IP-XACT Is Back, For All The Right Reasons

The intent behind IP-XACT has always been to provide a bridge between system-on-chip (SoC) assembly and larger considerations. This standard has additionally been used to adapt to multi-sourced and constantly evolving intellectual property (IP) that design and product teams build, often in different companies. Moreover, it was used to interface with product development beyond the specialized needs of logic design. Admittedly, it was developed early, offering a solution to a problem not yet widely recognized. It was early, but it was not wrong. Market changes are pushing more and more SoC builders in this direction in mature and emerging semiconductor and systems companies. Production needs are finally catching up with this standard.

What changed?

For a long time, SoC design was viewed as little more than an extension of lower-level register transfer level (RTL) design. Whatever might be needed beyond RTL could be handled through in-house script-ware. That worked very well for a long time until it started to break down for multiple reasons: a familiar nemesis — size and complexity; more IPs from many more sources, each issuing updates to fix bugs and add features; many more connections between all these IPs; and more control layers – power, safety, and security. Imperfections in the scripts start to appear the more the system is stressed.

In part, design has become more collaborative between geographically diverse teams, in-house and across companies. There is also more collaboration between design companies and systems companies. Tier 1s and OEMs now actively partner with semiconductor companies in the automotive industry, fighting to stay ahead of existing and emerging competitors. Hyperscalers are building more of their own SoCs to gain an edge in throughput and features. 5G infrastructure is again wide open for competition. And governments all over the world are investing in semiconductor development at all levels, from chips to electric vehicles, to supercomputers and mobile networks. All of which push inefficiencies and barriers to collaboration under a microscope. Collaborating teams demand a single and reliable source of truth for the design — an electronic document detailing the information they need. IP-XACT was always intended to provide this support.

Implications for SoC integration

Although teams will need to build an IP-XACT model for each IP, nothing changes at the IP level. The SoC level will benefit most by switching to an IP-XACT infrastructure. Subsystems benefit, too, if teams want to spin and adapt product families more quickly. No one hand-creates SoC top levels these days, so they are already not using RTL directly. Most SoC teams use some mix of spreadsheets and script-ware, but they are not standardized. Each team cooks their own or adapts another team’s code. They do the same for memory maps, header files, and software macros — more script-ware. Product managers want to keep product documentation in sync with changes as the design evolves, requiring manual information transfer and manual checking.

These steps can be grounded in IP-XACT with significant automation to support creation, editing, and generation for design collateral, software collateral, documentation inserts, and standards-based formats. Designers can still bolt on favorite high-value custom features through application programming interfaces (APIs). There is no need to lose these. Other teams in the company can also take advantage of the cool customizations.

What is coming next?

Now, there is a groundswell of adoption for IP-XACT in the industry; it is being pushed harder. There is interest in integrating documentation tables (clocks, resets, memory maps, etc.) directly into the tech pubs production flow, so there is no chance for that data to be out of sync with the design.

There is an even bigger push for traceability in the entire design. Traceability is a routine expectation in automotive, aerospace, and medical systems design. Increasing dependency on advanced SoCs now means that traceability must cover the complete devices. Supporting this need requires establishing links between requirements documentation, design features, and tests that validate those features. There is a lot of manual work in developing this level of traceability today. It must become more automated to support larger systems, IPs, and subsystems coming from multiple sources.

As global companies both large and small adopt IP-XACT, it is clear that the potential of this standard is real. It is not much of a stretch to imagine enterprise management systems wanting to tie into some of this data for resources planning, among other things: IP licenses, compute farm resources, cloud compute resources, and much more. The IP-XACT potential is only starting to be tapped. Talk to us to learn more.