9 results about "Macro instruction" patented technology

The invention relates to the technical field of electrical automation, in particular to an electric cabinet automatic detection program construction method and system based on modular programming, an inspector firstly starts test application software written based on a VC language, a software interface displays an available test program list, the inspector selects a corresponding test program from the program list, and the test program is stored in the test application software; the method comprises detection steps and rules for a specific electric cabinet, after a program is started, a system prompts an inspector to input a cabinet body serial number and an inspector name, the information is used for generating a detection report, and if a cabinet body detection record with the same serial number exists in the system, the program pops up a prompt box, so that the inspection report is generated. According to the method and the system, an inspector is inquired whether to cover an original record or rewrite a new record, the inspector can choose to cover, rewrite or cancel operations, the modules are relatively independent, and the macro instruction module is utilized to compile abbreviation instructions of common actions, so that not only is repeated labor reduced, but also the error probability is reduced, and the programming efficiency is greatly improved.

The invention provides an integrated control method and system based on a whole satellite control macro instruction, and the method comprises the steps: generating a macro instruction in a unified format, and the macro instruction comprises a header, segment information and an auxiliary package; uploading the macro instruction to a satellite integrated electronic system; analyzing the macro instruction, and dynamically scheduling a task queue according to task analysis time in the header; when the task is executed, cooperative operation of the associated subsystems is automatically triggered, and the cooperative operation comprises attitude and orbit control pointing, load startup and shutdown, solid memory recording or playback and link establishment; task conflicts are processed through a task queue conflict detection mechanism, the planned tasks cover the planned tasks of the same time window, and insertion of new tasks is shielded by the protection time window for the tasks in execution. According to the method, collaborative management and task scheduling of multiple subsystems such as satellite attitude and orbit control, load, solid storage and link are realized, the instruction complexity is remarkably reduced, and the whole satellite control efficiency and robustness are improved.

The invention discloses a processing method and device. The method comprises the steps of obtaining historical execution data from a reasoning process of a target model; identifying a target tool calling sequence from the historical execution data; the target tool calling sequence comprises a plurality of target tools which are continuously called according to a time sequence; each target tool is used for executing a single target task related to target model reasoning; constructing a macro instruction tool based on the target tool calling sequence; the macro instruction tool is used for defining an execution sequence and a data dependency relationship of a plurality of target tools;

The invention discloses a central control screen instruction set processing method and system based on a scene mode, and relates to the technical field of instruction set processing, and the method comprises the steps: carrying out the real-time fusion of multi-dimensional information flows from voice, a driver monitoring system, a vehicle bus and an external environment, and forming the depth perception of the current state and abstract intention of a driver; based on this, a large-scale language model is introduced as a decision core, the extracted high-order intention is taken as a target, and a preset vehicle function atomic operation library is taken as an available resource to carry out dynamic logic arrangement and combination. And finally, the initial instruction sequence generated by the model is combined with the real-time vehicle state to carry out instant verification and safety optimization, and an accurate and executable composite macro instruction is output. Therefore, a decision closed loop from multi-modal scene perception to adaptive task execution can be realized, so that the active service capability and seamless interaction experience of the intelligent cabin are improved.

The application relates to a heterogeneous AI workflow unified scheduling and execution system based on a macro instruction protocol and belongs to the technical field of artificial intelligence and workflow automation. The system comprises the following: a macro instruction analysis module, which is used for analyzing a plurality of macro instruction steps in a workflow definition file and constructing a directed acyclic graph according to the inter-step dependency relationship; a compiling module, which is used for generating an execution plan through a logical workflow according to the macro instruction steps and the dependency relationship; a unified scheduling module, which is used for distributing corresponding step tasks to corresponding types of execution units according to the execution plan and promoting the execution of subsequent steps according to execution state feedback; a plurality of heterogeneous execution units, which are used for executing the step tasks respectively; and a resource deployment coordination module, which is used for triggering resource deployment when the resources corresponding to the model calling type step are in a non-ready state and feeding back the resource available state to the unified scheduling module after the resources are ready. The seamless integration and efficient collaboration of heterogeneous AI capabilities are realized.

This disclosure describes systems, methods, and devices related to using artificial intelligence to validate performance of integrated circuit features. A device may extract, from instruction files, microinstructions source and destination registers; generate a dependency graph including macroinstructions as nodes and dependencies between macroinstructions as edges between the nodes; generate, based on the dependency graph, a frequency distribution of instructions from trace files, performance univariate autoregressive conditionally heteroscedastic (Perf uarch) stat files, and register transfer language (RTL) stat files, predictors for a machine learning model; generate, based on the Perf uarch stat files and the RTL stat files, ratios of Perf uarch stats to RTL stats as target stat ratios; generate, using the predictors and the machine learning model, predicted ratios of Perf uarch stats to RTL stats; and generate, using greedy constrained optimization, based on the target stat ratios and the predicted ratios, recommended traces for debugging.

The invention discloses a man-machine interaction welding method and system based on multi-modal fusion and natural language, and the method comprises the steps: synchronously collecting voice, gesture and environment signals in real time, and preprocessing the obtained multi-modal data; based on the preprocessed multi-modal data, deep fusion of the multi-modal data is carried out, and the intention of an operator is accurately identified and confirmed; when the user intention is determined, continuously optimizing an adaptive speech recognition model, performing deep analysis on the user intention in a natural language form, and extracting task logic and parameters; based on the analyzed intention logic and parameters, an executable macro instruction sequence is generated, and the welding robot is driven to complete a task; in the task execution process, the task execution condition is continuously monitored to provide safety feedback, subsequent optimization is performed, and it is guaranteed that the task is safely and efficiently completed. The method provided by the invention can adapt to industrial noise and operator accent, solve the problem of single modal ambiguity, understand complex task instructions, and realize low delay and data security.

The invention discloses an ISA instruction format splicing method based on SystemRDL, and relates to the technical field of ISA instruction splicing, which comprises the following steps: establishing bit field mapping, analyzing operation code and operand information, definition field attribute, operation code field bit width of 7-10 bits, and immediate number field of 8-20 bits; instruction behaviors are expanded, operation logic and time sequence constraints are supplemented, and automatic assembling is carried out; the attribute of an operand is defined, the bit width of a register is 16-64 bits, and the bit width of an immediate digit is 1 bit to the maximum support width; splicing according to function classification, and setting corresponding bit field structures for arithmetic operation, memory access and control instructions; processing macro and pseudo instructions, decomposing and sequencing the macro instructions, and mapping the pseudo instructions; and scanning for three times to generate assembler codes, and sequentially analyzing, associating, integrating and outputting. According to the method, the instruction assembling efficiency and accuracy are improved, the expansibility and compatibility are enhanced, verification and maintenance are simplified, and automation and standardization of processor development are promoted.

The invention discloses a multi-level task allocation and decision support system oriented to man-machine symbiosis, which relates to the technical field of task strategy optimization and comprises a man-machine interaction module, a multi-level task allocation module, a decision support module, a system control center and a data bus, according to the method, accurate execution from strategic intention to specific operation is realized by constructing multi-level task analysis and dynamic allocation of target planning, process scheduling and specific execution, and a task analysis unit decomposes a macroscopic instruction into a three-layer task system with a clear target; the resource coordination unit quantifies the influence of external emergencies and internal task chain bottlenecks on the system under the driving of the decision support module, and the model prediction unit predicts execution consequences of different allocation schemes based on the influence, adjusts a task sequence in real time, forms a feedback optimization mechanism of environmental perception, task analysis and dynamic allocation, and improves the system performance. The mechanism ensures that strategic thinking of the user is converted into automatic operation of a machine through a system.