Esoteric Pull and Esoteric Push: Two Simple In-Place Streaming Schemes for the Lattice Boltzmann Method on GPUs
Round 1
Reviewer 1 Report
- Line 84 and in general: Rather than 'I' everywhere in the paper, you can state in terms of 'paper proposes' or 'paper, etc.'
- Use some other text format/size for the appendix?
- Any mention of computation cost, and resource cost for proposed methods?
- Any experiments, or results for the new algorithms?
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The authors present two new in-place Lattice Boltzmann Method algorithms, Esoteric Pull and Esoteric Push, based on a modification of the Esoteric-Twist algorithm, and demonstrate their use in a free surface LBM solver.
I suggest publication of the manuscript after the following minor revisions:
- In section 4.2, the authors compare their proposed streaming scheme to the existing one, focusing on its ease of implementation and compatibility with all velocity sets. Rather than advising the reader to compare the pseudo-code implementations, I recommend improving the reasons for these two advantages.
- According to the title of the paper, the streaming algorithm is only (or mostly) applicable to free surface LBM. I recommend revising the introduction and/or title to better describe the purpose of the manuscript and why FSLM was chosen to demonstrate the new streaming algorithm.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The author presents new streaming kernels for LB implementations on GPUs .
Although interesting, No performance numbers are reported (how much bandwidth is used,how does method compare to other schemes). This is essential,as a new pattern will only be used if it is advantageous. ..
I can therefore not recommend the paper for publication in its present form
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
The paper devoted to investigate an optimized method to use the GPU in the LBM simulation. The author developed Esoteric-Twist’s method to a new method to save the memory. The result is interesting and can be published after a revision.
1- The introduction is too short the author can first explain using of LBM in general structure on CPU and explain different application of it. The following references are useful
https://0-doi-org.brum.beds.ac.uk/10.1080/10407782.2014.949187
https://0-doi-org.brum.beds.ac.uk/10.1016/j.camwa.2015.07.008
10.1109/BMEI.2011.6098676
At the next step the author should explain a wide range of the previous methods used the GPU or both CPU-GPU and compare their advantages and disadvantages. It is recommended to add a table for comparison.
The following reference are useful.
https://0-doi-org.brum.beds.ac.uk/10.1016/j.camwa.2009.08.052
https://0-doi-org.brum.beds.ac.uk/10.1155/2017/1205892
https://0-doi-org.brum.beds.ac.uk/10.48550/arXiv.1904.02108
2- The paper type can be changed to a short paper rather than a full research paper.
3- What are the advantage of the current method to the previous published methods.
4- Since the coding is important in this case and the only point of this paper is to use a rather new developed method to enhance the simulation time, it is recommended to add the whole source code file of the figure 7’s simulation as supplementary materials.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 5 Report
The lattice Boltzmann method (LBM) is a popular method for the simulations of fluid flow. The paper of Lehmann proposes a novel method to reduce the memory demand of the parallelized LBM code on GPU. I think that the work is useful and important. The paper is also well documented. I recommend the publication of the paper in computation in its present form.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 6 Report
The report presents the outcome of a study on esoteric pull and esoteric push. The contribution of the report to the body of knowledge is significant and novel. Also, the aim and objectives of the study are within the scope of the Journal of Computation, MDPI. However, the present form of the report needs revision. Authors should bear in mind that peer review is the quality-control gateway that allows us to trust and build on published research. The author should consider the following points:
Q1. The title was written as, "Esoteric Pull and Esoteric Push – two simple in-place streaming
schemes for the (Free Surface) Lattice Boltzmann Method on
GPUs"
Comment: This title is not objective. Revise. It is very obvious that there is a gap between the title and the discussion of results. Also, wrong usage of double bracket for, "Free Surface" is worth noticing. Revise the title.
Q2. The first sentence of the abstract reads, "The lattice Boltzmann method (LBM) can be parallelized on graphics processing ..."
Comment: Start the abstract with the significance of the study.
Q3. The last sentence of the abstract reads, "Their simplicity greatly facilitates modifying existing codes to in-place streaming, even with extensions already in place, such as demonstrated for the Free Surface LBM implementation FluidX3D."
Comment: Major conclusion of the study are expected at the end of the abstract.
Q4. Line 14 - 32, section 1 - introduction, the exact point the author is trying to introduce is not clear after reading the two paragraphs.
Comment: Start the abstract with the definition of term which the two paragraphs introduce.
Q5. Line 33 - 42, the first paragraph is made of just two sentences while the second paragraph is made of a small number of sentence.
Comment: Merge To form a paragraph made up of more than five (5) sentences.
Q6. The caption of Figure 1 is faulty. It was written, "Figure 1. One-Step-Pull streaming scheme. DDFs are pulled in from neighbors (copy A of the DDFs), collided, and stored at the center node (copy B of the DDFs). After every time step, the pointers to A and B are swapped."
Comment: Are you trying to discuss in caption? The caption should be concise. Cite the figure in-text.
Q7. Line 43, it was written:
3. State-of-the-art methods for in-place streaming on GPUs 43
3.1. AA-Pattern
Comment: It is faulty that a subsection should directly follow a section. Immediately after section 3, insert a leading sentence to announce subsection 3.1. AA-Pattern and 3.2. Esoteric-Twist. Thereafter, then, present 3.1. AA-Pattern.
Q8. Line 44 - 60, merge all the sentences to form a paragraph made up of more than four (4) sentences.
Q9. Section 3.1 is not loaded enough to announce AA-Pattern.
Q10. The caption of Table 1 in page 5 of 15 is too long.
Comment: A concise caption is required.
Q11. The caption of Figure 7 in page 7 of 15 is too long.
Comment: A concise caption is required. Improve the in-text citation of the figure.
Q12. Start the conclusion section with the the aim of the study in a broad form. Remarked that you have established all the research objectives. Then, outline major conclusion. Ensure that conclusive statements are used. In a short Conclusion, state the most important outcome of the work based on the interpreted findings. Do NOT just summarize.
Comment: Report your success in addressing the research questions.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
The authors have thoroughly revised the paper, quality has severely improved.
I only have few minor comments, that should be considered prior to publication:
- P. 2: "Only few works consider in-place streaming on GPU so far [2-12]."
-> If the author can immediately cite 11 works on this, I would not consider this to be "few works" for a narrow field such as LBM. Please rephrase!
- P. 2, Fig. 1 and 2: remove "on parallel hardware" since two copies are even required in the sequential case!
- P. 7, Fig. 7: this is a very interesting figure! I suggest to improve its quality by resizing it to a full page size, so that all labels are readable. Besides, it should be extended by an experiment description, i.e. what scenario was computed (e.g. cavity) and how big was the respective domain size chosen per device.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
