On 06/25/2014 06:35 AM, Kai Tietz wrote:
> Hello,
>
> so there seems to be a fallout caused by moving peephole2 pass. See PR/61608.
> So we need indeed 2 peephole2 passes.
We don't need a second peephole pass. Please try this.
I think there's room for cleanup here, depending on when we leave cfglayout
mode for the last time.
r~
* bb-reorder.c (pass_duplicate_computed_gotos::execute): Cleanup
the cfg if there were any changes.
* passes.def: Revert move of peephole2 after reorder_blocks;
move duplicate_computed_gotos before peephole2.
diff --git a/gcc/bb-reorder.c b/gcc/bb-reorder.c
index 61b0cab..cd68fee 100644
--- a/gcc/bb-reorder.c
+++ b/gcc/bb-reorder.c
@@ -2520,13 +2520,20 @@ pass_duplicate_computed_gotos::execute (function *fun)
changed = true;
}
-done:
- /* Duplicating blocks above will redirect edges and may cause hot blocks
- previously reached by both hot and cold blocks to become dominated only
- by cold blocks. */
+ done:
if (changed)
- fixup_partitions ();
- cfg_layout_finalize ();
+ {
+ /* Duplicating blocks above will redirect edges and may cause hot
+ blocks previously reached by both hot and cold blocks to become
+ dominated only by cold blocks. */
+ fixup_partitions ();
+
+ /* Merge the duplicated blocks into predecessors, when possible. */
+ cfg_layout_finalize ();
+ cleanup_cfg (0);
+ }
+ else
+ cfg_layout_finalize ();
BITMAP_FREE (candidates);
return 0;
diff --git a/gcc/passes.def b/gcc/passes.def
index 280cf16..f13df6c 100644
--- a/gcc/passes.def
+++ b/gcc/passes.def
@@ -396,20 +396,13 @@ along with GCC; see the file COPYING3. If not see
NEXT_PASS (pass_rtl_dse2);
NEXT_PASS (pass_stack_adjustments);
NEXT_PASS (pass_jump2);
+ NEXT_PASS (pass_duplicate_computed_gotos);
+ NEXT_PASS (pass_peephole2);
NEXT_PASS (pass_if_after_reload);
NEXT_PASS (pass_regrename);
NEXT_PASS (pass_cprop_hardreg);
NEXT_PASS (pass_fast_rtl_dce);
- NEXT_PASS (pass_duplicate_computed_gotos);
NEXT_PASS (pass_reorder_blocks);
- /* We need to run peephole2 pass after the duplicate-
- compute-gotos and the reorder-blocks pass (PR/39284).
- We have a single indirect branch in the entire function
- before duplicate-compute-gotos pass. This vastly reduces
- the size of the CFG.
- For preventing to run peephole2 pass twice, its run after
- the jump2 got removed. */
- NEXT_PASS (pass_peephole2);
NEXT_PASS (pass_branch_target_load_optimize2);
NEXT_PASS (pass_leaf_regs);
NEXT_PASS (pass_split_before_sched2);
