This is the mail archive of the gcc-patches@gcc.gnu.org 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]

[PATCH] PR tree-optimize/19042 Scalarize complex types


	The general consensus on IRC was that complex types should be
scalarized.  decide_block_copy was not checking for complex types and the
other heuristics sometimes missed them, so a patch is appended.

Okay for mainline?

David


	* tree-sra.c (decide_block_copy): Try to use element copy for
	complex types.

Index: tree-sra.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/tree-sra.c,v
retrieving revision 2.47
diff -c -p -r2.47 tree-sra.c
*** tree-sra.c	10 Dec 2004 21:54:42 -0000	2.47
--- tree-sra.c	17 Dec 2004 02:35:33 -0000
*************** decide_block_copy (struct sra_elt *elt)
*** 1316,1324 ****
  	     is much more of a factor.  */
  
  	  /* If the structure is small, and we've made copies, go ahead
! 	     and instantiate, hoping that the copies will go away.  */
! 	  if (full_size <= max_size
! 	      && elt->n_copies > elt->n_uses)
  	    use_block_copy = false;
  	  else
  	    {
--- 1316,1325 ----
  	     is much more of a factor.  */
  
  	  /* If the structure is small, and we've made copies, go ahead
! 	     and instantiate, hoping that the copies will go away.
! 	     Always instantiate complex types.  */
! 	  if ((full_size <= max_size && elt->n_copies > elt->n_uses)
! 	      || TREE_CODE (elt->type) == COMPLEX_TYPE)
  	    use_block_copy = false;
  	  else
  	    {


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