Search

Out-of-core systems rely on high-performance cache sub-systems to reduce the number of I/O operations. Although the page cache in modern operating systems enables transparent access to memory and storage devices, it suffers from efficiency and scalability issues on cache misses, forcing out-of-core systems to design and implement their own cache components, which is a non-trivial task. This study proposes TriCache, a cache mechanism that enables in-memory programs to efficiently process out-of-core datasets without requiring any code rewrite. It provides a virtual memory interface on top of the conventional block interface to simultaneously achieve user transparency and sufficient out-of-core performance. A multi-level block cache design is proposed to address the challenge of per-access address translations required by a memory interface. It can exploit spatial and temporal localities in memory or storage accesses to render storage-to-memory address translation and page-level concurrency control adequately efficient for the virtual memory interface. Our evaluation shows that in-memory systems operating on top of TriCache can outperform Linux OS page cache by more than one order of magnitude, and can deliver performance comparable to or even better than that of corresponding counterparts designed specifically for out-of-core scenarios.

Apache Spark is a widely deployed big data analytics framework that offers such attractive features as resiliency, load-balancing, and a rich ecosystem. However, there is still plenty of room for improvement in its performance. Although a data-parallel system in a native programming language significantly improves performance, it may require re-implementing many functionalities of Spark to become a full-featured system. It is desirable for native big data systems to just write a compute engine in native languages to ensure high efficiency, and reuse other mature features provided by Spark rather than re-implement everything. But the interaction between the JVM and the native world risks becoming a bottleneck. This paper proposes Chukonu, a native big data framework that re-uses critical big data features provided by Spark. Owing to our novel DAG-splitting approach, the potential Spark integration overhead is alleviated, and its even outperforms existing pure native big data frameworks. Chukonu splits DAG programs into run-time parts and compile-time parts: The run-time parts are delegated to Spark to offload the complexities due to feature implementations. The compile-time parts are natively compiled. We propose a series of optimization techniques to be applied to the compile-time parts, such as operator fusion, vectorization, and compaction, to significantly reduce the Spark integration overhead. The results of evaluation show that Chukonu has a speedup of up to 71.58X (geometric mean 6.09X) over Apache Spark, and up to 7.20X (geometric mean 2.30X) over pure-native frameworks on six commonly-used big data applications. By translating the physical plan produced by SparkSQL into Chukonu programs, Chukonu accelerates Spark-SQL's TPC-DS performance by 2.29X.

Vapor deposition of perovskite solar cells (PSCs) has attracted considerable interest for its dry processing characteristics. However, a two-step sequential vapor deposition method suffers from ineffective conversion of PbI2 to perovskite with reasons still unclear. In this report, we carefully investigated the crystallization orientation of PbI2 films deposited by physical vapor deposition via synchrotron grazing-incidence wide-angle X-ray scattering (GIWAXS) and observed an asymmetric scattering pattern with respect to the qz-axis. The observed oriented morphology and texture hinder the diffusion of MAI molecules in the PbI2 films synthesized by vapor deposition, resulting in over 15% PbI2 remaining at the buried interface after reaction with MAI vapor. As a result, the MAPbI3 synthesized in this way was also highly oriented, especially in the surface layers. Surface fumigation (SF) step was introduced to decrease the orientational anisotropy of PbI2, which successfully breaks the diffusion barriers of MAI molecules by forming a complex layer on the PbI2 surface with polar solvent vapors, like dimethyl sulfoxide or 1,3-dimethyl-2-imidazolidinone. We infer that the SF treatment changes the vapor-solid reaction mechanism from reaction-crystallization to dissolution-recrystallization, which largely promotes the conversion of PbI2 to perovskite. Defects were reduced in perovskite synthesized in this way, and a p-i-n device with 19.56% efficiency was fabricated, which is among the highest efficiencies reported for sequential-vapor-deposited PSCs. Notably, this method enables the fabrication of conformal perovskite layers on uneven substrates. An exemplary PSC showing efficiency of 8.93% was fabricated on a precurved substrate. We believe that the method is applicable to the fabrication of tandem or curved PSCs that are compatible with wearable or building/autocar-integrated photovoltaics in the future.

