Uncovering Actionable Chip Design Insights Through ML-Driven Big Data Analytics
Look deep underneath the covers of any digital chip design flow, and you will find a vast, untapped wealth of details about the and status of the SoC design. Let's suppose you can turn this information-seamlessly and efficiently-into actionable insights that increase your productivity and, ultimately, the caliber of your SoCs.
We are in an environment of growing SoC and system complexity, persistently aggressive market windows, as well as an increasingly challenging engineering resource landscape. This likely won't change in the near future. So, anything that can help you tap into the data richness of the digital design flow could give you a benefit over your competitors.
That's the aim of the Synopsys DesignDash>Better Observability and Visibility for All
The myriad engines that comprise the Synopsys RTL-to-signoff design flow operate as a constant interplay -a waltz-like dance, if you will -of golden signoff-enabled analysis (including timing, power, area, IR-drop, and DRCs) alongside targeted, highly tuned optimizations. Their singular goal is to co-enable the achievement of power, performance, and area (PPA) targets as smoothly and efficiently as possible. This constant analysis generates vast volumes of data, much of that is highly unstructured. However, this mass of disconnected data points, when taken together, can provide a detailed knowledge of the \”health\” from the design. For instance, where are additional opportunities for design improvement? So what can be addressed earlier in the flow to create this design to closure?
Log files do go some way to expose this data. Still, using the size and complexity of the current crop of designs, only small views in to the vast underlying data are possible before they become too unwieldy to understand. Engineers, therefore, have long had to try to second-guess what data might be useful to them after which use proactive>The Synopsys DesignDash solution presents such a holistic view of all project data after which drives it many steps further (Figure 1). The answer efficiently and autonomously siphons metrics data while also intelligently curating the associated analysis data from Synopsys' unique single data model. Then, it transforms and loads the information into always-on, industry-standard databases. Handling flow metrics from a third-party tool is every bit simple. With the data available in tool-decoupled databases, searching, filtering, graphing, comparing, and trending are pretty straight forward, intuitive tasks. Weight loss runs progress, that data -indeed, the information in the entire project team -can also be absorbed, compared, and cross-referenced via a responsive, web-based interface (UI) after which distributed to ease.
Figure 1: Synopsys DesignDash technology enables comprehensive observability, bringing better progress tracking across blocks, sub-systems, or entire SoCs. Build the best view for the team or organization. Quickly identify areas for additional focus.
This concept of seamless sharing is a central thrust of the Synopsys DesignDash solution. The old, manual method of piecemeal data capture has made the look process extremely opaque. Not having an extensive way to measure underlying design processes has made it extremely difficult to enhance and more effectively manage them.
A live, 360-degree look at all design activities allows standardization to drive project-wide efficiencies – based on the proven fact that everything is now being measured. Engineers no more have to extract data, come in into spreadsheets, format it (with maybe somewhat arbitrary colors), and paste it into slides to report their daily or weekly progress. How many hours are the team currently paying for this kind of task? Too many! Instead, with our unique>Generating Rich Analytics to show True Prescriptive Chip Design Insights
Harvesting and curating the large data is just the start of the story. Synopsys DesignDash technology mines the breadth of incoming data to exhibit not only what's happening -the state of the underlying engines at any point throughout the flow -but also why it is happening. These machine learning- (ML-) based, augmented analytics allow it to be faster and easier to quickly obtain a deep, intimate knowledge of the look by autonomously classifying design trends, identifying design limitations, and providing guided root-cause analysis over the entire design flow. With the analytics running automatically without anyone's knowledge, it's like an army of extra engineers performing expert-level debugging after which distilling that real-time analysis into easily consumable (i.e., intuitive, customizable, interactive), cross-comparable visualizations.
How did my critical path evolve structurally through every stage from the flow? What was the dominant reason (logic depth, detouring, layer choice, skew, insertion delay) causing its failure? How does this run rival my other three experiments? What app option or flow change most likely caused it? Solutions to these types of questions and much more are quickly resolved as everyone on the team becomes equipped with deep domain expertise. It's a path toward better, more informed, more>
Figure 2: The Synopsys DesignDash solution enables the cross-visualization of data all flow stages – or perhaps across different design flows – in easy-to-comprehend views to hurry design understanding.
The deep, augmented analytics from Synopsys DesignDash technology also opens the door to prescriptive guidance, telling teams some different ways to fix the problems that have been automatically identified (Figure 3). Statistical- and ML-based models that capture how the tool best addresses specific design issues, accompanied by the generation of tool-consumable scripts to directly target the issues available, greatly speed design closure and unlock otherwise untapped PPA.
Figure 3: The Synopsys DesignDash solution fuses data from numerous, disparate cross-domain sources right into a combined view for seamless cross-probing. This view shows cross-talk timing impact versus location.
Helping Silicon Engineering Teams Work Faster
Working hand-in-hand using the Synopsys Digital Design Family of products, the Synopsys DesignDash solution smartly complements the industry-first Synopsys DSO.ai autonomous AI application for chip design. Together, these technologies address the growing shortfalls in productivity that lots of design teams face today. Through easier debugging and optimization, Synopsys DesignDash technology accelerates the achievement of more \”implementable\” designs across a much broader architectural solution space. These more optimized designs make up the starting points of the vast, multi-dimensional solution space search supplied by DSO.ai technology, significantly speeding the path to truly optimum system PPA. In essence, the Synopsys DesignDash solution will speed the identification of the best mountain in the best mountain range, and also the DSO.ai solution uses its many smarts to quickly find the 'top' of the highest peak -across an infinitely more comprehensive set of chip design goals.
Author: Mark Richards Sr. Staff Product Marketing Manager, Synopsys Silicon Realization Group.