/* Vectorizer Specific Loop Manipulations
- Copyright (C) 2003-2013 Free Software Foundation, Inc.
+ Copyright (C) 2003-2016 Free Software Foundation, Inc.
Contributed by Dorit Naishlos <dorit@il.ibm.com>
and Ira Rosen <irar@il.ibm.com>
#include "config.h"
#include "system.h"
#include "coretypes.h"
-#include "dumpfile.h"
-#include "tm.h"
-#include "ggc.h"
+#include "backend.h"
#include "tree.h"
-#include "basic-block.h"
-#include "gimple-pretty-print.h"
#include "gimple.h"
-#include "gimple-ssa.h"
+#include "cfghooks.h"
+#include "tree-pass.h"
+#include "ssa.h"
+#include "fold-const.h"
+#include "cfganal.h"
+#include "gimplify.h"
+#include "gimple-iterator.h"
+#include "gimplify-me.h"
#include "tree-cfg.h"
-#include "tree-phinodes.h"
-#include "ssa-iterators.h"
-#include "tree-ssanames.h"
#include "tree-ssa-loop-manip.h"
#include "tree-into-ssa.h"
#include "tree-ssa.h"
-#include "tree-pass.h"
#include "cfgloop.h"
-#include "diagnostic-core.h"
#include "tree-scalar-evolution.h"
#include "tree-vectorizer.h"
-#include "langhooks.h"
/*************************************************************************
Simple Loop Peeling Utilities
}
-/* Renames the variables in basic block BB. */
+/* Renames the variables in basic block BB. Allow renaming of PHI argumnets
+ on edges incoming from outer-block header if RENAME_FROM_OUTER_LOOP is
+ true. */
static void
-rename_variables_in_bb (basic_block bb)
+rename_variables_in_bb (basic_block bb, bool rename_from_outer_loop)
{
- gimple_stmt_iterator gsi;
- gimple stmt;
+ gimple *stmt;
use_operand_p use_p;
ssa_op_iter iter;
edge e;
edge_iterator ei;
struct loop *loop = bb->loop_father;
+ struct loop *outer_loop = NULL;
- for (gsi = gsi_start_bb (bb); !gsi_end_p (gsi); gsi_next (&gsi))
+ if (rename_from_outer_loop)
+ {
+ gcc_assert (loop);
+ outer_loop = loop_outer (loop);
+ }
+
+ for (gimple_stmt_iterator gsi = gsi_start_bb (bb); !gsi_end_p (gsi);
+ gsi_next (&gsi))
{
stmt = gsi_stmt (gsi);
FOR_EACH_SSA_USE_OPERAND (use_p, stmt, iter, SSA_OP_ALL_USES)
FOR_EACH_EDGE (e, ei, bb->preds)
{
- if (!flow_bb_inside_loop_p (loop, e->src))
+ if (!flow_bb_inside_loop_p (loop, e->src)
+ && (!rename_from_outer_loop || e->src != outer_loop->header))
continue;
- for (gsi = gsi_start_phis (bb); !gsi_end_p (gsi); gsi_next (&gsi))
- rename_use_op (PHI_ARG_DEF_PTR_FROM_EDGE (gsi_stmt (gsi), e));
+ for (gphi_iterator gsi = gsi_start_phis (bb); !gsi_end_p (gsi);
+ gsi_next (&gsi))
+ rename_use_op (PHI_ARG_DEF_PTR_FROM_EDGE (gsi.phi (), e));
}
}
-typedef struct
+struct adjust_info
{
tree from, to;
basic_block bb;
-} adjust_info;
+};
/* A stack of values to be adjusted in debug stmts. We have to
process them LIFO, so that the closest substitution applies. If we
tree orig_def = ai->from;
tree new_def = ai->to;
imm_use_iterator imm_iter;
- gimple stmt;
+ gimple *stmt;
basic_block bbdef = gimple_bb (SSA_NAME_DEF_STMT (orig_def));
gcc_assert (dom_info_available_p (CDI_DOMINATORS));
transformations. */
static void
-adjust_phi_and_debug_stmts (gimple update_phi, edge e, tree new_def)
+adjust_phi_and_debug_stmts (gimple *update_phi, edge e, tree new_def)
{
tree orig_def = PHI_ARG_DEF_FROM_EDGE (update_phi, e);
next_bb
The SSA names defined in the original loop have a current
- reaching definition that that records the corresponding new
- ssa-name used in the new duplicated loop copy.
+ reaching definition that records the corresponding new ssa-name
+ used in the new duplicated loop copy.
*/
/* Function slpeel_update_phi_nodes_for_guard1
slpeel_update_phi_nodes_for_guard1 (edge guard_edge, struct loop *loop,
bool is_new_loop, basic_block *new_exit_bb)
{
- gimple orig_phi, new_phi;
- gimple update_phi, update_phi2;
+ gphi *orig_phi, *new_phi;
+ gphi *update_phi, *update_phi2;
tree guard_arg, loop_arg;
basic_block new_merge_bb = guard_edge->dest;
edge e = EDGE_SUCC (new_merge_bb, 0);
basic_block orig_bb = loop->header;
edge new_exit_e;
tree current_new_name;
- gimple_stmt_iterator gsi_orig, gsi_update;
+ gphi_iterator gsi_orig, gsi_update;
/* Create new bb between loop and new_merge_bb. */
*new_exit_bb = split_edge (single_exit (loop));
{
source_location loop_locus, guard_locus;
tree new_res;
- orig_phi = gsi_stmt (gsi_orig);
- update_phi = gsi_stmt (gsi_update);
+ orig_phi = gsi_orig.phi ();
+ update_phi = gsi_update.phi ();
/** 1. Handle new-merge-point phis **/
/* 1.1. Generate new phi node in NEW_MERGE_BB: */
- new_res = copy_ssa_name (PHI_RESULT (orig_phi), NULL);
+ new_res = copy_ssa_name (PHI_RESULT (orig_phi));
new_phi = create_phi_node (new_res, new_merge_bb);
/* 1.2. NEW_MERGE_BB has two incoming edges: GUARD_EDGE and the exit-edge
continue;
/* 2.1. Generate new phi node in NEW_EXIT_BB: */
- new_res = copy_ssa_name (PHI_RESULT (orig_phi), NULL);
+ new_res = copy_ssa_name (PHI_RESULT (orig_phi));
new_phi = create_phi_node (new_res, *new_exit_bb);
/* 2.2. NEW_EXIT_BB has one incoming edge: the exit-edge of the loop. */
if (!current_new_name)
continue;
}
- gcc_assert (get_current_def (current_new_name) == NULL_TREE);
+ tree new_name = get_current_def (current_new_name);
+ /* Because of peeled_chrec optimization it is possible that we have
+ set this earlier. Verify the PHI has the same value. */
+ if (new_name)
+ {
+ gimple *phi = SSA_NAME_DEF_STMT (new_name);
+ gcc_assert (gimple_code (phi) == GIMPLE_PHI
+ && gimple_bb (phi) == *new_exit_bb
+ && (PHI_ARG_DEF_FROM_EDGE (phi, single_exit (loop))
+ == loop_arg));
+ continue;
+ }
set_current_def (current_new_name, PHI_RESULT (new_phi));
