ACE16k: The Third Generation of Mixed-Signal SIMD-CNN ACE Chips Toward VSoCs

Rodríguez-Vázquez, Á.; Liñán‐Cembrano, Gustavo; Carranza, Luis E.; Roca, E.; Carmona-Galán, Ricardo; Jimenez-Garrido, F.; Domínguez-Castro, R.; Espejo, S.

doi:10.1109/tcsi.2004.827621

Cited by 203 publications

(114 citation statements)

References 10 publications

Supporting

Mentioning

113

Contrasting

Unclassified

Order By: Relevance

“…Since these algorithms rely on local properties, diffusions and multiscale representations play a prominent role for their implementation. This is actually the very domain of analog and mixedsignal circuits [8][27] [27]. 3D sensor architectures for the extraction of features and the calculation of interest points within imagers have been previously proposed by the authors in [27] and [28].…”

Section: Mapping Smart-pixel Cvis Onto 3-d Architecturesmentioning

confidence: 99%

Form factor improvement of smart-pixels for vision sensors through 3-D vertically-integrated technologies

Rodríguez-Vázquez

Carmona-Galán

Berni

et al. 2014

2014 IEEE 5th Latin American Symposium on Circuits and Systems

View full text Add to dashboard Cite

While conventional CMOS active pixel sensors embed only the circuitry required for photo-detection, pixel addressing and voltage buffering, smart pixels incorporate also circuitry for data processing, data storage and control of data interchange. This additional circuitry enables data processing be realized concurrently with the acquisition of images which is instrumental to reduce the number of data needed to carry to information contained into images. This way, more efficient vision systems can be built at the cost of larger pixel pitch. Vertically-integrated 3D technologies enable to keep the advnatges of smart pixels while improving the form factor of smart pixels.

show abstract

Section: Mapping Smart-pixel Cvis Onto 3-d Architecturesmentioning

confidence: 99%

Form factor improvement of smart-pixels for vision sensors through 3-D vertically-integrated technologies

Rodríguez-Vázquez

Carmona-Galán

Berni

et al. 2014

2014 IEEE 5th Latin American Symposium on Circuits and Systems

View full text Add to dashboard Cite

show abstract

“…Already in 2004, a state-of-the-art programmable, commercially available CNN in a 0.35-µm standard CMOS technology exhibited peak computing figures of 330 GOPS [23] (or 3.6 GOPS/mm 2 and 82.5 GOPS/W in terms of area and power consumption). These numbers are yet excelled by non-programmable CNNs, which we propose for use as SIMPL systems.…”

Section: Introduction and General Ideamentioning

confidence: 99%

“…In specialized tasks, it is known that CNNs can outperform digital computers by a factor of up to 1,000 [24,25]. CNNs are the largest analog circuits, with the CNN referred to above [23] …”

Section: Introduction and General Ideamentioning

confidence: 99%

Towards Electrical, Integrated Implementations of SIMPL Systems

Rührmair¹,

Chen²,

Stutzmann

et al. 2010

Information Security Theory and Practices. Security and Privacy of Pervasive Systems and Smart Devices

View full text Add to dashboard Cite

This paper discusses the practical implementation of a novel security tool termed SIMPL system, which was introduced in [1]. SIMPL systems can be regarded as a public key version of physical unclonable functions (PUFs). Like the latter, a SIMPL system S is physically unique and nonreproducible, and implements an individual function F S . In opposition to a PUF, however, a SIMPL system S possesses a publicly known numerical description, which allows its digital simulation and prediction. At the same time, any such simulation must work at a detectably lower speed than the real-time behavior of S. As argued in [1], SIMPL systems have certain practicality and security advantages in comparison to PUFs, certificates of authenticity, physically obfuscated keys, and also to standard mathematical cryptotechniques.In [1], definitions, protocols, and optical implementations of SIMPL systems were presented. This manuscript focuses on concrete electrical, integrated realizations of SIMPL systems, and proposes two potential candidates: SIMPL systems derived from special SRAM-architectures (socalled "skew designs" of SRAM cells), and implementations based on Cellular Non-Linear Networks (CNNs). IntroductionPhysical Unclonable Functions (PUFs) are a relatively young, emerging cryptographic primitive [2,3,4,5,6,7]. However, one potential downside of PUF-based protocols is that they usually require a previously shared piece of information (typically some challenge-response-pairs) that was established in a joint set-up phase between the communicants. Alternatively, an online connection to a trusted authority at the time of the protocol execution must be employed. In this particular structural aspect, PUFs are resemblant of classical private key systems.In this paper, we are concerned with an alternative security tool called SIMPL systems, which is a public key version of standard PUFs. SIMPL systems have been introduced in [1]. The acronym SIMPL stands for "SIMulation Possible, but Laborious", and hints at the critical security feature of these structures. A physical system S is called a SIMPL system if the following holds:1. It is possible for everyone to numerically simulate and, thus, to predict the physical behaviour of S with very high accuracy. The basis of the simulation is an individual description D(S) of S, and a generic simulation algorithm Sim, which are both publicly known.

show abstract

“…Many low-level vision problems can be solved efficiently by the use of pixel-per-processor cellular SIMD architectures [1], [2], [3], [4], [5]. But these cellular processor arrays lack fast long distance communication, as the cells are only connected in the nearest neighborhood.…”

Section: Introductionmentioning

confidence: 99%

Autonomous long distance transfer on SIMD cellular processor arrays

Geese

Dudek

2010

2010 12th International Workshop on Cellular Nanoscale Networks and Their Applications (CNNA 2010)

View full text Add to dashboard Cite

Abstract-In image processing the tasks of rotating, mirroring and scaling the image are often required. These operations necessitate data transfer between distant elements in the image. SIMD processor arrays commonly support communication within the direct neighborhood only. In this paper a method for implementing long-distance data communication is proposed. Inspired by lattice gas cellular automata models and marching pixels, the developed method leads to an efficient pixel routing algorithm. Further, the use of autonomous elements leads to an emergent behavior that has been studied. Promising results for the above tasks have been achieved for simulations on nearestneighbor connected processor arrays.

show abstract

ACE16k: The Third Generation of Mixed-Signal SIMD-CNN ACE Chips Toward VSoCs

Cited by 203 publications

References 10 publications

Form factor improvement of smart-pixels for vision sensors through 3-D vertically-integrated technologies

Form factor improvement of smart-pixels for vision sensors through 3-D vertically-integrated technologies

Towards Electrical, Integrated Implementations of SIMPL Systems

Autonomous long distance transfer on SIMD cellular processor arrays

Contact Info

Product

Resources

About