37 results about "Program optimization" patented technology

The invention relates to the technical field of industrial internet, in particular to an automatic scheduling system for multi-chip parallel burning, which is characterized in that a program optimization and deployment module dynamically adapts a basic program according to hardware resources of a target equipment cluster and process parameters of chips, generates an efficient optimized burning program, and sends the efficient optimized burning program to a server; verifying the digital digest through an integrated block chain network and performing safe deployment; the intelligent scheduling decision module utilizes a dynamic digital twinborn model to perceive the equipment state and deployed optimization program characteristics in real time, and intelligently generates a global burning task optimal allocation scheme and a scheduling instruction; the cooperative control module compares actual efficiency with expected efficiency in real time, quantifies deviation and then feeds back the deviation to the reinforcement learning scheduling engine, and the engine automatically adjusts and continuously iteratively optimizes the overall scheduling strategy. According to the invention, the scheduling efficiency of multi-chip parallel burning can be improved.

The invention provides a GPU program optimization method in a parallel environment, a plurality of target models are configured in a GPU cluster for parallel training, and the method comprises the following steps: performing strategy configuration and memory bottleneck pre-judgment on parallel training configuration of the target models; carrying out throughput optimization through parallel strategy combination and instruction-level performance monitoring by utilizing a pre-judgment result; kernel re-optimization is carried out on an instruction level bottleneck appearing in the optimization process, so that the model training efficiency and the GPU instruction execution efficiency are synchronously improved.

The embodiment of the invention provides an application program optimization method and device, electronic equipment and a storage medium. The method comprises the following steps: collecting resource calling data of an application program on user equipment; identifying a to-be-optimized resource set based on the resource calling data; executing optimization processing on the to-be-optimized resource set to obtain an optimized target application program; wherein the optimization processing comprises resource removal or resource relocation. Compared with a traditional static analysis method, the method has the advantages that through a data analysis mode driven by user behaviors, the risk of function missing possibly caused by blind deletion of resources is avoided, and more accurate and safer application program optimization is realized. Besides, for the resources which are used at low frequency but are still necessary to be reserved, the optimization of the initial packet volume is realized while the functional integrity is maintained through a processing mode of resource relocation instead of simple deletion, and the relationship between the user experience and the resource efficiency is balanced.

The invention relates to a digital twin virtual debugging method for jet printing manufacturing control optimization, and belongs to the technical field related to intelligent manufacturing and industrial control. Comprising the following steps: configuring a control logic program for each equipment sub-module in a jet printing manufacturing equipment twin model through an equipment master controller, and writing the control logic program into a corresponding virtual controller; establishing communication connection between each virtual controller and the twin model, verifying a control logic program in the virtual controller, and verifying and optimizing the control logic program through general control; and inputting each optimized control logic program into the entity controller of the corresponding equipment sub-module, sending a control instruction to the entity controller corresponding to the equipment sub-module of which the current motion is to be verified through general control, driving the equipment sub-module to execute an action, optimizing the control logic program in the current entity controller through the general control, and verifying the current motion of the equipment sub-module. And through multi-round program optimization, a control program of the jet printing manufacturing equipment is obtained. The debugging time and cost of the ink-jet printing equipment can be reduced.

The invention discloses a program compiling device and method, and belongs to the technical field of computers. The device comprises a receiving module used for receiving a log collected by an instrumentation program in the running process of a target program, the instrumentation program is a part of the target program, and the log comprises popularity data of at least one function in the target program; and the compiling module is used for performing biased compiling on the target program based on the popularity data of the at least one function to obtain an updated target program, and the biased compiling comprises compiling of a biased performance optimization operation or compiling of a biased code area compression operation. In this way, the program can be optimized and upgraded better, and user requirements are met.

The invention provides a dynamic instruction replacement method and device, equipment and a storage medium, and the method comprises the steps: determining a to-be-optimized instruction group according to instruction operation information of a plurality of instruction groups, and obtaining a replacement instruction group corresponding to the to-be-optimized instruction group, the performance of the replacement instruction group is better than that of the to-be-optimized instruction group, and the performance of the replacement instruction group is better than that of the to-be-optimized instruction group; and storing the replacement instruction group to a pre-occupied virtual address field, if the to-be-processed current instruction stream comprises the to-be-optimized instruction group, skipping to the virtual address field to execute the replacement instruction group when executing to the to-be-optimized instruction group in the current instruction stream, and skipping to the current instruction stream after the execution of the replacement instruction group is finished, and executing instructions behind the instruction group to be optimized. According to the method and the device, function-level instruction replacement is realized, the program optimization cost is reduced, and the program running efficiency can be effectively improved.

