[PATCH] tree-optimization/101120 - fix compile-time issue with SLP groups

[PATCH] tree-optimization/101120 - fix compile-time issue with SLP groups

[Richard Biener]

This places two hacks to avoid an old compile-time issue when
vectorizing large permuted SLP groups with gaps where we end up
emitting loads and IV adjustments for the gap as well and those
have quite a high cost until they are eventually cleaned up.

The first hack is to fold the auto-inc style IV updates early
in the vectorizer rather than in the next forwprop pass which
shortens the SSA use-def chains of the used IV.

The second hack is to remove the unused loads after we've picked
all that we possibly use.

Bootstrap / regtest running on x86_64-unknown-linux-gnu.

I wonder if this is too gross (and I have to check the one or two
bug duplicates), but it should be at least easy to backport ...

Thanks,
Richard.

2021-06-18  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/101120
	* tree-vect-data-refs.c (bump_vector_ptr): Fold the
	built increment.
	* tree-vect-stmts.c (vectorizable_load): Remove unused
	loads in the DR chain for SLP.
---
 gcc/tree-vect-data-refs.c | 12 +++++++++++-
 gcc/tree-vect-stmts.c     | 12 ++++++++++++
 2 files changed, 23 insertions(+), 1 deletion(-)

diff --git a/gcc/tree-vect-data-refs.c b/gcc/tree-vect-data-refs.c
index bb086c6ac1c..be067c8923b 100644
--- a/gcc/tree-vect-data-refs.c
+++ b/gcc/tree-vect-data-refs.c
@@ -53,6 +53,7 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-hash-traits.h"
 #include "vec-perm-indices.h"
 #include "internal-fn.h"
+#include "gimple-fold.h"
 
 /* Return true if load- or store-lanes optab OPTAB is implemented for
    COUNT vectors of type VECTYPE.  NAME is the name of OPTAB.  */
@@ -5026,7 +5027,7 @@ bump_vector_ptr (vec_info *vinfo,
   struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
   tree vectype = STMT_VINFO_VECTYPE (stmt_info);
   tree update = TYPE_SIZE_UNIT (vectype);
-  gassign *incr_stmt;
+  gimple *incr_stmt;
   ssa_op_iter iter;
   use_operand_p use_p;
   tree new_dataref_ptr;
@@ -5041,6 +5042,15 @@ bump_vector_ptr (vec_info *vinfo,
   incr_stmt = gimple_build_assign (new_dataref_ptr, POINTER_PLUS_EXPR,
 				   dataref_ptr, update);
   vect_finish_stmt_generation (vinfo, stmt_info, incr_stmt, gsi);
+  /* Fold the increment, avoiding excessive chains use-def chains of
+     those, leading to compile-time issues for passes until the next
+     forwprop pass which would do this as well.  */
+  gimple_stmt_iterator fold_gsi = gsi_for_stmt (incr_stmt);
+  if (fold_stmt (&fold_gsi, follow_all_ssa_edges))
+    {
+      incr_stmt = gsi_stmt (fold_gsi);
+      update_stmt (incr_stmt);
+    }
 
   /* Copy the points-to information if it exists. */
   if (DR_PTR_INFO (dr))
diff --git a/gcc/tree-vect-stmts.c b/gcc/tree-vect-stmts.c
index eeef96a2eb6..1636e6716df 100644
--- a/gcc/tree-vect-stmts.c
+++ b/gcc/tree-vect-stmts.c
@@ -9765,6 +9765,18 @@ vectorizable_load (vec_info *vinfo,
 	  bool ok = vect_transform_slp_perm_load (vinfo, slp_node, dr_chain,
 						  gsi, vf, false, &n_perms);
 	  gcc_assert (ok);
+	  /* For SLP we know we've seen all possible uses of dr_chain.
+	     See to remove stmts we didn't need.
+	     ???  This is a hack to prevent compile-time issues as seen
+	     in PR101120 and friends.  */
+	  for (tree op : dr_chain)
+	    if (has_zero_uses (op))
+	      {
+		gimple *stmt = SSA_NAME_DEF_STMT (op);
+		gimple_stmt_iterator rgsi = gsi_for_stmt (stmt);
+		gsi_remove (&rgsi, true);
+		release_defs (stmt);
+	      }
         }
       else
         {
-- 
2.26.2

Re: [PATCH] tree-optimization/101120 - fix compile-time issue with SLP groups

[Richard Biener]

[-- Attachment #1: Type: text/plain, Size: 892 bytes --]

On Fri, Jun 18, 2021 at 2:23 PM Richard Biener <rguenther@suse.de> wrote:
>
> This places two hacks to avoid an old compile-time issue when
> vectorizing large permuted SLP groups with gaps where we end up
> emitting loads and IV adjustments for the gap as well and those
> have quite a high cost until they are eventually cleaned up.
>
> The first hack is to fold the auto-inc style IV updates early
> in the vectorizer rather than in the next forwprop pass which
> shortens the SSA use-def chains of the used IV.
>
> The second hack is to remove the unused loads after we've picked
> all that we possibly use.
>
> Bootstrap / regtest running on x86_64-unknown-linux-gnu.
>
> I wonder if this is too gross (and I have to check the one or two
> bug duplicates), but it should be at least easy to backport ...

