On Sep 18, 2:41 pm, steven.bossc...@
gmail.com wrote:
> On Sep 18, 4:13 pm, Arjen Markus
wrote:
>>> Perhaps you can describe in some detail how people can help out:
>> developing a compiler seems a rather daunting task and if you
>> can tell about the less daunting aspects and tasks you may get more
>> people involved.
>
> Well, daunting, daunting... Compiler construction as a (black) art is
> overrated ;-)
>
> You have to realize that not a single one of the gfortran developers
> (current and past, including g95) have a computer science background.
> At least, not as far as I know.
>
> All gfortran developers are physicists, engineers, mathematicians,
> etc., in their Real Life, just like most people that follow this
> newsgroup. These developers started as Fortran users, with no more
> knowledge of compiler construction than most people here.
>
> If they can learn to develop a compiler, then, certainly, you can do
> it too! All it takes is a willing brain and an itch to scratch, such
> as some bug that prevents you from compiling your program with
> gfortran.
>
> To illustrate this, let me give an example: One of the most productive
> developers of this moment was a Google Summer of Code student with
> little (if any) knowledge about gcc/gfortran internals and development
> processes. He only needed a couple of months to become a quite
> proficient gfortran developer.
>
> Tobias already said the rest...
>
> Gr.
> Steven
I too was daunted, but would love to contribute if I can come up to
speed.
I have about 15,000 lines of code that (at least to my way of
thinking) is very demanding in the following senses: lots of dynamic
memory allocation, lots of use of pointers, lots of defined operators
and assignments, deeply nested derived types (probably at least five
levels deep). The problem for me (which is an absolutely great
problem to have) is that gfortran runs the code perfectly, so I can't
start by reporting any bugs. On the contrary, compiling with gfortran
helped me to discover some very hard to trace bugs in my own code --
bugs that other compilers had not uncovered.
Since I can't report any new bugs, I'm glad to volunteer to work on
the things that excite me the most. I would be very excited to see
the following implemented and will contribute if someone can give me
specific guidance on where to start with these:
1. allocatable scalars
2. final procedures
3. polymorphic variables
4. type guard statements
These are the most important things I need from gfortran to get it to
compile a new code I'm working on. Of the compilers I've tried, it
currently only compiles with IBM XL Fortran.
Damian
