public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed
* [Bug middle-end/20982] alignment attribute ignired for vector pointers types
       [not found] <bug-20982-10424@http.gcc.gnu.org/bugzilla/>
@ 2007-11-16 19:33 ` manu at gcc dot gnu dot org
  2009-02-07 20:47 ` manu at gcc dot gnu dot org
  1 sibling, 0 replies; 2+ messages in thread
From: manu at gcc dot gnu dot org @ 2007-11-16 19:33 UTC (permalink / raw)
  To: gcc-bugs



------- Comment #2 from manu at gcc dot gnu dot org  2007-11-16 19:32 -------
(In reply to comment #1)
> aligned can never decrease the required alignment.
> 

Can we detect this and emit a diagnostic? We do it for declarations already but
not for types. I tried the following patch as a wild guess but it generates
regressions in gcc.dg/compat/struct-layout-1 and similar.

Index: gcc/testsuite/gcc.dg/pr20982.c
===================================================================
--- gcc/testsuite/gcc.dg/pr20982.c      (revision 0)
+++ gcc/testsuite/gcc.dg/pr20982.c      (revision 0)
@@ -0,0 +1,11 @@
+/* PR 20982 alignment attribute ignored for vector pointers types */
+/* { dg-do compile } */
+/* { dg-options "-msse " } */
+typedef float v4sf __attribute__ ((mode(V4SF),aligned(8)));
+/* { dg-warning "specifying vector types with __attribute__ ..mode.. is
deprecated" } "" { target *-*-* } 4 }*/
+/* { dg-warning " use .* instead" "" { target *-*-* } 4 } */
+/* { dg-error "alignment for .* must be at least" "" { target *-*-* } 4 } */
+
+v4sf foo(v4sf *p, v4sf *q) {
+  return *p + *q;
+}
Index: gcc/c-common.c
===================================================================
--- gcc/c-common.c      (revision 130174)
+++ gcc/c-common.c      (working copy)
@@ -5438,23 +5438,38 @@
     }
   else if (is_type)
     {
-      /* If we have a TYPE_DECL, then copy the type, so that we
-        don't accidentally modify a builtin type.  See pushdecl.  */
-      if (decl && TREE_TYPE (decl) != error_mark_node
-         && DECL_ORIGINAL_TYPE (decl) == NULL_TREE)
-       {
-         tree tt = TREE_TYPE (decl);
-         *type = build_variant_type_copy (*type);
-         DECL_ORIGINAL_TYPE (decl) = tt;
-         TYPE_NAME (*type) = decl;
-         TREE_USED (*type) = TREE_USED (decl);
-         TREE_TYPE (decl) = *type;
-       }
-      else if (!(flags & (int) ATTR_FLAG_TYPE_IN_PLACE))
-       *type = build_variant_type_copy (*type);
-
-      TYPE_ALIGN (*type) = (1 << i) * BITS_PER_UNIT;
-      TYPE_USER_ALIGN (*type) = 1;
+      if (TYPE_ALIGN (*type) > (unsigned) (1 << i) * BITS_PER_UNIT)
+        {
+          if (TYPE_USER_ALIGN (*type))
+            error ("alignment for %qT was previously specified as %d "
+                   "and may not be decreased", *type,
+                   TYPE_ALIGN (*type) / BITS_PER_UNIT);
+          else
+            error ("alignment for %qT must be at least %d", *type,
+                   TYPE_ALIGN (*type) / BITS_PER_UNIT);
+          *no_add_attrs = true;
+        }
+      else
+        {
+          /* If we have a TYPE_DECL, then copy the type, so that we
+             don't accidentally modify a builtin type.  See pushdecl.  */
+          if (decl && TREE_TYPE (decl) != error_mark_node
+              && DECL_ORIGINAL_TYPE (decl) == NULL_TREE)
+            {
+              tree tt = TREE_TYPE (decl);
+              *type = build_variant_type_copy (*type);
+              DECL_ORIGINAL_TYPE (decl) = tt;
+              TYPE_NAME (*type) = decl;
+              TREE_USED (*type) = TREE_USED (decl);
+              TREE_TYPE (decl) = *type;
+            }
+          else if (!(flags & (int) ATTR_FLAG_TYPE_IN_PLACE))
+            *type = build_variant_type_copy (*type);
+
+
+          TYPE_ALIGN (*type) = (1 << i) * BITS_PER_UNIT;
+          TYPE_USER_ALIGN (*type) = 1;
+        }
     }
   else if (! VAR_OR_FUNCTION_DECL_P (decl)
           && TREE_CODE (decl) != FIELD_DECL)


-- 

manu at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |manu at gcc dot gnu dot org
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2007-11-16 19:32:54
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20982


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

* [Bug middle-end/20982] alignment attribute ignired for vector pointers types
       [not found] <bug-20982-10424@http.gcc.gnu.org/bugzilla/>
  2007-11-16 19:33 ` [Bug middle-end/20982] alignment attribute ignired for vector pointers types manu at gcc dot gnu dot org
@ 2009-02-07 20:47 ` manu at gcc dot gnu dot org
  1 sibling, 0 replies; 2+ messages in thread
From: manu at gcc dot gnu dot org @ 2009-02-07 20:47 UTC (permalink / raw)
  To: gcc-bugs



------- Comment #3 from manu at gcc dot gnu dot org  2009-02-07 20:46 -------
This is a missing diagnostic.


-- 

manu at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |diagnostic
   Last reconfirmed|2007-11-16 19:32:54         |2009-02-07 20:46:55
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=20982


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

end of thread, other threads:[~2009-02-07 20:47 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
     [not found] <bug-20982-10424@http.gcc.gnu.org/bugzilla/>
2007-11-16 19:33 ` [Bug middle-end/20982] alignment attribute ignired for vector pointers types manu at gcc dot gnu dot org
2009-02-07 20:47 ` manu at gcc dot gnu dot 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).