[PATCH] Fix PR33302, make domwalk from the exit bb always do something

Richard Guenther rguenther@suse.de
Wed Sep 5 14:25:00 GMT 2007


This special-cases EXIT_BLOCK_PTR just like it does ENTRY_BLOCK_PTR for
testing of interestingness because of unreachability.  Which allows DSE
to do its work for functions that do not terminate.

Bootstrap and regtest on x86_64-unknown-linux-gnu in progress.

Richard.

2007-08-05  Richard Guenther  <rguenther@suse.de>

	PR tree-optimization/33302
	* domwalk.c (walk_dominator_tree): The exit block is
	interesting even if it is not reachable.

	* gcc.dg/tree-ssa/ssa-dse-11.c: New testcase.

Index: testsuite/gcc.dg/tree-ssa/ssa-dse-11.c
===================================================================
*** testsuite/gcc.dg/tree-ssa/ssa-dse-11.c	(revision 0)
--- testsuite/gcc.dg/tree-ssa/ssa-dse-11.c	(revision 0)
***************
*** 0 ****
--- 1,21 ----
+ /* { dg-do compile } */
+ /* { dg-options "-O -fdump-tree-dse1-details" } */
+ 
+ extern void abort(void);
+ void foo(int *p)
+ {
+   while (1)
+    {
+       *p = 0;
+       *p = 0;
+    }
+ }
+ void bar(int *p)
+ {
+   *p = 0;
+   *p = 0;
+   abort ();
+ }
+ 
+ /* { dg-final { scan-tree-dump-times "Deleted dead store" 2 "dse1" } } */
+ /* { dg-final { cleanup-tree-dump "dse1" } } */
Index: domwalk.c
===================================================================
*** domwalk.c	(revision 128123)
--- domwalk.c	(working copy)
*************** walk_dominator_tree (struct dom_walk_dat
*** 151,157 ****
    while (true)
      {
        /* Don't worry about unreachable blocks.  */
!       if (EDGE_COUNT (bb->preds) > 0 || bb == ENTRY_BLOCK_PTR)
  	{
  	  /* If block BB is not interesting to the caller, then none of the
  	     callbacks that walk the statements in BB are going to be
--- 151,159 ----
    while (true)
      {
        /* Don't worry about unreachable blocks.  */
!       if (EDGE_COUNT (bb->preds) > 0
! 	  || bb == ENTRY_BLOCK_PTR
! 	  || bb == EXIT_BLOCK_PTR)
  	{
  	  /* If block BB is not interesting to the caller, then none of the
  	     callbacks that walk the statements in BB are going to be



More information about the Gcc-patches mailing list