[Bug fortran/100989] New: Bogus internal VOLATILE attribute for ASYNCHRONOUS

[Bug fortran/100989] New: Bogus internal VOLATILE attribute for ASYNCHRONOUS

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100989

            Bug ID: 100989
           Summary: Bogus internal VOLATILE attribute for ASYNCHRONOUS
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
                CC: tkoenig at gcc dot gnu.org
  Target Milestone: ---

For a long time, there was the believe that due to GCC's pointer-escape
semantic, which follows C, no special attribute is needed for asynchronous:


  ...
  integer, asynchronous :: A
  call init_action(A)  ! pointer escape due to by-ref and GCC's C semantic 
  ...                  ! and due to 'asynchronous' due to Fortran semantic
  A = 5   ! (probably a bad idea if really asynchronous)
  ...
  call wait()   ! arrives here, can access A due to escaped pointer
  ! <-- value of 'A' could have changed

There might be an issue with the handling of the 'fn attr', though as the
variables have to be regarded as escaping for certain function calls.

In any case, VOLATILE prevents too much optimizations.

The volatile was added by commit:
  commit 2b4c90656132abb8b8ad155d345c7d4fbf1687c9
via
  https://gcc.gnu.org/pipermail/gcc-patches/2018-June/499870.html
  https://gcc.gnu.org/pipermail/gcc-patches/2018-August/503896.html
without any discussion of this point.

[Bug fortran/100989] Bogus internal VOLATILE attribute for ASYNCHRONOUS

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100989

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2021-06-09
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |WAITING

--- Comment #1 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
F2018, 8.5.4:

" An entity with the ASYNCHRONOUS attribute is a variable, and may be subject
to asynchronous input/output or asynchronous communication."

So, the variable can be changed without going through normal program
flow.  A pointer marked as escaping isn't enough, because there can
also be a change during code which does not invoke any other procedures.
(such as by an MPI call).

If volatile isn't the right model, what is?

[Bug fortran/100989] Bogus internal VOLATILE attribute for ASYNCHRONOUS

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100989

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |enhancement