Google Git
Sign in
gnu / gcc / 926d9947847a0683cf34de6fc231209747576088 / . / gcc / tree-vect-loop.c
blob: fac1d39b289065d736bfc256531cf81e414f143e [file] [log] [blame]
/* Loop Vectorization
Copyright (C) 2003-2015 Free Software Foundation, Inc.
Contributed by Dorit Naishlos <dorit@il.ibm.com> and
Ira Rosen <irar@il.ibm.com>
This file is part of GCC.
GCC is free software; you can redistribute it and/or modify it under
the terms of the GNU General Public License as published by the Free
Software Foundation; either version 3, or (at your option) any later
version.
GCC is distributed in the hope that it will be useful, but WITHOUT ANY
WARRANTY; without even the implied warranty of MERCHANTABILITY or
FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
for more details.
You should have received a copy of the GNU General Public License
along with GCC; see the file COPYING3. If not see
<http://www.gnu.org/licenses/>. */
#include "config.h"
#include "system.h"
#include "coretypes.h"
#include "dumpfile.h"
#include "tm.h"
#include "hash-set.h"
#include "machmode.h"
#include "vec.h"
#include "double-int.h"
#include "input.h"
#include "alias.h"
#include "symtab.h"
#include "wide-int.h"
#include "inchash.h"
#include "tree.h"
#include "fold-const.h"
#include "stor-layout.h"
#include "predict.h"
#include "hard-reg-set.h"
#include "function.h"
#include "dominance.h"
#include "cfg.h"
#include "cfganal.h"
#include "basic-block.h"
#include "gimple-pretty-print.h"
#include "tree-ssa-alias.h"
#include "internal-fn.h"
#include "gimple-expr.h"
#include "is-a.h"
#include "gimple.h"
#include "gimplify.h"
#include "gimple-iterator.h"
#include "gimplify-me.h"
#include "gimple-ssa.h"
#include "tree-phinodes.h"
#include "ssa-iterators.h"
#include "stringpool.h"
#include "tree-ssanames.h"
#include "tree-ssa-loop-ivopts.h"
#include "tree-ssa-loop-manip.h"
#include "tree-ssa-loop-niter.h"
#include "tree-pass.h"
#include "cfgloop.h"
#include "hashtab.h"
#include "rtl.h"
#include "flags.h"
#include "statistics.h"
#include "real.h"
#include "fixed-value.h"
#include "insn-config.h"
#include "expmed.h"
#include "dojump.h"
#include "explow.h"
#include "calls.h"
#include "emit-rtl.h"
#include "varasm.h"
#include "stmt.h"
#include "expr.h"
#include "recog.h"
#include "insn-codes.h"
#include "optabs.h"
#include "params.h"
#include "diagnostic-core.h"
#include "tree-chrec.h"
#include "tree-scalar-evolution.h"
#include "tree-vectorizer.h"
#include "target.h"
/* Loop Vectorization Pass.
This pass tries to vectorize loops.
For example, the vectorizer transforms the following simple loop:
short a[N]; short b[N]; short c[N]; int i;
for (i=0; i<N; i++){
a[i] = b[i] + c[i];
}
as if it was manually vectorized by rewriting the source code into:
typedef int __attribute__((mode(V8HI))) v8hi;
short a[N]; short b[N]; short c[N]; int i;
v8hi *pa = (v8hi*)a, *pb = (v8hi*)b, *pc = (v8hi*)c;
v8hi va, vb, vc;
for (i=0; i<N/8; i++){
vb = pb[i];
vc = pc[i];
va = vb + vc;
pa[i] = va;
}
The main entry to this pass is vectorize_loops(), in which
the vectorizer applies a set of analyses on a given set of loops,
followed by the actual vectorization transformation for the loops that
had successfully passed the analysis phase.
Throughout this pass we make a distinction between two types of
data: scalars (which are represented by SSA_NAMES), and memory references
("data-refs"). These two types of data require different handling both
during analysis and transformation. The types of data-refs that the
vectorizer currently supports are ARRAY_REFS which base is an array DECL
(not a pointer), and INDIRECT_REFS through pointers; both array and pointer
accesses are required to have a simple (consecutive) access pattern.
Analysis phase:
===============
The driver for the analysis phase is vect_analyze_loop().
It applies a set of analyses, some of which rely on the scalar evolution
analyzer (scev) developed by Sebastian Pop.
During the analysis phase the vectorizer records some information
per stmt in a "stmt_vec_info" struct which is attached to each stmt in the
loop, as well as general information about the loop as a whole, which is
recorded in a "loop_vec_info" struct attached to each loop.
Transformation phase:
=====================
The loop transformation phase scans all the stmts in the loop, and
creates a vector stmt (or a sequence of stmts) for each scalar stmt S in
the loop that needs to be vectorized. It inserts the vector code sequence
just before the scalar stmt S, and records a pointer to the vector code
in STMT_VINFO_VEC_STMT (stmt_info) (stmt_info is the stmt_vec_info struct
attached to S). This pointer will be used for the vectorization of following
stmts which use the def of stmt S. Stmt S is removed if it writes to memory;
otherwise, we rely on dead code elimination for removing it.
For example, say stmt S1 was vectorized into stmt VS1:
VS1: vb = px[i];
S1: b = x[i]; STMT_VINFO_VEC_STMT (stmt_info (S1)) = VS1
S2: a = b;
To vectorize stmt S2, the vectorizer first finds the stmt that defines
the operand 'b' (S1), and gets the relevant vector def 'vb' from the
vector stmt VS1 pointed to by STMT_VINFO_VEC_STMT (stmt_info (S1)). The
resulting sequence would be:
VS1: vb = px[i];
S1: b = x[i]; STMT_VINFO_VEC_STMT (stmt_info (S1)) = VS1
VS2: va = vb;
S2: a = b; STMT_VINFO_VEC_STMT (stmt_info (S2)) = VS2
Operands that are not SSA_NAMEs, are data-refs that appear in
load/store operations (like 'x[i]' in S1), and are handled differently.
Target modeling:
=================
Currently the only target specific information that is used is the
size of the vector (in bytes) - "TARGET_VECTORIZE_UNITS_PER_SIMD_WORD".
Targets that can support different sizes of vectors, for now will need
to specify one value for "TARGET_VECTORIZE_UNITS_PER_SIMD_WORD". More
flexibility will be added in the future.
Since we only vectorize operations which vector form can be
expressed using existing tree codes, to verify that an operation is
supported, the vectorizer checks the relevant optab at the relevant
machine_mode (e.g, optab_handler (add_optab, V8HImode)). If
the value found is CODE_FOR_nothing, then there's no target support, and
we can't vectorize the stmt.
For additional information on this project see:
http://gcc.gnu.org/projects/tree-ssa/vectorization.html
*/
static void vect_estimate_min_profitable_iters (loop_vec_info, int *, int *);
/* Function vect_determine_vectorization_factor
Determine the vectorization factor (VF). VF is the number of data elements
that are operated upon in parallel in a single iteration of the vectorized
loop. For example, when vectorizing a loop that operates on 4byte elements,
on a target with vector size (VS) 16byte, the VF is set to 4, since 4
elements can fit in a single vector register.
We currently support vectorization of loops in which all types operated upon
are of the same size. Therefore this function currently sets VF according to
the size of the types operated upon, and fails if there are multiple sizes
in the loop.
VF is also the factor by which the loop iterations are strip-mined, e.g.:
original loop:
for (i=0; i<N; i++){
a[i] = b[i] + c[i];
}
vectorized loop:
for (i=0; i<N; i+=VF){
a[i:VF] = b[i:VF] + c[i:VF];
}
*/
static bool
vect_determine_vectorization_factor (loop_vec_info loop_vinfo)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block *bbs = LOOP_VINFO_BBS (loop_vinfo);
int nbbs = loop->num_nodes;
unsigned int vectorization_factor = 0;
tree scalar_type;
gphi *phi;
tree vectype;
unsigned int nunits;
stmt_vec_info stmt_info;
int i;
HOST_WIDE_INT dummy;
gimple stmt, pattern_stmt = NULL;
gimple_seq pattern_def_seq = NULL;
gimple_stmt_iterator pattern_def_si = gsi_none ();
bool analyze_pattern_stmt = false;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"=== vect_determine_vectorization_factor ===\n");
for (i = 0; i < nbbs; i++)
{
basic_block bb = bbs[i];
for (gphi_iterator si = gsi_start_phis (bb); !gsi_end_p (si);
gsi_next (&si))
{
phi = si.phi ();
stmt_info = vinfo_for_stmt (phi);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "==> examining phi: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, phi, 0);
dump_printf (MSG_NOTE, "\n");
}
gcc_assert (stmt_info);
if (STMT_VINFO_RELEVANT_P (stmt_info))
{
gcc_assert (!STMT_VINFO_VECTYPE (stmt_info));
scalar_type = TREE_TYPE (PHI_RESULT (phi));
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"get vectype for scalar type: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, scalar_type);
dump_printf (MSG_NOTE, "\n");
}
vectype = get_vectype_for_scalar_type (scalar_type);
if (!vectype)
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: unsupported "
"data-type ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM,
scalar_type);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
STMT_VINFO_VECTYPE (stmt_info) = vectype;
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "vectype: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, vectype);
dump_printf (MSG_NOTE, "\n");
}
nunits = TYPE_VECTOR_SUBPARTS (vectype);
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location, "nunits = %d\n",
nunits);
if (!vectorization_factor
|| (nunits > vectorization_factor))
vectorization_factor = nunits;
}
}
for (gimple_stmt_iterator si = gsi_start_bb (bb);
!gsi_end_p (si) || analyze_pattern_stmt;)
{
tree vf_vectype;
if (analyze_pattern_stmt)
stmt = pattern_stmt;
else
stmt = gsi_stmt (si);
stmt_info = vinfo_for_stmt (stmt);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"==> examining statement: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, stmt, 0);
dump_printf (MSG_NOTE, "\n");
}
gcc_assert (stmt_info);
/* Skip stmts which do not need to be vectorized. */
if ((!STMT_VINFO_RELEVANT_P (stmt_info)
&& !STMT_VINFO_LIVE_P (stmt_info))
|| gimple_clobber_p (stmt))
{
if (STMT_VINFO_IN_PATTERN_P (stmt_info)
&& (pattern_stmt = STMT_VINFO_RELATED_STMT (stmt_info))
&& (STMT_VINFO_RELEVANT_P (vinfo_for_stmt (pattern_stmt))
|| STMT_VINFO_LIVE_P (vinfo_for_stmt (pattern_stmt))))
{
stmt = pattern_stmt;
stmt_info = vinfo_for_stmt (pattern_stmt);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"==> examining pattern statement: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, stmt, 0);
dump_printf (MSG_NOTE, "\n");
}
}
else
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location, "skip.\n");
gsi_next (&si);
continue;
}
}
else if (STMT_VINFO_IN_PATTERN_P (stmt_info)
&& (pattern_stmt = STMT_VINFO_RELATED_STMT (stmt_info))
&& (STMT_VINFO_RELEVANT_P (vinfo_for_stmt (pattern_stmt))
|| STMT_VINFO_LIVE_P (vinfo_for_stmt (pattern_stmt))))
analyze_pattern_stmt = true;
/* If a pattern statement has def stmts, analyze them too. */
if (is_pattern_stmt_p (stmt_info))
{
if (pattern_def_seq == NULL)
{
pattern_def_seq = STMT_VINFO_PATTERN_DEF_SEQ (stmt_info);
pattern_def_si = gsi_start (pattern_def_seq);
}
else if (!gsi_end_p (pattern_def_si))
gsi_next (&pattern_def_si);
if (pattern_def_seq != NULL)
{
gimple pattern_def_stmt = NULL;
stmt_vec_info pattern_def_stmt_info = NULL;
while (!gsi_end_p (pattern_def_si))
{
pattern_def_stmt = gsi_stmt (pattern_def_si);
pattern_def_stmt_info
= vinfo_for_stmt (pattern_def_stmt);
if (STMT_VINFO_RELEVANT_P (pattern_def_stmt_info)
|| STMT_VINFO_LIVE_P (pattern_def_stmt_info))
break;
gsi_next (&pattern_def_si);
}
if (!gsi_end_p (pattern_def_si))
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"==> examining pattern def stmt: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM,