}
slpeel_update_phi_nodes_for_guard2 (edge guard_edge, struct loop *loop,
bool is_new_loop, basic_block *new_exit_bb)
{
- gimple orig_phi, new_phi;
- gimple update_phi, update_phi2;
+ gphi *orig_phi, *new_phi;
+ gphi *update_phi, *update_phi2;
tree guard_arg, loop_arg;
basic_block new_merge_bb = guard_edge->dest;
edge e = EDGE_SUCC (new_merge_bb, 0);
tree orig_def, orig_def_new_name;
tree new_name, new_name2;
tree arg;
- gimple_stmt_iterator gsi;
+ gphi_iterator gsi;
/* Create new bb between loop and new_merge_bb. */
*new_exit_bb = split_edge (single_exit (loop));
for (gsi = gsi_start_phis (update_bb); !gsi_end_p (gsi); gsi_next (&gsi))
{
tree new_res;
- update_phi = gsi_stmt (gsi);
+ update_phi = gsi.phi ();
orig_phi = update_phi;
orig_def = PHI_ARG_DEF_FROM_EDGE (orig_phi, e);
/* This loop-closed-phi actually doesn't represent a use
/** 1. Handle new-merge-point phis **/
/* 1.1. Generate new phi node in NEW_MERGE_BB: */
- new_res = copy_ssa_name (PHI_RESULT (orig_phi), NULL);
+ new_res = copy_ssa_name (PHI_RESULT (orig_phi));
new_phi = create_phi_node (new_res, new_merge_bb);
/* 1.2. NEW_MERGE_BB has two incoming edges: GUARD_EDGE and the exit-edge
/** 2. Handle loop-closed-ssa-form phis **/
/* 2.1. Generate new phi node in NEW_EXIT_BB: */
- new_res = copy_ssa_name (PHI_RESULT (orig_phi), NULL);
+ new_res = copy_ssa_name (PHI_RESULT (orig_phi));
new_phi = create_phi_node (new_res, *new_exit_bb);
/* 2.2. NEW_EXIT_BB has one incoming edge: the exit-edge of the loop. */
arg = guard_arg;
/* 3.2. Generate new phi node in GUARD_BB: */
- new_res = copy_ssa_name (PHI_RESULT (orig_phi), NULL);
+ new_res = copy_ssa_name (PHI_RESULT (orig_phi));
new_phi = create_phi_node (new_res, guard_edge->src);
/* 3.3. GUARD_BB has one incoming edge: */
slpeel_make_loop_iterate_ntimes (struct loop *loop, tree niters)
{
tree indx_before_incr, indx_after_incr;
- gimple cond_stmt;
- gimple orig_cond;
+ gcond *cond_stmt;
+ gcond *orig_cond;
edge exit_edge = single_exit (loop);
gimple_stmt_iterator loop_cond_gsi;
gimple_stmt_iterator incr_gsi;
bool insert_after;
tree init = build_int_cst (TREE_TYPE (niters), 0);
tree step = build_int_cst (TREE_TYPE (niters), 1);
- LOC loop_loc;
+ source_location loop_loc;
enum tree_code code;
orig_cond = get_loop_exit_condition (loop);
loop_loc = find_loop_location (loop);
if (dump_enabled_p ())
{
- if (LOCATION_LOCUS (loop_loc) != UNKNOWN_LOC)
- dump_printf (MSG_NOTE, "\nloop at %s:%d: ", LOC_FILE (loop_loc),
- LOC_LINE (loop_loc));
+ if (LOCATION_LOCUS (loop_loc) != UNKNOWN_LOCATION)
+ dump_printf (MSG_NOTE, "\nloop at %s:%d: ", LOCATION_FILE (loop_loc),
+ LOCATION_LINE (loop_loc));
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, cond_stmt, 0);
- dump_printf (MSG_NOTE, "\n");
}
loop->nb_iterations = niters;
}
+/* Helper routine of slpeel_tree_duplicate_loop_to_edge_cfg.
+ For all PHI arguments in FROM->dest and TO->dest from those
+ edges ensure that TO->dest PHI arguments have current_def
+ to that in from. */
+
+static void
+slpeel_duplicate_current_defs_from_edges (edge from, edge to)
+{
+ gimple_stmt_iterator gsi_from, gsi_to;
+
+ for (gsi_from = gsi_start_phis (from->dest),
+ gsi_to = gsi_start_phis (to->dest);
+ !gsi_end_p (gsi_from) && !gsi_end_p (gsi_to);)
+ {
+ gimple *from_phi = gsi_stmt (gsi_from);
+ gimple *to_phi = gsi_stmt (gsi_to);
+ tree from_arg = PHI_ARG_DEF_FROM_EDGE (from_phi, from);
+ if (TREE_CODE (from_arg) != SSA_NAME)
+ {
+ gsi_next (&gsi_from);
+ continue;
+ }
+ tree to_arg = PHI_ARG_DEF_FROM_EDGE (to_phi, to);
+ if (TREE_CODE (to_arg) != SSA_NAME)
+ {
+ gsi_next (&gsi_to);
+ continue;
+ }
+ if (get_current_def (to_arg) == NULL_TREE)
+ set_current_def (to_arg, get_current_def (from_arg));
+ gsi_next (&gsi_from);
+ gsi_next (&gsi_to);
+ }
+}
+
/* Given LOOP this function generates a new copy of it and puts it
- on E which is either the entry or exit of LOOP. */
+ on E which is either the entry or exit of LOOP. If SCALAR_LOOP is
+ non-NULL, assume LOOP and SCALAR_LOOP are equivalent and copy the
+ basic blocks from SCALAR_LOOP instead of LOOP, but to either the
+ entry or exit of LOOP. */
struct loop *
-slpeel_tree_duplicate_loop_to_edge_cfg (struct loop *loop, edge e)
+slpeel_tree_duplicate_loop_to_edge_cfg (struct loop *loop,
+ struct loop *scalar_loop, edge e)
{
struct loop *new_loop;
basic_block *new_bbs, *bbs;
bool was_imm_dom;
basic_block exit_dest;
edge exit, new_exit;
+ bool duplicate_outer_loop = false;
exit = single_exit (loop);
at_exit = (e == exit);
if (!at_exit && e != loop_preheader_edge (loop))
return NULL;
- bbs = XNEWVEC (basic_block, loop->num_nodes + 1);
- get_loop_body_with_size (loop, bbs, loop->num_nodes);
+ if (scalar_loop == NULL)
+ scalar_loop = loop;
+ bbs = XNEWVEC (basic_block, scalar_loop->num_nodes + 1);
+ get_loop_body_with_size (scalar_loop, bbs, scalar_loop->num_nodes);
+ /* Allow duplication of outer loops. */
+ if (scalar_loop->inner)
+ duplicate_outer_loop = true;
/* Check whether duplication is possible. */
- if (!can_copy_bbs_p (bbs, loop->num_nodes))
+ if (!can_copy_bbs_p (bbs, scalar_loop->num_nodes))
{
free (bbs);
return NULL;
}
/* Generate new loop structure. */
- new_loop = duplicate_loop (loop, loop_outer (loop));
- duplicate_subloops (loop, new_loop);
+ new_loop = duplicate_loop (scalar_loop, loop_outer (scalar_loop));
+ duplicate_subloops (scalar_loop, new_loop);
exit_dest = exit->dest;
was_imm_dom = (get_immediate_dominator (CDI_DOMINATORS,
/* Also copy the pre-header, this avoids jumping through hoops to
