Test Access Control for an Embedded MAC Unit Based Core using IEEE Std 1687

Ahstract-A major move today in developing and deploying embedded electronic systems is to include trained machine learning (ML) models that can be used for predicting new data in order to include added value to the embedded system functionality. The ML model can be implemented in software or hardware, and would typically require a highly parallelized architecture in order to reduce computation times. The design effort to create a suitable implementation is non-trivial. However, an aspect of the overall system development and deployment requirements that can be overlooked is the testability of the design hardware. In this paper, the test access for a trained ML model implemented in hardware as a MAC unit, and realized within a field programmable gate array (FPGA), is presented. The IEEE Std 1687 to provide access to embedded instrumentation is utilized to support the test requirements. The ML trained model is considered as an embedded computation core connected to a host processor.