pattern_def_stmt, 0);
dump_printf (MSG_NOTE, "\n");
}
stmt = pattern_def_stmt;
stmt_info = pattern_def_stmt_info;
}
else
{
pattern_def_si = gsi_none ();
analyze_pattern_stmt = false;
}
}
else
analyze_pattern_stmt = false;
}
if (gimple_get_lhs (stmt) == NULL_TREE
/* MASK_STORE has no lhs, but is ok. */
&& (!is_gimple_call (stmt)
|| !gimple_call_internal_p (stmt)
|| gimple_call_internal_fn (stmt) != IFN_MASK_STORE))
{
if (is_gimple_call (stmt))
{
/* Ignore calls with no lhs. These must be calls to
#pragma omp simd functions, and what vectorization factor
it really needs can't be determined until
vectorizable_simd_clone_call. */
if (!analyze_pattern_stmt && gsi_end_p (pattern_def_si))
{
pattern_def_seq = NULL;
gsi_next (&si);
}
continue;
}
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: irregular stmt.");
dump_gimple_stmt (MSG_MISSED_OPTIMIZATION, TDF_SLIM, stmt,
0);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
if (VECTOR_MODE_P (TYPE_MODE (gimple_expr_type (stmt))))
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: vector stmt in loop:");
dump_gimple_stmt (MSG_MISSED_OPTIMIZATION, TDF_SLIM, stmt, 0);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
if (STMT_VINFO_VECTYPE (stmt_info))
{
/* The only case when a vectype had been already set is for stmts
that contain a dataref, or for "pattern-stmts" (stmts
generated by the vectorizer to represent/replace a certain
idiom). */
gcc_assert (STMT_VINFO_DATA_REF (stmt_info)
|| is_pattern_stmt_p (stmt_info)
|| !gsi_end_p (pattern_def_si));
vectype = STMT_VINFO_VECTYPE (stmt_info);
}
else
{
gcc_assert (!STMT_VINFO_DATA_REF (stmt_info));
if (is_gimple_call (stmt)
&& gimple_call_internal_p (stmt)
&& gimple_call_internal_fn (stmt) == IFN_MASK_STORE)
scalar_type = TREE_TYPE (gimple_call_arg (stmt, 3));
else
scalar_type = TREE_TYPE (gimple_get_lhs (stmt));
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"get vectype for scalar type: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, scalar_type);
dump_printf (MSG_NOTE, "\n");
}
vectype = get_vectype_for_scalar_type (scalar_type);
if (!vectype)
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: unsupported "
"data-type ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM,
scalar_type);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
STMT_VINFO_VECTYPE (stmt_info) = vectype;
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "vectype: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, vectype);
dump_printf (MSG_NOTE, "\n");
}
}
/* The vectorization factor is according to the smallest
scalar type (or the largest vector size, but we only
support one vector size per loop). */
scalar_type = vect_get_smallest_scalar_type (stmt, &dummy,
&dummy);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"get vectype for scalar type: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, scalar_type);
dump_printf (MSG_NOTE, "\n");
}
vf_vectype = get_vectype_for_scalar_type (scalar_type);
if (!vf_vectype)
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: unsupported data-type ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM,
scalar_type);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
if ((GET_MODE_SIZE (TYPE_MODE (vectype))
!= GET_MODE_SIZE (TYPE_MODE (vf_vectype))))
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: different sized vector "
"types in statement, ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM,
vectype);
dump_printf (MSG_MISSED_OPTIMIZATION, " and ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM,
vf_vectype);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "vectype: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, vf_vectype);
dump_printf (MSG_NOTE, "\n");
}
nunits = TYPE_VECTOR_SUBPARTS (vf_vectype);
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location, "nunits = %d\n", nunits);
if (!vectorization_factor
|| (nunits > vectorization_factor))
vectorization_factor = nunits;
if (!analyze_pattern_stmt && gsi_end_p (pattern_def_si))
{
pattern_def_seq = NULL;
gsi_next (&si);
}
}
}
/* TODO: Analyze cost. Decide if worth while to vectorize. */
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location, "vectorization factor = %d\n",
vectorization_factor);
if (vectorization_factor <= 1)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: unsupported data-type\n");
return false;
}
LOOP_VINFO_VECT_FACTOR (loop_vinfo) = vectorization_factor;
return true;
}
/* Function vect_is_simple_iv_evolution.
FORNOW: A simple evolution of an induction variables in the loop is
considered a polynomial evolution. */
static bool
vect_is_simple_iv_evolution (unsigned loop_nb, tree access_fn, tree * init,
tree * step)
{
tree init_expr;
tree step_expr;
tree evolution_part = evolution_part_in_loop_num (access_fn, loop_nb);
basic_block bb;
/* When there is no evolution in this loop, the evolution function
is not "simple". */
if (evolution_part == NULL_TREE)
return false;
/* When the evolution is a polynomial of degree >= 2
the evolution function is not "simple". */
if (tree_is_chrec (evolution_part))
return false;
step_expr = evolution_part;
init_expr = unshare_expr (initial_condition_in_loop_num (access_fn, loop_nb));
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "step: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, step_expr);
dump_printf (MSG_NOTE, ", init: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, init_expr);
dump_printf (MSG_NOTE, "\n");
}
*init = init_expr;
*step = step_expr;
if (TREE_CODE (step_expr) != INTEGER_CST
&& (TREE_CODE (step_expr) != SSA_NAME
|| ((bb = gimple_bb (SSA_NAME_DEF_STMT (step_expr)))
&& flow_bb_inside_loop_p (get_loop (cfun, loop_nb), bb))
|| (!INTEGRAL_TYPE_P (TREE_TYPE (step_expr))
&& (!SCALAR_FLOAT_TYPE_P (TREE_TYPE (step_expr))
|| !flag_associative_math)))
&& (TREE_CODE (step_expr) != REAL_CST
|| !flag_associative_math))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"step unknown.\n");
return false;
}
return true;
}
/* Function vect_analyze_scalar_cycles_1.
Examine the cross iteration def-use cycles of scalar variables
in LOOP. LOOP_VINFO represents the loop that is now being
considered for vectorization (can be LOOP, or an outer-loop
enclosing LOOP). */
static void
vect_analyze_scalar_cycles_1 (loop_vec_info loop_vinfo, struct loop *loop)
{
basic_block bb = loop->header;
tree init, step;
auto_vec<gimple, 64> worklist;
gphi_iterator gsi;
bool double_reduc;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"=== vect_analyze_scalar_cycles ===\n");
/* First - identify all inductions. Reduction detection assumes that all the
inductions have been identified, therefore, this order must not be
changed. */
for (gsi = gsi_start_phis (bb); !gsi_end_p (gsi); gsi_next (&gsi))
{
gphi *phi = gsi.phi ();
tree access_fn = NULL;
tree def = PHI_RESULT (phi);
stmt_vec_info stmt_vinfo = vinfo_for_stmt (phi);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "Analyze phi: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, phi, 0);
dump_printf (MSG_NOTE, "\n");
}
/* Skip virtual phi's. The data dependences that are associated with
virtual defs/uses (i.e., memory accesses) are analyzed elsewhere. */
if (virtual_operand_p (def))
continue;
STMT_VINFO_DEF_TYPE (stmt_vinfo) = vect_unknown_def_type;
/* Analyze the evolution function. */
access_fn = analyze_scalar_evolution (loop, def);
if (access_fn)
{
STRIP_NOPS (access_fn);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"Access function of PHI: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, access_fn);
dump_printf (MSG_NOTE, "\n");
}
STMT_VINFO_LOOP_PHI_EVOLUTION_PART (stmt_vinfo)
= evolution_part_in_loop_num (access_fn, loop->num);
}
if (!access_fn
|| !vect_is_simple_iv_evolution (loop->num, access_fn, &init, &step)
|| (LOOP_VINFO_LOOP (loop_vinfo) != loop
&& TREE_CODE (step) != INTEGER_CST))
{
worklist.safe_push (phi);
continue;
}
gcc_assert (STMT_VINFO_LOOP_PHI_EVOLUTION_PART (stmt_vinfo) != NULL_TREE);
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location, "Detected induction.\n");
STMT_VINFO_DEF_TYPE (stmt_vinfo) = vect_induction_def;
}
/* Second - identify all reductions and nested cycles. */
while (worklist.length () > 0)
{
gimple phi = worklist.pop ();
tree def = PHI_RESULT (phi);
stmt_vec_info stmt_vinfo = vinfo_for_stmt (phi);
gimple reduc_stmt;
bool nested_cycle;
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "Analyze phi: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, phi, 0);
dump_printf (MSG_NOTE, "\n");
}
gcc_assert (!virtual_operand_p (def)
&& STMT_VINFO_DEF_TYPE (stmt_vinfo) == vect_unknown_def_type);
nested_cycle = (loop != LOOP_VINFO_LOOP (loop_vinfo));
reduc_stmt = vect_force_simple_reduction (loop_vinfo, phi, !nested_cycle,
&double_reduc);
if (reduc_stmt)
{
if (double_reduc)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"Detected double reduction.\n");
STMT_VINFO_DEF_TYPE (stmt_vinfo) = vect_double_reduction_def;
STMT_VINFO_DEF_TYPE (vinfo_for_stmt (reduc_stmt)) =
vect_double_reduction_def;
}
else
{
if (nested_cycle)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"Detected vectorizable nested cycle.\n");
STMT_VINFO_DEF_TYPE (stmt_vinfo) = vect_nested_cycle;
STMT_VINFO_DEF_TYPE (vinfo_for_stmt (reduc_stmt)) =
vect_nested_cycle;
}
else
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"Detected reduction.\n");
STMT_VINFO_DEF_TYPE (stmt_vinfo) = vect_reduction_def;
STMT_VINFO_DEF_TYPE (vinfo_for_stmt (reduc_stmt)) =
vect_reduction_def;
/* Store the reduction cycles for possible vectorization in
loop-aware SLP. */
LOOP_VINFO_REDUCTIONS (loop_vinfo).safe_push (reduc_stmt);
}
}
}
else
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"Unknown def-use cycle pattern.\n");
}
}
/* Function vect_analyze_scalar_cycles.
Examine the cross iteration def-use cycles of scalar variables, by
analyzing the loop-header PHIs of scalar variables. Classify each
cycle as one of the following: invariant, induction, reduction, unknown.
We do that for the loop represented by LOOP_VINFO, and also to its
inner-loop, if exists.
Examples for scalar cycles:
Example1: reduction:
loop1:
for (i=0; i<N; i++)
sum += a[i];
Example2: induction:
loop2:
for (i=0; i<N; i++)
a[i] = i; */
static void
vect_analyze_scalar_cycles (loop_vec_info loop_vinfo)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
vect_analyze_scalar_cycles_1 (loop_vinfo, loop);
/* When vectorizing an outer-loop, the inner-loop is executed sequentially.
Reductions in such inner-loop therefore have different properties than
the reductions in the nest that gets vectorized:
1. When vectorized, they are executed in the same order as in the original
scalar loop, so we can't change the order of computation when
vectorizing them.
2. FIXME: Inner-loop reductions can be used in the inner-loop, so the
current checks are too strict. */
if (loop->inner)
vect_analyze_scalar_cycles_1 (loop_vinfo, loop->inner);
}
/* Function vect_get_loop_niters.
Determine how many iterations the loop is executed and place it
in NUMBER_OF_ITERATIONS. Place the number of latch iterations
in NUMBER_OF_ITERATIONSM1.
Return the loop exit condition. */
static gcond *
vect_get_loop_niters (struct loop *loop, tree *number_of_iterations,
tree *number_of_iterationsm1)
{
tree niters;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"=== get_loop_niters ===\n");
niters = number_of_latch_executions (loop);
*number_of_iterationsm1 = niters;
/* We want the number of loop header executions which is the number
of latch executions plus one.
