A Combined Parallel-in-time Direct Inverse (ParaDIn)-Parareal Method for Nonlinear Differential Equations

Nail K. Yamaleev; Subhash Paudel

REVIEW 3 major objections 4 minor 31 references

A Combined Parallel-in-time Direct Inverse (ParaDIn)-Parareal Method for Nonlinear Differential Equations

T0 review · 3 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read A combined ParaDIn-Parareal method parallelizes both coarse and fine propagators, yielding speedups up to 124 on 480 cores for nonlinear heat and Burgers equations while reproducing the sequential BDF1 solution exactly.

desk verdict A genuinely new and promising combination, but the printed stopping criterion, exact-solution error, and missing k_P scaling data make the headline speedups unverifiable as written. read the letter →

arxiv 2506.10820 v1 pith:7N2DGRKO submitted 2025-06-12 math.NA cs.NA

Subhash Paudel , Nail K. Yamaleev This is my paper · ORCID

classification math.NAcs.NA MSC 65M0665M5565Y05

keywords parallel-in-timeintegrationParaDInmethodPararealblock-JacobipreconditionernonlinearheatequationBurgersBDF1schemespeedupscaling

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 tries to establish that the parallel-in-time ParaDIn method can be combined with Parareal so that nonlinear PDEs discretized by the first-order implicit BDF1 scheme run on hundreds of cores with speedups that grow with the number of time steps. The obstacle it removes is double: plain ParaDIn only works when the total number of time steps stays below a condition-number limit, and classical Parareal needs a sequential coarse-grid correction that caps speedup. The proposed method partitions time into blocks, solves each block in parallel with ParaDIn, and uses Parareal only to couple the blocks, with both coarse and fine solves themselves parallelized by ParaDIn. The paper reports speedups up to 124 on 480 cores for the 2-D nonlinear heat equation and 79 for the 2-D Burgers equation, with discrete errors identical to the sequential scheme.

What carries the argument

The load-bearing object is the ParaDIn decoupling identity, Eq. (6): multiplying the successive bidiagonal equations of the all-at-once Newton system by products of the Jacobian matrices $A_i$ turns the coupled block-bidiagonal system into a diagonal system in which the update at each time level is solved independently. The combined method uses this identity twice: inside each fine block of length $J$, and on the coarse grid of $M$ points for the Parareal correction. The coarse grid supplies the block-initial conditions that block-Jacobi lacks, and the block partition keeps the product-matrix condition number bounded by limiting the number of factors to $J$.

What would settle it

Take the 2-D nonlinear heat case, fix the block count $M$, and double the total time steps $N_t$ repeatedly, recording how many Parareal iterations are needed to hit the stated tolerance; if that count grows with $N_t$, or if lowering the viscosity or steepening the Burgers shock forces $k_P$ well above 3, the claimed $N_t/(2k_P+1)$ scaling fails. The paper already reports that classical Parareal diverges on these problems, so the same experiment also decides how wide the nonlinear range of the combined method is.

Watch

Extended reading notes

Core claim

At each Newton iteration of the all-at-once space-time system, the paper replaces the sequential time marching with a block-Jacobi preconditioner in time: M blocks of J time steps each, where each block is solved by the ParaDIn decoupling. The block-Jacobi coupling term, the last time level of the previous block, is then corrected by a Parareal coarse-grid iteration on M coarse time points. The central claim is that because ParaDIn makes both the fine-grid and coarse-grid solves fully decoupled across time levels, the coarse-grid correction is no longer a sequential bottleneck, and every time level can sit on its own computing core. If the Parareal loop converges in $k_P$ iterations, the speedup follows $S \approx N_t/(2k_P+1)$, or $S \approx N_t/k_P$ with spatial coarsening, while the discrete solution is exactly the BDF1 solution.

