On Feb 27, 2004, at 6:07 PM, Aleksandar Donev wrote:
> Drew McCormack wrote:
>> Any new language could simply
>> be built on top of the current fortran backend.
> I was talking about things like array statements, assumed-shape
> arrays, array
> pointers, allocatables, etc. Just ensuring that no temporary arrays are
> generated needlessly, for example, takes a certain sophistication. The
> backend is as Dan said commong to all languages, but the role of the
> syntax
> and how it is parsed to intermediate representations should not be
> underestimated.
>
The point I'm trying to make is that this work has already been done,
and the resulting code is possessed by current fortran compiler makers.
If they have written their compilers well, there should not be too
great a barrier to extracting the array processing code for a new
language compiler.
Drew
|