The efficiency and stability of perovskite solar cells (PSCs) can be either deteriorated or enhanced by strain at interfaces, which is sensitive to various external conditions, particularly light illumination. Here we investigated the vertical strain distribution in perovskite films synthesized under light illumination with various wavelengths. The films were formed by reacting formamidinium iodide (FAI)/methylammonium chloride (MACl) vapor with vapor-deposited PbI

With the development of renewable energy systems, clean hydrogen is burgeoning as an optimal alternative to fossil fuels, in which its application is promising to retarding the global energy and environmental crisis. The hydrogen evolution reaction (HER), capable of producing high-purity hydrogen rapidly in electrocatalytic water splitting, has received much attention. Abundant research about HER has been done, focusing on advanced electrocatalyst design with high efficiency and robust stability. As potential HER catalysts, metal nanoclusters (MNCs) have been studied extensively. They are composed of several to a hundred metal atoms, with sizes being comparable to the Fermi wavelength of electrons, that is,2.0 nm. Different from metal atoms/nanoparticles, they exhibit unique catalytic properties due to their quantum size effect and low-coordination environment. In this review, the activity-enhancing approaches of MNCs applied in HER electrocatalysis are mainly summarized. Furthermore, recent progress in MNCs classified with different stabilization strategies, that is, the freestanding MNCs, MNCs with organic, metal and carbon supports, are introduced. Finally, the current challenges and deficiencies of these MNCs for HER are prospected.

Sputtering with copper indium gallium selenide (CIGS) ceramic targets could produce smooth CIGS thin films that are preferred for preparing two-terminal tandem devices. However, grain sizes prepared in this way are small and device efficiency was low. To increase the grain size, in this report, an Ag layer was pre-sputtered beneath CIGS. The Ag doping layer increased the grain size and improved the crystalline alignment. Consequently, the Ag-doped films exhibited improved charge mobility. From X-ray diffraction, scanning electron microscopy and X-ray photoelectron spectroscopy characterizations, we obtained an optimized Ag thickness of 15 nm. Short-circuit current density (JSC), open-circuit voltage (VOC), and fill factor (FF) were all improved after doping with 15-nm Ag. Increasing the annealing temperature from 550 °C to 575 °C, the grains was enlarged further, with the power conversion efficiency (PCE) increasing to 14.33% and VOC to 545 mV. Upon the smooth CIGS film, a thin conformal perovskite layer was fabricated without polishing. This work demonstrates a simple way to fabricate smooth and highly-crystalline CIGS films that can be used for tandem solar cells.

Due to mass diffusion issues, it is challenging to prepare black-phase thick formamidinium-based perovskite (FAPbI3) films via vapor approaches. Precursor engineering is employed here to overcome the dilemma of thorough reaction and black-phase stabilization of FAPbI3 in a sequential vapor approach. For the first time, FAPbBr3 was used as an additive in the precursor to promote the formation of FAPbI3 perovskite. To balance off the increased crystallization degree of precursor films due to the addition of FAPbBr3, CsI dissolved in dimethyl sulfoxide (DMSO) was further added. It is indicated that the simultaneous incorporation of FAPbBr3 and CsI-DMSO successfully accelerated the formation rate of perovskite and inhibited the formation of FAPbI3 yellow phase. The power conversion efficiency of the as-prepared devices of different areas (0.1125 or 1 cm2) reached 20%, the first report of large-area 20%-efficiency PSCs based on a vapor approach, highlighting its applicability to large-area manufacture in the future. Furthermore, when blade coating is used in preparing the precursor film, the efficiency reached 19%. When the precursor film was prepared by dip coating, we could prepare conformal FAPbI3 coatings on carbon fibers, suggesting possible future applications in fabricating wearable PSCs.

