public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug fortran/93148] Pointer-function-result as LVALUE variable – function called multiple times with copy-in/out
       [not found] <bug-93148-4@http.gcc.gnu.org/bugzilla/>
@ 2021-10-04 19:49 ` burnus at gcc dot gnu.org
  2021-10-04 20:10 ` burnus at gcc dot gnu.org
  1 sibling, 0 replies; 2+ messages in thread
From: burnus at gcc dot gnu.org @ 2021-10-04 19:49 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93148

--- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Also when looking at gfc_conv_subref_array_arg – the function is called
seemingly up to 3 times – once per:

* gfc_conv_loop_setup (&loop, &expr->where);
   which calls gfc_conv_loop_setup → gfc_conv_loop_setup →
gfc_conv_procedure_call
   → gfc_trans_create_temp_array

* gfc_conv_expr (&lse, expr);

* gfc_conv_is_contiguous_expr (&cont_se, expr);
  (which calls: "gfc_conv_expr_descriptor (&array_se, expr);")

^ permalink raw reply	[flat|nested] 2+ messages in thread

* [Bug fortran/93148] Pointer-function-result as LVALUE variable – function called multiple times with copy-in/out
       [not found] <bug-93148-4@http.gcc.gnu.org/bugzilla/>
  2021-10-04 19:49 ` [Bug fortran/93148] Pointer-function-result as LVALUE variable – function called multiple times with copy-in/out burnus at gcc dot gnu.org
@ 2021-10-04 20:10 ` burnus at gcc dot gnu.org
  1 sibling, 0 replies; 2+ messages in thread
From: burnus at gcc dot gnu.org @ 2021-10-04 20:10 UTC (permalink / raw)
  To: gcc-bugs

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93148

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Additionally, it seems as if the contiguous check used for
gfortran.dg/contiguous_10.f90 currently unconditionally copies in the data –
which causes a problem when the pointer address escapes.

(Okay, whether that works is implementation defined per F2018:15.5.2.4
paragraph 9, but it is still a missed optimizion.)  – Lightly related to the
issue above as the issue shows up with copy-in/copy-out. Possible fix - which
causes a multi-call issue for gfortran.dg/contiguous_10.f90, see previous
comment:

--- a/gcc/fortran/trans-expr.c
+++ b/gcc/fortran/trans-expr.c
@@ -6528,5 +6528,7 @@ gfc_conv_procedure_call (gfc_se * se, gfc_symbol * sym,
                  gfc_conv_subref_array_arg (&parmse, e, nodesc_arg,
                                             fsym->attr.intent,
-                                            fsym->attr.pointer);
+                                            fsym->attr.pointer, fsym,
+                                            sym->name, NULL,
+                                            /* check_contiguous= */ true);
                }
              else

^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2021-10-04 20:10 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <bug-93148-4@http.gcc.gnu.org/bugzilla/>
2021-10-04 19:49 ` [Bug fortran/93148] Pointer-function-result as LVALUE variable – function called multiple times with copy-in/out burnus at gcc dot gnu.org
2021-10-04 20:10 ` 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).