X-Git-Url: http://git.sourceforge.jp/view?a=blobdiff_plain;f=gcc%2Ftree-ssa-loop-manip.c;h=f6e2e9c937f358e412a63a571e3f6a5fdb3b2240;hb=9e229ec708d7e1c812330cd2eb941b6c4f50bb9c;hp=6f787575d729f1607856d60778d8d005837d936e;hpb=44a461035a32653210a2df4ea4f1cc0630b4e537;p=pf3gnuchains%2Fgcc-fork.git diff --git a/gcc/tree-ssa-loop-manip.c b/gcc/tree-ssa-loop-manip.c index 6f787575d72..f6e2e9c937f 100644 --- a/gcc/tree-ssa-loop-manip.c +++ b/gcc/tree-ssa-loop-manip.c @@ -1,34 +1,31 @@ /* High-level loop manipulation functions. - Copyright (C) 2004 Free Software Foundation, Inc. - + Copyright (C) 2004, 2005, 2006, 2007, 2008, 2010 + Free Software Foundation, Inc. + 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 2, or (at your option) any +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 COPYING. If not, write to the Free -Software Foundation, 59 Temple Place - Suite 330, Boston, MA -02111-1307, USA. */ +along with GCC; see the file COPYING3. If not see +. */ #include "config.h" #include "system.h" #include "coretypes.h" #include "tm.h" #include "tree.h" -#include "rtl.h" #include "tm_p.h" -#include "hard-reg-set.h" #include "basic-block.h" #include "output.h" -#include "diagnostic.h" #include "tree-flow.h" #include "tree-dump.h" #include "timevar.h" @@ -36,34 +33,41 @@ Software Foundation, 59 Temple Place - Suite 330, Boston, MA #include "tree-pass.h" #include "cfglayout.h" #include "tree-scalar-evolution.h" +#include "params.h" +#include "tree-inline.h" +#include "langhooks.h" /* Creates an induction variable with value BASE + STEP * iteration in LOOP. It is expected that neither BASE nor STEP are shared with other expressions (unless the sharing rules allow this). Use VAR as a base var_decl for it (if NULL, a new temporary will be created). The increment will occur at - INCR_POS (after it if AFTER is true, before it otherwise). The ssa versions + INCR_POS (after it if AFTER is true, before it otherwise). INCR_POS and + AFTER can be computed using standard_iv_increment_position. The ssa versions of the variable before and after increment will be stored in VAR_BEFORE and VAR_AFTER (unless they are NULL). */ void create_iv (tree base, tree step, tree var, struct loop *loop, - block_stmt_iterator *incr_pos, bool after, + gimple_stmt_iterator *incr_pos, bool after, tree *var_before, tree *var_after) { - tree stmt, initial, step1, stmts; + gimple stmt; + tree initial, step1; + gimple_seq stmts; tree vb, va; enum tree_code incr_op = PLUS_EXPR; + edge pe = loop_preheader_edge (loop); if (!var) { var = create_tmp_var (TREE_TYPE (base), "ivtmp"); - add_referenced_tmp_var (var); + add_referenced_var (var); } - vb = make_ssa_name (var, NULL_TREE); + vb = make_ssa_name (var, NULL); if (var_before) *var_before = vb; - va = make_ssa_name (var, NULL_TREE); + va = make_ssa_name (var, NULL); if (var_after) *var_after = va; @@ -73,7 +77,7 @@ create_iv (tree base, tree step, tree var, struct loop *loop, { if (TYPE_UNSIGNED (TREE_TYPE (step))) { - step1 = fold (build1 (NEGATE_EXPR, TREE_TYPE (step), step)); + step1 = fold_build1 (NEGATE_EXPR, TREE_TYPE (step), step); if (tree_int_cst_lt (step1, step)) { incr_op = MINUS_EXPR; @@ -82,36 +86,45 @@ create_iv (tree base, tree step, tree var, struct loop *loop, } else { - if (!tree_expr_nonnegative_p (step) + bool ovf; + + if (!tree_expr_nonnegative_warnv_p (step, &ovf) && may_negate_without_overflow_p (step)) { incr_op = MINUS_EXPR; - step = fold (build1 (NEGATE_EXPR, TREE_TYPE (step), step)); + step = fold_build1 (NEGATE_EXPR, TREE_TYPE (step), step); } } } + if (POINTER_TYPE_P (TREE_TYPE (base))) + { + if (TREE_CODE (base) == ADDR_EXPR) + mark_addressable (TREE_OPERAND (base, 0)); + step = convert_to_ptrofftype (step); + if (incr_op == MINUS_EXPR) + step = fold_build1 (NEGATE_EXPR, TREE_TYPE (step), step); + incr_op = POINTER_PLUS_EXPR; + } + /* Gimplify the step if necessary. We put the computations in front of the + loop (i.e. the step should be loop invariant). */ + step = force_gimple_operand (step, &stmts, true, NULL_TREE); + if (stmts) + gsi_insert_seq_on_edge_immediate (pe, stmts); - stmt = build2 (MODIFY_EXPR, void_type_node, va, - build2 (incr_op, TREE_TYPE (base), - vb, step)); - SSA_NAME_DEF_STMT (va) = stmt; + stmt = gimple_build_assign_with_ops (incr_op, va, vb, step); if (after) - bsi_insert_after (incr_pos, stmt, BSI_NEW_STMT); + gsi_insert_after (incr_pos, stmt, GSI_NEW_STMT); else - bsi_insert_before (incr_pos, stmt, BSI_NEW_STMT); + gsi_insert_before (incr_pos, stmt, GSI_NEW_STMT); initial = force_gimple_operand (base, &stmts, true, var); if (stmts) - { - edge pe = loop_preheader_edge (loop); - - bsi_insert_on_edge_immediate_loop (pe, stmts); - } + gsi_insert_seq_on_edge_immediate (pe, stmts); stmt = create_phi_node (vb, loop->header); SSA_NAME_DEF_STMT (vb) = stmt; - add_phi_arg (&stmt, initial, loop_preheader_edge (loop)); - add_phi_arg (&stmt, va, loop_latch_edge (loop)); + add_phi_arg (stmt, initial, loop_preheader_edge (loop), UNKNOWN_LOCATION); + add_phi_arg (stmt, va, loop_latch_edge (loop), UNKNOWN_LOCATION); } /* Add exit phis for the USE on EXIT. */ @@ -119,8 +132,8 @@ create_iv (tree base, tree step, tree var, struct loop *loop, static void add_exit_phis_edge (basic_block exit, tree use) { - tree phi, def_stmt = SSA_NAME_DEF_STMT (use); - basic_block def_bb = bb_for_stmt (def_stmt); + gimple phi, def_stmt = SSA_NAME_DEF_STMT (use); + basic_block def_bb = gimple_bb (def_stmt); struct loop *def_loop; edge e; edge_iterator ei; @@ -138,11 +151,10 @@ add_exit_phis_edge (basic_block exit, tree use) return; phi = create_phi_node (use, exit); - + create_new_def_for (gimple_phi_result (phi), phi, + gimple_phi_result_ptr (phi)); FOR_EACH_EDGE (e, ei, exit->preds) - add_phi_arg (&phi, use, e); - - SSA_NAME_DEF_STMT (use) = def_stmt; + add_phi_arg (phi, use, e, UNKNOWN_LOCATION); } /* Add exit phis for VAR that is used in LIVEIN. @@ -152,16 +164,19 @@ static void add_exit_phis_var (tree var, bitmap livein, bitmap exits) { bitmap def; - int index; - basic_block def_bb = bb_for_stmt (SSA_NAME_DEF_STMT (var)); + unsigned index; + basic_block def_bb = gimple_bb (SSA_NAME_DEF_STMT (var)); bitmap_iterator bi; - bitmap_clear_bit (livein, def_bb->index); + if (is_gimple_reg (var)) + bitmap_clear_bit (livein, def_bb->index); + else + bitmap_set_bit (livein, def_bb->index); - def = BITMAP_XMALLOC (); + def = BITMAP_ALLOC (NULL); bitmap_set_bit (def, def_bb->index); compute_global_livein (livein, def); - BITMAP_XFREE (def); + BITMAP_FREE (def); EXECUTE_IF_AND_IN_BITMAP (exits, livein, 0, index, bi) { @@ -190,7 +205,7 @@ add_exit_phis (bitmap names_to_rename, bitmap *use_blocks, bitmap loop_exits) static bitmap get_loops_exits (void) { - bitmap exits = BITMAP_XMALLOC (); + bitmap exits = BITMAP_ALLOC (NULL); basic_block bb; edge e; edge_iterator ei; @@ -211,10 +226,11 @@ get_loops_exits (void) /* For USE in BB, if it is used outside of the loop it is defined in, mark it for rewrite. Record basic block BB where it is used - to USE_BLOCKS. */ + to USE_BLOCKS. Record the ssa name index to NEED_PHIS bitmap. */ static void -find_uses_to_rename_use (basic_block bb, tree use, bitmap *use_blocks) +find_uses_to_rename_use (basic_block bb, tree use, bitmap *use_blocks, + bitmap need_phis) { unsigned ver; basic_block def_bb; @@ -223,62 +239,97 @@ find_uses_to_rename_use (basic_block bb, tree use, bitmap *use_blocks) if (TREE_CODE (use) != SSA_NAME) return; + /* We don't need to keep virtual operands in loop-closed form. */ + if (!is_gimple_reg (use)) + return; + ver = SSA_NAME_VERSION (use); - def_bb = bb_for_stmt (SSA_NAME_DEF_STMT (use)); + def_bb = gimple_bb (SSA_NAME_DEF_STMT (use)); if (!def_bb) return; def_loop = def_bb->loop_father; - /* If the definition is not inside loop, it is not interesting. */ - if (!def_loop->outer) + /* If the definition is not inside a loop, it is not interesting. */ + if (!loop_outer (def_loop)) + return; + + /* If the use is not outside of the loop it is defined in, it is not + interesting. */ + if (flow_bb_inside_loop_p (def_loop, bb)) return; if (!use_blocks[ver]) - use_blocks[ver] = BITMAP_XMALLOC (); + use_blocks[ver] = BITMAP_ALLOC (NULL); bitmap_set_bit (use_blocks[ver], bb->index); - if (!flow_bb_inside_loop_p (def_loop, bb)) - mark_for_rewrite (use); + bitmap_set_bit (need_phis, ver); } /* For uses in STMT, mark names that are used outside of the loop they are defined to rewrite. Record the set of blocks in that the ssa - names are defined to USE_BLOCKS. */ + names are defined to USE_BLOCKS and the ssa names themselves to + NEED_PHIS. */ static void -find_uses_to_rename_stmt (tree stmt, bitmap *use_blocks) +find_uses_to_rename_stmt (gimple stmt, bitmap *use_blocks, bitmap need_phis) { ssa_op_iter iter; tree var; - basic_block bb = bb_for_stmt (stmt); + basic_block bb = gimple_bb (stmt); - get_stmt_operands (stmt); + if (is_gimple_debug (stmt)) + return; - FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_ALL_USES | SSA_OP_ALL_KILLS) - find_uses_to_rename_use (bb, var, use_blocks); + FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_ALL_USES) + find_uses_to_rename_use (bb, var, use_blocks, need_phis); +} + +/* Marks names that are used in BB and outside of the loop they are + defined in for rewrite. Records the set of blocks in that the ssa + names are defined to USE_BLOCKS. Record the SSA names that will + need exit PHIs in NEED_PHIS. */ + +static void +find_uses_to_rename_bb (basic_block bb, bitmap *use_blocks, bitmap need_phis) +{ + gimple_stmt_iterator bsi; + edge e; + edge_iterator ei; + + FOR_EACH_EDGE (e, ei, bb->succs) + for (bsi = gsi_start_phis (e->dest); !gsi_end_p (bsi); gsi_next (&bsi)) + find_uses_to_rename_use (bb, PHI_ARG_DEF_FROM_EDGE (gsi_stmt (bsi), e), + use_blocks, need_phis); + + for (bsi = gsi_start_bb (bb); !gsi_end_p (bsi); gsi_next (&bsi)) + find_uses_to_rename_stmt (gsi_stmt (bsi), use_blocks, need_phis); } /* Marks names that are used outside of the loop they are defined in for rewrite. Records the set of blocks in that the ssa - names are defined to USE_BLOCKS. */ + names are defined to USE_BLOCKS. If CHANGED_BBS is not NULL, + scan only blocks in this set. */ static void -find_uses_to_rename (bitmap *use_blocks) +find_uses_to_rename (bitmap changed_bbs, bitmap *use_blocks, bitmap need_phis) { basic_block bb; - block_stmt_iterator bsi; - tree phi; - unsigned i; + unsigned index; + bitmap_iterator bi; - FOR_EACH_BB (bb) + if (changed_bbs && !bitmap_empty_p (changed_bbs)) { - for (phi = phi_nodes (bb); phi; phi = TREE_CHAIN (phi)) - for (i = 0; i < (unsigned) PHI_NUM_ARGS (phi); i++) - find_uses_to_rename_use (PHI_ARG_EDGE (phi, i)->src, - PHI_ARG_DEF (phi, i), use_blocks); - - for (bsi = bsi_start (bb); !bsi_end_p (bsi); bsi_next (&bsi)) - find_uses_to_rename_stmt (bsi_stmt (bsi), use_blocks); + EXECUTE_IF_SET_IN_BITMAP (changed_bbs, 0, index, bi) + { + find_uses_to_rename_bb (BASIC_BLOCK (index), use_blocks, need_phis); + } + } + else + { + FOR_EACH_BB (bb) + { + find_uses_to_rename_bb (bb, use_blocks, need_phis); + } } } @@ -306,36 +357,52 @@ find_uses_to_rename (bitmap *use_blocks) Looking from the outer loop with the normal SSA form, the first use of k is not well-behaved, while the second one is an induction variable with - base 99 and step 1. */ + base 99 and step 1. + + If CHANGED_BBS is not NULL, we look for uses outside loops only in + the basic blocks in this set. + + UPDATE_FLAG is used in the call to update_ssa. See + TODO_update_ssa* for documentation. */ void -rewrite_into_loop_closed_ssa (void) +rewrite_into_loop_closed_ssa (bitmap changed_bbs, unsigned update_flag) { - bitmap loop_exits = get_loops_exits (); + bitmap loop_exits; bitmap *use_blocks; - unsigned i; + unsigned i, old_num_ssa_names; bitmap names_to_rename; - gcc_assert (!any_marked_for_rewrite_p ()); + loops_state_set (LOOP_CLOSED_SSA); + if (number_of_loops () <= 1) + return; - use_blocks = xcalloc (num_ssa_names, sizeof (bitmap)); + loop_exits = get_loops_exits (); + names_to_rename = BITMAP_ALLOC (NULL); + + /* If the pass has caused the