When we talk about low-power verification its easy to equate it with simulation. For certain, simulation is the heart of a low-power verification solution. Simulation enables engineers to run their design in the context of power intent. The challenge is that a simulation-only approach is inadequate. For example, if engineers could achieve SoC quality by verifying the individual function of each power control module (PCM), then simulation could be enough. For a single power domain, simulation can be enough.
However, when the SoC has multiple power domains -- and we have seen SoCs with hundreds of them -- engineers have to check the PCMs and all of the arcs between the power modes. These SoCs often synchronize some of the domain switching to reduce overall complexity, creating the potential for signal skew errors on the control signals for the connected domains. Managing these complexities requires verification methodologies including advanced debug, verification planning, assertion-based verification, Universal Verification Methodology - Low Power (UVM-LP), and more (see Figure 1).
Figure 1: Comprehensive Low-Power Verification
But even advanced verification methodologies on top of simulation aren't enough. For example, the state machine that defines the legal and illegal power mode transitions is often written in software. The speed and capacity of the Palladium emulation platform is ideal to verify in this context, and it is integrated with simulation sharing debug, UVM acceleration, and static checks for low-power. And, it reports verification progress into a holistic plan for the SoC. Another example is the ability to compare the design in the implementation flow with the design running in simulation to make sure that what we verify is what we intend to build.
Taken together, verification across multiple engines provides the comprehensive low-power verification needed for today's advanced node SoCs. That's the heart of this low-power verification announcement.
Another point you may have noticed is the extension of the Common Power Format (CPF) based power-aware support in the Incisive Enterprise Simulator to IEEE 1801. We chose to bring IEEE 1801 to simulation first because users like you sometimes need to mix vendors for regression flows. Over time, Cadence will extend the low-power capabilities throughout its product suite to IEEE 1801.
If you are using CPF today, you already have the best low-power solution. The evidence is clear: the upcoming IEEE 1801-2013 update includes many of the CPF features contributed to 1801/UPF to enable methodology convergence. Since you already have those features in the CPF flow, any migration before you have a mature IEEE 1801-2013 tool flow would reduce the functionality you have today.
If you are using Unified Power Format (UPF) 1.0 today, you want to start planning your move toward the IEEE 1801-2013 standard. A good first step would be to move to the IEEE 1801-2009 standard. It fills holes in the earlier UPF 1.0 definition. While it does lack key features in -2013, it is an improvement that will make the migration to -2013 easier. The Incisive 13.1 release will run both UPF 1.0 and IEEE 1801-2009 power intent today.
Over the next few weeks you'll see more technical blogs about the low-power capabilities coming in the Incisive 13.1 release. You can also join us on June 19 for a webinar that will introduce those capabilities using the reference design supplied with the Incisive Enterprise Simulator release.
=Adam "The Jouler" Sherer
(Yes, "Sherilog" is still here. :-) )