??? For UINT_MAX latch executions this number overflows to zero
for loops like do { n++; } while (n != 0); */
if (niters && !chrec_contains_undetermined (niters))
niters = fold_build2 (PLUS_EXPR, TREE_TYPE (niters), unshare_expr (niters),
build_int_cst (TREE_TYPE (niters), 1));
*number_of_iterations = niters;
return get_loop_exit_condition (loop);
}
/* Function bb_in_loop_p
Used as predicate for dfs order traversal of the loop bbs. */
static bool
bb_in_loop_p (const_basic_block bb, const void *data)
{
const struct loop *const loop = (const struct loop *)data;
if (flow_bb_inside_loop_p (loop, bb))
return true;
return false;
}
/* Function new_loop_vec_info.
Create and initialize a new loop_vec_info struct for LOOP, as well as
stmt_vec_info structs for all the stmts in LOOP. */
static loop_vec_info
new_loop_vec_info (struct loop *loop)
{
loop_vec_info res;
basic_block *bbs;
gimple_stmt_iterator si;
unsigned int i, nbbs;
res = (loop_vec_info) xcalloc (1, sizeof (struct _loop_vec_info));
LOOP_VINFO_LOOP (res) = loop;
bbs = get_loop_body (loop);
/* Create/Update stmt_info for all stmts in the loop. */
for (i = 0; i < loop->num_nodes; i++)
{
basic_block bb = bbs[i];
/* BBs in a nested inner-loop will have been already processed (because
we will have called vect_analyze_loop_form for any nested inner-loop).
Therefore, for stmts in an inner-loop we just want to update the
STMT_VINFO_LOOP_VINFO field of their stmt_info to point to the new
loop_info of the outer-loop we are currently considering to vectorize
(instead of the loop_info of the inner-loop).
For stmts in other BBs we need to create a stmt_info from scratch. */
if (bb->loop_father != loop)
{
/* Inner-loop bb. */
gcc_assert (loop->inner && bb->loop_father == loop->inner);
for (si = gsi_start_phis (bb); !gsi_end_p (si); gsi_next (&si))
{
gimple phi = gsi_stmt (si);
stmt_vec_info stmt_info = vinfo_for_stmt (phi);
loop_vec_info inner_loop_vinfo =
STMT_VINFO_LOOP_VINFO (stmt_info);
gcc_assert (loop->inner == LOOP_VINFO_LOOP (inner_loop_vinfo));
STMT_VINFO_LOOP_VINFO (stmt_info) = res;
}
for (si = gsi_start_bb (bb); !gsi_end_p (si); gsi_next (&si))
{
gimple stmt = gsi_stmt (si);
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
loop_vec_info inner_loop_vinfo =
STMT_VINFO_LOOP_VINFO (stmt_info);
gcc_assert (loop->inner == LOOP_VINFO_LOOP (inner_loop_vinfo));
STMT_VINFO_LOOP_VINFO (stmt_info) = res;
}
}
else
{
/* bb in current nest. */
for (si = gsi_start_phis (bb); !gsi_end_p (si); gsi_next (&si))
{
gimple phi = gsi_stmt (si);
gimple_set_uid (phi, 0);
set_vinfo_for_stmt (phi, new_stmt_vec_info (phi, res, NULL));
}
for (si = gsi_start_bb (bb); !gsi_end_p (si); gsi_next (&si))
{
gimple stmt = gsi_stmt (si);
gimple_set_uid (stmt, 0);
set_vinfo_for_stmt (stmt, new_stmt_vec_info (stmt, res, NULL));
}
}
}
/* CHECKME: We want to visit all BBs before their successors (except for
latch blocks, for which this assertion wouldn't hold). In the simple
case of the loop forms we allow, a dfs order of the BBs would the same
as reversed postorder traversal, so we are safe. */
free (bbs);
bbs = XCNEWVEC (basic_block, loop->num_nodes);
nbbs = dfs_enumerate_from (loop->header, 0, bb_in_loop_p,
bbs, loop->num_nodes, loop);
gcc_assert (nbbs == loop->num_nodes);
LOOP_VINFO_BBS (res) = bbs;
LOOP_VINFO_NITERSM1 (res) = NULL;
LOOP_VINFO_NITERS (res) = NULL;
LOOP_VINFO_NITERS_UNCHANGED (res) = NULL;
LOOP_VINFO_COST_MODEL_MIN_ITERS (res) = 0;
LOOP_VINFO_COST_MODEL_THRESHOLD (res) = 0;
LOOP_VINFO_VECTORIZABLE_P (res) = 0;
LOOP_VINFO_PEELING_FOR_ALIGNMENT (res) = 0;
LOOP_VINFO_VECT_FACTOR (res) = 0;
LOOP_VINFO_LOOP_NEST (res).create (3);
LOOP_VINFO_DATAREFS (res).create (10);
LOOP_VINFO_DDRS (res).create (10 * 10);
LOOP_VINFO_UNALIGNED_DR (res) = NULL;
LOOP_VINFO_MAY_MISALIGN_STMTS (res).create (
PARAM_VALUE (PARAM_VECT_MAX_VERSION_FOR_ALIGNMENT_CHECKS));
LOOP_VINFO_MAY_ALIAS_DDRS (res).create (
PARAM_VALUE (PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS));
LOOP_VINFO_GROUPED_STORES (res).create (10);
LOOP_VINFO_REDUCTIONS (res).create (10);
LOOP_VINFO_REDUCTION_CHAINS (res).create (10);
LOOP_VINFO_SLP_INSTANCES (res).create (10);
LOOP_VINFO_SLP_UNROLLING_FACTOR (res) = 1;
LOOP_VINFO_TARGET_COST_DATA (res) = init_cost (loop);
LOOP_VINFO_PEELING_FOR_GAPS (res) = false;
LOOP_VINFO_PEELING_FOR_NITER (res) = false;
LOOP_VINFO_OPERANDS_SWAPPED (res) = false;
return res;
}
/* Function destroy_loop_vec_info.
Free LOOP_VINFO struct, as well as all the stmt_vec_info structs of all the
stmts in the loop. */
void
destroy_loop_vec_info (loop_vec_info loop_vinfo, bool clean_stmts)
{
struct loop *loop;
basic_block *bbs;
int nbbs;
gimple_stmt_iterator si;
int j;
vec<slp_instance> slp_instances;
slp_instance instance;
bool swapped;
if (!loop_vinfo)
return;
loop = LOOP_VINFO_LOOP (loop_vinfo);
bbs = LOOP_VINFO_BBS (loop_vinfo);
nbbs = clean_stmts ? loop->num_nodes : 0;
swapped = LOOP_VINFO_OPERANDS_SWAPPED (loop_vinfo);
for (j = 0; j < nbbs; j++)
{
basic_block bb = bbs[j];
for (si = gsi_start_phis (bb); !gsi_end_p (si); gsi_next (&si))
free_stmt_vec_info (gsi_stmt (si));
for (si = gsi_start_bb (bb); !gsi_end_p (si); )
{
gimple stmt = gsi_stmt (si);
/* We may have broken canonical form by moving a constant
into RHS1 of a commutative op. Fix such occurrences. */
if (swapped && is_gimple_assign (stmt))
{
enum tree_code code = gimple_assign_rhs_code (stmt);
if ((code == PLUS_EXPR
|| code == POINTER_PLUS_EXPR
|| code == MULT_EXPR)
&& CONSTANT_CLASS_P (gimple_assign_rhs1 (stmt)))
swap_ssa_operands (stmt,
gimple_assign_rhs1_ptr (stmt),
gimple_assign_rhs2_ptr (stmt));
}
/* Free stmt_vec_info. */
free_stmt_vec_info (stmt);
gsi_next (&si);
}
}
free (LOOP_VINFO_BBS (loop_vinfo));
vect_destroy_datarefs (loop_vinfo, NULL);
free_dependence_relations (LOOP_VINFO_DDRS (loop_vinfo));
LOOP_VINFO_LOOP_NEST (loop_vinfo).release ();
LOOP_VINFO_MAY_MISALIGN_STMTS (loop_vinfo).release ();
LOOP_VINFO_MAY_ALIAS_DDRS (loop_vinfo).release ();
slp_instances = LOOP_VINFO_SLP_INSTANCES (loop_vinfo);
FOR_EACH_VEC_ELT (slp_instances, j, instance)
vect_free_slp_instance (instance);
LOOP_VINFO_SLP_INSTANCES (loop_vinfo).release ();
LOOP_VINFO_GROUPED_STORES (loop_vinfo).release ();
LOOP_VINFO_REDUCTIONS (loop_vinfo).release ();
LOOP_VINFO_REDUCTION_CHAINS (loop_vinfo).release ();
delete LOOP_VINFO_PEELING_HTAB (loop_vinfo);
LOOP_VINFO_PEELING_HTAB (loop_vinfo) = NULL;
destroy_cost_data (LOOP_VINFO_TARGET_COST_DATA (loop_vinfo));
free (loop_vinfo);
loop->aux = NULL;
}
/* Function vect_analyze_loop_1.
Apply a set of analyses on LOOP, and create a loop_vec_info struct
for it. The different analyses will record information in the
loop_vec_info struct. This is a subset of the analyses applied in
vect_analyze_loop, to be applied on an inner-loop nested in the loop
that is now considered for (outer-loop) vectorization. */
static loop_vec_info
vect_analyze_loop_1 (struct loop *loop)
{
loop_vec_info loop_vinfo;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"===== analyze_loop_nest_1 =====\n");
/* Check the CFG characteristics of the loop (nesting, entry/exit, etc. */
loop_vinfo = vect_analyze_loop_form (loop);
if (!loop_vinfo)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad inner-loop form.\n");
return NULL;
}
return loop_vinfo;
}
/* Function vect_analyze_loop_form.
Verify that certain CFG restrictions hold, including:
- the loop has a pre-header
- the loop has a single entry and exit
- the loop exit condition is simple enough, and the number of iterations
can be analyzed (a countable loop). */
loop_vec_info
vect_analyze_loop_form (struct loop *loop)
{
loop_vec_info loop_vinfo;
gcond *loop_cond;
tree number_of_iterations = NULL, number_of_iterationsm1 = NULL;
loop_vec_info inner_loop_vinfo = NULL;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"=== vect_analyze_loop_form ===\n");
/* Different restrictions apply when we are considering an inner-most loop,
vs. an outer (nested) loop.
(FORNOW. May want to relax some of these restrictions in the future). */
if (!loop->inner)
{
/* Inner-most loop. We currently require that the number of BBs is
exactly 2 (the header and latch). Vectorizable inner-most loops
look like this:
(pre-header)
|
header <--------+
| | |
| +--> latch --+
|
(exit-bb) */
if (loop->num_nodes != 2)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: control flow in loop.\n");
return NULL;
}
if (empty_block_p (loop->header))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: empty loop.\n");
return NULL;
}
}
else
{
struct loop *innerloop = loop->inner;
edge entryedge;
/* Nested loop. We currently require that the loop is doubly-nested,
contains a single inner loop, and the number of BBs is exactly 5.
