industry Standard Benchmark Suite for Machine Learning Performance”. In:IEEE Micro40.2 (Mar. 2020), pp. 8–16.issn: 1937-4143.doi:10.1109/mm.2020.2974843.url:http: //dx.doi.org/10.1109/MM.2020.2974843

    arXiv 2020

  28. [28]

    The Density Matrix Renor- malization Group in Quantum Chemistry

    Garnet Kin-Lic Chan and Sandeep Sharma. “The Density Matrix Renor- malization Group in Quantum Chemistry”. In:Annual Review of Physical Chemistry62.1 (May 2011), pp. 465–481.issn: 1545-1593.doi:10.1146/ annurev-physchem-032210-103338

    2011

  29. [29]

    The density matrix renormalization group in chemistry and molecular physics: Recent developments and new challenges

    Alberto Baiardi and Markus Reiher. “The density matrix renormalization group in chemistry and molecular physics: Recent developments and new challenges”. In:The Journal of Chemical Physics152.4 (Jan. 2020).issn: 1089-7690.doi:10.1063/1.5129672

    doi:10.1063/1.5129672 2020

  30. [30]

    The ITensor Software Library for Tensor Network Calculations

    Matthew Fishman, Steven White, and Edwin Stoudenmire. “The ITensor Software Library for Tensor Network Calculations”. In:SciPost Physics Codebases(Aug. 2022).doi:10.21468/scipostphyscodeb.4

    doi:10.21468/scipostphyscodeb.4 2022

  31. [31]

    Block2: A comprehensive open source frame- work to develop and apply state-of-the-art DMRG algorithms in electronic structure and beyond

    Huanchen Zhai et al. “Block2: A comprehensive open source frame- work to develop and apply state-of-the-art DMRG algorithms in electronic structure and beyond”. In:The Journal of Chemical Physics159.23 (Dec. 2023).issn: 1089-7690.doi:10.1063/5.0180424

    doi:10.1063/5.0180424 2023

  32. [32]

    Tensor network Python (TeNPy) version 1

    Johannes Hauschild et al. “Tensor network Python (TeNPy) version 1”. In:SciPost Physics Codebases(Nov. 2024).issn: 2949-804X.doi:10 . 21468/scipostphyscodeb.41

    2024

  33. [33]

    YASTN: Yet another symmetric tensor networks; A Python library for Abelian symmetric tensor network calculations

    Marek Rams et al. “YASTN: Yet another symmetric tensor networks; A Python library for Abelian symmetric tensor network calculations”. In: SciPost Physics Codebases(Feb. 2025).issn: 2949-804X.doi:10.21468/ scipostphyscodeb.52

    2025

  34. [34]

    The Cytnx library for tensor networks

    Kai-Hsin Wu et al. “The Cytnx library for tensor networks”. In:Sci- Post Physics Codebases(Mar. 2025).issn: 2949-804X.doi:10.21468/ scipostphyscodeb.53. 26

    2025

  35. [35]

    pyTTN: An open-source toolbox for open and closed system quantum dynamics simulations using tree tensor networks

    Lachlan P. Lindoy et al. “pyTTN: An open-source toolbox for open and closed system quantum dynamics simulations using tree tensor networks”. In:The Journal of Chemical Physics163.20 (Nov. 2025).issn: 1089-7690. doi:10.1063/5.0301775

    work page doi:10.1063/5.0301775 2025

  36. [36]

    2025.doi:10

    Lukas Devos and Jutho Haegeman.TensorKit.jl: A Julia package for large- scale tensor computations, with a hint of category theory. 2025.doi:10. 48550/ARXIV.2508.10076

    Pith/arXiv arXiv doi:10.48550/arxiv.2508.10076 2025

  37. [37]

    High-performance ab initio density matrix renormalization group method: Applicability to large-scale mul- tireference problems for metal compounds

