Skip to content

Commit 2c103dc

Browse files
committed
deploy: 0914469
1 parent 9e36f0f commit 2c103dc

File tree

2 files changed

+28
-123
lines changed

2 files changed

+28
-123
lines changed

methods/paradiag.html

Lines changed: 13 additions & 13 deletions
Original file line numberDiff line numberDiff line change
@@ -304,28 +304,28 @@ <h2>ParaDiag</h2>
304304
<section class="modal-card-body">
305305
In 2008, Maday and Rønquist introduced an interesting new approach for the direct parallel-in-time (PinT) solution of
306306
time-dependent PDEs. The idea is to diagonalize the time stepping matrix, keeping the matrices for the space discretization
307-
unchanged, and then to solve all time steps in parallel. Since then, several variants appeared, and we
308-
call these closely related algorithms \em ParaDiag algorithms. ParaDiag algorithms in the literature can
307+
unchanged, and then to solve all time steps in parallel. Since then, several variants appeared, and we
308+
call these closely related algorithms \em ParaDiag algorithms. ParaDiag algorithms in the literature can
309309
be classified into two groups:
310310
\beginitemize
311311
\item ParaDiag-I: direct standalone solvers,
312312
\item ParaDiag-II: iterative solvers.
313313
\enditemize
314-
We will explain the basic features of each group in this note. To have concrete examples, we will
315-
introduce ParaDiag-I and ParaDiag-II for the advection-diffusion equation. We will also introduce
316-
ParaDiag-II for the wave equation and an optimal control problem for the wave equation.
314+
We will explain the basic features of each group in this note. To have concrete examples, we will
315+
introduce ParaDiag-I and ParaDiag-II for the advection-diffusion equation. We will also introduce
316+
ParaDiag-II for the wave equation and an optimal control problem for the wave equation.
317317
We could have used the advection-diffusion equation as well to illustrate ParaDiag-II,
318-
but wave equations are known to cause problems for certain PinT algorithms and thus constitute an
318+
but wave equations are known to cause problems for certain PinT algorithms and thus constitute an
319319
especially interesting example for which ParaDiag algorithms were tested. We show the main known
320-
theoretical results in each case, and also provide Matlab codes for testing. The goal of the Matlab
321-
codes is to help the interested reader understand the key features of the ParaDiag algorithms,
320+
theoretical results in each case, and also provide Matlab codes for testing. The goal of the Matlab
321+
codes is to help the interested reader understand the key features of the ParaDiag algorithms,
322322
without intention to be highly tuned for efficiency and/or low memory use.
323323

324-
We also provide speedup measurements of ParaDiag algorithms for a 2D linear advection-diffusion equation.
325-
These results are obtained on the Tianhe-1 supercomputer in China and the SIUE Campus Cluster in the US,
326-
which is a multi-array, configurable and cooperative parallel system, and we compare these results to the
327-
performance of parareal and MGRiT, two widely used PinT algorithms. In a forthcoming update of this note,
328-
we will provide more material on ParaDiag algorithms, in particular further Matlab codes and parallel
324+
We also provide speedup measurements of ParaDiag algorithms for a 2D linear advection-diffusion equation.
325+
These results are obtained on the Tianhe-1 supercomputer in China and the SIUE Campus Cluster in the US,
326+
which is a multi-array, configurable and cooperative parallel system, and we compare these results to the
327+
performance of parareal and MGRiT, two widely used PinT algorithms. In a forthcoming update of this note,
328+
we will provide more material on ParaDiag algorithms, in particular further Matlab codes and parallel
329329
computing results, also for more realistic applications.
330330
</section>
331331
</div>

references/index.html

Lines changed: 15 additions & 110 deletions
Original file line numberDiff line numberDiff line change
@@ -5120,101 +5120,6 @@ <h2 id="year2025" class="year">2025 <span class="count-stat"></span></h2>
51205120
<h2 id="year2024" class="year">2024 <span class="count-stat"></span></h2>
51215121