Load-bearing premise

The method's projected speedup is $N_t$ divided by $(2k_P+1)$, so everything rests on the Parareal loop finishing in a small, roughly constant number of sweeps regardless of time-step count, block size, and stiffness; the paper reports 2--3 sweeps numerically but supplies no convergence theory for the nonlinear problems.

Editorial extensions

If this is right

Adding time steps adds parallel work without a serial coarse solve, so the $N_t/(2k_P+1)$ estimate predicts speedup roughly proportional to the number of time levels, not the spatial grid size.
The discrete solution is bit-for-bit the sequential BDF1 solution, so all accuracy and convergence properties of BDF1 carry over; the numerical tables show identical $L^2$ and $L^1$ errors.
Spatial coarsening of the coarse-grid correction roughly doubles the projected speedup to $N_t/k_P$ for the parabolic test case, reaching 124 on 480 cores.
Baseline Parareal on the same nonlinear problems saturates near 4.3 (heat) and 3.2 (Burgers) speedup, so the combination is what opens up time-parallel scalability for these nonlinear and shock-containing cases.

Reading between the lines

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

Beyond the paper, if $k_P$ stays bounded when $N_t$ and block size grow, the same construction should scale to thousands of cores for long time horizons; the natural test is strong scaling with $M$ fixed and $N_t$ doubled.
Beyond the paper, the block-Jacobi-plus-ParaDIn structure is not tied to BDF1: it could wrap higher-order implicit Runge-Kutta or BDF2 stages, where the same decoupling identity applies to each stage's all-at-once system.
Beyond the paper, a convergence bound for the linearized Parareal iteration on the all-at-once system would turn the empirical $k_P \approx 2\text{--}3$ into a provable complexity statement; until then, the speedup formula rests on the observed iteration counts.
Beyond the paper, the reported failure of spatial coarsening for Burgers suggests that for convection-dominated problems only temporal coarsening is safe, which may limit memory savings but not the parallel scaling.

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

Two things to know. First, the paper's actual idea is good: plug ParaDIn into both the fine and coarse propagators of Parareal so the coarse-grid correction stops being sequential. I checked Theorem 1 and Proposition 1; the block Gaussian decoupling and the M-iteration block-Jacobi convergence are algebraically correct, and the authors are upfront that classical Parareal diverges on these nonlinear problems. If the speedups are real, this is a useful step for parallel-in-time methods on nonlinear PDEs.

Second, do not trust the headline numbers until three things are fixed. The stopping criterion in Sec. 5.4 is printed as ||(Δu_c)_{k−1} − (Δu_c)_{k−1}|| < ε_P, which is identically zero. I assume it is a typo for (Δu_c)_k − (Δu_c)_{k−1} or a fine-grid residual norm, but as written it gives no reproducible basis for k_P = 2, and there is no table or log showing how k_P behaves as N_t, M, or J grow. Every speedup formula is dominated by the 1/(2k_P+1) factor, so k_P's scaling is the core empirical question. The paper does not answer it.

Third, the claimed consistency with Eq. (21) uses p = 5, while the text allows p ≥ 3. With p = 3 the predicted speedup on the 480-core heat run would be roughly 38, not 88. That discrepancy needs an explanation, not a silent choice of exponent.

Fourth, the printed exact solution for the nonlinear heat test does not satisfy Eq. (1) unless μ0 = α; the stated values μ0 = 1e−6 and α = 1 are inconsistent. The L2 comparison between sequential and parallel schemes is still meaningful as a code-to-code check, but the 'exact solution' label is wrong.

Minor: the baseline Parareal speedup caps of 4.3 and 3.2 are asserted without a table, and no code or data are shipped. For a paper whose central claim is a measured speedup, that is a real gap.

The underlying method is worth taking seriously. The math is checkable and the limitations section is honest. But the paper needs a revision that fixes the stopping criterion, adds an iteration-count scaling study, reconciles the p inconsistency, and ships at least the iteration logs. I would send it to peer review, with the expectation of major revision.

