It does seem like the basic definition of IP as a building block is the big problem. This locks you in to a bottom up approach. We need a "compiler" that can pull from the "library" of building block components and customize then as required to meet the high level requirements (speed, power, latency, etc). The tool would search thru the solution space to find a particular combination that fits the requirements (or comes back with a 'no solution' result). That's the direction I'd like to see things move...
In that case, please check us out, Space Codesign. We have the capability of going to implementation after algorithm design and validation, and architecture design exploration. Implementation can either map to libraries (e.g., Xilinx for FPGA) or to HLS (high level synthesis - Forte, Calypto, Xilinx, etc.) after design exploration figures out meeting your requirements.
By the way, Guy Bois, our president has a new blog article on this site discussing our approach.
As we unveil EE Times’ 2015 Silicon 60 list, journalist & Silicon 60 researcher Peter Clarke hosts a conversation on startups in the electronics industry. Panelists Dan Armbrust (investment firm Silicon Catalyst), Andrew Kau (venture capital firm Walden International), and Stan Boland (successful serial entrepreneur, former CEO of Neul, Icera) join in the live debate.