pith. machine review for the scientific record. sign in

arxiv: 2605.04773 · v1 · submitted 2026-05-06 · 💻 cs.GR · cs.PF

Recognition: unknown

AGIPC: Adaptive In-Solve Algebraic Coarsening for GPU IPC

Kemeng Huang, Minchen Li, Taku Komura, Xuan Wang, Zhaofeng Luo

Pith reviewed 2026-05-08 16:37 UTC · model grok-4.3

classification 💻 cs.GR cs.PF
keywords adaptive coarseningGPU simulationimplicit time integrationcontact mechanicsalgebraic reductionIPCNewton solverphysics-based animation
0
0 comments X

The pith

Algebraic coarsening inside Newton solves cuts GPU contact simulation time by up to 3x.

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

The paper introduces a GPU-friendly technique that dynamically reduces the size of the linear system during Newton iterations for implicit time integration in physics simulations involving contact. It achieves this by treating adaptivity as selective edge collapses on the fine mesh, using parallel warp-level hashing to aggregate vertices into coarse super-nodes while leaving protected edges intact to retain detail. Gradients and Hessians from the fine scale are then mapped and reduced algebraically through efficient GPU kernels to build a smaller coarse system, which is solved with preconditioned conjugate gradient before prolongating the result back to the original mesh. This process avoids explicit remeshing that would disrupt parallelism and memory layout, and it works directly with barrier energies for robust contact handling. A sympathetic reader would care because it targets the dominant cost in stiff-material and large-deformation simulations, promising practical speed gains on consumer GPUs without visible changes in behavior.

Core claim

The central claim is that expressing adaptivity through per-edge tags and implementing selective aggregation via warp-level hash mapping and GPU reduction kernels produces a coarse linear system whose PCG solution, when prolonged to the fine mesh, yields dynamics and contact forces equivalent to the full fine-scale solve for IPC, delivering up to 3x speedup over prior GPU IPC solvers across challenging scenarios with visually indistinguishable output.

What carries the argument

The algebraic in-solve coarsening operator that groups fine vertices into super-nodes using parallel edge aggregation and reduction kernels, then prolongates the coarse PCG solution without any topological mesh change.

If this is right

  • The technique integrates directly into existing GPU IPC pipelines without requiring changes to mesh data structures.
  • Protected edges allow preservation of fine geometric features while still reducing overall degrees of freedom.
  • End-to-end GPU execution becomes feasible for stiff and large-deformation contact problems that were previously CPU-bound.
  • The approach maintains visual equivalence while lowering the cost of repeated linear solves inside implicit integration.

Where Pith is reading between the lines

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

  • The same algebraic aggregation idea could apply to other nonlinear solvers in simulation that currently rely on fixed fine meshes.
  • Combining the coarsening with adaptive error-driven tag selection might further improve efficiency on varying problem stiffness.
  • Real-time animation pipelines in games or VR could become viable for previously expensive stiff materials if the speedup scales to interactive rates.
  • Similar in-solve reduction might help scale contact simulations to higher resolutions on current hardware without proportional time increases.

Load-bearing premise

That the algebraically reduced coarse system, after prolongation, produces identical dynamics and contact behavior to the full fine-scale system in every relevant simulation case.

What would settle it

A specific high-curvature cloth folding or multi-object stacking test where the coarsened solver produces visibly different deformations or interpenetrations compared to the fine solver.

Figures

Figures reproduced from arXiv: 2605.04773 by Kemeng Huang, Minchen Li, Taku Komura, Xuan Wang, Zhaofeng Luo.