SSA form to be out-of-date, update it + now. */ + update_ssa (update_flag); + + old_num_ssa_names = num_ssa_names; + use_blocks = XCNEWVEC (bitmap, old_num_ssa_names); /* Find the uses outside loops. */ - find_uses_to_rename (use_blocks); + find_uses_to_rename (changed_bbs, use_blocks, names_to_rename); - /* Add the phi nodes on exits of the loops for the names we need to + /* Add the PHI nodes on exits of the loops for the names we need to rewrite. */ - names_to_rename = marked_ssa_names (); add_exit_phis (names_to_rename, use_blocks, loop_exits); - for (i = 0; i < num_ssa_names; i++) - BITMAP_XFREE (use_blocks[i]); + for (i = 0; i < old_num_ssa_names; i++) + BITMAP_FREE (use_blocks[i]); free (use_blocks); - BITMAP_XFREE (loop_exits); - BITMAP_XFREE (names_to_rename); + BITMAP_FREE (loop_exits); + BITMAP_FREE (names_to_rename); - /* Do the rewriting. */ - rewrite_ssa_into_ssa (); + /* Fix up all the names found to be used outside their original + loops. */ + update_ssa (TODO_update_ssa); } /* Check invariants of the loop closed ssa form for the USE in BB. */ @@ -343,14 +410,14 @@ rewrite_into_loop_closed_ssa (void) static void check_loop_closed_ssa_use (basic_block bb, tree use) { - tree def; + gimple def; basic_block def_bb; - - if (TREE_CODE (use) != SSA_NAME) + + if (TREE_CODE (use) != SSA_NAME || !is_gimple_reg (use)) return; def = SSA_NAME_DEF_STMT (use); - def_bb = bb_for_stmt (def); + def_bb = gimple_bb (def); gcc_assert (!def_bb || flow_bb_inside_loop_p (def_bb->loop_father, bb)); } @@ -358,97 +425,92 @@ check_loop_closed_ssa_use (basic_block bb, tree use) /* Checks invariants of loop closed ssa form in statement STMT in BB. */ static void -check_loop_closed_ssa_stmt (basic_block bb, tree stmt) +check_loop_closed_ssa_stmt (basic_block bb, gimple stmt) { ssa_op_iter iter; tree var; - get_stmt_operands (stmt); + if (is_gimple_debug (stmt)) + return; FOR_EACH_SSA_TREE_OPERAND (var, stmt, iter, SSA_OP_ALL_USES) check_loop_closed_ssa_use (bb, var); } -/* Checks that invariants of the loop closed ssa form are preserved. */ +/* Checks that invariants of the loop closed ssa form are preserved. + Call verify_ssa when VERIFY_SSA_P is true. */ -void -verify_loop_closed_ssa (void) +DEBUG_FUNCTION void +verify_loop_closed_ssa (bool verify_ssa_p) { basic_block bb; - block_stmt_iterator bsi; - tree phi; - unsigned i; + gimple_stmt_iterator bsi; + gimple phi; + edge e; + edge_iterator ei; - verify_ssa (); + if (number_of_loops () <= 1) + return; + + if (verify_ssa_p) + verify_ssa (false); + + timevar_push (TV_VERIFY_LOOP_CLOSED); FOR_EACH_BB (bb) { - for (phi = phi_nodes (bb); phi; phi = TREE_CHAIN (phi)) - for (i = 0; i < (unsigned) PHI_NUM_ARGS (phi); i++) - check_loop_closed_ssa_use (PHI_ARG_EDGE (phi, i)->src, - PHI_ARG_DEF (phi, i)); + for (bsi = gsi_start_phis (bb); !gsi_end_p (bsi); gsi_next (&bsi)) + { + phi = gsi_stmt (bsi); + FOR_EACH_EDGE (e, ei, bb->preds) + check_loop_closed_ssa_use (e->src, + PHI_ARG_DEF_FROM_EDGE (phi, e)); + } - for (bsi = bsi_start (bb); !bsi_end_p (bsi); bsi_next (&bsi)) - check_loop_closed_ssa_stmt (bb, bsi_stmt (bsi)); + for (bsi = gsi_start_bb (bb); !gsi_end_p (bsi); gsi_next (&bsi)) + check_loop_closed_ssa_stmt (bb, gsi_stmt (bsi)); } + + timevar_pop (TV_VERIFY_LOOP_CLOSED); } /* Split loop exit edge EXIT. The things are a bit complicated by a need to - preserve the loop closed ssa form. */ + preserve the loop closed ssa form. The newly created block is returned. */ -void +basic_block split_loop_exit_edge (edge exit) { basic_block dest = exit->dest; - basic_block bb = loop_split_edge_with (exit, NULL); - tree phi, new_phi, new_name, name; + basic_block bb = split_edge (exit); + gimple phi, new_phi; + tree new_name, name; use_operand_p op_p; + gimple_stmt_iterator psi; + source_location locus; - for (phi = phi_nodes (dest); phi; phi = TREE_CHAIN (phi)) + for (psi = gsi_start_phis (dest); !gsi_end_p (psi); gsi_next (&psi)) { - op_p = PHI_ARG_DEF_PTR_FROM_EDGE (phi, EDGE_SUCC (bb, 0)); + phi = gsi_stmt (psi); + op_p = PHI_ARG_DEF_PTR_FROM_EDGE (phi, single_succ_edge (bb)); + locus = gimple_phi_arg_location_from_edge (phi, single_succ_edge (bb)); name = USE_FROM_PTR (op_p); - /* If the argument of the phi node is a constant, we do not need + /* If the argument of the PHI node is a constant, we do not need to keep it inside loop. */ if (TREE_CODE (name) != SSA_NAME) continue; /* Otherwise create an auxiliary phi node that will copy the value - of the ssa name out of the loop. */ + of the SSA name out of the loop. */ new_name = duplicate_ssa_name (name, NULL); new_phi = create_phi_node (new_name, bb); SSA_NAME_DEF_STMT (new_name) = new_phi; - add_phi_arg (&new_phi, name, exit); + add_phi_arg (new_phi, name, exit, locus); SET_USE (op_p, new_name); } -} - -/* Insert statement STMT to the edge E and update the loop structures. - Returns the newly created block (if any). */ - -basic_block -bsi_insert_on_edge_immediate_loop (edge e, tree stmt) -{ - basic_block src, dest, new_bb; - struct loop *loop_c; - - src = e->src; - dest = e->dest; - loop_c = find_common_loop (src->loop_father, dest->loop_father); - - new_bb = bsi_insert_on_edge_immediate (e, stmt); - - if (!new_bb) - return NULL; - - add_bb_to_loop (new_bb, loop_c); - if (dest->loop_father->latch == src) - dest->loop_father->latch = new_bb; - - return new_bb; + return bb; } /* Returns the basic block in that statements should be emitted for induction @@ -466,16 +528,17 @@ ip_end_pos (struct loop *loop) basic_block ip_normal_pos (struct loop *loop) { - tree last; + gimple last; basic_block bb; edge exit; - if (EDGE_COUNT (loop->latch->preds) > 1) + if (!single_pred_p (loop->latch)) return NULL; - bb = EDGE_PRED (loop->latch, 0)->src; + bb = single_pred (loop->latch); last = last_stmt (bb); - if (TREE_CODE (last) != COND_EXPR) + if (!last + || gimple_code (last) != GIMPLE_COND) return NULL; exit = EDGE_SUCC (bb, 0); @@ -494,21 +557,21 @@ ip_normal_pos (struct loop *loop) the increment should be inserted after *BSI. */ void -standard_iv_increment_position (struct loop *loop, block_stmt_iterator *bsi, +standard_iv_increment_position (struct loop *loop, gimple_stmt_iterator *bsi, bool *insert_after) { basic_block bb = ip_normal_pos (loop), latch = ip_end_pos (loop); - tree last = last_stmt (latch); + gimple last = last_stmt (latch); if (!bb - || (last && TREE_CODE (last) != LABEL_EXPR)) + || (last && gimple_code (last) != GIMPLE_LABEL)) { - *bsi = bsi_last (latch); + *bsi = gsi_last_bb (latch); *insert_after = true; } else { - *bsi = bsi_last (bb); + *bsi = gsi_last_bb (bb); *insert_after = false; } } @@ -522,89 +585,46 @@ copy_phi_node_args (unsigned first_new_block) unsigned i; for (i = first_new_block; i < (unsigned) last_basic_block; i++) - BASIC_BLOCK (i)->rbi->duplicated = 1; + BASIC_BLOCK (i)->flags |= BB_DUPLICATED; for (i = first_new_block; i < (unsigned) last_basic_block; i++) add_phi_args_after_copy_bb (BASIC_BLOCK (i)); for (i = first_new_block; i < (unsigned) last_basic_block; i++) - BASIC_BLOCK (i)->rbi->duplicated = 0; + BASIC_BLOCK (i)->flags &= ~BB_DUPLICATED; } -/* Renames variables in the area copied by tree_duplicate_loop_to_header_edge. - FIRST_NEW_BLOCK is the first block in the copied area. DEFINITIONS is - a bitmap of all ssa names defined inside the loop. */ -static void -rename_variables (unsigned first_new_block, bitmap definitions) -{ - unsigned i, copy_number = 0; - basic_block bb; - htab_t ssa_name_map = NULL; +/* The same as cfgloopmanip.c:duplicate_loop_to_header_edge, but also + updates the PHI nodes at start of the copied region. In order to + achieve this, only loops whose exits all lead to the same location + are handled. - for (i = first_new_block; i < (unsigned) last_basic_block; i++) - { - bb = BASIC_BLOCK (i); - - /* We assume that first come all blocks from the first copy, then all - blocks from the second copy, etc. */ - if (copy_number != (unsigned) bb->rbi->copy_number) - { - allocate_ssa_names (definitions, &ssa_name_map); - copy_number = bb->rbi->copy_number; - } - - rewrite_to_new_ssa_names_bb (bb, ssa_name_map); - } - - htab_delete (ssa_name_map); -} - -/* Sets SSA_NAME_DEF_STMT for results of all phi nodes in BB. */ - -static void -set_phi_def_stmts (basic_block bb) -{ - tree phi; - - for (phi = phi_nodes (bb); phi; phi = TREE_CHAIN (phi)) - SSA_NAME_DEF_STMT (PHI_RESULT (phi)) = phi; -} - -/* The same ad cfgloopmanip.c:duplicate_loop_to_header_edge, but also updates - ssa. In order to achieve this, only loops whose exits all lead to the same - location are handled. - - FIXME: we create some degenerate phi nodes that could be avoided by copy - propagating them instead. Unfortunately this is not completely - straightforward due to problems with constant folding. */ + Notice that we do not completely update the SSA web after + duplication. The caller is responsible for calling update_ssa + after the loop has been duplicated. */ bool -tree_duplicate_loop_to_header_edge (struct loop *loop, edge e, - struct loops *loops, +gimple_duplicate_loop_to_header_edge (struct loop *loop, edge e, unsigned int ndupl, sbitmap wont_exit, - edge orig, edge *to_remove, - unsigned int *n_to_remove, int flags) + edge orig, VEC (edge, heap) **to_remove, + int flags) { unsigned first_new_block; - basic_block bb; - unsigned i; - bitmap definitions; - if (!(loops->state & LOOPS_HAVE_SIMPLE_LATCHES)) + if (!loops_state_satisfies_p (LOOPS_HAVE_SIMPLE_LATCHES)) return false; - if (!(loops->state & LOOPS_HAVE_PREHEADERS)) + if (!loops_state_satisfies_p (LOOPS_HAVE_PREHEADERS)) return false; #ifdef ENABLE_CHECKING - verify_loop_closed_ssa (); + if (loops_state_satisfies_p (LOOP_CLOSED_SSA)) + verify_loop_closed_ssa (true); #endif - gcc_assert (!any_marked_for_rewrite_p ()); - first_new_block = last_basic_block; - if (!duplicate_loop_to_header_edge (loop, e, loops, ndupl, wont_exit, - orig, to_remove, n_to_remove, flags)) + if (!duplicate_loop_to_header_edge (loop, e, ndupl, wont_exit, + orig, to_remove, flags)) return false; /* Readd the removed phi args for e. */ @@ -613,209 +633,635 @@ tree_duplicate_loop_to_header_edge (struct loop *loop, edge e, /* Copy the phi node arguments. */ copy_phi_node_args (first_new_block); - /* Rename the variables. */ - definitions = marked_ssa_names (); - rename_variables (first_new_block, definitions); - unmark_all_for_rewrite (); - BITMAP_XFREE (definitions); + scev_reset (); - /* For some time we have the identical ssa names as results in multiple phi - nodes. When phi node is resized, it sets SSA_NAME_DEF_STMT of its result - to the new copy. This means that we cannot easily ensure that the ssa - names defined in those phis are pointing to the right one -- so just - recompute SSA_NAME_DEF_STMT for them. */ + return true; +} - for (i = first_new_block; i < (unsigned) last_basic_block; i++) +/* Returns true if we can unroll LOOP FACTOR times. Number + of iterations of the loop is returned in NITER. */ + +bool +can_unroll_loop_p (struct loop *loop, unsigned factor, + struct tree_niter_desc *niter) +{ + edge exit; + + /* Check whether unrolling is possible. We only want to unroll loops + for that we are able to determine number of iterations. We also + want to split the extra iterations of the loop from its end, + therefore we require that the loop has precisely one + exit. */ + + exit = single_dom_exit (loop); + if (!exit) + return false; + + if (!number_of_iterations_exit (loop, exit, niter, false) + || niter->cmp == ERROR_MARK + /* Scalar evolutions analysis might have copy propagated + the abnormal ssa names into these expressions, hence + emitting the computations based on them during loop + unrolling might create overlapping life ranges for + them, and failures in out-of-ssa. */ + || contains_abnormal_ssa_name_p (niter->may_be_zero) + || contains_abnormal_ssa_name_p (niter->control.base) + || contains_abnormal_ssa_name_p (niter->control.step) + || contains_abnormal_ssa_name_p (niter->bound)) + return false; + + /* And of course, we must be able to duplicate the loop. */ + if (!can_duplicate_loop_p (loop)) + return false; + + /* The final loop should be small enough. */ + if (tree_num_loop_insns (loop, &eni_size_weights) * factor + > (unsigned) PARAM_VALUE (PARAM_MAX_UNROLLED_INSNS)) + return false; + + return true; +} + +/* Determines the conditions that control execution of LOOP unrolled FACTOR + times. DESC is number of iterations of LOOP. ENTER_COND is set to + condition that must be true if the main loop can be entered. + EXIT_BASE, EXIT_STEP, EXIT_CMP and EXIT_BOUND are set to values describing + how the exit from the unrolled loop should be controlled. */ + +static void +determine_exit_conditions (struct loop *loop, struct tree_niter_desc *desc, + unsigned factor, tree *enter_cond, + tree *exit_base, tree *exit_step, + enum tree_code *exit_cmp, tree *exit_bound) +{ + gimple_seq stmts; + tree base = desc->control.base; + tree step = desc->control.step; + tree bound = desc->bound; + tree type = TREE_TYPE (step); + tree bigstep, delta; + tree min = lower_bound_in_type (type, type); + tree max = upper_bound_in_type (type, type); + enum tree_code cmp = desc->cmp; + tree cond = boolean_true_node, assum; + + /* For pointers, do the arithmetics in the type of step. */ + base = fold_convert (type, base); + bound = fold_convert (type, bound); + + *enter_cond = boolean_false_node; + *exit_base = NULL_TREE; + *exit_step = NULL_TREE; + *exit_cmp = ERROR_MARK; + *exit_bound = NULL_TREE; + gcc_assert (cmp != ERROR_MARK); + + /* We only need to be correct when we answer question + "Do at least FACTOR more iterations remain?" in the unrolled loop. + Thus, transforming BASE + STEP * i <> BOUND to + BASE + STEP * i < BOUND is ok. */ + if (cmp == NE_EXPR) { - bb = BASIC_BLOCK (i); - set_phi_def_stmts (bb); - if (bb->rbi->copy_number == 1) - set_phi_def_stmts (bb->rbi->original); + if (tree_int_cst_sign_bit (step)) + cmp = GT_EXPR; + else + cmp = LT_EXPR; } + else if (cmp == LT_EXPR) + { + gcc_assert (!tree_int_cst_sign_bit (step)); + } + else if (cmp == GT_EXPR) + { + gcc_assert (tree_int_cst_sign_bit (step)); + } + else + gcc_unreachable (); + + /* The main body of the loop may be entered iff: + + 1) desc->may_be_zero is false. + 2) it is possible to check that there are at least FACTOR iterations + of the loop, i.e., BOUND - step * FACTOR does not overflow. + 3) # of iterations is at least FACTOR */ + + if (!integer_zerop (desc->may_be_zero)) + cond = fold_build2 (TRUTH_AND_EXPR, boolean_type_node, + invert_truthvalue (desc->may_be_zero), + cond); + + bigstep = fold_build2 (MULT_EXPR, type, step, + build_int_cst_type (type, factor)); + delta = fold_build2 (MINUS_EXPR, type, bigstep, step); + if (cmp == LT_EXPR) + assum = fold_build2 (GE_EXPR, boolean_type_node, + bound, + fold_build2 (PLUS_EXPR, type, min, delta)); + else + assum = fold_build2 (LE_EXPR, boolean_type_node, + bound, + fold_build2 (PLUS_EXPR, type, max, delta)); + cond = fold_build2 (TRUTH_AND_EXPR, boolean_type_node, assum, cond); - scev_reset (); -#ifdef ENABLE_CHECKING - verify_loop_closed_ssa (); -#endif + bound = fold_build2 (MINUS_EXPR, type, bound, delta); + assum = fold_build2 (cmp, boolean_type_node, base, bound); + cond = fold_build2 (TRUTH_AND_EXPR, boolean_type_node, assum, cond); - return true; + cond = force_gimple_operand (unshare_expr (cond), &stmts, false, NULL_TREE); + if (stmts) + gsi_insert_seq_on_edge_immediate (loop_preheader_edge (loop), stmts); + /* cond now may be a gimple comparison, which would be OK, but also any + other gimple rhs (say a && b). In this case we need to force it to + operand. */ + if (!is_gimple_condexpr (cond)) + { + cond = force_gimple_operand (cond, &stmts, true, NULL_TREE); + if (stmts) + gsi_insert_seq_on_edge_immediate (loop_preheader_edge (loop), stmts); + } + *enter_cond = cond; + + base = force_gimple_operand (unshare_expr (base), &stmts, true, NULL_TREE); + if (stmts) + gsi_insert_seq_on_edge_immediate (loop_preheader_edge (loop), stmts); + bound = force_gimple_operand (unshare_expr (bound), &stmts, true, NULL_TREE); + if (stmts) + gsi_insert_seq_on_edge_immediate (loop_preheader_edge (loop), stmts); + + *exit_base = base; + *exit_step = bigstep; + *exit_cmp = cmp; + *exit_bound = bound; } -/*--------------------------------------------------------------------------- - Loop versioning - ---------------------------------------------------------------------------*/ - -/* Adjust phi nodes for 'first' basic block. 'second' basic block is a copy - of 'first'. Both of them are dominated by 'new_head' basic block. When - 'new_head' was created by 'second's incoming edge it received phi arguments - on the edge by split_edge(). Later, additional edge 'e' was created to - connect 'new_head' and 'first'. Now this routine adds phi args on this - additional edge 'e' that new_head to second edge received as part of edge - splitting. -*/ +/* Scales the frequencies of all basic blocks in LOOP that are strictly + dominated by BB by NUM/DEN. */ static void -lv_adjust_loop_header_phi (basic_block first, basic_block second, - basic_block new_head, edge e) +scale_dominated_blocks_in_loop (struct loop *loop, basic_block bb, + int num, int den) { - tree phi1, phi2; + basic_block son; + + if (den == 0) + return; + + for (son = first_dom_son (CDI_DOMINATORS, bb); + son; + son = next_dom_son (CDI_DOMINATORS, son)) + { + if (!flow_bb_inside_loop_p (loop, son)) + continue; + scale_bbs_frequencies_int (&son, 1, num, den); + scale_dominated_blocks_in_loop (loop, son, num, den); + } +} + +/* Unroll LOOP FACTOR times. DESC describes number of iterations of LOOP. + EXIT is the exit of the loop to that DESC corresponds. + + If N is number of iterations of the loop and MAY_BE_ZERO is the condition + under that loop exits in the first iteration even if N != 0, - /* Browse all 'second' basic block phi nodes and add phi args to - edge 'e' for 'first' head. PHI args are always in correct order. */ + while (1) + { + x = phi (init, next); + + pre; + if (st) + break; + post; + } + + becomes (with possibly the exit conditions formulated a