/* Loop header copying on trees.
- Copyright (C) 2004 Free Software Foundation, Inc.
+ Copyright (C) 2004, 2005 Free Software Foundation, Inc.
This file is part of GCC.
/* Duplicates headers of loops if they are small enough, so that the statements
in the loop body are always executed when the loop is entered. This
- increases effectivity of code motion optimizations, and reduces the need
+ increases effectiveness of code motion optimizations, and reduces the need
for loop preconditioning. */
/* Check whether we should duplicate HEADER of LOOP. At most *LIMIT
if (header->aux)
return false;
- if (!header->succ)
- abort ();
- if (!header->succ->succ_next)
+ gcc_assert (EDGE_COUNT (header->succs) > 0);
+ if (single_succ_p (header))
return false;
- if (header->succ->succ_next->succ_next)
- return false;
- if (flow_bb_inside_loop_p (loop, header->succ->dest)
- && flow_bb_inside_loop_p (loop, header->succ->succ_next->dest))
+ if (flow_bb_inside_loop_p (loop, EDGE_SUCC (header, 0)->dest)
+ && flow_bb_inside_loop_p (loop, EDGE_SUCC (header, 1)->dest))
return false;
/* If this is not the original loop header, we want it to have just
one predecessor in order to match the && pattern. */
- if (header != loop->header
- && header->pred->pred_next)
+ if (header != loop->header && !single_pred_p (header))
return false;
last = last_stmt (header);
return true;
}
-/* Duplicates destinations of edges in BBS_TO_DUPLICATE. */
-
-static void
-duplicate_blocks (varray_type bbs_to_duplicate)
-{
- unsigned i;
- edge preheader_edge, e, e1;
- basic_block header, new_header;
- tree phi, new_phi, var;
-
- /* TODO: It should be quite easy to keep the dominance information
- up-to-date. */
- free_dominance_info (CDI_DOMINATORS);
-
- for (i = 0; i < VARRAY_ACTIVE_SIZE (bbs_to_duplicate); i++)
- {
- preheader_edge = VARRAY_GENERIC_PTR_NOGC (bbs_to_duplicate, i);
- header = preheader_edge->dest;
-
- if (!header->aux)
- abort ();
- header->aux = NULL;
-
- new_header = duplicate_block (header, preheader_edge);
-
- /* Create the phi nodes on on entry to new_header. */
- for (phi = phi_nodes (header), var = PENDING_STMT (preheader_edge);
- phi;
- phi = TREE_CHAIN (phi), var = TREE_CHAIN (var))
- {
- new_phi = create_phi_node (PHI_RESULT (phi), new_header);
- add_phi_arg (&new_phi, TREE_VALUE (var), preheader_edge);
- }
- PENDING_STMT (preheader_edge) = NULL;
-
- /* Add the phi arguments to the outgoing edges. */
- for (e = header->succ; e; e = e->succ_next)
- {
- for (e1 = new_header->succ; e1->dest != e->dest; e1 = e1->succ_next)
- continue;
-
- for (phi = phi_nodes (e->dest); phi; phi = TREE_CHAIN (phi))
- {
- tree def = PHI_ARG_DEF_FROM_EDGE (phi, e);
- add_phi_arg (&phi, def, e1);
- }
- }
- }
-
- calculate_dominance_info (CDI_DOMINATORS);
-
- rewrite_ssa_into_ssa ();
-}
-
/* Checks whether LOOP is a do-while style loop. */
static bool
unsigned i;
struct loop *loop;
basic_block header;
- edge preheader_edge;
- varray_type bbs_to_duplicate = NULL;
+ edge exit, entry;
+ basic_block *bbs, *copied_bbs;
+ unsigned n_bbs;
+ unsigned bbs_size;
+ gcov_type entry_count, body_count, total_count;
loops = loop_optimizer_init (dump_file);
if (!loops)
return;
+ rewrite_into_loop_closed_ssa (NULL);
/* We do not try to keep the information about irreducible regions
up-to-date. */
verify_loop_structure (loops);
#endif
+ bbs = xmalloc (sizeof (basic_block) * n_basic_blocks);
+ copied_bbs = xmalloc (sizeof (basic_block) * n_basic_blocks);
+ bbs_size = n_basic_blocks;
+
for (i = 1; i < loops->num; i++)
{
/* Copy at most 20 insns. */
int limit = 20;
loop = loops->parray[i];
- preheader_edge = loop_preheader_edge (loop);
- header = preheader_edge->dest;
+ if (!loop)
+ continue;
+ header = loop->header;
/* If the loop is already a do-while style one (either because it was
written as such, or because jump threading transformed it into one),
like while (a && b) {...}, where we want to have both of the conditions
copied. TODO -- handle while (a || b) - like cases, by not requiring
the header to have just a single successor and copying up to
- postdominator.