Figure 1
Figure 1. Figure 1: Soft domino. We simulate a domino scene with deformable cuboids (Young’s Modulus is 3 × 105Pa) and rigid holders (simulated with ABD). As the dominoes fall sequentially, the kinetic and elastic energy distribution of the scene evolves over time. Our algebraic coarsening method adaptively coarsens inactive degrees of freedom at each Newton iteration, achieving a 2× speedup over StiffGIPC while producing vis… view at source ↗
Figure 2
Figure 2. Figure 2: A fine mesh coarsening example. In the left figure, three consecutive nodes are grouped. Blue disks are mesh nodes; blue lines (right) are their edges. Fully connected nodes within a group are aggregated into a single, coarser node (orange disks). This grouping and aggregation are applied recursively. The final coarsened mesh is shown as a large pink disk, the coarse supernode, achieving the target minimum… view at source ↗
Figure 3
Figure 3. Figure 3: A mesh DoF recovery example. The red-crossed edge indicates that this edge is protected from collapse during the mesh coarsening process. node positions. For the gradient (force vector), the contribution of each fine node g𝑓 is accumulated to its mapped coarse node g𝑐 : g𝑐 = Í 𝑓 ∈𝑐ℎ𝑖𝑙𝑑 (𝑐 ) g𝑓 . The Hessian is assembled similarly: each 3 × 3 fine-level Hessian block, stored in BCOO format with row/column i… view at source ↗
Figure 4
Figure 4. Figure 4: Adaptive coarsening criterion. We use Green strain increments to drive collapse decisions, capturing high-frequency elastic waves in both triangle and tetrahedral meshes. (Left) Initial three Newton iterations of a vibrating string; (Right) first two iterations of a hanging cloth. Because deformation states evolve significantly even within a single Newton step, we re-evaluate our criterion per iteration to… view at source ↗
Figure 8
Figure 8. Figure 8: Comparison with AmgX. Solver-level comparison with AmgX under identical convergence criteria. When used as a standalone solver (AmgX-Solver), AmgX converges slowly. When used as a PCG precondi￾tioner (AmgX-PCG), it significantly improves convergence; however, its high construction costs result in slower overall performance than both StiffGIPC and our method. A detailed analysis of the V-cycle settings for … view at source ↗
Figure 7
Figure 7. Figure 7: Comparison with GMG. Framework-level comparison with a geo￾metric multigrid solver under the same Newton tolerance. GMG requires substantially more Newton iterations to converge. Although individual lin￾ear solves are cheaper, the total simulation time for 60 frames is 7.5× longer than ours. Comparison with Multigrid Methods. From the perspective of lin￾ear solvers, our method is conceptually related to mu… view at source ↗
Figure 12
Figure 12. Figure 12: (a) Three squishy balls. Three squishy balls with Young’s modulus 107 , 5 × 106 , and 106 , respectively, falling inside a cylinder. (b) Xmas. A Christmas scene featuring a tree with Young’s modulus 108 and toys with Young’s modulus 107 . The ribbon is a triangular mesh, while the snowflakes are ABD. We set the IPC relative distance ˆ𝑑 to 3𝑒 −4 in both scenes. 5.2 Performance Evaluation We compare our met… view at source ↗
Figure 13
Figure 13. Figure 13: Varying material stiffness. A squishy ball falling onto the ground simulated with different Young’s modulus. The top row shows full-resolution results, while the bottom row shows results produced by our adaptive method. The dynamics are visually comparable across all stiffnesses. For the stiffest material, our method achieves over 3× speedup compared to StiffGIPC. Varying Material Stiffness. We evaluate o… view at source ↗
Figure 15
Figure 15. Figure 15: Scalability test. A cloth with resolutions ranging from 10K to 174K vertices is dropped onto an ABD sphere. As resolution increases, the total runtime of the full-resolution simulation grows superlinearly, while our adaptive method exhibits a much milder increase view at source ↗
Figure 16
Figure 16. Figure 16: Domino SIGGRAPH THANK YOU scene. A large domino scene with over 9K deformable cards and 172K vertices. As the cards fall sequen￾tially, the pattern transitions from SIGGRAPH to THANK YOU. Overall Performance view at source ↗
Figure 1
Figure 1. Figure 1: A direct neighbor hash encoding example. We illustrate the procedure of Algorithm 1 using group 0 from view at source ↗
Figure 2
Figure 2. Figure 2: A hash encoding example under non-mutual connectivity. We illustrate the same construction procedure of Algorithm 1 using an example where connectivity is not mutual. The global coarse index is obtained by adding a prefix-sum over the number of coarse nodes in each group (lines 29 and 34–38, Algorithm 2). Since the first group contains one coarse node 1 , the final global indices for {𝑎3, 𝑎4, 𝑎5} become {1… view at source ↗
Figure 3
Figure 3. Figure 3: Per-timestep speedup analysis. We plot the per-timestep speedup of our framework relative to StiffGIPC. (a) In the dolphin scenario ( view at source ↗
Figure 5
Figure 5. Figure 5: Impact of post-coarsening iterations on solver performance. We sweep the cap on post-coarsening PCG iterations for the squishy-ball scene with Young’s modulus 107 ( view at source ↗
Figure 7
Figure 7. Figure 7: Different adaptive criteria. We evaluate the deformation gradi￾ent rate criterion against our Green strain increment criterion on the 1e7 squishy ball example. (Left) Green strain increment criterion. (Middle) The deformation gradient rate criterion, tuned to the best threshold we could find for performance, exhibits artificial stiffening in spiky regions. (Right) Deformation gradient rate criterion with a… view at source ↗
Figure 6
Figure 6. Figure 6: Analysis of Newton convergence metrics. (Top) 𝐿∞ norm of displacement at convergence (dragon scenario with Δ𝑡 = 0.01𝑠). (Bottom) The corresponding root-mean-square of the energy gradient. Analysis of Corner Cases for Preconditioners. In most scenarios, MAS preconditioner effectively reduces the total simulation time (see view at source ↗
Figure 9
Figure 9. Figure 9: Localized mass-ratio analysis. We evaluate the mass distribution during the Newton iteration exhibiting the highest dynamic range. (a) Global mass distribution. (b) Mass gradient. Despite a maximum-to-minimum mass ratio exceeding 105 , the distribution follows a continuous spatial trend. Third, the aggregated mass field exhibits a remarkably smooth spatial gradient, even in the most extreme cases. As shown… view at source ↗
Figure 8
Figure 8. Figure 8: AmgX iterations when used as a PCG preconditioner. We eval￾uate the effect of the number of AmgX multigrid V-cycles when used as a PCG preconditioner. The x-axis shows the maximum number of V-cycles per preconditioning step. (Left) Total computation time for 100 frames of the armadillo scene. (Right) Total number of PCG iterations. Increasing the number of V-cycles improves PCG convergence, but incurs high… view at source ↗
read the original abstract