duplicate the loop entry PHI arguments. Create an empty
pre-header unconditionally for this. */
- basic_block preheader = split_edge (loop_preheader_edge (loop));
+ basic_block preheader = split_edge (loop_preheader_edge (scalar_loop));
edge entry_e = single_pred_edge (preheader);
- bbs[loop->num_nodes] = preheader;
- new_bbs = XNEWVEC (basic_block, loop->num_nodes + 1);
+ bbs[scalar_loop->num_nodes] = preheader;
+ new_bbs = XNEWVEC (basic_block, scalar_loop->num_nodes + 1);
- copy_bbs (bbs, loop->num_nodes + 1, new_bbs,
+ exit = single_exit (scalar_loop);
+ copy_bbs (bbs, scalar_loop->num_nodes + 1, new_bbs,
&exit, 1, &new_exit, NULL,
e->src, true);
- basic_block new_preheader = new_bbs[loop->num_nodes];
+ exit = single_exit (loop);
+ basic_block new_preheader = new_bbs[scalar_loop->num_nodes];
- add_phi_args_after_copy (new_bbs, loop->num_nodes + 1, NULL);
+ add_phi_args_after_copy (new_bbs, scalar_loop->num_nodes + 1, NULL);
+
+ if (scalar_loop != loop)
+ {
+ /* If we copied from SCALAR_LOOP rather than LOOP, SSA_NAMEs from
+ SCALAR_LOOP will have current_def set to SSA_NAMEs in the new_loop,
+ but LOOP will not. slpeel_update_phi_nodes_for_guard{1,2} expects
+ the LOOP SSA_NAMEs (on the exit edge and edge from latch to
+ header) to have current_def set, so copy them over. */
+ slpeel_duplicate_current_defs_from_edges (single_exit (scalar_loop),
+ exit);
+ slpeel_duplicate_current_defs_from_edges (EDGE_SUCC (scalar_loop->latch,
+ 0),
+ EDGE_SUCC (loop->latch, 0));
+ }
if (at_exit) /* Add the loop copy at exit. */
{
+ if (scalar_loop != loop)
+ {
+ gphi_iterator gsi;
+ new_exit = redirect_edge_and_branch (new_exit, exit_dest);
+
+ for (gsi = gsi_start_phis (exit_dest); !gsi_end_p (gsi);
+ gsi_next (&gsi))
+ {
+ gphi *phi = gsi.phi ();
+ tree orig_arg = PHI_ARG_DEF_FROM_EDGE (phi, e);
+ location_t orig_locus
+ = gimple_phi_arg_location_from_edge (phi, e);
+
+ add_phi_arg (phi, orig_arg, new_exit, orig_locus);
+ }
+ }
redirect_edge_and_branch_force (e, new_preheader);
flush_pending_stmts (e);
set_immediate_dominator (CDI_DOMINATORS, new_preheader, e->src);
- if (was_imm_dom)
- set_immediate_dominator (CDI_DOMINATORS, exit_dest, new_loop->header);
+ if (was_imm_dom || duplicate_outer_loop)
+ set_immediate_dominator (CDI_DOMINATORS, exit_dest, new_exit->src);
/* And remove the non-necessary forwarder again. Keep the other
one so we have a proper pre-header for the loop at the exit edge. */
- redirect_edge_pred (single_succ_edge (preheader), single_pred (preheader));
+ redirect_edge_pred (single_succ_edge (preheader),
+ single_pred (preheader));
delete_basic_block (preheader);
- set_immediate_dominator (CDI_DOMINATORS, loop->header,
- loop_preheader_edge (loop)->src);
+ set_immediate_dominator (CDI_DOMINATORS, scalar_loop->header,
+ loop_preheader_edge (scalar_loop)->src);
}
else /* Add the copy at entry. */
{
+ if (scalar_loop != loop)
+ {
+ /* Remove the non-necessary forwarder of scalar_loop again. */
+ redirect_edge_pred (single_succ_edge (preheader),
+ single_pred (preheader));
+ delete_basic_block (preheader);
+ set_immediate_dominator (CDI_DOMINATORS, scalar_loop->header,
+ loop_preheader_edge (scalar_loop)->src);
+ preheader = split_edge (loop_preheader_edge (loop));
+ entry_e = single_pred_edge (preheader);
+ }
+
redirect_edge_and_branch_force (entry_e, new_preheader);
flush_pending_stmts (entry_e);
set_immediate_dominator (CDI_DOMINATORS, new_preheader, entry_e->src);
/* And remove the non-necessary forwarder again. Keep the other
one so we have a proper pre-header for the loop at the exit edge. */
- redirect_edge_pred (single_succ_edge (new_preheader), single_pred (new_preheader));
+ redirect_edge_pred (single_succ_edge (new_preheader),
+ single_pred (new_preheader));
delete_basic_block (new_preheader);
set_immediate_dominator (CDI_DOMINATORS, new_loop->header,
loop_preheader_edge (new_loop)->src);
}
- for (unsigned i = 0; i < loop->num_nodes+1; i++)
- rename_variables_in_bb (new_bbs[i]);
+ for (unsigned i = 0; i < scalar_loop->num_nodes + 1; i++)
+ rename_variables_in_bb (new_bbs[i], duplicate_outer_loop);
+
+ if (scalar_loop != loop)
+ {
+ /* Update new_loop->header PHIs, so that on the preheader
+ edge they are the ones from loop rather than scalar_loop. */
+ gphi_iterator gsi_orig, gsi_new;
+ edge orig_e = loop_preheader_edge (loop);
+ edge new_e = loop_preheader_edge (new_loop);
+
+ for (gsi_orig = gsi_start_phis (loop->header),
+ gsi_new = gsi_start_phis (new_loop->header);
+ !gsi_end_p (gsi_orig) && !gsi_end_p (gsi_new);
+ gsi_next (&gsi_orig), gsi_next (&gsi_new))
+ {
+ gphi *orig_phi = gsi_orig.phi ();
+ gphi *new_phi = gsi_new.phi ();
+ tree orig_arg = PHI_ARG_DEF_FROM_EDGE (orig_phi, orig_e);
+ location_t orig_locus
+ = gimple_phi_arg_location_from_edge (orig_phi, orig_e);
+
+ add_phi_arg (new_phi, orig_arg, new_e, orig_locus);
+ }
+ }
free (new_bbs);
free (bbs);
-#ifdef ENABLE_CHECKING
- verify_dominators (CDI_DOMINATORS);
-#endif
+ checking_verify_dominators (CDI_DOMINATORS);
return new_loop;
}
{
gimple_stmt_iterator gsi;
edge new_e, enter_e;
- gimple cond_stmt;
+ gcond *cond_stmt;
gimple_seq gimplify_stmt_list = NULL;
enter_e = EDGE_SUCC (guard_bb, 0);
/* This function verifies that the following restrictions apply to LOOP:
- (1) it is innermost
- (2) it consists of exactly 2 basic blocks - header, and an empty latch.
- (3) it is single entry, single exit
- (4) its exit condition is the last stmt in the header
- (5) E is the entry/exit edge of LOOP.
