This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [Patch, Fortran] PR56907 - do not 'pack' arrays passed to C_LOC

Janus Weil:
What I don't quite understand is:
@@ -6317,8 +6317,13 @@ conv_isocbinding_function (gfc_se *se, gfc_expr *expr)
        if (arg->expr->rank == 0)
      gfc_conv_expr_reference (se, arg->expr);
-      else
+      else if (gfc_is_simply_contiguous (arg->expr, false))
      gfc_conv_array_parameter (se, arg->expr, true, NULL, NULL, NULL);
+      else
+    {
+      gfc_conv_expr_descriptor (se, arg->expr);
+      se->expr = gfc_conv_descriptor_data_get (se->expr);
+    }

Why doesn't 'gfc_conv_array_parameter' handle this situation properly?

Well, it does: As it doesn't know whether the array is contiguous or not - it packs the array. That's what one usually wants.

However, for C_LOC one knows that the array is contiguous - the user promises this to the compiler - and, hence, no packing is needed.

After all there is code like this inside:
   no_pack = (...)
          gfc_is_simply_contiguous (expr, false));
Wouldn't it be better to fix the logic in there, instead of avoiding
to call it alltogether? This might also help in other situations ...

I don't know of any other situation where the *user* guarantees that the argument is contiguous while the compiler cannot check this. And even if the user lied to the compiler - doing no packing gives at least the address of the first element instead of the address of an immediately feed temporary array.

(Actually, there are other places in the compiler, where the user guarantees this: Pointer assignment to a "contiguous" pointer; there, the user promises that the target is contiguous. Maybe there are other spots as well. On the other hand, if the dummy argument has the CONTIGUOUS attribute, an automatic copy-in/out is required if it is not contiguous.)



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]