Referee Report

3 major / 4 minor

Summary. This paper proposes a hybrid parallel-in-time method for the implicit BDF1 discretization of nonlinear PDEs. It partitions the time domain into M blocks, solves each block with the ParaDIn direct-inverse method, and accelerates the resulting block-Jacobi iteration with a Parareal correction in which both the coarse- and fine-grid propagators are themselves implemented in parallel via ParaDIn. The paper proves the algebraic decoupling identity Eq. (6) and shows that pure block Jacobi with exact block solves converges in exactly M iterations (Proposition 1). Speedup estimates are derived in Eqs. (20)-(24), and numerical experiments on 2-D nonlinear heat and Burgers equations report speedups up to 124 on 480 cores.

Significance. If the reported iteration counts are stable, the paper offers a practical way around ParaDIn's time-level constraint and improves on classical Parareal by parallelizing the coarse-grid correction. The two algebraic results are clean and correct, and the accuracy tables show the method reproduces the sequential BDF1 solution. However, the speedup claims rest on k_P staying small, and the manuscript currently provides no verifiable stopping criterion and no iteration-count scaling data, so the central quantitative claim is not yet established.

major comments (3)

[Sec. 5.4] The stopping criterion printed in Section 5.4 is ||(Δu_c)_{k-1} − (Δu_c)_{k-1}|| < ε_P, which compares the vector with itself and is identically zero for every k. As written, the Parareal iteration would terminate immediately after initialization, so the reported k_P = 2 for the heat equation and the speedup estimates in Eqs. (21) and (24), which depend on 1/(2k_P+1) or 1/k_P, have no reproducible convergence basis. Please correct the criterion, for example to ||(Δu_c)_k − (Δu_c)_{k-1}|| < ε_P or a coarse/fine mismatch norm, and report k_P for every run, including its dependence on N_t, M, and J at fixed tolerances.
[Secs. 5.5 and 7] The central speedup formulas S ≈ N_t/(2k_P+1) and S ≈ N_t/k_P require k_P to be small and roughly independent of N_t, M, and J. The paper provides no convergence theory for the nonlinear problems (Section 5 notes that none exists for classical Parareal, and Section 7 says classical Parareal diverges on these benchmarks), and no table or log of k_P versus N_t, M, or J is given. Moreover, in Tables 2, 3, and 5 the block size J = N_t/M is held at 30 for all runs, so the experiments cannot separate how k_P depends on J versus M. Please add scaling studies with fixed tolerances and varying N_t, M, and J independently, together with convergence histories.
[Eqs. (20)-(21) and (23)-(24), Tables 2-3] The speedup model contains an unquantified exponent p (the text before Eq. (20) only states p ≥ 3), and the reported agreement with measurements is not parameter-free. With the stated values c_f = 4, c_s = 2, d = 2, and k_P = 2, the 480-core speedup of 87.6 in Table 2 matches Eq. (21) only for p ≈ 5, whereas the 480-core speedup of 124.2 in Table 3 matches Eq. (24) only for p ≈ 4.4; no value of p is specified or measured. Please define p, derive it from the actual coarse-grid solver cost, and state the constants used when comparing the formulas with the measured speedups.

minor comments (4)

[Eq. (18)] The labels in Eq. (18) appear swapped: the coarse-grid solve should carry the O(M^2 N_s) product cost and the fine-grid solve the O(J^2 N_s) cost, since M is the number of coarse time levels and J is the number of fine time levels per block.
[Eqs. (15) and (17)] There is a stray 'Δ' before A_c^i in the product notation in Eqs. (15) and (17).
[Proof of Theorem 1 and Sec. 7.1] In the proof of Theorem 1, 'multiplied by by' should be 'multiplied by'; in Section 7.1, 'Form these results' should be 'From these results'.
[Sec. 7] The comparison with baseline Parareal, which is reported only as maximum speedups of 4.3 and 3.2, omits the implementation details and convergence histories needed to reproduce those numbers; please provide them or state where the implementation is available.