+ (1) it consists of exactly 2 basic blocks - header, and an empty latch
+ for innermost loop and 5 basic blocks for outer-loop.
+ (2) it is single entry, single exit
+ (3) its exit condition is the last stmt in the header
+ (4) E is the entry/exit edge of LOOP.
*/
bool
{
edge exit_e = single_exit (loop);
edge entry_e = loop_preheader_edge (loop);
- gimple orig_cond = get_loop_exit_condition (loop);
+ gcond *orig_cond = get_loop_exit_condition (loop);
gimple_stmt_iterator loop_exit_gsi = gsi_last_bb (exit_e->src);
+ unsigned int num_bb = loop->inner? 5 : 2;
- if (loop->inner
/* All loops have an outer scope; the only case loop->outer is NULL is for
the function itself. */
- || !loop_outer (loop)
- || loop->num_nodes != 2
+ if (!loop_outer (loop)
+ || loop->num_nodes != num_bb
|| !empty_block_p (loop->latch)
|| !single_exit (loop)
/* Verify that new loop exit condition can be trivially modified. */
return true;
}
-#ifdef ENABLE_CHECKING
static void
-slpeel_verify_cfg_after_peeling (struct loop *first_loop,
- struct loop *second_loop)
+slpeel_checking_verify_cfg_after_peeling (struct loop *first_loop,
+ struct loop *second_loop)
{
+ if (!flag_checking)
+ return;
+
basic_block loop1_exit_bb = single_exit (first_loop)->dest;
basic_block loop2_entry_bb = loop_preheader_edge (second_loop)->src;
basic_block loop1_entry_bb = loop_preheader_edge (first_loop)->src;
second_loop. */
/* TODO */
}
-#endif
/* If the run time cost model check determines that vectorization is
not profitable and hence scalar loop should be generated then set
basic_block cond_bb, then_bb;
tree var, prologue_after_cost_adjust_name;
gimple_stmt_iterator gsi;
- gimple newphi;
+ gphi *newphi;
edge e_true, e_false, e_fallthru;
- gimple cond_stmt;
+ gcond *cond_stmt;
gimple_seq stmts = NULL;
tree cost_pre_condition = NULL_TREE;
tree scalar_loop_iters =
Input:
- LOOP: the loop to be peeled.
+ - SCALAR_LOOP: if non-NULL, the alternate loop from which basic blocks
+ should be copied.
- E: the exit or entry edge of LOOP.
If it is the entry edge, we peel the first iterations of LOOP. In this
case first-loop is LOOP, and second-loop is the newly created loop.
FORNOW the resulting code will not be in loop-closed-ssa form.
*/
-static struct loop*
-slpeel_tree_peel_loop_to_edge (struct loop *loop,
+static struct loop *
+slpeel_tree_peel_loop_to_edge (struct loop *loop, struct loop *scalar_loop,
edge e, tree *first_niters,
tree niters, bool update_first_loop_count,
unsigned int th, bool check_profitability,
basic_block bb_before_first_loop;
basic_block bb_between_loops;
basic_block new_exit_bb;
- gimple_stmt_iterator gsi;
+ gphi_iterator gsi;
edge exit_e = single_exit (loop);
- LOC loop_loc;
- tree cost_pre_condition = NULL_TREE;
+ source_location loop_loc;
/* There are many aspects to how likely the first loop is going to be executed.
Without histogram we can't really do good job. Simply set it to
2/3, so the first loop is not reordered to the end of function and
for (gsi = gsi_start_phis (loop->header); !gsi_end_p (gsi); gsi_next (&gsi))
if (virtual_operand_p (gimple_phi_result (gsi_stmt (gsi))))
{
- gimple phi = gsi_stmt (gsi);
+ gphi *phi = gsi.phi ();
for (gsi = gsi_start_phis (exit_e->dest);
!gsi_end_p (gsi); gsi_next (&gsi))
if (virtual_operand_p (gimple_phi_result (gsi_stmt (gsi))))
break;
if (gsi_end_p (gsi))
{
- tree new_vop = copy_ssa_name (PHI_RESULT (phi), NULL);
- gimple new_phi = create_phi_node (new_vop, exit_e->dest);
+ tree new_vop = copy_ssa_name (PHI_RESULT (phi));
+ gphi *new_phi = create_phi_node (new_vop, exit_e->dest);
tree vop = PHI_ARG_DEF_FROM_EDGE (phi, EDGE_SUCC (loop->latch, 0));
imm_use_iterator imm_iter;
- gimple stmt;
+ gimple *stmt;
use_operand_p use_p;
add_phi_arg (new_phi, vop, exit_e, UNKNOWN_LOCATION);
gimple_phi_set_result (new_phi, new_vop);
FOR_EACH_IMM_USE_STMT (stmt, imm_iter, vop)
- if (stmt != new_phi && gimple_bb (stmt) != loop->header)
+ if (stmt != new_phi
+ && !flow_bb_inside_loop_p (loop, gimple_bb (stmt)))
FOR_EACH_IMM_USE_ON_STMT (use_p, imm_iter)
SET_USE (use_p, new_vop);
}
orig_exit_bb:
*/
- if (!(new_loop = slpeel_tree_duplicate_loop_to_edge_cfg (loop, e)))
+ if (!(new_loop = slpeel_tree_duplicate_loop_to_edge_cfg (loop, scalar_loop,
+ e)))
{
loop_loc = find_loop_location (loop);
dump_printf_loc (MSG_MISSED_OPTIMIZATION, loop_loc,
/* Epilogue peeling. */
if (!update_first_loop_count)
{
+ loop_vec_info loop_vinfo = loop_vec_info_for_loop (loop);
+ tree scalar_loop_iters = LOOP_VINFO_NITERSM1 (loop_vinfo);
+ unsigned limit = LOOP_VINFO_VECT_FACTOR (loop_vinfo) - 1;
+ if (LOOP_VINFO_PEELING_FOR_GAPS (loop_vinfo))
+ limit = limit + 1;
+ if (check_profitability
+ && th > limit)
+ limit = th;
pre_condition =
- fold_build2 (LE_EXPR, boolean_type_node, *first_niters,
- build_int_cst (TREE_TYPE (*first_niters), 0));
- if (check_profitability)
- {
- tree scalar_loop_iters
- = unshare_expr (LOOP_VINFO_NITERS_UNCHANGED
- (loop_vec_info_for_loop (loop)));
- cost_pre_condition =
- fold_build2 (LE_EXPR, boolean_type_node, scalar_loop_iters,
- build_int_cst (TREE_TYPE (scalar_loop_iters), th));
-
- pre_condition = fold_build2 (TRUTH_OR_EXPR, boolean_type_node,
- cost_pre_condition, pre_condition);
- }
+ fold_build2 (LT_EXPR, boolean_type_node, scalar_loop_iters,
+ build_int_cst (TREE_TYPE (scalar_loop_iters), limit));
if (cond_expr)
{
pre_condition =
location is calculated.
