This PR brings some miscellaneous quality of life improvements from the performance_estimates branch that are not related to the performance estimates themselves.
Don't pad trees automatically when unroll is False for xilinxhls backend
Add Vivado synthesis option to xilinxhls backend
Add some utilities for model inspection: depth, sparsity
Make HLS/HDL report reading more robust
Don't create the vivadohls and vitishls backends, just use xilinxhls. Tool discovery is still automatic (so either Vivado HLS or Vitis HLS will be used automatically). This creation of the backends previously prevents XilinxHLSModel from being pickle-able, which is useful in some cases
This PR brings some miscellaneous quality of life improvements from the
performance_estimates
branch that are not related to the performance estimates themselves.unroll
isFalse
forxilinxhls
backendxilinxhls
backendvivadohls
andvitishls
backends, just usexilinxhls
. Tool discovery is still automatic (so either Vivado HLS or Vitis HLS will be used automatically). This creation of the backends previously preventsXilinxHLSModel
from beingpickle
-able, which is useful in some cases