Vapor deposition of perovskite solar cells (PSCs) has attracted considerable interest for its dry processing characteristics. However, a two-step sequential vapor deposition method suffers from ineffective conversion of PbI

Machine learning applications on Spark suffers from poor scalability. In this paper, we reveal that the key reasons is the non-scalable reduction, which is restricted by the non-splittable object programming interface in Spark. This insight guides us to propose Sparker, Spark with Efficient Reduction. By providing a split aggregation interface, Sparker is able to perform split aggregation with scalable reduction while being backward compatible with existing applications. We implemented Sparker in 2,534 lines of code. Sparker can improve the aggregation performance by up to 6.47 × and can improve the end-to-end performance of MLlib model training by up to 3.69 × with a geometric mean of 1.81 × .

Defects lower the performance of perovskite solar cells by playing as recombination centers of carriers and migration channels of ions under illumination. Here we used various kinds of lights to as...

Air-stable precursor films are critical to reproducibly fabricating the large-area perovskite solar cells (PSCs) in sequential deposition. In traditional sequential vapor approaches, perovskite fil...

Herein, 1,6-Diaminohexane Dihydrochloride (1,6-DD) is introduced into perovskite precursors to fabricate the inverted planar perovskite solar cells. By regulating the concentration of 1,6-DD additive, the average power conversion efficiency (PCE) of perovskite solar cells is enhanced by 20%. The champion device achieves a relatively high PCE of 17% and an excellent fill factor of 80.1%. The PCE of the large-area (1 cm2) device also reaches to 13.68%. After exposure to the air for 16 days, the device with 1,6-DD additive still retains above 90% of the initial efficiency, exhibiting good stability. We demonstrate that a small amount of 1,6-DD affects the crystallization dynamic, yielding ideal perovskite film with enhanced crystallinity and enlarged grain size. The two terminal -NH3+ groups passivates the vacancy defects at the perovskite crystal surface, suppressing charge recombination and facilitating charge transportation effectively. Meanwhile, adjacent crystal surfaces are linked through the hexane alkyl chain of 1,6-DD molecule, which enhances the interaction between perovskite grains and anchors the microstructure of perovskite to some degree. Hydrophobic hexane alkyl chains also increase the moisture resistance of perovskite film. Thus, an easy and effective way is provided for fabricating efficient and stable perovskite solar cells.

Pre-trained Language Models (PLMs) have proven to be beneficial for various downstream NLP tasks. Recently, GPT-3, with 175 billion parameters and 570 GB training data, drew a lot of attention due to the capacity of few-shot (even zero-shot) learning. However, applying GPT-3 to address Chinese NLP tasks is still challenging, as the training corpus of GPT-3 is primarily English, and the parameters are not publicly available. In this technical report, we release the Chinese Pre-trained Language Model (CPM) with generative pre-training on large-scale Chinese training data. To the best of our knowledge, CPM, with 2.6 billion parameters and 100 GB Chinese training data, is the largest Chinese pre-trained language model, which could facilitate several downstream Chinese NLP tasks, such as conversation, essay generation, cloze test, and language understanding. Extensive experiments demonstrate that CPM achieves strong performance on many NLP tasks in the settings of few-shot (even zero-shot) learning. The code and parameters are available at https://github.com/TsinghuaAI/CPM .

Due to mass diffusion issues, it is challenging to prepare black-phase thick formamidinium-based perovskite (FAPbI

We employed mixed PbCl2:PbAc2 as lead sources and followed the two-step sequential deposition method to prepare the perovskite films. We focused on the influence of substrate-preheating treatment on crystal growth, morphology, optical and electrical properties of the final perovskite film, as well as the device performance. A combination of characterizations including scanning electron microscope (SEM), conducting atomic force microscopy (C-AFM), energy-dispersive X-ray (EDX), X-ray diffraction (XRD), UV–vis absorption, time-resolved photoluminescence spectra (TRPL), and surface photovoltage spectroscopy (SPS) was utilized to investigate the effect of substrate-preheating treatment. It was demonstrated that appropriate substrate-preheating treatment could realize controllable perovskite crystallization, leading to high-quality perovskite film with improved optical and electrical properties. The devices with the optimized perovskite films exhibited good reproducibility and stability, delivering an ideal PCE of 15.37%, with decent FF of 73.84%, high Voc of 0.98 V, and Jsc of 21.24 mA/cm2.