Return the loop location if succeed and NULL if not. */
-LOC
+source_location
find_loop_location (struct loop *loop)
{
- gimple stmt = NULL;
+ gimple *stmt = NULL;
basic_block bb;
gimple_stmt_iterator si;
if (!loop)
- return UNKNOWN_LOC;
+ return UNKNOWN_LOCATION;
stmt = get_loop_exit_condition (loop);
try to estimate the loop location */
if (!loop->header)
- return UNKNOWN_LOC;
+ return UNKNOWN_LOCATION;
bb = loop->header;
return gimple_location (stmt);
}
- return UNKNOWN_LOC;
-}
-
-
-/* This function builds ni_name = number of iterations loop executes
- on the loop preheader. If SEQ is given the stmt is instead emitted
- there. */
-
-static tree
-vect_build_loop_niters (loop_vec_info loop_vinfo, gimple_seq seq)
-{
- tree ni_name, var;
- gimple_seq stmts = NULL;
- edge pe;
- struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
- tree ni = unshare_expr (LOOP_VINFO_NITERS (loop_vinfo));
-
- var = create_tmp_var (TREE_TYPE (ni), "niters");
- ni_name = force_gimple_operand (ni, &stmts, false, var);
-
- pe = loop_preheader_edge (loop);
- if (stmts)
- {
- if (seq)
- gimple_seq_add_seq (&seq, stmts);
- else
- {
- basic_block new_bb = gsi_insert_seq_on_edge_immediate (pe, stmts);
- gcc_assert (!new_bb);
- }
- }
-
- return ni_name;
+ return UNKNOWN_LOCATION;
}
-/* This function generates the following statements:
-
- ni_name = number of iterations loop executes
- ratio = ni_name / vf
- ratio_mult_vf_name = ratio * vf
-
- and places them at the loop preheader edge or in COND_EXPR_STMT_LIST
- if that is non-NULL. */
-
-void
-vect_generate_tmps_on_preheader (loop_vec_info loop_vinfo,
- tree *ni_name_ptr,
- tree *ratio_mult_vf_name_ptr,
- tree *ratio_name_ptr,
- gimple_seq cond_expr_stmt_list)
-{
-
- edge pe;
- basic_block new_bb;
- gimple_seq stmts;
- tree ni_name, ni_minus_gap_name;
- tree var;
- tree ratio_name;
- tree ratio_mult_vf_name;
- struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
- tree ni = LOOP_VINFO_NITERS (loop_vinfo);
- int vf = LOOP_VINFO_VECT_FACTOR (loop_vinfo);
- tree log_vf;
-
- pe = loop_preheader_edge (loop);
-
- /* Generate temporary variable that contains
- number of iterations loop executes. */
-
- ni_name = vect_build_loop_niters (loop_vinfo, cond_expr_stmt_list);
- log_vf = build_int_cst (TREE_TYPE (ni), exact_log2 (vf));
-
- /* If epilogue loop is required because of data accesses with gaps, we
- subtract one iteration from the total number of iterations here for
- correct calculation of RATIO. */
- if (LOOP_VINFO_PEELING_FOR_GAPS (loop_vinfo))
- {
- ni_minus_gap_name = fold_build2 (MINUS_EXPR, TREE_TYPE (ni_name),
- ni_name,
- build_one_cst (TREE_TYPE (ni_name)));
- if (!is_gimple_val (ni_minus_gap_name))
- {
- var = create_tmp_var (TREE_TYPE (ni), "ni_gap");
-
- stmts = NULL;
- ni_minus_gap_name = force_gimple_operand (ni_minus_gap_name, &stmts,
- true, var);
- if (cond_expr_stmt_list)
- gimple_seq_add_seq (&cond_expr_stmt_list, stmts);
- else
- {
- pe = loop_preheader_edge (loop);
- new_bb = gsi_insert_seq_on_edge_immediate (pe, stmts);
- gcc_assert (!new_bb);
- }
- }
- }
- else
- ni_minus_gap_name = ni_name;
-
- /* Create: ratio = ni >> log2(vf) */
-
- ratio_name = fold_build2 (RSHIFT_EXPR, TREE_TYPE (ni_minus_gap_name),
- ni_minus_gap_name, log_vf);
- if (!is_gimple_val (ratio_name))
- {
- var = create_tmp_var (TREE_TYPE (ni), "bnd");
-
- stmts = NULL;
- ratio_name = force_gimple_operand (ratio_name, &stmts, true, var);
- if (cond_expr_stmt_list)
- gimple_seq_add_seq (&cond_expr_stmt_list, stmts);
- else
- {
- pe = loop_preheader_edge (loop);
- new_bb = gsi_insert_seq_on_edge_immediate (pe, stmts);
- gcc_assert (!new_bb);
- }
- }
-
- /* Create: ratio_mult_vf = ratio << log2 (vf). */
-
- ratio_mult_vf_name = fold_build2 (LSHIFT_EXPR, TREE_TYPE (ratio_name),
- ratio_name, log_vf);
- if (!is_gimple_val (ratio_mult_vf_name))
- {
- var = create_tmp_var (TREE_TYPE (ni), "ratio_mult_vf");
-
- stmts = NULL;
- ratio_mult_vf_name = force_gimple_operand (ratio_mult_vf_name, &stmts,
- true, var);
- if (cond_expr_stmt_list)
- gimple_seq_add_seq (&cond_expr_stmt_list, stmts);
- else
- {
- pe = loop_preheader_edge (loop);
- new_bb = gsi_insert_seq_on_edge_immediate (pe, stmts);
- gcc_assert (!new_bb);
- }
- }
-
- *ni_name_ptr = ni_name;
- *ratio_mult_vf_name_ptr = ratio_mult_vf_name;
- *ratio_name_ptr = ratio_name;
-
- return;
-}
-
/* Function vect_can_advance_ivs_p
In case the number of iterations that LOOP iterates is unknown at compile
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block bb = loop->header;
- gimple phi;
- gimple_stmt_iterator gsi;
+ gimple *phi;
+ gphi_iterator gsi;
/* Analyze phi functions of the loop header. */
{
tree evolution_part;
- phi = gsi_stmt (gsi);
+ phi = gsi.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
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
basic_block exit_bb = single_exit (loop)->dest;
- gimple phi, phi1;
- gimple_stmt_iterator gsi, gsi1;
+ gphi *phi, *phi1;
+ gphi_iterator gsi, gsi1;
basic_block update_bb = update_e->dest;
- /* gcc_assert (vect_can_advance_ivs_p (loop_vinfo)); */
+ gcc_checking_assert (vect_can_advance_ivs_p (loop_vinfo));
/* Make sure there exists a single-predecessor exit bb: */
gcc_assert (single_pred_p (exit_bb));
gimple_stmt_iterator last_gsi;
stmt_vec_info stmt_info;
- phi = gsi_stmt (gsi);
- phi1 = gsi_stmt (gsi1);
+ phi = gsi.phi ();
+ phi1 = gsi1.phi ();
if (dump_enabled_p ())
{
dump_printf_loc (MSG_NOTE, vect_location,