Circularity Check

0 steps flagged · score 2.0 of 10

No significant circularity: the measured speedups are independent runtime data, though the printed Parareal stopping criterion is vacuous and the constant k_P is not documented across grid sizes.

full rationale

The derivation chain is not circular. The central assertion—that the combined ParaDIn-Parareal method reaches speedups of about 88–124 on 480 cores—is an empirical runtime measurement (Tables 2 and 3), not a quantity obtained by substituting the target result into the model. The algebraic equivalence of the all-at-once system (5) and the decoupled system (6) is proved in Theorem 1 by block Gaussian elimination, so the ParaDIn decoupling is a self-contained transformation rather than an ansatz that assumes the result. The cost bounds borrowed from the authors' earlier paper [29] are parameter-free complexity estimates whose assumptions (N_t < N_s^{1/2}; M, J much less than N_s^{1/2}) do not include the target speedup; they are self-citations but not load-bearing in the sense of reducing the claim to a self-asserted premise. Speedup formulas (21) and (24) are post-hoc performance models: they take measured k_P, c_f, c_s, and p as inputs and are checked loosely against measured runtimes, so they are not fitted predictions of the measured speedups. Flagged but not counted as circularity: Section 5.4's only stated stopping criterion, "‖(Delta u_c)_{k-1} - (Delta u_c)_{k-1}‖ < epsilon_P", compares the vector with itself and is identically zero. As printed, it cannot produce the reported k_P=2, and the paper gives no table showing k_P versus N_t, M, or J. This is a serious reproducibility/correctness defect in the convergence claim, but it is a typo or omission rather than a circular derivation: the speedups themselves are measured data. The absence of a nonlinear convergence theory (noted in Section 5) and the reported divergence of classical Parareal on the same benchmarks further mean the constant-k_P assumption is not theoretically justified, but again that is a support gap, not circularity.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The algorithm rests on the ParaDIn machinery imported from the authors' prior paper [29]: the condition-number growth constraint that motivates the block decomposition, and the O(J^2 N_s) per-core cost estimate for forming product matrices. The paper adds the empirical assumption that the outer Parareal iteration converges in a small constant number of iterations, which is the most fragile premise. M, J, c_f, c_s and the tolerances are hand-chosen; the cost-scaling exponent p is left as an unspecified p>=3, though matching the measured speedups requires p=5. No new physical entities are introduced.

free parameters (6)

k_P (number of Parareal iterations) = 2 for the 2D heat problem on all grids; not tabulated for Burgers
The speedup estimates (21) and (24) are linear in 1/(2k_P+1) or 1/k_P; k_P is measured per test case after convergence to tolerance, not predicted. No study of how k_P scales with N_t, M, or stiffness is given.
c_f (space-time coarsening factor for initial guess) = 4 for heat, 3 for Burgers
Hand-chosen constant; the initial guess is the sequential BDF1 solution on a grid coarsened by c_f in time and each spatial direction. The cost of this initialization, N_t/c_f^p W_sol, enters the denominator of Eqs. (20) and (23).
c_s (spatial coarsening factor for coarse-grid correction) = 2 (heat case only)
Hand-chosen; reduces coarse-grid correction cost by c_s^d and therefore appears in Eq. (23). Spatial coarsening was not used for Burgers because it diverges.
p (cost-scaling exponent of the direct solver) = unspecified; stated p >= 3 for 2-D, consistency with measurements requires p = 5
The claimed consistency of Table 2 with Eq. (21) and of Table 3 with Eq. (24) holds only if p = 5 (c_f^p = 1024 for c_f=4); the paper never fixes p, so the speedup model is not a tight predictor.
M and J (number of blocks and time levels per block) = M up to 16, J = 30 at N_t = 480
Algorithm configuration chosen by hand; the constraints M, J << N_s^(1/2) = 64 justify the O((M^2+J^2)N_s) per-core cost, but at N_t=480, (M^2+J^2)N_s is comparable to a banded-direct solve cost, so the assumed dominance of W_sol is borderline.
tolerances eps_N, eps_P, safety factor C_sf = not reported numerically
The number of Parareal and Newton iterations (and hence the measured k_P feeding the speedup estimate) depends on these; only the relation eps_P = C_sf * eps_N with C_sf << 1 is given (Section 5.4).

