https://gcc.gnu.org/g:8bb076ac0d887a6a67263e280b17aacd9e5062e7
commit r16-3570-g8bb076ac0d887a6a67263e280b17aacd9e5062e7
Author: Robin Dapp <rd...@ventanamicro.com>
Date: Thu Aug 28 11:52:51 2025 +0200
vect: Use poly_int64 for prolog bound.
Since peeling and version for alignment for VLA modes was introduced
(r16-3065-geee51f9a4b6) we have been seeing a lot of test suite failures
like
internal compiler error: in apply_scale, at profile-count.h:1187
This is because vect_gen_prolog_loop_niters sets the prolog bound to -1
in case align_in_elems is a non-constant poly_int.
bound - 1 is later used to scale the loop profile in scale_loop_profile
so we try to calculate with an assumed -2 iterations.
This patch changes bound_prolog to poly_int64, using a poly estimate for
frequency scaling but only records an iteration bound for the prolog if
the bound is a scalar.
PR/tree-optimization 121523
gcc/ChangeLog:
* tree-vect-loop-manip.cc (vect_gen_prolog_loop_niters):
Change prolog bound to poly_int64.
(vect_gen_scalar_loop_niters): Ditto.
(vect_do_peeling): Use poly estimate for frequency scaling.
Diff:
---
gcc/tree-vect-loop-manip.cc | 23 +++++++++++------------
1 file changed, 11 insertions(+), 12 deletions(-)
diff --git a/gcc/tree-vect-loop-manip.cc b/gcc/tree-vect-loop-manip.cc
index 566308f4fe5d..20141dbc2e54 100644
--- a/gcc/tree-vect-loop-manip.cc
+++ b/gcc/tree-vect-loop-manip.cc
@@ -2507,7 +2507,7 @@ get_misalign_in_elems (gimple **seq, loop_vec_info
loop_vinfo)
static tree
vect_gen_prolog_loop_niters (loop_vec_info loop_vinfo,
- basic_block bb, int *bound)
+ basic_block bb, poly_int64 *bound)
{
dr_vec_info *dr_info = LOOP_VINFO_UNALIGNED_DR (loop_vinfo);
tree var;
@@ -2555,11 +2555,7 @@ vect_gen_prolog_loop_niters (loop_vec_info loop_vinfo,
misalign_in_elems);
iters = fold_build2 (BIT_AND_EXPR, type, iters, align_in_elems_minus_1);
iters = fold_convert (niters_type, iters);
- unsigned HOST_WIDE_INT align_in_elems_c;
- if (align_in_elems.is_constant (&align_in_elems_c))
- *bound = align_in_elems_c - 1;
- else
- *bound = -1;
+ *bound = align_in_elems;
}
if (dump_enabled_p ())
@@ -2728,8 +2724,8 @@ vect_build_loop_niters (loop_vec_info loop_vinfo, bool
*new_var_p)
static tree
vect_gen_scalar_loop_niters (tree niters_prolog, int int_niters_prolog,
- int bound_prolog, poly_int64 bound_epilog, int th,
- poly_uint64 *bound_scalar,
+ poly_int64 bound_prolog, poly_int64 bound_epilog,
+ int th, poly_uint64 *bound_scalar,
bool check_profitability)
{
tree type = TREE_TYPE (niters_prolog);
@@ -3257,11 +3253,11 @@ vect_do_peeling (loop_vec_info loop_vinfo, tree niters,
tree nitersm1,
/* Generate the number of iterations for the prolog loop. We do this here
so that we can also get the upper bound on the number of iterations. */
tree niters_prolog;
- int bound_prolog = 0;
+ poly_int64 bound_prolog = 0;
if (prolog_peeling)
{
niters_prolog = vect_gen_prolog_loop_niters (loop_vinfo, anchor,
- &bound_prolog);
+ &bound_prolog);
/* If algonment peeling is known, we will always execute prolog. */
if (TREE_CODE (niters_prolog) == INTEGER_CST)
prob_prolog = profile_probability::always ();
@@ -3404,7 +3400,8 @@ vect_do_peeling (loop_vec_info loop_vinfo, tree niters,
tree nitersm1,
slpeel_update_phi_nodes_for_guard1 (prolog, loop, guard_e, e);
scale_bbs_frequencies (&bb_after_prolog, 1, prob_prolog);
- scale_loop_profile (prolog, prob_prolog, bound_prolog - 1);
+ scale_loop_profile (prolog, prob_prolog,
+ estimated_poly_value (bound_prolog) - 1);
}
/* Update init address of DRs. */
@@ -3429,7 +3426,9 @@ vect_do_peeling (loop_vec_info loop_vinfo, tree niters,
tree nitersm1,
/* Prolog iterates at most bound_prolog times, latch iterates at
most bound_prolog - 1 times. */
- record_niter_bound (prolog, bound_prolog - 1, false, true);
+ if (bound_prolog.is_constant ())
+ record_niter_bound (prolog, bound_prolog.to_constant () - 1, false,
+ true);
delete_update_ssa ();
adjust_vec_debug_stmts ();
scev_reset ();