    Yuki Kurashige and Takeshi Yanai. “High-performance ab initio density matrix renormalization group method: Applicability to large-scale mul- tireference problems for metal compounds”. In:The Journal of Chemical Physics130.23 (June 2009).issn: 1089-7690.doi:10.1063/1.3152576

    work page doi:10.1063/1.3152576 2009

  38. [38]

    Matrix product state applications for the ALPS project

    Michele Dolfi et al. “Matrix product state applications for the ALPS project”. In:Computer Physics Communications185.12 (Dec. 2014), pp. 3430– 3440.issn: 0010-4655.doi:10.1016/j.cpc.2014.08.019

    work page doi:10.1016/j.cpc.2014.08.019 2014

  39. [39]

    CheMPS2: A free open-source spin-adapted im- plementation of the density matrix renormalization group for ab initio quantum chemistry

    Sebastian Wouters et al. “CheMPS2: A free open-source spin-adapted im- plementation of the density matrix renormalization group for ab initio quantum chemistry”. In:Computer Physics Communications185.6 (June 2014), pp. 1501–1514.issn: 0010-4655.doi:10.1016/j.cpc.2014.01. 019

    work page doi:10.1016/j.cpc.2014.01 2014

  40. [40]

    Massively parallel quantum chemical density matrix renormalization group method

    Jiri Brabec et al. “Massively parallel quantum chemical density matrix renormalization group method”. In:Journal of Computational Chemistry 42.8 (Dec. 2020), pp. 534–544.issn: 1096-987X.doi:10.1002/jcc.26476

    doi:10.1002/jcc.26476 2020

  41. [41]

    Recent implementations in kylin 1.3: Improved com- putational efficiency ofab initioDMRG and a spin-adapted version of EC-MRCI

    Yinxuan Song et al. “Recent implementations in kylin 1.3: Improved com- putational efficiency ofab initioDMRG and a spin-adapted version of EC-MRCI”. In:Chinese Journal of Chemical Physics38.4 (Aug. 2025), pp. 447–456.issn: 2327-2244.doi:10.1063/1674-0068/cjcp2506083

    work page doi:10.1063/1674-0068/cjcp2506083 2025

  42. [42]

    QCMaquis 4.0: Multipurpose Electronic, Vibra- tional, and Vibronic Structure and Dynamics Calculations with the Den- sity Matrix Renormalization Group

    Kalman Szenes et al. “QCMaquis 4.0: Multipurpose Electronic, Vibra- tional, and Vibronic Structure and Dynamics Calculations with the Den- sity Matrix Renormalization Group”. In:The Journal of Physical Chem- istry A(Aug. 2025).issn: 1520-5215.doi:10.1021/acs.jpca.5c02970

    work page doi:10.1021/acs.jpca.5c02970 2025

  43. [43]

    The density matrix renormalization group for strongly corre- lated electron systems: A generic implementation

    G. Alvarez. “The density matrix renormalization group for strongly corre- lated electron systems: A generic implementation”. In:Computer Physics Communications180.9 (Sept. 2009), pp. 1572–1578.issn: 0010-4655.doi: 10.1016/j.cpc.2009.02.016

    work page doi:10.1016/j.cpc.2009.02.016 2009

  44. [44]

    BAGEL: Brilliantly Advanced General Electronic-structure Library

    Toru Shiozaki. “BAGEL: Brilliantly Advanced General Electronic-structure Library”. In:WIREs Computational Molecular Science8.1 (Aug. 2017). issn: 1759-0884.doi:10.1002/wcms.1331

    doi:10.1002/wcms.1331 2017

  45. [45]

    Open source Matrix Product States: Opening ways to simulate entangled many-body quantum systems in one dimension

