/* Reassociation for trees. Copyright (C) 2005-2014 Free Software Foundation, Inc. Contributed by Daniel Berlin This file is part of GCC. GCC is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 3, or (at your option) any later version. GCC is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with GCC; see the file COPYING3. If not see . */ #include "config.h" #include "system.h" #include "coretypes.h" #include "hash-table.h" #include "tm.h" #include "rtl.h" #include "tm_p.h" #include "tree.h" #include "stor-layout.h" #include "basic-block.h" #include "gimple-pretty-print.h" #include "tree-inline.h" #include "pointer-set.h" #include "tree-ssa-alias.h" #include "internal-fn.h" #include "gimple-fold.h" #include "tree-eh.h" #include "gimple-expr.h" #include "is-a.h" #include "gimple.h" #include "gimple-iterator.h" #include "gimplify-me.h" #include "gimple-ssa.h" #include "tree-cfg.h" #include "tree-phinodes.h" #include "ssa-iterators.h" #include "stringpool.h" #include "tree-ssanames.h" #include "tree-ssa-loop-niter.h" #include "tree-ssa-loop.h" #include "expr.h" #include "tree-dfa.h" #include "tree-ssa.h" #include "tree-iterator.h" #include "tree-pass.h" #include "alloc-pool.h" #include "langhooks.h" #include "cfgloop.h" #include "flags.h" #include "target.h" #include "params.h" #include "diagnostic-core.h" /* This is a simple global reassociation pass. It is, in part, based on the LLVM pass of the same name (They do some things more/less than we do, in different orders, etc). It consists of five steps: 1. Breaking up subtract operations into addition + negate, where it would promote the reassociation of adds. 2. Left linearization of the expression trees, so that (A+B)+(C+D) becomes (((A+B)+C)+D), which is easier for us to rewrite later. During linearization, we place the operands of the binary expressions into a vector of operand_entry_t 3. Optimization of the operand lists, eliminating things like a + -a, a & a, etc. 3a. Combine repeated factors with the same occurrence counts into a __builtin_powi call that will later be optimized into an optimal number of multiplies. 4. Rewrite the expression trees we linearized and optimized so they are in proper rank order. 5. Repropagate negates, as nothing else will clean it up ATM. A bit of theory on #4, since nobody seems to write anything down about why it makes sense to do it the way they do it: We could do this much nicer theoretically, but don't (for reasons explained after how to do it theoretically nice :P). In order to promote the most redundancy elimination, you want binary expressions whose operands are the same rank (or preferably, the same value) exposed to the redundancy eliminator, for possible elimination. So the way to do this if we really cared, is to build the new op tree from the leaves to the roots, merging as you go, and putting the new op on the end of the worklist, until you are left with one thing on the worklist. IE if you have to rewrite the following set of operands (listed with rank in parentheses), with opcode PLUS_EXPR: a (1), b (1), c (1), d (2), e (2) We start with our merge worklist empty, and the ops list with all of those on it. You want to first merge all leaves of the same rank, as much as possible. So first build a binary op of mergetmp = a + b, and put "mergetmp" on the merge worklist. Because there is no three operand form of PLUS_EXPR, c is not going to be exposed to redundancy elimination as a rank 1 operand. So you might as well throw it on the merge worklist (you could also consider it to now be a rank two operand, and merge it with d and e, but in this case, you then have evicted e from a binary op. So at least in this situation, you can't win.) Then build a binary op of d + e mergetmp2 = d + e and put mergetmp2 on the merge worklist. so merge worklist = {mergetmp, c, mergetmp2} Continue building binary ops of these operations until you have only one operation left on the worklist. So we have build binary op mergetmp3 = mergetmp + c worklist = {mergetmp2, mergetmp3} mergetmp4 = mergetmp2 + mergetmp3 worklist = {mergetmp4} because we have one operation left, we can now just set the original statement equal to the result of that operation. This will at least expose a + b and d + e to redundancy elimination as binary operations. For extra points, you can reuse the old statements to build the mergetmps, since you shouldn't run out. So why don't we do this? Because it's expensive, and rarely will help. Most trees we are reassociating have 3 or less ops. If they have 2 ops, they already will be written into a nice single binary op. If you have 3 ops, a single simple check suffices to tell you whether the first two are of the same rank. If so, you know to order it mergetmp = op1 + op2 newstmt = mergetmp + op3 instead of mergetmp = op2 + op3 newstmt = mergetmp + op1 If all three are of the same rank, you can't expose them all in a single binary operator anyway, so the above is *still* the best you can do. Thus, this is what we do. When we have three ops left, we check to see what order to put them in, and call it a day. As a nod to vector sum reduction, we check if any of the ops are really a phi node that is a destructive update for the associating op, and keep the destructive update together for vector sum reduction recognition. */ /* Statistics */ static struct { int linearized; int constants_eliminated; int ops_eliminated; int rewritten; int pows_encountered; int pows_created; } reassociate_stats; /* Operator, rank pair. */ typedef struct operand_entry { unsigned int rank; int id; tree op; unsigned int count; } *operand_entry_t; static alloc_pool operand_entry_pool; /* This is used to assign a unique ID to each struct operand_entry so that qsort results are identical on different hosts. */ static int next_operand_entry_id; /* Starting rank number for a given basic block, so that we can rank operations using unmovable instructions in that BB based on the bb depth. */ static long *bb_rank; /* Operand->rank hashtable. */ static struct pointer_map_t *operand_rank; /* Forward decls. */ static long get_rank (tree); static bool reassoc_stmt_dominates_stmt_p (gimple, gimple); /* Wrapper around gsi_remove, which adjusts gimple_uid of debug stmts possibly added by gsi_remove. */ bool reassoc_remove_stmt (gimple_stmt_iterator *gsi) { gimple stmt = gsi_stmt (*gsi); if (!MAY_HAVE_DEBUG_STMTS || gimple_code (stmt) == GIMPLE_PHI) return gsi_remove (gsi, true); gimple_stmt_iterator prev = *gsi; gsi_prev (&prev); unsigned uid = gimple_uid (stmt); basic_block bb = gimple_bb (stmt); bool ret = gsi_remove (gsi, true); if (!gsi_end_p (prev)) gsi_next (&prev); else prev = gsi_start_bb (bb); gimple end_stmt = gsi_stmt (*gsi); while ((stmt = gsi_stmt (prev)) != end_stmt) { gcc_assert (stmt && is_gimple_debug (stmt) && gimple_uid (stmt) == 0); gimple_set_uid (stmt, uid); gsi_next (&prev); } return ret; } /* Bias amount for loop-carried phis. We want this to be larger than the depth of any reassociation tree we can see, but not larger than the rank difference between two blocks. */ #define PHI_LOOP_BIAS (1 << 15) /* Rank assigned to a phi statement. If STMT is a loop-carried phi of an innermost loop, and the phi has only a single use which is inside the loop, then the rank is the block rank of the loop latch plus an extra bias for the loop-carried dependence. This causes expressions calculated into an accumulator variable to be independent for each iteration of the loop. If STMT is some other phi, the rank is the block rank of its containing block. */ static long phi_rank (gimple stmt) { basic_block bb = gimple_bb (stmt); struct loop *father = bb->loop_father; tree res; unsigned i; use_operand_p use; gimple use_stmt; /* We only care about real loops (those with a latch). */ if (!father->latch) return bb_rank[bb->index]; /* Interesting phis must be in headers of innermost loops. */ if (bb != father->header || father->inner) return bb_rank[bb->index]; /* Ignore virtual SSA_NAMEs. */ res = gimple_phi_result (stmt); if (virtual_operand_p (res)) return bb_rank[bb->index]; /* The phi definition must have a single use, and that use must be within the loop. Otherwise this isn't an accumulator pattern. */ if (!single_imm_use (res, &use, &use_stmt) || gimple_bb (use_stmt)->loop_father != father) return bb_rank[bb->index]; /* Look for phi arguments from within the loop. If found, bias this phi. */ for (i = 0; i < gimple_phi_num_args (stmt); i++) { tree arg = gimple_phi_arg_def (stmt, i); if (TREE_CODE (arg) == SSA_NAME && !SSA_NAME_IS_DEFAULT_DEF (arg)) { gimple def_stmt = SSA_NAME_DEF_STMT (arg); if (gimple_bb (def_stmt)->loop_father == father) return bb_rank[father->latch->index] + PHI_LOOP_BIAS; } } /* Must be an uninteresting phi. */ return bb_rank[bb->index]; } /* If EXP is an SSA_NAME defined by a PHI statement that represents a loop-carried dependence of an innermost loop, return TRUE; else return FALSE. */ static bool loop_carried_phi (tree exp) { gimple phi_stmt; long block_rank; if (TREE_CODE (exp) != SSA_NAME || SSA_NAME_IS_DEFAULT_DEF (exp)) return false; phi_stmt = SSA_NAME_DEF_STMT (exp); if (gimple_code (SSA_NAME_DEF_STMT (exp)) != GIMPLE_PHI) return false; /* Non-loop-carried phis have block rank. Loop-carried phis have an additional bias added in. If this phi doesn't have block rank, it's biased and should not be propagated. */ block_rank = bb_rank[gimple_bb (phi_stmt)->index]; if (phi_rank (phi_stmt) != block_rank) return true; return false; } /* Return the maximum of RANK and the rank that should be propagated from expression OP. For most operands, this is just the rank of OP. For loop-carried phis, the value is zero to avoid undoing the bias in favor of the phi. */ static long propagate_rank (long rank, tree op) { long op_rank; if (loop_carried_phi (op)) return rank; op_rank = get_rank (op); return MAX (rank, op_rank); } /* Look up the operand rank structure for expression E. */ static inline long find_operand_rank (tree e) { void **slot = pointer_map_contains (operand_rank, e); return slot ? (long) (intptr_t) *slot : -1; } /* Insert {E,RANK} into the operand rank hashtable. */ static inline void insert_operand_rank (tree e, long rank) { void **slot; gcc_assert (rank > 0); slot = pointer_map_insert (operand_rank, e); gcc_assert (!*slot); *slot = (void *) (intptr_t) rank; } /* Given an expression E, return the rank of the expression. */ static long get_rank (tree e) { /* Constants have rank 0. */ if (is_gimple_min_invariant (e)) return 0; /* SSA_NAME's have the rank of the expression they are the result of. For globals and uninitialized values, the rank is 0. For function arguments, use the pre-setup rank. For PHI nodes, stores, asm statements, etc, we use the rank of the BB. For simple operations, the rank is the maximum rank of any of its operands, or the bb_rank, whichever is less. I make no claims that this is optimal, however, it gives good results. */ /* We make an exception to the normal ranking system to break dependences of accumulator variables in loops. Suppose we have a simple one-block loop containing: x_1 = phi(x_0, x_2) b = a + x_1 c = b + d x_2 = c + e As shown, each iteration of the calculation into x is fully dependent upon the iteration before it. We would prefer to see this in the form: x_1 = phi(x_0, x_2) b = a + d c = b + e x_2 = c + x_1 If the loop is unrolled, the calculations of b and c from different iterations can be interleaved. To obtain this result during reassociation, we bias the rank of the phi definition x_1 upward, when it is recognized as an accumulator pattern. The artificial rank causes it to be added last, providing the desired independence. */ if (TREE_CODE (e) == SSA_NAME) { gimple stmt; long rank; int i, n; tree op; if (SSA_NAME_IS_DEFAULT_DEF (e)) return find_operand_rank (e); stmt = SSA_NAME_DEF_STMT (e); if (gimple_code (stmt) == GIMPLE_PHI) return phi_rank (stmt); if (!is_gimple_assign (stmt) || gimple_vdef (stmt)) return bb_rank[gimple_bb (stmt)->index]; /* If we already have a rank for this expression, use that. */ rank = find_operand_rank (e); if (rank != -1) return rank; /* Otherwise, find the maximum rank for the operands. As an exception, remove the bias from loop-carried phis when propagating the rank so that dependent operations are not also biased. */ rank = 0; if (gimple_assign_single_p (stmt)) { tree rhs = gimple_assign_rhs1 (stmt); n = TREE_OPERAND_LENGTH (rhs); if (n == 0) rank = propagate_rank (rank, rhs); else { for (i = 0; i < n; i++) { op = TREE_OPERAND (rhs, i); if (op != NULL_TREE) rank = propagate_rank (rank, op); } } } else { n = gimple_num_ops (stmt); for (i = 1; i < n; i++) { op = gimple_op (stmt, i); gcc_assert (op); rank = propagate_rank (rank, op); } } if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Rank for "); print_generic_expr (dump_file, e, 0); fprintf (dump_file, " is %ld\n", (rank + 1)); } /* Note the rank in the hashtable so we don't recompute it. */ insert_operand_rank (e, (rank + 1)); return (rank + 1); } /* Globals, etc, are rank 0 */ return 0; } /* We want integer ones to end up last no matter what, since they are the ones we can do the most with. */ #define INTEGER_CONST_TYPE 1 << 3 #define FLOAT_CONST_TYPE 1 << 2 #define OTHER_CONST_TYPE 1 << 1 /* Classify an invariant tree into integer, float, or other, so that we can sort them to be near other constants of the same type. */ static inline int constant_type (tree t) { if (INTEGRAL_TYPE_P (TREE_TYPE (t))) return INTEGER_CONST_TYPE; else if (SCALAR_FLOAT_TYPE_P (TREE_TYPE (t))) return FLOAT_CONST_TYPE; else return OTHER_CONST_TYPE; } /* qsort comparison function to sort operand entries PA and PB by rank so that the sorted array is ordered by rank in decreasing order. */ static int sort_by_operand_rank (const void *pa, const void *pb) { const operand_entry_t oea = *(const operand_entry_t *)pa; const operand_entry_t oeb = *(const operand_entry_t *)pb; /* It's nicer for optimize_expression if constants that are likely to fold when added/multiplied//whatever are put next to each other. Since all constants have rank 0, order them by type. */ if (oeb->rank == 0 && oea->rank == 0) { if (constant_type (oeb->op) != constant_type (oea->op)) return constant_type (oeb->op) - constant_type (oea->op); else /* To make sorting result stable, we use unique IDs to determine order. */ return oeb->id - oea->id; } /* Lastly, make sure the versions that are the same go next to each other. */ if ((oeb->rank - oea->rank == 0) && TREE_CODE (oea->op) == SSA_NAME && TREE_CODE (oeb->op) == SSA_NAME) { /* As SSA_NAME_VERSION is assigned pretty randomly, because we reuse versions of removed SSA_NAMEs, so if possible, prefer to sort based on basic block and gimple_uid of the SSA_NAME_DEF_STMT. See PR60418. */ if (!SSA_NAME_IS_DEFAULT_DEF (oea->op) && !SSA_NAME_IS_DEFAULT_DEF (oeb->op) && SSA_NAME_VERSION (oeb->op) != SSA_NAME_VERSION (oea->op)) { gimple stmta = SSA_NAME_DEF_STMT (oea->op); gimple stmtb = SSA_NAME_DEF_STMT (oeb->op); basic_block bba = gimple_bb (stmta); basic_block bbb = gimple_bb (stmtb); if (bbb != bba) { if (bb_rank[bbb->index] != bb_rank[bba->index]) return bb_rank[bbb->index] - bb_rank[bba->index]; } else { bool da = reassoc_stmt_dominates_stmt_p (stmta, stmtb); bool db = reassoc_stmt_dominates_stmt_p (stmtb, stmta); if (da != db) return da ? 