Was apparently too simple - the following passes bootstrap and
regtest.

Richard.

[-- Attachment #2: p --]
[-- Type: application/octet-stream, Size: 6678 bytes --]

From 14d83407b265f1b5026c4d2db327d5cc58bda074 Mon Sep 17 00:00:00 2001
From: Richard Biener <rguenther@suse.de>
Date: Fri, 18 Jun 2021 14:07:00 +0200
Subject: [PATCH] tree-optimization/101120 - fix compile-time issue with SLP
 groups
To: gcc-patches@gcc.gnu.org

This places two hacks to avoid an old compile-time issue when
vectorizing large permuted SLP groups with gaps where we end up
emitting loads and IV adjustments for the gap as well and those
have quite a high cost until they are eventually cleaned up.

The first hack is to fold the auto-inc style IV updates early
in the vectorizer rather than in the next forwprop pass which
shortens the SSA use-def chains of the used IV.

The second hack is to remove the unused loads after we've picked
all that we possibly use.

2021-06-18  Richard Biener  <rguenther@suse.de>

	PR tree-optimization/101120
	* tree-vect-data-refs.c (bump_vector_ptr): Fold the
	built increment.
	* tree-vect-slp.c (vect_transform_slp_perm_load): Add
	DR chain DCE capability.
	* tree-vectorizer.h (vect_transform_slp_perm_load): Adjust.
	* tree-vect-stmts.c (vectorizable_load): Remove unused
	loads in the DR chain for SLP.
---
 gcc/tree-vect-data-refs.c | 12 +++++++++++-
 gcc/tree-vect-slp.c       | 31 ++++++++++++++++++++++++++-----
 gcc/tree-vect-stmts.c     |  7 ++++++-
 gcc/tree-vectorizer.h     |  2 +-
 4 files changed, 44 insertions(+), 8 deletions(-)

diff --git a/gcc/tree-vect-data-refs.c b/gcc/tree-vect-data-refs.c
index bb086c6ac1c..be067c8923b 100644
--- a/gcc/tree-vect-data-refs.c
+++ b/gcc/tree-vect-data-refs.c
@@ -53,6 +53,7 @@ along with GCC; see the file COPYING3.  If not see
 #include "tree-hash-traits.h"
 #include "vec-perm-indices.h"
 #include "internal-fn.h"
+#include "gimple-fold.h"
 
 /* Return true if load- or store-lanes optab OPTAB is implemented for
    COUNT vectors of type VECTYPE.  NAME is the name of OPTAB.  */
@@ -5026,7 +5027,7 @@ bump_vector_ptr (vec_info *vinfo,
   struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
   tree vectype = STMT_VINFO_VECTYPE (stmt_info);
   tree update = TYPE_SIZE_UNIT (vectype);
-  gassign *incr_stmt;
+  gimple *incr_stmt;
   ssa_op_iter iter;
   use_operand_p use_p;
   tree new_dataref_ptr;
@@ -5041,6 +5042,15 @@ bump_vector_ptr (vec_info *vinfo,
   incr_stmt = gimple_build_assign (new_dataref_ptr, POINTER_PLUS_EXPR,
 				   dataref_ptr, update);
   vect_finish_stmt_generation (vinfo, stmt_info, incr_stmt, gsi);
+  /* Fold the increment, avoiding excessive chains use-def chains of
+     those, leading to compile-time issues for passes until the next
+     forwprop pass which would do this as well.  */
+  gimple_stmt_iterator fold_gsi = gsi_for_stmt (incr_stmt);
+  if (fold_stmt (&fold_gsi, follow_all_ssa_edges))
+    {
+      incr_stmt = gsi_stmt (fold_gsi);
+      update_stmt (incr_stmt);
+    }
 
   /* Copy the points-to information if it exists. */
   if (DR_PTR_INFO (dr))
diff --git a/gcc/tree-vect-slp.c b/gcc/tree-vect-slp.c
index e2e5a54a87b..396b2bee4d8 100644
--- a/gcc/tree-vect-slp.c
+++ b/gcc/tree-vect-slp.c
@@ -6317,14 +6317,15 @@ vect_get_slp_defs (vec_info *,
    If ANALYZE_ONLY is TRUE, only check that it is possible to create valid
    permute statements for the SLP node NODE.  Store the number of vector
    permute instructions in *N_PERMS and the number of vector load
-   instructions in *N_LOADS.  */
+   instructions in *N_LOADS.  If DCE_CHAIN is true, remove all definitions
+   that were not needed.  */
 
 bool
 vect_transform_slp_perm_load (vec_info *vinfo,
 			      slp_tree node, vec<tree> dr_chain,
 			      gimple_stmt_iterator *gsi, poly_uint64 vf,
 			      bool analyze_only, unsigned *n_perms,
-			      unsigned int *n_loads)
+			      unsigned int *n_loads, bool dce_chain)
 {
   stmt_vec_info stmt_info = SLP_TREE_SCALAR_STMTS (node)[0];
   int vec_index = 0;
@@ -6403,6 +6404,7 @@ vect_transform_slp_perm_load (vec_info *vinfo,
     }
   auto_sbitmap used_in_lanes (in_nlanes);
   bitmap_clear (used_in_lanes);
+  auto_bitmap used_defs;
 
