The partitioning process identifies the sections of code that can be run, in parallel, across individual, and potentially different types of, processor cores. The process of mapping then allows the developer to decide which software partition to allocate to each processor core. Accelerators or other hardware in the architecture can be modeled by providing performance information. Code partitions can then be mapped onto the hardware for more accurate performance estimation.
Once the partitioning and mapping decisions are complete, vfEmbedded provides a set of step-by-step instructions, referred to as a recipe, from which those decisions can be implemented without introducing unexpected errors. The time saved by both accelerating the partitioning and mapping process and reducing debug time through the avoidance of bugs makes it possible to complete the process in days instead of months.
The target systems for vfEmbedded are traditional embedded applications and system-on-chip (SoC) devices. vfEmbedded is ideal for software designers trying to implement complex algorithms and codecs onto a given platform. By using the tool they can optimize the application by squeezing more code into the available spaces without wasting time trying different partitions on a trial and error basis. In addition, vfEmbedded suits software architects trying to match the various algorithms and code being used against potential hardware architectures. In this way, a better and more efficient match can be made between software and hardware resources. This approach avoids resource over-provisioning and results in tighter control of end-product cost.
vfEmbedded is available as a cloud-based software service from March 1st. The tool is accessed using a standard web browser through the Vector Fabrics web site. Pricing is from ?450 per month. A free trial, including several example designs, is available.