public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug fortran/100991] New: [OpenMP] firstprivate for optional arguments is mishandled
@ 2021-06-09 11:48 burnus at gcc dot gnu.org
  2021-10-31 17:37 ` [Bug fortran/100991] " aldot at gcc dot gnu.org
  2023-05-12 14:25 ` burnus at gcc dot gnu.org
  0 siblings, 2 replies; 3+ messages in thread
From: burnus at gcc dot gnu.org @ 2021-06-09 11:48 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100991

            Bug ID: 100991
           Summary: [OpenMP] firstprivate for optional arguments is
                    mishandled
           Product: gcc
           Version: 12.0
            Status: UNCONFIRMED
          Keywords: openmp
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
  Target Milestone: ---

Split off from my defaultmap patch.

FAILING is either:

(A) FIRST TEST:
  libgomp/testsuite/libgomp.fortran/optional-map.f90
for the implicitly mapped
  subroutine sub2(ii, ...
    integer, optional :: ii, ...
but that requires the trans-openmp.c patch shown below (otherwise, ii is mapped
fromto)


(B) SECOND TEST

program main
 implicit none
 integer :: ii
 ii = 7
 call foo(ii)
 call foo()
 contains
subroutine foo (ii1)
  integer, optional :: ii1

  !$omp target firstprivate(ii1)
    if (present (ii1)) then
      if (ii1 /= 7) stop 1
      ii1 = 5
    end if
  !$omp end target
  if (present (ii1)) then
     if (ii1 /= 7) stop 1
  end if
end
end


Patch mentioned for (A):

gfc_omp_scalar_p (tree decl, bool ptr_alloc_ok)
 {
   tree type = TREE_TYPE (decl);
+  if (gfc_omp_is_optional_argument (decl))
+    type = TREE_TYPE (type);
   if (TREE_CODE (type) == REFERENCE_TYPE)
     type = TREE_TYPE (type);
   if (TREE_CODE (type) == POINTER_TYPE)

 * * * *

The code needed is similar to:

--- a/gcc/omp-low.c
+++ b/gcc/omp-low.c
@@ -12804,7 +12804,8 @@ lower_omp_target (gimple_stmt_iterator *gsi_p,
omp_context *ctx)
            if (omp_is_reference (ovar))
              type = TREE_TYPE (type);
            if ((INTEGRAL_TYPE_P (type)
-                && TYPE_PRECISION (type) <= POINTER_SIZE)
+                && TYPE_PRECISION (type) <= POINTER_SIZE
+                && !omp_check_optional_argument (ovar, false))
                || TREE_CODE (type) == POINTER_TYPE)
              {
                tkind = GOMP_MAP_FIRSTPRIVATE_INT;
@@ -13026,13 +13027,15 @@ lower_omp_target (gimple_stmt_iterator *gsi_p,
omp_context *ctx)
                || is_gimple_reg_type (TREE_TYPE (var)))
              {
                tree new_var = lookup_decl (var, ctx);
+               bool is_optional = omp_check_optional_argument (var, false);
                tree type;
                type = TREE_TYPE (var);
                if (omp_is_reference (var))
                  type = TREE_TYPE (type);
                if ((INTEGRAL_TYPE_P (type)
                     && TYPE_PRECISION (type) <= POINTER_SIZE)
-                   || TREE_CODE (type) == POINTER_TYPE)
+                   || TREE_CODE (type) == POINTER_TYPE
+                   || is_optional)
                  {
                    x = build_receiver_ref (var, false, ctx);
                    if (TREE_CODE (type) != POINTER_TYPE)
@@ -13040,11 +13043,14 @@ lower_omp_target (gimple_stmt_iterator *gsi_p,
omp_context *ctx)
                    x = fold_convert (type, x);
                    gimplify_expr (&x, &new_body, NULL, is_gimple_val,
                                   fb_rvalue);
-                   if (omp_is_reference (var))
+                   if (omp_is_reference (var) || is_optional)
                      {
+                       tree present = (is_optional
+                                       ? omp_check_optional_argument (ovar,
true)
+                                       : NULL_TREE);
...
(Cf. other code in this file.)

Except, of course, that this does not handle all the fun stuff which is not
trivially assignable. (allocatable arrays, polymorphic ...)

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [Bug fortran/100991] [OpenMP] firstprivate for optional arguments is mishandled
  2021-06-09 11:48 [Bug fortran/100991] New: [OpenMP] firstprivate for optional arguments is mishandled burnus at gcc dot gnu.org
@ 2021-10-31 17:37 ` aldot at gcc dot gnu.org
  2023-05-12 14:25 ` burnus at gcc dot gnu.org
  1 sibling, 0 replies; 3+ messages in thread
From: aldot at gcc dot gnu.org @ 2021-10-31 17:37 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100991

Bernhard Reutner-Fischer <aldot at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
                 CC|                            |aldot at gcc dot gnu.org
   Last reconfirmed|                            |2021-10-31

^ permalink raw reply	[flat|nested] 3+ messages in thread

* [Bug fortran/100991] [OpenMP] firstprivate for optional arguments is mishandled
  2021-06-09 11:48 [Bug fortran/100991] New: [OpenMP] firstprivate for optional arguments is mishandled burnus at gcc dot gnu.org
  2021-10-31 17:37 ` [Bug fortran/100991] " aldot at gcc dot gnu.org
@ 2023-05-12 14:25 ` burnus at gcc dot gnu.org
  1 sibling, 0 replies; 3+ messages in thread
From: burnus at gcc dot gnu.org @ 2023-05-12 14:25 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=100991

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |INVALID

--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Close as INVALID as firstprivate is no longer regarded as valid for absent
optional arguments in OpenMP

See PR 109831 for the remaining issues and what is valid.

^ permalink raw reply	[flat|nested] 3+ messages in thread

end of thread, other threads:[~2023-05-12 14:25 UTC | newest]

Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-06-09 11:48 [Bug fortran/100991] New: [OpenMP] firstprivate for optional arguments is mishandled burnus at gcc dot gnu.org
2021-10-31 17:37 ` [Bug fortran/100991] " aldot at gcc dot gnu.org
2023-05-12 14:25 ` burnus at gcc dot gnu.org

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).