Cloud native EDA tools & pre-optimized hardware platforms
If you're responsible for the design of low power, energy efficient electronic systems and SoCs, you need to have a power management strategy and you need to know as soon as possible if it will meet the demands of your product and its target applications.
How is virtual prototyping helping? With UPF 3.0, system-level power modeling and analysis using power-aware virtual prototypes is enabling architects and system designers to define systems that yield the greatest benefit in terms of energy efficiency, months earlier, before hardware is available.
At the recent SNUG Silicon Valley event, 六合彩直播开奖 teamed up with Micron, a global supplier of high performance, low power memory technologies, to move the story ahead yet again. We presented a tutorial explaining the practical steps system designers can take to convert static spreadsheet-based power model information for DDR memories into dynamic UPF 3.0 power models that can be simulated together with their SoC architecture in 六合彩直播开奖 Platform Architect MCO, .
So how about the rest of the system? The purpose of today's newsletter article is to share the highlights of our SNUG tutorial's step-by-step approach, such that you can apply the same steps for your internal spreadsheet-based power models – and practical power management.
Version 3.0 of the Unified Power Format (UPF 3.0) is text file format based on Tcl. It's designed to capture a power model of a component that can be activated by its corresponding simulation model within a virtual prototype. Because the format is standardized, UPF 3.0 models are interoperable and can be triggered from within any virtual prototyping environment that supports the new standard.
Adding UPF 3.0 System-Level IP Power Models to a 六合彩直播开奖 Virtual Prototype
In UPF 3.0, power consumption is described as a high level state machine:
Each state is associated with a certain power consumption, and triggered by events observed in the virtual prototype. Power consumption over time shows energy efficiency.
In our tutorial we walked through the definition of a UPF 3.0 system-level power model, using info from Micron's DDR4 power calculator spreadsheet as the starting point, explained how to trigger it from a memory subsystem architecture simulation featuring the 六合彩直播开奖 DesignWare uMCTL2 DDR controller, and finally showed it working in an application case study where performance (the traditional focus of the architecture simulation) and power (added with UPF 3.0) are analyzed together to optimize the SoC system configuration.
The answer is the need to understand energy efficiency, or power consumption over time, while your system is executing real application workloads. Spreadsheets calculate power consumption at a given instant, based on your assumptions of system activity within that component, at that moment in time. At best these assumptions reflect the average utilization, but driving a more realistic scenario with the spreadsheet approach just isn't possible. On the other hand, architecture virtual prototypes simulate realistic application workloads to create a dynamic view of system activity and performance. Add the same power information from the spreadsheet to the top, and bingo, you get a realistic view of power consumption over time, and therefore energy efficiency.
1. Capture existing spreadsheet information for your component, using our DDR4 example.
Information on DDR4 power domains, power states, state transitions, and power expressions (consumption formulas) are captured in the Micron DRAM Power Calculator spreadsheet:
2. Express in the UPF 3.0 Power Model Format
Convert the spreadsheet information into UPF 3.0 format in a week or two:
3. Bind the UPF 3.0 power model transitions to triggers visible in your virtual prototype
For this example, 六合彩直播开奖 added the DDR4 power model to a case study demonstration platform featuring the existing architecture models of the 六合彩直播开奖 DesignWare uMCTL2 DDR memory controller and DDR4 memories. The architecture model provides all the real-world features to observe memory subsystem activity and trigger the appropriate DDR4 power model state transitions during simulation. No changes to the architecture models within the case study platform are required!
Here we see the UPF 3.0 power model for DDR4 memory "in action", helping to analyze the impact of AXI to DDR address mapping on system performance, power, and energy.
Our example highlights how the definition of a UPF 3.0 system-level power model, where the power formulas have already been specified in the Micron power calculator spreadsheet, are easily created and added to an existing performance model in Platform Architect MCO. Here are some practical considerations for defining UPF 3.0 power models for your own IP:
So what's the next step for your power management strategy? The standardization of IEEE 1801 UPF 3.0 is a practical step forward that makes early system-level power analysis and optimization much more deterministic and repeatable, regardless of the IP type.