mirror/gcc
1
0
Fork 0
mirror of https://github.com/gcc-mirror/gcc.git synced 2024-11-21 13:40:47 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

tree-optimization/117606 - SLP and single element interleaving

Browse Source 
The following tries to reduce the amount of difference between
SLP and non-SLP for single-element interleaving load classification.

This fixes another fallout of --param vect-force-slp=1

	PR tree-optimization/117606
	* tree-vect-stmts.cc (get_group_load_store_type): For single
	element interleaving also fall back to VMAT_ELEMENTWISE if
	a left-over permutation isn't supported.
This commit is contained in:
Richard Biener 2024-11-15 10:04:23 +01:00 committed by Richard Biener
parent 4621b684ae
commit 39e763ca87
1 changed files with 25 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
gcc/tree-vect-stmts.cc
View File

@ -2082,8 +2082,9 @@ get_group_load_store_type (vec_info *vinfo, stmt_vec_info stmt_info,
*memory_access_type = VMAT_CONTIGUOUS;
/* If this is single-element interleaving with an element
distance that leaves unused vector loads around punt - we
at least create very sub-optimal code in that case (and
distance that leaves unused vector loads around fall back
to elementwise access if possible - we otherwise least
create very sub-optimal code in that case (and
blow up memory, see PR65518). */
if (loop_vinfo
&& single_element_p
@ -2110,6 +2111,28 @@ get_group_load_store_type (vec_info *vinfo, stmt_vec_info stmt_info,
}
}
/* For single-element interleaving also fall back to elementwise
access in case we did not lower a permutation and cannot
code generate it. */
auto_vec<tree> temv;
unsigned n_perms;
if (loop_vinfo
&& single_element_p
&& SLP_TREE_LANES (slp_node) == 1
&& (*memory_access_type == VMAT_CONTIGUOUS
|| *memory_access_type == VMAT_CONTIGUOUS_REVERSE)
&& SLP_TREE_LOAD_PERMUTATION (slp_node).exists ()
&& !vect_transform_slp_perm_load
(loop_vinfo, slp_node, temv, NULL,
LOOP_VINFO_VECT_FACTOR (loop_vinfo), true, &n_perms))
{
*memory_access_type = VMAT_ELEMENTWISE;
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"single-element interleaving permutation not "
"supported, using elementwise access\n");
}
overrun_p = (loop_vinfo && gap != 0
&& *memory_access_type != VMAT_ELEMENTWISE);
if (overrun_p && vls_type != VLS_LOAD)