/* Reassociation for trees.
- Copyright (C) 2005, 2007, 2008, 2009, 2010 Free Software Foundation, Inc.
+ Copyright (C) 2005, 2007, 2008, 2009, 2010, 2011
+ Free Software Foundation, Inc.
Contributed by Daniel Berlin <dan@dberlin.org>
This file is part of GCC.
#include "pointer-set.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
/* Operand->rank hashtable. */
static struct pointer_map_t *operand_rank;
+/* Forward decls. */
+static long get_rank (tree);
+
+
+/* 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 (!is_gimple_reg (SSA_NAME_VAR (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. */
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, maxrank;
+ long rank;
int i, n;
+ tree op;
if (TREE_CODE (SSA_NAME_VAR (e)) == PARM_DECL
&& SSA_NAME_IS_DEFAULT_DEF (e))
if (gimple_bb (stmt) == NULL)
return 0;
+ 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 (rank != -1)
return rank;
- /* Otherwise, find the maximum rank for the operands, or the bb
- rank, whichever is less. */
+ /* 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;
- maxrank = bb_rank[gimple_bb(stmt)->index];
if (gimple_assign_single_p (stmt))
{
tree rhs = gimple_assign_rhs1 (stmt);
n = TREE_OPERAND_LENGTH (rhs);
if (n == 0)
- rank = MAX (rank, get_rank (rhs));
+ rank = propagate_rank (rank, rhs);
else
{
- for (i = 0;
- i < n && TREE_OPERAND (rhs, i) && rank != maxrank; i++)
- rank = MAX(rank, get_rank (TREE_OPERAND (rhs, i)));
+ 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 && rank != maxrank; i++)
+ for (i = 1; i < n; i++)
{
- gcc_assert (gimple_op (stmt, i));
- rank = MAX(rank, get_rank (gimple_op (stmt, i)));
+ op = gimple_op (stmt, i);
+ gcc_assert (op);
+ rank = propagate_rank (rank, op);
}
}
{
VEC_free (operand_entry_t, heap, *ops);
*ops = NULL;
- add_to_ops_vec (ops, fold_convert (TREE_TYPE (last->op),
- integer_zero_node));
+ add_to_ops_vec (ops, build_zero_cst (TREE_TYPE (last->op)));
*all_done = true;
}
else
}
VEC_ordered_remove (operand_entry_t, *ops, i);
- add_to_ops_vec (ops, fold_convert(TREE_TYPE (oe->op),
- integer_zero_node));
+ add_to_ops_vec (ops, build_zero_cst (TREE_TYPE (oe->op)));
VEC_ordered_remove (operand_entry_t, *ops, currindex);
reassociate_stats.ops_eliminated ++;
}
if (opcode == BIT_AND_EXPR)
- oe->op = fold_convert (TREE_TYPE (oe->op), integer_zero_node);
+ oe->op = build_zero_cst (TREE_TYPE (oe->op));
else if (opcode == BIT_IOR_EXPR)
oe->op = build_low_bits_mask (TREE_TYPE (oe->op),
TYPE_PRECISION (TREE_TYPE (oe->op)));
candidates = sbitmap_alloc (length);
sbitmap_zero (candidates);
nr_candidates = 0;
- for (i = 0; VEC_iterate (operand_entry_t, *ops, i, oe1); ++i)
+ FOR_EACH_VEC_ELT (operand_entry_t, *ops, i, oe1)
{
enum tree_code dcode;
gimple oe1def;
linearize_expr_tree (&subops[i], oedef,
associative_tree_code (oecode), false);
- for (j = 0; VEC_iterate (operand_entry_t, subops[i], j, oe1); ++j)
+ FOR_EACH_VEC_ELT (operand_entry_t, subops[i], j, oe1)
{
oecount c;
void **slot;
htab_delete (ctable);
/* Sort the counting table. */
- qsort (VEC_address (oecount, cvec), VEC_length (oecount, cvec),
- sizeof (oecount), oecount_cmp);
+ VEC_qsort (oecount, cvec, oecount_cmp);
if (dump_file && (dump_flags & TDF_DETAILS))
{
oecount *c;