"vect_update_ivs_after_vectorizer: phi: ");
dump_gimple_stmt (MSG_NOTE, TDF_SLIM, phi, 0);
- dump_printf (MSG_NOTE, "\n");
}
/* Skip virtual phi's. */
/* Skip reduction phis. */
stmt_info = vinfo_for_stmt (phi);
- if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_reduction_def)
+ if (STMT_VINFO_DEF_TYPE (stmt_info) == vect_reduction_def
+ || STMT_VINFO_DEF_TYPE (stmt_info) == vect_double_reduction_def)
{
if (dump_enabled_p ())
dump_printf_loc (MSG_MISSED_OPTIMIZATION, vect_location,
test. */
void
-vect_do_peeling_for_loop_bound (loop_vec_info loop_vinfo, tree *ratio,
+vect_do_peeling_for_loop_bound (loop_vec_info loop_vinfo,
+ tree ni_name, tree ratio_mult_vf_name,
unsigned int th, bool check_profitability)
{
- tree ni_name, ratio_mult_vf_name;
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
+ struct loop *scalar_loop = LOOP_VINFO_SCALAR_LOOP (loop_vinfo);
struct loop *new_loop;
edge update_e;
basic_block preheader;
initialize_original_copy_tables ();
- /* Generate the following variables on the preheader of original loop:
-
- ni_name = number of iteration the original loop executes
- ratio = ni_name / vf
- ratio_mult_vf_name = ratio * vf */
- vect_generate_tmps_on_preheader (loop_vinfo, &ni_name,
- &ratio_mult_vf_name, ratio,
- cond_expr_stmt_list);
-
loop_num = loop->num;
- new_loop = slpeel_tree_peel_loop_to_edge (loop, single_exit (loop),
- &ratio_mult_vf_name, ni_name, false,
- th, check_profitability,
- cond_expr, cond_expr_stmt_list,
- 0, LOOP_VINFO_VECT_FACTOR (loop_vinfo));
+ new_loop
+ = slpeel_tree_peel_loop_to_edge (loop, scalar_loop, single_exit (loop),
+ &ratio_mult_vf_name, ni_name, false,
+ th, check_profitability,
+ cond_expr, cond_expr_stmt_list,
+ 0, LOOP_VINFO_VECT_FACTOR (loop_vinfo));
gcc_assert (new_loop);
gcc_assert (loop_num == loop->num);
-#ifdef ENABLE_CHECKING
- slpeel_verify_cfg_after_peeling (loop, new_loop);
-#endif
+ slpeel_checking_verify_cfg_after_peeling (loop, new_loop);
/* A guard that controls whether the new_loop is to be executed or skipped
is placed in LOOP->exit. LOOP->exit therefore has two successors - one
: LOOP_VINFO_VECT_FACTOR (loop_vinfo)) - 2;
if (check_profitability)
max_iter = MAX (max_iter, (int) th - 1);
- record_niter_bound (new_loop, double_int::from_shwi (max_iter), false, true);
+ record_niter_bound (new_loop, max_iter, false, true);
dump_printf (MSG_NOTE,
"Setting upper bound of nb iterations for epilogue "
"loop to %d\n", max_iter);
tree iters, iters_name;
edge pe;
basic_block new_bb;
- gimple dr_stmt = DR_STMT (dr);
+ gimple *dr_stmt = DR_STMT (dr);
stmt_vec_info stmt_info = vinfo_for_stmt (dr_stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_info);
int vectype_align = TYPE_ALIGN (vectype) / BITS_PER_UNIT;
pe = loop_preheader_edge (loop);
- if (LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo) > 0)
+ if (LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo) > 0)
{
- int npeel = LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo);
+ int npeel = LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo);
if (dump_enabled_p ())
dump_printf_loc (MSG_NOTE, vect_location,
"known peeling = %d.\n", npeel);
iters = build_int_cst (niters_type, npeel);
- *bound = LOOP_PEELING_FOR_ALIGNMENT (loop_vinfo);
+ *bound = LOOP_VINFO_PEELING_FOR_ALIGNMENT (loop_vinfo);
}
else
{
gimple_seq new_stmts = NULL;
bool negative = tree_int_cst_compare (DR_STEP (dr), size_zero_node) < 0;
tree offset = negative
- ? size_int (-TYPE_VECTOR_SUBPARTS (vectype) + 1) : NULL_TREE;
+ ? size_int (-TYPE_VECTOR_SUBPARTS (vectype) + 1) : size_zero_node;
tree start_addr = vect_create_addr_base_for_vector_ref (dr_stmt,
&new_stmts, offset, loop);
tree type = unsigned_type_for (TREE_TYPE (start_addr));
peeling is recorded in LOOP_VINFO_UNALIGNED_DR. */
void
-vect_do_peeling_for_alignment (loop_vec_info loop_vinfo,
+vect_do_peeling_for_alignment (loop_vec_info loop_vinfo, tree ni_name,
unsigned int th, bool check_profitability)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
- tree niters_of_prolog_loop, ni_name;
- tree n_iters;
+ struct loop *scalar_loop = LOOP_VINFO_SCALAR_LOOP (loop_vinfo);
+ tree niters_of_prolog_loop;
tree wide_prolog_niters;
struct loop *new_loop;
int max_iter;
initialize_original_copy_tables ();
- ni_name = vect_build_loop_niters (loop_vinfo, NULL);
+ gimple_seq stmts = NULL;
+ gsi_insert_seq_on_edge_immediate (loop_preheader_edge (loop), stmts);
niters_of_prolog_loop = vect_gen_niters_for_prolog_loop (loop_vinfo,
ni_name,
&bound);
/* Peel the prolog loop and iterate it niters_of_prolog_loop. */
new_loop =
- slpeel_tree_peel_loop_to_edge (loop, loop_preheader_edge (loop),
+ slpeel_tree_peel_loop_to_edge (loop, scalar_loop,
+ loop_preheader_edge (loop),
&niters_of_prolog_loop, ni_name, true,
th, check_profitability, NULL_TREE, NULL,
- bound,
- 0);
+ bound, 0);
gcc_assert (new_loop);
-#ifdef ENABLE_CHECKING
- slpeel_verify_cfg_after_peeling (new_loop, loop);
-#endif
+ slpeel_checking_verify_cfg_after_peeling (new_loop, loop);
/* For vectorization factor N, we need to copy at most N-1 values
for alignment and this means N-2 loopback edge executions. */
max_iter = LOOP_VINFO_VECT_FACTOR (loop_vinfo) - 2;
if (check_profitability)
max_iter = MAX (max_iter, (int) th - 1);
- record_niter_bound (new_loop, double_int::from_shwi (max_iter), false, true);
+ record_niter_bound (new_loop, max_iter, false, true);
dump_printf (MSG_NOTE,
"Setting upper bound of nb iterations for prologue "
"loop to %d\n", max_iter);
