The following tackles another source of slow bitmap operations,
namely populating blocks_to_update. We already have that in
tree view around PHI insertion but also the initial population is
slow. There's unfortunately a conditional inbetween list view
requirement and the bitmap API doesn't allow opportunistic
switching but rejects tree -> tree or list -> list transitions.
So the following patch wraps the early population in a tree view
section with possibly one redundant tree -> list -> tree view
transition.
This cuts tree SSA incremental from 228.25s (21%) to 65.05s (7%).
Bootstrapped and tested on x86_64-unknown-linux-gnu, pushed.
PR tree-optimization/114855
* tree-into-ssa.cc (update_ssa): Use tree view for the
initial population of blocks_to_update.
---
gcc/tree-into-ssa.cc | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/gcc/tree-into-ssa.cc b/gcc/tree-into-ssa.cc
index 1cce9d62809..fc61d47ca77 100644
--- a/gcc/tree-into-ssa.cc
+++ b/gcc/tree-into-ssa.cc
@@ -3445,6 +3445,7 @@ update_ssa (unsigned update_flags)
blocks_with_phis_to_rewrite = BITMAP_ALLOC (NULL);
bitmap_tree_view (blocks_with_phis_to_rewrite);
blocks_to_update = BITMAP_ALLOC (NULL);
+ bitmap_tree_view (blocks_to_update);
insert_phi_p = (update_flags != TODO_update_ssa_no_phi);
@@ -3492,6 +3493,8 @@ update_ssa (unsigned update_flags)
placement heuristics. */
prepare_block_for_update (start_bb, insert_phi_p);
+ bitmap_list_view (blocks_to_update);
+
tree name;
if (flag_checking)
@@ -3517,6 +3520,8 @@ update_ssa (unsigned update_flags)
}
else
{
+ bitmap_list_view (blocks_to_update);
+
/* Otherwise, the entry block to the region is the nearest
common dominator for the blocks in BLOCKS. */
start_bb = nearest_common_dominator_for_set (CDI_DOMINATORS,
--
2.43.0