Vectorizable outer-loops look like this:
(pre-header)
|
header <---+
| |
inner-loop |
| |
tail ------+
|
(exit-bb)
The inner-loop has the properties expected of inner-most loops
as described above. */
if ((loop->inner)->inner || (loop->inner)->next)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: multiple nested loops.\n");
return NULL;
}
/* Analyze the inner-loop. */
inner_loop_vinfo = vect_analyze_loop_1 (loop->inner);
if (!inner_loop_vinfo)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: Bad inner loop.\n");
return NULL;
}
if (!expr_invariant_in_loop_p (loop,
LOOP_VINFO_NITERS (inner_loop_vinfo)))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: inner-loop count not"
" invariant.\n");
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
if (loop->num_nodes != 5)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: control flow in loop.\n");
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
gcc_assert (EDGE_COUNT (innerloop->header->preds) == 2);
entryedge = EDGE_PRED (innerloop->header, 0);
if (EDGE_PRED (innerloop->header, 0)->src == innerloop->latch)
entryedge = EDGE_PRED (innerloop->header, 1);
if (entryedge->src != loop->header
|| !single_exit (innerloop)
|| single_exit (innerloop)->dest != EDGE_PRED (loop->latch, 0)->src)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: unsupported outerloop form.\n");
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"Considering outer-loop vectorization.\n");
}
if (!single_exit (loop)
|| EDGE_COUNT (loop->header->preds) != 2)
{
if (dump_enabled_p ())
{
if (!single_exit (loop))
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: multiple exits.\n");
else if (EDGE_COUNT (loop->header->preds) != 2)
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: too many incoming edges.\n");
}
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
/* We assume that the loop exit condition is at the end of the loop. i.e,
that the loop is represented as a do-while (with a proper if-guard
before the loop if needed), where the loop header contains all the
executable statements, and the latch is empty. */
if (!empty_block_p (loop->latch)
|| !gimple_seq_empty_p (phi_nodes (loop->latch)))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: latch block not empty.\n");
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
/* Make sure there exists a single-predecessor exit bb: */
if (!single_pred_p (single_exit (loop)->dest))
{
edge e = single_exit (loop);
if (!(e->flags & EDGE_ABNORMAL))
{
split_loop_exit_edge (e);
if (dump_enabled_p ())
dump_printf (MSG_NOTE, "split exit edge.\n");
}
else
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: abnormal loop exit edge.\n");
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
}
loop_cond = vect_get_loop_niters (loop, &number_of_iterations,
&number_of_iterationsm1);
if (!loop_cond)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: complicated exit condition.\n");
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
if (!number_of_iterations
|| chrec_contains_undetermined (number_of_iterations))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: number of iterations cannot be "
"computed.\n");
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
if (integer_zerop (number_of_iterations))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: number of iterations = 0.\n");
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, true);
return NULL;
}
loop_vinfo = new_loop_vec_info (loop);
LOOP_VINFO_NITERSM1 (loop_vinfo) = number_of_iterationsm1;
LOOP_VINFO_NITERS (loop_vinfo) = number_of_iterations;
LOOP_VINFO_NITERS_UNCHANGED (loop_vinfo) = number_of_iterations;
if (!LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo))
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"Symbolic number of iterations is ");
dump_generic_expr (MSG_NOTE, TDF_DETAILS, number_of_iterations);
dump_printf (MSG_NOTE, "\n");
}
}
STMT_VINFO_TYPE (vinfo_for_stmt (loop_cond)) = loop_exit_ctrl_vec_info_type;
/* CHECKME: May want to keep it around it in the future. */
if (inner_loop_vinfo)
destroy_loop_vec_info (inner_loop_vinfo, false);
gcc_assert (!loop->aux);
loop->aux = loop_vinfo;
return loop_vinfo;
}
/* Function vect_analyze_loop_operations.
Scan the loop stmts and make sure they are all vectorizable. */
static bool
vect_analyze_loop_operations (loop_vec_info loop_vinfo, bool slp)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block *bbs = LOOP_VINFO_BBS (loop_vinfo);
int nbbs = loop->num_nodes;
unsigned int vectorization_factor = 0;
int i;
stmt_vec_info stmt_info;
bool need_to_vectorize = false;
int min_profitable_iters;
int min_scalar_loop_bound;
unsigned int th;
bool only_slp_in_loop = true, ok;
HOST_WIDE_INT max_niter;
HOST_WIDE_INT estimated_niter;
int min_profitable_estimate;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"=== vect_analyze_loop_operations ===\n");
gcc_assert (LOOP_VINFO_VECT_FACTOR (loop_vinfo));
vectorization_factor = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
if (slp)
{
/* If all the stmts in the loop can be SLPed, we perform only SLP, and
vectorization factor of the loop is the unrolling factor required by
the SLP instances. If that unrolling factor is 1, we say, that we
perform pure SLP on loop - cross iteration parallelism is not
exploited. */
for (i = 0; i < nbbs; i++)
{
basic_block bb = bbs[i];
for (gimple_stmt_iterator si = gsi_start_bb (bb); !gsi_end_p (si);
gsi_next (&si))
{
gimple stmt = gsi_stmt (si);
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
gcc_assert (stmt_info);
if ((STMT_VINFO_RELEVANT_P (stmt_info)
|| VECTORIZABLE_CYCLE_DEF (STMT_VINFO_DEF_TYPE (stmt_info)))
&& !PURE_SLP_STMT (stmt_info))
/* STMT needs both SLP and loop-based vectorization. */
only_slp_in_loop = false;
}
}
if (only_slp_in_loop)
vectorization_factor = LOOP_VINFO_SLP_UNROLLING_FACTOR (loop_vinfo);
else
vectorization_factor = least_common_multiple (vectorization_factor,
LOOP_VINFO_SLP_UNROLLING_FACTOR (loop_vinfo));
LOOP_VINFO_VECT_FACTOR (loop_vinfo) = vectorization_factor;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"Updating vectorization factor to %d\n",
vectorization_factor);
}
for (i = 0; i < nbbs; i++)
{
basic_block bb = bbs[i];
for (gphi_iterator si = gsi_start_phis (bb); !gsi_end_p (si);
gsi_next (&si))
{
gphi *phi = si.phi ();
ok = true;
stmt_info = vinfo_for_stmt (phi);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "examining phi: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, phi, 0);
dump_printf (MSG_NOTE, "\n");
}
/* Inner-loop loop-closed exit phi in outer-loop vectorization
(i.e., a phi in the tail of the outer-loop). */
if (! is_loop_header_bb_p (bb))
{
/* FORNOW: we currently don't support the case that these phis
are not used in the outerloop (unless it is double reduction,
i.e., this phi is vect_reduction_def), cause this case
requires to actually do something here. */
if ((!STMT_VINFO_RELEVANT_P (stmt_info)
|| STMT_VINFO_LIVE_P (stmt_info))
&& STMT_VINFO_DEF_TYPE (stmt_info)
!= vect_double_reduction_def)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"Unsupported loop-closed phi in "
"outer-loop.\n");
return false;
}
/* If PHI is used in the outer loop, we check that its operand
is defined in the inner loop. */
if (STMT_VINFO_RELEVANT_P (stmt_info))
{
tree phi_op;
gimple op_def_stmt;
if (gimple_phi_num_args (phi) != 1)
return false;
phi_op = PHI_ARG_DEF (phi, 0);
if (TREE_CODE (phi_op) != SSA_NAME)
return false;
op_def_stmt = SSA_NAME_DEF_STMT (phi_op);
if (gimple_nop_p (op_def_stmt)
|| !flow_bb_inside_loop_p (loop, gimple_bb (op_def_stmt))
|| !vinfo_for_stmt (op_def_stmt))
return false;
if (STMT_VINFO_RELEVANT (vinfo_for_stmt (op_def_stmt))
!= vect_used_in_outer
&& STMT_VINFO_RELEVANT (vinfo_for_stmt (op_def_stmt))
!= vect_used_in_outer_by_reduction)
return false;
}
continue;
}
gcc_assert (stmt_info);
if (STMT_VINFO_LIVE_P (stmt_info))
{
/* FORNOW: not yet supported. */
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: value used after loop.\n");
return false;
}
if (STMT_VINFO_RELEVANT (stmt_info) == vect_used_in_scope
&& STMT_VINFO_DEF_TYPE (stmt_info) != vect_induction_def)
{
/* A scalar-dependence cycle that we don't support. */
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: scalar dependence cycle.\n");
return false;
}
if (STMT_VINFO_RELEVANT_P (stmt_info))
{
need_to_vectorize = true;
if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_induction_def)
ok = vectorizable_induction (phi, NULL, NULL);
}
if (!ok)
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: relevant phi not "
"supported: ");
dump_gimple_stmt (MSG_MISSED_OPTIMIZATION, TDF_SLIM, phi, 0);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
}
for (gimple_stmt_iterator si = gsi_start_bb (bb); !gsi_end_p (si);
gsi_next (&si))
{
gimple stmt = gsi_stmt (si);
if (!gimple_clobber_p (stmt)
&& !vect_analyze_stmt (stmt, &need_to_vectorize, NULL))
return false;
}
} /* bbs */
/* All operations in the loop are either irrelevant (deal with loop
control, or dead), or only used outside the loop and can be moved
out of the loop (e.g. invariants, inductions). The loop can be
optimized away by scalar optimizations. We're better off not
touching this loop. */
if (!need_to_vectorize)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"All the computation can be taken out of the loop.\n");
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: redundant loop. no profit to "
"vectorize.\n");
return false;
}
if (LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo) && dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"vectorization_factor = %d, niters = "
HOST_WIDE_INT_PRINT_DEC "\n", vectorization_factor,
LOOP_VINFO_INT_NITERS (loop_vinfo));
if ((LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
&& (LOOP_VINFO_INT_NITERS (loop_vinfo) < vectorization_factor))
|| ((max_niter = max_stmt_executions_int (loop)) != -1
&& (unsigned HOST_WIDE_INT) max_niter < vectorization_factor))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: iteration count too small.\n");
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: iteration count smaller than "
"vectorization factor.\n");
return false;
}
/* Analyze cost. Decide if worth while to vectorize. */
/* Once VF is set, SLP costs should be updated since the number of created
vector stmts depends on VF. */
vect_update_slp_costs_according_to_vf (loop_vinfo);
vect_estimate_min_profitable_iters (loop_vinfo, &min_profitable_iters,
&min_profitable_estimate);
LOOP_VINFO_COST_MODEL_MIN_ITERS (loop_vinfo) = min_profitable_iters;
if (min_profitable_iters < 0)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: vectorization not profitable.\n");
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: vector version will never be "
"profitable.\n");
return false;
}
min_scalar_loop_bound = ((PARAM_VALUE (PARAM_MIN_VECT_LOOP_BOUND)
* vectorization_factor) - 1);
/* Use the cost model only if it is more conservative than user specified
threshold. */
th = (unsigned) min_scalar_loop_bound;
if (min_profitable_iters
&& (!min_scalar_loop_bound
|| min_profitable_iters > min_scalar_loop_bound))
th = (unsigned) min_profitable_iters;
LOOP_VINFO_COST_MODEL_THRESHOLD (loop_vinfo) = th;
if (LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
&& LOOP_VINFO_INT_NITERS (loop_vinfo) <= th)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: vectorization not profitable.\n");
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"not vectorized: iteration count smaller than user "
"specified loop bound parameter or minimum profitable "
"iterations (whichever is more conservative).\n");
return false;
}
if ((estimated_niter = estimated_stmt_executions_int (loop)) != -1
&& ((unsigned HOST_WIDE_INT) estimated_niter
<= MAX (th, (unsigned)min_profitable_estimate)))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: estimated iteration count too "
"small.\n");
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"not vectorized: estimated iteration count smaller "
"than specified loop bound parameter or minimum "
"profitable iterations (whichever is more "
"conservative).\n");
return false;
}
return true;
}
/* Function vect_analyze_loop_2.
Apply a set of analyses on LOOP, and create a loop_vec_info struct
for it. The different analyses will record information in the
loop_vec_info struct. */
static bool
vect_analyze_loop_2 (loop_vec_info loop_vinfo)
{
bool ok, slp = false;
int max_vf = MAX_VECTORIZATION_FACTOR;
int min_vf = 2;
unsigned int th;
unsigned int n_stmts = 0;
/* Find all data references in the loop (which correspond to vdefs/vuses)
and analyze their evolution in the loop. Also adjust the minimal
vectorization factor according to the loads and stores.
FORNOW: Handle only simple, array references, which
alignment can be forced, and aligned pointer-references. */
ok = vect_analyze_data_refs (loop_vinfo, NULL, &min_vf, &n_stmts);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad data references.\n");
return false;
}
/* Classify all cross-iteration scalar data-flow cycles.
Cross-iteration cycles caused by virtual phis are analyzed separately. */
vect_analyze_scalar_cycles (loop_vinfo);
vect_pattern_recog (loop_vinfo, NULL);
/* Analyze the access patterns of the data-refs in the loop (consecutive,
complex, etc.). FORNOW: Only handle consecutive access pattern. */
ok = vect_analyze_data_ref_accesses (loop_vinfo, NULL);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad data access.\n");
return false;
}
/* Data-flow analysis to detect stmts that do not need to be vectorized. */
ok = vect_mark_stmts_to_be_vectorized (loop_vinfo);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"unexpected pattern.\n");
return false;
}
/* Analyze data dependences between the data-refs in the loop
and adjust the maximum vectorization factor according to
the dependences.