    Daniel Jaschke, Michael L. Wall, and Lincoln D. Carr. “Open source Matrix Product States: Opening ways to simulate entangled many-body quantum systems in one dimension”. In:Computer Physics Communica- tions225 (Apr. 2018), pp. 59–91.issn: 0010-4655.doi:10.1016/j.cpc. 2017.12.015. 27

    doi:10.1016/j.cpc 2018

  46. [46]

    quimb: A python package for quantum information and many-body calculations

    Johnnie Gray. “quimb: A python package for quantum information and many-body calculations”. In:Journal of Open Source Software3.29 (Sept. 2018), p. 819.issn: 2475-9066.doi:10.21105/joss.00819

    doi:10.21105/joss.00819 2018

  47. [47]

    PyTeNet: A concise Python implementation of quan- tum tensor network algorithms

    Christian Mendl. “PyTeNet: A concise Python implementation of quan- tum tensor network algorithms”. In:Journal of Open Source Software3.30 (Oct. 2018), p. 948.issn: 2475-9066.doi:10.21105/joss.00948

    work page doi:10.21105/joss.00948 2018

  48. [48]

    QSpace - An open-source tensor library for Abelian and non-Abelian symmetries

    Andreas Weichselbaum. “QSpace - An open-source tensor library for Abelian and non-Abelian symmetries”. In:SciPost Physics Codebases(Nov. 2024). issn: 2949-804X.doi:10.21468/scipostphyscodeb.40

    doi:10.21468/scipostphyscodeb.40 2024

  49. [49]

    SeeMPS: A Python-based Matrix Product State and Tensor Train Library

    Paula Garc ´ ıa-Molina et al.SeeMPS: A Python-based Matrix Product State and Tensor Train Library. 2026.doi:10.48550/ARXIV.2601.16734

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2601.16734 2026

  50. [50]

    Implementation of the SU(2) Hamiltonian symmetry for the DMRG algorithm

    Gonzalo Alvarez. “Implementation of the SU(2) Hamiltonian symmetry for the DMRG algorithm”. In:Computer Physics Communications183.10 (Oct. 2012), pp. 2226–2232.issn: 0010-4655.doi:10.1016/j.cpc.2012. 04.025

    work page doi:10.1016/j.cpc.2012 2012

  51. [51]

    Parallelization strategies for density matrix renormaliza- tion group algorithms on shared-memory systems

    G. Hager et al. “Parallelization strategies for density matrix renormaliza- tion group algorithms on shared-memory systems”. In:Journal of Com- putational Physics194.2 (Mar. 2004), pp. 795–808.issn: 0021-9991.doi: 10.1016/j.jcp.2003.09.018

    work page doi:10.1016/j.jcp.2003.09.018 2004

  52. [52]

    The density matrix renormalization group algorithm on kilo-processor architectures: Implementation and trade-offs

    Csaba Nemes et al. “The density matrix renormalization group algorithm on kilo-processor architectures: Implementation and trade-offs”. In:Com- puter Physics Communications185.6 (June 2014), pp. 1570–1581.issn: 0010-4655.doi:10.1016/j.cpc.2014.02.021

    doi:10.1016/j.cpc.2014.02.021 2014

  53. [53]

    A massively parallel tensor contraction framework for coupled-cluster computations

    Edgar Solomonik et al. “A massively parallel tensor contraction framework for coupled-cluster computations”. In:Journal of Parallel and Distributed Computing74.12 (Dec. 2014), pp. 3176–3190.issn: 0743-7315.doi:10. 1016/j.jpdc.2014.06.002

    2014

  54. [54]

    An algorithm for large scale density matrix renor- malization group calculations

    Garnet Kin-Lic Chan. “An algorithm for large scale density matrix renor- malization group calculations”. In:The Journal of Chemical Physics120.7 (Feb. 2004), pp. 3172–3178.issn: 1089-7690.doi:10.1063/1.1638734

    doi:10.1063/1.1638734 2004

  55. [55]

    The density matrix renormal- ization group for ab initio quantum chemistry

    Sebastian Wouters and Dimitri Van Neck. “The density matrix renormal- ization group for ab initio quantum chemistry”. In:The European Physical Journal D68.9 (Sept. 2014).issn: 1434-6079.doi:10.1140/epjd/e2014- 50500-1

    work page doi:10.1140/epjd/e2014- 2014

  56. [56]