fprintf (dump_file, "Candidates:\n");
- for (j = 0; VEC_iterate (oecount, cvec, j, c); ++j)
+ FOR_EACH_VEC_ELT (oecount, cvec, j, c)
{
fprintf (dump_file, " %u %s: ", c->cnt,
c->oecode == MULT_EXPR
if (oecode != c->oecode)
continue;
- for (j = 0; VEC_iterate (operand_entry_t, subops[i], j, oe1); ++j)
+ FOR_EACH_VEC_ELT (operand_entry_t, subops[i], j, oe1)
{
if (oe1->op == c->op)
{
}
zero_one_operation (&oe2->op, c->oecode, c->op);
sum = build_and_add_sum (tmpvar, oe1->op, oe2->op, opcode);
- oe2->op = fold_convert (TREE_TYPE (oe2->op), integer_zero_node);
+ oe2->op = build_zero_cst (TREE_TYPE (oe2->op));
oe2->rank = 0;
oe1->op = gimple_get_lhs (sum);
}
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: ");
enum tree_code subcode;
tree newop1;
tree newop2;
+ gcc_assert (COMPARISON_CLASS_P (t));
tmpvar = create_tmp_var (TREE_TYPE (t), NULL);
add_referenced_var (tmpvar);
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 (tmpvar, newop1, newop2, subcode);
curr->op = gimple_get_lhs (sum);
}
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. */
+
+static void
+init_range_entry (struct range_entry *r, tree exp)
+{
+ 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 (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 = build_int_cst (TREE_TYPE (exp), 0);
+ high = low;
+ in_p = 0;
+ strict_overflow_p = false;
+ is_bool = false;
+ 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)
+ {
+ gimple stmt;
+ enum tree_code code;
+ tree arg0, arg1, exp_type;
+ tree nexp;
+ location_t loc;
+
+ if (TREE_CODE (exp) != SSA_NAME)
+ break;
+
+ stmt = SSA_NAME_DEF_STMT (exp);
+ if (!is_gimple_assign (stmt))
+ break;
+
+ code = gimple_assign_rhs_code (stmt);
+ arg0 = gimple_assign_rhs1 (stmt);
+ if (TREE_CODE (arg0) != SSA_NAME)
+ break;
+ arg1 = gimple_assign_rhs2 (stmt);
+ exp_type = TREE_TYPE (exp);
+ loc = gimple_location (stmt);
+ switch (code)
+ {
+ case BIT_NOT_EXPR:
+ if (TREE_CODE (TREE_TYPE (exp)) == BOOLEAN_TYPE)
+ {
+ 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 (p->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. */
+
+static bool
+update_range_test (struct range_entry *range, struct range_entry *otherrange,
+ unsigned int count, enum tree_code opcode,
+ VEC (operand_entry_t, heap) **ops, tree exp, bool in_p,
+ tree low, tree high, bool strict_overflow_p)
+{
+ tree op = VEC_index (operand_entry_t, *ops, range->idx)->op;
+ location_t loc = gimple_location (SSA_NAME_DEF_STMT (op));
+ tree tem = build_range_check (loc, TREE_TYPE (op), 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)
+ tem = invert_truthvalue_loc (loc, tem);
+
+ tem = fold_convert_loc (loc, TREE_TYPE (op), tem);
+ gsi = gsi_for_stmt (SSA_NAME_DEF_STMT (op));
+ tem = force_gimple_operand_gsi (&gsi, tem, true, NULL_TREE, true,
+ GSI_SAME_STMT);
+
+ VEC_index (operand_entry_t, *ops, range->idx)->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++)
+ {
+ VEC_index (operand_entry_t, *ops, range->idx)->op = error_mark_node;
+ range->exp = NULL_TREE;
+ }
+ return true;
+}
+
+/* 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. */
+
+static void
+optimize_range_tests (enum tree_code opcode,
+ VEC (operand_entry_t, heap) **ops)
+{
+ unsigned int length = VEC_length (operand_entry_t, *ops), i, j, first;