assumptions (5)

domain assumption The condition number of the ParaDIn product matrices grows with the number of time levels, so the number of time levels must satisfy N_t < N_s^(1/2) for the O(N_t^2 N_s) per-core cost estimate of Theorem 2 in [29] to hold.
This is the load-bearing premise that motivates the block decomposition; it is imported from the authors' prior work [29] and not re-derived or numerically verified here. The paper does not report condition numbers or inner-solve residuals for the J=30 blocks used in Section 7.
domain assumption The per-core cost of forming the product matrices and right-hand sides, O((M^2+J^2)N_s), is negligible compared with W_sol, the cost of one banded direct solve (Eqs. (18)-(19)).
Used to drop the O((M^2+J^2)N_s) term in Eqs. (20) and (23). With N_s=4096, M=16, J=30, this term is about 4.7e6 operations, the same order as a dense-band solve, so the neglect is questionable.
domain assumption Communication cost W_com is negligible compared with W_sol (Section 5.5).
Required for the speedup formulas; the runtime jump from 648.3 s (240 cores) to 926.9 s (480 cores) in Table 2 indicates visible communication or load imbalance, so the assumption is only approximately true.
ad hoc to paper The outer Parareal iteration converges to the tolerance in a small, roughly constant number of iterations for the Newton-linearized all-at-once systems.
Section 5 states there is no convergence theory for nonlinear Parareal, and Section 7 reports classical Parareal diverges on these test problems. The combined method's convergence (k_P = 2 for heat) is shown on two test cases only, with no scaling study; this is the most fragile premise of the speedup argument.
domain assumption The initial guess for the global Newton method is obtained by a sequential BDF1 solve on a grid coarsened by c_f in time and space, interpolated to the fine grid (Section 7.1).
The cost of this sequential initialization is part of the denominator of Eqs. (20) and (23). Its feasibility and accuracy are asserted, not analyzed; for the Burgers case with c_f = 3 the coarse-grid solve still has N_t/3 sequential steps.

how reviews work

0 comments

Cite this review

Pith. "Pith review of A Combined Parallel-in-time Direct Inverse (ParaDIn)-Parareal Method for Nonlinear Differential Equations." pith.science (2026). https://pith.science/paper/7N2DGRKO

@misc{pith2026250610820,
  author       = {Pith},
  title        = {Pith review of: A Combined Parallel-in-time Direct Inverse (ParaDIn)-Parareal Method for Nonlinear Differential Equations},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7N2DGRKO}},
  note         = {Machine review of arXiv:2506.10820}
}

abstract

As has been shown in our previous work, the parallel-in-time direct inverse (ParaDIn) method introduced by Yamaleev and Paudel in (arXiv: 2406.00878v1, 2024) imposes some constraint on the maximum number of time levels, $N_t$, that can be integrated in parallel. To circumvent this problem and further increase the speedup, we combine the ParaDIn method with the Parareal algorithm to efficiently parallelize the first-order time derivative term in nonlinear partial differential equations discretized by the method of lines. The main idea of the proposed approach is to use a block-Jacobi preconditioner, so that each block is solved by using the ParaDIn method. To accelerate the convergence of Jacobi iterations, we use the Parareal method which can be interpreted as a two-level multigrid method in time. In contrast to the conventional Parareal algorithm whose coarse grid correction step is performed sequentially, both the coarse- and fine-grid propagators in the proposed approach are implemented in parallel by using the ParaDIn method, thus significantly increasing the parallel performance of the combined algorithm. Numerical results show that the new combined ParaDIn-Parareal method provides the speedup of up to 124 on 480 computing cores as compared with the sequential first-order implicit backward difference (BDF1) scheme for the 2-D nonlinear heat and Burgers equations with both smooth and discontinuous solutions.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 30 canonical work pages