Implicit time integration is key to robustly simulating stiff materials and large deformations, but its performance is often dominated by repeatedly solving large linear systems. Adaptive coarsening can reduce this cost by concentrating degrees of freedom (DoF) to where it is most needed, yet conventional explicit remeshing changes connectivity (and often vertex ordering), complicating parallel implementations, harming memory locality, and sometimes being disallowed when it may introduce local geometry intersections. Adaptive subspace approaches avoid topological changes, but basis construction and updates incur irregular data access patterns and typically produce dense system matrices, limiting GPU efficiency and keeping many practical systems CPU-centric. We present algebraic adaptive in-solve coarsening, a GPU-oriented method that dynamically reduces DoF within the Newton solve of implicit time integration without explicit topological modification. Starting from a fine mesh, we express adaptivity as a selective edge-collapse process governed by per-edge tags. Collapsible edges are aggregated in parallel using a warp-level hash mapping scheme that groups fine vertices into coarse super-nodes, while protected edges preserve local detail. This defines an implicit coarse mesh whose linear system is assembled algebraically by mapping and reducing fine-scale gradients and Hessians via efficient GPU reduction kernels. We solve the resulting coarse system with a preconditioned conjugate gradient (PCG) method and then prolongate the solution back to the fine mesh. Our approach integrates seamlessly with IPC's barrier energy and exploits GPU parallelism end-to-end. Across a range of challenging scenarios, we achieve up to 3x speedup over a state-of-the-art GPU IPC solver while producing visually indistinguishable results.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript presents AGIPC, an adaptive in-solve algebraic coarsening method for GPU IPC simulations. It performs selective edge aggregation on a fine mesh using per-edge tags and warp-level hashing to form coarse super-nodes without explicit remeshing, assembles the reduced Newton system algebraically by mapping and reducing fine-scale gradients and Hessians via GPU kernels, solves the coarse system with PCG, and prolongates the solution back to the fine mesh. The central claim is up to 3x speedup over a state-of-the-art GPU IPC solver while producing visually indistinguishable results across challenging scenarios.

