Over the past decade, Graphics Processing Units (GPUs) have revolutionized high-performance computing, playing pivotal roles in advancing fields like IoT, autonomous vehicles, and exascale computing. Despite these adv...Over the past decade, Graphics Processing Units (GPUs) have revolutionized high-performance computing, playing pivotal roles in advancing fields like IoT, autonomous vehicles, and exascale computing. Despite these advancements, efficiently programming GPUs remains a daunting challenge, often relying on trial-and-error optimization methods. This paper introduces an optimization technique for CUDA programs through a novel Data Layout strategy, aimed at restructuring memory data arrangement to significantly enhance data access locality. Focusing on the dynamic programming algorithm for chained matrix multiplication—a critical operation across various domains including artificial intelligence (AI), high-performance computing (HPC), and the Internet of Things (IoT)—this technique facilitates more localized access. We specifically illustrate the importance of efficient matrix multiplication in these areas, underscoring the technique’s broader applicability and its potential to address some of the most pressing computational challenges in GPU-accelerated applications. Our findings reveal a remarkable reduction in memory consumption and a substantial 50% decrease in execution time for CUDA programs utilizing this technique, thereby setting a new benchmark for optimization in GPU computing.展开更多
Alumina(Al_(2)O_(3))is widely used in the chemical industry as the catalyst and support due to its high specific surface area,abundant pore size distribution and chemical stability.However,the occurrence of hydration ...Alumina(Al_(2)O_(3))is widely used in the chemical industry as the catalyst and support due to its high specific surface area,abundant pore size distribution and chemical stability.However,the occurrence of hydration in water environment,result in outstanding decrease in specific surface area and collapse of pore structure.In this work,dodecyl phosphoric acid(PA)is used to modify the surface of Al_(2)O_(3)to obtain a series of hydrophobic material(Al_(2)O_(3)-PA).Based on XPS and NMR analysis,PA is chemically bonded on Al_(2)O_(3)to form PAOAAl bond.Furthermore,BET and WCA results display that Al_(2)O_(3)-1PA exhibits excellent the hydrophobicity and hydrothermal stability while maintains the pore structure.Take it as the substrate to support the Pd nanoparticles,the as-prepared Pd/Al_(2)O_(3)-PA shows the superior catalytic performance in the hydrogenation of phenol and anthraquinone relative to Pd/Al_(2)O_(3),indicating the accessibility of Pd sites after PA modification.Especially,the significantly enhanced stability is also obtained in four cycles for aqueous phenol hydrogenation.This can be ascribed that the PA modification inhibits the aggregation of Pd nanoparticles and the products adhesion in the reaction process.The extension of PA coatings to monolithic catalysts could expand their current capabilities in industrial applications and warrants ongoing investigation.展开更多
Electric-field control of perpendicular magnetic anisotropy(PMA) is a feasible way to manipulate perpendicular magnetization,which is of great importance for realizing energy-efficient spintronics.Here,we propose a no...Electric-field control of perpendicular magnetic anisotropy(PMA) is a feasible way to manipulate perpendicular magnetization,which is of great importance for realizing energy-efficient spintronics.Here,we propose a novel approach to accomplish this task at room temperature by resistive switching(RS) via electrochemical metallization(ECM) in a device with the stack of Si/SiO_(2)/Ta/Pt/Ag/Mn-doped ZnO(MZO)/Pt/Co/Pt/ITO.By applying certain voltages,the device could be set at high-resistance-state(HRS) and low-resistance-state(LRS),accompanied with a larger and a smaller coercivity(H_(C)),respectively,which demonstrates a nonvolatile E-field control of PMA.Based on our previous studies and the present control experiments,the electric modulation of PMA can be briefly explained as follows.At LRS,the Ag conductive filaments form and pass through the entire MZO layer and finally reach the Pt/Co/Pt sandwich,leading to weakening of PMA and reduction of H_(C).In contrast,at HRS,most of the Ag filaments dissolve and leave away from the Pt/Co/Pt sandwich,causing partial recovery of PMA and an increase of H_(C).This work provides a new clue to designing low-power spintronic devices based on PMA films.展开更多
文摘Over the past decade, Graphics Processing Units (GPUs) have revolutionized high-performance computing, playing pivotal roles in advancing fields like IoT, autonomous vehicles, and exascale computing. Despite these advancements, efficiently programming GPUs remains a daunting challenge, often relying on trial-and-error optimization methods. This paper introduces an optimization technique for CUDA programs through a novel Data Layout strategy, aimed at restructuring memory data arrangement to significantly enhance data access locality. Focusing on the dynamic programming algorithm for chained matrix multiplication—a critical operation across various domains including artificial intelligence (AI), high-performance computing (HPC), and the Internet of Things (IoT)—this technique facilitates more localized access. We specifically illustrate the importance of efficient matrix multiplication in these areas, underscoring the technique’s broader applicability and its potential to address some of the most pressing computational challenges in GPU-accelerated applications. Our findings reveal a remarkable reduction in memory consumption and a substantial 50% decrease in execution time for CUDA programs utilizing this technique, thereby setting a new benchmark for optimization in GPU computing.