   unsigned int count = mask.encoded_nelts ();
   mask.quick_grow (count);
@@ -6510,11 +6512,20 @@ vect_transform_slp_perm_load (vec_info *vinfo,
 					       mask_vec);
 		      vect_finish_stmt_generation (vinfo, stmt_info, perm_stmt,
 						   gsi);
+		      if (dce_chain)
+			{
+			  bitmap_set_bit (used_defs, first_vec_index + ri);
+			  bitmap_set_bit (used_defs, second_vec_index + ri);
+			}
 		    }
 		  else
-		    /* If mask was NULL_TREE generate the requested
-		       identity transform.  */
-		    perm_stmt = SSA_NAME_DEF_STMT (first_vec);
+		    {
+		      /* If mask was NULL_TREE generate the requested
+			 identity transform.  */
+		      perm_stmt = SSA_NAME_DEF_STMT (first_vec);
+		      if (dce_chain)
+			bitmap_set_bit (used_defs, first_vec_index + ri);
+		    }
 
 		  /* Store the vector statement in NODE.  */
 		  SLP_TREE_VEC_STMTS (node)[vect_stmts_counter++] = perm_stmt;
@@ -6554,6 +6565,16 @@ vect_transform_slp_perm_load (vec_info *vinfo,
 	}
     }
 
+  if (dce_chain)
+    for (unsigned i = 0; i < dr_chain.length (); ++i)
+      if (!bitmap_bit_p (used_defs, i))
+	{
+	  gimple *stmt = SSA_NAME_DEF_STMT (dr_chain[i]);
+	  gimple_stmt_iterator rgsi = gsi_for_stmt (stmt);
+	  gsi_remove (&rgsi, true);
+	  release_defs (stmt);
+	}
+
   return true;
 }
 
diff --git a/gcc/tree-vect-stmts.c b/gcc/tree-vect-stmts.c
index 28d9984318f..d95e359daae 100644
--- a/gcc/tree-vect-stmts.c
+++ b/gcc/tree-vect-stmts.c
@@ -9762,8 +9762,13 @@ vectorizable_load (vec_info *vinfo,
       if (slp_perm)
         {
 	  unsigned n_perms;
+	  /* For SLP we know we've seen all possible uses of dr_chain so
+	     direct vect_transform_slp_perm_load to DCE the unused parts.
+	     ???  This is a hack to prevent compile-time issues as seen
+	     in PR101120 and friends.  */
 	  bool ok = vect_transform_slp_perm_load (vinfo, slp_node, dr_chain,
-						  gsi, vf, false, &n_perms);
+						  gsi, vf, false, &n_perms,
+						  nullptr, true);
 	  gcc_assert (ok);
         }
       else
diff --git a/gcc/tree-vectorizer.h b/gcc/tree-vectorizer.h
index b9824623ad9..fa28336d429 100644
--- a/gcc/tree-vectorizer.h
+++ b/gcc/tree-vectorizer.h
@@ -2012,7 +2012,7 @@ extern void vect_free_slp_instance (slp_instance);
 extern bool vect_transform_slp_perm_load (vec_info *, slp_tree, vec<tree>,
 					  gimple_stmt_iterator *, poly_uint64,
 					  bool, unsigned *,
-					  unsigned * = nullptr);
+					  unsigned * = nullptr, bool = false);
 extern bool vect_slp_analyze_operations (vec_info *);
 extern void vect_schedule_slp (vec_info *, vec<slp_instance>);
 extern opt_result vect_analyze_slp (vec_info *, unsigned);
-- 
2.26.2

Re: [PATCH] tree-optimization/101120 - fix compile-time issue with SLP groups

[Richard Biener]

On Fri, Jun 18, 2021 at 4:24 PM Richard Biener
<richard.guenther@gmail.com> wrote:
>
> On Fri, Jun 18, 2021 at 2:23 PM Richard Biener <rguenther@suse.de> wrote:
> >
> > This places two hacks to avoid an old compile-time issue when
> > vectorizing large permuted SLP groups with gaps where we end up
> > emitting loads and IV adjustments for the gap as well and those
> > have quite a high cost until they are eventually cleaned up.
> >
> > The first hack is to fold the auto-inc style IV updates early
> > in the vectorizer rather than in the next forwprop pass which
> > shortens the SSA use-def chains of the used IV.
> >
> > The second hack is to remove the unused loads after we've picked
> > all that we possibly use.
> >
> > Bootstrap / regtest running on x86_64-unknown-linux-gnu.
> >
> > I wonder if this is too gross (and I have to check the one or two
> > bug duplicates), but it should be at least easy to backport ...
>
> Was apparently too simple - the following passes bootstrap and
> regtest.

I've pushed this now after thinking about better solutions.

Richard.

>
> Richard.