Use of shared memory in the context of embedded multi-core processors: exploration of the technology and its limits

Modern embedded systems embrace many-core shared-memory designs. Due to constrained power and area budgets, most of them feature software-managed scratchpad memories instead of data caches to increase the data locality. It is therefore programmers’ responsibility to explicitly manage the memory transfers, and this make programming these platform cumbersome. Moreover, complex modern applications must be adequately parallelized before they can the parallel potential of the platform into actual performance. To support this, programming languages were proposed, which work at a high level of abstraction, and rely on a runtime whose cost hinders performance, especially in embedded systems, where resources and power budget are constrained. This dissertation explores the applicability of the shared-memory paradigm on modern many-core systems, focusing on the ease-of-programming. It focuses on OpenMP, the de-facto standard for shared memory programming. In a first part, the cost of algorithms for synchronization and data partitioning are analyzed, and they are adapted to modern embedded many-cores. Then, the original design of an OpenMP runtime library is presented, which supports complex forms of parallelism such as multi-level and irregular parallelism. In the second part of the thesis, the focus is on heterogeneous systems, where hardware accelerators are coupled to (many-)cores to implement key functional kernels with orders-of-magnitude of speedup and energy efficiency compared to the “pure software” version. However, three main issues rise, namely i) platform design complexity, ii) architectural scalability and iii) programmability. To tackle them, a template for a generic hardware processing unit (HWPU) is proposed, which share the memory banks with cores, and the template for a scalable architecture is shown, which integrates them through the shared-memory system. Then, a full software stack and toolchain are developed to support platform design and to let programmers exploiting the accelerators of the platform. The OpenMP frontend is extended to interact with it.

Novel hydroxystearoyl-, heterocyclic and carbocyclic derivatives: synthesis and applications in biological field

This PhD project has been mainly focused on the synthesis of novel organic compounds containing heterocyclic and/or carbocyclic scaffold and on the study of stearic acid derivatives and their applications in biological field. The synthesis of novel derivatives of 9-hydroxystearic acid (9-HSA) evidenced how the presence of substituents on C9, able to make hydrogen bonds is of crucial importance for the biological activity. Also the position of the hydroxy group along the chain of hydroxystearic acids was investigated: regioisomers with the hydroxy group bound to odd carbons resulted more active than those bearing the hydroxy group on even carbons. Further, the insertion of (R)-9-HSA in magnetic nanoparticles gave a novel material which characterization remarked its suitability for drug delivery. Structural hybrids between amino aza-heterocycles and azelaic acid have been synthesized and some of them showed a selective activity towards osteosarcoma cell line U2OS. Several Apcin analogues bearing indole, benzothiazole, benzofurazan moieties connected to tryptaminyl-, amino pyridinyl-, pyrimidinyl- and pyrazinyl ring through a 1,1,1-trichloroethyl group were synthesized. Biological tests showed the importance of both the tryptaminyl and the pyrimidinyl moieties, confirming the effectiveness against acute leukemia models. The SNAr between 2-aminothiazole derivatives and 7-chlorodinitrobenzofuroxan revealed different behaviour depending from amino substituent of the thiazole. The reaction with 2-N-piperidinyl-, 2-N-morpholinyl-, or 2-N-pyrrolidinyl thiazole gave two isomeric species derived from the attack on C-5 of thiazole ring. Thiazoles substituted with primary- or not-cyclic secondary amines reacted with the exocyclic amino nitrogen atom giving a series of compounds whose biological activity have highlighted as they might be promising candidates for further development of antitumor agents. A series of 9-fluorenylidene derivatives, of interest in medical and optoelectronic field as organic scintillators, was synthesized through Wittig or Suzuky reaction and will be analyzed to test their potential scintillatory properties.

Variational and learning models for image and time series inverse problems

Inverse problems are at the core of many challenging applications. Variational and learning models provide estimated solutions of inverse problems as the outcome of specific reconstruction maps. In the variational approach, the result of the reconstruction map is the solution of a regularized minimization problem encoding information on the acquisition process and prior knowledge on the solution. In the learning approach, the reconstruction map is a parametric function whose parameters are identified by solving a minimization problem depending on a large set of data. In this thesis, we go beyond this apparent dichotomy between variational and learning models and we show they can be harmoniously merged in unified hybrid frameworks preserving their main advantages. We develop several highly efficient methods based on both these model-driven and data-driven strategies, for which we provide a detailed convergence analysis. The arising algorithms are applied to solve inverse problems involving images and time series. For each task, we show the proposed schemes improve the performances of many other existing methods in terms of both computational burden and quality of the solution. In the first part, we focus on gradient-based regularized variational models which are shown to be effective for segmentation purposes and thermal and medical image enhancement. We consider gradient sparsity-promoting regularized models for which we develop different strategies to estimate the regularization strength. Furthermore, we introduce a novel gradient-based Plug-and-Play convergent scheme considering a deep learning based denoiser trained on the gradient domain. In the second part, we address the tasks of natural image deblurring, image and video super resolution microscopy and positioning time series prediction, through deep learning based methods. We boost the performances of supervised, such as trained convolutional and recurrent networks, and unsupervised deep learning strategies, such as Deep Image Prior, by penalizing the losses with handcrafted regularization terms.