The invention provides a particle program optimization method based on a Shenwei architecture. The method comprises the following steps: analyzing program hotspots, finding and calculating a hotspot function, and optimizing the function by using five means of particle control, step number parameter control, data classification transmission, function expansion and function inline by utilizing the advantages of an SW26010pro architecture; the execution efficiency of the particle program is improved, the computing performance of the Shenwei architecture is fully played, and rapid and effective operation of the large-scale particle computing program is achieved.

The application relates to an automatic program optimization method and device based on a flow network generation model and a storage medium, wherein the method comprises the following steps: S1, an initial tensor program is acquired, a plurality of candidate tensor programs with the same logical function are obtained through calculation graph extraction, subgraph segmentation and transformation based on the acquired initial tensor program, and a data set composed of a plurality of samples is constructed based on the obtained candidate tensor programs, wherein the sample comprises a binary tuple composed of a calculation subgraph and a candidate tensor program, and the hardware execution time corresponding to the binary tuple; S2, a plurality of samples are selected from the data set, and a GFlowNet sampling model is trained offline; and S3, the trained GFlowNet sampling model is used to optimize a program to be optimized. Compared with the prior art, the GFlowNet adopts a non-iterative probabilistic sampling strategy, generates a plurality of high-performance programs, apportions the sampling overhead from one mode to another mode, and accelerates the convergence speed.

The application relates to the technical field of program optimization, and particularly discloses a small program code volume reduction method and device, a storage medium and a computer device. The method comprises the following steps: in response to a small program code volume reduction instruction, determining a target small program to be subjected to volume reduction; acquiring a global configuration file of the target small program, and parsing declared valid page paths from the global configuration file; scanning an engineering directory corresponding to the target small program, and identifying a plurality of page directories contained in the engineering directory; comparing each page directory with the declared valid page paths respectively, so as to screen out redundant page directories not contained in the valid page paths from the plurality of page directories; and deleting the redundant page directories and all files under the redundant page directories, so as to obtain a target small program after code reduction.

The invention discloses a converged media program list arrangement system and a generation method, and belongs to the technical field of media content management and intelligent scheduling. The system comprises a content acquisition module, a feature extraction module, a user portrait and platform model module, a program optimization and arrangement module, an intelligent conflict detection module, a self-adaptive generation module and a feedback optimization module. Unified management of program resources is realized through multi-source content collection and semantic tagging analysis; a multi-objective optimization model is constructed based on the user portrait and the platform strategy, and program lists adapted to different terminals are automatically generated; and dynamically adjusting a programming result by utilizing a conflict detection and self-learning feedback mechanism. According to the method, the intelligent level of convergence media program management can be remarkably improved, the cross-platform content distribution efficiency is optimized, and automatic generation and accurate pushing of the program list are realized.

Embodiments of the invention disclose an instruction fusion method and apparatus, an electronic device, a storage medium and a program product. The method comprises the steps of obtaining a to-be-optimized program; when it is determined that all instructions in the to-be-optimized program belong to the same basic block and the instructions with the dependency relationship in the to-be-optimized program are adjacently arranged, all continuous instruction sequences in the to-be-optimized program are obtained, to-be-fused instruction sequences are determined in all the continuous instruction sequences, and each set of to-be-fused instruction sequences is replaced with a corresponding fusion instruction; when it is determined that all the instructions in the to-be-optimized program do not belong to the same basic block or the instructions with the dependency relationship in the to-be-optimized program are not adjacently arranged, obtaining candidate instructions in the to-be-optimized program, determining to-be-fused instructions in the candidate instructions according to the fixed value-use chains corresponding to the candidate instructions, and fusing the to-be-fused instructions in the to-be-optimized program according to the determined to-be-fused instructions in the to-be-optimized program. And the target instruction sequence corresponding to each instruction to be fused is replaced with the corresponding fusion instruction, so that the program optimization efficiency and the program optimization capability are balanced.