Significance. If the algebraic coarsening reliably preserves IPC contact fidelity and dynamics upon prolongation, the method would be a meaningful advance for GPU-accelerated implicit simulation by avoiding topological changes, dense matrices, and irregular access patterns that limit current adaptive approaches. The end-to-end GPU design and seamless integration with barrier energies are practical strengths. However, the significance is limited by the absence of quantitative fidelity metrics, which leaves the speedup claim resting on visual inspection alone.

major comments (2)
  1. [Abstract] Abstract and results description: the claim of 'visually indistinguishable results' and equivalent contact behavior is load-bearing for the 3x speedup assertion, yet no quantitative metrics (e.g., maximum penetration depth, barrier energy residuals, or contact force differences between fine and prolonged solutions) are reported to substantiate that the algebraic reduction of gradients/Hessians preserves IPC constraints.
  2. [Method description] The description of the prolongation step and its effect on nonlinear barrier energies: because IPC energies are evaluated on the fine mesh and are highly nonlinear in penetration depth and normal direction, it is unclear whether simple averaging via edge aggregation (governed by per-edge tags and hashing) guarantees that the prolonged solution satisfies the same barrier constraints as the original fine system, particularly for sharp or thin geometry.
minor comments (1)
  1. [Abstract] The abstract mentions 'a range of challenging scenarios' but does not specify the exact test cases, mesh resolutions, or material parameters used to obtain the 3x figure.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the thoughtful and constructive comments. We address each major comment below and describe the revisions we will incorporate.

read point-by-point responses

  1. Referee: [Abstract] Abstract and results description: the claim of 'visually indistinguishable results' and equivalent contact behavior is load-bearing for the 3x speedup assertion, yet no quantitative metrics (e.g., maximum penetration depth, barrier energy residuals, or contact force differences between fine and prolonged solutions) are reported to substantiate that the algebraic reduction of gradients/Hessians preserves IPC constraints.

    Authors: We agree that quantitative metrics would strengthen the fidelity claims. In the revised manuscript we will add a dedicated results subsection with tables reporting maximum penetration depth, barrier energy residuals, and contact force differences between the fine-scale solver and the prolonged AGIPC solutions for all benchmark scenes. These measurements, computed from our existing simulation data, show deviations well below visual and perceptual thresholds, confirming that the algebraic coarsening preserves IPC contact behavior. revision: yes

  2. Referee: [Method description] The description of the prolongation step and its effect on nonlinear barrier energies: because IPC energies are evaluated on the fine mesh and are highly nonlinear in penetration depth and normal direction, it is unclear whether simple averaging via edge aggregation (governed by per-edge tags and hashing) guarantees that the prolonged solution satisfies the same barrier constraints as the original fine system, particularly for sharp or thin geometry.

    Authors: The prolongation operator replicates the coarse super-node displacement uniformly to every fine vertex belonging to that super-node; it is not an averaging operation. Because barrier energies continue to be evaluated on the original fine mesh after prolongation, any local constraint violation is immediately visible to the subsequent Newton iteration. The algebraic reduction approximates the fine-scale Newton direction, and the fine-mesh energy evaluation plus line search restores fidelity. We will expand the method section with a precise description of the prolongation operator, its interaction with the barrier energy, and a short discussion of behavior on sharp or thin features, supported by the new quantitative metrics. revision: partial

Circularity Check

0 steps flagged

No circularity: algorithmic construction is self-contained

full rationale

The paper describes a GPU-parallel algorithmic pipeline for dynamic edge aggregation, algebraic reduction of fine-scale gradients/Hessians into a coarse Newton system, PCG solve, and prolongation back to the fine mesh. No load-bearing step reduces by construction to a fitted parameter, self-defined quantity, or prior self-citation chain; the method is presented as an explicit sequence of parallel GPU kernels and mappings whose correctness is asserted via empirical timing and visual results rather than tautological re-derivation of inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no explicit free parameters, axioms, or invented entities; the method appears to compose existing GPU primitives and the IPC framework.

pith-pipeline@v0.9.0 · 5594 in / 1101 out tokens · 67590 ms · 2026-05-08T16:37:38.914541+00:00 · methodology

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

62 extracted references · 56 canonical work pages

  1. [1]

    Witkin , editor =

    David Baraff and Andrew P. Witkin , editor =. Large Steps in Cloth Simulation , booktitle =. 1998 , url =. doi:10.1145/280814.280821 , timestamp =

    work page doi:10.1145/280814.280821 1998

  2. [2]

    and Williams, J

    Pentland, A. and Williams, J. , title =. 1989 , isbn =. doi:10.1145/74333.74355 , booktitle =

    work page doi:10.1145/74333.74355 1989

  3. [3]