FORNOW: fail at the first data dependence that we encounter. */
ok = vect_analyze_data_ref_dependences (loop_vinfo, &max_vf);
if (!ok
|| max_vf < min_vf)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad data dependence.\n");
return false;
}
ok = vect_determine_vectorization_factor (loop_vinfo);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"can't determine vectorization factor.\n");
return false;
}
if (max_vf < LOOP_VINFO_VECT_FACTOR (loop_vinfo))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad data dependence.\n");
return false;
}
/* Analyze the alignment of the data-refs in the loop.
Fail if a data reference is found that cannot be vectorized. */
ok = vect_analyze_data_refs_alignment (loop_vinfo, NULL);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad data alignment.\n");
return false;
}
/* Prune the list of ddrs to be tested at run-time by versioning for alias.
It is important to call pruning after vect_analyze_data_ref_accesses,
since we use grouping information gathered by interleaving analysis. */
ok = vect_prune_runtime_alias_test_list (loop_vinfo);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"number of versioning for alias "
"run-time tests exceeds %d "
"(--param vect-max-version-for-alias-checks)\n",
PARAM_VALUE (PARAM_VECT_MAX_VERSION_FOR_ALIAS_CHECKS));
return false;
}
/* This pass will decide on using loop versioning and/or loop peeling in
order to enhance the alignment of data references in the loop. */
ok = vect_enhance_data_refs_alignment (loop_vinfo);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad data alignment.\n");
return false;
}
/* Check the SLP opportunities in the loop, analyze and build SLP trees. */
ok = vect_analyze_slp (loop_vinfo, NULL, n_stmts);
if (ok)
{
/* Decide which possible SLP instances to SLP. */
slp = vect_make_slp_decision (loop_vinfo);
/* Find stmts that need to be both vectorized and SLPed. */
vect_detect_hybrid_slp (loop_vinfo);
}
else
return false;
/* Scan all the operations in the loop and make sure they are
vectorizable. */
ok = vect_analyze_loop_operations (loop_vinfo, slp);
if (!ok)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad operation or unsupported loop bound.\n");
return false;
}
/* Decide whether we need to create an epilogue loop to handle
remaining scalar iterations. */
th = ((LOOP_VINFO_COST_MODEL_THRESHOLD (loop_vinfo) + 1)
/ LOOP_VINFO_VECT_FACTOR (loop_vinfo))
* LOOP_VINFO_VECT_FACTOR (loop_vinfo);
if (LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
&& LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo) > 0)
{
if (ctz_hwi (LOOP_VINFO_INT_NITERS (loop_vinfo)
- LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo))
< exact_log2 (LOOP_VINFO_VECT_FACTOR (loop_vinfo)))
LOOP_VINFO_PEELING_FOR_NITER (loop_vinfo) = true;
}
else if (LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo)
|| (tree_ctz (LOOP_VINFO_NITERS (loop_vinfo))
< (unsigned)exact_log2 (LOOP_VINFO_VECT_FACTOR (loop_vinfo))
/* In case of versioning, check if the maximum number of
iterations is greater than th. If they are identical,
the epilogue is unnecessary. */
&& ((!LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo)
&& !LOOP_REQUIRES_VERSIONING_FOR_ALIGNMENT (loop_vinfo))
|| (unsigned HOST_WIDE_INT)max_stmt_executions_int
(LOOP_VINFO_LOOP (loop_vinfo)) > th)))
LOOP_VINFO_PEELING_FOR_NITER (loop_vinfo) = true;
/* If an epilogue loop is required make sure we can create one. */
if (LOOP_VINFO_PEELING_FOR_GAPS (loop_vinfo)
|| LOOP_VINFO_PEELING_FOR_NITER (loop_vinfo))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location, "epilog loop required\n");
if (!vect_can_advance_ivs_p (loop_vinfo)
|| !slpeel_can_duplicate_loop_p (LOOP_VINFO_LOOP (loop_vinfo),
single_exit (LOOP_VINFO_LOOP
(loop_vinfo))))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"not vectorized: can't create required "
"epilog loop\n");
return false;
}
}
return true;
}
/* Function vect_analyze_loop.
Apply a set of analyses on LOOP, and create a loop_vec_info struct
for it. The different analyses will record information in the
loop_vec_info struct. */
loop_vec_info
vect_analyze_loop (struct loop *loop)
{
loop_vec_info loop_vinfo;
unsigned int vector_sizes;
/* Autodetect first vector size we try. */
current_vector_size = 0;
vector_sizes = targetm.vectorize.autovectorize_vector_sizes ();
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"===== analyze_loop_nest =====\n");
if (loop_outer (loop)
&& loop_vec_info_for_loop (loop_outer (loop))
&& LOOP_VINFO_VECTORIZABLE_P (loop_vec_info_for_loop (loop_outer (loop))))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"outer-loop already vectorized.\n");
return NULL;
}
while (1)
{
/* Check the CFG characteristics of the loop (nesting, entry/exit). */
loop_vinfo = vect_analyze_loop_form (loop);
if (!loop_vinfo)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"bad loop form.\n");
return NULL;
}
if (vect_analyze_loop_2 (loop_vinfo))
{
LOOP_VINFO_VECTORIZABLE_P (loop_vinfo) = 1;
return loop_vinfo;
}
destroy_loop_vec_info (loop_vinfo, true);
vector_sizes &= ~current_vector_size;
if (vector_sizes == 0
|| current_vector_size == 0)
return NULL;
/* Try the next biggest vector size. */
current_vector_size = 1 << floor_log2 (vector_sizes);
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"***** Re-trying analysis with "
"vector size %d\n", current_vector_size);
}
}
/* Function reduction_code_for_scalar_code
Input:
CODE - tree_code of a reduction operations.
Output:
REDUC_CODE - the corresponding tree-code to be used to reduce the
vector of partial results into a single scalar result, or ERROR_MARK
if the operation is a supported reduction operation, but does not have
such a tree-code.
Return FALSE if CODE currently cannot be vectorized as reduction. */
static bool
reduction_code_for_scalar_code (enum tree_code code,
enum tree_code *reduc_code)
{
switch (code)
{
case MAX_EXPR:
*reduc_code = REDUC_MAX_EXPR;
return true;
case MIN_EXPR:
*reduc_code = REDUC_MIN_EXPR;
return true;
case PLUS_EXPR:
*reduc_code = REDUC_PLUS_EXPR;
return true;
case MULT_EXPR:
case MINUS_EXPR:
case BIT_IOR_EXPR:
case BIT_XOR_EXPR:
case BIT_AND_EXPR:
*reduc_code = ERROR_MARK;
return true;
default:
return false;
}
}
/* Error reporting helper for vect_is_simple_reduction below. GIMPLE statement
STMT is printed with a message MSG. */
static void
report_vect_op (int msg_type, gimple stmt, const char *msg)
{
dump_printf_loc (msg_type, vect_location, "%s", msg);
dump_gimple_stmt (msg_type, TDF_SLIM, stmt, 0);
dump_printf (msg_type, "\n");
}
/* Detect SLP reduction of the form:
#a1 = phi <a5, a0>
a2 = operation (a1)
a3 = operation (a2)
a4 = operation (a3)
a5 = operation (a4)
#a = phi <a5>
PHI is the reduction phi node (#a1 = phi <a5, a0> above)
FIRST_STMT is the first reduction stmt in the chain
(a2 = operation (a1)).
Return TRUE if a reduction chain was detected. */
static bool
vect_is_slp_reduction (loop_vec_info loop_info, gimple phi, gimple first_stmt)
{
struct loop *loop = (gimple_bb (phi))->loop_father;
struct loop *vect_loop = LOOP_VINFO_LOOP (loop_info);
enum tree_code code;
gimple current_stmt = NULL, loop_use_stmt = NULL, first, next_stmt;
stmt_vec_info use_stmt_info, current_stmt_info;
tree lhs;
imm_use_iterator imm_iter;
use_operand_p use_p;
int nloop_uses, size = 0, n_out_of_loop_uses;
bool found = false;
if (loop != vect_loop)
return false;
lhs = PHI_RESULT (phi);
code = gimple_assign_rhs_code (first_stmt);
while (1)
{
nloop_uses = 0;
n_out_of_loop_uses = 0;
FOR_EACH_IMM_USE_FAST (use_p, imm_iter, lhs)
{
gimple use_stmt = USE_STMT (use_p);
if (is_gimple_debug (use_stmt))
continue;
/* Check if we got back to the reduction phi. */
if (use_stmt == phi)
{
loop_use_stmt = use_stmt;
found = true;
break;
}
if (flow_bb_inside_loop_p (loop, gimple_bb (use_stmt)))
{
if (vinfo_for_stmt (use_stmt)
&& !STMT_VINFO_IN_PATTERN_P (vinfo_for_stmt (use_stmt)))
{
loop_use_stmt = use_stmt;
nloop_uses++;
}
}
else
n_out_of_loop_uses++;
/* There are can be either a single use in the loop or two uses in
phi nodes. */
if (nloop_uses > 1 || (n_out_of_loop_uses && nloop_uses))
return false;
}
if (found)
break;
/* We reached a statement with no loop uses. */
if (nloop_uses == 0)
return false;
/* This is a loop exit phi, and we haven't reached the reduction phi. */
if (gimple_code (loop_use_stmt) == GIMPLE_PHI)
return false;
if (!is_gimple_assign (loop_use_stmt)
|| code != gimple_assign_rhs_code (loop_use_stmt)
|| !flow_bb_inside_loop_p (loop, gimple_bb (loop_use_stmt)))
return false;
/* Insert USE_STMT into reduction chain. */
use_stmt_info = vinfo_for_stmt (loop_use_stmt);
if (current_stmt)
{
current_stmt_info = vinfo_for_stmt (current_stmt);
GROUP_NEXT_ELEMENT (current_stmt_info) = loop_use_stmt;
GROUP_FIRST_ELEMENT (use_stmt_info)
= GROUP_FIRST_ELEMENT (current_stmt_info);
}
else
GROUP_FIRST_ELEMENT (use_stmt_info) = loop_use_stmt;
lhs = gimple_assign_lhs (loop_use_stmt);
current_stmt = loop_use_stmt;
size++;
}
if (!found || loop_use_stmt != phi || size < 2)
return false;
/* Swap the operands, if needed, to make the reduction operand be the second
operand. */
lhs = PHI_RESULT (phi);
next_stmt = GROUP_FIRST_ELEMENT (vinfo_for_stmt (current_stmt));
while (next_stmt)
{
if (gimple_assign_rhs2 (next_stmt) == lhs)
{
tree op = gimple_assign_rhs1 (next_stmt);
gimple def_stmt = NULL;
if (TREE_CODE (op) == SSA_NAME)
def_stmt = SSA_NAME_DEF_STMT (op);
/* Check that the other def is either defined in the loop
("vect_internal_def"), or it's an induction (defined by a
loop-header phi-node). */
if (def_stmt
&& gimple_bb (def_stmt)
&& flow_bb_inside_loop_p (loop, gimple_bb (def_stmt))
&& (is_gimple_assign (def_stmt)
|| is_gimple_call (def_stmt)
|| STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def_stmt))
== vect_induction_def
|| (gimple_code (def_stmt) == GIMPLE_PHI
&& STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def_stmt))
== vect_internal_def
&& !is_loop_header_bb_p (gimple_bb (def_stmt)))))
{
lhs = gimple_assign_lhs (next_stmt);
next_stmt = GROUP_NEXT_ELEMENT (vinfo_for_stmt (next_stmt));
continue;
}
return false;
}
else
{
tree op = gimple_assign_rhs2 (next_stmt);
gimple def_stmt = NULL;
if (TREE_CODE (op) == SSA_NAME)
def_stmt = SSA_NAME_DEF_STMT (op);
/* Check that the other def is either defined in the loop
("vect_internal_def"), or it's an induction (defined by a
loop-header phi-node). */
if (def_stmt
&& gimple_bb (def_stmt)
&& flow_bb_inside_loop_p (loop, gimple_bb (def_stmt))
&& (is_gimple_assign (def_stmt)
|| is_gimple_call (def_stmt)
|| STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def_stmt))
== vect_induction_def
|| (gimple_code (def_stmt) == GIMPLE_PHI
&& STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def_stmt))
== vect_internal_def
&& !is_loop_header_bb_p (gimple_bb (def_stmt)))))
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "swapping oprnds: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, next_stmt, 0);
dump_printf (MSG_NOTE, "\n");
}
swap_ssa_operands (next_stmt,
gimple_assign_rhs1_ptr (next_stmt),
gimple_assign_rhs2_ptr (next_stmt));
update_stmt (next_stmt);
if (CONSTANT_CLASS_P (gimple_assign_rhs1 (next_stmt)))
LOOP_VINFO_OPERANDS_SWAPPED (loop_info) = true;
}
else
return false;
}
lhs = gimple_assign_lhs (next_stmt);
next_stmt = GROUP_NEXT_ELEMENT (vinfo_for_stmt (next_stmt));
}
/* Save the chain for further analysis in SLP detection. */
first = GROUP_FIRST_ELEMENT (vinfo_for_stmt (current_stmt));
LOOP_VINFO_REDUCTION_CHAINS (loop_info).safe_push (first);
GROUP_SIZE (vinfo_for_stmt (first)) = size;
return true;
}
/* Function vect_is_simple_reduction_1
(1) Detect a cross-iteration def-use cycle that represents a simple
reduction computation. We look for the following pattern:
loop_header:
a1 = phi < a0, a2 >
a3 = ...
a2 = operation (a3, a1)
or
a3 = ...
loop_header:
a1 = phi < a0, a2 >
a2 = operation (a3, a1)
such that:
1. operation is commutative and associative and it is safe to
change the order of the computation (if CHECK_REDUCTION is true)
2. no uses for a2 in the loop (a2 is used out of the loop)
3. no uses of a1 in the loop besides the reduction operation
4. no uses of a1 outside the loop.
