On Thu, Dec 2, 2010 at 12:03 AM, Jon Harrop <
jonathandeanharrop@googlemail.com> wrote:
>
> >
> > As does assembler, so even more reasons to emit assembler?
>
> LLVM makes it a *lot* easier to generate efficient code, of course.


Just like the way gcc is using a hierarchy of program representations from
high-level to low-level, an ocaml compiler could in principle perform type
and language specific optimization passes in the high-level (cmm for
instance) and then emit llvm code, without any loss of efficiency, I think.

Still, I have to agree with other posters stating that a compiler backend
written in ocaml is preferable. I agree, if not only for the reason that I
know from experience it would be 10 times easier to write in ocaml some of
the passes I have had to deal with in C++. Even working with symbols is
quite awkward in C++. While on the other hand, libraries like boost can help
at times, but it is nothing we cannot replicate in ocaml. As a compiler
developer, I would be most intrigued in profiling and static analysis
information represented in ocaml.

I don't see why hybrids are a bad idea, though. If you see LLVM becoming an
industry standard, then, of course it's worth supporting. It means you'll
target new architectures with no effort.

Best,

-- 
Eray Ozkural, PhD candidate.  Comp. Sci. Dept., Bilkent University, Ankara
http://groups.yahoo.com/group/ai-philosophy
http://myspace.com/arizanesil http://myspace.com/malfunct