Herein, a low-melting point tin-bismuth (SnBi) alloy film combined with a silver layer was introduced to fabricate the inverted planar perovskite solar cells. SnBi alloy was prepared by high temperature melting method and was characterized by metallurgical microscope, XRD, SEM, DTA, XPS methods. The influence of silver thickness on the performance of perovskite solar cells was studied. The optimized perovskite solar cells based on 20 nm Ag and 80 nm SnBi show a power conversion efficiency of 13.52%, involving a 19.6% increase compared to the control device. The double-layer electrode can not only reduce the charge transfer resistance, increase the extraction efficiency of the charge, improve the photovoltaic performance of the device, but also improve the stability of the device in the environment.

Phenyl-C61-butyric acid methyl ester (PCBM) doped with bathocuproine (BCP) is employed as a novel electron transport layer (ETL) for the first time. We demonstrate that this solution-processed PCBM:BCP blend ETL could significantly improve perovskite solar cell performance in a wide range of BCP doping amounts. The effect of BCP doping has been systematically investigated through a combination of characterizations, including atomic force microscopy, UV–vis absorption spectra, Kelvin probe measurement, surface photovoltage spectroscopy, impedance spectroscopy, and the contact angle test. It is found that the incorporation of BCP not only ameliorates the film formation property of PCBM and the interfacial contact but facilitates the charge transport and separation at the interface through effectively passivating the surface trap states of perovskite as well. As a result, the power conversion efficiency is improved from 7.05% to nearly doubled 13.11%, mainly benefiting from the enhanced short-circuit current density ( J sc) of 18.96 mA/cm2 and fill factor (FF) of 73%. Solution-processed PCBM ETL with BCP as additive endows the devices with excellent performance comparable to or even better than that of traditional devices with evaporated BCP or C60/BCP. It is also an effective and convenient approach to simplify the device fabrication procedure, with promising application in large-scale production.

We present a language and compilation framework for productively generating high-performance systolic arrays for dense tensor kernels on spatial architectures, including FPGAs and CGRAs. It decouples a functional specification from a spatial mapping, allowing programmers to quickly explore various spatial optimizations for the same function. The actual implementation of these optimizations is left to a compiler. Thus, productivity and performance are achieved at the same time. We used this framework to implement several important dense tensor kernels. We implemented dense matrix multiply for an Arria-10 FPGA and a research CGRA, achieving 88% and 92% of the performance of manually written, and highly optimized expert (ninja") implementations in just 3% of their engineering time. Three other tensor kernels, including MTTKRP, TTM and TTMc, were also implemented with high performance and low design effort, and for the first time on spatial architectures."

We report a doping method to enhance the performance of solution processed planar heterojunction perovskite solar cells. By doping a small percentage (3 wt%) of polyethylenimine (PEI) as additive into the PCBM electron transport layer of an inverted perovskite solar cell, which led to significant enhancements of power conversion efficiency from (5.9 ± 0.2) % to (10.4 ± 0.2) %. The AFM images show that the PEI doped PCBM layer can help to form a high quality, homogeneous and compact electron transporting layer on the rough CH3NH3PbI3 layer, which results in enhanced hole blocking ability and reduced leakage current at the interfaces between the CH3NH3PbI3, PCBM films and the top Al electrode. Organic field-effect transistors (OFETs) measurements reveal that the addition of 1–3 wt% PEI into PCBM layer can improve device performance without any negative effect on the electron transport property of PCBM. Steady-state PL analysis shows that the electron-rich PEI may also act as an effective interfacial modifier to passivate the trap states at the perovskite surface or crystal boundaries and to avoid the undesired charge recombination often observed in perovskite solar cells. PEI will also improve performance as a cathode interfacial modifier because the PCE of the device with PEI deposited between PCBM and Al is superior to the device without PEI. This work demonstrated that amine-containing polymer materials can be used as an efficient dual functional additive in perovskite solar cells. This study provides an efficient way of developing highly efficient CH3NH3PbI3-based perovskite solar cells.