Conditions 1,4 are tested here.
Conditions 2,3 are tested in vect_mark_stmts_to_be_vectorized.
(2) Detect a cross-iteration def-use cycle in nested loops, i.e.,
nested cycles, if CHECK_REDUCTION is false.
(3) Detect cycles of phi nodes in outer-loop vectorization, i.e., double
reductions:
a1 = phi < a0, a2 >
inner loop (def of a3)
a2 = phi < a3 >
If MODIFY is true it tries also to rework the code in-place to enable
detection of more reduction patterns. For the time being we rewrite
"res -= RHS" into "rhs += -RHS" when it seems worthwhile.
*/
static gimple
vect_is_simple_reduction_1 (loop_vec_info loop_info, gimple phi,
bool check_reduction, bool *double_reduc,
bool modify)
{
struct loop *loop = (gimple_bb (phi))->loop_father;
struct loop *vect_loop = LOOP_VINFO_LOOP (loop_info);
edge latch_e = loop_latch_edge (loop);
tree loop_arg = PHI_ARG_DEF_FROM_EDGE (phi, latch_e);
gimple def_stmt, def1 = NULL, def2 = NULL;
enum tree_code orig_code, code;
tree op1, op2, op3 = NULL_TREE, op4 = NULL_TREE;
tree type;
int nloop_uses;
tree name;
imm_use_iterator imm_iter;
use_operand_p use_p;
bool phi_def;
*double_reduc = false;
/* If CHECK_REDUCTION is true, we assume inner-most loop vectorization,
otherwise, we assume outer loop vectorization. */
gcc_assert ((check_reduction && loop == vect_loop)
|| (!check_reduction && flow_loop_nested_p (vect_loop, loop)));
name = PHI_RESULT (phi);
/* ??? If there are no uses of the PHI result the inner loop reduction
won't be detected as possibly double-reduction by vectorizable_reduction
because that tries to walk the PHI arg from the preheader edge which
can be constant. See PR60382. */
if (has_zero_uses (name))
return NULL;
nloop_uses = 0;
FOR_EACH_IMM_USE_FAST (use_p, imm_iter, name)
{
gimple use_stmt = USE_STMT (use_p);
if (is_gimple_debug (use_stmt))
continue;
if (!flow_bb_inside_loop_p (loop, gimple_bb (use_stmt)))
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"intermediate value used outside loop.\n");
return NULL;
}
if (vinfo_for_stmt (use_stmt)
&& !is_pattern_stmt_p (vinfo_for_stmt (use_stmt)))
nloop_uses++;
if (nloop_uses > 1)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"reduction used in loop.\n");
return NULL;
}
}
if (TREE_CODE (loop_arg) != SSA_NAME)
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"reduction: not ssa_name: ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM, loop_arg);
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return NULL;
}
def_stmt = SSA_NAME_DEF_STMT (loop_arg);
if (!def_stmt)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"reduction: no def_stmt.\n");
return NULL;
}
if (!is_gimple_assign (def_stmt) && gimple_code (def_stmt) != GIMPLE_PHI)
{
if (dump_enabled_p ())
{
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, def_stmt, 0);
dump_printf (MSG_NOTE, "\n");
}
return NULL;
}
if (is_gimple_assign (def_stmt))
{
name = gimple_assign_lhs (def_stmt);
phi_def = false;
}
else
{
name = PHI_RESULT (def_stmt);
phi_def = true;
}
nloop_uses = 0;
FOR_EACH_IMM_USE_FAST (use_p, imm_iter, name)
{
gimple use_stmt = USE_STMT (use_p);
if (is_gimple_debug (use_stmt))
continue;
if (flow_bb_inside_loop_p (loop, gimple_bb (use_stmt))
&& vinfo_for_stmt (use_stmt)
&& !is_pattern_stmt_p (vinfo_for_stmt (use_stmt)))
nloop_uses++;
if (nloop_uses > 1)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"reduction used in loop.\n");
return NULL;
}
}
/* If DEF_STMT is a phi node itself, we expect it to have a single argument
defined in the inner loop. */
if (phi_def)
{
op1 = PHI_ARG_DEF (def_stmt, 0);
if (gimple_phi_num_args (def_stmt) != 1
|| TREE_CODE (op1) != SSA_NAME)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"unsupported phi node definition.\n");
return NULL;
}
def1 = SSA_NAME_DEF_STMT (op1);
if (gimple_bb (def1)
&& flow_bb_inside_loop_p (loop, gimple_bb (def_stmt))
&& loop->inner
&& flow_bb_inside_loop_p (loop->inner, gimple_bb (def1))
&& is_gimple_assign (def1))
{
if (dump_enabled_p ())
report_vect_op (MSG_NOTE, def_stmt,
"detected double reduction: ");
*double_reduc = true;
return def_stmt;
}
return NULL;
}
code = orig_code = gimple_assign_rhs_code (def_stmt);
/* We can handle "res -= x[i]", which is non-associative by
simply rewriting this into "res += -x[i]". Avoid changing
gimple instruction for the first simple tests and only do this
if we're allowed to change code at all. */
if (code == MINUS_EXPR
&& modify
&& (op1 = gimple_assign_rhs1 (def_stmt))
&& TREE_CODE (op1) == SSA_NAME
&& SSA_NAME_DEF_STMT (op1) == phi)
code = PLUS_EXPR;
if (check_reduction
&& (!commutative_tree_code (code) || !associative_tree_code (code)))
{
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: not commutative/associative: ");
return NULL;
}
if (get_gimple_rhs_class (code) != GIMPLE_BINARY_RHS)
{
if (code != COND_EXPR)
{
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: not binary operation: ");
return NULL;
}
op3 = gimple_assign_rhs1 (def_stmt);
if (COMPARISON_CLASS_P (op3))
{
op4 = TREE_OPERAND (op3, 1);
op3 = TREE_OPERAND (op3, 0);
}
op1 = gimple_assign_rhs2 (def_stmt);
op2 = gimple_assign_rhs3 (def_stmt);
if (TREE_CODE (op1) != SSA_NAME && TREE_CODE (op2) != SSA_NAME)
{
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: uses not ssa_names: ");
return NULL;
}
}
else
{
op1 = gimple_assign_rhs1 (def_stmt);
op2 = gimple_assign_rhs2 (def_stmt);
if (TREE_CODE (op1) != SSA_NAME && TREE_CODE (op2) != SSA_NAME)
{
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: uses not ssa_names: ");
return NULL;
}
}
type = TREE_TYPE (gimple_assign_lhs (def_stmt));
if ((TREE_CODE (op1) == SSA_NAME
&& !types_compatible_p (type,TREE_TYPE (op1)))
|| (TREE_CODE (op2) == SSA_NAME
&& !types_compatible_p (type, TREE_TYPE (op2)))
|| (op3 && TREE_CODE (op3) == SSA_NAME
&& !types_compatible_p (type, TREE_TYPE (op3)))
|| (op4 && TREE_CODE (op4) == SSA_NAME
&& !types_compatible_p (type, TREE_TYPE (op4))))
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"reduction: multiple types: operation type: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM, type);
dump_printf (MSG_NOTE, ", operands types: ");
dump_generic_expr (MSG_NOTE, TDF_SLIM,
TREE_TYPE (op1));
dump_printf (MSG_NOTE, ",");
dump_generic_expr (MSG_NOTE, TDF_SLIM,
TREE_TYPE (op2));
if (op3)
{
dump_printf (MSG_NOTE, ",");
dump_generic_expr (MSG_NOTE, TDF_SLIM,
TREE_TYPE (op3));
}
if (op4)
{
dump_printf (MSG_NOTE, ",");
dump_generic_expr (MSG_NOTE, TDF_SLIM,
TREE_TYPE (op4));
}
dump_printf (MSG_NOTE, "\n");
}
return NULL;
}
/* Check that it's ok to change the order of the computation.
