UC Santa Cruz

Constrained random verification (CRV) at system level is often considered difficult due to the size of the solution-space due to the number of parameters and their inter-dependencies dictated by the requirements. The overwhelming complexity lies not only in understanding the system level requirements, but also in creating valid constraints and generating a valid scenario via a constraint-solver. In this paper, we present a multi-staged methodology to generate practical constraints that can be easily solved. Our methodology has been applied to a 5G-NR (5th Generation - New Radio) communication link by effectively mapping the 5G-NR specification to the SystemVerilog constraints and delivering a constraint-random (CR) generated test scenario for effective verification of any uplink/downlink system. The constraint-random engine focuses on creating a 5G-NR resource-grid. The resource-grid is a representation of the time and frequency domain. The CR infrastructure maps the resource allocation requirements present in the 5G-NR specification into SystemVerilog constraints and builds the slot-map with the desired channeltypes. The key challenge this paper addresses is to identify the intersection region for all the requirements per channel type abstracted as parameters along with inter-dependency between channels. The solution focuses on breaking down the randomization load effectively to make randomization at this grand scale a possibility. In the end, we successfully demonstrate a practical 5G NR system level CR test scenario generation framework that can randomize a full scenario with approximately 1% of verification execution time overhead.