On Sun, Mar 30, 2025 at 02:06:31PM +0000, Yuao Ma wrote:
>
> I have uploaded a draft proposal for the 'Fortran – 2018/202x' project to the
> platform. You can also find a PDF version at this Google Drive link:
> https://drive.google.com/file/d/1cSRPuWpXWHMwHdNItQluEYspLfylqZcU.
I rarely follow random links. I did, however, take a a very
brief look at the doc.
> Could you please review the proposal and provide feedback on how it can be
> improved? Thank you for your time and advice.
The technical description is somewhat muddled. First, I think
you should chose either the half-cycle trig functions or the token
manipulation functions not both. For either, you will need to
edit several files.
In gcc/gcc/fortran:
gfortran.h: define prototypes and GFC_ISYM_*.
check.cc: write functions for argument checking and -std=f2023.
intrinsic.texi: document the new functions.
intrinsic.h: prototypes
intrinsic.cc: Add functions to table of intrinsic subprograms
iresolve.cc: Resolution of function names/args to libgfortran names
simplify.cc: Constant folding (using MPFR for half-cyclei trig).
trans-intrinsic.cc: Need to ensure the right functions is called
and possibly in-line the function.
In gcc/libgfortran:
gfortran.map: Add symbols for symbol versioning.
configure.ac: Add checks for HAVE_SINPIF, etc.
intrinsics/trigpi.cc: Implementation details of half-cycle
intrinsics/split.cc: Implementation details of SPLIT()
intrinsics/tokenize.cc: Implementation details of TOKENIZE()
Again, if you choose the half-cycle trig function, look at the
bug report where I have already done some of the work.
--
Steve
