https://hgpu.org/?p=26220
Improving Loop Parallelization by a Combination of Static and Dynamic Analyses in HLS