The invention relates to an automatic program optimization method and system for synthesizing and eliminating an intermediate data structure by utilizing an induction program. The method comprises the following steps: acquiring a low-efficiency program which is input by a user and contains an intermediate data structure, and appointing the intermediate data structure which needs to be eliminated; a specified intermediate data structure is replaced by a scalar type, and related program fragments in the original program are rewritten using efficient program fragments of constant time complexity. The invention provides a novel induction program synthesis method aiming at the problem of eliminating an intermediate data structure, the induction program synthesis problem can be efficiently decomposed into simpler subtasks, so that the confronted efficiency problem is overcome, and the method has stronger expression ability than an existing fusion method, and can be widely applied to the field of data fusion. And the time for a programmer to carry out efficient system development can be greatly reduced.

According to an implementation, a computer-implemented method to determine an optimal configuration of a benchmark program is proposed. The method includes iteratively selecting a unique parameter set from a parameter space, each parameter set corresponding to a unique configuration of the benchmark program; running the benchmark program as configured for each parameter set at least once; determining an evaluation score for each parameter set from one or more runs using an evaluation heuristic; recording an optimal parameter set having a best evaluation score; and configuring the benchmark program based on the optimal parameter set.

The invention relates to an automatic program optimization method and device for an application program, computer equipment and a storage medium. The automatic program optimization method comprises the following steps: acquiring execution information and initial performance data of each instruction in a plurality of instructions when the application program runs; determining a hotspot instruction according to the execution information of each instruction in the plurality of instructions; obtaining an instruction type of the hotspot instruction, judging an instruction optimization mode corresponding to the instruction type, and if the instruction optimization mode is judged to be based on the configuration template, obtaining a matched target optimization template from a preset template library according to the instruction type; performing code optimization on the hotspot instruction according to an optimization strategy and a code mode configured in the target optimization template to generate an optimization code of the hotspot instruction; and performing program optimization on the application program according to the optimization code and the initial performance data of the application program. According to the method, the accuracy, the automation degree and the efficiency of software performance optimization are improved.

The invention discloses a real-time monitoring system for a dry net cleaning vacuum valve. The real-time monitoring system comprises a detection module, an alarm module, a feedback module, a program optimization module, a bracket structure and a data storage module, according to the invention, a dry net cleaning program is optimized by using off-machine limiting and self-processing of the fixing bracket, and the real-time use condition of the vacuum valve is efficiently detected without cost. And the conditions of wastage reduction and dry screen printing caused by blockage of the vacuum valve are effectively avoided, and the quality of finished paper is improved. Meanwhile, by adding real-time feedback and alarm functions on a DCS picture, the abnormity of the vacuum valve can be found in time, the failure rate is reduced, the workload of personnel is reduced, and the market competitiveness of a company is further improved. In addition, the detachable support facilitates later maintenance, the troubleshooting time is shortened, and the data storage module facilitates later analysis and summarization of equipment operation conditions.

The invention discloses a tensor program functionalization method, tensor program functionalization equipment and a tensor program product. The method comprises the following steps: acquiring a tensor program, and constructing a graph-level intermediate representation program of the tensor program; performing mutation rewriting and tensor version relation labeling on the graph-level intermediate representation program based on a memory dependency graph and an immutable operator of the graph-level intermediate representation program to obtain a rewritten graph-level intermediate representation program; the immutable operator is used for replacing a mutation operator in a graph-level intermediate representation program, and a new version tensor output by the immutable operator and an input source tensor have independent memories; and performing program optimization on the rewritten graph-level intermediate representation program to obtain a functional tensor intermediate representation program based on the static single assignment. The Graph-Level IR program is combined with the static single assignment SSA, so that the side effect of tensor mutation in the Graph-Level IR program is thoroughly eliminated, and the optimization performance of a deep learning compiler on the tensor program is improved.

Various embodiments of the present technology generally relate to systems and methods for providing a framework for optimizing configuration settings of application instances. In certain embodiments, a method may comprise operating an optimizer service to implement an application optimizer process to improve performance of an application instance. The process may include receiving a plurality of checks from an application development system for the application instance, the plurality of checks including scans and fixes for configuration settings of the application instance. The process may further include executing the scans on the application instance, determining a selection of fixes configured to improve the performance of the application instance in response to a result of the scans, providing a notification to the application instance recommending user implementation of the selection of fixes, and providing analytics data corresponding to the user implementation of the selection of fixes to the application development system.

