the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Apr 2022
Block-structured, equal workload, multigrid nesting interface for Boussinesq wave model FUNWAVE-TVD
Abstract. We describe the development of a block-structured, equal CPU-load, multigrid nesting interface for the Boussinesq wave model FUNWAVE-TVD. The new model framework does not interfere with the core solver, and thus the core program, FUNWAVE-TVD, is still a stand-alone model used for a single grid. The nesting interface manages the time sequencing and two-way nesting processes between the parent grid and child grid with grid refinement in a hierarchical manner. Workload balance in the MPI-based parallelization is handled by an equal-load scheme. A strategy of shared array allocation is applied for data management, that allows a large number of nested grids without creating additional memory allocations. Four model tests are conducted to verify the nesting algorithm, model accuracy, wetting-drying treatment, and the robustness in the application to modeling transoceanic tsunamis and coastal effects.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(3540 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3540 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
backboneof the nesting framework, handling data input, output, time sequencing, and internal interactions between grids at different scales.
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-35', Anonymous Referee #1, 10 Apr 2022
The manuscript describes an extension of a popular wave model with the aim to increase applicability and computational speed through multi-grid nesting and MPI parallelization. The paper is well-written and most sections are properly explained. However, there are still a few section, where more information in necessary. This applies mostly to memory sharing, synchonizaion and several other small details, which could really help a less experienced researcher in understanding the details of this presented technique. Below are a few comments, which can help to improve the overall quality and message of this paper.
Page 4 line 101-104 :
“is 4 or better” meaning refinement factor >=4 or 4<= ?
Page 7 line 176-183: Any particular reason why the spherical mode solve the weakly nonlinear equations and not the fully nonlinear equations?
Page 10 line 225-228: Is it necessary to exchange the dispersive terms: how does it increase the model efficiency.
How does the exchange work for the tridiagonal solver (child grid)?
Page 12 line 275-276: The restriction operator (The update operator) is not detailed. Which parent grid cells are actually updated? What about the ghost cells? What variables are updated in the Parent grid? Free surface/ velocity/ dispersion terms?
Workload balance and data management
Page 13 line 293-295: This statement is not clear.
What terms are pre-computed before the model run?
The child-parent proximity (if proximity means boundaries) changes at each Parent time step – Same for the restriction process.
Personally I think the implementation is not very detailed considering that it’s the main contribution from the paper.
I dont understand how the communication between the child and the parent grid is straightforward...
When the authors talk about shared array allocation, do all processors have a copy of all the model variables (Parents and child grids) + the boundary condition of child grid ? How does the synchronization work?
How is the MPI implementation optimized for nested grid?
How do the authors synchronize the parent and child grid after each parent time step? Do the authors use a fixed time step for the Child grid? If they use the CFL condition for the parent and the child solutions, does this involve that some type of synchronization is required before the update step.
Application
4.1 Evolution of an initial rectangular-shaped hump
Symmetry test is okay
Figure 7 : maybe include a diagonal transect and plot (a) (b) and (c) in the same figure. For better comparison.
Page18 line 349-352: The whole solution depends on grid resolution not only dispersive effects.
Citation: https://doi.org/10.5194/egusphere-2022-35-RC1 - AC1: 'Reply on RC1', Fengyan Shi Shi, 06 May 2022
-
RC2: 'Comment on egusphere-2022-35', Anonymous Referee #2, 06 May 2022
- AC2: 'Reply on RC2', Fengyan Shi Shi, 21 May 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-35', Anonymous Referee #1, 10 Apr 2022
The manuscript describes an extension of a popular wave model with the aim to increase applicability and computational speed through multi-grid nesting and MPI parallelization. The paper is well-written and most sections are properly explained. However, there are still a few section, where more information in necessary. This applies mostly to memory sharing, synchonizaion and several other small details, which could really help a less experienced researcher in understanding the details of this presented technique. Below are a few comments, which can help to improve the overall quality and message of this paper.
Page 4 line 101-104 :
“is 4 or better” meaning refinement factor >=4 or 4<= ?
Page 7 line 176-183: Any particular reason why the spherical mode solve the weakly nonlinear equations and not the fully nonlinear equations?
Page 10 line 225-228: Is it necessary to exchange the dispersive terms: how does it increase the model efficiency.
How does the exchange work for the tridiagonal solver (child grid)?
Page 12 line 275-276: The restriction operator (The update operator) is not detailed. Which parent grid cells are actually updated? What about the ghost cells? What variables are updated in the Parent grid? Free surface/ velocity/ dispersion terms?
Workload balance and data management
Page 13 line 293-295: This statement is not clear.
What terms are pre-computed before the model run?
The child-parent proximity (if proximity means boundaries) changes at each Parent time step – Same for the restriction process.
Personally I think the implementation is not very detailed considering that it’s the main contribution from the paper.
I dont understand how the communication between the child and the parent grid is straightforward...
When the authors talk about shared array allocation, do all processors have a copy of all the model variables (Parents and child grids) + the boundary condition of child grid ? How does the synchronization work?
How is the MPI implementation optimized for nested grid?
How do the authors synchronize the parent and child grid after each parent time step? Do the authors use a fixed time step for the Child grid? If they use the CFL condition for the parent and the child solutions, does this involve that some type of synchronization is required before the update step.
Application
4.1 Evolution of an initial rectangular-shaped hump
Symmetry test is okay
Figure 7 : maybe include a diagonal transect and plot (a) (b) and (c) in the same figure. For better comparison.
Page18 line 349-352: The whole solution depends on grid resolution not only dispersive effects.
Citation: https://doi.org/10.5194/egusphere-2022-35-RC1 - AC1: 'Reply on RC1', Fengyan Shi Shi, 06 May 2022
-
RC2: 'Comment on egusphere-2022-35', Anonymous Referee #2, 06 May 2022
- AC2: 'Reply on RC2', Fengyan Shi Shi, 21 May 2022
Peer review completion
Journal article(s) based on this preprint
backboneof the nesting framework, handling data input, output, time sequencing, and internal interactions between grids at different scales.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 382 | 101 | 7 | 490 | 3 | 4 |
- HTML: 382
- PDF: 101
- XML: 7
- Total: 490
- BibTeX: 3
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Young-Kwang Choi
Fengyan Shi
Matt Malej
Jane M. Smith
James T. Kirby
Stephan Grilli
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(3540 KB) - Metadata XML