Fast digital design space exploration with high-level synthesis: A case study with approximate conjugate gradient pursuit

摘要

The rapid architecture design paradigm of highlevel synthesis (HLS) facilitates extensive design space exploration without much effort. Modern HLS compilers, like Xilinx Vivado HLS, support the evaluation of different architectures by so-called directives. We propose to extend the exploration capabilities through a data type-agnostic programming methodology. Automated by shell scripts, a link between the HLS tool and Mathworks Matlab facilitates the generation of large numbers of test vectors as stimuli to our designs for bit-accurate verification purposes. Additionally, we report on the first implementation of the Approximate Conjugate Gradient Pursuit algorithm, which is exemplarily used to demonstrate the augmented design space exploration capabilities.