bit differently, + avoiding the need to create a new iv): + + if (MAY_BE_ZERO || N < FACTOR) + goto rest; + + do + { + x = phi (init, next); + + pre; + post; + pre; + post; + ... + pre; + post; + N -= FACTOR; + + } while (N >= FACTOR); + + rest: + init' = phi (init, x); + + while (1) + { + x = phi (init', next); + + pre; + if (st) + break; + post; + } + + Before the loop is unrolled, TRANSFORM is called for it (only for the + unrolled loop, but not for its versioned copy). DATA is passed to + TRANSFORM. */ + +/* Probability in % that the unrolled loop is entered. Just a guess. */ +#define PROB_UNROLLED_LOOP_ENTERED 90 + +void +tree_transform_and_unroll_loop (struct loop *loop, unsigned factor, + edge exit, struct tree_niter_desc *desc, + transform_callback transform, + void *data) +{ + gimple exit_if; + tree ctr_before, ctr_after; + tree enter_main_cond, exit_base, exit_step, exit_bound; + enum tree_code exit_cmp; + gimple phi_old_loop, phi_new_loop, phi_rest; + gimple_stmt_iterator psi_old_loop, psi_new_loop; + tree init, next, new_init, var; + struct loop *new_loop; + basic_block rest, exit_bb; + edge old_entry, new_entry, old_latch, precond_edge, new_exit; + edge new_nonexit, e; + gimple_stmt_iterator bsi; + use_operand_p op; + bool ok; + unsigned est_niter, prob_entry, scale_unrolled, scale_rest, freq_e, freq_h; + unsigned new_est_niter, i, prob; + unsigned irr = loop_preheader_edge (loop)->flags & EDGE_IRREDUCIBLE_LOOP; + sbitmap wont_exit; + VEC (edge, heap) *to_remove = NULL; + + est_niter = expected_loop_iterations (loop); + determine_exit_conditions (loop, desc, factor, + &enter_main_cond, &exit_base, &exit_step, + &exit_cmp, &exit_bound); + + /* Let us assume that the unrolled loop is quite likely to be entered. */ + if (integer_nonzerop (enter_main_cond)) + prob_entry = REG_BR_PROB_BASE; + else + prob_entry = PROB_UNROLLED_LOOP_ENTERED * REG_BR_PROB_BASE / 100; + + /* The values for scales should keep profile consistent, and somewhat close + to correct. + + TODO: The current value of SCALE_REST makes it appear that the loop that + is created by splitting the remaining iterations of the unrolled loop is + executed the same number of times as the original loop, and with the same + frequencies, which is obviously wrong. This does not appear to cause + problems, so we do not bother with fixing it for now. To make the profile + correct, we would need to change the probability of the exit edge of the + loop, and recompute the distribution of frequencies in its body because + of this change (scale the frequencies of blocks before and after the exit + by appropriate factors). */ + scale_unrolled = prob_entry; + scale_rest = REG_BR_PROB_BASE; + + new_loop = loop_version (loop, enter_main_cond, NULL, + prob_entry, scale_unrolled, scale_rest, true); + gcc_assert (new_loop != NULL); + update_ssa (TODO_update_ssa); + + /* Determine the probability of the exit edge of the unrolled loop. */ + new_est_niter = est_niter / factor; + + /* Without profile feedback, loops for that we do not know a better estimate + are assumed to roll 10 times. When we unroll such loop, it appears to + roll too little, and it may even seem to be cold. To avoid this, we + ensure that the created loop appears to roll at least 5 times (but at + most as many times as before unrolling). */ + if (new_est_niter < 5) + { + if (est_niter < 5) + new_est_niter = est_niter; + else + new_est_niter = 5; + } - for (phi2 = phi_nodes (second), phi1 = phi_nodes (first); - phi2 && phi1; - phi2 = TREE_CHAIN (phi2), phi1 = TREE_CHAIN (phi1)) + /* Prepare the cfg and update the phi nodes. Move the loop exit to the + loop latch (and make its condition dummy, for the moment). */ + rest = loop_preheader_edge (new_loop)->src; + precond_edge = single_pred_edge (rest); + split_edge (loop_latch_edge (loop)); + exit_bb = single_pred (loop->latch); + + /* Since the exit edge will be removed, the frequency of all the blocks + in the loop that are dominated by it must be scaled by + 1 / (1 - exit->probability). */ + scale_dominated_blocks_in_loop (loop, exit->src, + REG_BR_PROB_BASE, + REG_BR_PROB_BASE - exit->probability); + + bsi = gsi_last_bb (exit_bb); + exit_if = gimple_build_cond (EQ_EXPR, integer_zero_node, + integer_zero_node, + NULL_TREE, NULL_TREE); + + gsi_insert_after (&bsi, exit_if, GSI_NEW_STMT); + new_exit = make_edge (exit_bb, rest, EDGE_FALSE_VALUE | irr); + rescan_loop_exit (new_exit, true, false); + + /* Set the probability of new exit to the same of the old one. Fix + the frequency of the latch block, by scaling it back by + 1 - exit->probability. */ + new_exit->count = exit->count; + new_exit->probability = exit->probability; + new_nonexit = single_pred_edge (loop->latch); + new_nonexit->probability = REG_BR_PROB_BASE - exit->probability; + new_nonexit->flags = EDGE_TRUE_VALUE; + new_nonexit->count -= exit->count; + if (new_nonexit->count < 0) + new_nonexit->count = 0; + scale_bbs_frequencies_int (&loop->latch, 1, new_nonexit->probability, + REG_BR_PROB_BASE); + + old_entry = loop_preheader_edge (loop); + new_entry = loop_preheader_edge (new_loop); + old_latch = loop_latch_edge (loop); + for (psi_old_loop = gsi_start_phis (loop->header), + psi_new_loop = gsi_start_phis (new_loop->header); + !gsi_end_p (psi_old_loop); + gsi_next (&psi_old_loop), gsi_next (&psi_new_loop)) { - int i; - for (i = 0; i < PHI_NUM_ARGS (phi2); i++) + phi_old_loop = gsi_stmt (psi_old_loop); + phi_new_loop = gsi_stmt (psi_new_loop); + + init = PHI_ARG_DEF_FROM_EDGE (phi_old_loop, old_entry); + op = PHI_ARG_DEF_PTR_FROM_EDGE (phi_new_loop, new_entry); + gcc_assert (operand_equal_for_phi_arg_p (init, USE_FROM_PTR (op))); + next = PHI_ARG_DEF_FROM_EDGE (phi_old_loop, old_latch); + + /* Prefer using original variable as a base for the new ssa name. + This is necessary for virtual ops, and useful in order to avoid + losing debug info for real ops. */ + if (TREE_CODE (next) == SSA_NAME + && useless_type_conversion_p (TREE_TYPE (next), + TREE_TYPE (init))) + var = SSA_NAME_VAR (next); + else