Design and analysis of the worlds most advanced microprocessors using Jupyter notebooks
Kerim Kalafala and Nicholai L'Esperance share their experiences using Jupyter notebooks as a critical aid in designing the next generation of IBM Power and Z processors, focusing on analytics on graphs consisting of hundreds of millions of nodes. Along the way, Kerim and Nicholai explain how they leverage Jupyter notebooks as part of their overall design system.
|Talk Title||Design and analysis of the worlds most advanced microprocessors using Jupyter notebooks|
|Speakers||Kerim Kalafala (IBM), NICHOLAI L’ESPERANCE (IBM)|
|Conference||JupyterCon in New York 2018|
|Conf Tag||The Official Jupyter Conference|
|Location||New York, New York|
|Date||August 22-24, 2018|
When it comes to advanced technology, time to market is a critical factor. Processor manufacturers are under immense pressure to deliver new generations of chips that offer faster processing and greater reliability with higher efficiency, which means that each new generation is more complex and has more (and smaller) components, with four times the number of internal connections. Throughout a microprocessor design project lifecycle, terabytes of data are produced, which must be efficiently analyzed and acted upon within ever shrinking time-to-market windows. Within this mountain of data, relationships are crucial entities (e.g., Which circuits talk to each other? How does high-level logic design correspond to detailed physical implementation? What are the shortest and longest paths through a given network of electrical components?). The combination of Python analytics with graph databases is uniquely well suited to the analysis tasks at hand. Kerim Kalafala and Nicholai L’Esperance explain how IBM is using Jupyter notebooks as part of its overall analytics stack to tackle one of the world’s most complex computer design and engineering problems and share their ongoing experiences using Jupyter notebooks to design the next generation of Power and Z processors. In particular, they showcase how Python-based analytics, in conjunction with graph databases, are becoming a critical component to the analysis and design of the world’s fastest microprocessors (the “brains” of every single computer designed by IBM, including those that sit at the heart of the IBM Watson Cognitive Computing platform). The complexity at hand includes analysis of ~10+ billion transistors to an accuracy tolerance of a trillionth of a second (roughly the time taken for light to travel a fraction of a millimeter). Along the way, Kerim and Nicholai explain how they leverage Jupyter notebooks as part of their overall design system.