[1]

50 years of time parallel time integration,

M. J. Gander, “50 years of time parallel time integration,” in Multi- ple Shooting and Time Domain Decomposition Methods, T. Carraro, M. 21 Geiger, S. K¨ orkel, and R. Rannacher, eds., Springer, 2015, pp. 69–114

work page 2015
[2]

Applications of time parallelization,

B. W. Ong and J. B. Schroder, “Applications of time parallelization,” Com- put. Visual. Sci.,Vol. 23, No. 11, 2020

work page 2020
[3]

A parareal in time discretization of PDEs,

J.-L. Lions, Y. Maday, and G. Turinici, “A parareal in time discretization of PDEs,” C. R. Acad. Sci. Paris, Serie I, Vol. 332, 2001, pp. 661-668

work page 2001
[4]

Stable parareal in time method for first- and second- order hyperbolic systems,

X. Dai and Y. Maday, “Stable parareal in time method for first- and second- order hyperbolic systems, ” SIAM J. Sci. Comput., Vol. 35 2013, pp. A52- A78

work page 2013
[5]

A hybrid parareal spectral deferred corrections method,

M. L. Minion, “A hybrid parareal spectral deferred corrections method, ” Comm. App. Math. and Comp. Sci.,Vol. 5, 2010, pp. 265-301

work page 2010
[6]

On the Use of Reduced Basis Methods to Accelerate and Stabilize the Parareal Method

F. Chen, J.S. Hesthaven, and X. Zhu, “On the Use of Reduced Basis Methods to Accelerate and Stabilize the Parareal Method”, Reduced Order Methods for Modeling and Computational Reduction, Springer Inter. Publ. 2014, pp. 187–214

work page 2014
[7]

Stability of the Parareal Algorithm,

G. A. Staff and E. M. Rønquist, “Stability of the Parareal Algorithm,” Do- main Decomposition Methods in Science and Engineering, Springer Berlin Heidelberg, 2005, pp. 449–456

work page 2005
[8]

Toward parallel coarse grid correction for parareal algorithm,

S.-L. Wu, “Toward parallel coarse grid correction for parareal algorithm,” SIAM J. Sci. Comput., Vol. 40, No. 3 2018, pp. A1446 - A1472

work page 2018

Show all 31 references

[9]

A nonlinear paraexp algorithm

M. J. Gander, S. G¨ uttel, M. Petcu, “A nonlinear paraexp algorithm”. In: P.E. Bjorstad, S.C. Brenner, L. Halpern, H.H. Kim, R. Kornhuber, T. Rahman, O.B. Widlund (eds.) Domain Decomposition Methods in Science and Engineering XXIV, pp. 261–270. Springer International Publishin...

2018
[10]

Optimal Schwarz waveform relax- ation for the one-dimensional wave equation,

M. J. Gander, L. Hallern, and F. Nataf, “Optimal Schwarz waveform relax- ation for the one-dimensional wave equation, ” SIAM J. Numer. Analysis, Vol. 41, 2003, pp. 1643-1681

2003
[11]

The parareal in time iterative solver: A further direction to parallel implementation,

Y. Maday and G. Turinici, “The parareal in time iterative solver: A further direction to parallel implementation,” in Domain decomposition methods in science and engineering, Springer, 2005, pp. 441-448

2005
[12]

Time spectral method for periodic and quasi-periodic unsteady computations on unstructured meshes,