The embodiment of the invention discloses a program tuning code marking method and device, a terminal and a medium, and the method comprises the steps: marking an optimization code in a code editor; obtaining a first program performance information file corresponding to the to-be-optimized code; obtaining a second program performance information file corresponding to the modified and optimized code; comparing the first program performance information file with the second program performance information file to obtain a difference function between the first program performance information file and the second program performance information file; and searching the difference function by utilizing a code editor difference tool, and determining a modified source code according to the mark. The change condition of the program performance in the whole source code modification process can be recorded, and a developer can carry out optimization, backtracking and iteration on the program performance conveniently.

According to an implementation, a computer-implemented method to determine an optimal configuration of a benchmark program is proposed. The method includes iteratively selecting a unique parameter set from a parameter space, each parameter set corresponding to a unique configuration of the benchmark program; running the benchmark program as configured for each parameter set at least once; determining an evaluation score for each parameter set from one or more runs using an evaluation heuristic; recording an optimal parameter set having a best evaluation score; and configuring the benchmark program based on the optimal parameter set.

The invention discloses a program optimization method and device, a program product and electronic equipment, and relates to the field of financial science and technology, the method comprises the steps that static characteristics and dynamic characteristics of a target program are collected, the static characteristics are at least used for representing control flow information, data dependence information and function call information of the target program, and the dynamic characteristics are at least used for representing control flow information of the target program; the dynamic characteristics are at least used for representing function execution information in the target program; performing feature fusion on the static features and the dynamic features to obtain target features of the target program; based on the target feature, determining a program label of the target program, the program label being used for representing a program type of the target program; and performing performance optimization on the target program based on the target feature through a preset model corresponding to the program tag. The technical problem that in the prior art, the program optimization effect is poor due to the fact that the program is optimized only based on the static scanning result of the program is solved.

The invention provides a runtime program optimization method and device oriented to pipeline parallelism, and the method comprises the steps: recognizing a time window in a heterogeneous parallel computing process, the time window being an idle time period generated between accelerator equipment and a host processor due to task dependence or data synchronization; within the time window, collecting runtime information in a computing process from an accelerator device; dynamically selecting a compilation optimization strategy according to the characteristics of the time window and the runtime information; and based on the selected compilation optimization strategy, executing compilation optimization in the time window and generating an optimization code. According to the method, the idle time window in the calculation process can be effectively identified and utilized, and real-time performance optimization is achieved.

The invention provides a program optimization method and device, electronic equipment and a storage medium, and relates to the technical field of computers, in particular to the field of program page loading. According to the specific implementation scheme, the method comprises the following steps: counting completion time consumption and midway exit rate of each initialization loading operation in a plurality of serial initialization loading operations in a page loading process of a first program; and under the condition that the completion time consumption and the midway exit rate of each initialization loading operation do not meet preset conditions, combining and optimizing the subprograms corresponding to the plurality of initialization loading operations in the first program to obtain a second program, the second program comprises a subprogram corresponding to a target initialization loading operation obtained by combining and optimizing a plurality of initialization loading operations.

The present application belongs to the field of computer system construction, and particularly relates to a CPU bottleneck analysis model based on performance counter. The present application solves the technical problem that in the prior art, the average delay evaluation of the failure event required for performance evaluation is limited, and the pipeline technology of modern processors makes the failure delay difficult to be fully exposed, thereby affecting the CPI stack to discover the performance bottleneck of the CPU. In the present application, each clock cycle pause of the program instruction is assigned to the corresponding failure event, the relationship between the instruction pause and the failure event is obtained based on the interval analysis model by extracting the key performance signal in the program execution process, so as to construct the performance model and guide the program optimization and processor hardware iteration. The algorithm for constructing the CPI stack by using the performance counter can be deployed on the prototype verification platform and accurately evaluate the processor performance.

The application discloses a GPU resource-aware matrix multiplication parallel performance analysis model construction method, and relates to the technical field of high-performance computing, which is based on the Roofline principle, combines a bandwidth model, a resource load model and an instruction dependency model to construct a performance analysis model, and is successfully applied to a matrix multiplication scene to realize parallel performance quantification under resource-aware matrix multiplication application.The application can realize resource awareness without platform limitation, can distinguish parameter settings for maximizing resource use and optimizing performance in the matrix multiplication scene through the performance analysis model, and lays a foundation for subsequent parallel program optimization work; the hardware parameters used by the application are completely based on publicly available GPU hardware parameters, so that non-GPU professionals can also use the model to evaluate the performance of programs.