51225122
<ol class="bibliography"><li><div class="bibtex-entry-container">
5123-
<div>
5124-
<div class="col-md-3 col-sm-12 bibtex-ref-meta hidden">
5125-
<div class="row">
5126-
<div class="col-md-12 ref-label label label-default">
5127-
AppelEtAl2024
5128-
</div>
5129-
</div>
5130-
</div>
5131-
5132-
<div class="bibtex-ref-entry">
5133-
<span id="AppelEtAl2024">M. Appel and J. Alexandersen, “One-shot Parareal Approach for Topology Optimisation of Transient Heat Flow,” arXiv:2411.19030v1 [cs.CE], 2024 [Online]. Available at: <a href="http://arxiv.org/abs/2411.19030v1" target="_blank">http://arxiv.org/abs/2411.19030v1</a></span>
5134-
</div>
5135-
5136-
5137-
<div class="row mt-4">
5138-
<div class="btn-group btn-group-xs btn-group-justified" role="group">
5139-
<div class="btn-group btn-group-xs" role="group">
5140-
5141-
<button class="button" data-toggle="modal" data-target="#modalAppelEtAl2024Bibtex" aria-expanded="false"
5142-
aria-controls="modalAppelEtAl2024Bibtex">
5143-
BibTeX
5144-
</button>
5145-
5146-
</div>
5147-
<div class="btn-group btn-group-xs" role="group">
5148-
5149-
<button class="button" data-toggle="modal" data-target="#modalAppelEtAl2024Abstract" aria-expanded="false"
5150-
aria-controls="modalAppelEtAl2024Abstract">
5151-
Abstract
5152-
</button>
5153-
5154-
</div>
5155-
</div>
5156-
</div>
5157-
</div>
5158-
5159-
<div class="modal" id="modalAppelEtAl2024Bibtex" tabindex="-1" role="dialog"
5160-
aria-labelledby="modalAppelEtAl2024BibtexTitle" aria-hidden="true">
5161-
<div class="modal-background">
5162-
<div class="modal-card">
5163-
<div class="modal-content">
5164-
<header class="modal-card-head">
5165-
<div class="modal-card-head-content">
5166-
<div>
5167-
<p class="modal-card-title" id="modalAppelEtAl2024BibtexTitle">
5168-
BibTeX entry <code>AppelEtAl2024</code>
5169-
</p>
5170-
</div>
5171-
<button type="button" class="delete" data-dismiss="modal" aria-label="Close"><span
5172-
aria-hidden="true">&times;</span></button>
5173-
</div>
5174-
</header>
5175-
<section class="modal-card-body">
5176-
<pre class="abstract">@unpublished{AppelEtAl2024,
5177-
author = {Appel, Magnus and Alexandersen, Joe},
5178-
howpublished = {arXiv:2411.19030v1 [cs.CE]},
5179-
title = {One-shot Parareal Approach for Topology Optimisation of Transient Heat Flow},
5180-
url = {http://arxiv.org/abs/2411.19030v1},
5181-
year = {2024}
5182-
}
5183-
</pre>
5184-
</section>
5185-
</div>
5186-
</div>
5187-
</div>
5188-
</div>
5189-
5190-
5191-
<div class="modal" id="modalAppelEtAl2024Abstract" tabindex="-1" role="dialog"
5192-
aria-labelledby="modalAppelEtAl2024AbstractTitle" aria-hidden="true">
5193-
<div class="modal-background">
5194-
<div class="modal-card">
5195-
<div class="modal-content">
5196-
<header class="modal-card-head">
5197-
<div class="modal-card-head-content">
5198-
<div>
5199-
<p class="modal-card-title" id="modalAppelEtAl2024AbstractTitle">
5200-
Abstract for BibTeX entry <code>AppelEtAl2024</code>
5201-
</p>
5202-
</div>
5203-
<button type="button" class="delete" data-dismiss="modal" aria-label="Close"><span
5204-
aria-hidden="true">&times;</span></button>
5205-
</div>
5206-
</header>
5207-
<section class="modal-card-body">
5208-
