GPU@SAT DevKit: Empowering Edge Computing Development Onboard Satellites in the Space-IoT Era

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The work is interesting since it presents a novel DevKit, which represents an advanced tool in the field of edge computing for satellites and offers a robust and user-friendly ecosystem for accelerating the development of space-based IoT applications. The benifits of the proposed kit are explained through a benchmark. 

The paper is well structured. It is believed that the proposed kit is promising in the satellite IoT. Introduction provides sufficient research background and make clear the contributions of the paper. The proposed kit is described adequately. 

My only concern is the evalution of the proposed kit. From the table, the advantages stated by the authors, including "robust", "comprehensive performance metrics" etc., cannot be recognized. The only index we can see is the execcution time. It is expected that more detailed comparison be provided.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

• The authors' idea to improve the onboard capabilities of new satellites is interesting and relevant; however, they should better motivate it in comparison to existing applications, solutions, or previously presented works.
• For example, https://doi.org/10.3390/aerospace10020101 proposes accelerators onboard.
• Several papers have already addressed the topic of enhancing satellites' onboard computing capabilities, and these should be included as key references in the background and analysis of this work.
• The abstract should be rewritten, as the advantages and achievements of this work are not clearly understood.
• Given that neuromorphic processors are now being proposed, offering even more significant advantages over GPUs, what is the continued interest or contribution of GPU processing?
• The results and future directions should be expanded further.

Author Response

See the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

This paper proposes a system that integrates the GPU@SAT IP core with the SoC-embedded processor, suitable for edge computing development onboard satellites in the space-IoT era. The work is interesting, but there are some points to be considered.

1. The author introduces why FPGAs are suitable for space applications. But as far as I know, FPGAs are still very sensitive to SEU, SET, and SEL, even if they are radiation-hardened. The author needs to provide a more detailed reason for selecting FPGAs for the acceleration of ML or DL algorithms.

2. The author only provides data about the basic operations, such as convolution operations, making it difficult to evaluate the work's performance and value. More data, such as the FPGA resource usage and comparison with similar works, should be supplied. 

3. If the author provides a real application case, it will better demonstrate the value and feasibility of this work.

Author Response

see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

Thank you to the authors for addressing my comments and clarifying my doubts. The contribution has improved and is now more precise. For future revisions, I recommend that the authors include the changes directly in the manuscript to make them easier to identify.

Reviewer 3 Report

Comments and Suggestions for Authors

The authors have addressed my questions.