    2012 , url =

    Alec Jacobson and Ilya Baran and Ladislav Kavan and Jovan Popovic and Olga Sorkine , title =. 2012 , url =. doi:10.1145/2185520.2185573 , timestamp =

    work page doi:10.1145/2185520.2185573 2012

  4. [4]

    2015 , url =

    Yu Wang and Alec Jacobson and Jernej Barbic and Ladislav Kavan , title =. 2015 , url =. doi:10.1145/2766952 , timestamp =

    work page doi:10.1145/2766952 2015

  5. [5]

    2018 , url =

    Christopher Brandt and Elmar Eisemann and Klaus Hildebrandt , title =. 2018 , url =. doi:10.1145/3197517.3201387 , timestamp =

    work page doi:10.1145/3197517.3201387 2018

  6. [6]

    2023 , url =

    Otman Benchekroun and Jiayi Eris Zhang and Siddartha Chaudhuri and Eitan Grinspun and Yi Zhou and Alec Jacobson , title =. 2023 , url =. doi:10.1145/3592404 , timestamp =

    work page doi:10.1145/3592404 2023

  7. [7]

    Ty Trusty and Yun (Raymond) Fei and David I. W. Levin and Danny M. Kaufman , title =. 2024 , url =. doi:10.1145/3687946 , timestamp =

    work page doi:10.1145/3687946 2024

  8. [8]

    Adaptive Physically Based Models in Computer Graphics , journal =

    Pierre. Adaptive Physically Based Models in Computer Graphics , journal =. 2017 , url =. doi:10.1111/CGF.12941 , timestamp =

    work page doi:10.1111/cgf.12941 2017

  9. [9]

    Kaufman and Daniele Panozzo , title =

    Zachary Ferguson and Teseo Schneider and Danny M. Kaufman and Daniele Panozzo , title =. 2023 , url =. doi:10.1145/3592428 , timestamp =

    work page doi:10.1145/3592428 2023

  10. [10]

    O'Brien , title =

    Martin Wicke and Daniel Ritchie and Bryan Matthew Klingner and Sebastian Burke and Jonathan Richard Shewchuk and James F. O'Brien , title =. 2010 , url =. doi:10.1145/1778765.1778786 , timestamp =

    work page doi:10.1145/1778765.1778786 2010

  11. [11]

    O'Brien , title =

    Rahul Narain and Armin Samii and James F. O'Brien , title =. 2012 , url =. doi:10.1145/2366145.2366171 , timestamp =

    work page doi:10.1145/2366145.2366171 2012

  12. [12]

    Kaufman and Chenfanfu Jiang , title =

    Minchen Li and Danny M. Kaufman and Chenfanfu Jiang , title =. ACM Trans. Graph. (SIGGRAPH) , year =

  13. [13]

    , title =

    Pfaff, Tobias and Narain, Rahul and de Joya, Juan Miguel and O'Brien, James F. , title =. ACM Trans. Graph. , month = jul, articleno =. 2014 , issue_date =

    2014

  14. [14]

    O'Brien , title =

    Rahul Narain and Tobias Pfaff and James F. O'Brien , title =. 2013 , url =. doi:10.1145/2461912.2462010 , timestamp =

    work page doi:10.1145/2461912.2462010 2013

  15. [15]

    An implicit frictional contact solver for adaptive cloth simulation , journal =

    Jie Li and Gilles Daviet and Rahul Narain and Florence Bertails. An implicit frictional contact solver for adaptive cloth simulation , journal =. 2018 , url =. doi:10.1145/3197517.3201308 , timestamp =

    work page doi:10.1145/3197517.3201308 2018

  16. [16]

    Designing inflatable structures , journal =

    M. Designing inflatable structures , journal =. 2014 , url =. doi:10.1145/2601097.2601166 , timestamp =

    work page doi:10.1145/2601097.2601166 2014

  17. [17]

    2017 , url =

    Zhongshi Jiang and Scott Schaefer and Daniele Panozzo , title =. 2017 , url =. doi:10.1145/3130800.3130895 , timestamp =

    work page doi:10.1145/3130800.3130895 2017

  18. [18]

    Kaufman , title =

    Jiahao Wen and Jernej Barbic and Danny M. Kaufman , title =. 2025 , url =. doi:10.1145/3731204 , timestamp =

    work page doi:10.1145/3731204 2025

  19. [19]

