Real-Time Speeds Reach Embedded

Using virtual platforms, VaST Systems Technology Corp. announced that the cycle-accurate development of embedded systems can now be performed at real-time speeds. For developers of software-rich chips, such as wireless devices and automotive electronics, this means that they can perform architectural analysis and cycle-accurate simulation of their embedded software up to 20 times faster. Raw processor simulation performance claims to be five times faster.

These performance comparisons are based on the newest versions of VaST's Comet and Meteor embedded-system development tools. Compared to other approaches, such as in-house models from processor vendors and models generated from architectural descriptions, Comet and Meteor claim to be 25 to 250 times faster. The performance for real-world platforms involving peripheral devices is now much closer to raw processor performance when there is no hardware interaction. Examples of such peripheral devices include automotive power-train control or 3G cell phones.

To execute embedded software, Comet and Meteor incorporated VaST's virtual-platform simulation technology. They could then accurately predict its eventual behavior. Comet is used to construct and modify virtual platforms. It then analyzes them while running a realistic software load to provide the cycle-accurate simulation of virtual platforms including processors, buses, and peripherals.

Cycle accuracy means that these virtual platforms can be used for hardware-knowledgeable software like device drivers. They also can be utilized in the development of software with real-time constraints, such as wireless and automotive devices. In many applications, VaST's virtual platforms execute code faster than actual hardware, making them suitable for developing almost any type of embedded software.

Version 5.0 of Comet and Meteor, which contain these performance improvements, are available now from VaST. U.S. pricing for Comet starts at $50,000 per year. Pricing for Meteor starts at $15,000 per year. The price of processor models depends on their complexity.