45 results about "High-throughput computing" patented technology

The invention discloses a multi-scale simulation method for mechanical behaviors of a multi-phase composite material, and belongs to the field of multidisciplinary intersection of computational material science, analog simulation, high-throughput calculation and the like. Various physical property parameters needed by finite element simulation of the macroscopic scale can be obtained through calculation of the first principle of the nanoscopic scale, molecular dynamics simulation of the microcosmic scale and thermodynamic calculation of the mesoscopic scale, and the physical property parameters comprise elastic-plastic physical parameters of all phases in the composite material at different temperatures and different grain sizes and the like. A real material tissue structure is obtained byutilizing a focused ion beam experiment and image processing introduction. Through parameter coupling and parameter transmission among scale calculation results and combination with the structure ofthe material, mechanical behaviors such as stress-strain relationship, stress distribution and evolution and plastic deformation of the multi-phase composite material under complex stress and different temperatures can be simulated.

Embodiments of the present invention manage workloads in a high-throughput computing environment for a hybrid processing system. A set of high-throughput computing service level agreements (SLAs) is retrieved. The set of SLAs is associated with a hybrid processing system including a server system and a set of accelerator systems, where each system has a different architecture. A first set of data-parallel workload tasks scheduled on the server system and a second set of data-parallel workload tasks scheduled with the set of accelerator systems are identified. At least a portion of one of the first set of data-parallel workload tasks and the second set of data-parallel workload tasks is dynamically rescheduled on a second one of the server system and the set of accelerator systems. The dynamically rescheduling substantially satisfies the set of high-throughput computing SLAs.

The invention belongs to the technical field of material calculation, and provides a high-throughput calculation screening method for a doped energy material, which is realized through four modules. The method specifically comprises the steps of obtaining a crystal structure information file of an intrinsic material through an intrinsic material structure extraction module; processing the crystalstructure information file by a doped structure construction module to obtain a doped structure information file; reading the structure information file through a doped material structure optimizationand property calculation module; generating structure optimization-property calculation work flow in batches and performing high-throughput calculation; and finally storing calculation results of energy, electronic properties and the like in a private database through a material database importing module. The purpose of high-throughput calculation screening can be efficiently achieved; the working efficiency of material screening is improved; and relatively high universality and expandability are achieved.

The invention discloses a system and method for calculating the surface adsorption diversity of a two-dimensional material under experimental condition. The method comprises analyzing the experimental environment of the two-dimensional material, establishing a relationship between the formation energy of functional groups of the two-dimensional material and the corresponding chemical potential, then enumerating all possible surface configurations, storing each structure with different surface functional group proportions and arrangements, carrying out relaxation on the structures, and calculating corresponding formation energy. Through high-throughput calculation, a formation energy database of the two-dimensional material under the experimental condition is established, and the proportion of the functional group with the lowest formation energy within the same chemical potential range under the experimental condition and the corresponding configuration are screened out. According to the system and the method, a method for unifying formation energy of different functional groups into a calculation formula is realized, and the formation energy of the functional groups on the surface of the two-dimensional material under the experimental condition is calculated by applying a high-throughput technology; the surface configuration of the two-dimensional material with excellent thermodynamic stability can be more accurately screened by the method, and beneficial help is provided for new material design.

This disclosure describes methods and systems for high-throughput computations in a fully-connected deep neural network. Specifically, a hardware-based deep neural network architecture including a set of parallel node processors is used to process node value transition between layers of the deep neural network, which usually involves a large-scale matrix multiplication. The set of parallel node processors are configured to decompose the large-scale matrix multiplication into sub-matrix multiplications with smaller sizes and thus reducing the hardware-complexity and making feasible direct implementation in hardware. With this implementation deep neural network may achieve a very high throughput and can handle a large number of processing layers.