-
- We do not really copy the blocks immediately, so that we do not have
- to worry about updating loop structures, and also so that we do not
- have to rewrite variables out of and into ssa form for each block.
- Instead we just record the block into worklist and duplicate all of
- them at once. */
+ postdominator. */
+
+ exit = NULL;
+ n_bbs = 0;
while (should_duplicate_loop_header_p (header, loop, &limit))
{
- if (!bbs_to_duplicate)
- VARRAY_GENERIC_PTR_NOGC_INIT (bbs_to_duplicate, 10,
- "bbs_to_duplicate");
- VARRAY_PUSH_GENERIC_PTR_NOGC (bbs_to_duplicate, preheader_edge);
- header->aux = &header->aux;
-
- if (dump_file && (dump_flags & TDF_DETAILS))
- fprintf (dump_file,
- "Scheduled basic block %d for duplication.\n",
- header->index);
-
/* Find a successor of header that is inside a loop; i.e. the new
header after the condition is copied. */
- if (flow_bb_inside_loop_p (loop, header->succ->dest))
- preheader_edge = header->succ;
+ if (flow_bb_inside_loop_p (loop, EDGE_SUCC (header, 0)->dest))
+ exit = EDGE_SUCC (header, 0);
else
- preheader_edge = header->succ->succ_next;
- header = preheader_edge->dest;
+ exit = EDGE_SUCC (header, 1);
+ bbs[n_bbs++] = header;
+ gcc_assert (bbs_size > n_bbs);
+ header = exit->dest;
}
- }
- loop_optimizer_finalize (loops, NULL);
+ if (!exit)
+ continue;
- if (bbs_to_duplicate)
- {
- duplicate_blocks (bbs_to_duplicate);
- VARRAY_FREE (bbs_to_duplicate);
+ if (dump_file && (dump_flags & TDF_DETAILS))
+ fprintf (dump_file,
+ "Duplicating header of the loop %d up to edge %d->%d.\n",
+ loop->num, exit->src->index, exit->dest->index);
+
+ /* Ensure that the header will have just the latch as a predecessor
+ inside the loop. */
+ if (!single_pred_p (exit->dest))
+ exit = single_succ_edge (loop_split_edge_with (exit, NULL));
+
+ entry = loop_preheader_edge (loop);
+ entry_count = entry->src->count;
+ body_count = exit->dest->count;
+
+ if (!tree_duplicate_sese_region (entry, exit, bbs, n_bbs, copied_bbs))
+ {
+ fprintf (dump_file, "Duplication failed.\n");
+ continue;
+ }
+
+ /* Fix profiling info. Scaling is done in gcov_type arithmetic to
+ avoid losing information; this is slow, but is done at most
+ once per loop. We special case 0 to avoid division by 0;
+ probably other special cases exist. */
+ total_count = body_count + entry_count;
+ if (total_count == 0LL)
+ {
+ scale_bbs_frequencies_int (bbs, n_bbs, 0, 1);
+ scale_bbs_frequencies_int (copied_bbs, n_bbs, 0, 1);
+ }
+ else
+ {
+ scale_bbs_frequencies_gcov_type (bbs, n_bbs, body_count, total_count);
+ scale_bbs_frequencies_gcov_type (copied_bbs, n_bbs, entry_count,
+ total_count);
+ }
+
+ /* Ensure that the latch and the preheader is simple (we know that they
+ are not now, since there was the loop exit condition. */
+ loop_split_edge_with (loop_preheader_edge (loop), NULL);
+ loop_split_edge_with (loop_latch_edge (loop), NULL);
}
- /* Run cleanup_tree_cfg here regardless of whether we have done anything, so
- that we cleanup the blocks created in order to get the loops into a
- canonical shape. */
- cleanup_tree_cfg ();
+ free (bbs);
+ free (copied_bbs);
+
+#ifdef ENABLE_CHECKING
+ verify_loop_closed_ssa ();
+#endif
+
+ loop_optimizer_finalize (loops, NULL);
}
static bool
NULL, /* next */
0, /* static_pass_number */
TV_TREE_CH, /* tv_id */
- PROP_cfg | PROP_ssa | PROP_alias, /* properties_required */
+ PROP_cfg | PROP_ssa, /* properties_required */
0, /* properties_provided */
0, /* properties_destroyed */
0, /* todo_flags_start */
- (TODO_dump_func
- | TODO_verify_ssa) /* todo_flags_finish */
+ TODO_cleanup_cfg | TODO_dump_func
+ | TODO_verify_ssa, /* todo_flags_finish */
+ 0 /* letter */
};