loops: Invoke lim after successful loop interchange

This patch makes the entry point to loop invariant motion public, so
that it can be called after loop interchange when that pass has
swapped loops.  This avoids the non-LTO -Ofast run-time regressions of
410.bwaves and 503.bwaves_r (which are 19% and 15% faster than current
master on an AMD zen2 machine) while not introducing a full LIM pass
into the pass pipeline.

The patch also adds a parameter which allows not to perform any store
motion so that it is not done after an interchange.

gcc/ChangeLog:

2020-11-12  Martin Jambor  <mjambor@suse.cz>

	PR tree-optimization/94406
	* tree-ssa-loop-im.c (tree_ssa_lim): Renamed to
	loop_invariant_motion_in_fun, added a parameter to control store
	motion.
	(pass_lim::execute): Adjust call to tree_ssa_lim, now
	loop_invariant_motion_in_fun.
	* tree-ssa-loop-manip.h (loop_invariant_motion_in_fun): Declare.
	* gimple-loop-interchange.cc (pass_linterchange::execute): Call
	loop_invariant_motion_in_fun if any interchange has been done.

diff --git a/gcc/gimple-loop-interchange.cc b/gcc/gimple-loop-interchange.cc
index 1656004ecf0b6bc325069675f9724fde8a9fb8a9..a36dbb49b1fa1c59bf41b6feb6e96fe4d4125912 100644 (file)
--- a/gcc/gimple-loop-interchange.cc
+++ b/gcc/gimple-loop-interchange.cc
@@ -2085,8 +2085,13 @@ pass_linterchange::execute (function *fun)
     }
 
   if (changed_p)
-    scev_reset ();
-  return changed_p ? (TODO_update_ssa_only_virtuals) : 0;
+    {
+      unsigned todo = TODO_update_ssa_only_virtuals;
+      todo |= loop_invariant_motion_in_fun (cfun, false);
+      scev_reset ();
+      return todo;
+    }
+  return 0;
 }
 
 } // anon namespace

diff --git a/gcc/tree-ssa-loop-im.c b/gcc/tree-ssa-loop-im.c
index 6bb07e133cdb316adc09f17d0d5a4b9e1b7adc41..3c7412737f02e0d5b80bd26a5c33e5d18307718b 100644 (file)
--- a/gcc/tree-ssa-loop-im.c
+++ b/gcc/tree-ssa-loop-im.c
@@ -3089,10 +3089,11 @@ tree_ssa_lim_finalize (void)
 }
 
 /* Moves invariants from loops.  Only "expensive" invariants are moved out --
-   i.e. those that are likely to be win regardless of the register pressure.  */
+   i.e. those that are likely to be win regardless of the register pressure.
+   Only perform store motion if STORE_MOTION is true.  */
 
-static unsigned int
-tree_ssa_lim (function *fun)
+unsigned int
+loop_invariant_motion_in_fun (function *fun, bool store_motion)
 {
   unsigned int todo = 0;
 
@@ -3114,7 +3115,8 @@ tree_ssa_lim (function *fun)
 
   /* Execute store motion.  Force the necessary invariants to be moved
      out of the loops as well.  */
-  do_store_motion ();
+  if (store_motion)
+    do_store_motion ();
 
   free (rpo);
   rpo = XNEWVEC (int, last_basic_block_for_fn (fun));
@@ -3175,7 +3177,7 @@ pass_lim::execute (function *fun)
 
   if (number_of_loops (fun) <= 1)
     return 0;
-  unsigned int todo = tree_ssa_lim (fun);
+  unsigned int todo = loop_invariant_motion_in_fun (fun, true);
 
   if (!in_loop_pipeline)
     loop_optimizer_finalize ();

diff --git a/gcc/tree-ssa-loop-manip.h b/gcc/tree-ssa-loop-manip.h
index e789e4fcb0b3dcb4e8dbbca417ae7db9eb9223b0..d8e918ef7c9177547e0287831c01b9c0968392a9 100644 (file)
--- a/gcc/tree-ssa-loop-manip.h
+++ b/gcc/tree-ssa-loop-manip.h
@@ -55,7 +55,7 @@ extern void tree_transform_and_unroll_loop (class loop *, unsigned,
 extern void tree_unroll_loop (class loop *, unsigned,
                              edge, class tree_niter_desc *);
 extern tree canonicalize_loop_ivs (class loop *, tree *, bool);
-
+extern unsigned int loop_invariant_motion_in_fun (function *, bool);
 
 
 #endif /* GCC_TREE_SSA_LOOP_MANIP_H */