    Instant neural graphics primitives with a multiresolution hash encoding , year =

    Alexandre Mercier. Adaptive rigidification of elastic solids , journal =. 2022 , url =. doi:10.1145/3528223.3530124 , timestamp =

    work page doi:10.1145/3528223.3530124 2022

  20. [20]

    Adaptive Rigidification of Discrete Shells , journal =

    Alexandre Mercier. Adaptive Rigidification of Discrete Shells , journal =. 2023 , url =. doi:10.1145/3606932 , timestamp =

    work page doi:10.1145/3606932 2023

  21. [21]

    Chitalu and Huancheng Lin and Taku Komura , title =

    Kemeng Huang and Floyd M. Chitalu and Huancheng Lin and Taku Komura , title =. 2024 , url =. doi:10.1145/3643028 , timestamp =

    work page doi:10.1145/3643028 2024

  22. [22]

    2025 , url =

    Kemeng Huang and Xinyu Lu and Huancheng Lin and Taku Komura and Minchen Li , title =. 2025 , url =. doi:10.1145/3735126 , timestamp =

    work page doi:10.1145/3735126 2025

  23. [23]

    , title =

    Li, Minchen and Ferguson, Zachary and Schneider, Teseo and Langlois, Timothy and Zorin, Denis and Panozzo, Daniele and Jiang, Chenfanfu and Kaufman, Danny M. , title =. ACM Trans. Graph. , articleno =. 2020 , issue_date =

    2020

  24. [24]

    Jiayi Eris Zhang and Seungbae Bang and David I. W. Levin and Alec Jacobson , title =. 2020 , url =. doi:10.1145/3414685.3417819 , timestamp =

    work page doi:10.1145/3414685.3417819 2020

  25. [25]

    Kaufman and Minchen Li and Chenfanfu Jiang and Yin Yang , title =

    Lei Lan and Danny M. Kaufman and Minchen Li and Chenfanfu Jiang and Yin Yang , title =. 2022 , url =. doi:10.1145/3528223.3530064 , timestamp =

    work page doi:10.1145/3528223.3530064 2022

  26. [26]

    2019 , url =

    Zangyueyang Xian and Xin Tong and Tiantian Liu , title =. 2019 , url =. doi:10.1145/3355089.3356486 , timestamp =

    work page doi:10.1145/3355089.3356486 2019

  27. [27]

    2015, SIAM Journal on Scientific Computing, 37, S602, doi: 10.1137/140980260

    Maxim Naumov and M. Arsaev and Patrice Castonguay and Jonathan M. Cohen and Julien Demouth and Joe Eaton and Simon K. Layton and N. Markovskiy and Istv. AmgX:. 2015 , url =. doi:10.1137/140980260 , timestamp =

    work page doi:10.1137/140980260 2015

  28. [28]

    ACM Transactions on Graphics (TOG) , volume=

    A GPU-based multilevel additive schwarz preconditioner for cloth and deformable body simulation , author=. ACM Transactions on Graphics (TOG) , volume=. 2022 , publisher=

    2022

  29. [29]

    High-order differentiable autoencoder for nonlinear model reduction,

    Lei Lan and Yin Yang and Danny M. Kaufman and Junfeng Yao and Minchen Li and Chenfanfu Jiang , title =. 2021 , url =. doi:10.1145/3450626.3459753 , timestamp =

    work page doi:10.1145/3450626.3459753 2021

  30. [30]

    2024 , url =

    Wenxin Du and Siqiong Yao and Xinlei Wang and Yuhang Xu and Wenqiang Xu and Cewu Lu , title =. 2024 , url =. doi:10.1109/LRA.2024.3381012 , timestamp =

    work page doi:10.1109/lra.2024.3381012 2024

  31. [31]

    Photo tourism: exploring photo collections in 3d,

    Bryan Matthew Klingner and Bryan E. Feldman and Nuttapong Chentanez and James F. O'Brien , title =. 2006 , url =. doi:10.1145/1141911.1141961 , timestamp =

    work page doi:10.1145/1141911.1141961 2006

  32. [32]

    Bargteil and Christopher Wojtan and Jessica K

    Adam W. Bargteil and Christopher Wojtan and Jessica K. Hodgins and Greg Turk , title =. 2007 , url =. doi:10.1145/1276377.1276397 , timestamp =

    work page doi:10.1145/1276377.1276397 2007

  33. [33]

