LEAP: A Virtual Platform Architecture for FPGAs

FPGAs are known to be very effective at accelerating certain classes of algorithms. A variety of FPGA platforms are available today, but because of the absence of a standardized platform architecture, each platform comes in the form of a board with a diverse set of devices and communication endpoints. Therefore, FPGA programmers typically have to spend significant effort in building interfaces to devices and adapting their applications to work with the semantics of these devices. Further, an FPGA board by itself is in many cases incapable running full real-world applications – software support is required. Working out communication protocols between the FPGA and software is another unnecessary time sink for programmers who would rather focus on the high-level functionalities of their applications. Finally, there is little support for building and allocating flexible memory hierarchies on FPGA platforms. All of these problems are further exacerbated by the fact that switching FPGA platforms usually requires the programmer to re-do a significant portion of this work. These are all non-issues for software programmers who live in a world of block and character devices, hardware-managed memory hierarchies with rich memory management libraries, and a plethora of portable communication protocols. We attempt to bridge this gap between platform support for software and FPGA application development by proposing LEAP Virtual Platforms. LEAP (Logic-based Environment for Application Programming) provides an FPGA application with a consistent set of useful services and device abstractions, a memory hierarchy and a flexible communication protocol across a range of FPGA physical platforms. Tying these functionalities together is a modular development and build infrastructure. In this paper we describe the services provided by LEAP and explain how they are implemented using a multi-layered stack of abstractions.