+ operand_entry_t oe;
+ struct range_entry *ranges;
+ bool any_changes = false;
+
+ if (length == 1)
+ return;
+
+ ranges = XNEWVEC (struct range_entry, length);
+ for (i = 0; i < length; i++)
+ {
+ ranges[i].idx = i;
+ init_range_entry (ranges + i, VEC_index (operand_entry_t, *ops, i)->op);
+ /* For | invert it now, we will invert it again before emitting
+ the optimized expression. */
+ if (opcode == 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;
+ }
+
+ /* 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 above loop into
+ (X - 2U) <= 1U && (X - 10U) <= 1U
+ and this loop can transform that into
+ ((X & ~8) - 2U) <= 1U. */
+ for (i = first; i < length; i++)
+ {
+ tree lowi, highi, lowj, highj, type, lowxor, highxor, tem, exp;
+
+ 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++)
+ {
+ if (ranges[j].exp == NULL_TREE)
+ continue;
+ if (ranges[i].exp != ranges[j].exp)
+ break;
+ if (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);
+ tem = fold_binary (GT_EXPR, boolean_type_node,
+ lowj, highi);
+ if (tem == NULL_TREE || !integer_onep (tem))
+ continue;
+ lowxor = fold_binary (BIT_XOR_EXPR, type, lowi, lowj);
+ if (lowxor == NULL_TREE || TREE_CODE (lowxor) != INTEGER_CST)
+ continue;
+ gcc_checking_assert (!integer_zerop (lowxor));
+ tem = fold_binary (MINUS_EXPR, type, lowxor,
+ build_int_cst (type, 1));
+ if (tem == NULL_TREE)
+ continue;
+ tem = fold_binary (BIT_AND_EXPR, type, lowxor, tem);
+ if (tem == NULL_TREE || !integer_zerop (tem))
+ continue;
+ highxor = fold_binary (BIT_XOR_EXPR, type, highi, highj);
+ if (!tree_int_cst_equal (lowxor, highxor))
+ continue;
+ tem = fold_build1 (BIT_NOT_EXPR, type, lowxor);
+ exp = fold_build2 (BIT_AND_EXPR, type, ranges[i].exp, tem);
+ lowj = fold_build2 (BIT_AND_EXPR, type, lowi, tem);
+ highj = fold_build2 (BIT_AND_EXPR, type, highi, tem);
+ if (update_range_test (ranges + i, ranges + j, 1, opcode, ops, exp,
+ ranges[i].in_p, lowj, highj,
+ ranges[i].strict_overflow_p
+ || ranges[j].strict_overflow_p))
+ {
+ any_changes = true;
+ break;
+ }
+ }
+ }
+
+ if (any_changes)
+ {
+ j = 0;
+ FOR_EACH_VEC_ELT (operand_entry_t, *ops, i, oe)
+ {
+ if (oe->op == error_mark_node)
+ continue;
+ else if (i != j)
+ VEC_replace (operand_entry_t, *ops, j, oe);
+ j++;
+ }
+ VEC_truncate (operand_entry_t, *ops, j);
+ }
+
+ XDELETEVEC (ranges);
+}
+
/* 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. */
}
}
+/* 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(operand_entry_t, heap) * ops,
+ unsigned int opindex, gimple stmt)
+{
+ operand_entry_t oe1, oe2, oe3;
+
+ oe1 = VEC_index (operand_entry_t, ops, opindex);
+ oe2 = VEC_index (operand_entry_t, ops, opindex + 1);
+ oe3 = VEC_index (operand_entry_t, 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;
+ }
+}
+
/* Recursively rewrite our linearized statements so that the operators
match those in OPS[OPINDEX], putting the computation in rank
order. */
tree rhs2 = gimple_assign_rhs2 (stmt);
operand_entry_t oe;
- /* If we have three operands left, then we want to make sure the one
- that gets the double binary op are the ones with the same rank.
-
- The alternative we try is to see if this 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.
