CubeDMA – Optimizing three-dimensional DMA transfers for hyperspectral imaging applications

Fjeldtvedt, Johan Austlid; Orlandić, Milica

doi:10.1016/j.micpro.2018.12.009

Cited by 10 publications

(6 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…DMAEs can be grouped according to their system binding: register, transfer-descriptor, and instruction-based. Engines requiring a high degree of agility [11], [30] or featuring a small footprint [33], [38], [39] tend to use a registerbased interface. PEs write the transfer information in a dedicated register space and use a read or write operation to a special register location to launch the transfer.…”

Section: Related Workmentioning

confidence: 99%

“…SoC DMAEs, e.g., Fjeldtvedt et al [30], Rossi et al [11], and Morales et al [39] feature a fixed configuration of onchip protocol(s). Some controllers allow selectively adding a simple interface to connect to peripherals.…”

Section: Related Workmentioning

confidence: 99%

“…Fjeldtvedt et al's CubeDMA can even handle three-dimensional transfers. Any higher dimensional transfer is handled in software either by repetitively launching simpler transfers [30] [11] or by employing transfer descriptor chaining [1], [26]. Our tensor ND mid-end can execute arbitrary high dimensional transfers in hardware.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

A High-Performance, Energy-Efficient Modular DMA Engine Architecture

Benz,

Rogenmoser,

Scheffler

et al. 2024

IEEE Trans. Comput.

View full text Add to dashboard Cite

Data transfers are essential in today's computing systems as latency and complex memory access patterns are increasingly challenging to manage. Direct memory access engines (DMAEs) are critically needed to transfer data independently of the processing elements, hiding latency and achieving high throughput even for complex access patterns to high-latency memory. With the prevalence of heterogeneous systems, DMAEs must operate efficiently in increasingly diverse environments. This work proposes a modular and highly configurable open-source DMAE architecture called intelligent DMA (iDMA), split into three parts that can be composed and customized independently. The front-end implements the control plane binding to the surrounding system. The mid-end accelerates complex data transfer patterns such as multi-dimensional transfers, scattering, or gathering. The back-end interfaces with the on-chip communication fabric (data plane). We assess the efficiency of iDMA in various instantiations: In high-performance systems, we achieve speedups of up to 15.8× with only 1 % additional area compared to a base system without a DMAE. We achieve an area reduction of 10 % while improving ML inference performance by 23 % in ultra-low-energy edge AI systems over an existing DMAE solution. We provide area, timing, latency, and performance characterization to guide its instantiation in various systems.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A High-Performance, Energy-Efficient Modular DMA Engine Architecture

Benz,

Rogenmoser,

Scheffler

et al. 2024

IEEE Trans. Comput.

View full text Add to dashboard Cite

show abstract

“…After image capture and binning, with a full binned cube in main memory, a DMA Intellectual Property Core (IP Core) has been developed, cube DMA [70], specialized to efficiently stream hyperspectral data to the FPGA processing modules in the sample order required by the FPGA module irrespective of whether the HSI data are stored in BIP, BIL, BSQ in the main memory.…”

Section: Minimal On-board Image Processing Pipelinementioning

confidence: 99%

Robust and Reconfigurable On-Board Processing for a Hyperspectral Imaging Small Satellite

et al. 2023

Self Cite

View full text Add to dashboard Cite

Hyperspectral imaging is a powerful remote sensing technology, but its use in space is limited by the large volume of data it produces, which leads to a downlink bottleneck. Therefore, most payloads to date have been oriented towards demonstrating the scientific usefulness of hyperspectral data sporadically over diverse areas rather than detailed monitoring of spatio-spectral dynamics. The key to overcoming the data bandwidth limitation is to process the data on-board the satellite prior to downlink. In this article, the design, implementation, and in-flight demonstration of the on-board processing pipeline of the HYPSO-1 cube-satellite are presented. The pipeline provides not only flexible image processing but also reliability and resilience, characterized by robust booting and updating procedures. The processing time and compression rate of the simplest pipeline, which includes capturing, binning, and compressing the image, are analyzed in detail. Based on these analyses, the implications of the pipeline performance on HYPSO-1’s mission are discussed.

show abstract

“…DMA can characterize the viscoelastic properties of materials and could be used to determine storage modulus (E 0 ), loss modulus (E 00 ) and loss factor (tand) of carbon fabric/PP composites. [15][16][17] A single cantilever bending pattern of DMA was carried out by using DMA850/Q850 equipment with a heating rate of 3 C/min from -70 to 70 C. The amplitude and frequency of the tests were 20.0 lm and 1.0 Hz, respectively.…”

Section: Dynamic Mechanical Thermal Analysis (Dma)mentioning

confidence: 99%

A new method of grafting multi-walled carbon nanotubes on carbon fibers for improving the mechanical and thermal properties of woven fabric composites

Liu

Rao

Wang

et al. 2021

Journal of Composite Materials

View full text Add to dashboard Cite

A new method of grafting multi-walled carbon nanotubes (MWCNTs) onto carbon fiber surface to improve the thermo-mechanical properties of woven carbon fabric reinforced composites was proposed. In this method, both carbon woven fabrics and MWCNTs were oxidized by sulfuric acid to generate carboxyl groups on their surfaces, respectively. Then silane coupling agent was used to react with the carboxyl groups to graft MWCNTs onto the carbon fiber surfaces of the woven fabric. The untreated, acid treated and MWCNTs grafted carbon woven fabrics were separately combined with polypropylene films to form composite plates by thermal-stamping. Fourier transform infrared spectroscopy and X-ray photoelectron spectroscopy were conducted to estimate the changes of element contents and functional groups on surfaces of carbon fibers and MWCNTs. Atomic force microscope was used to estimate the roughness of carbon fiber surfaces. Scanning electron microscopy, differential scanning calorimeter, dynamic mechanical thermal analysis and tensile tests were carried out to analyze the surface morphology, thermal, and mechanical properties of carbon fabrics and their composites. Testing results showed that MWCNTs could be successfully grafted onto the carbon fibers by using silane as an intermediate bridge. Compared with the untreated and acid treated composites, the in-plane shearing stiffness and fracture strength of the composites were increased significantly by MWCNTs grafting. In terms of thermal properties, acid treatment and MWCNTs grafting have little effect on melting point of composites. MWCNTs can promote the recrystallization process of the PP and reduce the numbers of imperfect crystals. As for thermo-mechanical properties, acid treatment deteriorated the bending storage modulus of the composite, while MWCNTs grafting could compensate it.

show abstract

CubeDMA – Optimizing three-dimensional DMA transfers for hyperspectral imaging applications

Cited by 10 publications

References 9 publications

A High-Performance, Energy-Efficient Modular DMA Engine Architecture

A High-Performance, Energy-Efficient Modular DMA Engine Architecture

Robust and Reconfigurable On-Board Processing for a Hyperspectral Imaging Small Satellite

A new method of grafting multi-walled carbon nanotubes on carbon fibers for improving the mechanical and thermal properties of woven fabric composites

Contact Info

Product

Resources

About