This paper presents a method of performing topology optimisation of transient heat conduction problems using the parallel-in-time method Parareal. To accommodate the adjoint analysis, the Parareal method was modified to store intermediate time steps. Preliminary tests revealed that Parareal requires many iterations to achieve accurate results and, thus, achieves no appreciable speedup. To mitigate this, a one-shot approach was used, where the time history is iteratively refined over the optimisation process. The method estimates objectives and sensitivities by introducing cumulative objectives and sensitivities and solving for these using a single iteration of Parareal, after which it updates the design using the Method of Moving Asymptotes. The resulting method was applied to a test problem where a power mean of the temperature was minimised. It achieved a peak speedup relative to a sequential reference method of 5\times using 16 threads. The resulting designs were similar to the one found by the reference method, both in terms of objective values and qualitative appearance. The one-shot Parareal method was compared to the Parallel Local-in-Time method of topology optimisation. This revealed that the Parallel Local-in-Time method was unstable for the considered test problem, but it achieved a peak speedup of 12\times using 32 threads. It was determined that the dominant bottleneck in the one-shot Parareal method was the time spent on computing coarse propagators.
5209-
</section>
5210-
</div>
5211-
</div>
5212-
</div>
5213-
</div>
5214-
5215-
</div>
5216-
</li>
5217-
<li><div class="bibtex-entry-container">
52185123
<div>
52195124
<div class="col-md-3 col-sm-12 bibtex-ref-meta hidden">
52205125
<div class="row">
@@ -20217,28 +20122,28 @@ <h2 id="year2021" class="year">2021 <span class="count-stat"></span></h2>
2021720122
<section class="modal-card-body">
2021820123
In 2008, Maday and Rønquist introduced an interesting new approach for the direct parallel-in-time (PinT) solution of
2021920124
time-dependent PDEs. The idea is to diagonalize the time stepping matrix, keeping the matrices for the space discretization
20220-
unchanged, and then to solve all time steps in parallel. Since then, several variants appeared, and we
20221-
call these closely related algorithms \em ParaDiag algorithms. ParaDiag algorithms in the literature can
20125+
unchanged, and then to solve all time steps in parallel. Since then, several variants appeared, and we
20126+
call these closely related algorithms \em ParaDiag algorithms. ParaDiag algorithms in the literature can
2022220127
be classified into two groups:
2022320128
\beginitemize
2022420129
\item ParaDiag-I: direct standalone solvers,
2022520130
\item ParaDiag-II: iterative solvers.
2022620131
\enditemize
20227-
We will explain the basic features of each group in this note. To have concrete examples, we will
20228-
introduce ParaDiag-I and ParaDiag-II for the advection-diffusion equation. We will also introduce
20229-
ParaDiag-II for the wave equation and an optimal control problem for the wave equation.
20132+
We will explain the basic features of each group in this note. To have concrete examples, we will
20133+
introduce ParaDiag-I and ParaDiag-II for the advection-diffusion equation. We will also introduce
20134+
ParaDiag-II for the wave equation and an optimal control problem for the wave equation.
2023020135
We could have used the advection-diffusion equation as well to illustrate ParaDiag-II,
20231-
but wave equations are known to cause problems for certain PinT algorithms and thus constitute an
20136+
but wave equations are known to cause problems for certain PinT algorithms and thus constitute an
2023220137
especially interesting example for which ParaDiag algorithms were tested. We show the main known
20233-
theoretical results in each case, and also provide Matlab codes for testing. The goal of the Matlab
20234-
codes is to help the interested reader understand the key features of the ParaDiag algorithms,
20138+
theoretical results in each case, and also provide Matlab codes for testing. The goal of the Matlab
20139+
codes is to help the interested reader understand the key features of the ParaDiag algorithms,
2023520140
without intention to be highly tuned for efficiency and/or low memory use.
2023620141