A new vapor–solid reaction (VSR) method is demonstrated, in which two parallel hot plates were used to synthesize high-quality organic–inorganic hybrid perovskite thin films over an area of 8 × 8 cm2. We first probed the VSR process in detail, and found side reactions between methylammonium iodide (MAI) vapor and lead iodide (PbI2) occurring, which could be largely suppressed by inserting a PEDOT:PSS layer beneath. Furthermore, we employed an in situ resistance measurement method to monitor the evolution of perovskite thin films, which provided valuable insights for precise control of film growth. Inverted planar perovskite solar cells with a decent efficiency of 12.2% have been achieved. Our work provides not only a strategy to fabricate high-quality perovskite thin films over a large area, but also a reliable way to optimize the reaction parameters of perovskite. In addition, our finding reveals that both the precise control of the ratio of MAI to PbI2 and the proper choice of substrates are decisive for achieving high-performance devices during the VSR process.

Normally, it is difficult to directly measure the bandgaps of perovskite based on methylammonium (MA) or formamidinium (FA) at high temperatures due to material decomposition. We prevent the decomposition by keeping the synthesized perovskite films (MAPbI3 and MAPbI3) in organic iodide vapors, then measure the in-situ resistance of the films at varied temperatures, and further evaluate the bandgaps of these two materials. The evaluated bandgaps are consistent with the results from ultraviolet-visible (UV-vis) absorption spectrum. The bandgap of MAPbI3 decreases with temperature above 95 °C, whereas that of FAPbI3 first increases with temperature from 95 °C to 107 °C and then decreases with temperature above 107 °C.

This paper presents analysis and optimizations for High Performance Conjugate Gradient benchmark (HPCG) on the Sunway many-core processor. For modern multi-core and many-core processors, HPCG always presents a poor performance and under-utilizes computation resource because of its low arithmetic intensity and fine-grain parallelism. We apply two conventional methods to parallel Gauss-Seidel smoother the most time consumer kernel in HPCG, including Level-Scheduling (LS) and Multi-Coloring (MC). These strategies are effective and achieve 1.54x and 5.52x performance improvement. For overcoming the poor locality for MC and limited parallelism for LS, we propose a novel Hierarchical Grid (HG) algorithm and our algorithmic and architecture-aware optimizations achieve an aggregated performance of 3.54 Gflops, which is around 0.475% of the peak performance and 15.4x higher than reference on the single core-group of SW26010 processor. With MPI parallelize, we balance the parallelism, pre-processing, convergence rate and communication overheads, we achieved 192 TFlops (70% parallelization efficiency) when scaling to 81920 CGs (5,324,800 cores) on Sunway Taihulight System. Moreover, we analyze the adaptability of our parallel method and optimization strategies and summarize several key points when refactoring and optimizing HPC applications on the Sunway heterogeneous many-core processor.

Silver nanoprisms (AgNPs) affect the performance of organic solar cells (OSCs) in different ways depending on their positions in the device. To investigate this issue, we incorporate AgNPs in different positions of OSCs and compare their performance. The power conversion efficiency (PCE) is improved by 23.60% to 3.98% when the AgNPs are incorporated in front of the active layer. On the other hand, when AgNPs are incorporated in the back of the active layer, the short-circuit current density (J SC) is improved by 17.44% to 10.84 mA/cm2. However, if AgNPs are incorporated in the active layer, both open-circuit voltage (V OC) and J SC are decreased. We discuss the position effect on the device performance, clarify the absorption shadow and exciton recombination caused by AgNPs, and finally indicate that the optimal position of plasmonic AgNPs is in front of the active layer.

We fabricated silver nanoprisms (AgNPs) on ITO by immersing ITO substrates in AgNPs solution for a series of immersion times. The amount of adsorbed AgNPs increased with immersion time. The AgNPs showed plasmonic absorption in the range of 400 – 600 nm and were used in organic solar cells (OSCs). The device performance was the best when the immersion time was 30 min, corresponding to AgNPs coverage of 68%. Under this condition, the device showed short-circuit current density (JSC) of 10.10 mA cm-2(18% improved), and power conversion efficiency (PCE) of 3.88% (23% improved).