    Matrix product operators, matrix product states, and ab initio density matrix renormalization group algorithms

    Garnet Kin-Lic Chan et al. “Matrix product operators, matrix product states, and ab initio density matrix renormalization group algorithms”. In:The Journal of Chemical Physics145.1 (July 2016).issn: 1089-7690. doi:10.1063/1.4955108

    doi:10.1063/1.4955108 2016

  57. [57]

    Real-space parallel density matrix renormalization group

    E. M. Stoudenmire and Steven R. White. “Real-space parallel density matrix renormalization group”. In:Physical Review B87.15 (Apr. 2013). issn: 1550-235X.doi:10.1103/physrevb.87.155137. 28

    doi:10.1103/physrevb.87.155137 2013

  58. [58]

    Real-space parallel density matrix renormalization group with adaptive boundaries

    Fu-Zhou Chen, Chen Cheng, and Hong-Gang Luo. “Real-space parallel density matrix renormalization group with adaptive boundaries”. In:Chi- nese Physics B30.8 (July 2021), p. 080202.issn: 1674-1056.doi:10 . 1088/1674-1056/abeb08

    2021

  59. [59]

    Numerical assessment for accuracy and GPU acceleration of TD-DMRG time evolution schemes

    Weitang Li, Jiajun Ren, and Zhigang Shuai. “Numerical assessment for accuracy and GPU acceleration of TD-DMRG time evolution schemes”. In:The Journal of Chemical Physics152.2 (Jan. 2020).issn: 1089-7690. doi:10.1063/1.5135363

    doi:10.1063/1.5135363 2020

  60. [60]

    Improved hybrid par- allel strategy for density matrix renormalization group method*

    Fu-Zhou Chen, Chen Cheng, and Hong-Gang Luo. “Improved hybrid par- allel strategy for density matrix renormalization group method*”. In: Chinese Physics B29.7 (Aug. 2020), p. 070202.issn: 1674-1056.doi: 10.1088/1674-1056/ab8a42

    work page doi:10.1088/1674-1056/ab8a42 2020

  61. [61]

    Density Matrix Renormalization Group with Tensor Processing Units

    Martin Ganahl et al. “Density Matrix Renormalization Group with Tensor Processing Units”. In:PRX Quantum4.1 (Feb. 2023).issn: 2691-3399. doi:10.1103/prxquantum.4.010317

    doi:10.1103/prxquantum.4.010317 2023

  62. [62]

    High Performance Single-Site Finite DMRG on GPUs

    Hao Hong et al. “High Performance Single-Site Finite DMRG on GPUs”. In:CAIBDA 2022; 2nd International Conference on Artificial Intelligence, Big Data and Algorithms. 2022, pp. 1–5

    2022

  63. [63]

    High-Performance Tensor-Train Primitives Using GPU Tensor Cores

    Xiao-Yang Liu et al. “High-Performance Tensor-Train Primitives Using GPU Tensor Cores”. In:IEEE Transactions on Computers73.11 (Nov. 2024), pp. 2634–2648.issn: 2326-3814.doi:10.1109/tc.2024.3441831

    arXiv 2024

  64. [64]

    Distributed-Memory DMRG via Sparse and Dense Parallel Tensor Contractions

    Ryan Levy, Edgar Solomonik, and Bryan K. Clark. “Distributed-Memory DMRG via Sparse and Dense Parallel Tensor Contractions”. In:SC20: International Conference for High Performance Computing, Networking, Storage and Analysis. IEEE, Nov. 2020, pp. 1–14.doi:10.1109/sc41405. 2020.00028

    arXiv 2020

  65. [65]

    Massively Parallel Tensor Network State Algorithms on Hybrid CPU-GPU Based Architectures

