Distributed model compilation

摘要

Methods and systems are provided for partitioning and distributing the model processing of a graphical model to provide an executable form of the graphical model. The methods and systems determine tasks of the model processing that can be processed separately and concurrently from other tasks. A distribution scheme for distributing the model processing tasks is determined that considers the characteristics of the model processing tasks in view of the characteristics of available computing resources. Variations of partitioning and distributing the model processing tasks can be determined to provide for distributed processing in an optimized or desired manner. The results of the distributed processing of the model processing tasks are obtained and combined to provide an executable form for the graphical model.

主权项

1. A computer-implemented method, comprising:
receiving a graphical model to be compiled,
the graphical model comprising elements including a plurality of signals; identifying a first type of signal propagation that is performed in compiling the graphical model; identifying a second type of signal propagation that is performed in compiling the graphical model; analyzing a model processing graph for the graphical model, wherein the model processing graph includes:
a first branch identifying a first set of elements of the graphical model that are subject to the first type of signal propagation, anda second branch identifying a second set of elements of the graphical model that are subject to the second type of signal propagation; based on the analyzing of the model processing graph, determining that the first type of signal propagation for the first set of elements is to be performed concurrently with the second type of signal propagation for the second set of elements; partitioning the first type of signal propagation of the first set of elements into a first task; partitioning the second type of signal propagation of the second set of elements into a second task; and sending the first task and the second task to separate distributed model compilers for performing the tasks.