Re: Reverse compilation of DSP code

MicroAPL, a British company I believe, sells a 68k to PowerPC assembly
language conversion tool. I know a few commercial Macintosh to Power
Macintosh products used this. This sort of thing isn't very common in
the PC/Mac developer market because people usually either write in a
high level language in the first place or take the opportunity to
rewrite when an architecture change forces them too. Or they just kill
the product :-)


(Note I didn't say anything about the high-level language code being
portable above. Basically the computer companies making the
architecture changes are forced to handle that. If PowerPC was not
big-endian or didn't allow unaligned memory accesses, the architecture
transition likely would have been unsuccessful. We see the same sorts
of things with architecture transitions in the Windows
world. Everything is little-endian. I expect IA-64 will have efficient
support for unaligned accesses somehow. The 64-bit extension of the
Windows API leaves most of the existing integer types the same size
and adds a new types to handle 64-bit. Etc.)


I think the attitude in the DSP market has been to optimize for a
single family of DSPs and stick with it until something forces a
change to a new architecture, at which point one does the work over
again. Sometimes one can leverage highly optimized vendor libraries
for common functions like math primitives, FFTs and common
filters. And in fact reengineering the code to use an optimized
library might be a better way to go than translating the underlying
code. All depends on what the application does...


-Z-

Return to the comp.compilers page.
Search the comp.compilers archives again.