    Andor Menczer and ¨Ors Legeza. “Massively Parallel Tensor Network State Algorithms on Hybrid CPU-GPU Based Architectures”. In:Journal of Chemical Theory and Computation21.4 (Feb. 2025), pp. 1572–1587.issn: 1549-9626.doi:10.1021/acs.jctc.4c00661

    doi:10.1021/acs.jctc.4c00661 2025

  66. [67]

    Two-dimensional quantum lattice models via mode optimized hybrid CPU-GPU density matrix renormalization group method

    Andor Menczer et al. “Two-dimensional quantum lattice models via mode optimized hybrid CPU-GPU density matrix renormalization group method”. In:Physical Review B109.19 (May 2024).issn: 2469-9969.doi:10.1103/ physrevb.109.195148

    2024

  67. [68]

    Parallel Implementation of the Density Matrix Renormalization Group Method Achieving a Quarter petaFLOPS Per- formance on a Single DGX-H100 GPU Node

    Andor Menczer et al. “Parallel Implementation of the Density Matrix Renormalization Group Method Achieving a Quarter petaFLOPS Per- formance on a Single DGX-H100 GPU Node”. In:Journal of Chemical Theory and Computation20.19 (Sept. 2024), pp. 8397–8404.issn: 1549- 9626.doi:10.1021/acs.jctc.4c00903. 29

    doi:10.1021/acs.jctc.4c00903 2024

  68. [69]

    2024.doi:10.48550/ARXIV.2412

    Andor Menczer and ¨Ors Legeza.Cost optimized ab initio tensor network state methods: industrial perspectives. 2024.doi:10.48550/ARXIV.2412. 04676

    doi:10.48550/arxiv.2412 2024

  69. [70]

    Orbital Optimization of Large Active Spaces via AI- Accelerators

    ¨Ors Legeza et al. “Orbital Optimization of Large Active Spaces via AI- Accelerators”. In:Journal of Chemical Theory and Computation21.13 (June 2025), pp. 6545–6558.issn: 1549-9626.doi:10.1021/acs.jctc. 5c00571

    doi:10.1021/acs.jctc 2025

  70. [71]

    Distributed Multi-GPU Ab Initio Density Matrix Renormalization Group Algorithm with Applications to the P-Cluster of Nitrogenase

    Chunyang Xiang et al. “Distributed Multi-GPU Ab Initio Density Matrix Renormalization Group Algorithm with Applications to the P-Cluster of Nitrogenase”. In:Journal of Chemical Theory and Computation20.2 (Jan. 2024), pp. 775–786.issn: 1549-9626.doi:10.1021/acs.jctc.3c01228

    doi:10.1021/acs.jctc.3c01228 2024

  71. [72]

    Benchmarking Quantum Red TEA on CPUs, GPUs, and TPUs

    Daniel Jaschke et al. “Benchmarking Quantum Red TEA on CPUs, GPUs, and TPUs”. In:Proceedings of the 10th bwHPC Symposium. KIT Scientific Publishing, 2026, pp. 45–60.isbn: 9783731514015.doi:10.58895/ksp/ 1000169488-4.url:http://dx.doi.org/10.58895/ksp/1000169488- 4

    arXiv 2026

  72. [73]

    Error estimates for extrap- olations with matrix-product states

    C. Hubig, J. Haegeman, and U. Schollw¨ ock. “Error estimates for extrap- olations with matrix-product states”. In:Physical Review B97.4 (Jan. 2018).issn: 2469-9969.doi:10.1103/physrevb.97.045125. [74]Block2. 2025.url:https : / / github . com / block - hczhai / block2 - preview. [75]ITensorMPS.jl. 2025.url:https://github.com/ITensor/ITensorMPS. jl. [76]ITe...

    doi:10.1103/physrevb.97.045125 2018

  73. [81]

    Time-Dependent Density Matrix Renormalization Group Algorithms for Nearly Exact Ab- sorption and Fluorescence Spectra of Molecular Aggregates at Both Zero and Finite Temperature

