Prof Imai said that he had been working with VHDL for over ten years, and that in a way he finds it a perfect language. However, he accepted that it is difficult to use and that there is a case for using a simplified, higher level language. After seeing Colin’s presentation of Handel-C, Prof Imai suggested that it seemed similar to NTT’s "Structured Function description Language" (SFL) [http://www.kecl.ntt.co.jp/parthenon/index_e.htm], especially because of its synchronous semantics. He agreed that Handel-C is closer to being a Register Transfer Language than some other C-based languages, notably Spec-C [http://www.specc.gr.jp/eng/] and System-C [http://www.systemc.org/], while still having the look and feel of a high level language.
Prof Imai stressed the importance of sound semantics, and suggested that "lessons learned over the last ten years have indicated that extending a language is a bad idea" - implying (I think) that C or C++, or a subset of either, can themselves be used for hardware specification and that he wasn’t convinced of the necessity to incorporate extensions. He added that if extensions are necessary they should be kept to a minimum. (Handel-C extends a subset of C by incorporating an explicit par construct, with inter-process communication based on channels as in Hoare’s CSP [http://www.afm.sbu.ac.uk/csp/]).
Prof Imai pointed out that with pure C or C++ it is possible to preserve backwards and forwards movement between the hardware and software domains, and he suggested that this is an important issue in hardware/software co-design. On the other hand, he accepted the argument that the extensions within Handel-C give designers more control over the structure of the hardware. (C and C++ lack any explicit mechanism to define parallel processes; it is up to the EDA software to identify any parallelism that is implicit in the specification.) Colin said in reply that many researchers have concluded that auto-parallelisation is a largely intractable problem. Prof Imai concluded that Handel-C would/might be a success only if the advantages associated with being able to define parallelism and inter-process communication explicitly were found to outweigh the disadvantages associated with language extension.
Although he seemed impressed by the demonstrations and enthusiastic about the prospect of C-based design, Prof Imai said that he had some doubts about the general quality of the hardware that would result from using a C-based language. His view was that even if it becomes popular to use a C-based language to design hardware VHDL will still be required because experienced designers would be able to produce better (small scale) designs using VHDL. By way of an analogy, he suggested that a skillful software engineer should be able to a produce better small scale program by writing directly in assembler than could be achieved by using a high level language and a compiler. Colin said he doubted this, and I do too, but the idea that VHDL (or Verilog HDL) will still be necessary in the future seems beyond dispute.