The unprecedented progress in Artificial Intelligence (AI), particularly in deep learning algorithms with ubiquitous internet connected smart devices, has created a high demand for AI computing on the edge devices. This review studied commercially available edge processors, and the processors that are still in industrial research stages. We categorized state-of-the-art edge processors based on the underlying architecture, such as dataflow, neuromorphic, and Processing in-Memory (PIM) architecture. The processors are analyzed based on their performance, chip area, energy efficiency, and application domains. The supported programming frameworks, model compression, data precision, and the CMOS fabrication process technology are discussed. Currently, most of the commercial edge processors utilize dataflow architectures. However, emerging non-von Neumann computing architectures have attracted the industry in recent years. Neuromorphic processors are highly efficient for performing computation with fewer synaptic operations, and several neuromorphic processors offer online training for secured and personalized AI applications. This review found that the PIM processors show significant energy efficiency and consume less power compared to dataflow and neuromorphic processors. The future direction of the industry would be to implement state-of-the-art deep learning algorithms in emerging non-Von Neumann computing paradigms for low power computing on the edge devices.

Potassium 6-oxo-7,13,16,22-tetraazatetracyclo[12.6.2.18,12.017,21]tricosa-1(20),8(23),9,11,14,16,18,21-octaen-2-yne-15-carboxylate was synthesized through a multi-step pathway starting from commercially available 3-iodo-1,2-phenylenediamine. Structure characterization of this new substituted macrocyclic quinoxaline compound was achieved by using 1H NMR, 13C NMR, and HRMS spectral analysis. This new macrocyclic derivative demonstrated submicromolar potency on both Pim-1 and Pim-2 isoforms, with an interesting selectivity profile against a selected panel of human kinases.

Random copolymers made of both (PIM-polyimide) and (6FDA-durene-PI) were prepared for the first time by a facile one-step polycondensation reaction. By combining the highly porous and contorted structure of PIM (polymers with intrinsic microporosity) and high thermomechanical properties of PI (polyimide), the membranes obtained from these random copolymers [(PIM-PI)x-(6FDA-durene-PI)y] showed high CO₂ permeability (> 1047 Barrer) with moderate CO₂/N₂ (> 16.5) and CO₂/CH₄ (> 18) selectivity, together with excellent thermal and mechanical properties. The membranes prepared from three different compositions of two comonomers (1:4, 1:6 and 1:10 of x:y), all showed similar morphological and physical properties, and gas separation performance, indicating ease of synthesis and practicability for large-scale production. The gas separation performance of these membranes at various pressure ranges (100–1500 torr) was also investigated.

A polymer inclusion membrane (PIM) composed of 50 wt% poly(vinylidenefluoride-co-hexafluoropropylene) (PVDF-HFP) as its base polymer, 40 wt% Aliquat® 336 as its extractant and 10 wt% dibutyl phthalate (DBP) as plasticizer provided efficient extraction of vanadium(V) from its sulfate solutions adjusted to pH 2.5. It was suggested that V(V) was extracted as VO2SO4− via an anion exchange mechanism. Quantitative back-extraction was achieved in a sulfuric acid solution (6 mol L-1) containing 1 v/v% of hydrogen peroxide. It was assumed that the back-extraction process involved the oxidation of VO2+ to VO(O2)+ by hydrogen peroxide. The newly developed PIM with the optimized composition mentioned above exhibited excellent selectivity for V(V) in the presence of metallic species present in digests of spent alumina hydrodesulfurization catalysts (i.e., Al(III), Co(II), Cu(II), Fe(III), Mn(II), and Ni(II)). The co-extraction of Mo(VI) with V(V) was eliminated by its selective extraction at pH 1.1. The optimized PIM was characterized by contact angle measurements, atomic-force microscopy (AFM), energy dispersive X-ray spectroscopy (EDS), thermogravimetric analysis (TGA)/derivatives thermogravimetric analysis (DTGA), and the stress-strain measurements.

This comprehensive review explores the advancements in processing-in-memory (PIM) techniques for deep learning applications. It addresses the challenges faced by monolithic chip architectures and highlights the benefits of chiplet-based designs in terms of scalability, modularity, and flexibility. The review emphasizes the importance of dataflow-awareness, communication optimization, and thermal considerations in designing PIM-enabled manycore architectures. It discusses different machine learning workloads and their tailored dataflow requirements. Additionally, the review presents a heterogeneous PIM system for energy-efficient neural network training and discusses thermally efficient dataflow-aware monolithic 3D (M3D) NoC architectures for accelerating CNN inferencing. The advantages of TEFLON (Thermally Efficient Dataflow-Aware 3D NoC) over performance-optimized SFC-based counterparts are highlighted. Overall, this review provides valuable insights into the development and evaluation of chiplet and PIM architectures, emphasizing improved performance, energy efficiency, and inference accuracy in deep learning applications.