    Jiajun Ren, Zhigang Shuai, and Garnet Kin-Lic Chan. “Time-Dependent Density Matrix Renormalization Group Algorithms for Nearly Exact Ab- sorption and Fluorescence Spectra of Molecular Aggregates at Both Zero and Finite Temperature”. In:Journal of Chemical Theory and Computa- tion14.10 (Aug. 2018), pp. 5027–5039.issn: 1549-9626.doi:10.1021/ acs.jctc.8b00628

    2018

  74. [82]

    Finite Temperature Dynamical Density Matrix Renor- malization Group for Spectroscopy in Frequency Domain

    Tong Jiang et al. “Finite Temperature Dynamical Density Matrix Renor- malization Group for Spectroscopy in Frequency Domain”. In:The Jour- nal of Physical Chemistry Letters11.10 (Apr. 2020), pp. 3761–3768.issn: 1948-7185.doi:10.1021/acs.jpclett.0c00905. 30

    work page doi:10.1021/acs.jpclett.0c00905 2020

  75. [83]

    Time-dependent density matrix renormalization group method for quantum dynamics in complex systems

    Jiajun Ren et al. “Time-dependent density matrix renormalization group method for quantum dynamics in complex systems”. In:WIREs Compu- tational Molecular Science12.6 (Mar. 2022).issn: 1759-0884.doi:10 . 1002/wcms.1614. [84]Renormalizer. 2025.url:https://github.com/shuaigroup/Renormalizer

    2022

  76. [85]

    Efficient numerical simulations with Tensor Networks: Tensor Network Python (TeNPy)

    Johannes Hauschild and Frank Pollmann. “Efficient numerical simulations with Tensor Networks: Tensor Network Python (TeNPy)”. In:SciPost Physics Lecture Notes(Oct. 2018).doi:10.21468/scipostphyslectnotes. 5. [86]TeNPy. 2025.url:https://github.com/tenpy/tenpy. [87]YASTN. 2025.url:https://github.com/yastn/yastn

    doi:10.21468/scipostphyslectnotes 2018

  77. [88]

    The one-dimensional Ising model with a transverse field

    Pierre Pfeuty. “The one-dimensional Ising model with a transverse field”. In:Annals of Physics57.1 (Mar. 1970), pp. 79–90.issn: 0003-4916.doi: 10.1016/0003-4916(70)90270-8

    doi:10.1016/0003-4916(70)90270-8 1970

  78. [89]

    Nonintegrability and quantum spin chains

    G. M¨ uller, J. C. Bonner, and J. B. Parkinson. “Nonintegrability and quantum spin chains”. In:Journal of Applied Physics61.8 (Apr. 1987), pp. 3950–3952.issn: 1089-7550.doi:10.1063/1.338594

    work page doi:10.1063/1.338594 1987

  79. [90]

    Numerical renormalization-group study of low-lying eigenstates of the antiferromagneticS=1 Heisenberg chain

    Steven R. White and David A. Huse. “Numerical renormalization-group study of low-lying eigenstates of the antiferromagneticS=1 Heisenberg chain”. In:Physical Review B48.6 (Aug. 1993), pp. 3844–3852.issn: 1095- 3795.doi:10.1103/physrevb.48.3844

    doi:10.1103/physrevb.48.3844 1993

  80. [91]

    Finite-lattice extrapolations for a Haldane-gap antiferromagnet

    O. Golinelli, Th. Jolicoeur, and R. Lacaze. “Finite-lattice extrapolations for a Haldane-gap antiferromagnet”. In:Physical Review B50.5 (Aug. 1994), pp. 3037–3044.issn: 1095-3795.doi:10.1103/physrevb.50.3037

    doi:10.1103/physrevb.50.3037 1994

Showing first 80 references.

This paper was first reviewed by grok-4.5 on July 31, 2026.