D. J. Mavriplis and Z. Yang, “Time spectral method for periodic and quasi-periodic unsteady computations on unstructured meshes,” AIAA Pa- per 2006-5034, 2006

2006
[13]

Towards ef- ficient parallel-in-time simulation of periodic flows,

J. Leffell, J. Sitaraman, V. Lakshminarayan, and A. Wissink, “Towards ef- ficient parallel-in-time simulation of periodic flows,” em AIAA Paper 2016- 0066, 2016. 22

2016
[14]

Parabolic multigrid methods,

W. Hackbusch, “Parabolic multigrid methods, ” In: Computing methods in applied sciences and engineering, VI (Versailles, 1983), North-Holland, Amsterdam, 1984, pp. 189–197

1983
[15]

Analysis of a new space-time parallel multigrid algorithm for parabolic problems,

M. J. Gander and M. Neumuller, “Analysis of a new space-time parallel multigrid algorithm for parabolic problems,” SIAM J. Sci. Comp., Vol. 38, No. 4, 2016, pp. A2173-A2208

2016
[16]

Parallelization in time through tensor- product space–time solvers,

Y. Maday and E. M. Ronquist, “Parallelization in time through tensor- product space–time solvers,” C. R. Acad. Sci. Paris, Ser. I,Vol. 346, 2008, pp. 113–118

2008
[17]

Preconditioning and iterative solution of all-at-once systems for evolutionary partial differential equa- tions,

E. McDonald, J. Pestana, and A. Wathen, “Preconditioning and iterative solution of all-at-once systems for evolutionary partial differential equa- tions,” SIAM J. Sci. Comput., Vol. 40, 2018, pp. A1012-A1033

2018
[18]

Time parallelization for nonlinear problems based on diagonalization,

M. J. Gander and L. Halpern, “Time parallelization for nonlinear problems based on diagonalization,” Lect. Notes Comput. Sci. Eng., Vol. 116, 2017, pp. 163-170

2017
[19]

Space-Time Methods: Fast Solvers and Applications,

M. Neum¨ uller, “Space-Time Methods: Fast Solvers and Applications,” PhD thesis, University of Graz, 2013

2013
[20]

Multigrid methods with space–time concurrency,

R. D. Falgout, S. Friedho, T. V. Kolev, S. P. MacLachlan, J. B. Schroder, S. Vandewalle, “Multigrid methods with space–time concurrency,” Comput. and Visual. Sci.,Vol. 18, No. 4, 2017, pp. 123–143

2017
[21]

Analysis of the parareal time-parallel time- integration method,

J. Gander and S. Vandewalle, “Analysis of the parareal time-parallel time- integration method, ” SIAM J. Sci. Comput., Vol. 29, 2007, pp. 556–578

2007
[22]

Convergence analysis of the Parareal-Euler algorithm for sys- tems of ODEs with complex eigenvalues,

S.-L. Wu, “Convergence analysis of the Parareal-Euler algorithm for sys- tems of ODEs with complex eigenvalues, ” J. Sci. Comput., Vol. 67, 2016, pp. 644–668

2016
[23]

A space-time multigrid method for parabolic partial differential equations,

G. Horton, S. Vandewalle, “A space-time multigrid method for parabolic partial differential equations,” SIAM J. Sci. Comput., Vol. 16, No. 4, 1995, pp. 848–864

1995
[24]

Parallel time integration with multigrid,

R. D. Falgout, S. Friedho, T.V. Kolev, S. P. MacLachlan, J. B. Schroder, “Parallel time integration with multigrid,” SIAM J. Sci. Comput. , Vol. 36, No. 6, 2014, pp. C635–C661

2014
[25]

ParaDiag: parallel-in- time algorithms based on the diagonalization technique,

M. J. Gander, J. Liu, S. -L. Wu, X. Yue, T. Zhou, “ParaDiag: parallel-in- time algorithms based on the diagonalization technique,” arXiv preprint, arXiv:2005.09158, 2020