基金supported by National Key Research&Development Program of China(2021YFB3801600)Fundamental Research Funds for the Central University(buctrc201921,JD2223,12060093063)Innovative Achievement Commercialization Service-Platform of Industrial Catalysis(2019-00900-2-1).
文摘Alumina(Al_(2)O_(3))is widely used in the chemical industry as the catalyst and support due to its high specific surface area,abundant pore size distribution and chemical stability.However,the occurrence of hydration in water environment,result in outstanding decrease in specific surface area and collapse of pore structure.In this work,dodecyl phosphoric acid(PA)is used to modify the surface of Al_(2)O_(3)to obtain a series of hydrophobic material(Al_(2)O_(3)-PA).Based on XPS and NMR analysis,PA is chemically bonded on Al_(2)O_(3)to form PAOAAl bond.Furthermore,BET and WCA results display that Al_(2)O_(3)-1PA exhibits excellent the hydrophobicity and hydrothermal stability while maintains the pore structure.Take it as the substrate to support the Pd nanoparticles,the as-prepared Pd/Al_(2)O_(3)-PA shows the superior catalytic performance in the hydrogenation of phenol and anthraquinone relative to Pd/Al_(2)O_(3),indicating the accessibility of Pd sites after PA modification.Especially,the significantly enhanced stability is also obtained in four cycles for aqueous phenol hydrogenation.This can be ascribed that the PA modification inhibits the aggregation of Pd nanoparticles and the products adhesion in the reaction process.The extension of PA coatings to monolithic catalysts could expand their current capabilities in industrial applications and warrants ongoing investigation.
基金Project supported by the National Key Research and Development Program of China (Grant No. 2022YFA1403602)the National Natural Science Foundation of China (Grant Nos. 51971109, 52025012, and 52001169)。
文摘Electric-field control of perpendicular magnetic anisotropy(PMA) is a feasible way to manipulate perpendicular magnetization,which is of great importance for realizing energy-efficient spintronics.Here,we propose a novel approach to accomplish this task at room temperature by resistive switching(RS) via electrochemical metallization(ECM) in a device with the stack of Si/SiO_(2)/Ta/Pt/Ag/Mn-doped ZnO(MZO)/Pt/Co/Pt/ITO.By applying certain voltages,the device could be set at high-resistance-state(HRS) and low-resistance-state(LRS),accompanied with a larger and a smaller coercivity(H_(C)),respectively,which demonstrates a nonvolatile E-field control of PMA.Based on our previous studies and the present control experiments,the electric modulation of PMA can be briefly explained as follows.At LRS,the Ag conductive filaments form and pass through the entire MZO layer and finally reach the Pt/Co/Pt sandwich,leading to weakening of PMA and reduction of H_(C).In contrast,at HRS,most of the Ag filaments dissolve and leave away from the Pt/Co/Pt sandwich,causing partial recovery of PMA and an increase of H_(C).This work provides a new clue to designing low-power spintronic devices based on PMA films.