Jim,
I agree with you about the value of mixing and matching HLL and HDL solutions in your system. You might want to design the core of your algorithm using an HLL tool, then link it up to control and memory systems that you've designed using HDL.
I also can see the constant changing of the underlying hardware as being a challenge to those who are developing C-to-hardware tools. I think perhaps it favours those who target their systems at a single architecture, like SRC with their Carte programming environment. HandelC has seen a lot of real success, the most notable in my mind is it being used in an effort by Lockheed Martin to create a space system to dock with Hubble
With each new generation of FPGA, you'll need to update your underlying library routines. As I recall Peter Alfke saying at this year's FPL, to get the best out of FPGAs, you need to target your architectures, generic just won't cut it (apologies to Peter if that's not what he was getting at).
So I guess I agree with you Jim, in that C -> registers is not the best approach.
Apologies for my ignorance, but can I ask you to expand on "alternative of HLL -> FPGA Running HLL amd the best tool set". I wasn't sure what you meant.
Cheers,
Robin