PCM-based in-memory computing: architectures, circuits and applications

Analog In-memory Computing (AIMC) has been proposed in the context of Beyond Von Neumann architectures as a valid strategy to reduce internal data transfers energy consumption and latency, and to improve compute efficiency. The aim of AIMC is to perform computations within the memory unit, typically leveraging the physical features of memory devices. Among resistive Non-volatile Memories (NVMs), Phase-change Memory (PCM) has become a promising technology due to its intrinsic capability to store multilevel data. Hence, PCM technology is currently investigated to enhance the possibilities and the applications of AIMC. This thesis aims at exploring the potential of new PCM-based architectures as in-memory computational accelerators. In a first step, a preliminar experimental characterization of PCM devices has been carried out in an AIMC perspective. PCM cells non-idealities, such as time-drift, noise, and non-linearity have been studied to develop a dedicated multilevel programming algorithm. Measurement-based simulations have been then employed to evaluate the feasibility of PCM-based operations in the fields of Deep Neural Networks (DNNs) and Structural Health Monitoring (SHM). Moreover, a first testchip has been designed and tested to evaluate the hardware implementation of Multiply-and-Accumulate (MAC) operations employing PCM cells. This prototype experimentally demonstrates the possibility to reach a 95% MAC accuracy with a circuit-level compensation of cells time drift and non-linearity. Finally, empirical circuit behavior models have been included in simulations to assess the use of this technology in specific DNN applications, and to enhance the potentiality of this innovative computation approach.

A robotic platform for precision agriculture and applications

Agricultural techniques have been improved over the centuries to match with the growing demand of an increase in global population. Farming applications are facing new challenges to satisfy global needs and the recent technology advancements in terms of robotic platforms can be exploited. As the orchard management is one of the most challenging applications because of its tree structure and the required interaction with the environment, it was targeted also by the University of Bologna research group to provide a customized solution addressing new concept for agricultural vehicles. The result of this research has blossomed into a new lightweight tracked vehicle capable of performing autonomous navigation both in the open-filed scenario and while travelling inside orchards for what has been called in-row navigation. The mechanical design concept, together with customized software implementation has been detailed to highlight the strengths of the platform and some further improvements envisioned to improve the overall performances. Static stability testing has proved that the vehicle can withstand steep slopes scenarios. Some improvements have also been investigated to refine the estimation of the slippage that occurs during turning maneuvers and that is typical of skid-steering tracked vehicles. The software architecture has been implemented using the Robot Operating System (ROS) framework, so to exploit community available packages related to common and basic functions, such as sensor interfaces, while allowing dedicated custom implementation of the navigation algorithm developed. Real-world testing inside the university’s experimental orchards have proven the robustness and stability of the solution with more than 800 hours of fieldwork. The vehicle has also enabled a wide range of autonomous tasks such as spraying, mowing, and on-the-field data collection capabilities. The latter can be exploited to automatically estimate relevant orchard properties such as fruit counting and sizing, canopy properties estimation, and autonomous fruit harvesting with post-harvesting estimations.

Synthesis, multiscale-multiphase characterization and applications of thiophene-based biohybrids

Biohybrid derivatives of π-conjugated materials are emerging as powerful tools to study biological events through the (opto)electronic variations of the π-conjugated moieties, as well as to direct and govern the self-assembly properties of the organic materials through the organization principles of the bio component. So far, very few examples of thiophene-based biohybrids have been reported. The aim of this Ph. D thesis has been the development of oligothiophene-oligonucleotide hybrid derivatives as tools, on one side, to detect DNA hybridisation events and, on the other, as model compounds to investigate thiophene-nucleobase interactions in the solid state. To obtain oligothiophene bioconjugates with the required high level of purity, we first developed new synthetic ecofriendly protocols for the synthesis of thiophene oligomers. Our innovative heterogeneous Suzuki coupling methodology, carried out in EtOH/water or isopropanol under microwave irradiation, allowed us to obtain alkyl substituted oligothiophenes and thiophene based co-oligomers in high yields and very short reaction times, free from residual metals and with improved film forming properties. These methodologies were subsequently applied in the synthesis of oligothiophene-oligonucleotide conjugates. Oligothiophene-5-labeled deoxyuridines were synthesized and incorporated into 19-meric oligonucletide sequences. We showed that the oligothiophene-labeled oligonucletide sequences obtained can be used as probes to detect a single nucleotide polymorphism (SNP) in complementary DNA target sequences. In fact, all the probes showed marked variations in emission intensity upon hybridization with a complementary target sequence. The observed variations in emitted light were comparable or even superior to those reported in similar studies, showing that the biohybrids can potentially be useful to develop biosensors for the detection of DNA mismatches. Finally, water-soluble, photoluminescent and electroactive dinucleotide-hybrid derivatives of quaterthiophene and quinquethiophene were synthesized. By means of a combination of spectroscopy and microscopy techniques, electrical characterizations, microfluidic measurements and theoretical calculations, we were able to demonstrate that the self-assembly modalities of the biohybrids in thin films are driven by the interplay of intra and intermolecular interactions in which the π-stacking between the oligothiophene and nucleotide bases plays a major role.