The application provides a GPU program optimization method in a parallel environment. A plurality of target models are configured in a GPU cluster for parallel training. The method comprises: performing strategy configuration and memory bottleneck prediction on the parallel training configuration of the target model; using the prediction result, performing throughput optimization through parallel strategy combination and instruction-level performance monitoring; and performing kernel re-optimization on the instruction-level bottleneck occurring in the optimization process, so that the model training and GPU instruction execution efficiency are simultaneously improved.

Machine learning using a frequency image is accurately executed. An information processing apparatus (100) acquires a second frequency image by inputting output of an encoder, which has input a first frequency image, to a decoder. The information processing apparatus (100) trains the encoder and the decoder based on a loss function, the first frequency image, and the second frequency image. In the loss function, a weight related to a first frequency is smaller than a weight related to a second frequency higher than the first frequency.

The invention provides a CGRA compiling method and device based on multistage intermediate representation and a polyhedral model, and relates to the technical field of compiler design. The method comprises the following steps: performing grammar and semantic analysis on a calculation task code or model to obtain an initial intermediate representation, converting to obtain an affine intermediate representation, and executing hardware-independent optimization; performing cyclic transformation and tensor optimization based on a polyhedral model, and realizing mapping by combining regularized mapping and a search strategy; and generating configuration information and an executable file on the reconfigurable architecture according to a mapping result. According to the method, the end-to-end compiling framework is constructed, and the loop program optimization capability of the polyhedral model and a rule-driven rapid mapping mechanism are combined, so that the execution performance and the compiling efficiency of the artificial intelligence task on a coarse-grained reconfigurable architecture are remarkably improved, the problems of optimization hierarchy splitting and low mapping speed of a traditional compiling method are effectively solved, and the method is suitable for large-scale popularization and application. The method is suitable for high-complexity and large-scale parallel artificial intelligence application deployment requirements.

The invention discloses a control program optimization method, a detection system, an electronic device and a storage medium, the control program optimization method is applied to a detection platform, the detection platform is connected with an elevator core component, when a to-be-tested demand is received, a test case is determined according to parameter information of the to-be-tested demand, and the test case is stored in the detection platform; the to-be-tested demand comprises at least one test demand of a to-be-tested function and a target fault; burning the elevator control program to be tested to the detection platform; the test case is executed on the detection platform, response parameters of the elevator core component are obtained, and a test result is obtained; and performing optimization iteration on the elevator control program based on the test result and an expected result, wherein the expected result shows that the elevator is in a normal state. The test case is determined according to the parameter information of the to-be-tested requirement and executed on the detection platform, the test can be carried out without waiting for the idle elevator, the development test duration can be greatly shortened, the development progress of the elevator function is accelerated, and meanwhile the elevator operation safety is guaranteed.

The application discloses a method and system for obtaining a control flow graph representation of a HIP kernel function basic block call relationship, extracts a kernel function of a HIP program and compiles the kernel function into an LLVM IR intermediate code, divides the kernel function into a plurality of basic blocks, allocates a unique serial number to each basic block, inserts a new edge basic block to represent a call relationship on a jump edge, generates a basic block level kernel function control flow graph, uses an LLVM insertion Pass to insert the newly added edge basic block, designs two light-weight thread insertion modes to reduce insertion overhead, the two light-weight thread insertion modes are ProfileCTA and ProfileWavefront, adopts a compilation process combining device end and host end insertion, runs on a domestic DCU platform, collects a call frequency between the basic blocks, takes the dynamic information as an edge weight of the CFG, and finally generates a basic block level HIP program kernel function CFG with dynamic information. The application effectively reduces the insertion overhead, improves the acquisition efficiency of the dynamic call relationship, and provides strong support for high-performance computing and program optimization.

The invention provides an optimization method for multi-parameter function calling, which comprises the following steps of: storing all parameters called by a program or parameters except for the first four parameters of a function parameter list in a section of memory space; the first address of the section of memory space replaces the original parameter needing to be transmitted to serve as a new parameter to participate in the function calling process, redundant or even all parameters are stored in the memory in advance, and only the first address of the memory storing the parameters is transmitted when the function is called. Therefore, the number of parameters needing to be transmitted during function calling is effectively reduced, repeated storage and reading operations of redundant parameters during function calling are omitted, the number of instructions for function implementation is reduced, the execution efficiency of a program is improved, and the purposes of optimizing the program, improving the program efficiency and finally improving the running efficiency of the whole system are achieved.