Hi,

the CONSTRUCTOR_NO_CLEARING flag was invented to avoid generating a memset for 
CONSTRUCTORS that lack elements, but it turns out that the gimplifier can 
generate a memcpy for them instead, which is even worse performance-wise, so 
this prevents it from doing that for them.

Tested on x86-64/Linux, OK for the mainline?


2020-05-05  Eric Botcazou  <ebotca...@adacore.com>

        * gimplify.c (gimplify_init_constructor): Do not put the constructor
        into static memory if it is not complete.


2020-05-05  Eric Botcazou  <ebotca...@adacore.com>

        * gnat.dg/aggr29.adb: New test.

-- 
Eric Botcazou
diff --git a/gcc/gimplify.c b/gcc/gimplify.c
index 2f2c51b2d89..7807e30fb2e 100644
--- a/gcc/gimplify.c
+++ b/gcc/gimplify.c
@@ -4995,8 +4995,8 @@ gimplify_init_constructor (tree *expr_p, gimple_seq *pre_p, gimple_seq *post_p,
 	   all-zero initializers (which aren't big enough to merit
 	   clearing), and don't try to make bitwise copies of
 	   TREE_ADDRESSABLE types.  */
-
 	if (valid_const_initializer
+	    && complete_p
 	    && !(cleared || num_nonzero_elements == 0)
 	    && !TREE_ADDRESSABLE (type))
 	  {
-- { dg-do compile }
-- { dg-options "-fdump-tree-gimple" }

with Ada.Streams; use Ada.Streams;

procedure Aggr29 is
  A : aliased Stream_Element_Array := (1 .. 512 => <>);
begin
  null;
end;

-- { dg-final { scan-tree-dump-not "a___UNC = \\*" "gimple" } }

Reply via email to