High Level Synthesis for Intel and Xilinx FPGAs
High Level Synthesis for Intel and Xilinx FPGAs
Missing Link Electronics (MLE) has been an early adopter of High Level Synthesis (HLS) for FPGAs. In particular for Domain-Specific Architectures which aim to accelerate algorithms and communication protocols HLS delivers on the promised benefits:
- Increased productivity as we can focus on the behavior and let HLS do the scheduling and resource mapping
- Better portability across FPGA device families and even across FPGA device vendors
This MLE Technical Brief describes our findings when using HLS to accelerate a telecommunications network protocol accelerator with FPGA. Driven by the project’s need for short Time-to-Market major portions have been implemented in C/C++ using HLS. And given the application’s large unit volume it was important to evaluate cost/performance across a set of Intel and Xilinx FPGA devices.
Our example uses Intel and Xilinx HLS to implement a specialized Packet FIFO. This Packet FIFO is then integrated as a particular design block into a block-based top-level design. Despite the fact that Intel HLS and Xilinx HLS behave quite differently, and do require special code, we did see a benefit from using HLS compared to “classical” RTL design using VHDL and/or Verilog HDL.
Hence, we encourage the reader to follow a similar approach.