Stability is of paramount importance in organic semiconductor devices, especially in organic solar cells (OSCs). Serious degradation in air limits wide applications of these flexible, light-weight and low-cost power-generation devices. Studying the stability of organic solar cells will help us understand degradation mechanisms and further improve the stability of these devices. There are many investigations into the efficiency and stability of OSCs. The efficiency and stability of devices even of the same photoactive materials are scattered in different papers. In particular, the extrinsic degradation that mainly occurs near the interface between the organic layer and the cathode is a major stability concern. In the past few years, researchers have developed many new cathodes and cathode buffer layers, some of which have astonishingly improved the stability of OSCs. In this review article, we discuss the recent developments of these materials and summarize recent progresses in the study of the degradation/stability of OSCs, with emphasis on the extrinsic degradation/stability that is related to the intrusion of oxygen and water. The review provides detailed insight into the current status of research on the stability of OSCs and seeks to facilitate the development of highly-efficient OSCs with enhanced stability.

Compared to solution approaches, vapor–solid reaction (VSR) processes have advantages in fabricating multistack perovskite solar cells (PSCs). Here, we modeled the formation of perovskite by VSR through combined experimental and simulation analyses. Scanning electron microscopy and X-ray diffraction characterizations at the turning points of the experimentally observed W-shaped in situ electrical resistance curves suggest a diffusion-dominated growth mechanism. Simulation reveals that to obtain stoichiometric perovskite films, we need to accelerate the diffusion of methylammonium iodide (MAI) in PbI2 and slowly supply MAI vapor. Experiments were conducted to verify the diffusion-dominated formation model. This work sheds important light on rational design of new VSR approaches to fabricate more effective PSCs.

Since Penn et al. first discovered the oriented attachment growth of crystals, the oriented attachment mechanism has now become a major research focus in the crystal field, and extensive efforts have been carried out over the past decade to systematically investigate the growth mechanism and the statistical kinetic models. However, most of the work mainly focuses on the experimental results on the oriented attachment growth. In contrast to the previous reviews, our review provides an overview of the recent theoretical advances in oriented attachment kinetics combined with experimental evidences. After a brief introduction to the van der Waals interaction and Coulombic interaction in a colloidal system, the correlation between the kinetic models of oriented attachment growth and the interactions is then our focus. The impact of in situ experimental observation techniques on the study of oriented attachment growth is examined with insightful examples. In addition, the advances in theoretical simulations mainly investigating the thermodynamic origin of these interactions at the atomic level are reviewed. This review seeks to understand the oriented attachment crystal growth from a kinetic point of view and provide a quantitative methodology to rationally design an oriented attachment system with pre-evaluated crystal growth parameters.

Taking the advantage of nanomaterials to protect the environment and avoiding the side effect need a fundamental understanding of the growth mechanism of the nanomaterials. Here, the van der Waals interaction between a nanoparticle and a nanotube in the oriented-attachment growth of nanotubes is quantitatively evaluated for the first time. In particular, the correlation between van der Waals interaction and the growth parameters is investigated in depth. Our work opens up the opportunity of studying the important interparticle interactions in the oriented attachment growth of nanotubes.

We demonstrate an efficient approach by using a new type of ionic liquid, 1-ethylpyridinium chloride (1-EC) with a relative low melting point of 100 °C, to control the morphological growth of CH3NH3PbI3 during the one-step deposition method for preparing efficient planar heterojunction perovskite solar cells, leading to a continuous and dense morphology without large voids and pinhole. With the optimized concentration, 11.8% power conversion efficiency was obtained. Our work paves the way to optimize the morphology of perovskite by adopting a new kind of ionic liquid as additive. This method has great benefits for exploiting new technologies for large-scale perovskite solar cells on the flexible substrate.

The in-air stability of unencapsulated small-molecular organic solar cells (OSCs) was improved by 3 orders of magnitude by using a Ti (inner)–Al (exposed to air) bilayer cathode, in comparison with that of Al-monolayer-cathode OSCs. The stability of the unencapsulated bilayer-cathode OSCs (AlTi-OSCs) was comparable to that of encapsulated ones. These devices were degradation-free in air for more than 3 months. Under weathering tests, the unencapsulated AlTi-OSCs recovered their performance in the idle periods. This provides us with an effective and economical approach to prevent degradation of the extensively studied aluminum-cathode OSCs.