29 - 30 October, 2019

Munich, Germany

Event Details

MP Associates, Inc.
THURSDAY October 25, 10:45am - 12:15pm | Forum 6
EVENT TYPE: REGULAR SESSION
SESSION 3
UVM I
Chair:
Alexander Rath - Infineon Technologies
New Ideas on UVM.

3.1Extending Functionality of UVM Components by using Visitor Design Pattern
Visitor is a software design pattern used to add a new operation to each class in an existing class structure, suitable to be implemented for classes within a well-defined class hierarchy, such is the UVM based environment. All the necessary infrastructure for the usage of Visitor is provided within UVM library. The paper will briefly introduce the pattern elements and show the examples how the pattern can be incorporated into the verification environment, with benefits it brings.
 Speaker: Darko M. Tomusilovic - VTool Ltd.
 Author: Darko M. Tomusilovic - VTool Ltd.
3.2UVM Register Map Dynamic Configuration
UVM register map provides data structures and testsuites that allow checking the data integrity of the DUT register map during simulations. This object can be extended to allow the synchronization between the register model and any verification component present in the environment. Beside ad-hoc approaches based on code customization or custom code generation from IP-XACT, we present a solution that simplifies and automatizes such interaction.
 Speaker: Matteo Barbati - STMicroelectronics
 Authors: Matteo Barbati - STMicroelectronics
Alberto Allara - STMicroelectronics
3.3Clustering and Classification of UVM Test Failures using Machine Learning Techniques
When verifying complex hardware designs we are often faced with the problem of analyzing failures in large test suites. This paper describes how we applied machine learning techniques to automate the tasks of clustering and classifying test failures. Different algorithms were implemented, and their performance was evaluated using real life data from our UVM test benches. This paper also discusses the pre-processing of simulation log files to transform them into data suitable as inputs to machine learning algorithms.
 Speaker: Lars Viklund - Axis Communications AB
 Authors: Andy Truong - Axis Communications AB
Daniel Hellström - Axis Communications AB
Harry Duque - Axis Communications AB
Lars Viklund - Axis Communications AB