Calibre SONR leads the semiconductor industry into the machine learning era. It converts all kinds of information, chiefly layout information, to features with excellent scalability and memory consumption, giving users unprecedented insight into their chips. It performs pattern clustering for unsupervised or semi-supervised machine-learning models, grouping billions, even tens of billions, of coordinates into distinct clusters based on their features. Calibre SONR transforms the landscape, making applications like full-chip down sampling and comprehensive chip comparisons a reality.
Calibre SONR uses a unique database format (MLDB) for data saving and querying. Compared to the traditional CSV and SQL, MLDB file size can be compressed to tens or even hundreds of times smaller. Querying information from the database is very fast (seconds) and can use multiple threads for further speed up.