Learnings from a HLS-based High-Productivity Digital VLSI Flow