1 : -1; } } if (SSA_NAME_VERSION (oeb->op) != SSA_NAME_VERSION (oea->op)) return SSA_NAME_VERSION (oeb->op) - SSA_NAME_VERSION (oea->op); else return oeb->id - oea->id; } if (oeb->rank != oea->rank) return oeb->rank - oea->rank; else return oeb->id - oea->id; } /* Add an operand entry to *OPS for the tree operand OP. */ static void add_to_ops_vec (vec *ops, tree op) { operand_entry_t oe = (operand_entry_t) pool_alloc (operand_entry_pool); oe->op = op; oe->rank = get_rank (op); oe->id = next_operand_entry_id++; oe->count = 1; ops->safe_push (oe); } /* Add an operand entry to *OPS for the tree operand OP with repeat count REPEAT. */ static void add_repeat_to_ops_vec (vec *ops, tree op, HOST_WIDE_INT repeat) { operand_entry_t oe = (operand_entry_t) pool_alloc (operand_entry_pool); oe->op = op; oe->rank = get_rank (op); oe->id = next_operand_entry_id++; oe->count = repeat; ops->safe_push (oe); reassociate_stats.pows_encountered++; } /* Return true if STMT is reassociable operation containing a binary operation with tree code CODE, and is inside LOOP. */ static bool is_reassociable_op (gimple stmt, enum tree_code code, struct loop *loop) { basic_block bb = gimple_bb (stmt); if (gimple_bb (stmt) == NULL) return false; if (!flow_bb_inside_loop_p (loop, bb)) return false; if (is_gimple_assign (stmt) && gimple_assign_rhs_code (stmt) == code && has_single_use (gimple_assign_lhs (stmt))) return true; return false; } /* Given NAME, if NAME is defined by a unary operation OPCODE, return the operand of the negate operation. Otherwise, return NULL. */ static tree get_unary_op (tree name, enum tree_code opcode) { gimple stmt = SSA_NAME_DEF_STMT (name); if (!is_gimple_assign (stmt)) return NULL_TREE; if (gimple_assign_rhs_code (stmt) == opcode) return gimple_assign_rhs1 (stmt); return NULL_TREE; } /* If CURR and LAST are a pair of ops that OPCODE allows us to eliminate through equivalences, do so, remove them from OPS, and return true. Otherwise, return false. */ static bool eliminate_duplicate_pair (enum tree_code opcode, vec *ops, bool *all_done, unsigned int i, operand_entry_t curr, operand_entry_t last) { /* If we have two of the same op, and the opcode is & |, min, or max, we can eliminate one of them. If we have two of the same op, and the opcode is ^, we can eliminate both of them. */ if (last && last->op == curr->op) { switch (opcode) { case MAX_EXPR: case MIN_EXPR: case BIT_IOR_EXPR: case BIT_AND_EXPR: if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Equivalence: "); print_generic_expr (dump_file, curr->op, 0); fprintf (dump_file, " [&|minmax] "); print_generic_expr (dump_file, last->op, 0); fprintf (dump_file, " -> "); print_generic_stmt (dump_file, last->op, 0); } ops->ordered_remove (i); reassociate_stats.ops_eliminated ++; return true; case BIT_XOR_EXPR: if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Equivalence: "); print_generic_expr (dump_file, curr->op, 0); fprintf (dump_file, " ^ "); print_generic_expr (dump_file, last->op, 0); fprintf (dump_file, " -> nothing\n"); } reassociate_stats.ops_eliminated += 2; if (ops->length () == 2) { ops->create (0); add_to_ops_vec (ops, build_zero_cst (TREE_TYPE (last->op))); *all_done = true; } else { ops->ordered_remove (i-1); ops->ordered_remove (i-1); } return true; default: break; } } return false; } static vec plus_negates; /* If OPCODE is PLUS_EXPR, CURR->OP is a negate expression or a bitwise not expression, look in OPS for a corresponding positive operation to cancel it out. If we find one, remove the other from OPS, replace OPS[CURRINDEX] with 0 or -1, respectively, and return true. Otherwise, return false. */ static bool eliminate_plus_minus_pair (enum tree_code opcode, vec *ops, unsigned int currindex, operand_entry_t curr) { tree negateop; tree notop; unsigned int i; operand_entry_t oe; if (opcode != PLUS_EXPR || TREE_CODE (curr->op) != SSA_NAME) return false; negateop = get_unary_op (curr->op, NEGATE_EXPR); notop = get_unary_op (curr->op, BIT_NOT_EXPR); if (negateop == NULL_TREE && notop == NULL_TREE) return false; /* Any non-negated version will have a rank that is one less than the current rank. So once we hit those ranks, if we don't find one, we can stop. */ for (i = currindex + 1; ops->iterate (i, &oe) && oe->rank >= curr->rank - 1 ; i++) { if (oe->op == negateop) { if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Equivalence: "); print_generic_expr (dump_file, negateop, 0); fprintf (dump_file, " + -"); print_generic_expr (dump_file, oe->op, 0); fprintf (dump_file, " -> 0\n"); } ops->ordered_remove (i); add_to_ops_vec (ops, build_zero_cst (TREE_TYPE (oe->op))); ops->ordered_remove (currindex); reassociate_stats.ops_eliminated ++; return true; } else if (oe->op == notop) { tree op_type = TREE_TYPE (oe->op); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Equivalence: "); print_generic_expr (dump_file, notop, 0); fprintf (dump_file, " + ~"); print_generic_expr (dump_file, oe->op, 0); fprintf (dump_file, " -> -1\n"); } ops->ordered_remove (i); add_to_ops_vec (ops, build_int_cst_type (op_type, -1)); ops->ordered_remove (currindex); reassociate_stats.ops_eliminated ++; return true; } } /* CURR->OP is a negate expr in a plus expr: save it for later inspection in repropagate_negates(). */ if (negateop != NULL_TREE) plus_negates.safe_push (curr->op); return false; } /* If OPCODE is BIT_IOR_EXPR, BIT_AND_EXPR, and, CURR->OP is really a bitwise not expression, look in OPS for a corresponding operand to cancel it out. If we find one, remove the other from OPS, replace OPS[CURRINDEX] with 0, and return true. Otherwise, return false. */ static bool eliminate_not_pairs (enum tree_code opcode, vec *ops, unsigned int currindex, operand_entry_t curr) { tree notop; unsigned int i; operand_entry_t oe; if ((opcode != BIT_IOR_EXPR && opcode != BIT_AND_EXPR) || TREE_CODE (curr->op) != SSA_NAME) return false; notop = get_unary_op (curr->op, BIT_NOT_EXPR); if (notop == NULL_TREE) return false; /* Any non-not version will have a rank that is one less than the current rank. So once we hit those ranks, if we don't find one, we can stop. */ for (i = currindex + 1; ops->iterate (i, &oe) && oe->rank >= curr->rank - 1; i++) { if (oe->op == notop) { if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Equivalence: "); print_generic_expr (dump_file, notop, 0); if (opcode == BIT_AND_EXPR) fprintf (dump_file, " & ~"); else if (opcode == BIT_IOR_EXPR) fprintf (dump_file, " | ~"); print_generic_expr (dump_file, oe->op, 0); if (opcode == BIT_AND_EXPR) fprintf (dump_file, " -> 0\n"); else if (opcode == BIT_IOR_EXPR) fprintf (dump_file, " -> -1\n"); } if (opcode == BIT_AND_EXPR) oe->op = build_zero_cst (TREE_TYPE (oe->op)); else if (opcode == BIT_IOR_EXPR) oe->op = build_all_ones_cst (TREE_TYPE (oe->op)); reassociate_stats.ops_eliminated += ops->length () - 1; ops->truncate (0); ops->quick_push (oe); return true; } } return false; } /* Use constant value that may be present in OPS to try to eliminate operands. Note that this function is only really used when we've eliminated ops for other reasons, or merged constants. Across single statements, fold already does all of this, plus more. There is little point in duplicating logic, so I've only included the identities that I could ever construct testcases to trigger. */ static void eliminate_using_constants (enum tree_code opcode, vec *ops) { operand_entry_t oelast = ops->last (); tree type = TREE_TYPE (oelast->op); if (oelast->rank == 0 && (INTEGRAL_TYPE_P (type) || FLOAT_TYPE_P (type))) { switch (opcode) { case BIT_AND_EXPR: if (integer_zerop (oelast->op)) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found & 0, removing all other ops\n"); reassociate_stats.ops_eliminated += ops->length () - 1; ops->truncate (0); ops->quick_push (oelast); return; } } else if (integer_all_onesp (oelast->op)) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found & -1, removing\n"); ops->pop (); reassociate_stats.ops_eliminated++; } } break; case BIT_IOR_EXPR: if (integer_all_onesp (oelast->op)) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found | -1, removing all other ops\n"); reassociate_stats.ops_eliminated += ops->length () - 1; ops->truncate (0); ops->quick_push (oelast); return; } } else if (integer_zerop (oelast->op)) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found | 0, removing\n"); ops->pop (); reassociate_stats.ops_eliminated++; } } break; case MULT_EXPR: if (integer_zerop (oelast->op) || (FLOAT_TYPE_P (type) && !HONOR_NANS (TYPE_MODE (type)) && !HONOR_SIGNED_ZEROS (TYPE_MODE (type)) && real_zerop (oelast->op))) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found * 0, removing all other ops\n"); reassociate_stats.ops_eliminated += ops->length () - 1; ops->truncate (1); ops->quick_push (oelast); return; } } else if (integer_onep (oelast->op) || (FLOAT_TYPE_P (type) && !HONOR_SNANS (TYPE_MODE (type)) && real_onep (oelast->op))) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found * 1, removing\n"); ops->pop (); reassociate_stats.ops_eliminated++; return; } } break; case BIT_XOR_EXPR: case PLUS_EXPR: case MINUS_EXPR: if (integer_zerop (oelast->op) || (FLOAT_TYPE_P (type) && (opcode == PLUS_EXPR || opcode == MINUS_EXPR) && fold_real_zero_addition_p (type, oelast->op, opcode == MINUS_EXPR))) { if (ops->length () != 1) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Found [|^+] 0, removing\n"); ops->pop (); reassociate_stats.ops_eliminated++; return; } } break; default: break; } } } static void linearize_expr_tree (vec *, gimple, bool, bool); /* Structure for tracking and counting operands. */ typedef struct oecount_s { int cnt; int id; enum tree_code oecode; tree op; } oecount; /* The heap for the oecount hashtable and the sorted list of operands. */ static vec cvec; /* Oecount hashtable helpers. */ struct oecount_hasher : typed_noop_remove { /* Note that this hash table stores integers, not pointers. So, observe the casting in the member functions. */ typedef void value_type; typedef void compare_type; static inline hashval_t hash (const value_type *); static inline bool equal (const value_type *, const compare_type *); }; /* Hash function for oecount. */ inline hashval_t oecount_hasher::hash (const value_type *p) { const oecount *c = &cvec[(size_t)p - 42]; return htab_hash_pointer (c->op) ^ (hashval_t)c->oecode; } /* Comparison function for oecount. */ inline bool oecount_hasher::equal (const value_type *p1, const compare_type *p2) { const oecount *c1 = &cvec[(size_t)p1 - 42]; const oecount *c2 = &cvec[(size_t)p2 - 42]; return (c1->oecode == c2->oecode && c1->op == c2->op); } /* Comparison function for qsort sorting oecount elements by count. */ static int oecount_cmp (const void *p1, const void *p2) { const oecount *c1 = (const oecount *)p1; const oecount *c2 = (const oecount *)p2; if (c1->cnt != c2->cnt) return c1->cnt - c2->cnt; else /* If counts are identical, use unique IDs to stabilize qsort. */ return c1->id - c2->id; } /* Return TRUE iff STMT represents a builtin call that raises OP to some exponent. */ static bool stmt_is_power_of_op (gimple stmt, tree op) { tree fndecl; if (!is_gimple_call (stmt)) return false; fndecl = gimple_call_fndecl (stmt); if (!fndecl || DECL_BUILT_IN_CLASS (fndecl) != BUILT_IN_NORMAL) return false; switch (DECL_FUNCTION_CODE (gimple_call_fndecl (stmt))) { CASE_FLT_FN (BUILT_IN_POW): CASE_FLT_FN (BUILT_IN_POWI): return (operand_equal_p (gimple_call_arg (stmt, 0), op, 0)); default: return false; } } /* Given STMT which is a __builtin_pow* call, decrement its exponent in place and return the result. Assumes that stmt_is_power_of_op was previously called for STMT and returned TRUE. */ static HOST_WIDE_INT decrement_power (gimple stmt) { REAL_VALUE_TYPE c, cint; HOST_WIDE_INT power; tree arg1; switch (DECL_FUNCTION_CODE (gimple_call_fndecl (stmt))) { CASE_FLT_FN (BUILT_IN_POW): arg1 = gimple_call_arg (stmt, 1); c = TREE_REAL_CST (arg1); power = real_to_integer (&c) - 1; real_from_integer (&cint, VOIDmode, power, 0, 0); gimple_call_set_arg (stmt, 1, build_real (TREE_TYPE (arg1), cint)); return power; CASE_FLT_FN (BUILT_IN_POWI): arg1 = gimple_call_arg (stmt, 1); power = TREE_INT_CST_LOW (arg1) - 1; gimple_call_set_arg (stmt, 1, build_int_cst (TREE_TYPE (arg1), power)); return power; default: gcc_unreachable (); } } /* Find the single immediate use of STMT's LHS, and replace it with OP. Remove STMT. If STMT's LHS is the same as *DEF, replace *DEF with OP as well. */ static void propagate_op_to_single_use (tree op, gimple stmt, tree *def) { tree lhs; gimple use_stmt; use_operand_p use; gimple_stmt_iterator gsi; if (is_gimple_call (stmt)) lhs = gimple_call_lhs (stmt); else lhs = gimple_assign_lhs (stmt); gcc_assert (has_single_use (lhs)); single_imm_use (lhs, &use, &use_stmt); if (lhs == *def) *def = op; SET_USE (use, op); if (TREE_CODE (op) != SSA_NAME) update_stmt (use_stmt); gsi = gsi_for_stmt (stmt); unlink_stmt_vdef (stmt); reassoc_remove_stmt (&gsi); release_defs (stmt); } /* Walks the linear chain with result *DEF searching for an operation with operand OP and code OPCODE removing that from the chain. *DEF is updated if there is only one operand but no operation left. */ static void zero_one_operation (tree *def, enum tree_code opcode, tree op) { gimple stmt = SSA_NAME_DEF_STMT (*def); do { tree name; if (opcode == MULT_EXPR && stmt_is_power_of_op (stmt, op)) { if (decrement_power (stmt) == 1) propagate_op_to_single_use (op, stmt, def); return; } name = gimple_assign_rhs1 (stmt); /* If this is the operation we look for and one of the operands is ours simply propagate the other operand into the stmts single use. */ if (gimple_assign_rhs_code (stmt) == opcode && (name == op || gimple_assign_rhs2 (stmt) == op)) { if (name == op) name = gimple_assign_rhs2 (stmt); propagate_op_to_single_use (name, stmt, def); return; } /* We might have a multiply of two __builtin_pow* calls, and the operand might be hiding in the rightmost one. */ if (opcode == MULT_EXPR && gimple_assign_rhs_code (stmt) == opcode && TREE_CODE (gimple_assign_rhs2 (stmt)) == SSA_NAME && has_single_use (gimple_assign_rhs2 (stmt))) { gimple stmt2 = SSA_NAME_DEF_STMT (gimple_assign_rhs2 (stmt)); if (stmt_is_power_of_op (stmt2, op)) { if (decrement_power (stmt2) == 1) propagate_op_to_single_use (op, stmt2, def); return; } } /* Continue walking the chain. */ gcc_assert (name != op && TREE_CODE (name) == SSA_NAME); stmt = SSA_NAME_DEF_STMT (name); } while (1); } /* Returns true if statement S1 dominates statement S2. Like stmt_dominates_stmt_p, but uses stmt UIDs to optimize. */ static bool reassoc_stmt_dominates_stmt_p (gimple s1, gimple s2) { basic_block bb1 = gimple_bb (s1), bb2 = gimple_bb (s2); /* If bb1 is NULL, it should be a GIMPLE_NOP def stmt of an (D) SSA_NAME. Assume it lives at the beginning of function and thus dominates everything. */ if (!bb1 || s1 == s2) return true; /* If bb2 is NULL, it doesn't dominate any stmt with a bb. */ if (!bb2) return false; if (bb1 == bb2) { /* PHIs in the same basic block are assumed to be executed all in parallel, if only one stmt is a PHI, it dominates the other stmt in the same basic block. */ if (gimple_code (s1) == GIMPLE_PHI) return true; if (gimple_code (s2) == GIMPLE_PHI) return false; gcc_assert (gimple_uid (s1) && gimple_uid (s2)); if (gimple_uid (s1) < gimple_uid (s2)) return true; if (gimple_uid (s1) > gimple_uid (s2)) return false; gimple_stmt_iterator gsi = gsi_for_stmt (s1); unsigned int uid = gimple_uid (s1); for (gsi_next (&gsi); !gsi_end_p (gsi); gsi_next (&gsi)) { gimple s = gsi_stmt (gsi); if (gimple_uid (s) != uid) break; if (s == s2) return true; } return false; } return dominated_by_p (CDI_DOMINATORS, bb2, bb1); } /* Insert STMT after INSERT_POINT. */ static void insert_stmt_after (gimple stmt, gimple insert_point) { gimple_stmt_iterator gsi; basic_block bb; if (gimple_code (insert_point) == GIMPLE_PHI) bb = gimple_bb (insert_point); else if (!stmt_ends_bb_p (insert_point)) { gsi = gsi_for_stmt (insert_point); gimple_set_uid (stmt, gimple_uid (insert_point)); gsi_insert_after (&gsi, stmt, GSI_NEW_STMT); return; } else /* We assume INSERT_POINT is a SSA_NAME_DEF_STMT of some SSA_NAME, thus if it must end a basic block, it should be a call that can throw, or some assignment that can throw. If it throws, the LHS of it will not be initialized though, so only valid places using the SSA_NAME should be dominated by the fallthru edge. */ bb = find_fallthru_edge (gimple_bb (insert_point)->succs)->dest; gsi = gsi_after_labels (bb); if (gsi_end_p (gsi)) { gimple_stmt_iterator gsi2 = gsi_last_bb (bb); gimple_set_uid (stmt, gsi_end_p (gsi2) ? 