In recent years, the most powerful supercomputers have already reached megawatt power consumption levels, an important issue that challenges sustainability and shows the impossibility of maintaining this trend. To this date, the prevalent approach to super-computing is dominated by CPUs and GPUs. Given their fixed architectures with generic instruction sets, they have been favored with lots of tools and mature workflows which led to mass adoption and further growth. However, reconfigurable hardware such as FPGAs has repeatedly proven that it offers substantial advantages over this super-computing approach concerning performance and power consumption. In this survey, we review the most relevant works that advanced the field of heterogeneous super-computing using FPGAs focusing on their architectural characteristics. Each work was divided into three main parts: network, hardware, and software tools. All implementations face challenges that involve all three parts. These dependencies result in compromises that designers must take into account. The advantages and limitations of each approach are discussed and compared in detail. The classification and study of the architectures illustrate the trade-offs of the solutions and help identify open problems and research lines.

We present an high-resolution 3D (X,Y,t) imager for time-resolved experiments based on time-to-space conversion. The system uses Cross Delay-Line detectors (CDL) for particle identification and a fully-configurable digital processor based on Field Programmable Gate Arrays (FPGAs) for the 3D image reconstruction. The instrument reaches a spatial resolution of 45 μm FWHM (i.e.,18 μm r.m.s.) and a temporal precision of 15 ps r.m.s.. The detection rate achieved is 10 Mcps, with a dead-time below 7 ns, leading to a global throughput up to 6 Gb/s. In addition to the state-of-the-art performance, the innovative aspect of the presented contribution resides in the complete reconfigurability of the instrument: it is, in fact, the first time that the Time-to-Digital Converter (TDC) used for the time-to-space conversion, has been fully implemented in programmable logic, without the use of dedicated Application Specific Integrated Circuit (ASIC) components. This advancement allows to adapt the instrument to the experimental setup without undergoing impractical hardware modifications.

We report here an experimental setup to perform three-pulse pump–probe measurements over a wide wavelength and temperature range. By combining two pump pulses in the visible (650 nm–900 nm) and mid-IR (5 μm–20 μm) range, with a broadband supercontinuum white-light probe, our apparatus enables both the combined selective excitation of different material degrees of freedom and a full time-dependent reconstruction of the non-equilibrium dielectric function of the sample. We describe here the optical setup, the cryogenic sample environment, and the custom-made acquisition electronics capable of referenced single-pulse detection of broadband spectra at the maximum repetition rate of 50 kHz, achieving a sensitivity of the order of 10⁻⁴ over an integration time of 1 s. We demonstrate the performance of the setup by reporting data on a mid-IR pump, optical push, and broadband probe in a single crystal of Bi₂Sr₂Y_0.08Ca_0.92Cu₂O_8+δ across the superconducting and pseudogap phases.

X-ray ptychography is an advanced computational microscopy technique, which is delivering exceptionally detailed quantitative imaging of biological and nanotechnology specimens, which can be used for high-precision X-ray measurements. However, coarse parametrisation in propagation distance, position errors and partial coherence frequently threaten the experimental viability. In this work, we formally introduce these actors, solving the whole reconstruction as an optimisation problem. A modern deep learning framework was used to autonomously correct the setup incoherences, thus improving the quality of a ptychography reconstruction. Automatic procedures are indeed crucial to reduce the time for a reliable analysis, which has a significant impact on all the fields that use this kind of microscopy. We implemented our algorithm in our software framework, SciComPty, releasing it as open-source. We tested our system on both synthetic datasets, as well as on real data acquired at the TwinMic beamline of the Elettra synchrotron facility.

The dynamic response of gas sensors based on poly(3-hexylthiophene) (P3HT) nanofibers (NFs) to gaseous acetone was assessed using a setup based on flow-injection analysis, aimed at emulating actual breath exhalation. The setup was validated by using a commercially available sensor. The P3HT NFs sensors tested in dynamic flow conditions showed satisfactory reproducibility down to about 3.5 ppm acetone concentration, a linear response over a clinically relevant concentration range (3.5-35 ppm), excellent baseline recovery and reversibility upon repeated exposures to the analyte, short pulse rise and fall times (less than 1 s and about 2 s, respectively) and low power consumption (few nW), with no relevant response to water. Comparable responses’ decay times under either nitrogen or dry air suggest that the mechanisms at work is mainly attributable to specific analyte-semiconducting polymer interactions. These results open the way to the use of P3HT NFs-based sensing elements for the realization of portable, real-time electronic noses for on-the-fly exhaled breath analysis.

Face recognition functions are today exploited through biometric sensors in many applications, from extended security systems to inclusion devices; deep neural network methods are reaching in this field stunning performances. The main limitation of the deep learning approach is an inconvenient relation between the accuracy of the results and the needed computing power. When a personal device is employed, in particular, many algorithms require a cloud computing approach to achieve the expected performances; other algorithms adopt models that are simple by design. A third viable option consists of model (oracle) distillation. This is the most intriguing among the compression techniques since it permits to devise of the minimal structure that will enforce the same I/O relation as the original model. In this paper, a distillation technique is applied to a complex model, enabling the introduction of fast state-of-the-art recognition capabilities on a low-end hardware face recognition sensor module. Two distilled models are presented in this contribution: the former can be directly used in place of the original oracle, while the latter incarnates better the end-to-end approach, removing the need for a separate alignment procedure. The presented biometric systems are examined on the two problems of face verification and face recognition in an open set by using well-agreed training/testing methodologies and datasets.