Patents
Literature
1 results about "Macro instruction" patented technology
Efficacy Topic
Property
Owner
Technical Advancement
Application Domain
Technology Topic
Technology Field Word
Patent Country/Region
Patent Type
Patent Status
Application Year
Inventor
A Macro instruction is a line of computer program coding that results in one or more lines of program coding in the target programming language, sets variables for use by other statements, etc.. In the mid-1950s, when assembly language programming was commonly used to write programs for digital computers, the use of macro instructions was initiated for two main purposes: to reduce the amount of program coding that had to be written by generating several assembly language statements from one macro instruction and to enforce program writing standards, e.g. specifying input/output commands in standard ways. Macro instructions were effectively a middle step between assembly language programming and the high-level programming languages that followed, such as FORTRAN and COBOL. Two of the earliest programming installations to develop "macro languages" for the IBM 705 computer were at Dow Chemical Corp. in Delaware and the Air Material Command, Ballistics Missile Logistics Office in California.
Heterogeneous ai workflow unified orchestration and execution system based on macro instruction protocol
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.
Owner:SHANGHAI SHENGCHUN SUHUIYUN TECHNOLOGY CO LTD