1 : gimple_uid (gsi_stmt (gsi2))); } else gimple_set_uid (stmt, gimple_uid (gsi_stmt (gsi))); gsi_insert_before (&gsi, stmt, GSI_SAME_STMT); } /* Builds one statement performing OP1 OPCODE OP2 using TMPVAR for the result. Places the statement after the definition of either OP1 or OP2. Returns the new statement. */ static gimple build_and_add_sum (tree type, tree op1, tree op2, enum tree_code opcode) { gimple op1def = NULL, op2def = NULL; gimple_stmt_iterator gsi; tree op; gimple sum; /* Create the addition statement. */ op = make_ssa_name (type, NULL); sum = gimple_build_assign_with_ops (opcode, op, op1, op2); /* Find an insertion place and insert. */ if (TREE_CODE (op1) == SSA_NAME) op1def = SSA_NAME_DEF_STMT (op1); if (TREE_CODE (op2) == SSA_NAME) op2def = SSA_NAME_DEF_STMT (op2); if ((!op1def || gimple_nop_p (op1def)) && (!op2def || gimple_nop_p (op2def))) { gsi = gsi_after_labels (single_succ (ENTRY_BLOCK_PTR_FOR_FN (cfun))); if (gsi_end_p (gsi)) { gimple_stmt_iterator gsi2 = gsi_last_bb (single_succ (ENTRY_BLOCK_PTR_FOR_FN (cfun))); gimple_set_uid (sum, gsi_end_p (gsi2) ? 1 : gimple_uid (gsi_stmt (gsi2))); } else gimple_set_uid (sum, gimple_uid (gsi_stmt (gsi))); gsi_insert_before (&gsi, sum, GSI_NEW_STMT); } else { gimple insert_point; if ((!op1def || gimple_nop_p (op1def)) || (op2def && !gimple_nop_p (op2def) && reassoc_stmt_dominates_stmt_p (op1def, op2def))) insert_point = op2def; else insert_point = op1def; insert_stmt_after (sum, insert_point); } update_stmt (sum); return sum; } /* Perform un-distribution of divisions and multiplications. A * X + B * X is transformed into (A + B) * X and A / X + B / X to (A + B) / X for real X. The algorithm is organized as follows. - First we walk the addition chain *OPS looking for summands that are defined by a multiplication or a real division. This results in the candidates bitmap with relevant indices into *OPS. - Second we build the chains of multiplications or divisions for these candidates, counting the number of occurrences of (operand, code) pairs in all of the candidates chains. - Third we sort the (operand, code) pairs by number of occurrence and process them starting with the pair with the most uses. * For each such pair we walk the candidates again to build a second candidate bitmap noting all multiplication/division chains that have at least one occurrence of (operand, code). * We build an alternate addition chain only covering these candidates with one (operand, code) operation removed from their multiplication/division chain. * The first candidate gets replaced by the alternate addition chain multiplied/divided by the operand. * All candidate chains get disabled for further processing and processing of (operand, code) pairs continues. The alternate addition chains built are re-processed by the main reassociation algorithm which allows optimizing a * x * y + b * y * x to (a + b ) * x * y in one invocation of the reassociation pass. */ static bool undistribute_ops_list (enum tree_code opcode, vec *ops, struct loop *loop) { unsigned int length = ops->length (); operand_entry_t oe1; unsigned i, j; sbitmap candidates, candidates2; unsigned nr_candidates, nr_candidates2; sbitmap_iterator sbi0; vec *subops; hash_table ctable; bool changed = false; int next_oecount_id = 0; if (length <= 1 || opcode != PLUS_EXPR) return false; /* Build a list of candidates to process. */ candidates = sbitmap_alloc (length); bitmap_clear (candidates); nr_candidates = 0; FOR_EACH_VEC_ELT (*ops, i, oe1) { enum tree_code dcode; gimple oe1def; if (TREE_CODE (oe1->op) != SSA_NAME) continue; oe1def = SSA_NAME_DEF_STMT (oe1->op); if (!is_gimple_assign (oe1def)) continue; dcode = gimple_assign_rhs_code (oe1def); if ((dcode != MULT_EXPR && dcode != RDIV_EXPR) || !is_reassociable_op (oe1def, dcode, loop)) continue; bitmap_set_bit (candidates, i); nr_candidates++; } if (nr_candidates < 2) { sbitmap_free (candidates); return false; } if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "searching for un-distribute opportunities "); print_generic_expr (dump_file, (*ops)[bitmap_first_set_bit (candidates)]->op, 0); fprintf (dump_file, " %d\n", nr_candidates); } /* Build linearized sub-operand lists and the counting table. */ cvec.create (0); ctable.create (15); /* ??? Macro arguments cannot have multi-argument template types in them. This typedef is needed to workaround that limitation. */ typedef vec vec_operand_entry_t_heap; subops = XCNEWVEC (vec_operand_entry_t_heap, ops->length ()); EXECUTE_IF_SET_IN_BITMAP (candidates, 0, i, sbi0) { gimple oedef; enum tree_code oecode; unsigned j; oedef = SSA_NAME_DEF_STMT ((*ops)[i]->op); oecode = gimple_assign_rhs_code (oedef); linearize_expr_tree (&subops[i], oedef, associative_tree_code (oecode), false); FOR_EACH_VEC_ELT (subops[i], j, oe1) { oecount c; void **slot; size_t idx; c.oecode = oecode; c.cnt = 1; c.id = next_oecount_id++; c.op = oe1->op; cvec.safe_push (c); idx = cvec.length () + 41; slot = ctable.find_slot ((void *)idx, INSERT); if (!*slot) { *slot = (void *)idx; } else { cvec.pop (); cvec[(size_t)*slot - 42].cnt++; } } } ctable.dispose (); /* Sort the counting table. */ cvec.qsort (oecount_cmp); if (dump_file && (dump_flags & TDF_DETAILS)) { oecount *c; fprintf (dump_file, "Candidates:\n"); FOR_EACH_VEC_ELT (cvec, j, c) { fprintf (dump_file, " %u %s: ", c->cnt, c->oecode == MULT_EXPR ? "*" : c->oecode == RDIV_EXPR ? "/" : "?"); print_generic_expr (dump_file, c->op, 0); fprintf (dump_file, "\n"); } } /* Process the (operand, code) pairs in order of most occurrence. */ candidates2 = sbitmap_alloc (length); while (!cvec.is_empty ()) { oecount *c = &cvec.last (); if (c->cnt < 2) break; /* Now collect the operands in the outer chain that contain the common operand in their inner chain. */ bitmap_clear (candidates2); nr_candidates2 = 0; EXECUTE_IF_SET_IN_BITMAP (candidates, 0, i, sbi0) { gimple oedef; enum tree_code oecode; unsigned j; tree op = (*ops)[i]->op; /* If we undistributed in this chain already this may be a constant. */ if (TREE_CODE (op) != SSA_NAME) continue; oedef = SSA_NAME_DEF_STMT (op); oecode = gimple_assign_rhs_code (oedef); if (oecode != c->oecode) continue; FOR_EACH_VEC_ELT (subops[i], j, oe1) { if (oe1->op == c->op) { bitmap_set_bit (candidates2, i); ++nr_candidates2; break; } } } if (nr_candidates2 >= 2) { operand_entry_t oe1, oe2; gimple prod; int first = bitmap_first_set_bit (candidates2); /* Build the new addition chain. */ oe1 = (*ops)[first]; if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Building ("); print_generic_expr (dump_file, oe1->op, 0); } zero_one_operation (&oe1->op, c->oecode, c->op); EXECUTE_IF_SET_IN_BITMAP (candidates2, first+1, i, sbi0) { gimple sum; oe2 = (*ops)[i]; if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " + "); print_generic_expr (dump_file, oe2->op, 0); } zero_one_operation (&oe2->op, c->oecode, c->op); sum = build_and_add_sum (TREE_TYPE (oe1->op), oe1->op, oe2->op, opcode); oe2->op = build_zero_cst (TREE_TYPE (oe2->op)); oe2->rank = 0; oe1->op = gimple_get_lhs (sum); } /* Apply the multiplication/division. */ prod = build_and_add_sum (TREE_TYPE (oe1->op), oe1->op, c->op, c->oecode); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, ") %s ", c->oecode == MULT_EXPR ? "*" : "/"); print_generic_expr (dump_file, c->op, 0); fprintf (dump_file, "\n"); } /* Record it in the addition chain and disable further undistribution with this op. */ oe1->op = gimple_assign_lhs (prod); oe1->rank = get_rank (oe1->op); subops[first].release (); changed = true; } cvec.pop (); } for (i = 0; i < ops->length (); ++i) subops[i].release (); free (subops); cvec.release (); sbitmap_free (candidates); sbitmap_free (candidates2); return changed; } /* If OPCODE is BIT_IOR_EXPR or BIT_AND_EXPR and CURR is a comparison expression, examine the other OPS to see if any of them are comparisons of the same values, which we may be able to combine or eliminate. For example, we can rewrite (a < b) | (a == b) as (a <= b). */ static bool eliminate_redundant_comparison (enum tree_code opcode, vec *ops, unsigned int currindex, operand_entry_t curr) { tree op1, op2; enum tree_code lcode, rcode; gimple def1, def2; int i; operand_entry_t oe; if (opcode != BIT_IOR_EXPR && opcode != BIT_AND_EXPR) return false; /* Check that CURR is a comparison. */ if (TREE_CODE (curr->op) != SSA_NAME) return false; def1 = SSA_NAME_DEF_STMT (curr->op); if (!is_gimple_assign (def1)) return false; lcode = gimple_assign_rhs_code (def1); if (TREE_CODE_CLASS (lcode) != tcc_comparison) return false; op1 = gimple_assign_rhs1 (def1); op2 = gimple_assign_rhs2 (def1); /* Now look for a similar comparison in the remaining OPS. */ for (i = currindex + 1; ops->iterate (i, &oe); i++) { tree t; if (TREE_CODE (oe->op) != SSA_NAME) continue; def2 = SSA_NAME_DEF_STMT (oe->op); if (!is_gimple_assign (def2)) continue; rcode = gimple_assign_rhs_code (def2); if (TREE_CODE_CLASS (rcode) != tcc_comparison) continue; /* If we got here, we have a match. See if we can combine the two comparisons. */ if (opcode == BIT_IOR_EXPR) t = maybe_fold_or_comparisons (lcode, op1, op2, rcode, gimple_assign_rhs1 (def2), gimple_assign_rhs2 (def2)); else t = maybe_fold_and_comparisons (lcode, op1, op2, rcode, gimple_assign_rhs1 (def2), gimple_assign_rhs2 (def2)); if (!t) continue; /* maybe_fold_and_comparisons and maybe_fold_or_comparisons always give us a boolean_type_node value back. If the original BIT_AND_EXPR or BIT_IOR_EXPR was of a wider integer type, we need to convert. */ if (!useless_type_conversion_p (TREE_TYPE (curr->op), TREE_TYPE (t))) t = fold_convert (TREE_TYPE (curr->op), t); if (TREE_CODE (t) != INTEGER_CST && !operand_equal_p (t, curr->op, 0)) { enum tree_code subcode; tree newop1, newop2; if (!COMPARISON_CLASS_P (t)) continue; extract_ops_from_tree (t, &subcode, &newop1, &newop2); STRIP_USELESS_TYPE_CONVERSION (newop1); STRIP_USELESS_TYPE_CONVERSION (newop2); if (!is_gimple_val (newop1) || !is_gimple_val (newop2)) continue; } if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Equivalence: "); print_generic_expr (dump_file, curr->op, 0); fprintf (dump_file, " %s ", op_symbol_code (opcode)); print_generic_expr (dump_file, oe->op, 0); fprintf (dump_file, " -> "); print_generic_expr (dump_file, t, 0); fprintf (dump_file, "\n"); } /* Now we can delete oe, as it has been subsumed by the new combined expression t. */ ops->ordered_remove (i); reassociate_stats.ops_eliminated ++; /* If t is the same as curr->op, we're done. Otherwise we must replace curr->op with t. Special case is if we got a constant back, in which case we add it to the end instead of in place of the current entry. */ if (TREE_CODE (t) == INTEGER_CST) { ops->ordered_remove (currindex); add_to_ops_vec (ops, t); } else if (!operand_equal_p (t, curr->op, 0)) { gimple sum; enum tree_code subcode; tree newop1; tree newop2; gcc_assert (COMPARISON_CLASS_P (t)); extract_ops_from_tree (t, &subcode, &newop1, &newop2); STRIP_USELESS_TYPE_CONVERSION (newop1); STRIP_USELESS_TYPE_CONVERSION (newop2); gcc_checking_assert (is_gimple_val (newop1) && is_gimple_val (newop2)); sum = build_and_add_sum (TREE_TYPE (t), newop1, newop2, subcode); curr->op = gimple_get_lhs (sum); } return true; } return false; } /* Perform various identities and other optimizations on the list of operand entries, stored in OPS. The tree code for the binary operation between all the operands is OPCODE. */ static void optimize_ops_list (enum tree_code opcode, vec *ops) { unsigned int length = ops->length (); unsigned int i; operand_entry_t oe; operand_entry_t oelast = NULL; bool iterate = false; if (length == 1) return; oelast = ops->last (); /* If the last two are constants, pop the constants off, merge them and try the next two. */ if (oelast->rank == 0 && is_gimple_min_invariant (oelast->op)) { operand_entry_t oelm1 = (*ops)[length - 2]; if (oelm1->rank == 0 && is_gimple_min_invariant (oelm1->op) && useless_type_conversion_p (TREE_TYPE (oelm1->op), TREE_TYPE (oelast->op))) { tree folded = fold_binary (opcode, TREE_TYPE (oelm1->op), oelm1->op, oelast->op); if (folded && is_gimple_min_invariant (folded)) { if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Merging constants\n"); ops->pop (); ops->pop (); add_to_ops_vec (ops, folded); reassociate_stats.constants_eliminated++; optimize_ops_list (opcode, ops); return; } } } eliminate_using_constants (opcode, ops); oelast = NULL; for (i = 0; ops->iterate (i, &oe);) { bool done = false; if (eliminate_not_pairs (opcode, ops, i, oe)) return; if (eliminate_duplicate_pair (opcode, ops, &done, i, oe, oelast) || (!done && eliminate_plus_minus_pair (opcode, ops, i, oe)) || (!done && eliminate_redundant_comparison (opcode, ops, i, oe))) { if (done) return; iterate = true; oelast = NULL; continue; } oelast = oe; i++; } length = ops->length (); oelast = ops->last (); if (iterate) optimize_ops_list (opcode, ops); } /* The following functions are subroutines to optimize_range_tests and allow it to try to change a logical combination of comparisons into a range test. For example, both X == 2 || X == 5 || X == 3 || X == 4 and X >= 2 && X <= 5 are converted to (unsigned) (X - 2) <= 3 For more information see comments above fold_test_range in fold-const.c, this implementation is for GIMPLE. */ struct range_entry { tree exp; tree low; tree high; bool in_p; bool strict_overflow_p; unsigned int idx, next; }; /* This is similar to make_range in fold-const.c, but on top of GIMPLE instead of trees. If EXP is non-NULL, it should be an SSA_NAME and STMT argument is ignored, otherwise STMT argument should be a GIMPLE_COND. */ static void init_range_entry (struct range_entry *r, tree exp, gimple stmt) { int in_p; tree low, high; bool is_bool, strict_overflow_p; r->exp = NULL_TREE; r->in_p = false; r->strict_overflow_p = false; r->low = NULL_TREE; r->high = NULL_TREE; if (exp != NULL_TREE && (TREE_CODE (exp) != SSA_NAME || !INTEGRAL_TYPE_P (TREE_TYPE (exp)))) return; /* Start with simply saying "EXP != 0" and then look at the code of EXP and see if we can refine the range. Some of the cases below may not happen, but it doesn't seem worth worrying about this. We "continue" the outer loop when we've changed something; otherwise we "break" the switch, which will "break" the while. */ low = exp ? build_int_cst (TREE_TYPE (exp), 0) : boolean_false_node; high = low; in_p = 0; strict_overflow_p = false; is_bool = false; if (exp == NULL_TREE) is_bool = true; else if (TYPE_PRECISION (TREE_TYPE (exp)) == 1) { if (TYPE_UNSIGNED (TREE_TYPE (exp))) is_bool = true; else return; } else if (TREE_CODE (TREE_TYPE (exp)) == BOOLEAN_TYPE) is_bool = true; while (1) { enum tree_code code; tree arg0, arg1, exp_type; tree nexp; location_t loc; if (exp != NULL_TREE) { if (TREE_CODE (exp) != SSA_NAME || SSA_NAME_OCCURS_IN_ABNORMAL_PHI (exp)) break; stmt = SSA_NAME_DEF_STMT (exp); if (!is_gimple_assign (stmt)) break; code = gimple_assign_rhs_code (stmt); arg0 = gimple_assign_rhs1 (stmt); arg1 = gimple_assign_rhs2 (stmt); exp_type = TREE_TYPE (exp); } else { code = gimple_cond_code (stmt); arg0 = gimple_cond_lhs (stmt); arg1 = gimple_cond_rhs (stmt); exp_type = boolean_type_node; } if (TREE_CODE (arg0) != SSA_NAME) break; loc = gimple_location (stmt); switch (code) { case BIT_NOT_EXPR: if (TREE_CODE (TREE_TYPE (exp)) == BOOLEAN_TYPE /* Ensure the range is either +[-,0], +[0,0], -[-,0], -[0,0] or +[1,-], +[1,1], -[1,-] or -[1,1]. If it is e.g. +[-,-] or -[-,-] or similar expression of unconditional true or false, it should not be negated. */ && ((high && integer_zerop (high)) || (low && integer_onep (low)))) { in_p = !in_p; exp = arg0; continue; } break; case SSA_NAME: exp = arg0; continue; CASE_CONVERT: if (is_bool) goto do_default; if (TYPE_PRECISION (TREE_TYPE (arg0)) == 1) { if (TYPE_UNSIGNED (TREE_TYPE (arg0))) is_bool = true; else return; } else if (TREE_CODE (TREE_TYPE (arg0)) == BOOLEAN_TYPE) is_bool = true; goto do_default; case EQ_EXPR: case NE_EXPR: case LT_EXPR: case LE_EXPR: case GE_EXPR: case GT_EXPR: is_bool = true; /* FALLTHRU */ default: if (!is_bool) return; do_default: nexp = make_range_step (loc, code, arg0, arg1, exp_type, &low, &high, &in_p, &strict_overflow_p); if (nexp != NULL_TREE) { exp = nexp; gcc_assert (TREE_CODE (exp) == SSA_NAME); continue; } break; } break; } if (is_bool) { r->exp = exp; r->in_p = in_p; r->low = low; r->high = high; r->strict_overflow_p = strict_overflow_p; } } /* Comparison function for qsort. Sort entries without SSA_NAME exp first, then with SSA_NAMEs sorted by increasing SSA_NAME_VERSION, and for the same SSA_NAMEs by increasing ->low and if ->low is the same, by increasing ->high. ->low == NULL_TREE means minimum, ->high == NULL_TREE maximum. */ static int range_entry_cmp (const void *a, const void *b) { const struct range_entry *p = (const struct range_entry *) a; const struct range_entry *q = (const struct range_entry *) b; if (p->exp != NULL_TREE && TREE_CODE (p->exp) == SSA_NAME) { if (q->exp != NULL_TREE && TREE_CODE (q->exp) == SSA_NAME) { /* Group range_entries for the same SSA_NAME together. */ if (SSA_NAME_VERSION (p->exp) < SSA_NAME_VERSION (q->exp)) return -1; else if (SSA_NAME_VERSION (p->exp) > SSA_NAME_VERSION (q->exp)) return 1; /* If ->low is different, NULL low goes first, then by ascending low. */ if (p->low != NULL_TREE) { if (q->low != NULL_TREE) { tree tem = fold_binary (LT_EXPR, boolean_type_node, p->low, q->low); if (tem && integer_onep (tem)) return -1; tem = fold_binary (GT_EXPR, boolean_type_node, p->low, q->low); if (tem && integer_onep (tem)) return 1; } else return 1; } else if (q->low != NULL_TREE) return -1; /* If ->high is different, NULL high goes last, before that by ascending high. */ if (p->high != NULL_TREE) { if (q->high != NULL_TREE) { tree tem = fold_binary (LT_EXPR, boolean_type_node, p->high, q->high); if (tem && integer_onep (tem)) return -1; tem = fold_binary (GT_EXPR, boolean_type_node, p->high, q->high); if (tem && integer_onep (tem)) return 1; } else return -1; } else if (q->high != NULL_TREE) return 1; /* If both ranges are the same, sort below by ascending idx. */ } else return 1; } else if (q->exp != NULL_TREE && TREE_CODE (q->exp) == SSA_NAME) return -1; if (p->idx < q->idx) return -1; else { gcc_checking_assert (p->idx > q->idx); return 1; } } /* Helper routine of optimize_range_test. [EXP, IN_P, LOW, HIGH, STRICT_OVERFLOW_P] is a merged range for RANGE and OTHERRANGE through OTHERRANGE + COUNT - 1 ranges, OPCODE and OPS are arguments of optimize_range_tests. Return true if the range merge has been successful. If OPCODE is ERROR_MARK, this is called from within maybe_optimize_range_tests and is performing inter-bb range optimization. In that case, whether an op is BIT_AND_EXPR or BIT_IOR_EXPR is found in oe->rank. */ static bool update_range_test (struct range_entry *range, struct range_entry *otherrange, unsigned int count, enum tree_code opcode, vec *ops, tree exp, bool in_p, tree low, tree high, bool strict_overflow_p) { operand_entry_t oe = (*ops)[range->idx]; tree op = oe->op; gimple stmt = op ? SSA_NAME_DEF_STMT (op) : last_stmt (BASIC_BLOCK_FOR_FN (cfun, oe->id)); location_t loc = gimple_location (stmt); tree optype = op ? TREE_TYPE (op) : boolean_type_node; tree tem = build_range_check (loc, optype, exp, in_p, low, high); enum warn_strict_overflow_code wc = WARN_STRICT_OVERFLOW_COMPARISON; gimple_stmt_iterator gsi; if (tem == NULL_TREE) return false; if (strict_overflow_p && issue_strict_overflow_warning (wc)) warning_at (loc, OPT_Wstrict_overflow, "assuming signed overflow does not occur " "when simplifying range test"); if (dump_file && (dump_flags & TDF_DETAILS)) { struct range_entry *r; fprintf (dump_file, "Optimizing range tests "); print_generic_expr (dump_file, range->exp, 0); fprintf (dump_file, " %c[", range->in_p ? '+' : '-'); print_generic_expr (dump_file, range->low, 0); fprintf (dump_file, ", "); print_generic_expr (dump_file, range->high, 0); fprintf (dump_file, "]"); for (r = otherrange; r < otherrange + count; r++) { fprintf (dump_file, " and %c[", r->in_p ? '+' : '-'); print_generic_expr (dump_file, r->low, 0); fprintf (dump_file, ", "); print_generic_expr (dump_file, r->high, 0); fprintf (dump_file, "]"); } fprintf (dump_file, "\n into "); print_generic_expr (dump_file, tem, 0); fprintf (dump_file, "\n"); } if (opcode == BIT_IOR_EXPR || (opcode == ERROR_MARK && oe->rank == BIT_IOR_EXPR)) tem = invert_truthvalue_loc (loc, tem); tem = fold_convert_loc (loc, optype, tem); gsi = gsi_for_stmt (stmt); /* In rare cases range->exp can be equal to lhs of stmt. In that case we have to insert after the stmt rather then before it. */ if (op == range->exp) tem = force_gimple_operand_gsi (&gsi, tem, true, NULL_TREE, false, GSI_CONTINUE_LINKING); else { tem = force_gimple_operand_gsi (&gsi, tem, true, NULL_TREE, true, GSI_SAME_STMT); gsi_prev (&gsi); } for (; !gsi_end_p (gsi); gsi_prev (&gsi)) if (gimple_uid (gsi_stmt (gsi))) break; else gimple_set_uid (gsi_stmt (gsi), gimple_uid (stmt)); oe->op = tem; range->exp = exp; range->low = low; range->high = high; range->in_p = in_p; range->strict_overflow_p = false; for (range = otherrange; range < otherrange + count; range++) { oe = (*ops)[range->idx]; /* Now change all the other range test immediate uses, so that those tests will be optimized away. */ if (opcode == ERROR_MARK) { if (oe->op) oe->op = build_int_cst (TREE_TYPE (oe->op), oe->rank == BIT_IOR_EXPR ? 0 : 1); else oe->op = (oe->rank == BIT_IOR_EXPR ? boolean_false_node : boolean_true_node); } else oe->op = error_mark_node; range->exp = NULL_TREE; } return true; } /* Optimize X == CST1 || X == CST2 if popcount (CST1 ^ CST2) == 1 into (X & ~(CST1 ^ CST2)) == (CST1 & ~(CST1 ^ CST2)). Similarly for ranges. E.g. X != 2 && X != 3 && X != 10 && X != 11 will be transformed by the previous optimization into !((X - 2U) <= 1U || (X - 10U) <= 1U) and this loop can transform that into !(((X & ~8) - 2U) <= 1U). */ static bool optimize_range_tests_xor (enum tree_code opcode, tree type, tree lowi, tree lowj, tree highi, tree highj, vec *ops, struct range_entry *rangei, struct range_entry *rangej) { tree lowxor, highxor, tem, exp; /* Check highi ^ lowi == highj ^ lowj and popcount (highi ^ lowi) == 1. */ lowxor = fold_binary (BIT_XOR_EXPR, type, lowi, lowj); if (lowxor == NULL_TREE || TREE_CODE (lowxor) != INTEGER_CST) return false; if (!integer_pow2p (lowxor)) return false; highxor = fold_binary (BIT_XOR_EXPR, type, highi, highj); if (!tree_int_cst_equal (lowxor, highxor)) return false; tem = fold_build1 (BIT_NOT_EXPR, type, lowxor); exp = fold_build2 (BIT_AND_EXPR, type, rangei->exp, tem); lowj = fold_build2 (BIT_AND_EXPR, type, lowi, tem); highj = fold_build2 (BIT_AND_EXPR, type, highi, tem); if (update_range_test (rangei, rangej, 1, opcode, ops, exp, rangei->in_p, lowj, highj, rangei->strict_overflow_p || rangej->strict_overflow_p)) return true; return false; } /* Optimize X == CST1 || X == CST2 if popcount (CST2 - CST1) == 1 into ((X - CST1) & ~(CST2 - CST1)) == 0. Similarly for ranges. E.g. X == 43 || X == 76 || X == 44 || X == 78 || X == 77 || X == 46 || X == 75 || X == 45 will be transformed by the previous optimization into (X - 43U) <= 3U || (X - 75U) <= 3U and this loop can transform that into ((X - 43U) & ~(75U - 43U)) <= 3U. */ static bool optimize_range_tests_diff (enum tree_code opcode, tree type, tree lowi, tree lowj, tree highi, tree highj, vec *ops, struct range_entry *rangei, struct range_entry *rangej) { tree tem1, tem2, mask; /* Check highi - lowi == highj - lowj. */ tem1 = fold_binary (MINUS_EXPR, type, highi, lowi); if (tem1 == NULL_TREE || TREE_CODE (tem1) != INTEGER_CST) return false; tem2 = fold_binary (MINUS_EXPR, type, highj, lowj); if (!tree_int_cst_equal (tem1, tem2)) return false; /* Check popcount (lowj - lowi) == 1. */ tem1 = fold_binary (MINUS_EXPR, type, lowj, lowi); if (tem1 == NULL_TREE || TREE_CODE (tem1) != INTEGER_CST) return false; if (!integer_pow2p (tem1)) return false; mask = fold_build1 (BIT_NOT_EXPR, type, tem1); tem1 = fold_binary (MINUS_EXPR, type, rangei->exp, lowi); tem1 = fold_build2 (BIT_AND_EXPR, type, tem1, mask); lowj = build_int_cst (type, 0); if (update_range_test (rangei, rangej, 1, opcode, ops, tem1, rangei->in_p, lowj, tem2, rangei->strict_overflow_p || rangej->strict_overflow_p)) return true; return false; } /* It does some common checks for function optimize_range_tests_xor and optimize_range_tests_diff. If OPTIMIZE_XOR is TRUE, it calls optimize_range_tests_xor. Else it calls optimize_range_tests_diff. */ static bool optimize_range_tests_1 (enum tree_code opcode, int first, int length, bool optimize_xor, vec *ops, struct range_entry *ranges) { int i, j; bool any_changes = false; for (i = first; i < length; i++) { tree lowi, highi, lowj, highj, type, tem; if (ranges[i].exp == NULL_TREE || ranges[i].in_p) continue; type = TREE_TYPE (ranges[i].exp); if (!INTEGRAL_TYPE_P (type)) continue; lowi = ranges[i].low; if (lowi == NULL_TREE) lowi = TYPE_MIN_VALUE (type); highi = ranges[i].high; if (highi == NULL_TREE) continue; for (j = i + 1; j < length && j < i + 64; j++) { bool changes; if (ranges[i].exp != ranges[j].exp || ranges[j].in_p) continue; lowj = ranges[j].low; if (lowj == NULL_TREE) continue; highj = ranges[j].high; if (highj == NULL_TREE) highj = TYPE_MAX_VALUE (type); /* Check lowj > highi. */ tem = fold_binary (GT_EXPR, boolean_type_node, lowj, highi); if (tem == NULL_TREE || !integer_onep (tem)) continue; if (optimize_xor) changes = optimize_range_tests_xor (opcode, type, lowi, lowj, highi, highj, ops, ranges + i, ranges + j); else changes = optimize_range_tests_diff (opcode, type, lowi, lowj, highi, highj, ops, ranges + i, ranges + j); if (changes) { any_changes = true; break; } } } return any_changes; } /* Optimize range tests, similarly how fold_range_test optimizes it on trees. The tree code for the binary operation between all the operands is OPCODE. If OPCODE is ERROR_MARK, optimize_range_tests is called from within maybe_optimize_range_tests for inter-bb range optimization. In that case if oe->op is NULL, oe->id is bb->index whose GIMPLE_COND is && or ||ed into the test, and oe->rank says the actual opcode. */ static bool optimize_range_tests (enum tree_code opcode, vec *ops) { unsigned int length = ops->length (), i, j, first; operand_entry_t oe; struct range_entry *ranges; bool any_changes = false; if (length == 1) return false; ranges = XNEWVEC (struct range_entry, length); for (i = 0; i < length; i++) { oe = (*ops)[i]; ranges[i].idx = i; init_range_entry (ranges + i, oe->op, oe->op ? NULL : last_stmt (BASIC_BLOCK_FOR_FN (cfun, oe->id))); /* For | invert it now, we will invert it again before emitting the optimized expression. */ if (opcode == BIT_IOR_EXPR || (opcode == ERROR_MARK && oe->rank == BIT_IOR_EXPR)) ranges[i].in_p = !ranges[i].in_p; } qsort (ranges, length, sizeof (*ranges), range_entry_cmp); for (i = 0; i < length; i++) if (ranges[i].exp != NULL_TREE && TREE_CODE (ranges[i].exp) == SSA_NAME) break; /* Try to merge ranges. */ for (first = i; i < length; i++) { tree low = ranges[i].low; tree high = ranges[i].high; int in_p = ranges[i].in_p; bool strict_overflow_p = ranges[i].strict_overflow_p; int update_fail_count = 0; for (j = i + 1; j < length; j++) { if (ranges[i].exp != ranges[j].exp) break; if (!merge_ranges (&in_p, &low, &high, in_p, low, high, ranges[j].in_p, ranges[j].low, ranges[j].high)) break; strict_overflow_p |= ranges[j].strict_overflow_p; } if (j == i + 1) continue; if (update_range_test (ranges + i, ranges + i + 1, j - i - 1, opcode, ops, ranges[i].exp, in_p, low, high, strict_overflow_p)) { i = j - 1; any_changes = true; } /* Avoid quadratic complexity if all merge_ranges calls would succeed, while update_range_test would fail. */ else if (update_fail_count == 64) i = j - 1; else ++update_fail_count; } any_changes |= optimize_range_tests_1 (opcode, first, length, true, ops, ranges); if (BRANCH_COST (optimize_function_for_speed_p (cfun), false) >= 2) any_changes |= optimize_range_tests_1 (opcode, first, length, false, ops, ranges); if (any_changes && opcode != ERROR_MARK) { j = 0; FOR_EACH_VEC_ELT (*ops, i, oe) { if (oe->op == error_mark_node) continue; else if (i != j) (*ops)[j] = oe; j++; } ops->truncate (j); } XDELETEVEC (ranges); return any_changes; } /* Return true if STMT is a cast like: : ... _123 = (int) _234; : # _345 = PHI <_123(N), 1(...), 1(...)