/* Update number of times loop executes. */
- n_iters = LOOP_VINFO_NITERS (loop_vinfo);
LOOP_VINFO_NITERS (loop_vinfo) = fold_build2 (MINUS_EXPR,
- TREE_TYPE (n_iters), n_iters, niters_of_prolog_loop);
+ TREE_TYPE (ni_name), ni_name, niters_of_prolog_loop);
+ LOOP_VINFO_NITERSM1 (loop_vinfo) = fold_build2 (MINUS_EXPR,
+ TREE_TYPE (ni_name),
+ LOOP_VINFO_NITERSM1 (loop_vinfo), niters_of_prolog_loop);
if (types_compatible_p (sizetype, TREE_TYPE (niters_of_prolog_loop)))
wide_prolog_niters = niters_of_prolog_loop;
gimple_seq *cond_expr_stmt_list)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
- vec<gimple> may_misalign_stmts
+ vec<gimple *> may_misalign_stmts
= LOOP_VINFO_MAY_MISALIGN_STMTS (loop_vinfo);
- gimple ref_stmt;
+ gimple *ref_stmt;
int mask = LOOP_VINFO_PTR_MASK (loop_vinfo);
tree mask_cst;
unsigned int i;
char tmp_name[20];
tree or_tmp_name = NULL_TREE;
tree and_tmp_name;
- gimple and_stmt;
+ gimple *and_stmt;
tree ptrsize_zero;
tree part_cond_expr;
tree addr_base;
tree addr_tmp_name;
tree new_or_tmp_name;
- gimple addr_stmt, or_stmt;
+ gimple *addr_stmt, *or_stmt;
stmt_vec_info stmt_vinfo = vinfo_for_stmt (ref_stmt);
tree vectype = STMT_VINFO_VECTYPE (stmt_vinfo);
bool negative = tree_int_cst_compare
(DR_STEP (STMT_VINFO_DATA_REF (stmt_vinfo)), size_zero_node) < 0;
tree offset = negative
- ? size_int (-TYPE_VECTOR_SUBPARTS (vectype) + 1) : NULL_TREE;
+ ? size_int (-TYPE_VECTOR_SUBPARTS (vectype) + 1) : size_zero_node;
/* create: addr_tmp = (int)(address_of_first_vector) */
addr_base =
sprintf (tmp_name, "addr2int%d", i);
addr_tmp_name = make_temp_ssa_name (int_ptrsize_type, NULL, tmp_name);
- addr_stmt = gimple_build_assign_with_ops (NOP_EXPR, addr_tmp_name,
- addr_base, NULL_TREE);
+ addr_stmt = gimple_build_assign (addr_tmp_name, NOP_EXPR, addr_base);
gimple_seq_add_stmt (cond_expr_stmt_list, addr_stmt);
/* The addresses are OR together. */
/* create: or_tmp = or_tmp | addr_tmp */
sprintf (tmp_name, "orptrs%d", i);
new_or_tmp_name = make_temp_ssa_name (int_ptrsize_type, NULL, tmp_name);
- or_stmt = gimple_build_assign_with_ops (BIT_IOR_EXPR,
- new_or_tmp_name,
- or_tmp_name, addr_tmp_name);
+ or_stmt = gimple_build_assign (new_or_tmp_name, BIT_IOR_EXPR,
+ or_tmp_name, addr_tmp_name);
gimple_seq_add_stmt (cond_expr_stmt_list, or_stmt);
or_tmp_name = new_or_tmp_name;
}
/* create: and_tmp = or_tmp & mask */
and_tmp_name = make_temp_ssa_name (int_ptrsize_type, NULL, "andmask");
- and_stmt = gimple_build_assign_with_ops (BIT_AND_EXPR, and_tmp_name,
- or_tmp_name, mask_cst);
+ and_stmt = gimple_build_assign (and_tmp_name, BIT_AND_EXPR,
+ or_tmp_name, mask_cst);
gimple_seq_add_stmt (cond_expr_stmt_list, and_stmt);
/* Make and_tmp the left operand of the conditional test against zero.
void
vect_create_cond_for_alias_checks (loop_vec_info loop_vinfo, tree * cond_expr)
{
- vec<dr_addr_with_seg_len_pair_t> comp_alias_ddrs =
+ vec<dr_with_seg_len_pair_t> comp_alias_ddrs =
LOOP_VINFO_COMP_ALIAS_DDRS (loop_vinfo);
tree part_cond_expr;
for (size_t i = 0, s = comp_alias_ddrs.length (); i < s; ++i)
{
- const dr_addr_with_seg_len& dr_a = comp_alias_ddrs[i].first;
- const dr_addr_with_seg_len& dr_b = comp_alias_ddrs[i].second;
+ const dr_with_seg_len& dr_a = comp_alias_ddrs[i].first;
+ const dr_with_seg_len& dr_b = comp_alias_ddrs[i].second;
tree segment_length_a = dr_a.seg_len;
tree segment_length_b = dr_b.seg_len;
+ tree addr_base_a = DR_BASE_ADDRESS (dr_a.dr);
+ tree addr_base_b = DR_BASE_ADDRESS (dr_b.dr);
+ tree offset_a = DR_OFFSET (dr_a.dr), offset_b = DR_OFFSET (dr_b.dr);
- tree addr_base_a
- = fold_build_pointer_plus (dr_a.basic_addr, dr_a.offset);
- tree addr_base_b
- = fold_build_pointer_plus (dr_b.basic_addr, dr_b.offset);
+ offset_a = fold_build2 (PLUS_EXPR, TREE_TYPE (offset_a),
+ offset_a, DR_INIT (dr_a.dr));
+ offset_b = fold_build2 (PLUS_EXPR, TREE_TYPE (offset_b),
+ offset_b, DR_INIT (dr_b.dr));
+ addr_base_a = fold_build_pointer_plus (addr_base_a, offset_a);
+ addr_base_b = fold_build_pointer_plus (addr_base_b, offset_b);
if (dump_enabled_p ())
{
tree seg_a_min = addr_base_a;
tree seg_a_max = fold_build_pointer_plus (addr_base_a, segment_length_a);
+ /* For negative step, we need to adjust address range by TYPE_SIZE_UNIT
+ bytes, e.g., int a[3] -> a[1] range is [a+4, a+16) instead of
+ [a, a+12) */
if (tree_int_cst_compare (DR_STEP (dr_a.dr), size_zero_node) < 0)
- seg_a_min = seg_a_max, seg_a_max = addr_base_a;
+ {
+ tree unit_size = TYPE_SIZE_UNIT (TREE_TYPE (DR_REF (dr_a.dr)));
+ seg_a_min = fold_build_pointer_plus (seg_a_max, unit_size);
+ seg_a_max = fold_build_pointer_plus (addr_base_a, unit_size);
+ }
tree seg_b_min = addr_base_b;
tree seg_b_max = fold_build_pointer_plus (addr_base_b, segment_length_b);
if (tree_int_cst_compare (DR_STEP (dr_b.dr), size_zero_node) < 0)
- seg_b_min = seg_b_max, seg_b_max = addr_base_b;
+ {
+ tree unit_size = TYPE_SIZE_UNIT (TREE_TYPE (DR_REF (dr_b.dr)));
+ seg_b_min = fold_build_pointer_plus (seg_b_max, unit_size);
+ seg_b_max = fold_build_pointer_plus (addr_base_b, unit_size);
+ }
part_cond_expr =
fold_build2 (TRUTH_OR_EXPR, boolean_type_node,
dump_printf_loc (MSG_NOTE, vect_location,
"created %u versioning for alias checks.\n",
comp_alias_ddrs.length ());
-
- comp_alias_ddrs.release ();
}
unsigned int th, bool check_profitability)
{
struct loop *loop = LOOP_VINFO_LOOP (loop_vinfo);
+ struct loop *scalar_loop = LOOP_VINFO_SCALAR_LOOP (loop_vinfo);
basic_block condition_bb;
