The following consistently keeps SLP_TREE_REPRESENTATIVE cleared
for VEC_PERM_EXPR SLP nodes and verifies this.
Bootstrapped and tested on x86_64-unknown-linux-gnu, now also
testing on aarch64-linux-gnu.
* tree-vect-slp.cc (vect_slp_build_two_operator_nodes):
Set SLP_TREE_REPRESENTATIVE to NULL for the permute nodes.
(vect_build_slp_tree_2): Likewise.
(vect_build_slp_store_interleaving): Likewise.
(vect_lower_load_permutations): Likewise.
(vect_optimize_slp_pass::get_result_with_layout): Likewise.
(vect_cse_slp_node_parts): Deal with NULL SLP_TREE_REPRESENTATIVE
for VEC_PERM nodes.
(vect_slp_analyze_node_operations_1): Assert permute nodes
do not have SLP_TREE_REPRESENTATIVE set.
* tree-vect-slp-patterns.cc (complex_pattern::build): Adjust.
---
gcc/tree-vect-slp-patterns.cc | 2 +-
gcc/tree-vect-slp.cc | 18 ++++++++++--------
2 files changed, 11 insertions(+), 9 deletions(-)
diff --git a/gcc/tree-vect-slp-patterns.cc b/gcc/tree-vect-slp-patterns.cc
index 8a03a25475d..233a4bad3e6 100644
--- a/gcc/tree-vect-slp-patterns.cc
+++ b/gcc/tree-vect-slp-patterns.cc
@@ -535,7 +535,7 @@ complex_pattern::build (vec_info *vinfo)
FOR_EACH_VEC_ELT (this->m_workset, ix, node)
{
/* Calculate the location of the statement in NODE to replace. */
- stmt_info = SLP_TREE_REPRESENTATIVE (node);
+ stmt_info = SLP_TREE_SCALAR_STMTS (node)[0];
gimple* old_stmt = STMT_VINFO_STMT (stmt_info);
tree lhs_old_stmt = gimple_get_lhs (old_stmt);
tree type = TREE_TYPE (lhs_old_stmt);
diff --git a/gcc/tree-vect-slp.cc b/gcc/tree-vect-slp.cc
index b2f14e6528a..5b602bcfc71 100644
--- a/gcc/tree-vect-slp.cc
+++ b/gcc/tree-vect-slp.cc
@@ -2095,8 +2095,7 @@ vect_slp_build_two_operator_nodes (slp_tree perm, tree
vectype,
SLP_TREE_CODE (perm) = VEC_PERM_EXPR;
SLP_TREE_VECTYPE (perm) = vectype;
SLP_TREE_LANES (perm) = group_size;
- /* ??? We should set this NULL but that's not expected. */
- SLP_TREE_REPRESENTATIVE (perm) = oper1;
+ SLP_TREE_REPRESENTATIVE (perm) = NULL;
SLP_TREE_LANE_PERMUTATION (perm) = lperm;
SLP_TREE_CHILDREN (perm).quick_push (child1);
SLP_TREE_CHILDREN (perm).quick_push (child2);
@@ -2309,6 +2308,7 @@ vect_build_slp_tree_2 (vec_info *vinfo, slp_tree node,
SLP_TREE_CODE (node) = VEC_PERM_EXPR;
SLP_TREE_CHILDREN (node).safe_push (unperm_load);
SLP_TREE_LANE_PERMUTATION (node) = lperm;
+ SLP_TREE_REPRESENTATIVE (node) = NULL;
load_permutation.release ();
return node;
}
@@ -2386,6 +2386,7 @@ vect_build_slp_tree_2 (vec_info *vinfo, slp_tree node,
SLP_TREE_LANE_PERMUTATION (node) = lperm;
SLP_TREE_VECTYPE (node) = vectype;
SLP_TREE_CHILDREN (node).quick_push (vnode);
+ SLP_TREE_REPRESENTATIVE (node) = NULL;
return node;
}
/* When discovery reaches an associatable operation see whether we can
@@ -3316,6 +3317,7 @@ fail:
SLP_TREE_VECTYPE (pnode) = vectype;
SLP_TREE_CHILDREN (pnode).quick_push (child);