-
- 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. */
+ /* If we have three operands left, then we want to make sure the ones
+ that get the double binary op are chosen wisely. */
if (opindex + 3 == VEC_length (operand_entry_t, ops))
- {
- operand_entry_t oe1, oe2, oe3;
-
- oe1 = VEC_index (operand_entry_t, ops, opindex);
- oe2 = VEC_index (operand_entry_t, ops, opindex + 1);
- oe3 = VEC_index (operand_entry_t, ops, opindex + 2);
-
- if ((oe1->rank == oe2->rank
- && oe2->rank != oe3->rank)
- || (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)
- || (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;
- }
- }
+ swap_ops_for_binary_stmt (ops, opindex, stmt);
/* The final recursion case for this function is that you have
exactly two operations left.
rewrite_expr_tree (SSA_NAME_DEF_STMT (rhs1), opindex + 1, ops, moved);
}
+/* 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(operand_entry_t, heap) * ops)
+{
+ enum tree_code opcode = gimple_assign_rhs_code (stmt);
+ int op_num = VEC_length (operand_entry_t, ops);
+ 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);
+ tree lhs_var;
+
+ /* 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]));
+
+ lhs_var = create_tmp_reg (TREE_TYPE (last_rhs1), NULL);
+ add_referenced_var (lhs_var);
+
+ 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 = VEC_index (operand_entry_t, 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 = VEC_index (operand_entry_t, ops, op_index--)->op;
+ op1 = VEC_index (operand_entry_t, 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 (lhs_var, 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. */
if (TREE_CODE (binlhs) == SSA_NAME)
{
binlhsdef = SSA_NAME_DEF_STMT (binlhs);
- binlhsisreassoc = is_reassociable_op (binlhsdef, rhscode, loop);
+ 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);
+ 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
unsigned int i = 0;
tree negate;
- for (i = 0; VEC_iterate (tree, plus_negates, i, negate); i++)
+ FOR_EACH_VEC_ELT (tree, plus_negates, i, negate)
{
gimple user = get_single_immediate_use (negate);
{
gimple stmt = gsi_stmt (gsi);
- if (is_gimple_assign (stmt))
+ if (is_gimple_assign (stmt)
+ && !stmt_could_throw_p (stmt))
{
tree lhs, rhs1, rhs2;
enum tree_code rhs_code = gimple_assign_rhs_code (stmt);
gimple_set_visited (stmt, true);
linearize_expr_tree (&ops, stmt, true, true);
- qsort (VEC_address (operand_entry_t, ops),
- VEC_length (operand_entry_t, ops),
- sizeof (operand_entry_t),
- sort_by_operand_rank);
+ VEC_qsort (operand_entry_t, ops, sort_by_operand_rank);
optimize_ops_list (rhs_code, &ops);
if (undistribute_ops_list (rhs_code, &ops,
loop_containing_stmt (stmt)))
{
- qsort (VEC_address (operand_entry_t, ops),
- VEC_length (operand_entry_t, ops),
- sizeof (operand_entry_t),
- sort_by_operand_rank);
+ VEC_qsort (operand_entry_t, ops, 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 (VEC_length (operand_entry_t, ops) == 1)
{
if (dump_file && (dump_flags & TDF_DETAILS))
}
}
else
- rewrite_expr_tree (stmt, 0, ops, false);
+ {
+ enum machine_mode mode = TYPE_MODE (TREE_TYPE (lhs));
+ int ops_num = VEC_length (operand_entry_t, ops);
+ int width = get_reassociation_width (ops_num, rhs_code, mode);
+
+ if (dump_file && (dump_flags & TDF_DETAILS))
+ fprintf (dump_file,
+ "Width = %d was chosen for reassociation\n", width);
+
+ if (width > 1
+ && VEC_length (operand_entry_t, ops) > 3)
+ rewrite_expr_tree_parallel (stmt, width, ops);
+ else
+ rewrite_expr_tree (stmt, 0, ops, false);
+ }
VEC_free (operand_entry_t, heap, ops);
}
operand_entry_t oe;
unsigned int i;
- for (i = 0; VEC_iterate (operand_entry_t, ops, i, oe); i++)
+ FOR_EACH_VEC_ELT (operand_entry_t, ops, i, oe)
{
fprintf (file, "Op %d -> rank: %d, tree: ", i, oe->rank);
print_generic_expr (file, oe->op, 0);
0, /* properties_provided */
0, /* properties_destroyed */
0, /* todo_flags_start */
- TODO_dump_func | TODO_ggc_collect | TODO_verify_ssa /* todo_flags_finish */
+ TODO_verify_ssa
+ | TODO_verify_flow
+ | TODO_ggc_collect /* todo_flags_finish */
}
};
-
OSDN Git Service
