[PATCH, PR 42398] More generic type compatibility checking in sra_modify_expr

[Martin Jambor]

Hi,

the testcase in PR 42398 persuaded me that we need fully generic type
compatibility checking in sra_modify_expr rather than looking for
potential causes of incompatibilities in situations that are handled
by the function.  The downside is that the type conversion code leaves
the old aggregate around (it has to for TREE_ADDRESSABLE types) and we
will not be drawn to new causes of it being used by the assert if any
occur.

The alternative would be to add code generating new temporaries and
then an assignment from a VIEW_CONVERT_EXPRs but that hardly seems
worth it for assembly statements.

The patch leaves flag grp_different_types useless and so I removed it
so that it is not calculated needlessly.

I have successfully bootstrapped and regression tested the patch on
x86_64-linux.  OK for trunk?

Thanks,

Martin

2009-12-22  Martin Jambor  <mjambor@suse.cz>

	PR tree-optimization/42398
	* tree-sra.c (struct access): Removed flag grp_different_types.
	(dump_access): Do not dump the removed flag.
	(sort_and_splice_var_accesses): Do not set the removed flag.
	(sra_modify_expr): Check for type compatibility directly.

	* testsuite/gcc.c-torture/compile/pr42398.c: New test.


Index: mine/gcc/tree-sra.c
===================================================================
--- mine.orig/gcc/tree-sra.c
+++ mine/gcc/tree-sra.c
@@ -199,10 +199,6 @@ struct access
      BIT_FIELD_REF?  */
   unsigned grp_partial_lhs : 1;
 
-  /* Does this group contain accesses to different types? (I.e. through a union
-     or a similar mechanism).  */
-  unsigned grp_different_types : 1;
-
   /* Set when a scalar replacement should be created for this variable.  We do
      the decision and creation at different places because create_tmp_var
      cannot be called from within FOR_EACH_REFERENCED_VAR. */
@@ -343,14 +339,12 @@ dump_access (FILE *f, struct access *acc
     fprintf (f, ", grp_write = %d, grp_read = %d, grp_hint = %d, "
 	     "grp_covered = %d, grp_unscalarizable_region = %d, "
 	     "grp_unscalarized_data = %d, grp_partial_lhs = %d, "
-	     "grp_different_types = %d, grp_to_be_replaced = %d, "
-	     "grp_maybe_modified = %d, "
+	     "grp_to_be_replaced = %d, grp_maybe_modified = %d, "
 	     "grp_not_necessarilly_dereferenced = %d\n",
 	     access->grp_write, access->grp_read, access->grp_hint,
 	     access->grp_covered, access->grp_unscalarizable_region,
 	     access->grp_unscalarized_data, access->grp_partial_lhs,
-	     access->grp_different_types, access->grp_to_be_replaced,
-	     access->grp_maybe_modified,
+	     access->grp_to_be_replaced, access->grp_maybe_modified,
 	     access->grp_not_necessarilly_dereferenced);
   else
     fprintf (f, ", write = %d, grp_partial_lhs = %d\n", access->write,
@@ -1434,7 +1428,6 @@ sort_and_splice_var_accesses (tree var)
       bool grp_read = !access->write;
       bool multiple_reads = false;
       bool grp_partial_lhs = access->grp_partial_lhs;
-      bool grp_different_types = false;
       bool first_scalar = is_gimple_reg_type (access->type);
       bool unscalarizable_region = access->grp_unscalarizable_region;
 
@@ -1466,7 +1459,6 @@ sort_and_splice_var_accesses (tree var)
 		grp_read = true;
 	    }
 	  grp_partial_lhs |= ac2->grp_partial_lhs;
-	  grp_different_types |= !types_compatible_p (access->type, ac2->type);
 	  unscalarizable_region |= ac2->grp_unscalarizable_region;
 	  relink_to_new_repr (access, ac2);
 
@@ -1485,7 +1477,6 @@ sort_and_splice_var_accesses (tree var)
       access->grp_read = grp_read;
       access->grp_hint = multiple_reads;
       access->grp_partial_lhs = grp_partial_lhs;
-      access->grp_different_types = grp_different_types;
       access->grp_unscalarizable_region = unscalarizable_region;
       if (access->first_link)
 	add_access_to_work_queue (access);
@@ -2136,11 +2127,9 @@ sra_modify_expr (tree *expr, gimple_stmt
 
          We also want to use this when accessing a complex or vector which can
          be accessed as a different type too, potentially creating a need for
-         type conversion  (see PR42196).  */
-      if (!is_gimple_reg_type (type)
-	  || (access->grp_different_types
-	      && (TREE_CODE (type) == COMPLEX_TYPE
-		  || TREE_CODE (type) == VECTOR_TYPE)))
+         type conversion (see PR42196) and when scalarized unions are involved
+         in assembler statements (see PR42398).  */
+      if (!useless_type_conversion_p (type, access->type))
 	{
 	  tree ref = access->base;
 	  bool ok;
@@ -2171,10 +2160,7 @@ sra_modify_expr (tree *expr, gimple_stmt
 	    }
 	}
       else
-	{
-	  gcc_assert (useless_type_conversion_p (type, access->type));
-	  *expr = repl;
-	}
+	*expr = repl;
       sra_stats.exprs++;
     }
 
Index: mine/gcc/testsuite/gcc.c-torture/compile/pr42398.c
===================================================================
--- /dev/null
+++ mine/gcc/testsuite/gcc.c-torture/compile/pr42398.c
@@ -0,0 +1,6 @@
+int ptrace_setregs(void)
+{
+  union { unsigned int l; int t; } __gu_tmp;
+  __asm__ __volatile__("" : "=r" (__gu_tmp.l));
+  return __gu_tmp.t;
+}