https://gcc.gnu.org/g:0d4b254b20a9193ab261d02f8a063e21816f85e4
commit r15-5168-g0d4b254b20a9193ab261d02f8a063e21816f85e4
Author: Richard Biener <rguent...@suse.de>
Date: Tue Nov 12 10:31:30 2024 +0100
tree-optimization/116973 - SLP permute lower heuristic and single-lane SLP
When forcing single-lane SLP to emulate non-SLP behavior we need to
disable heuristics designed to optimize SLP loads and instead in
all cases resort to an interleaving scheme as requested by forcefully
doing single-lane SLP.
This fixes the remaining fallout for --param vect-force-slp=1 on x86.
PR tree-optimization/116973
* tree-vect-slp.cc (vect_lower_load_permutations): Add
force_single_lane parameter. Disable heuristic that keeps
some load-permutations.
(vect_analyze_slp): Pass force_single_lane to
vect_lower_load_permutations.
Diff:
---
gcc/tree-vect-slp.cc | 17 +++++++++++------
1 file changed, 11 insertions(+), 6 deletions(-)
diff --git a/gcc/tree-vect-slp.cc b/gcc/tree-vect-slp.cc
index eebac1955de7..d69fdc04b9df 100644
--- a/gcc/tree-vect-slp.cc
+++ b/gcc/tree-vect-slp.cc
@@ -4402,7 +4402,8 @@ vllp_cmp (const void *a_, const void *b_)
static void
vect_lower_load_permutations (loop_vec_info loop_vinfo,
scalar_stmts_to_slp_tree_map_t *bst_map,
- const array_slice<slp_tree> &loads)
+ const array_slice<slp_tree> &loads,
+ bool force_single_lane)
{
/* We at this point want to lower without a fixed VF or vector
size in mind which means we cannot actually compute whether we
@@ -4494,7 +4495,8 @@ vect_lower_load_permutations (loop_vec_info loop_vinfo,
extracting it from the larger load.
??? Long-term some of the lowering should move to where
the vector types involved are fixed. */
- if (ld_lanes_lanes == 0
+ if (!force_single_lane
+ && ld_lanes_lanes == 0
&& contiguous
&& (SLP_TREE_LANES (load) > 1 || loads.size () == 1)
&& pow2p_hwi (SLP_TREE_LANES (load))
@@ -4668,7 +4670,8 @@ vect_lower_load_permutations (loop_vec_info loop_vinfo,
static void
vect_lower_load_permutations (loop_vec_info loop_vinfo,
- scalar_stmts_to_slp_tree_map_t *bst_map)
+ scalar_stmts_to_slp_tree_map_t *bst_map,
+ bool force_single_lane)
{
/* Gather and sort loads across all instances. */
hash_set<slp_tree> visited;
@@ -4696,14 +4699,16 @@ vect_lower_load_permutations (loop_vec_info loop_vinfo,
if (STMT_VINFO_GROUPED_ACCESS (a0))
vect_lower_load_permutations (loop_vinfo, bst_map,
make_array_slice (&loads[firsti],
- i - firsti));
+ i - firsti),
+ force_single_lane);
firsti = i;
}
if (firsti < loads.length ()
&& STMT_VINFO_GROUPED_ACCESS (SLP_TREE_SCALAR_STMTS (loads[firsti])[0]))
vect_lower_load_permutations (loop_vinfo, bst_map,
make_array_slice (&loads[firsti],
- loads.length () - firsti));
+ loads.length () - firsti),
+ force_single_lane);
}
/* Check if there are stmts in the loop can be vectorized using SLP. Build SLP
@@ -5097,7 +5102,7 @@ vect_analyze_slp (vec_info *vinfo, unsigned max_tree_size,
like schemes. */
if (loop_vec_info loop_vinfo = dyn_cast <loop_vec_info> (vinfo))
{
- vect_lower_load_permutations (loop_vinfo, bst_map);
+ vect_lower_load_permutations (loop_vinfo, bst_map, force_single_lane);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