2005 arXiv
[26]

Towards developing robust algorithms for solv- ing partial differential equations on MIMD machines,

J. H. Saltz and V. K. Naik, “Towards developing robust algorithms for solv- ing partial differential equations on MIMD machines, ” Parallel computing, Vol. 6, 1988, pp. 19-44. 23

1988
[27]

A time-stepping algorithm for parallel computers,

D. E. Womble, “A time-stepping algorithm for parallel computers,” SIAM J. Sci. Stat. Comput., Vol. 11, 1990, vpp. 824-837

1990
[28]

A rigorous analysis of time domain parallelism,

A. Deshpande, S. Malhotra, and C. Douglas, “A rigorous analysis of time domain parallelism,” Parallel Algorithms and Applications, Vol. 6, 1995, pp. 53-62

1995
[29]

A new parallel-in time direct inverse method for nonlinear differential equations

N. K. Yamaleev and S. Paudel, “A new parallel-in time direct inverse method for nonlinear differential equations” arXiv preprint , arXiv:2406.00878v1, 2024

2024 arXiv
[30]

Convergence of Parareal with spatial coarsening,

D. Ruprecht, “Convergence of Parareal with spatial coarsening,” Proc. in Appl. Math. Mech., Vol., 2014, No. 1, pp. 1031–1034

2014
[31]

Impact of spatial coarsening on Parareal convergence,

J. Angel, S. G¨ otschel, and D. Ruprecht, “Impact of spatial coarsening on Parareal convergence,” arXiv preprint, arXiv:2111.10228v1, 2021. 24

2021

Pith tools

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

[1] [1]

50 years of time parallel time integration,

M. J. Gander, “50 years of time parallel time integration,” in Multi- ple Shooting and Time Domain Decomposition Methods, T. Carraro, M. 21 Geiger, S. K¨ orkel, and R. Rannacher, eds., Springer, 2015, pp. 69–114

work page 2015

[2] [2]

Applications of time parallelization,

B. W. Ong and J. B. Schroder, “Applications of time parallelization,” Com- put. Visual. Sci.,Vol. 23, No. 11, 2020

work page 2020

[3] [3]

A parareal in time discretization of PDEs,

J.-L. Lions, Y. Maday, and G. Turinici, “A parareal in time discretization of PDEs,” C. R. Acad. Sci. Paris, Serie I, Vol. 332, 2001, pp. 661-668

work page 2001

[4] [4]

Stable parareal in time method for first- and second- order hyperbolic systems,

X. Dai and Y. Maday, “Stable parareal in time method for first- and second- order hyperbolic systems, ” SIAM J. Sci. Comput., Vol. 35 2013, pp. A52- A78

work page 2013

[5] [5]

A hybrid parareal spectral deferred corrections method,

M. L. Minion, “A hybrid parareal spectral deferred corrections method, ” Comm. App. Math. and Comp. Sci.,Vol. 5, 2010, pp. 265-301

work page 2010

[6] [6]

On the Use of Reduced Basis Methods to Accelerate and Stabilize the Parareal Method

F. Chen, J.S. Hesthaven, and X. Zhu, “On the Use of Reduced Basis Methods to Accelerate and Stabilize the Parareal Method”, Reduced Order Methods for Modeling and Computational Reduction, Springer Inter. Publ. 2014, pp. 187–214

work page 2014

[7] [7]

Stability of the Parareal Algorithm,

G. A. Staff and E. M. Rønquist, “Stability of the Parareal Algorithm,” Do- main Decomposition Methods in Science and Engineering, Springer Berlin Heidelberg, 2005, pp. 449–456

work page 2005

[8] [8]

Toward parallel coarse grid correction for parareal algorithm,

S.-L. Wu, “Toward parallel coarse grid correction for parareal algorithm,” SIAM J. Sci. Comput., Vol. 40, No. 3 2018, pp. A1446 - A1472

work page 2018