Michael van Elst (mlel...@speckled.mpifr-bonn.mpg.de) wrote:
: In <D32KGA....@unx.sas.com> ja...@cdevil.unx.sas.com (James Cooper) writes:
: >Main Disadvantages:  Program load speed would be slower, due to the
: >translation step.  Also, the resultant code may not be as absolutely
: >highly optimized as possible by directly coding for the specific CPU.

: The second point is the real problem. While it is better than a simulator
: for the "CPU-X" it will cost you probably a factor 2 in performance and
: program size, especially when the native CPU and "CPU-X" do not match.

  Actually, anyone involved with this project should take the time to
  read ftp://ftp.inf.ethz.ch/doc/diss/th10497.ps.gz . It describes a
  very feasible way of doing a cross-platform binary system. I believe
  the machines used in the study were 68k based and PowerPC based. I
  haven't finished reading it yet (its long), but it seems like it might
  be very applicable to this project. Some one else (I am sorry but I lost
  the name) already suggested checking this out, and I did, and I second
  the suggestion.

: Regards,
: --
:                                 Michael van Elst

: Internet: mlel...@mpifr-bonn.mpg.de                mlel...@serpens.rhein.de
:                                 "A potential Snark may lurk in every tree."

--
thedw...@mail.utexas.edu <the Edward Blevins>