Internal Model Principle: extension to the switching case and applications

This thesis deals with a novel control approach based on the extension of the well-known Internal Model Principle to the case of periodic switched linear exosystems. This extension, inspired by power electronics applications, aims to provide an effective design method to robustly achieve the asymptotic tracking of periodic references with an infinite number of harmonics. In the first part of the thesis the basic components of the novel control scheme are described and preliminary results on stabilization are provided. In the second part, advanced control methods for two applications coming from the world high energy physics are presented.

3D probability tomography: theoretical developments and applications to high-resolution geophysical prospecting

The theory of the 3D multipole probability tomography method (3D GPT) to image source poles, dipoles, quadrupoles and octopoles, of a geophysical vector or scalar field dataset is developed. A geophysical dataset is assumed to be the response of an aggregation of poles, dipoles, quadrupoles and octopoles. These physical sources are used to reconstruct without a priori assumptions the most probable position and shape of the true geophysical buried sources, by determining the location of their centres and critical points of their boundaries, as corners, wedges and vertices. This theory, then, is adapted to the geoelectrical, gravity and self potential methods. A few synthetic examples using simple geometries and three field examples are discussed in order to demonstrate the notably enhanced resolution power of the new approach. At first, the application to a field example related to a dipole–dipole geoelectrical survey carried out in the archaeological park of Pompei is presented. The survey was finalised to recognize remains of the ancient Roman urban network including roads, squares and buildings, which were buried under the thick pyroclastic cover fallen during the 79 AD Vesuvius eruption. The revealed anomaly structures are ascribed to wellpreserved remnants of some aligned walls of Roman edifices, buried and partially destroyed by the 79 AD Vesuvius pyroclastic fall. Then, a field example related to a gravity survey carried out in the volcanic area of Mount Etna (Sicily, Italy) is presented, aimed at imaging as accurately as possible the differential mass density structure within the first few km of depth inside the volcanic apparatus. An assemblage of vertical prismatic blocks appears to be the most probable gravity model of the Etna apparatus within the first 5 km of depth below sea level. Finally, an experimental SP dataset collected in the Mt. Somma-Vesuvius volcanic district (Naples, Italy) is elaborated in order to define location and shape of the sources of two SP anomalies of opposite sign detected in the northwestern sector of the surveyed area. The modelled sources are interpreted as the polarization state induced by an intense hydrothermal convective flow mechanism within the volcanic apparatus, from the free surface down to about 3 km of depth b.s.l..

Metaheuristics for Search Problems in Genomics - New Algorithms and Applications

In this thesis we made the first steps towards the systematic application of a methodology for automatically building formal models of complex biological systems. Such a methodology could be useful also to design artificial systems possessing desirable properties such as robustness and evolvability. The approach we follow in this thesis is to manipulate formal models by means of adaptive search methods called metaheuristics. In the first part of the thesis we develop state-of-the-art hybrid metaheuristic algorithms to tackle two important problems in genomics, namely, the Haplotype Inference by parsimony and the Founder Sequence Reconstruction Problem. We compare our algorithms with other effective techniques in the literature, we show strength and limitations of our approaches to various problem formulations and, finally, we propose further enhancements that could possibly improve the performance of our algorithms and widen their applicability. In the second part, we concentrate on Boolean network (BN) models of gene regulatory networks (GRNs). We detail our automatic design methodology and apply it to four use cases which correspond to different design criteria and address some limitations of GRN modeling by BNs. Finally, we tackle the Density Classification Problem with the aim of showing the learning capabilities of BNs. Experimental evaluation of this methodology shows its efficacy in producing network that meet our design criteria. Our results, coherently to what has been found in other works, also suggest that networks manipulated by a search process exhibit a mixture of characteristics typical of different dynamical regimes.

