ATPG based functional test for data paths: application to a floating point unit

Application of an ATPG based functional test methodology that is tailored towards data paths to a floating point unit is described. The methodology employs the instruction set of the processor to control the inputs and to observe the outputs of the data path and utilizes an ATPG tool to generate test patterns. The test patterns are then converted to instruction sequences and applied as a functional test. This methodology provides high at-speed coverage without the performance and area overhead of the traditional structural testing. While we target stuck-at faults in this work, the methodology is applicable to other faults models, including delay faults.