SLP_TREE_CHILDREN (pnode).quick_push (child);
+ SLP_TREE_REPRESENTATIVE (pnode) = NULL;
lane_permutation_t& perm = SLP_TREE_LANE_PERMUTATION (pnode);
children.safe_push (pnode);
@@ -3345,6 +3347,7 @@ fail:
SLP_TREE_CODE (node) = VEC_PERM_EXPR;
SLP_TREE_CHILDREN (node).quick_push (one);
SLP_TREE_CHILDREN (node).quick_push (two);
+ SLP_TREE_REPRESENTATIVE (node) = NULL;
enum tree_code code0 = ERROR_MARK;
enum tree_code ocode = ERROR_MARK;
if (gassign *stmt = dyn_cast <gassign *> (stmts[0]->stmt))
@@ -4108,9 +4111,7 @@ vect_build_slp_store_interleaving (vec<slp_tree>
&rhs_nodes,
SLP_TREE_VECTYPE (perm) = SLP_TREE_VECTYPE (node);
perm->max_nunits = max_nunits;
SLP_TREE_LANES (perm) = group_size;
- /* ??? We should set this NULL but that's not expected. */
- SLP_TREE_REPRESENTATIVE (perm)
- = SLP_TREE_REPRESENTATIVE (SLP_TREE_CHILDREN (rhs_nodes[0])[l]);
+ SLP_TREE_REPRESENTATIVE (perm) = NULL;
for (unsigned j = 0; j < rhs_nodes.length (); ++j)
{
SLP_TREE_CHILDREN (perm)
@@ -5918,7 +5919,6 @@ vect_lower_load_permutations (loop_vec_info loop_vinfo,
SLP_TREE_LANE_PERMUTATION (p) = perm;
SLP_TREE_VECTYPE (p) = SLP_TREE_VECTYPE (load);
SLP_TREE_LANES (p) = perm.length ();
- SLP_TREE_REPRESENTATIVE (p) = SLP_TREE_REPRESENTATIVE (load);
/* ??? As we have scalar stmts for this intermediate permute we
could CSE it via bst_map but we do not want to pick up
another SLP node with a load permutation. We instead should
@@ -5937,6 +5937,7 @@ vect_lower_load_permutations (loop_vec_info loop_vinfo,
SLP_TREE_LANE_PERMUTATION (load) = final_perm;
SLP_TREE_CHILDREN (load).create (1);
SLP_TREE_CHILDREN (load).quick_push (l0);
+ SLP_TREE_REPRESENTATIVE (load) = NULL;
}
}
@@ -8077,7 +8078,6 @@ vect_optimize_slp_pass::get_result_with_layout (slp_tree
node,
if (to_layout_i != 0)
vect_slp_permute (m_perms[to_layout_i], stmts, true);
}
- SLP_TREE_REPRESENTATIVE (result) = SLP_TREE_REPRESENTATIVE (node);
SLP_TREE_LANES (result) = num_lanes;
SLP_TREE_VECTYPE (result) = SLP_TREE_VECTYPE (node);
result->vertex = -1;
@@ -8659,7 +8659,8 @@ vect_cse_slp_node_parts (hash_set<slp_tree> &visited,
&& SLP_TREE_SCALAR_STMTS (node)[0]
/* Avoid touching loads which need care with load permutations
and specialities like load-lane representations. */
- && !STMT_VINFO_DATA_REF (SLP_TREE_REPRESENTATIVE (node)))
+ && (SLP_TREE_PERMUTE_P (node)
+ || !STMT_VINFO_DATA_REF (SLP_TREE_REPRESENTATIVE (node))))
for (slp_tree cand
: part_starts[gimple_uid (SLP_TREE_SCALAR_STMTS (node)[0]->stmt)])
/* ??? There is a possible ordering/optimality problem in that
@@ -8960,6 +8961,7 @@ vect_slp_analyze_node_operations_1 (vec_info *vinfo,
slp_tree node,
/* Handle purely internal nodes. */
if (SLP_TREE_PERMUTE_P (node))
{
+ gcc_checking_assert (!SLP_TREE_REPRESENTATIVE (node));
if (!vectorizable_slp_permutation (vinfo, NULL, node, cost_vec))
return false;
--
2.51.0