The invention describes an explicit solvent all-atom molecular dynamics methodology (SILCS: Site Identification by Ligand Competitive Saturation) that uses no small aliphatic and aromatic molecules plus water molecules to map the affinity pattern of a large molecule for hydrophobic groups, aromatic groups, hydrogen bond donors, and hydrogen bond acceptors. By simultaneously incorporating ligands representative of all these functionalities, the method is an in silico free energy-based competition assay that generates three-dimensional probability maps of fragment binding (FragMaps) indicating favorable fragment:large molecule interactions. The FragMaps may be used to qualitatively inform the design of small-molecule ligands or as scoring grids for high-throughput in silico docking that incorporates both an atomic-level description of solvation and the large molecule's flexibility.

The invention discloses a data collection method and device, electronic equipment and a readable storage medium. The method comprises the following steps: constructing an input file which conforms to a material structure and is suitable for material simulation software according to material attribute parameters of original data in a target database; based on a preset convergence standard and the input file, performing optimization processing on the material structure through high-throughput calculation to obtain a target material structure; and processing the target material structure by using the target calculation method for calculating the to-be-demanded physical quantity to obtain the instance data of the to-be-demanded physical quantity, so that a large amount of instance data meeting the material informatics reality demand can be obtained.

Disclosed is a method of efficiently allocating resources in a cloud computing system of the High-Throughput Computing (HTC) field. In more detail, a cloud resource allocation model is proposed in which accumulated job histories are analyzed, a correlation between a performance factor and the job properties is extracted, a factor which is greatly influential in a task performance is identified based on the result of the analyzing and the extracting and is applied to subsequent resource allocations, and reanalysis is performed adaptively.

The invention discloses a workflow modeling method suitable for multi-scale high-throughput material calculation, particularly relates to the technical field of computers, and aims to deal with challenges of automatic construction of a microcosmic to macroscopic multi-scale processing process and efficient execution of a large number of high-concurrency calculation tasks in a material design process. The invention provides a workflow flexible configuration and hierarchical modeling method suitable for multi-scale high-throughput material calculation. Packaging of a high-concurrency computing task and flexible expression of a typical process control structure are achieved through a high-throughput computing component model; and a multi-level resource scheduling model oriented to material high-throughput calculation is established, and unified delivery, monitoring and result access of high-throughput scheduling tasks in a high-performance calculation environment are supported. The method is innovative in automatic control of the service flow facing high-throughput calculation, the multi-scale and high-throughput material calculation service workflow can be constructed through a visual means, and execution of the calculation process and transmission of calculation data are automatically achieved.

Embodiments of the present invention provide high-throughput computing in a hybrid processing system. A set of high-throughput computing service level agreements (SLAs) is analyzed. The set of high-throughput computing SLAs are associated with a hybrid processing system. The hybrid processing system includes at least one server system that includes a first computing architecture and a set of accelerator systems each including a second computing architecture that is different from the first computing architecture. A first set of resources at the server system and a second set of resources at the set of accelerator systems are monitored. A set of data-parallel workload tasks is dynamically scheduled across at least one resource in the first set of resources and at least one resource in the second set of resources. The dynamic scheduling of the set of data-parallel workload tasks substantially satisfies the set of high-throughput computing SLAs.

The invention discloses a high-throughput calculation method of surface properties based on automatic modeling technology, which belongs to the field of materials. Firstly, the structure file of the crystal Q to be studied is obtained, and the crystal Q is converted into a unit cell by using the structure file, and the symmetry is analyzed. Then convert the initial base vector of crystal Q according to the Miller index specified by the user, and obtain the new base vector matrix corresponding to the new crystal structure G. Determine the atomic coordinates corresponding to the new base vector matrix, and fill the new base vector with the atoms of crystal Q, Construct a surface model of crystal G. Finally, the surface energy and work function of the new surface model of crystal G constructed by DFT calculation. For several other crystals with different materials and different surface structures, high-throughput batch calculations are performed separately to obtain their respective surface energies and work functions to form a database of material surface properties. The invention realizes the parallel calculation of the surface properties of different surface structures and improves the calculation efficiency.