Ivan Godard <ivan@millcomputing.com> schrieb:

I have cross-posted this to comp.lang.fortran, maybe a volunteer will
step up there :-)

> We don't have the resources to add Fortan to the development list on our
> own. However, we'd welcome a Fortran expert to join the team part time
> and try it out using flang and LLVM. Hint, hint...

Thanks for the implied offer :-) but I think I will have to restrict
myself a bit there. I know gfortran fairly well, but getting into
flang (especially the old compiler, which is the only one that
produces workable code at the moment) is a bit more than I can do.

I can certainly provide you with *.ll files for the Polyhedron
benchmarks, for example, but the recipe for that is very simple:
Download the Polyhedron benchmarks (easy to google), unpack, cd to the
source directoy for Linux, run "flang -S -O2 -emit-llvm fatigue2.f90" (for
example), and you're set. Any of the *.f90 files are self-contained,
so you can inspect them and see what makes sense (which was enough
for me to submit a couple of github issues for My 66000).

As for running, this would require building the library, and I
have no experience at all setting up flang, which I believe is
separate from the mainline LLVM compiler.

However, if you get a gcc port running (you previously mentioned
that any compiler could, in principle, generate genAsm), I will
be happy to assist you in all matters gfortran. The same offer
also goes Mitch's My 66000, of course. I have a fondness for
unconventional architectures, which is why I hang out on comp.arch,
and which is also why I helped IBM in adapting gfortran on POWER to
IEEE 128-bit reals recently, which will be in the upcoming gcc 12.

On 3/1/2022 9:43 AM, Thomas Koenig wrote:
> Ivan Godard <ivan@millcomputing.com> schrieb:
>
> I have cross-posted this to comp.lang.fortran, maybe a volunteer will
> step up there :-)
>
>> We don't have the resources to add Fortan to the development list on our
>> own. However, we'd welcome a Fortran expert to join the team part time
>> and try it out using flang and LLVM. Hint, hint...
>
> Thanks for the implied offer :-) but I think I will have to restrict
> myself a bit there. I know gfortran fairly well, but getting into
> flang (especially the old compiler, which is the only one that
> produces workable code at the moment) is a bit more than I can do.
>
> I can certainly provide you with *.ll files for the Polyhedron
> benchmarks, for example, but the recipe for that is very simple:
> Download the Polyhedron benchmarks (easy to google), unpack, cd to the
> source directoy for Linux, run "flang -S -O2 -emit-llvm fatigue2.f90" (for
> example), and you're set. Any of the *.f90 files are self-contained,
> so you can inspect them and see what makes sense (which was enough
> for me to submit a couple of github issues for My 66000).
>
> As for running, this would require building the library, and I
> have no experience at all setting up flang, which I believe is
> separate from the mainline LLVM compiler.
>
> However, if you get a gcc port running (you previously mentioned
> that any compiler could, in principle, generate genAsm), I will
> be happy to assist you in all matters gfortran. The same offer
> also goes Mitch's My 66000, of course. I have a fondness for
> unconventional architectures, which is why I hang out on comp.arch,
> and which is also why I helped IBM in adapting gfortran on POWER to
> IEEE 128-bit reals recently, which will be in the upcoming gcc 12.

Someday, but not until more talent or more money :-)

On Tuesday, March 1, 2022 at 10:43:34 AM UTC-7, Thomas Koenig wrote:

> and which is also why I helped IBM in adapting gfortran on POWER to
> IEEE 128-bit reals recently, which will be in the upcoming gcc 12.

Thank you very much. While I don't have a computer with a PowerPC
processor, I'm sure that this will be very helpful to many.

John Savard