- gimple_stmt_iterator gsi, cond_exp_gsi;
+ gphi_iterator gsi;
+ gimple_stmt_iterator cond_exp_gsi;
basic_block merge_bb;
basic_block new_exit_bb;
edge new_exit_e, e;
- gimple orig_phi, new_phi;
+ gphi *orig_phi, *new_phi;
tree cond_expr = NULL_TREE;
gimple_seq cond_expr_stmt_list = NULL;
tree arg;
gimple_seq_add_seq (&cond_expr_stmt_list, gimplify_stmt_list);
initialize_original_copy_tables ();
- loop_version (loop, cond_expr, &condition_bb,
- prob, prob, REG_BR_PROB_BASE - prob, true);
+ if (scalar_loop)
+ {
+ edge scalar_e;
+ basic_block preheader, scalar_preheader;
+
+ /* We don't want to scale SCALAR_LOOP's frequencies, we need to
+ scale LOOP's frequencies instead. */
+ loop_version (scalar_loop, cond_expr, &condition_bb,
+ prob, REG_BR_PROB_BASE, REG_BR_PROB_BASE - prob, true);
+ scale_loop_frequencies (loop, prob, REG_BR_PROB_BASE);
+ /* CONDITION_BB was created above SCALAR_LOOP's preheader,
+ while we need to move it above LOOP's preheader. */
+ e = loop_preheader_edge (loop);
+ scalar_e = loop_preheader_edge (scalar_loop);
+ gcc_assert (empty_block_p (e->src)
+ && single_pred_p (e->src));
+ gcc_assert (empty_block_p (scalar_e->src)
+ && single_pred_p (scalar_e->src));
+ gcc_assert (single_pred_p (condition_bb));
+ preheader = e->src;
+ scalar_preheader = scalar_e->src;
+ scalar_e = find_edge (condition_bb, scalar_preheader);
+ e = single_pred_edge (preheader);
+ redirect_edge_and_branch_force (single_pred_edge (condition_bb),
+ scalar_preheader);
+ redirect_edge_and_branch_force (scalar_e, preheader);
+ redirect_edge_and_branch_force (e, condition_bb);
+ set_immediate_dominator (CDI_DOMINATORS, condition_bb,
+ single_pred (condition_bb));
+ set_immediate_dominator (CDI_DOMINATORS, scalar_preheader,
+ single_pred (scalar_preheader));
+ set_immediate_dominator (CDI_DOMINATORS, preheader,
+ condition_bb);
+ }
+ else
+ loop_version (loop, cond_expr, &condition_bb,
+ prob, prob, REG_BR_PROB_BASE - prob, true);
- if (LOCATION_LOCUS (vect_location) != UNKNOWN_LOC
+ if (LOCATION_LOCUS (vect_location) != UNKNOWN_LOCATION
&& dump_enabled_p ())
{
if (version_alias)
basic block (i.e. it has two predecessors). Just in order to simplify
following transformations in the vectorizer, we fix this situation
here by adding a new (empty) block on the exit-edge of the loop,
- with the proper loop-exit phis to maintain loop-closed-form. */
+ with the proper loop-exit phis to maintain loop-closed-form.
+ If loop versioning wasn't done from loop, but scalar_loop instead,
+ merge_bb will have already just a single successor. */
merge_bb = single_exit (loop)->dest;
- gcc_assert (EDGE_COUNT (merge_bb->preds) == 2);
- new_exit_bb = split_edge (single_exit (loop));
- new_exit_e = single_exit (loop);
- e = EDGE_SUCC (new_exit_bb, 0);
-
- for (gsi = gsi_start_phis (merge_bb); !gsi_end_p (gsi); gsi_next (&gsi))
- {
- tree new_res;
- orig_phi = gsi_stmt (gsi);
- new_res = copy_ssa_name (PHI_RESULT (orig_phi), NULL);
- new_phi = create_phi_node (new_res, new_exit_bb);
- arg = PHI_ARG_DEF_FROM_EDGE (orig_phi, e);
- add_phi_arg (new_phi, arg, new_exit_e,
- gimple_phi_arg_location_from_edge (orig_phi, e));
- adjust_phi_and_debug_stmts (orig_phi, e, PHI_RESULT (new_phi));
- }
-
-
- /* Extract load statements on memrefs with zero-stride accesses. */
-
- if (LOOP_REQUIRES_VERSIONING_FOR_ALIAS (loop_vinfo))
+ if (scalar_loop == NULL || EDGE_COUNT (merge_bb->preds) >= 2)
{
- /* In the loop body, we iterate each statement to check if it is a load.
- Then we check the DR_STEP of the data reference. If DR_STEP is zero,
- then we will hoist the load statement to the loop preheader. */
-
- basic_block *bbs = LOOP_VINFO_BBS (loop_vinfo);
- int nbbs = loop->num_nodes;
+ gcc_assert (EDGE_COUNT (merge_bb->preds) >= 2);
+ new_exit_bb = split_edge (single_exit (loop));
+ new_exit_e = single_exit (loop);
+ e = EDGE_SUCC (new_exit_bb, 0);
- for (int i = 0; i < nbbs; ++i)
+ for (gsi = gsi_start_phis (merge_bb); !gsi_end_p (gsi); gsi_next (&gsi))
{
- for (gimple_stmt_iterator si = gsi_start_bb (bbs[i]);
- !gsi_end_p (si);)
- {
- gimple stmt = gsi_stmt (si);
- stmt_vec_info stmt_info = vinfo_for_stmt (stmt);
- struct data_reference *dr = STMT_VINFO_DATA_REF (stmt_info);
-
- if (is_gimple_assign (stmt)
- && (!dr
- || (DR_IS_READ (dr) && integer_zerop (DR_STEP (dr)))))
- {
- bool hoist = true;
- ssa_op_iter iter;
- tree var;
-
- /* We hoist a statement if all SSA uses in it are defined
- outside of the loop. */
- FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_USE)
- {
- gimple def = SSA_NAME_DEF_STMT (var);
- if (!gimple_nop_p (def)
- && flow_bb_inside_loop_p (loop, gimple_bb (def)))
- {
- hoist = false;
- break;
- }
- }
-
- if (hoist)
- {
- if (dr)
- gimple_set_vuse (stmt, NULL);
-
- gsi_remove (&si, false);
- gsi_insert_on_edge_immediate (loop_preheader_edge (loop),
- stmt);
-
- if (dump_enabled_p ())
- {
- dump_printf_loc
- (MSG_NOTE, vect_location,
- "hoisting out of the vectorized loop: ");
- dump_gimple_stmt (MSG_NOTE, TDF_SLIM, stmt, 0);
- dump_printf (MSG_NOTE, "\n");
- }
- continue;
- }
- }
- gsi_next (&si);
- }
+ tree new_res;
+ orig_phi = gsi.phi ();
+ new_res = copy_ssa_name (PHI_RESULT (orig_phi));
+ new_phi = create_phi_node (new_res, new_exit_bb);
+ arg = PHI_ARG_DEF_FROM_EDGE (orig_phi, e);
+ add_phi_arg (new_phi, arg, new_exit_e,
+ gimple_phi_arg_location_from_edge (orig_phi, e));
+ adjust_phi_and_debug_stmts (orig_phi, e, PHI_RESULT (new_phi));
}
}
projects / gcc.git / blobdiff