    2011 , url =

    Jin Huang and Yiying Tong and Kun Zhou and Hujun Bao and Mathieu Desbrun , title =. 2011 , url =. doi:10.1109/TVCG.2010.109 , timestamp =

    work page doi:10.1109/tvcg.2010.109 2011

  34. [34]

    2011 , url =

    Jernej Barbic and Yili Zhao , title =. 2011 , url =. doi:10.1145/2010324.1964986 , timestamp =

    work page doi:10.1145/2010324.1964986 2011

  35. [35]

    James , editor =

    Theodore Kim and Doug L. James , editor =. Physics-based Character Skinningusing Multi-Domain Subspace Deformations , booktitle =. 2011 , url =. doi:10.2312/SCA/SCA11/063-072 , timestamp =

    work page doi:10.2312/sca/sca11/063-072 2011

  36. [36]

    Kaufman and David I

    Ty Trusty and Otman Benchekroun and Eitan Grinspun and Danny M. Kaufman and David I. W. Levin , editor =. Subspace Mixed Finite Elements for Real-Time Heterogeneous Elastodynamics , booktitle =. 2023 , url =. doi:10.1145/3610548.3618220 , timestamp =

    work page doi:10.1145/3610548.3618220 2023

  37. [37]

    COMPUTER GRAPHICS forum , volume =

    Affinification: A Fine Approximation of Deformations , author =. COMPUTER GRAPHICS forum , volume =

  38. [38]

    Modal Warping: Real-Time Simulation of Large Rotational Deformation and Manipulation , journal =

    Min Gyu Choi and Hyeong. Modal Warping: Real-Time Simulation of Large Rotational Deformation and Manipulation , journal =. 2005 , url =. doi:10.1109/TVCG.2005.13 , timestamp =

    work page doi:10.1109/tvcg.2005.13 2005

  39. [39]

    James , title =

    Jernej Barbic and Doug L. James , title =. 2005 , url =. doi:10.1145/1073204.1073300 , timestamp =

    work page doi:10.1145/1073204.1073300 2005

  40. [40]

    2015 , url =

    Zherong Pan and Hujun Bao and Jin Huang , title =. 2015 , url =. doi:10.1145/2816795.2818090 , timestamp =

    work page doi:10.1145/2816795.2818090 2015

  41. [41]

    Frame-based elastic models , journal =

    Benjamin Gilles and Guillaume Bousquet and Fran. Frame-based elastic models , journal =. 2011 , url =. doi:10.1145/1944846.1944855 , timestamp =

    work page doi:10.1145/1944846.1944855 2011

  42. [42]

    2011 , url =

    Alec Jacobson and Ilya Baran and Jovan Popovic and Olga Sorkine , title =. 2011 , url =. doi:10.1145/2010324.1964973 , timestamp =

    work page doi:10.1145/2010324.1964973 2011

  43. [43]

    James , title =

    Theodore Kim and Doug L. James , title =. 2009 , url =. doi:10.1145/1618452.1618469 , timestamp =

    work page doi:10.1145/1618452.1618469 2009

  44. [44]

    Christopher Brandt and Klaus Hildebrandt , title =. Comput. Aided Geom. Des. , volume =. 2017 , url =. doi:10.1016/J.CAGD.2017.03.004 , timestamp =

    work page doi:10.1016/j.cagd.2017.03.004 2017

  45. [45]

    Magnor and Christian Theobalt , title =

    Thomas Neumann and Kiran Varanasi and Stephan Wenger and Markus Wacker and Marcus A. Magnor and Christian Theobalt , title =. 2013 , url =. doi:10.1145/2508363.2508417 , timestamp =

    work page doi:10.1145/2508363.2508417 2013

  46. [46]

    Briggs and Van Emden Henson and Stephen F

    William L. Briggs and Van Emden Henson and Stephen F. McCormick , title =. 2000 , isbn =

    2000

  47. [47]

    Kaufman and Chenfanfu Jiang , title =

    Xinlei Wang and Minchen Li and Yu Fang and Xinxin Zhang and Ming Gao and Min Tang and Danny M. Kaufman and Chenfanfu Jiang , title =. 2020 , url =. doi:10.1145/3386760 , timestamp =

    work page doi:10.1145/3386760 2020

  48. [48]