Generally, when vectorizing a reduction we change the order of the
computation. This may change the behavior of the program in some
cases, so we need to check that this is ok. One exception is when
vectorizing an outer-loop: the inner-loop is executed sequentially,
and therefore vectorizing reductions in the inner-loop during
outer-loop vectorization is safe. */
/* CHECKME: check for !flag_finite_math_only too? */
if (SCALAR_FLOAT_TYPE_P (type) && !flag_associative_math
&& check_reduction)
{
/* Changing the order of operations changes the semantics. */
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: unsafe fp math optimization: ");
return NULL;
}
else if (INTEGRAL_TYPE_P (type) && TYPE_OVERFLOW_TRAPS (type)
&& check_reduction)
{
/* Changing the order of operations changes the semantics. */
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: unsafe int math optimization: ");
return NULL;
}
else if (SAT_FIXED_POINT_TYPE_P (type) && check_reduction)
{
/* Changing the order of operations changes the semantics. */
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: unsafe fixed-point math optimization: ");
return NULL;
}
/* If we detected "res -= x[i]" earlier, rewrite it into
"res += -x[i]" now. If this turns out to be useless reassoc
will clean it up again. */
if (orig_code == MINUS_EXPR)
{
tree rhs = gimple_assign_rhs2 (def_stmt);
tree negrhs = make_ssa_name (TREE_TYPE (rhs));
gimple negate_stmt = gimple_build_assign (negrhs, NEGATE_EXPR, rhs);
gimple_stmt_iterator gsi = gsi_for_stmt (def_stmt);
set_vinfo_for_stmt (negate_stmt, new_stmt_vec_info (negate_stmt,
loop_info, NULL));
gsi_insert_before (&gsi, negate_stmt, GSI_NEW_STMT);
gimple_assign_set_rhs2 (def_stmt, negrhs);
gimple_assign_set_rhs_code (def_stmt, PLUS_EXPR);
update_stmt (def_stmt);
}
/* Reduction is safe. We're dealing with one of the following:
1) integer arithmetic and no trapv
2) floating point arithmetic, and special flags permit this optimization
3) nested cycle (i.e., outer loop vectorization). */
if (TREE_CODE (op1) == SSA_NAME)
def1 = SSA_NAME_DEF_STMT (op1);
if (TREE_CODE (op2) == SSA_NAME)
def2 = SSA_NAME_DEF_STMT (op2);
if (code != COND_EXPR
&& ((!def1 || gimple_nop_p (def1)) && (!def2 || gimple_nop_p (def2))))
{
if (dump_enabled_p ())
report_vect_op (MSG_NOTE, def_stmt, "reduction: no defs for operands: ");
return NULL;
}
/* Check that one def is the reduction def, defined by PHI,
the other def is either defined in the loop ("vect_internal_def"),
or it's an induction (defined by a loop-header phi-node). */
if (def2 && def2 == phi
&& (code == COND_EXPR
|| !def1 || gimple_nop_p (def1)
|| !flow_bb_inside_loop_p (loop, gimple_bb (def1))
|| (def1 && flow_bb_inside_loop_p (loop, gimple_bb (def1))
&& (is_gimple_assign (def1)
|| is_gimple_call (def1)
|| STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def1))
== vect_induction_def
|| (gimple_code (def1) == GIMPLE_PHI
&& STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def1))
== vect_internal_def
&& !is_loop_header_bb_p (gimple_bb (def1)))))))
{
if (dump_enabled_p ())
report_vect_op (MSG_NOTE, def_stmt, "detected reduction: ");
return def_stmt;
}
if (def1 && def1 == phi
&& (code == COND_EXPR
|| !def2 || gimple_nop_p (def2)
|| !flow_bb_inside_loop_p (loop, gimple_bb (def2))
|| (def2 && flow_bb_inside_loop_p (loop, gimple_bb (def2))
&& (is_gimple_assign (def2)
|| is_gimple_call (def2)
|| STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def2))
== vect_induction_def
|| (gimple_code (def2) == GIMPLE_PHI
&& STMT_VINFO_DEF_TYPE (vinfo_for_stmt (def2))
== vect_internal_def
&& !is_loop_header_bb_p (gimple_bb (def2)))))))
{
if (check_reduction)
{
/* Swap operands (just for simplicity - so that the rest of the code
can assume that the reduction variable is always the last (second)
argument). */
if (dump_enabled_p ())
report_vect_op (MSG_NOTE, def_stmt,
"detected reduction: need to swap operands: ");
swap_ssa_operands (def_stmt, gimple_assign_rhs1_ptr (def_stmt),
gimple_assign_rhs2_ptr (def_stmt));
if (CONSTANT_CLASS_P (gimple_assign_rhs1 (def_stmt)))
LOOP_VINFO_OPERANDS_SWAPPED (loop_info) = true;
}
else
{
if (dump_enabled_p ())
report_vect_op (MSG_NOTE, def_stmt, "detected reduction: ");
}
return def_stmt;
}
/* Try to find SLP reduction chain. */
if (check_reduction && vect_is_slp_reduction (loop_info, phi, def_stmt))
{
if (dump_enabled_p ())
report_vect_op (MSG_NOTE, def_stmt,
"reduction: detected reduction chain: ");
return def_stmt;
}
if (dump_enabled_p ())
report_vect_op (MSG_MISSED_OPTIMIZATION, def_stmt,
"reduction: unknown pattern: ");
return NULL;
}
/* Wrapper around vect_is_simple_reduction_1, that won't modify code
in-place. Arguments as there. */
static gimple
vect_is_simple_reduction (loop_vec_info loop_info, gimple phi,
bool check_reduction, bool *double_reduc)
{
return vect_is_simple_reduction_1 (loop_info, phi, check_reduction,
double_reduc, false);
}
/* Wrapper around vect_is_simple_reduction_1, which will modify code
in-place if it enables detection of more reductions. Arguments
as there. */
gimple
vect_force_simple_reduction (loop_vec_info loop_info, gimple phi,
bool check_reduction, bool *double_reduc)
{
return vect_is_simple_reduction_1 (loop_info, phi, check_reduction,
double_reduc, true);
}
/* Calculate the cost of one scalar iteration of the loop. */
int
vect_get_single_scalar_iteration_cost (loop_vec_info loop_vinfo,
stmt_vector_for_cost *scalar_cost_vec)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block *bbs = LOOP_VINFO_BBS (loop_vinfo);
int nbbs = loop->num_nodes, factor, scalar_single_iter_cost = 0;
int innerloop_iters, i;
/* Count statements in scalar loop. Using this as scalar cost for a single
iteration for now.
TODO: Add outer loop support.
TODO: Consider assigning different costs to different scalar
statements. */
/* FORNOW. */
innerloop_iters = 1;
if (loop->inner)
innerloop_iters = 50; /* FIXME */
for (i = 0; i < nbbs; i++)
{
gimple_stmt_iterator si;
basic_block bb = bbs[i];
if (bb->loop_father == loop->inner)
factor = innerloop_iters;
else
factor = 1;
for (si = gsi_start_bb (bb); !gsi_end_p (si); gsi_next (&si))
{
gimple stmt = gsi_stmt (si);
stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
if (!is_gimple_assign (stmt) && !is_gimple_call (stmt))
continue;
/* Skip stmts that are not vectorized inside the loop. */
if (stmt_info
&& !STMT_VINFO_RELEVANT_P (stmt_info)
&& (!STMT_VINFO_LIVE_P (stmt_info)
|| !VECTORIZABLE_CYCLE_DEF (STMT_VINFO_DEF_TYPE (stmt_info)))
&& !STMT_VINFO_IN_PATTERN_P (stmt_info))
continue;
vect_cost_for_stmt kind;
if (STMT_VINFO_DATA_REF (vinfo_for_stmt (stmt)))
{
if (DR_IS_READ (STMT_VINFO_DATA_REF (vinfo_for_stmt (stmt))))
kind = scalar_load;
else
kind = scalar_store;
}
else
kind = scalar_stmt;
scalar_single_iter_cost
+= record_stmt_cost (scalar_cost_vec, factor, kind,
NULL, 0, vect_prologue);
}
}
return scalar_single_iter_cost;
}
/* Calculate cost of peeling the loop PEEL_ITERS_PROLOGUE times. */
int
vect_get_known_peeling_cost (loop_vec_info loop_vinfo, int peel_iters_prologue,
int *peel_iters_epilogue,
stmt_vector_for_cost *scalar_cost_vec,
stmt_vector_for_cost *prologue_cost_vec,
stmt_vector_for_cost *epilogue_cost_vec)
{
int retval = 0;
int vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
if (!LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo))
{
*peel_iters_epilogue = vf/2;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"cost model: epilogue peel iters set to vf/2 "
"because loop iterations are unknown .\n");
/* If peeled iterations are known but number of scalar loop
iterations are unknown, count a taken branch per peeled loop. */
retval = record_stmt_cost (prologue_cost_vec, 1, cond_branch_taken,
NULL, 0, vect_prologue);
retval = record_stmt_cost (prologue_cost_vec, 1, cond_branch_taken,
NULL, 0, vect_epilogue);
}
else
{
int niters = LOOP_VINFO_INT_NITERS (loop_vinfo);
peel_iters_prologue = niters < peel_iters_prologue ?
niters : peel_iters_prologue;
*peel_iters_epilogue = (niters - peel_iters_prologue) % vf;
/* If we need to peel for gaps, but no peeling is required, we have to
peel VF iterations. */
if (LOOP_VINFO_PEELING_FOR_GAPS (loop_vinfo) && !*peel_iters_epilogue)
*peel_iters_epilogue = vf;
}
stmt_info_for_cost *si;
int j;
if (peel_iters_prologue)
FOR_EACH_VEC_ELT (*scalar_cost_vec, j, si)
retval += record_stmt_cost (prologue_cost_vec,
si->count * peel_iters_prologue,
si->kind, NULL, si->misalign,
vect_prologue);
if (*peel_iters_epilogue)
FOR_EACH_VEC_ELT (*scalar_cost_vec, j, si)
retval += record_stmt_cost (epilogue_cost_vec,
si->count * *peel_iters_epilogue,
si->kind, NULL, si->misalign,
vect_epilogue);
return retval;
}
/* Function vect_estimate_min_profitable_iters
Return the number of iterations required for the vector version of the
loop to be profitable relative to the cost of the scalar version of the
loop. */
static void
vect_estimate_min_profitable_iters (loop_vec_info loop_vinfo,
int *ret_min_profitable_niters,
int *ret_min_profitable_estimate)
{
int min_profitable_iters;
int min_profitable_estimate;
int peel_iters_prologue;
int peel_iters_epilogue;
unsigned vec_inside_cost = 0;
int vec_outside_cost = 0;
unsigned vec_prologue_cost = 0;
unsigned vec_epilogue_cost = 0;
int scalar_single_iter_cost = 0;
int scalar_outside_cost = 0;
int vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
int npeel = LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo);
void *target_cost_data = LOOP_VINFO_TARGET_COST_DATA (loop_vinfo);
/* Cost model disabled. */
if (unlimited_cost_model (LOOP_VINFO_LOOP (loop_vinfo)))
{
dump_printf_loc (MSG_NOTE, vect_location, "cost model disabled.\n");
*ret_min_profitable_niters = 0;
*ret_min_profitable_estimate = 0;
return;
}
/* Requires loop versioning tests to handle misalignment. */
if (LOOP_REQUIRES_VERSIONING_FOR_ALIGNMENT (loop_vinfo))
{
/* FIXME: Make cost depend on complexity of individual check. */
unsigned len = LOOP_VINFO_MAY_MISALIGN_STMTS (loop_vinfo).length ();
(void) add_stmt_cost (target_cost_data, len, vector_stmt, NULL, 0,
vect_prologue);
dump_printf (MSG_NOTE,
"cost model: Adding cost of checks for loop "
"versioning to treat misalignment.\n");
}
/* Requires loop versioning with alias checks. */
if (LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
{
/* FIXME: Make cost depend on complexity of individual check. */
unsigned len = LOOP_VINFO_COMP_ALIAS_DDRS (loop_vinfo).length ();
(void) add_stmt_cost (target_cost_data, len, vector_stmt, NULL, 0,
vect_prologue);
dump_printf (MSG_NOTE,
"cost model: Adding cost of checks for loop "
"versioning aliasing.\n");
}
if (LOOP_REQUIRES_VERSIONING_FOR_ALIGNMENT (loop_vinfo)
|| LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
(void) add_stmt_cost (target_cost_data, 1, cond_branch_taken, NULL, 0,
vect_prologue);
/* Count statements in scalar loop. Using this as scalar cost for a single
iteration for now.
TODO: Add outer loop support.
TODO: Consider assigning different costs to different scalar
statements. */
auto_vec<stmt_info_for_cost> scalar_cost_vec;
scalar_single_iter_cost
= vect_get_single_scalar_iteration_cost (loop_vinfo, &scalar_cost_vec);
/* Add additional cost for the peeled instructions in prologue and epilogue
loop.
FORNOW: If we don't know the value of peel_iters for prologue or epilogue
at compile-time - we assume it's vf/2 (the worst would be vf-1).
TODO: Build an expression that represents peel_iters for prologue and
epilogue to be used in a run-time test. */
if (npeel < 0)
{
peel_iters_prologue = vf/2;
dump_printf (MSG_NOTE, "cost model: "
"prologue peel iters set to vf/2.\n");
/* If peeling for alignment is unknown, loop bound of main loop becomes
unknown. */
peel_iters_epilogue = vf/2;
dump_printf (MSG_NOTE, "cost model: "
"epilogue peel iters set to vf/2 because "
"peeling for alignment is unknown.\n");
/* If peeled iterations are unknown, count a taken branch and a not taken
branch per peeled loop. Even if scalar loop iterations are known,
vector iterations are not known since peeled prologue iterations are
not known. Hence guards remain the same. */
(void) add_stmt_cost (target_cost_data, 1, cond_branch_taken,
NULL, 0, vect_prologue);
(void) add_stmt_cost (target_cost_data, 1, cond_branch_not_taken,
NULL, 0, vect_prologue);
(void) add_stmt_cost (target_cost_data, 1, cond_branch_taken,
NULL, 0, vect_epilogue);
(void) add_stmt_cost (target_cost_data, 1, cond_branch_not_taken,
NULL, 0, vect_epilogue);
stmt_info_for_cost *si;
int j;
FOR_EACH_VEC_ELT (scalar_cost_vec, j, si)
{
struct _stmt_vec_info *stmt_info
= si->stmt ? vinfo_for_stmt (si->stmt) : NULL;
(void) add_stmt_cost (target_cost_data,
si->count * peel_iters_prologue,
si->kind, stmt_info, si->misalign,
vect_prologue);
(void) add_stmt_cost (target_cost_data,
si->count * peel_iters_epilogue,
si->kind, stmt_info, si->misalign,
vect_epilogue);
}
}
else
{
stmt_vector_for_cost prologue_cost_vec, epilogue_cost_vec;
stmt_info_for_cost *si;
int j;
void *data = LOOP_VINFO_TARGET_COST_DATA (loop_vinfo);
prologue_cost_vec.create (2);
epilogue_cost_vec.create (2);
peel_iters_prologue = npeel;
(void) vect_get_known_peeling_cost (loop_vinfo, peel_iters_prologue,
&peel_iters_epilogue,
&scalar_cost_vec,
&prologue_cost_vec,
&epilogue_cost_vec);
FOR_EACH_VEC_ELT (prologue_cost_vec, j, si)
{
struct _stmt_vec_info *stmt_info
= si->stmt ? vinfo_for_stmt (si->stmt) : NULL;
(void) add_stmt_cost (data, si->count, si->kind, stmt_info,
si->misalign, vect_prologue);
}
FOR_EACH_VEC_ELT (epilogue_cost_vec, j, si)
{
struct _stmt_vec_info *stmt_info
= si->stmt ? vinfo_for_stmt (si->stmt) : NULL;
(void) add_stmt_cost (data, si->count, si->kind, stmt_info,
si->misalign, vect_epilogue);
}
prologue_cost_vec.release ();
epilogue_cost_vec.release ();
}
/* FORNOW: The scalar outside cost is incremented in one of the
following ways:
1. The vectorizer checks for alignment and aliasing and generates
a condition that allows dynamic vectorization. A cost model
check is ANDED with the versioning condition. Hence scalar code
path now has the added cost of the versioning check.
if (cost > th & versioning_check)
jmp to vector code
Hence run-time scalar is incremented by not-taken branch cost.