> where _234 has bool type, _123 has single use and bb N has a single successor M. This is commonly used in the last block of a range test. */ static bool final_range_test_p (gimple stmt) { basic_block bb, rhs_bb; edge e; tree lhs, rhs; use_operand_p use_p; gimple use_stmt; if (!gimple_assign_cast_p (stmt)) return false; bb = gimple_bb (stmt); if (!single_succ_p (bb)) return false; e = single_succ_edge (bb); if (e->flags & EDGE_COMPLEX) return false; lhs = gimple_assign_lhs (stmt); rhs = gimple_assign_rhs1 (stmt); if (!INTEGRAL_TYPE_P (TREE_TYPE (lhs)) || TREE_CODE (rhs) != SSA_NAME || TREE_CODE (TREE_TYPE (rhs)) != BOOLEAN_TYPE) return false; /* Test whether lhs is consumed only by a PHI in the only successor bb. */ if (!single_imm_use (lhs, &use_p, &use_stmt)) return false; if (gimple_code (use_stmt) != GIMPLE_PHI || gimple_bb (use_stmt) != e->dest) return false; /* And that the rhs is defined in the same loop. */ rhs_bb = gimple_bb (SSA_NAME_DEF_STMT (rhs)); if (rhs_bb == NULL || !flow_bb_inside_loop_p (loop_containing_stmt (stmt), rhs_bb)) return false; return true; } /* Return true if BB is suitable basic block for inter-bb range test optimization. If BACKWARD is true, BB should be the only predecessor of TEST_BB, and *OTHER_BB is either NULL and filled by the routine, or compared with to find a common basic block to which all conditions branch to if true resp. false. If BACKWARD is false, TEST_BB should be the only predecessor of BB. */ static bool suitable_cond_bb (basic_block bb, basic_block test_bb, basic_block *other_bb, bool backward) { edge_iterator ei, ei2; edge e, e2; gimple stmt; gimple_stmt_iterator gsi; bool other_edge_seen = false; bool is_cond; if (test_bb == bb) return false; /* Check last stmt first. */ stmt = last_stmt (bb); if (stmt == NULL || (gimple_code (stmt) != GIMPLE_COND && (backward || !final_range_test_p (stmt))) || gimple_visited_p (stmt) || stmt_could_throw_p (stmt) || *other_bb == bb) return false; is_cond = gimple_code (stmt) == GIMPLE_COND; if (is_cond) { /* If last stmt is GIMPLE_COND, verify that one of the succ edges goes to the next bb (if BACKWARD, it is TEST_BB), and the other to *OTHER_BB (if not set yet, try to find it out). */ if (EDGE_COUNT (bb->succs) != 2) return false; FOR_EACH_EDGE (e, ei, bb->succs) { if (!(e->flags & (EDGE_TRUE_VALUE | EDGE_FALSE_VALUE))) return false; if (e->dest == test_bb) { if (backward) continue; else return false; } if (e->dest == bb) return false; if (*other_bb == NULL) { FOR_EACH_EDGE (e2, ei2, test_bb->succs) if (!(e2->flags & (EDGE_TRUE_VALUE | EDGE_FALSE_VALUE))) return false; else if (e->dest == e2->dest) *other_bb = e->dest; if (*other_bb == NULL) return false; } if (e->dest == *other_bb) other_edge_seen = true; else if (backward) return false; } if (*other_bb == NULL || !other_edge_seen) return false; } else if (single_succ (bb) != *other_bb) return false; /* Now check all PHIs of *OTHER_BB. */ e = find_edge (bb, *other_bb); e2 = find_edge (test_bb, *other_bb); for (gsi = gsi_start_phis (e->dest); !gsi_end_p (gsi); gsi_next (&gsi)) { gimple phi = gsi_stmt (gsi); /* If both BB and TEST_BB end with GIMPLE_COND, all PHI arguments corresponding to BB and TEST_BB predecessor must be the same. */ if (!operand_equal_p (gimple_phi_arg_def (phi, e->dest_idx), gimple_phi_arg_def (phi, e2->dest_idx), 0)) { /* Otherwise, if one of the blocks doesn't end with GIMPLE_COND, one of the PHIs should have the lhs of the last stmt in that block as PHI arg and that PHI should have 0 or 1 corresponding to it in all other range test basic blocks considered. */ if (!is_cond) { if (gimple_phi_arg_def (phi, e->dest_idx) == gimple_assign_lhs (stmt) && (integer_zerop (gimple_phi_arg_def (phi, e2->dest_idx)) || integer_onep (gimple_phi_arg_def (phi, e2->dest_idx)))) continue; } else { gimple test_last = last_stmt (test_bb); if (gimple_code (test_last) != GIMPLE_COND && gimple_phi_arg_def (phi, e2->dest_idx) == gimple_assign_lhs (test_last) && (integer_zerop (gimple_phi_arg_def (phi, e->dest_idx)) || integer_onep (gimple_phi_arg_def (phi, e->dest_idx)))) continue; } return false; } } return true; } /* Return true if BB doesn't have side-effects that would disallow range test optimization, all SSA_NAMEs set in the bb are consumed in the bb and there are no PHIs. */ static bool no_side_effect_bb (basic_block bb) { gimple_stmt_iterator gsi; gimple last; if (!gimple_seq_empty_p (phi_nodes (bb))) return false; last = last_stmt (bb); for (gsi = gsi_start_bb (bb); !gsi_end_p (gsi); gsi_next (&gsi)) { gimple stmt = gsi_stmt (gsi); tree lhs; imm_use_iterator imm_iter; use_operand_p use_p; if (is_gimple_debug (stmt)) continue; if (gimple_has_side_effects (stmt)) return false; if (stmt == last) return true; if (!is_gimple_assign (stmt)) return false; lhs = gimple_assign_lhs (stmt); if (TREE_CODE (lhs) != SSA_NAME) return false; if (gimple_assign_rhs_could_trap_p (stmt)) return false; FOR_EACH_IMM_USE_FAST (use_p, imm_iter, lhs) { gimple use_stmt = USE_STMT (use_p); if (is_gimple_debug (use_stmt)) continue; if (gimple_bb (use_stmt) != bb) return false; } } return false; } /* If VAR is set by CODE (BIT_{AND,IOR}_EXPR) which is reassociable, return true and fill in *OPS recursively. */ static bool get_ops (tree var, enum tree_code code, vec *ops, struct loop *loop) { gimple stmt = SSA_NAME_DEF_STMT (var); tree rhs[2]; int i; if (!is_reassociable_op (stmt, code, loop)) return false; rhs[0] = gimple_assign_rhs1 (stmt); rhs[1] = gimple_assign_rhs2 (stmt); gimple_set_visited (stmt, true); for (i = 0; i < 2; i++) if (TREE_CODE (rhs[i]) == SSA_NAME && !get_ops (rhs[i], code, ops, loop) && has_single_use (rhs[i])) { operand_entry_t oe = (operand_entry_t) pool_alloc (operand_entry_pool); oe->op = rhs[i]; oe->rank = code; oe->id = 0; oe->count = 1; ops->safe_push (oe); } return true; } /* Find the ops that were added by get_ops starting from VAR, see if they were changed during update_range_test and if yes, create new stmts. */ static tree update_ops (tree var, enum tree_code code, vec ops, unsigned int *pidx, struct loop *loop) { gimple stmt = SSA_NAME_DEF_STMT (var); tree rhs[4]; int i; if (!is_reassociable_op (stmt, code, loop)) return NULL; rhs[0] = gimple_assign_rhs1 (stmt); rhs[1] = gimple_assign_rhs2 (stmt); rhs[2] = rhs[0]; rhs[3] = rhs[1]; for (i = 0; i < 2; i++) if (TREE_CODE (rhs[i]) == SSA_NAME) { rhs[2 + i] = update_ops (rhs[i], code, ops, pidx, loop); if (rhs[2 + i] == NULL_TREE) { if (has_single_use (rhs[i])) rhs[2 + i] = ops[(*pidx)++]->op; else rhs[2 + i] = rhs[i]; } } if ((rhs[2] != rhs[0] || rhs[3] != rhs[1]) && (rhs[2] != rhs[1] || rhs[3] != rhs[0])) { gimple_stmt_iterator gsi = gsi_for_stmt (stmt); var = make_ssa_name (TREE_TYPE (var), NULL); gimple g = gimple_build_assign_with_ops (gimple_assign_rhs_code (stmt), var, rhs[2], rhs[3]); gimple_set_uid (g, gimple_uid (stmt)); gimple_set_visited (g, true); gsi_insert_before (&gsi, g, GSI_SAME_STMT); } return var; } /* Structure to track the initial value passed to get_ops and the range in the ops vector for each basic block. */ struct inter_bb_range_test_entry { tree op; unsigned int first_idx, last_idx; }; /* Inter-bb range test optimization. */ static void maybe_optimize_range_tests (gimple stmt) { basic_block first_bb = gimple_bb (stmt); basic_block last_bb = first_bb; basic_block other_bb = NULL; basic_block bb; edge_iterator ei; edge e; auto_vec ops; auto_vec bbinfo; bool any_changes = false; /* Consider only basic blocks that end with GIMPLE_COND or a cast statement satisfying final_range_test_p. All but the last bb in the first_bb .. last_bb range should end with GIMPLE_COND. */ if (gimple_code (stmt) == GIMPLE_COND) { if (EDGE_COUNT (first_bb->succs) != 2) return; } else if (final_range_test_p (stmt)) other_bb = single_succ (first_bb); else return; if (stmt_could_throw_p (stmt)) return; /* As relative ordering of post-dominator sons isn't fixed, maybe_optimize_range_tests can be called first on any bb in the range we want to optimize. So, start searching backwards, if first_bb can be set to a predecessor. */ while (single_pred_p (first_bb)) { basic_block pred_bb = single_pred (first_bb); if (!suitable_cond_bb (pred_bb, first_bb, &other_bb, true)) break; if (!no_side_effect_bb (first_bb)) break; first_bb = pred_bb; } /* If first_bb is last_bb, other_bb hasn't been computed yet. Before starting forward search in last_bb successors, find out the other_bb. */ if (first_bb == last_bb) { other_bb = NULL; /* As non-GIMPLE_COND last stmt always terminates the range, if forward search didn't discover anything, just give up. */ if (gimple_code (stmt) != GIMPLE_COND) return; /* Look at both successors. Either it ends with a GIMPLE_COND and satisfies suitable_cond_bb, or ends with a cast and other_bb is that cast's successor. */ FOR_EACH_EDGE (e, ei, first_bb->succs) if (!(e->flags & (EDGE_TRUE_VALUE | EDGE_FALSE_VALUE)) || e->dest == first_bb) return; else if (single_pred_p (e->dest)) { stmt = last_stmt (e->dest); if (stmt && gimple_code (stmt) == GIMPLE_COND && EDGE_COUNT (e->dest->succs) == 2) { if (suitable_cond_bb (first_bb, e->dest, &other_bb, true)) break; else other_bb = NULL; } else if (stmt && final_range_test_p (stmt) && find_edge (first_bb, single_succ (e->dest))) { other_bb = single_succ (e->dest); if (other_bb == first_bb) other_bb = NULL; } } if (other_bb == NULL) return; } /* Now do the forward search, moving last_bb to successor bbs that aren't other_bb. */ while (EDGE_COUNT (last_bb->succs) == 2) { FOR_EACH_EDGE (e, ei, last_bb->succs) if (e->dest != other_bb) break; if (e == NULL) break; if (!single_pred_p (e->dest)) break; if (!suitable_cond_bb (e->dest, last_bb, &other_bb, false)) break; if (!no_side_effect_bb (e->dest)) break; last_bb = e->dest; } if (first_bb == last_bb) return; /* Here basic blocks first_bb through last_bb's predecessor end with GIMPLE_COND, all of them have one of the edges to other_bb and another to another block in the range, all blocks except first_bb don't have side-effects and last_bb ends with either GIMPLE_COND, or cast satisfying final_range_test_p. */ for (bb = last_bb; ; bb = single_pred (bb)) { enum tree_code code; tree lhs, rhs; inter_bb_range_test_entry bb_ent; bb_ent.op = NULL_TREE; bb_ent.first_idx = ops.length (); bb_ent.last_idx = bb_ent.first_idx; e = find_edge (bb, other_bb); stmt = last_stmt (bb); gimple_set_visited (stmt, true); if (gimple_code (stmt) != GIMPLE_COND) { use_operand_p use_p; gimple phi; edge e2; unsigned int d; lhs = gimple_assign_lhs (stmt); rhs = gimple_assign_rhs1 (stmt); gcc_assert (bb == last_bb); /* stmt is _123 = (int) _234; followed by: : # _345 = PHI <_123(N), 1(...), 1(...)> or 0 instead of 1. If it is 0, the _234 range test is anded together with all the other range tests, if it is 1, it is ored with them. */ single_imm_use (lhs, &use_p, &phi); gcc_assert (gimple_code (phi) == GIMPLE_PHI); e2 = find_edge (first_bb, other_bb); d = e2->dest_idx; gcc_assert (gimple_phi_arg_def (phi, e->dest_idx) == lhs); if (integer_zerop (gimple_phi_arg_def (phi, d))) code = BIT_AND_EXPR; else { gcc_checking_assert (integer_onep (gimple_phi_arg_def (phi, d))); code = BIT_IOR_EXPR; } /* If _234 SSA_NAME_DEF_STMT is _234 = _567 | _789; (or &, corresponding to 1/0 in the phi arguments, push into ops the individual range test arguments of the bitwise or resp. and, recursively. */ if (!get_ops (rhs, code, &ops, loop_containing_stmt (stmt)) && has_single_use (rhs)) { /* Otherwise, push the _234 range test itself. */ operand_entry_t oe = (operand_entry_t) pool_alloc (operand_entry_pool); oe->op = rhs; oe->rank = code; oe->id = 0; oe->count = 1; ops.safe_push (oe); bb_ent.last_idx++; } else bb_ent.last_idx = ops.length (); bb_ent.op = rhs; bbinfo.safe_push (bb_ent); continue; } /* Otherwise stmt is GIMPLE_COND. */ code = gimple_cond_code (stmt); lhs = gimple_cond_lhs (stmt); rhs = gimple_cond_rhs (stmt); if (TREE_CODE (lhs) == SSA_NAME && INTEGRAL_TYPE_P (TREE_TYPE (lhs)) && ((code != EQ_EXPR && code != NE_EXPR) || rhs != boolean_false_node /* Either push into ops the individual bitwise or resp. and operands, depending on which edge is other_bb. */ || !get_ops (lhs, (((e->flags & EDGE_TRUE_VALUE) == 0) ^ (code == EQ_EXPR)) ? BIT_AND_EXPR : BIT_IOR_EXPR, &ops, loop_containing_stmt (stmt)))) { /* Or push the GIMPLE_COND stmt itself. */ operand_entry_t oe = (operand_entry_t) pool_alloc (operand_entry_pool); oe->op = NULL; oe->rank = (e->flags & EDGE_TRUE_VALUE) ? BIT_IOR_EXPR : BIT_AND_EXPR; /* oe->op = NULL signs that there is no SSA_NAME for the range test, and oe->id instead is the basic block number, at which's end the GIMPLE_COND is. */ oe->id = bb->index; oe->count = 1; ops.safe_push (oe); bb_ent.op = NULL; bb_ent.last_idx++; } else if (ops.length () > bb_ent.first_idx) { bb_ent.op = lhs; bb_ent.last_idx = ops.length (); } bbinfo.safe_push (bb_ent); if (bb == first_bb) break; } if (ops.length () > 1) any_changes = optimize_range_tests (ERROR_MARK, &ops); if (any_changes) { unsigned int idx; /* update_ops relies on has_single_use predicates returning the same values as it did during get_ops earlier. Additionally it never removes statements, only adds new ones and it should walk from the single imm use and check the predicate already before making those changes. On the other side, the handling of GIMPLE_COND directly can turn previously multiply used SSA_NAMEs into single use SSA_NAMEs, so it needs to be done in a separate loop afterwards. */ for (bb = last_bb, idx = 0; ; bb = single_pred (bb), idx++) { if (bbinfo[idx].first_idx < bbinfo[idx].last_idx && bbinfo[idx].op != NULL_TREE) { tree new_op; stmt = last_stmt (bb); new_op = update_ops (bbinfo[idx].op, (enum tree_code) ops[bbinfo[idx].first_idx]->rank, ops, &bbinfo[idx].first_idx, loop_containing_stmt (stmt)); if (new_op == NULL_TREE) { gcc_assert (bb == last_bb); new_op = ops[bbinfo[idx].first_idx++]->op; } if (bbinfo[idx].op != new_op) { imm_use_iterator iter; use_operand_p use_p; gimple use_stmt, cast_stmt = NULL; FOR_EACH_IMM_USE_STMT (use_stmt, iter, bbinfo[idx].op) if (is_gimple_debug (use_stmt)) continue; else if (gimple_code (use_stmt) == GIMPLE_COND || gimple_code (use_stmt) == GIMPLE_PHI) FOR_EACH_IMM_USE_ON_STMT (use_p, iter) SET_USE (use_p, new_op); else if (gimple_assign_cast_p (use_stmt)) cast_stmt = use_stmt; else gcc_unreachable (); if (cast_stmt) { gcc_assert (bb == last_bb); tree lhs = gimple_assign_lhs (cast_stmt); tree new_lhs = make_ssa_name (TREE_TYPE (lhs), NULL); enum tree_code rhs_code = gimple_assign_rhs_code (cast_stmt); gimple g; if (is_gimple_min_invariant (new_op)) { new_op = fold_convert (TREE_TYPE (lhs), new_op); g = gimple_build_assign (new_lhs, new_op); } else g = gimple_build_assign_with_ops (rhs_code, new_lhs, new_op, NULL_TREE); gimple_stmt_iterator gsi = gsi_for_stmt (cast_stmt); gimple_set_uid (g, gimple_uid (cast_stmt)); gimple_set_visited (g, true); gsi_insert_before (&gsi, g, GSI_SAME_STMT); FOR_EACH_IMM_USE_STMT (use_stmt, iter, lhs) if (is_gimple_debug (use_stmt)) continue; else if (gimple_code (use_stmt) == GIMPLE_COND || gimple_code (use_stmt) == GIMPLE_PHI) FOR_EACH_IMM_USE_ON_STMT (use_p, iter) SET_USE (use_p, new_lhs); else gcc_unreachable (); } } } if (bb == first_bb) break; } for (bb = last_bb, idx = 0; ; bb = single_pred (bb), idx++) { if (bbinfo[idx].first_idx < bbinfo[idx].last_idx && bbinfo[idx].op == NULL_TREE && ops[bbinfo[idx].first_idx]->op != NULL_TREE) { stmt = last_stmt (bb); if (integer_zerop (ops[bbinfo[idx].first_idx]->op)) gimple_cond_make_false (stmt); else if (integer_onep (ops[bbinfo[idx].first_idx]->op)) gimple_cond_make_true (stmt); else { gimple_cond_set_code (stmt, NE_EXPR); gimple_cond_set_lhs (stmt, ops[bbinfo[idx].first_idx]->op); gimple_cond_set_rhs (stmt, boolean_false_node); } update_stmt (stmt); } if (bb == first_bb) break; } } } /* Return true if OPERAND is defined by a PHI node which uses the LHS of STMT in it's operands. This is also known as a "destructive update" operation. */ static bool is_phi_for_stmt (gimple stmt, tree operand) { gimple def_stmt; tree lhs; use_operand_p arg_p; ssa_op_iter i; if (TREE_CODE (operand) != SSA_NAME) return false; lhs = gimple_assign_lhs (stmt); def_stmt = SSA_NAME_DEF_STMT (operand); if (gimple_code (def_stmt) != GIMPLE_PHI) return false; FOR_EACH_PHI_ARG (arg_p, def_stmt, i, SSA_OP_USE) if (lhs == USE_FROM_PTR (arg_p)) return true; return false; } /* Remove def stmt of VAR if VAR has zero uses and recurse on rhs1 operand if so. */ static void remove_visited_stmt_chain (tree var) { gimple stmt; gimple_stmt_iterator gsi; while (1) { if (TREE_CODE (var) != SSA_NAME || !has_zero_uses (var)) return; stmt = SSA_NAME_DEF_STMT (var); if (is_gimple_assign (stmt) && gimple_visited_p (stmt)) { var = gimple_assign_rhs1 (stmt); gsi = gsi_for_stmt (stmt); reassoc_remove_stmt (&gsi); release_defs (stmt); } else return; } } /* This function checks three consequtive operands in passed operands vector OPS starting from OPINDEX and swaps two operands if it is profitable for binary operation consuming OPINDEX + 1 abnd OPINDEX + 2 operands. We pair ops with the same rank if possible. The alternative we try is to see if STMT is a destructive update style statement, which is like: b = phi (a, ...) a = c + b; In that case, we want to use the destructive update form to expose the possible vectorizer sum reduction opportunity. In that case, the third operand will be the phi node. This check is not performed if STMT is null. We could, of course, try to be better as noted above, and do a lot of work to try to find these opportunities in >3 operand cases, but it is unlikely to be worth it. */ static void swap_ops_for_binary_stmt (vec ops, unsigned int opindex, gimple stmt) { operand_entry_t oe1, oe2, oe3; oe1 = ops[opindex]; oe2 = ops[opindex + 1]; oe3 = ops[opindex + 2]; if ((oe1->rank == oe2->rank && oe2->rank != oe3->rank) || (stmt && is_phi_for_stmt (stmt, oe3->op) && !is_phi_for_stmt (stmt, oe1->op) && !is_phi_for_stmt (stmt, oe2->op))) { struct operand_entry temp = *oe3; oe3->op = oe1->op; oe3->rank = oe1->rank; oe1->op = temp.op; oe1->rank= temp.rank; } else if ((oe1->rank == oe3->rank && oe2->rank != oe3->rank) || (stmt && is_phi_for_stmt (stmt, oe2->op) && !is_phi_for_stmt (stmt, oe1->op) && !is_phi_for_stmt (stmt, oe3->op))) { struct operand_entry temp = *oe2; oe2->op = oe1->op; oe2->rank = oe1->rank; oe1->op = temp.op; oe1->rank = temp.rank; } } /* If definition of RHS1 or RHS2 dominates STMT, return the later of those two definitions, otherwise return STMT. */ static inline gimple find_insert_point (gimple stmt, tree rhs1, tree rhs2) { if (TREE_CODE (rhs1) == SSA_NAME && reassoc_stmt_dominates_stmt_p (stmt, SSA_NAME_DEF_STMT (rhs1))) stmt = SSA_NAME_DEF_STMT (rhs1); if (TREE_CODE (rhs2) == SSA_NAME && reassoc_stmt_dominates_stmt_p (stmt, SSA_NAME_DEF_STMT (rhs2))) stmt = SSA_NAME_DEF_STMT (rhs2); return stmt; } /* Recursively rewrite our linearized statements so that the operators match those in OPS[OPINDEX], putting the computation in rank order. Return new lhs. */ static tree rewrite_expr_tree (gimple stmt, unsigned int opindex, vec ops, bool changed) { tree rhs1 = gimple_assign_rhs1 (stmt); tree rhs2 = gimple_assign_rhs2 (stmt); tree lhs = gimple_assign_lhs (stmt); operand_entry_t oe; /* The final recursion case for this function is that you have exactly two operations left. If we had one exactly one op in the entire list to start with, we would have never called this function, and the tail recursion rewrites them one at a time. */ if (opindex + 2 == ops.length ()) { operand_entry_t oe1, oe2; oe1 = ops[opindex]; oe2 = ops[opindex + 1]; if (rhs1 != oe1->op || rhs2 != oe2->op) { gimple_stmt_iterator gsi = gsi_for_stmt (stmt); unsigned int uid = gimple_uid (stmt); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Transforming "); print_gimple_stmt (dump_file, stmt, 0, 0); } if (changed) { gimple insert_point = find_insert_point (stmt, oe1->op, oe2->op); lhs = make_ssa_name (TREE_TYPE (lhs), NULL); stmt = gimple_build_assign_with_ops (gimple_assign_rhs_code (stmt), lhs, oe1->op, oe2->op); gimple_set_uid (stmt, uid); gimple_set_visited (stmt, true); if (insert_point == gsi_stmt (gsi)) gsi_insert_before (&gsi, stmt, GSI_SAME_STMT); else insert_stmt_after (stmt, insert_point); } else { gcc_checking_assert (find_insert_point (stmt, oe1->op, oe2->op) == stmt); gimple_assign_set_rhs1 (stmt, oe1->op); gimple_assign_set_rhs2 (stmt, oe2->op); update_stmt (stmt); } if (rhs1 != oe1->op && rhs1 != oe2->op) remove_visited_stmt_chain (rhs1); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " into "); print_gimple_stmt (dump_file, stmt, 0, 0); } } return lhs; } /* If we hit here, we should have 3 or more ops left. */ gcc_assert (opindex + 2 < ops.length ()); /* Rewrite the next operator. */ oe = ops[opindex]; /* Recurse on the LHS of the binary operator, which is guaranteed to be the non-leaf side. */ tree new_rhs1 = rewrite_expr_tree (SSA_NAME_DEF_STMT (rhs1), opindex + 1, ops, changed || oe->op != rhs2); if (oe->op != rhs2 || new_rhs1 != rhs1) { if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Transforming "); print_gimple_stmt (dump_file, stmt, 0, 0); } /* If changed is false, this is either opindex == 0 or all outer rhs2's were equal to corresponding oe->op, and powi_result is NULL. That means lhs is equivalent before and after reassociation. Otherwise ensure the old lhs SSA_NAME is not reused and create a new stmt as well, so that any debug stmts will be properly adjusted. */ if (changed) { gimple_stmt_iterator gsi = gsi_for_stmt (stmt); unsigned int uid = gimple_uid (stmt); gimple insert_point = find_insert_point (stmt, new_rhs1, oe->op); lhs = make_ssa_name (TREE_TYPE (lhs), NULL); stmt = gimple_build_assign_with_ops (gimple_assign_rhs_code (stmt), lhs, new_rhs1, oe->op); gimple_set_uid (stmt, uid); gimple_set_visited (stmt, true); if (insert_point == gsi_stmt (gsi)) gsi_insert_before (&gsi, stmt, GSI_SAME_STMT); else insert_stmt_after (stmt, insert_point); } else { gcc_checking_assert (find_insert_point (stmt, new_rhs1, oe->op) == stmt); gimple_assign_set_rhs1 (stmt, new_rhs1); gimple_assign_set_rhs2 (stmt, oe->op); update_stmt (stmt); } if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " into "); print_gimple_stmt (dump_file, stmt, 0, 0); } } return lhs; } /* Find out how many cycles we need to compute statements chain. OPS_NUM holds number os statements in a chain. CPU_WIDTH is a maximum number of independent statements we may execute per cycle. */ static int get_required_cycles (int ops_num, int cpu_width) { int res; int elog; unsigned int rest; /* While we have more than 2 * cpu_width operands we may reduce number of operands by cpu_width per cycle. */ res = ops_num / (2 * cpu_width); /* Remained operands count may be reduced twice per cycle until we have only one operand. */ rest = (unsigned)(ops_num - res * cpu_width); elog = exact_log2 (rest); if (elog >= 0) res += elog; else res += floor_log2 (rest) + 1; return res; } /* Returns an optimal number of registers to use for computation of given statements. */ static int get_reassociation_width (int ops_num, enum tree_code opc, enum machine_mode mode) { int param_width = PARAM_VALUE (PARAM_TREE_REASSOC_WIDTH); int width; int width_min; int cycles_best; if (param_width > 0) width = param_width; else width = targetm.sched.reassociation_width (opc, mode); if (width == 1) return width; /* Get the minimal time required for sequence computation. */ cycles_best = get_required_cycles (ops_num, width); /* Check if we may use less width and still compute sequence for the same time. It will allow us to reduce registers usage. get_required_cycles is monotonically increasing with lower width so we can perform a binary search for the minimal width that still results in the optimal cycle count. */ width_min = 1; while (width > width_min) { int width_mid = (width + width_min) / 2; if (get_required_cycles (ops_num, width_mid) == cycles_best) width = width_mid; else if (width_min < width_mid) width_min = width_mid; else break; } return width; } /* Recursively rewrite our linearized statements so that the operators match those in OPS[OPINDEX], putting the computation in rank order and trying to allow operations to be executed in parallel. */ static void rewrite_expr_tree_parallel (gimple stmt, int width, vec ops) { enum tree_code opcode = gimple_assign_rhs_code (stmt); int op_num = ops.length (); int stmt_num = op_num - 1; gimple *stmts = XALLOCAVEC (gimple, stmt_num); int op_index = op_num - 1; int stmt_index = 0; int ready_stmts_end = 0; int i = 0; tree last_rhs1 = gimple_assign_rhs1 (stmt); /* We start expression rewriting from the top statements. So, in this loop we create a full list of statements we will work with. */ stmts[stmt_num - 1] = stmt; for (i = stmt_num - 2; i >= 0; i--) stmts[i] = SSA_NAME_DEF_STMT (gimple_assign_rhs1 (stmts[i+1])); for (i = 0; i < stmt_num; i++) { tree op1, op2; /* Determine whether we should use results of already handled statements or not. */ if (ready_stmts_end == 0 && (i - stmt_index >= width || op_index < 1)) ready_stmts_end = i; /* Now we choose operands for the next statement. Non zero value in ready_stmts_end means here that we should use the result of already generated statements as new operand. */ if (ready_stmts_end > 0) { op1 = gimple_assign_lhs (stmts[stmt_index++]); if (ready_stmts_end > stmt_index) op2 = gimple_assign_lhs (stmts[stmt_index++]); else if (op_index >= 0) op2 = ops[op_index--]->op; else { gcc_assert (stmt_index < i); op2 = gimple_assign_lhs (stmts[stmt_index++]); } if (stmt_index >= ready_stmts_end) ready_stmts_end = 0; } else { if (op_index > 1) swap_ops_for_binary_stmt (ops, op_index - 2, NULL); op2 = ops[op_index--]->op; op1 = ops[op_index--]->op; } /* If we emit the last statement then we should put operands into the last statement. It will also break the loop. */ if (op_index < 0 && stmt_index == i) i = stmt_num - 1; if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Transforming "); print_gimple_stmt (dump_file, stmts[i], 0, 0); } /* We keep original statement only for the last one. All others are recreated. */ if (i == stmt_num - 1) { gimple_assign_set_rhs1 (stmts[i], op1); gimple_assign_set_rhs2 (stmts[i], op2); update_stmt (stmts[i]); } else stmts[i] = build_and_add_sum (TREE_TYPE (last_rhs1), op1, op2, opcode); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " into "); print_gimple_stmt (dump_file, stmts[i], 0, 0); } } remove_visited_stmt_chain (last_rhs1); } /* Transform STMT, which is really (A +B) + (C + D) into the left linear form, ((A+B)+C)+D. Recurse on D if necessary. */ static void linearize_expr (gimple stmt) { gimple_stmt_iterator gsi; gimple binlhs = SSA_NAME_DEF_STMT (gimple_assign_rhs1 (stmt)); gimple binrhs = SSA_NAME_DEF_STMT (gimple_assign_rhs2 (stmt)); gimple oldbinrhs = binrhs; enum tree_code rhscode = gimple_assign_rhs_code (stmt); gimple newbinrhs = NULL; struct loop *loop = loop_containing_stmt (stmt); tree lhs = gimple_assign_lhs (stmt); gcc_assert (is_reassociable_op (binlhs, rhscode, loop) && is_reassociable_op (binrhs, rhscode, loop)); gsi = gsi_for_stmt (stmt); gimple_assign_set_rhs2 (stmt, gimple_assign_rhs1 (binrhs)); binrhs = gimple_build_assign_with_ops (gimple_assign_rhs_code (binrhs), make_ssa_name (TREE_TYPE (lhs), NULL), gimple_assign_lhs (binlhs), gimple_assign_rhs2 (binrhs)); gimple_assign_set_rhs1 (stmt, gimple_assign_lhs (binrhs)); gsi_insert_before (&gsi, binrhs, GSI_SAME_STMT); gimple_set_uid (binrhs, gimple_uid (stmt)); if (TREE_CODE (gimple_assign_rhs2 (stmt)) == SSA_NAME) newbinrhs = SSA_NAME_DEF_STMT (gimple_assign_rhs2 (stmt)); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Linearized: "); print_gimple_stmt (dump_file, stmt, 0, 0); } reassociate_stats.linearized++; update_stmt (stmt); gsi = gsi_for_stmt (oldbinrhs); reassoc_remove_stmt (&gsi); release_defs (oldbinrhs); gimple_set_visited (stmt, true); gimple_set_visited (binlhs, true); gimple_set_visited (binrhs, true); /* Tail recurse on the new rhs if it still needs reassociation. */ if (newbinrhs && is_reassociable_op (newbinrhs, rhscode, loop)) /* ??? This should probably be linearize_expr (newbinrhs) but I don't want to change the algorithm while converting to tuples. */ linearize_expr (stmt); } /* If LHS has a single immediate use that is a GIMPLE_ASSIGN statement, return it. Otherwise, return NULL. */ static gimple get_single_immediate_use (tree lhs) { use_operand_p immuse; gimple immusestmt; if (TREE_CODE (lhs) == SSA_NAME && single_imm_use (lhs, &immuse, &immusestmt) && is_gimple_assign (immusestmt)) return immusestmt; return NULL; } /* Recursively negate the value of TONEGATE, and return the SSA_NAME representing the negated value. Insertions of any necessary instructions go before GSI. This function is recursive in that, if you hand it "a_5" as the value to negate, and a_5 is defined by "a_5 = b_3 + b_4", it will transform b_3 + b_4 into a_5 = -b_3 + -b_4. */ static tree negate_value (tree tonegate, gimple_stmt_iterator *gsip) { gimple negatedefstmt = NULL; tree resultofnegate; gimple_stmt_iterator gsi; unsigned int uid; /* If we are trying to negate a name, defined by an add, negate the add operands instead. */ if (TREE_CODE (tonegate) == SSA_NAME) negatedefstmt = SSA_NAME_DEF_STMT (tonegate); if (TREE_CODE (tonegate) == SSA_NAME && is_gimple_assign (negatedefstmt) && TREE_CODE (gimple_assign_lhs (negatedefstmt)) == SSA_NAME && has_single_use (gimple_assign_lhs (negatedefstmt)) && gimple_assign_rhs_code (negatedefstmt) == PLUS_EXPR) { tree rhs1 = gimple_assign_rhs1 (negatedefstmt); tree rhs2 = gimple_assign_rhs2 (negatedefstmt); tree lhs = gimple_assign_lhs (negatedefstmt); gimple g; gsi = gsi_for_stmt (negatedefstmt); rhs1 = negate_value (rhs1, &gsi); gsi = gsi_for_stmt (negatedefstmt); rhs2 = negate_value (rhs2, &gsi); gsi = gsi_for_stmt (negatedefstmt); lhs = make_ssa_name (TREE_TYPE (lhs), NULL); gimple_set_visited (negatedefstmt, true); g = gimple_build_assign_with_ops (PLUS_EXPR, lhs, rhs1, rhs2); gimple_set_uid (g, gimple_uid (negatedefstmt)); gsi_insert_before (&gsi, g, GSI_SAME_STMT); return lhs; } tonegate = fold_build1 (NEGATE_EXPR, TREE_TYPE (tonegate), tonegate); resultofnegate = force_gimple_operand_gsi (gsip, tonegate, true, NULL_TREE, true, GSI_SAME_STMT); gsi = *gsip; uid = gimple_uid (gsi_stmt (gsi)); for (gsi_prev (&gsi); !gsi_end_p (gsi); gsi_prev (&gsi)) { gimple stmt = gsi_stmt (gsi); if (gimple_uid (stmt) != 0) break; gimple_set_uid (stmt, uid); } return resultofnegate; } /* Return true if we should break up the subtract in STMT into an add with negate. This is true when we the subtract operands are really adds, or the subtract itself is used in an add expression. In either case, breaking up the subtract into an add with negate exposes the adds to reassociation. */ static bool should_break_up_subtract (gimple stmt) { tree lhs = gimple_assign_lhs (stmt); tree binlhs = gimple_assign_rhs1 (stmt); tree binrhs = gimple_assign_rhs2 (stmt); gimple immusestmt; struct loop *loop = loop_containing_stmt (stmt); if (TREE_CODE (binlhs) == SSA_NAME && is_reassociable_op (SSA_NAME_DEF_STMT (binlhs), PLUS_EXPR, loop)) return true; if (TREE_CODE (binrhs) == SSA_NAME && is_reassociable_op (SSA_NAME_DEF_STMT (binrhs), PLUS_EXPR, loop)) return true; if (TREE_CODE (lhs) == SSA_NAME && (immusestmt = get_single_immediate_use (lhs)) && is_gimple_assign (immusestmt) && (gimple_assign_rhs_code (immusestmt) == PLUS_EXPR || gimple_assign_rhs_code (immusestmt) == MULT_EXPR)) return true; return false; } /* Transform STMT from A - B into A + -B. */ static void break_up_subtract (gimple stmt, gimple_stmt_iterator *gsip) { tree rhs1 = gimple_assign_rhs1 (stmt); tree rhs2 = gimple_assign_rhs2 (stmt); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Breaking up subtract "); print_gimple_stmt (dump_file, stmt, 0, 0); } rhs2 = negate_value (rhs2, gsip); gimple_assign_set_rhs_with_ops (gsip, PLUS_EXPR, rhs1, rhs2); update_stmt (stmt); } /* Determine whether STMT is a builtin call that raises an SSA name to an integer power and has only one use. If so, and this is early reassociation and unsafe math optimizations are permitted, place the SSA name in *BASE and the exponent in *EXPONENT, and return TRUE. If any of these conditions does not hold, return FALSE. */ static bool acceptable_pow_call (gimple stmt, tree *base, HOST_WIDE_INT *exponent) { tree fndecl, arg1; REAL_VALUE_TYPE c, cint; if (!first_pass_instance || !flag_unsafe_math_optimizations || !is_gimple_call (stmt) || !has_single_use (gimple_call_lhs (stmt))) return false; fndecl = gimple_call_fndecl (stmt); if (!fndecl || DECL_BUILT_IN_CLASS (fndecl) != BUILT_IN_NORMAL) return false; switch (DECL_FUNCTION_CODE (fndecl)) { CASE_FLT_FN (BUILT_IN_POW): if (flag_errno_math) return false; *base = gimple_call_arg (stmt, 0); arg1 = gimple_call_arg (stmt, 1); if (TREE_CODE (arg1) != REAL_CST) return false; c = TREE_REAL_CST (arg1); if (REAL_EXP (&c) > HOST_BITS_PER_WIDE_INT) return false; *exponent = real_to_integer (&c); real_from_integer (&cint, VOIDmode, *exponent, *exponent < 0 ? -1 : 0, 0); if (!real_identical (&c, &cint)) return false; break; CASE_FLT_FN (BUILT_IN_POWI): *base = gimple_call_arg (stmt, 0); arg1 = gimple_call_arg (stmt, 1); if (!tree_fits_shwi_p (arg1)) return false; *exponent = tree_to_shwi (arg1); break; default: return false; } /* Expanding negative exponents is generally unproductive, so we don't complicate matters with those. Exponents of zero and one should have been handled by expression folding. */ if (*exponent < 2 || TREE_CODE (*base) != SSA_NAME) return false; return true; } /* Recursively linearize a binary expression that is the RHS of STMT. Place the operands of the expression tree in the vector named OPS. */ static void linearize_expr_tree (vec *ops, gimple stmt, bool is_associative, bool set_visited) { tree binlhs = gimple_assign_rhs1 (stmt); tree binrhs = gimple_assign_rhs2 (stmt); gimple binlhsdef = NULL, binrhsdef = NULL; bool binlhsisreassoc = false; bool binrhsisreassoc = false; enum tree_code rhscode = gimple_assign_rhs_code (stmt); struct loop *loop = loop_containing_stmt (stmt); tree base = NULL_TREE; HOST_WIDE_INT exponent = 0; if (set_visited) gimple_set_visited (stmt, true); if (TREE_CODE (binlhs) == SSA_NAME) { binlhsdef = SSA_NAME_DEF_STMT (binlhs); binlhsisreassoc = (is_reassociable_op (binlhsdef, rhscode, loop) && !stmt_could_throw_p (binlhsdef)); } if (TREE_CODE (binrhs) == SSA_NAME) { binrhsdef = SSA_NAME_DEF_STMT (binrhs); binrhsisreassoc = (is_reassociable_op (binrhsdef, rhscode, loop) && !stmt_could_throw_p (binrhsdef)); } /* If the LHS is not reassociable, but the RHS is, we need to swap them. If neither is reassociable, there is nothing we can do, so just put them in the ops vector. If the LHS is reassociable, linearize it. If both are reassociable, then linearize the RHS and the LHS. */ if (!binlhsisreassoc) { tree temp; /* If this is not a associative operation like division, give up. */ if (!is_associative) { add_to_ops_vec (ops, binrhs); return; } if (!binrhsisreassoc) { if (rhscode == MULT_EXPR && TREE_CODE (binrhs) == SSA_NAME && acceptable_pow_call (binrhsdef, &base, &exponent)) { add_repeat_to_ops_vec (ops, base, exponent); gimple_set_visited (binrhsdef, true); } else add_to_ops_vec (ops, binrhs); if (rhscode == MULT_EXPR && TREE_CODE (binlhs) == SSA_NAME && acceptable_pow_call (binlhsdef, &base, &exponent)) { add_repeat_to_ops_vec (ops, base, exponent); gimple_set_visited (binlhsdef, true); } else add_to_ops_vec (ops, binlhs); return; } if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "swapping operands of "); print_gimple_stmt (dump_file, stmt, 0, 0); } swap_ssa_operands (stmt, gimple_assign_rhs1_ptr (stmt), gimple_assign_rhs2_ptr (stmt)); update_stmt (stmt); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " is now "); print_gimple_stmt (dump_file, stmt, 0, 0); } /* We want to make it so the lhs is always the reassociative op, so swap. */ temp = binlhs; binlhs = binrhs; binrhs = temp; } else if (binrhsisreassoc) { linearize_expr (stmt); binlhs = gimple_assign_rhs1 (stmt); binrhs = gimple_assign_rhs2 (stmt); } gcc_assert (TREE_CODE (binrhs) != SSA_NAME || !is_reassociable_op (SSA_NAME_DEF_STMT (binrhs), rhscode, loop)); linearize_expr_tree (ops, SSA_NAME_DEF_STMT (binlhs), is_associative, set_visited); if (rhscode == MULT_EXPR && TREE_CODE (binrhs) == SSA_NAME && acceptable_pow_call (SSA_NAME_DEF_STMT (binrhs), &base, &exponent)) { add_repeat_to_ops_vec (ops, base, exponent); gimple_set_visited (SSA_NAME_DEF_STMT (binrhs), true); } else add_to_ops_vec (ops, binrhs); } /* Repropagate the negates back into subtracts, since no other pass currently does it. */ static void repropagate_negates (void) { unsigned int i = 0; tree negate; FOR_EACH_VEC_ELT (plus_negates, i, negate) { gimple user = get_single_immediate_use (negate); if (!user || !is_gimple_assign (user)) continue; /* The negate operand can be either operand of a PLUS_EXPR (it can be the LHS if the RHS is a constant for example). Force the negate operand to the RHS of the PLUS_EXPR, then transform the PLUS_EXPR into a MINUS_EXPR. */ if (gimple_assign_rhs_code (user) == PLUS_EXPR) { /* If the negated operand appears on the LHS of the PLUS_EXPR, exchange the operands of the PLUS_EXPR to force the negated operand to the RHS of the PLUS_EXPR. */ if (gimple_assign_rhs1 (user) == negate) { swap_ssa_operands (user, gimple_assign_rhs1_ptr (user), gimple_assign_rhs2_ptr (user)); } /* Now transform the PLUS_EXPR into a MINUS_EXPR and replace the RHS of the PLUS_EXPR with the operand of the NEGATE_EXPR. */ if (gimple_assign_rhs2 (user) == negate) { tree rhs1 = gimple_assign_rhs1 (user); tree rhs2 = get_unary_op (negate, NEGATE_EXPR); gimple_stmt_iterator gsi = gsi_for_stmt (user); gimple_assign_set_rhs_with_ops (&gsi, MINUS_EXPR, rhs1, rhs2); update_stmt (user); } } else if (gimple_assign_rhs_code (user) == MINUS_EXPR) { if (gimple_assign_rhs1 (user) == negate) { /* We have x = -a y = x - b which we transform into x = a + b y = -x . This pushes down the negate which we possibly can merge into some other operation, hence insert it into the plus_negates vector. */ gimple feed = SSA_NAME_DEF_STMT (negate); tree a = gimple_assign_rhs1 (feed); tree b = gimple_assign_rhs2 (user); gimple_stmt_iterator gsi = gsi_for_stmt (feed); gimple_stmt_iterator gsi2 = gsi_for_stmt (user); tree x = make_ssa_name (TREE_TYPE (gimple_assign_lhs (feed)), NULL); gimple g = gimple_build_assign_with_ops (PLUS_EXPR, x, a, b); gsi_insert_before (&gsi2, g, GSI_SAME_STMT); gimple_assign_set_rhs_with_ops (&gsi2, NEGATE_EXPR, x, NULL); user = gsi_stmt (gsi2); update_stmt (user); reassoc_remove_stmt (&gsi); release_defs (feed); plus_negates.safe_push (gimple_assign_lhs (user)); } else { /* Transform "x = -a; y = b - x" into "y = b + a", getting rid of one operation. */ gimple feed = SSA_NAME_DEF_STMT (negate); tree a = gimple_assign_rhs1 (feed); tree rhs1 = gimple_assign_rhs1 (user); gimple_stmt_iterator gsi = gsi_for_stmt (user); gimple_assign_set_rhs_with_ops (&gsi, PLUS_EXPR, rhs1, a); update_stmt (gsi_stmt (gsi)); } } } } /* Returns true if OP is of a type for which we can do reassociation. That is for integral or non-saturating fixed-point types, and for floating point type when associative-math is enabled. */ static bool can_reassociate_p (tree op) { tree type = TREE_TYPE (op); if ((INTEGRAL_TYPE_P (type) && TYPE_OVERFLOW_WRAPS (type)) || NON_SAT_FIXED_POINT_TYPE_P (type) || (flag_associative_math && FLOAT_TYPE_P (type))) return true; return false; } /* Break up subtract operations in block BB. We do this top down because we don't know whether the subtract is part of a possible chain of reassociation except at the top. IE given d = f + g c = a + e b = c - d q = b - r k = t - q we want to break up k = t - q, but we won't until we've transformed q = b - r, which won't be broken up until we transform b = c - d. En passant, clear the GIMPLE visited flag on every statement and set UIDs within each basic block. */ static void break_up_subtract_bb (basic_block bb) { gimple_stmt_iterator gsi; basic_block son; unsigned int uid = 1; for (gsi = gsi_start_bb (bb); !gsi_end_p (gsi); gsi_next (&gsi)) { gimple stmt = gsi_stmt (gsi); gimple_set_visited (stmt, false); gimple_set_uid (stmt, uid++); if (!is_gimple_assign (stmt) || !can_reassociate_p (gimple_assign_lhs (stmt))) continue; /* Look for simple gimple subtract operations. */ if (gimple_assign_rhs_code (stmt) == MINUS_EXPR) { if (!can_reassociate_p (gimple_assign_rhs1 (stmt)) || !can_reassociate_p (gimple_assign_rhs2 (stmt))) continue; /* Check for a subtract used only in an addition. If this is the case, transform it into add of a negate for better reassociation. IE transform C = A-B into C = A + -B if C is only used in an addition. */ if (should_break_up_subtract (stmt)) break_up_subtract (stmt, &gsi); } else if (gimple_assign_rhs_code (stmt) == NEGATE_EXPR && can_reassociate_p (gimple_assign_rhs1 (stmt))) plus_negates.safe_push (gimple_assign_lhs (stmt)); } for (son = first_dom_son (CDI_DOMINATORS, bb); son; son = next_dom_son (CDI_DOMINATORS, son)) break_up_subtract_bb (son); } /* Used for repeated factor analysis. */ struct repeat_factor_d { /* An SSA name that occurs in a multiply chain. */ tree factor; /* Cached rank of the factor. */ unsigned rank; /* Number of occurrences of the factor in the chain. */ HOST_WIDE_INT count; /* An SSA name representing the product of this factor and all factors appearing later in the repeated factor vector. */ tree repr; }; typedef struct repeat_factor_d repeat_factor, *repeat_factor_t; typedef const struct repeat_factor_d *const_repeat_factor_t; static vec repeat_factor_vec; /* Used for sorting the repeat factor vector. Sort primarily by ascending occurrence count, secondarily by descending rank. */ static int compare_repeat_factors (const void *x1, const void *x2) { const_repeat_factor_t rf1 = (const_repeat_factor_t) x1; const_repeat_factor_t rf2 = (const_repeat_factor_t) x2; if (rf1->count != rf2->count) return rf1->count - rf2->count; return rf2->rank - rf1->rank; } /* Look for repeated operands in OPS in the multiply tree rooted at STMT. Replace them with an optimal sequence of multiplies and powi builtin calls, and remove the used operands from OPS. Return an SSA name representing the value of the replacement sequence. */ static tree attempt_builtin_powi (gimple stmt, vec *ops) { unsigned i, j, vec_len; int ii; operand_entry_t oe; repeat_factor_t rf1, rf2; repeat_factor rfnew; tree result = NULL_TREE; tree target_ssa, iter_result; tree type = TREE_TYPE (gimple_get_lhs (stmt)); tree powi_fndecl = mathfn_built_in (type, BUILT_IN_POWI); gimple_stmt_iterator gsi = gsi_for_stmt (stmt); gimple mul_stmt, pow_stmt; /* Nothing to do if BUILT_IN_POWI doesn't exist for this type and target. */ if (!powi_fndecl) return NULL_TREE; /* Allocate the repeated factor vector. */ repeat_factor_vec.create (10); /* Scan the OPS vector for all SSA names in the product and build up a vector of occurrence counts for each factor. */ FOR_EACH_VEC_ELT (*ops, i, oe) { if (TREE_CODE (oe->op) == SSA_NAME) { FOR_EACH_VEC_ELT (repeat_factor_vec, j, rf1) { if (rf1->factor == oe->op) { rf1->count += oe->count; break; } } if (j >= repeat_factor_vec.length ()) { rfnew.factor = oe->op; rfnew.rank = oe->rank; rfnew.count = oe->count; rfnew.repr = NULL_TREE; repeat_factor_vec.safe_push (rfnew); } } } /* Sort the repeated factor vector by (a) increasing occurrence count, and (b) decreasing rank. */ repeat_factor_vec.qsort (compare_repeat_factors); /* It is generally best to combine as many base factors as possible into a product before applying __builtin_powi to the result. However, the sort order chosen for the repeated factor vector allows us to cache partial results for the product of the base factors for subsequent use. When we already have a cached partial result from a previous iteration, it is best to make use of it before looking for another __builtin_pow opportunity. As an example, consider x * x * y * y * y * z * z * z * z. We want to first compose the product x * y * z, raise it to the second power, then multiply