20237-
We also provide speedup measurements of ParaDiag algorithms for a 2D linear advection-diffusion equation.
20238-
These results are obtained on the Tianhe-1 supercomputer in China and the SIUE Campus Cluster in the US,
20239-
which is a multi-array, configurable and cooperative parallel system, and we compare these results to the
20240-
performance of parareal and MGRiT, two widely used PinT algorithms. In a forthcoming update of this note,
20241-
we will provide more material on ParaDiag algorithms, in particular further Matlab codes and parallel
20142+
We also provide speedup measurements of ParaDiag algorithms for a 2D linear advection-diffusion equation.
20143+
These results are obtained on the Tianhe-1 supercomputer in China and the SIUE Campus Cluster in the US,
20144+
which is a multi-array, configurable and cooperative parallel system, and we compare these results to the
20145+
performance of parareal and MGRiT, two widely used PinT algorithms. In a forthcoming update of this note,
20146+
we will provide more material on ParaDiag algorithms, in particular further Matlab codes and parallel
2024220147
computing results, also for more realistic applications.
2024320148
</section>
2024420149
</div>
@@ -22400,8 +22305,8 @@ <h2 id="year2021" class="year">2021 <span class="count-stat"></span></h2>
2240022305
</div>
2240122306
</header>
2240222307
<section class="modal-card-body">
22403-
Solving evolutionary equations in a parallel-in-time manner is an attractive topic and many algorithms are proposed in recent two decades. The algorithm based on the block α-circulant preconditioning technique has shown promising advantages, especially for wave propagation problems. By fast Fourier transform for factorizing the involved circulant matrices, the preconditioned iteration can be computed efficiently via the so-called diagonalization technique, which yields a direct parallel implementation across all time levels. In recent years, considerable efforts have been devoted to exploring the convergence of the preconditioned iteration
22404-
by studying the spectral radius of the iteration matrix, and this leads to many case-by-case studies depending on the used time-integrator. In this paper, we propose a unified convergence analysis for the algorithm applied to u’+Au=f, where σ(A)⊂\mathbbC^+ with σ(A) being the spectrum of A∈\mathbbC^m\times m. For any one-step method (such as the Runge-Kutta methods) with stability function \mathcalR(z), we prove that the decay rate of the global error is bounded by α/(1-α), provided the method is stable, i.e., \max_λ∈σ(A)|\mathcalR(∆tλ)|\leq1. For any linear multistep method, such a bound becomes cα/(1-cα), where c\geq1 is a constant specified by the multistep method itself.
22308+
Solving evolutionary equations in a parallel-in-time manner is an attractive topic and many algorithms are proposed in recent two decades. The algorithm based on the block α-circulant preconditioning technique has shown promising advantages, especially for wave propagation problems. By fast Fourier transform for factorizing the involved circulant matrices, the preconditioned iteration can be computed efficiently via the so-called diagonalization technique, which yields a direct parallel implementation across all time levels. In recent years, considerable efforts have been devoted to exploring the convergence of the preconditioned iteration
22309+
by studying the spectral radius of the iteration matrix, and this leads to many case-by-case studies depending on the used time-integrator. In this paper, we propose a unified convergence analysis for the algorithm applied to u’+Au=f, where σ(A)⊂\mathbbC^+ with σ(A) being the spectrum of A∈\mathbbC^m\times m. For any one-step method (such as the Runge-Kutta methods) with stability function \mathcalR(z), we prove that the decay rate of the global error is bounded by α/(1-α), provided the method is stable, i.e., \max_λ∈σ(A)|\mathcalR(∆tλ)|\leq1. For any linear multistep method, such a bound becomes cα/(1-cα), where c\geq1 is a constant specified by the multistep method itself.
2240522310
Our proof only relies on the stability of the time-integrator and the estimate is independent of the step size ∆t and the spectrum σ(A).
2240622311
</section>
2240722312
</div>

0 commit comments

Comments
 (0)