2. The vectorizer then checks if a prologue is required. If the
cost model check was not done before during versioning, it has to
be done before the prologue check.
if (cost <= th)
prologue = scalar_iters
if (prologue == 0)
jmp to vector code
else
execute prologue
if (prologue == num_iters)
go to exit
Hence the run-time scalar cost is incremented by a taken branch,
plus a not-taken branch, plus a taken branch cost.
3. The vectorizer then checks if an epilogue is required. If the
cost model check was not done before during prologue check, it
has to be done with the epilogue check.
if (prologue == 0)
jmp to vector code
else
execute prologue
if (prologue == num_iters)
go to exit
vector code:
if ((cost <= th) | (scalar_iters-prologue-epilogue == 0))
jmp to epilogue
Hence the run-time scalar cost should be incremented by 2 taken
branches.
TODO: The back end may reorder the BBS's differently and reverse
conditions/branch directions. Change the estimates below to
something more reasonable. */
/* If the number of iterations is known and we do not do versioning, we can
decide whether to vectorize at compile time. Hence the scalar version
do not carry cost model guard costs. */
if (!LOOP_VINFO_NITERS_KNOWN_P (loop_vinfo)
|| LOOP_REQUIRES_VERSIONING_FOR_ALIGNMENT (loop_vinfo)
|| LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
{
/* Cost model check occurs at versioning. */
if (LOOP_REQUIRES_VERSIONING_FOR_ALIGNMENT (loop_vinfo)
|| LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
scalar_outside_cost += vect_get_stmt_cost (cond_branch_not_taken);
else
{
/* Cost model check occurs at prologue generation. */
if (LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo) < 0)
scalar_outside_cost += 2 * vect_get_stmt_cost (cond_branch_taken)
+ vect_get_stmt_cost (cond_branch_not_taken);
/* Cost model check occurs at epilogue generation. */
else
scalar_outside_cost += 2 * vect_get_stmt_cost (cond_branch_taken);
}
}
/* Complete the target-specific cost calculations. */
finish_cost (LOOP_VINFO_TARGET_COST_DATA (loop_vinfo), &vec_prologue_cost,
&vec_inside_cost, &vec_epilogue_cost);
vec_outside_cost = (int)(vec_prologue_cost + vec_epilogue_cost);
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location, "Cost model analysis: \n");
dump_printf (MSG_NOTE, " Vector inside of loop cost: %d\n",
vec_inside_cost);
dump_printf (MSG_NOTE, " Vector prologue cost: %d\n",
vec_prologue_cost);
dump_printf (MSG_NOTE, " Vector epilogue cost: %d\n",
vec_epilogue_cost);
dump_printf (MSG_NOTE, " Scalar iteration cost: %d\n",
scalar_single_iter_cost);
dump_printf (MSG_NOTE, " Scalar outside cost: %d\n",
scalar_outside_cost);
dump_printf (MSG_NOTE, " Vector outside cost: %d\n",
vec_outside_cost);
dump_printf (MSG_NOTE, " prologue iterations: %d\n",
peel_iters_prologue);
dump_printf (MSG_NOTE, " epilogue iterations: %d\n",
peel_iters_epilogue);
}
/* Calculate number of iterations required to make the vector version
profitable, relative to the loop bodies only. The following condition
must hold true:
SIC * niters + SOC > VIC * ((niters-PL_ITERS-EP_ITERS)/VF) + VOC
where
SIC = scalar iteration cost, VIC = vector iteration cost,
VOC = vector outside cost, VF = vectorization factor,
PL_ITERS = prologue iterations, EP_ITERS= epilogue iterations
SOC = scalar outside cost for run time cost model check. */
if ((scalar_single_iter_cost * vf) > (int) vec_inside_cost)
{
if (vec_outside_cost <= 0)
min_profitable_iters = 1;
else
{
min_profitable_iters = ((vec_outside_cost - scalar_outside_cost) * vf
- vec_inside_cost * peel_iters_prologue
- vec_inside_cost * peel_iters_epilogue)
/ ((scalar_single_iter_cost * vf)
- vec_inside_cost);
if ((scalar_single_iter_cost * vf * min_profitable_iters)
<= (((int) vec_inside_cost * min_profitable_iters)
+ (((int) vec_outside_cost - scalar_outside_cost) * vf)))
min_profitable_iters++;
}
}
/* vector version will never be profitable. */
else
{
if (LOOP_VINFO_LOOP (loop_vinfo)->force_vectorize)
warning_at (vect_location, OPT_Wopenmp_simd, "vectorization "
"did not happen for a simd loop");
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"cost model: the vector iteration cost = %d "
"divided by the scalar iteration cost = %d "
"is greater or equal to the vectorization factor = %d"
".\n",
vec_inside_cost, scalar_single_iter_cost, vf);
*ret_min_profitable_niters = -1;
*ret_min_profitable_estimate = -1;
return;
}
dump_printf (MSG_NOTE,
" Calculated minimum iters for profitability: %d\n",
min_profitable_iters);
min_profitable_iters =
min_profitable_iters < vf ? vf : min_profitable_iters;
/* Because the condition we create is:
if (niters <= min_profitable_iters)
then skip the vectorized loop. */
min_profitable_iters--;
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
" Runtime profitability threshold = %d\n",
min_profitable_iters);
*ret_min_profitable_niters = min_profitable_iters;
/* Calculate number of iterations required to make the vector version
profitable, relative to the loop bodies only.
Non-vectorized variant is SIC * niters and it must win over vector
variant on the expected loop trip count. The following condition must hold true:
SIC * niters > VIC * ((niters-PL_ITERS-EP_ITERS)/VF) + VOC + SOC */
if (vec_outside_cost <= 0)
min_profitable_estimate = 1;
else
{
min_profitable_estimate = ((vec_outside_cost + scalar_outside_cost) * vf
- vec_inside_cost * peel_iters_prologue
- vec_inside_cost * peel_iters_epilogue)
/ ((scalar_single_iter_cost * vf)
- vec_inside_cost);
}
min_profitable_estimate --;
min_profitable_estimate = MAX (min_profitable_estimate, min_profitable_iters);
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
" Static estimate profitability threshold = %d\n",
min_profitable_iters);
*ret_min_profitable_estimate = min_profitable_estimate;
}
/* Writes into SEL a mask for a vec_perm, equivalent to a vec_shr by OFFSET
vector elements (not bits) for a vector of mode MODE. */
static void
calc_vec_perm_mask_for_shift (enum machine_mode mode, unsigned int offset,
unsigned char *sel)
{
unsigned int i, nelt = GET_MODE_NUNITS (mode);
for (i = 0; i < nelt; i++)
sel[i] = (i + offset) & (2*nelt - 1);
}
/* Checks whether the target supports whole-vector shifts for vectors of mode
MODE. This is the case if _either_ the platform handles vec_shr_optab, _or_
it supports vec_perm_const with masks for all necessary shift amounts. */
static bool
have_whole_vector_shift (enum machine_mode mode)
{
if (optab_handler (vec_shr_optab, mode) != CODE_FOR_nothing)
return true;
if (direct_optab_handler (vec_perm_const_optab, mode) == CODE_FOR_nothing)
return false;
unsigned int i, nelt = GET_MODE_NUNITS (mode);
unsigned char *sel = XALLOCAVEC (unsigned char, nelt);
for (i = nelt/2; i >= 1; i/=2)
{
calc_vec_perm_mask_for_shift (mode, i, sel);
if (!can_vec_perm_p (mode, false, sel))
return false;
}
return true;
}
/* TODO: Close dependency between vect_model_*_cost and vectorizable_*
functions. Design better to avoid maintenance issues. */
/* Function vect_model_reduction_cost.
Models cost for a reduction operation, including the vector ops
generated within the strip-mine loop, the initial definition before
the loop, and the epilogue code that must be generated. */
static bool
vect_model_reduction_cost (stmt_vec_info stmt_info, enum tree_code reduc_code,
int ncopies)
{
int prologue_cost = 0, epilogue_cost = 0;
enum tree_code code;
optab optab;
tree vectype;
gimple stmt, orig_stmt;
tree reduction_op;
machine_mode mode;
loop_vec_info loop_vinfo = STMT_VINFO_LOOP_VINFO (stmt_info);
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
void *target_cost_data = LOOP_VINFO_TARGET_COST_DATA (loop_vinfo);
/* Cost of reduction op inside loop. */
unsigned inside_cost = add_stmt_cost (target_cost_data, ncopies, vector_stmt,
stmt_info, 0, vect_body);
stmt = STMT_VINFO_STMT (stmt_info);
switch (get_gimple_rhs_class (gimple_assign_rhs_code (stmt)))
{
case GIMPLE_SINGLE_RHS:
gcc_assert (TREE_OPERAND_LENGTH (gimple_assign_rhs1 (stmt)) == ternary_op);
reduction_op = TREE_OPERAND (gimple_assign_rhs1 (stmt), 2);
break;
case GIMPLE_UNARY_RHS:
reduction_op = gimple_assign_rhs1 (stmt);
break;
case GIMPLE_BINARY_RHS:
reduction_op = gimple_assign_rhs2 (stmt);
break;
case GIMPLE_TERNARY_RHS:
reduction_op = gimple_assign_rhs3 (stmt);
break;
default:
gcc_unreachable ();
}
vectype = get_vectype_for_scalar_type (TREE_TYPE (reduction_op));
if (!vectype)
{
if (dump_enabled_p ())
{
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
"unsupported data-type ");
dump_generic_expr (MSG_MISSED_OPTIMIZATION, TDF_SLIM,
TREE_TYPE (reduction_op));
dump_printf (MSG_MISSED_OPTIMIZATION, "\n");
}
return false;
}
mode = TYPE_MODE (vectype);
orig_stmt = STMT_VINFO_RELATED_STMT (stmt_info);
if (!orig_stmt)
orig_stmt = STMT_VINFO_STMT (stmt_info);
code = gimple_assign_rhs_code (orig_stmt);
/* Add in cost for initial definition. */
prologue_cost += add_stmt_cost (target_cost_data, 1, scalar_to_vec,
stmt_info, 0, vect_prologue);
/* Determine cost of epilogue code.
We have a reduction operator that will reduce the vector in one statement.
Also requires scalar extract. */
if (!nested_in_vect_loop_p (loop, orig_stmt))
{
if (reduc_code != ERROR_MARK)
{
epilogue_cost += add_stmt_cost (target_cost_data, 1, vector_stmt,
stmt_info, 0, vect_epilogue);
epilogue_cost += add_stmt_cost (target_cost_data, 1, vec_to_scalar,
stmt_info, 0, vect_epilogue);
}
else
{
int