if (TREE_CODE (init) == SSA_NAME + && useless_type_conversion_p (TREE_TYPE (init), + TREE_TYPE (next))) + var = SSA_NAME_VAR (init); + else if (useless_type_conversion_p (TREE_TYPE (next), TREE_TYPE (init))) { - if (PHI_ARG_EDGE (phi2, i)->src == new_head) - { - tree def = PHI_ARG_DEF (phi2, i); - add_phi_arg (&phi1, def, e); - } + var = create_tmp_var (TREE_TYPE (next), "unrinittmp"); + add_referenced_var (var); } + else + { + var = create_tmp_var (TREE_TYPE (init), "unrinittmp"); + add_referenced_var (var); + } + + new_init = make_ssa_name (var, NULL); + phi_rest = create_phi_node (new_init, rest); + SSA_NAME_DEF_STMT (new_init) = phi_rest; + + add_phi_arg (phi_rest, init, precond_edge, UNKNOWN_LOCATION); + add_phi_arg (phi_rest, next, new_exit, UNKNOWN_LOCATION); + SET_USE (op, new_init); } + + remove_path (exit); + + /* Transform the loop. */ + if (transform) + (*transform) (loop, data); + + /* Unroll the loop and remove the exits in all iterations except for the + last one. */ + wont_exit = sbitmap_alloc (factor); + sbitmap_ones (wont_exit); + RESET_BIT (wont_exit, factor - 1); + + ok = gimple_duplicate_loop_to_header_edge + (loop, loop_latch_edge (loop), factor - 1, + wont_exit, new_exit, &to_remove, DLTHE_FLAG_UPDATE_FREQ); + free (wont_exit); + gcc_assert (ok); + + FOR_EACH_VEC_ELT (edge, to_remove, i, e) + { + ok = remove_path (e); + gcc_assert (ok); + } + VEC_free (edge, heap, to_remove); + update_ssa (TODO_update_ssa); + + /* Ensure that the frequencies in the loop match the new estimated + number of iterations, and change the probability of the new + exit edge. */ + freq_h = loop->header->frequency; + freq_e = EDGE_FREQUENCY (loop_preheader_edge (loop)); + if (freq_h != 0) + scale_loop_frequencies (loop, freq_e * (new_est_niter + 1), freq_h); + + exit_bb = single_pred (loop->latch); + new_exit = find_edge (exit_bb, rest); + new_exit->count = loop_preheader_edge (loop)->count; + new_exit->probability = REG_BR_PROB_BASE / (new_est_niter + 1); + + rest->count += new_exit->count; + rest->frequency += EDGE_FREQUENCY (new_exit); + + new_nonexit = single_pred_edge (loop->latch); + prob = new_nonexit->probability; + new_nonexit->probability = REG_BR_PROB_BASE - new_exit->probability; + new_nonexit->count = exit_bb->count - new_exit->count; + if (new_nonexit->count < 0) + new_nonexit->count = 0; + if (prob > 0) + scale_bbs_frequencies_int (&loop->latch, 1, new_nonexit->probability, + prob); + + /* Finally create the new counter for number of iterations and add the new + exit instruction. */ + bsi = gsi_last_nondebug_bb (exit_bb); + exit_if = gsi_stmt (bsi); + create_iv (exit_base, exit_step, NULL_TREE, loop, + &bsi, false, &ctr_before, &ctr_after); + gimple_cond_set_code (exit_if, exit_cmp); + gimple_cond_set_lhs (exit_if, ctr_after); + gimple_cond_set_rhs (exit_if, exit_bound); + update_stmt (exit_if); + +#ifdef ENABLE_CHECKING + verify_flow_info (); + verify_dominators (CDI_DOMINATORS); + verify_loop_structure (); + verify_loop_closed_ssa (true); +#endif } -/* Adjust entry edge for lv. - - e is a incoming edge. - - --- edge e ---- > [second_head] - - Split it and insert new conditional expression and adjust edges. - - --- edge e ---> [cond expr] ---> [first_head] - | - +---------> [second_head] - -*/ - -static basic_block -lv_adjust_loop_entry_edge (basic_block first_head, - basic_block second_head, - edge e, - tree cond_expr) -{ - block_stmt_iterator bsi; - basic_block new_head = NULL; - tree goto1 = NULL_TREE; - tree goto2 = NULL_TREE; - tree new_cond_expr = NULL_TREE; - edge e0, e1; - - gcc_assert (e->dest == second_head); - - /* Split edge 'e'. This will create a new basic block, where we can - insert conditional expr. */ - new_head = split_edge (e); - - /* Build new conditional expr */ - goto1 = build1 (GOTO_EXPR, void_type_node, tree_block_label (first_head)); - goto2 = build1 (GOTO_EXPR, void_type_node, tree_block_label (second_head)); - new_cond_expr = build3 (COND_EXPR, void_type_node, cond_expr, goto1, goto2); - - /* Add new cond. in new head. */ - bsi = bsi_start (new_head); - bsi_insert_after (&bsi, new_cond_expr, BSI_NEW_STMT); - - /* Adjust edges appropriately to connect new head with first head - as well as second head. */ - e0 = EDGE_SUCC (new_head, 0); - e0->flags &= ~EDGE_FALLTHRU; - e0->flags |= EDGE_FALSE_VALUE; - e1 = make_edge (new_head, first_head, EDGE_TRUE_VALUE); - set_immediate_dominator (CDI_DOMINATORS, first_head, new_head); - set_immediate_dominator (CDI_DOMINATORS, second_head, new_head); - - /* Adjust loop header phi nodes. */ - lv_adjust_loop_header_phi (first_head, second_head, new_head, e1); - - return new_head; +/* Wrapper over tree_transform_and_unroll_loop for case we do not + want to transform the loop before unrolling. The meaning + of the arguments is the same as for tree_transform_and_unroll_loop. */ + +void +tree_unroll_loop (struct loop *loop, unsigned factor, + edge exit, struct tree_niter_desc *desc) +{ + tree_transform_and_unroll_loop (loop, factor, exit, desc, + NULL, NULL); } -/* Main entry point for Loop Versioning transformation. - -This transformation given a condition and a loop, creates --if (condition) { loop_copy1 } else { loop_copy2 }, -where loop_copy1 is the loop transformed in one way, and loop_copy2 -is the loop transformed in another way (or unchanged). 