Re: [patch] Optimize empty class copies within a C++ return statement

[Aldy Hernandez]

On 03/06/2015 01:46 PM, Jason Merrill wrote:
On 03/06/2015 11:19 AM, Aldy Hernandez wrote:
We are hitting the MODIFY_EXPR case.  Indeed, _because_ we hit the
MODIFY_EXPR is that we return the uninitialized temporary.
For example, we start with:

     return retval = TARGET_EXPR <D.2347, ...>

which becomes:

     <stuff with &D.2347>
     return D.2349 = D.2347;

which the aforementioned MODIFY_EXPR case turns into:

     <stuff with &D.2347>
     return D.2349;

But doesn't this still involve a MODIFY_EXPR, i.e. return retval = D.2349?

If I understand you correct, no.

gimplify_return_expr creates a new temporary and uses that instead of 
<retval>:

  else if (gimplify_ctxp->return_temp)
    result = gimplify_ctxp->return_temp;
  else
    {
      result = create_tmp_reg (TREE_TYPE (result_decl));
      ...
    }
...
...
  /* Smash the lhs of the MODIFY_EXPR to the temporary we plan to use.
     Then gimplify the whole thing.  */
  if (result != result_decl)
    TREE_OPERAND (ret_expr, 0) = result;

Aldy