Novel tools for conservation genetics in marine fish: population structure and evolution of the European hake (Merluccius merluccius) inferred by SNP variation and applications to traceability

The research presented in my PhD thesis is part of a wider European project, FishPopTrace, focused on traceability of fish populations and products. My work was aimed at developing and analyzing novel genetic tools for a widely distributed marine fish species, the European hake (Merluccius merluccius), in order to investigate population genetic structure and explore potential applications to traceability scenarios. A total of 395 SNPs (Single Nucleotide Polymorphisms) were discovered from a massive collection of Expressed Sequence Tags, obtained by high-throughput sequencing, and validated on 19 geographic samples from Atlantic and Mediterranean. Genome-scan approaches were applied to identify polymorphisms on genes potentially under divergent selection (outlier SNPs), showing higher genetic differentiation among populations respect to the average observed across loci. Comparative analysis on population structure were carried out on putative neutral and outlier loci at wide (Atlantic and Mediterranean samples) and regional (samples within each basin) spatial scales, to disentangle the effects of demographic and adaptive evolutionary forces on European hake populations genetic structure. Results demonstrated the potential of outlier loci to unveil fine scale genetic structure, possibly identifying locally adapted populations, despite the weak signal showed from putative neutral SNPs. The application of outlier SNPs within the framework of fishery resources management was also explored. A minimum panel of SNP markers showing maximum discriminatory power was selected and applied to a traceability scenario aiming at identifying the basin (and hence the stock) of origin, Atlantic or Mediterranean, of individual fish. This case study illustrates how molecular analytical technologies have operational potential in real-world contexts, and more specifically, potential to support fisheries control and enforcement and fish and fish product traceability.

Low Correlated Sequences & Architectures and Algorithms for Analog-to-Information Converters: Theory, Design, Implementation and Applications.

Most electronic systems can be described in a very simplified way as an assemblage of analog and digital components put all together in order to perform a certain function. Nowadays, there is an increasing tendency to reduce the analog components, and to replace them by operations performed in the digital domain. This tendency has led to the emergence of new electronic systems that are more flexible, cheaper and robust. However, no matter the amount of digital process implemented, there will be always an analog part to be sorted out and thus, the step of converting digital signals into analog signals and vice versa cannot be avoided. This conversion can be more or less complex depending on the characteristics of the signals. Thus, even if it is desirable to replace functions carried out by analog components by digital processes, it is equally important to do so in a way that simplifies the conversion from digital to analog signals and vice versa. In the present thesis, we have study strategies based on increasing the amount of processing in the digital domain in such a way that the implementation of analog hardware stages can be simplified. To this aim, we have proposed the use of very low quantized signals, i.e. 1-bit, for the acquisition and for the generation of particular classes of signals.

Development and applications of hollow fiber flow field-flow fractionation in the bioanalytical field. Studies of aggregation phenomena in complex protein samples

Recent advances in the fast growing area of therapeutic/diagnostic proteins and antibodies - novel and highly specific drugs - as well as the progress in the field of functional proteomics regarding the correlation between the aggregation of damaged proteins and (immuno) senescence or aging-related pathologies, underline the need for adequate analytical methods for the detection, separation, characterization and quantification of protein aggregates, regardless of the their origin or formation mechanism. Hollow fiber flow field-flow fractionation (HF5), the miniaturized version of FlowFFF and integral part of the Eclipse DUALTEC FFF separation system, was the focus of this research; this flow-based separation technique proved to be uniquely suited for the hydrodynamic size-based separation of proteins and protein aggregates in a very broad size and molecular weight (MW) range, often present at trace levels. HF5 has shown to be (a) highly selective in terms of protein diffusion coefficients, (b) versatile in terms of bio-compatible carrier solution choice, (c) able to preserve the biophysical properties/molecular conformation of the proteins/protein aggregates and (d) able to discriminate between different types of protein aggregates. Thanks to the miniaturization advantages and the online coupling with highly sensitive detection techniques (UV/Vis, intrinsic fluorescence and multi-angle light scattering), HF5 had very low detection/quantification limits for protein aggregates. Compared to size-exclusion chromatography (SEC), HF5 demonstrated superior selectivity and potential as orthogonal analytical method in the extended characterization assays, often required by therapeutic protein formulations. In addition, the developed HF5 methods have proven to be rapid, highly selective, sensitive and repeatable. HF5 was ideally suitable as first dimension of separation of aging-related protein aggregates from whole cell lysates (proteome pre-fractionation method) and, by HF5-(UV)-MALS online coupling, important biophysical information on the fractionated proteins and protein aggregates was gathered: size (rms radius and hydrodynamic radius), absolute MW and conformation.