this by y * z, and finally multiply by z. This can be done in 5 multiplies provided we cache y * z for use in both expressions: t1 = y * z t2 = t1 * x t3 = t2 * t2 t4 = t1 * t3 result = t4 * z If we instead ignored the cached y * z and first multiplied by the __builtin_pow opportunity z * z, we would get the inferior: t1 = y * z t2 = t1 * x t3 = t2 * t2 t4 = z * z t5 = t3 * t4 result = t5 * y */ vec_len = repeat_factor_vec.length (); /* Repeatedly look for opportunities to create a builtin_powi call. */ while (true) { HOST_WIDE_INT power; /* First look for the largest cached product of factors from preceding iterations. If found, create a builtin_powi for it if the minimum occurrence count for its factors is at least 2, or just use this cached product as our next multiplicand if the minimum occurrence count is 1. */ FOR_EACH_VEC_ELT (repeat_factor_vec, j, rf1) { if (rf1->repr && rf1->count > 0) break; } if (j < vec_len) { power = rf1->count; if (power == 1) { iter_result = rf1->repr; if (dump_file && (dump_flags & TDF_DETAILS)) { unsigned elt; repeat_factor_t rf; fputs ("Multiplying by cached product ", dump_file); for (elt = j; elt < vec_len; elt++) { rf = &repeat_factor_vec[elt]; print_generic_expr (dump_file, rf->factor, 0); if (elt < vec_len - 1) fputs (" * ", dump_file); } fputs ("\n", dump_file); } } else { iter_result = make_temp_ssa_name (type, NULL, "reassocpow"); pow_stmt = gimple_build_call (powi_fndecl, 2, rf1->repr, build_int_cst (integer_type_node, power)); gimple_call_set_lhs (pow_stmt, iter_result); gimple_set_location (pow_stmt, gimple_location (stmt)); gsi_insert_before (&gsi, pow_stmt, GSI_SAME_STMT); if (dump_file && (dump_flags & TDF_DETAILS)) { unsigned elt; repeat_factor_t rf; fputs ("Building __builtin_pow call for cached product (", dump_file); for (elt = j; elt < vec_len; elt++) { rf = &repeat_factor_vec[elt]; print_generic_expr (dump_file, rf->factor, 0); if (elt < vec_len - 1) fputs (" * ", dump_file); } fprintf (dump_file, ")^"HOST_WIDE_INT_PRINT_DEC"\n", power); } } } else { /* Otherwise, find the first factor in the repeated factor vector whose occurrence count is at least 2. If no such factor exists, there are no builtin_powi opportunities remaining. */ FOR_EACH_VEC_ELT (repeat_factor_vec, j, rf1) { if (rf1->count >= 2) break; } if (j >= vec_len) break; power = rf1->count; if (dump_file && (dump_flags & TDF_DETAILS)) { unsigned elt; repeat_factor_t rf; fputs ("Building __builtin_pow call for (", dump_file); for (elt = j; elt < vec_len; elt++) { rf = &repeat_factor_vec[elt]; print_generic_expr (dump_file, rf->factor, 0); if (elt < vec_len - 1) fputs (" * ", dump_file); } fprintf (dump_file, ")^"HOST_WIDE_INT_PRINT_DEC"\n", power); } reassociate_stats.pows_created++; /* Visit each element of the vector in reverse order (so that high-occurrence elements are visited first, and within the same occurrence count, lower-ranked elements are visited first). Form a linear product of all elements in this order whose occurrencce count is at least that of element J. Record the SSA name representing the product of each element with all subsequent elements in the vector. */ if (j == vec_len - 1) rf1->repr = rf1->factor; else { for (ii = vec_len - 2; ii >= (int)j; ii--) { tree op1, op2; rf1 = &repeat_factor_vec[ii]; rf2 = &repeat_factor_vec[ii + 1]; /* Init the last factor's representative to be itself. */ if (!rf2->repr) rf2->repr = rf2->factor; op1 = rf1->factor; op2 = rf2->repr; target_ssa = make_temp_ssa_name (type, NULL, "reassocpow"); mul_stmt = gimple_build_assign_with_ops (MULT_EXPR, target_ssa, op1, op2); gimple_set_location (mul_stmt, gimple_location (stmt)); gsi_insert_before (&gsi, mul_stmt, GSI_SAME_STMT); rf1->repr = target_ssa; /* Don't reprocess the multiply we just introduced. */ gimple_set_visited (mul_stmt, true); } } /* Form a call to __builtin_powi for the maximum product just formed, raised to the power obtained earlier. */ rf1 = &repeat_factor_vec[j]; iter_result = make_temp_ssa_name (type, NULL, "reassocpow"); pow_stmt = gimple_build_call (powi_fndecl, 2, rf1->repr, build_int_cst (integer_type_node, power)); gimple_call_set_lhs (pow_stmt, iter_result); gimple_set_location (pow_stmt, gimple_location (stmt)); gsi_insert_before (&gsi, pow_stmt, GSI_SAME_STMT); } /* If we previously formed at least one other builtin_powi call, form the product of this one and those others. */ if (result) { tree new_result = make_temp_ssa_name (type, NULL, "reassocpow"); mul_stmt = gimple_build_assign_with_ops (MULT_EXPR, new_result, result, iter_result); gimple_set_location (mul_stmt, gimple_location (stmt)); gsi_insert_before (&gsi, mul_stmt, GSI_SAME_STMT); gimple_set_visited (mul_stmt, true); result = new_result; } else result = iter_result; /* Decrement the occurrence count of each element in the product by the count found above, and remove this many copies of each factor from OPS. */ for (i = j; i < vec_len; i++) { unsigned k = power; unsigned n; rf1 = &repeat_factor_vec[i]; rf1->count -= power; FOR_EACH_VEC_ELT_REVERSE (*ops, n, oe) { if (oe->op == rf1->factor) { if (oe->count <= k) { ops->ordered_remove (n); k -= oe->count; if (k == 0) break; } else { oe->count -= k; break; } } } } } /* At this point all elements in the repeated factor vector have a remaining occurrence count of 0 or 1, and those with a count of 1 don't have cached representatives. Re-sort the ops vector and clean up. */ ops->qsort (sort_by_operand_rank); repeat_factor_vec.release (); /* Return the final product computed herein. Note that there may still be some elements with single occurrence count left in OPS; those will be handled by the normal reassociation logic. */ return result; } /* Transform STMT at *GSI into a copy by replacing its rhs with NEW_RHS. */ static void transform_stmt_to_copy (gimple_stmt_iterator *gsi, gimple stmt, tree new_rhs) { tree rhs1; if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Transforming "); print_gimple_stmt (dump_file, stmt, 0, 0); } rhs1 = gimple_assign_rhs1 (stmt); gimple_assign_set_rhs_from_tree (gsi, new_rhs); update_stmt (stmt); remove_visited_stmt_chain (rhs1); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " into "); print_gimple_stmt (dump_file, stmt, 0, 0); } } /* Transform STMT at *GSI into a multiply of RHS1 and RHS2. */ static void transform_stmt_to_multiply (gimple_stmt_iterator *gsi, gimple stmt, tree rhs1, tree rhs2) { if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, "Transforming "); print_gimple_stmt (dump_file, stmt, 0, 0); } gimple_assign_set_rhs_with_ops (gsi, MULT_EXPR, rhs1, rhs2); update_stmt (gsi_stmt (*gsi)); remove_visited_stmt_chain (rhs1); if (dump_file && (dump_flags & TDF_DETAILS)) { fprintf (dump_file, " into "); print_gimple_stmt (dump_file, stmt, 0, 0); } } /* Reassociate expressions in basic block BB and its post-dominator as children. */ static void reassociate_bb (basic_block bb) { gimple_stmt_iterator gsi; basic_block son; gimple stmt = last_stmt (bb); if (stmt && !gimple_visited_p (stmt)) maybe_optimize_range_tests (stmt); for (gsi = gsi_last_bb (bb); !gsi_end_p (gsi); gsi_prev (&gsi)) { stmt = gsi_stmt (gsi); if (is_gimple_assign (stmt) && !stmt_could_throw_p (stmt)) { tree lhs, rhs1, rhs2; enum tree_code rhs_code = gimple_assign_rhs_code (stmt); /* If this is not a gimple binary expression, there is nothing for us to do with it. */ if (get_gimple_rhs_class (rhs_code) != GIMPLE_BINARY_RHS) continue; /* If this was part of an already processed statement, we don't need to touch it again. */ if (gimple_visited_p (stmt)) { /* This statement might have become dead because of previous reassociations. */ if (has_zero_uses (gimple_get_lhs (stmt))) { reassoc_remove_stmt (&gsi); release_defs (stmt); /* We might end up removing the last stmt above which places the iterator to the end of the sequence. Reset it to the last stmt in this case which might be the end of the sequence as well if we removed the last statement of the sequence. In which case we need to bail out. */ if (gsi_end_p (gsi)) { gsi = gsi_last_bb (bb); if (gsi_end_p (gsi)) break; } } continue; } lhs = gimple_assign_lhs (stmt); rhs1 = gimple_assign_rhs1 (stmt); rhs2 = gimple_assign_rhs2 (stmt); /* For non-bit or min/max operations we can't associate all types. Verify that here. */ if (rhs_code != BIT_IOR_EXPR && rhs_code != BIT_AND_EXPR && rhs_code != BIT_XOR_EXPR && rhs_code != MIN_EXPR && rhs_code != MAX_EXPR && (!can_reassociate_p (lhs) || !can_reassociate_p (rhs1) || !can_reassociate_p (rhs2))) continue; if (associative_tree_code (rhs_code)) { auto_vec ops; tree powi_result = NULL_TREE; /* There may be no immediate uses left by the time we get here because we may have eliminated them all. */ if (TREE_CODE (lhs) == SSA_NAME && has_zero_uses (lhs)) continue; gimple_set_visited (stmt, true); linearize_expr_tree (&ops, stmt, true, true); ops.qsort (sort_by_operand_rank); optimize_ops_list (rhs_code, &ops); if (undistribute_ops_list (rhs_code, &ops, loop_containing_stmt (stmt))) { ops.qsort (sort_by_operand_rank); optimize_ops_list (rhs_code, &ops); } if (rhs_code == BIT_IOR_EXPR || rhs_code == BIT_AND_EXPR) optimize_range_tests (rhs_code, &ops); if (first_pass_instance && rhs_code == MULT_EXPR && flag_unsafe_math_optimizations) powi_result = attempt_builtin_powi (stmt, &ops); /* If the operand vector is now empty, all operands were consumed by the __builtin_powi optimization. */ if (ops.length () == 0) transform_stmt_to_copy (&gsi, stmt, powi_result); else if (ops.length () == 1) { tree last_op = ops.last ()->op; if (powi_result) transform_stmt_to_multiply (&gsi, stmt, last_op, powi_result); else transform_stmt_to_copy (&gsi, stmt, last_op); } else { enum machine_mode mode = TYPE_MODE (TREE_TYPE (lhs)); int ops_num = ops.length (); int width = get_reassociation_width (ops_num, rhs_code, mode); tree new_lhs = lhs; if (dump_file && (dump_flags & TDF_DETAILS)) fprintf (dump_file, "Width = %d was chosen for reassociation\n", width); if (width > 1 && ops.length () > 3) rewrite_expr_tree_parallel (stmt, width, ops); else { /* When there are three operands left, we want to make sure the ones that get the double binary op are chosen wisely. */ int len = ops.length (); if (len >= 3) swap_ops_for_binary_stmt (ops, len - 3, stmt); new_lhs = rewrite_expr_tree (stmt, 0, ops, powi_result != NULL); } /* If we combined some repeated factors into a __builtin_powi call, multiply that result by the reassociated operands. */ if (powi_result) { gimple mul_stmt, lhs_stmt = SSA_NAME_DEF_STMT (lhs); tree type = TREE_TYPE (lhs); tree target_ssa = make_temp_ssa_name (type, NULL, "reassocpow"); gimple_set_lhs (lhs_stmt, target_ssa); update_stmt (lhs_stmt); if (lhs != new_lhs) target_ssa = new_lhs; mul_stmt = gimple_build_assign_with_ops (MULT_EXPR, lhs, powi_result, target_ssa); gimple_set_location (mul_stmt, gimple_location (stmt)); gsi_insert_after (&gsi, mul_stmt, GSI_NEW_STMT); } } } } } for (son = first_dom_son (CDI_POST_DOMINATORS, bb); son; son = next_dom_son (CDI_POST_DOMINATORS, son)) reassociate_bb (son); } void dump_ops_vector (FILE *file, vec ops); void debug_ops_vector (vec ops); /* Dump the operand entry vector OPS to FILE. */ void dump_ops_vector (FILE *file, vec ops) { operand_entry_t oe; unsigned int i; FOR_EACH_VEC_ELT (ops, i, oe) { fprintf (file, "Op %d -> rank: %d, tree: ", i, oe->rank); print_generic_expr (file, oe->op, 0); } } /* Dump the operand entry vector OPS to STDERR. */ DEBUG_FUNCTION void debug_ops_vector (vec ops) { dump_ops_vector (stderr, ops); } static void do_reassoc (void) { break_up_subtract_bb (ENTRY_BLOCK_PTR_FOR_FN (cfun)); reassociate_bb (EXIT_BLOCK_PTR_FOR_FN (cfun)); } /* Initialize the reassociation pass. */ static void init_reassoc (void) { int i; long rank = 2; int *bbs = XNEWVEC (int, n_basic_blocks_for_fn (cfun) - NUM_FIXED_BLOCKS); /* Find the loops, so that we can prevent moving calculations in them. */ loop_optimizer_init (AVOID_CFG_MODIFICATIONS); memset (&reassociate_stats, 0, sizeof (reassociate_stats)); operand_entry_pool = create_alloc_pool ("operand entry pool", sizeof (struct operand_entry), 30); next_operand_entry_id = 0; /* Reverse RPO (Reverse Post Order) will give us something where deeper loops come later. */ pre_and_rev_post_order_compute (NULL, bbs, false); bb_rank = XCNEWVEC (long, last_basic_block_for_fn (cfun)); operand_rank = pointer_map_create (); /* Give each default definition a distinct rank. This includes parameters and the static chain. Walk backwards over all SSA names so that we get proper rank ordering according to tree_swap_operands_p. */ for (i = num_ssa_names - 1; i > 0; --i) { tree name = ssa_name (i); if (name && SSA_NAME_IS_DEFAULT_DEF (name)) insert_operand_rank (name, ++rank); } /* Set up rank for each BB */ for (i = 0; i < n_basic_blocks_for_fn (cfun) - NUM_FIXED_BLOCKS; i++) bb_rank[bbs[i]] = ++rank << 16; free (bbs); calculate_dominance_info (CDI_POST_DOMINATORS); plus_negates = vNULL; } /* Cleanup after the reassociation pass, and print stats if requested. */ static void fini_reassoc (void) { statistics_counter_event (cfun, "Linearized", reassociate_stats.linearized); statistics_counter_event (cfun, "Constants eliminated", reassociate_stats.constants_eliminated); statistics_counter_event (cfun, "Ops eliminated", reassociate_stats.ops_eliminated); statistics_counter_event (cfun, "Statements rewritten", reassociate_stats.rewritten); statistics_counter_event (cfun, "Built-in pow[i] calls encountered", reassociate_stats.pows_encountered); statistics_counter_event (cfun, "Built-in powi calls created", reassociate_stats.pows_created); pointer_map_destroy (operand_rank); free_alloc_pool (operand_entry_pool); free (bb_rank); plus_negates.release (); free_dominance_info (CDI_POST_DOMINATORS); loop_optimizer_finalize (); } /* Gate and execute functions for Reassociation. */ static unsigned int execute_reassoc (void) { init_reassoc (); do_reassoc (); repropagate_negates (); fini_reassoc (); return 0; } static bool gate_tree_ssa_reassoc (void) { return flag_tree_reassoc != 0; } namespace { const pass_data pass_data_reassoc = { GIMPLE_PASS, /* type */ "reassoc", /* name */ OPTGROUP_NONE, /* optinfo_flags */ true, /* has_gate */ true, /* has_execute */ TV_TREE_REASSOC, /* tv_id */ ( PROP_cfg | PROP_ssa ), /* properties_required */ 0, /* properties_provided */ 0, /* properties_destroyed */ 0, /* todo_flags_start */ ( TODO_verify_ssa | TODO_update_ssa_only_virtuals | TODO_verify_flow ), /* todo_flags_finish */ }; class pass_reassoc : public gimple_opt_pass { public: pass_reassoc (gcc::context *ctxt) : gimple_opt_pass (pass_data_reassoc, ctxt) {} /* opt_pass methods: */ opt_pass * clone () { return new pass_reassoc (m_ctxt); } bool gate () { return gate_tree_ssa_reassoc (); } unsigned int execute () { return execute_reassoc (); } }; // class pass_reassoc } // anon namespace gimple_opt_pass * make_pass_reassoc (gcc::context *ctxt) { return new pass_reassoc (ctxt); }