Performance evaluation of halide auto-scheduler with directional cubic convolution interpolation

Nogami, Haruki; Oishi, Sou; Sasaki, Tsukasa; Maeda, Yoshihiro; Fukushima, Norishige

doi:10.1117/12.2666979

Cited by 3 publications

(3 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…HIPAcc [65] 4 is similar to Halide, and Forma [66] 5 has autoscheduling functionality. Currently, Halide also has mature autoscheduling [34]- [37], [67], [68], which has been verified in various applications [69], [70]. Poly-Mage [38] 6 uses the polyhedral model [71] for loop parallelization, which is more flexible for complex loop structures.…”

Section: Dsl For Image Processingmentioning

confidence: 99%

SlidingConv: Domain-Specific Description of Sliding Discrete Cosine Transform Convolution for Halide

Kanetaka,

Takagi,

Maeda

et al. 2024

IEEE Access

Self Cite

View full text Add to dashboard Cite

Section: Dsl For Image Processingmentioning

confidence: 99%

SlidingConv: Domain-Specific Description of Sliding Discrete Cosine Transform Convolution for Halide

Kanetaka,

Takagi,

Maeda

et al. 2024

IEEE Access

Self Cite

View full text Add to dashboard Cite

“…Halide [72] is a DSL for image processing, and the language can easily vectorize codes with a simple description. There are various effective implementations in Halide, such as interpolation [73], FIR [74], recursive [75], median [76], and variable-weighted [77] convolutions.…”

Section: B Optimization For Convolutionmentioning

confidence: 99%

SIMD-Constrained Lookup Table for Accelerating Variable-Weighted Convolution on x86/64 CPUs

Naganawa,

Kamei,

Kanetaka

et al. 2024

IEEE Access

View full text Add to dashboard Cite

Convolution is the inner product of the neighborhood signal and weights and plays a fundamental role in image processing; thus, acceleration of convolution is essential. Among convolutions, variable-weighted convolution is used in adaptive filters and edge-preserving smoothing to realize various applications. Some weights are replaced with lookup tables (LUTs) to accelerate these filters. LUT reference is a classical acceleration method. However, the difference between the growth rate in computing speed and memory I/O speed has limited the scope of utilization of LUT references. Speedup would be possible if registers could be used as LUTs, but their small size makes them difficult to utilize. Therefore, this study proposes a downsampling method to fit LUTs into SIMD registers, which are relatively large and an efficient reference method for register-LUTs. Experimental results show that the proposed method can reproduce an accuracy in PSNR of 65.52 (+25.11) dB, while a simple full-size LUT in the register size can only reproduce 40.41 dB. Using a wider register width, the PSNR was 78.63 (+38.22) dB with AVX-512 and 84.5 (+44.09) dB with bfloat16. The fastest proposed method was on average 4.82/3.72 times faster than direct vector computing, 2.99/3.10 times faster than vector addressing, and 3.79/7.80 times faster than scalar addressing on the AVX2/AVX-512 computers while exceeding the display limit of 60 dB for 8-bit displays. Taking into account these speed/accuracy trade-offs, the performance of the proposed method was superior. This paper shows that LUT references can be realized with small SIMD registers in convolution. The proposed method is expected to be extended to adaptive filters, convolutional neural networks, and other image processing applications by accelerating the approximation with this register-LUT. Our code is available at https://fukushimalab.github.io/registerLUT4conv/.

show abstract

“…Regarding the operation of SatNOGS, when a user wants to communicate with a satellite, first he needs to have a GS online and functioning, otherwise he will not be able to schedule observations across the network. Then, he can start scheduling, and he has two options: He can do this by accessing the SatNOGS network through a web interface [14] or using the SatNOGS auto-scheduler [16], which allows one to perform observations automatically. However, this will be determined considering that the observations are carried out on the web since it is easier for people who are not familiar with the command line.…”

Section: Satnogs Introductionmentioning

confidence: 99%

TinyGS vs. SatNOGS: A Comparative Analysis of Open-Source Satellite Ground Station Networks

Sá Gomes,

Ferreira da Silva

2024

Telecom

View full text Add to dashboard Cite

In recent years, two of the largest open-source ground station (GS) networks capable of enabling Earth–satellite communication have emerged: TinyGS and SatNOGS. These open-source projects enable anyone to build their own GS inexpensively and easily, integrate into a GS network, and receive data from satellites listed in the database. Additionally, it enables satellite developers to add satellites to the databases of these projects and take advantage of this GS network to receive data from the satellites. This article introduces the TinyGS and SatNOGS projects and conducts a comparative analysis between them. Generally, the TinyGS project seems to have simpler implementation as well as lower associated costs. In a deeper analysis, it was observed that on the 29 July 2023, the TinyGS project had a higher number of online GSs and a more favorable geographic distribution. On the other hand, the SatNOGS project managed to communicate and decode a larger number of satellites up to 29 July 2023. Additionally, in both projects, it was noted that frequencies between 436 and 437 had the highest number of satellites with decoded data. Ultimately, the choice between these projects depends on critical parameters defined by the reader.

show abstract

Performance evaluation of halide auto-scheduler with directional cubic convolution interpolation

Cited by 3 publications

References 13 publications

SlidingConv: Domain-Specific Description of Sliding Discrete Cosine Transform Convolution for Halide

SlidingConv: Domain-Specific Description of Sliding Discrete Cosine Transform Convolution for Halide

SIMD-Constrained Lookup Table for Accelerating Variable-Weighted Convolution on x86/64 CPUs

TinyGS vs. SatNOGS: A Comparative Analysis of Open-Source Satellite Ground Station Networks

Contact Info

Product

Resources

About