'condition' -may be a run time test for things that were not resolved by static -analysis (overlapping ranges (anti-aliasing), alignment, etc.). */ - -struct loop * -tree_ssa_loop_version (struct loops *loops, struct loop * loop, - tree cond_expr, basic_block *condition_bb) +/* Rewrite the phi node at position PSI in function of the main + induction variable MAIN_IV and insert the generated code at GSI. */ + +static void +rewrite_phi_with_iv (loop_p loop, + gimple_stmt_iterator *psi, + gimple_stmt_iterator *gsi, + tree main_iv) { - edge entry, latch_edge, exit, true_edge, false_edge; - basic_block first_head, second_head; - int irred_flag; - struct loop *nloop; + affine_iv iv; + gimple stmt, phi = gsi_stmt (*psi); + tree atype, mtype, val, res = PHI_RESULT (phi); - /* CHECKME: Loop versioning does not handle nested loop at this point. */ - if (loop->inner) - return NULL; + if (!is_gimple_reg (res) || res == main_iv) + { + gsi_next (psi); + return; + } - /* Record entry and latch edges for the loop */ - entry = loop_preheader_edge (loop); + if (!simple_iv (loop, loop, res, &iv, true)) + { + gsi_next (psi); + return; + } + + remove_phi_node (psi, false); + + atype = TREE_TYPE (res); + mtype = POINTER_TYPE_P (atype) ? sizetype : atype; + val = fold_build2 (MULT_EXPR, mtype, unshare_expr (iv.step), + fold_convert (mtype, main_iv)); + val = fold_build2 (POINTER_TYPE_P (atype) + ? POINTER_PLUS_EXPR : PLUS_EXPR, + atype, unshare_expr (iv.base), val); + val = force_gimple_operand_gsi (gsi, val, false, NULL_TREE, true, + GSI_SAME_STMT); + stmt = gimple_build_assign (res, val); + gsi_insert_before (gsi, stmt, GSI_SAME_STMT); + SSA_NAME_DEF_STMT (res) = stmt; +} - /* Note down head of loop as first_head. */ - first_head = entry->dest; +/* Rewrite all the phi nodes of LOOP in function of the main induction + variable MAIN_IV. */ - /* Duplicate loop. */ - irred_flag = entry->flags & EDGE_IRREDUCIBLE_LOOP; - entry->flags &= ~EDGE_IRREDUCIBLE_LOOP; - if (!tree_duplicate_loop_to_header_edge (loop, entry, loops, 1, - NULL, NULL, NULL, NULL, 0)) +static void +rewrite_all_phi_nodes_with_iv (loop_p loop, tree main_iv) +{ + unsigned i; + basic_block *bbs = get_loop_body_in_dom_order (loop); + gimple_stmt_iterator psi; + + for (i = 0; i < loop->num_nodes; i++) { - entry->flags |= irred_flag; - return NULL; + basic_block bb = bbs[i]; + gimple_stmt_iterator gsi = gsi_after_labels (bb); + + if (bb->loop_father != loop) + continue; + + for (psi = gsi_start_phis (bb); !gsi_end_p (psi); ) + rewrite_phi_with_iv (loop, &psi, &gsi, main_iv); } - /* After duplication entry edge now points to new loop head block. - Note down new head as second_head. */ - second_head = entry->dest; + free (bbs); +} - /* Split loop entry edge and insert new block with cond expr. */ - *condition_bb = lv_adjust_loop_entry_edge (first_head, second_head, entry, - cond_expr); +/* Bases all the induction variables in LOOP on a single induction + variable (unsigned with base 0 and step 1), whose final value is + compared with *NIT. When the IV type precision has to be larger + than *NIT type precision, *NIT is converted to the larger type, the + conversion code is inserted before the loop, and *NIT is updated to + the new definition. When BUMP_IN_LATCH is true, the induction + variable is incremented in the loop latch, otherwise it is + incremented in the loop header. Return the induction variable that + was created. */ + +tree +canonicalize_loop_ivs (struct loop *loop, tree *nit, bool bump_in_latch) +{ + unsigned precision = TYPE_PRECISION (TREE_TYPE (*nit)); + unsigned original_precision = precision; + tree type, var_before; + gimple_stmt_iterator gsi, psi; + gimple stmt; + edge exit = single_dom_exit (loop); + gimple_seq stmts; + enum machine_mode mode; + bool unsigned_p = false; + + for (psi = gsi_start_phis (loop->header); + !gsi_end_p (psi); gsi_next (&psi)) + { + gimple phi = gsi_stmt (psi); + tree res = PHI_RESULT (phi); + bool uns; + + type = TREE_TYPE (res); + if (!is_gimple_reg (res) + || (!INTEGRAL_TYPE_P (type) + && !POINTER_TYPE_P (type)) + || TYPE_PRECISION (type) < precision) + continue; - latch_edge = EDGE_SUCC (loop->latch->rbi->copy, 0); - - extract_true_false_edges_from_block (*condition_bb, &true_edge, &false_edge); - nloop = loopify (loops, - latch_edge, - EDGE_PRED (loop->header->rbi->copy, 0), - *condition_bb, true_edge, false_edge, - false /* Do not redirect all edges. */); + uns = POINTER_TYPE_P (type) | TYPE_UNSIGNED (type); - exit = loop->single_exit; - if (exit) - nloop->single_exit = find_edge (exit->src->rbi->copy, exit->dest); + if (TYPE_PRECISION (type) > precision) + unsigned_p = uns; + else + unsigned_p |= uns; - /* loopify redirected latch_edge. Update its PENDING_STMTS. */ - flush_pending_stmts (latch_edge); + precision = TYPE_PRECISION (type); + } - /* loopify redirected condition_bb's succ edge. Update its PENDING_STMTS. */ - extract_true_false_edges_from_block (*condition_bb, &true_edge, &false_edge); - flush_pending_stmts (false_edge); + mode = smallest_mode_for_size (precision, MODE_INT); + precision = GET_MODE_PRECISION (mode); + type = build_nonstandard_integer_type (precision, unsigned_p); - /* Adjust irreducible flag. */ - if (irred_flag) + if (original_precision != precision) { - (*condition_bb)->flags |= BB_IRREDUCIBLE_LOOP; - loop_preheader_edge (loop)->flags |= EDGE_IRREDUCIBLE_LOOP; - loop_preheader_edge (nloop)->flags |= EDGE_IRREDUCIBLE_LOOP; - EDGE_PRED ((*condition_bb), 0)->flags |= EDGE_IRREDUCIBLE_LOOP; + *nit = fold_convert (type, *nit); + *nit = force_gimple_operand (*nit, &stmts, true, NULL_TREE); + if (stmts) + gsi_insert_seq_on_edge_immediate (loop_preheader_edge (loop), stmts); } - /* At this point condition_bb is loop predheader with two successors, - first_head and second_head. Make sure that loop predheader has only - one successor. */ - loop_split_edge_with (loop_preheader_edge (loop), NULL); - loop_split_edge_with (loop_preheader_edge (nloop), NULL); + if (bump_in_latch) + gsi = gsi_last_bb (loop->latch); + else + gsi = gsi_last_nondebug_bb (loop->header); + create_iv (build_int_cst_type (type, 0), build_int_cst (type, 1), NULL_TREE, + loop, &gsi, bump_in_latch, &var_before, NULL); + + rewrite_all_phi_nodes_with_iv (loop, var_before); + + stmt = last_stmt (exit->src); + /* Make the loop exit if the control condition is not satisfied. */ + if (exit->flags & EDGE_TRUE_VALUE) + { + edge te, fe; + + extract_true_false_edges_from_block (exit->src, &te, &fe); + te->flags = EDGE_FALSE_VALUE; + fe->flags = EDGE_TRUE_VALUE; + } + gimple_cond_set_code (stmt, LT_EXPR); + gimple_cond_set_lhs (stmt, var_before); + gimple_cond_set_rhs (stmt, *nit); + update_stmt (stmt); - return nloop; + return var_before; }