2013-03-26 Richard Biener <rguenther@suse.de>

	* tree-cfg.c (execute_build_cfg): Build the loop tree.
	(pass_build_cfg): Provide PROP_loops.
	(move_sese_region_to_fn): Remove loops that are outlined into fn
	for now.
	* tree-inline.c: Include cfgloop.h.
	(initialize_cfun): Do not drop PROP_loops.
	(copy_loops): New function.
	(copy_cfg_body): Copy loop structure.
	(tree_function_versioning): Initialize destination loop tree.
	* tree-ssa-loop.c (pass_tree_loop_init): Do not provide PROP_loops.
	(pass_parallelize_loops): Do IL verification.
	* loop-init.c (loop_optimizer_init): Fixup loops if required.
	* tree-optimize.c (execute_fixup_cfg): If we need to cleanup
	the CFG make sure we fixup loops as well.
	* tree-ssa-tail-merge.c: Include cfgloop.h.
	(replace_block_by): When merging loop latches mark loops for fixup.
	* lto-streamer-out.c (output_struct_function_base): Drop
	PROP_loops for now.
	* tree-ssa-phiopt.c: Include tree-scalar-evolution.h.
	(tree_ssa_cs_elim): Initialize the loop optimizer and SCEV.
	* ipa-split.c: Include cfgloop.h.
	(split_function): Add the new return block to the loop tree root.
	* tree-cfgcleanup.c (remove_forwarder_block_with_phi): Return
	whether we have removed the forwarder block.
	(merge_phi_nodes): If we removed a forwarder mark loops for fixup.
	* cfgloop.h (place_new_loop): Declare.
	* cfgloopmanip.c (place_new_loop): Export.
	* Makefile.in (asan.o): Add $(CFGLOOP_H) dependency.
	(tree-switch-conversion.o): Likewise.
	(tree-complex.o): Likewise.
	(tree-inline.o): Likewise.
	(tree-ssa-tailmerge.o): Likewise.
	(ipa-split.o): Likewise.
	(tree-ssa-phiopt.o): Add $(SCEV_H) dependency.
	(tree-ssa-copy.o): Likewise.
	* tree-switch-conversion.c: Include cfgloop.h
	(process_switch): If we emit a bit-test cascade, schedule loops
	for fixup.
	* tree-complex.c: Include cfgloop.h.
	(expand_complex_div_wide): Properly add new basic-blocks to loops.
	* asan.c: Include cfgloop.h.
	(create_cond_insert_point): Properly add new basic-blocks to
	loops, schedule loop fixup.
	* cfgloop.c (verify_loop_structure): Check that looks are not
	marked for fixup.
	* omp-low.c (expand_parallel_call): Properly add new basic-blocks
	to loops.
	(expand_omp_for_generic): Likewise.
	(expand_omp_sections): Likewise.
	(expand_omp_atomic_pipeline): Schedule loops for fixup.
	* tree-ssa-copy.c: Include tree-scalar-evolution.h.
	(fini_copy_prop): Disable DCE in substitute_and_fold if SCEV
	is initialized, not when loops are present.
	* tree-parloops.c (parallelize_loops): Remove checking here.
	* passes.c (init_optimization_passes): Schedule a copy-propagation
	pass before complete unrolling of inner loops.

	* gcc.dg/tree-prof/update-loopch.c: Revert last change.
	* gcc.dg/graphite/pr33766.c: Fix undefined behavior.
	* gcc.dg/pr53265.c: Remove XFAILs.
	* gcc.dg/tree-ssa/loop-38.c: Remove unreliable dump scanning.
	* gcc.dg/tree-ssa/pr21559.c: Change back to two expected jump threads.


git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@198333 138bc75d-0d04-0410-961f-82ee72b054a4

1 files changed, 12 insertions, 0 deletions

diff --git a/gcc/tree-ssa-tail-merge.c b/gcc/tree-ssa-tail-merge.c
index b20d3067d66..f2ab744403a 100644
--- a/gcc/tree-ssa-tail-merge.c
+++ b/gcc/tree-ssa-tail-merge.c
@@ -197,6 +197,7 @@ along with GCC; see the file COPYING3.  If not see
 #include "gimple-pretty-print.h"
 #include "tree-ssa-sccvn.h"
 #include "tree-dump.h"
+#include "cfgloop.h"
 
 /* ??? This currently runs as part of tree-ssa-pre.  Why is this not
    a stand-alone GIMPLE pass?  */
@@ -1459,6 +1460,17 @@ replace_block_by (basic_block bb1, basic_block bb2)
   /* Mark the basic block as deleted.  */
   mark_basic_block_deleted (bb1);
 
+  /* ???  If we merge the loop preheader with the loop latch we are creating
+     additional entries into the loop, eventually rotating it.
+     Mark loops for fixup in this case.
+     ???  This is a completely unwanted transform and will wreck most
+     loops at this point - but with just not considering loop latches as
+     merge candidates we fail to commonize the two loops in gcc.dg/pr50763.c.
+     A better fix to avoid that regression is needed.  */
+  if (current_loops
+      && bb2->loop_father->latch == bb2)
+    loops_state_set (LOOPS_NEED_FIXUP);
+
   /* Redirect the incoming edges of bb1 to bb2.  */
   for (i = EDGE_COUNT (bb1->preds); i > 0 ; --i)
     {