    Surface multigrid via intrinsic prolongation , journal =

    Hsueh. Surface multigrid via intrinsic prolongation , journal =. 2021 , url =. doi:10.1145/3450626.3459768 , timestamp =

    work page doi:10.1145/3450626.3459768 2021

  49. [49]

    Constrainable Multigrid for Cloth , journal =

    In. Constrainable Multigrid for Cloth , journal =. 2013 , url =. doi:10.1111/CGF.12209 , timestamp =

    work page doi:10.1111/cgf.12209 2013

  50. [50]

    2025 , issue_date =

    Jia. Fast Galerkin Multigrid Method for Unstructured Meshes , journal =. 2025 , url =. doi:10.1145/3763327 , timestamp =

    work page doi:10.1145/3763327 2025

  51. [51]

    2024 , url =

    Liangwang Ruan and Bin Wang and Tiantian Liu and Baoquan Chen , title =. 2024 , url =. doi:10.1145/3687758 , timestamp =

    work page doi:10.1145/3687758 2024

  52. [52]

    McCormick , title =

    Rasmus Tamstorf and Toby Jones and Stephen F. McCormick , title =. 2015 , url =. doi:10.1145/2816795.2818081 , timestamp =

    work page doi:10.1145/2816795.2818081 2015

  53. [53]

    Michels , title =

    Han Shao and Libo Huang and Dominik L. Michels , title =. 2022 , url =. doi:10.1145/3528223.3530109 , timestamp =

    work page doi:10.1145/3528223.3530109 2022

  54. [54]

    2025 , url =

    Tetsuya Takahashi and Christopher Batty , title =. 2025 , url =. doi:10.1109/TVCG.2024.3378725 , timestamp =

    work page doi:10.1109/tvcg.2024.3378725 2025

  55. [55]

    Falgout and Ulrike Meier Yang , editor =

    Robert D. Falgout and Ulrike Meier Yang , editor =. hypre:. Computational Science -. 2002 , url =. doi:10.1007/3-540-47789-6\_66 , timestamp =

    work page doi:10.1007/3-540-47789-6 2002

  56. [56]

    Gropp and Lois Curfman McInnes and Barry F

    Satish Balay and William D. Gropp and Lois Curfman McInnes and Barry F. Smith , editor =. Efficient Management of Parallelism in Object-Oriented Numerical Software Libraries , booktitle =. 1996 , url =. doi:10.1007/978-1-4612-1986-6\_8 , timestamp =

    work page doi:10.1007/978-1-4612-1986-6 1996

  57. [57]

    A Multi-layer Solver for

    Alexandre Mercier. A Multi-layer Solver for. Comput. Graph. Forum , volume =. 2024 , url =. doi:10.1111/CGF.15186 , timestamp =

    work page doi:10.1111/cgf.15186 2024

  58. [58]

    Hierarchical Position Based Dynamics , booktitle =

    Matthias M. Hierarchical Position Based Dynamics , booktitle =. 2008 , url =. doi:10.2312/PE/VRIPHYS/VRIPHYS08/001-010 , timestamp =

    work page doi:10.2312/pe/vriphys/vriphys08/001-010 2008

  59. [59]

    2022 , issue_date =

    Jiayi Eris Zhang and J. Progressive Simulation for Cloth Quasistatics , journal =. 2022 , url =. doi:10.1145/3550454.3555510 , timestamp =

    work page doi:10.1145/3550454.3555510 2022

  60. [60]

    Progressive Shell Qasistatics for Unstructured Meshes , journal =

    Jiayi Eris Zhang and J. Progressive Shell Qasistatics for Unstructured Meshes , journal =. 2023 , url =. doi:10.1145/3618388 , timestamp =

    work page doi:10.1145/3618388 2023

  61. [61]

    James and Danny M

    Jiayi Eris Zhang and Doug L. James and Danny M. Kaufman , title =. 2024 , url =. doi:10.1145/3658214 , timestamp =

    work page doi:10.1145/3658214 2024

  62. [62]

    James and Danny M

    Jiayi Eris Zhang and Doug L. James and Danny M. Kaufman , title =. 2025 , url =. doi:10.1145/3731202 , timestamp =

    work page doi:10.1145/3731202 2025