#endif
struct obstack *graveyard_obst = NULL;
struct obstack *rebirth_obst = NULL;
- assert(! edges_activated(irg) && "dead node elimination requires disabled edges");
+
+ edges_deactivate(irg);
/* inform statistics that we started a dead-node elimination run */
hook_dead_node_elim(irg, 1);
* inlined procedure. The new entities must be in the link field of
* the entities.
*/
-static INLINE void
-copy_node_inline(ir_node *n, void *env) {
+static void copy_node_inline(ir_node *n, void *env) {
ir_node *nn;
ir_type *frame_tp = (ir_type *)env;
}
}
+/**
+ * Copies new predecessors of old node and move constants to
+ * the Start Block.
+ */
+static void copy_preds_inline(ir_node *n, void *env) {
+ ir_node *nn;
+
+ copy_preds(n, env);
+ nn = skip_Id(get_new_node(n));
+ if (is_irn_constlike(nn)) {
+ /* move Constants into the start block */
+ set_nodes_block(nn, get_irg_start_block(current_ir_graph));
+
+ n = identify_remember(current_ir_graph->value_table, nn);
+ if (nn != n) {
+ exchange(nn, n);
+ }
+ }
+}
+
/**
* Walker: checks if P_value_arg_base is used.
*/
ir_node *pre_call;
ir_node *post_call, *post_bl;
ir_node *in[pn_Start_max];
- ir_node *end, *end_bl;
+ ir_node *end, *end_bl, *block;
ir_node **res_pred;
ir_node **cf_pred;
+ ir_node **args_in;
ir_node *ret, *phi;
- int arity, n_ret, n_exc, n_res, i, n, j, rem_opt, irn_arity;
+ int arity, n_ret, n_exc, n_res, i, n, j, rem_opt, irn_arity, n_params;
enum exc_mode exc_handling;
- ir_type *called_frame, *curr_frame;
+ ir_type *called_frame, *curr_frame, *mtp, *ctp;
ir_entity *ent;
ir_graph *rem, *irg;
irg_inline_property prop = get_irg_inline_property(called_graph);
ent = get_irg_entity(called_graph);
- /* Do not inline variadic functions. */
- if (get_method_variadicity(get_entity_type(ent)) == variadicity_variadic) {
- /* Arg, KR functions are marked as variadic one's, so check further */
- ir_type *mtp = get_entity_type(ent);
- ir_type *ctp = get_Call_type(call);
- int n_params = get_method_n_params(mtp);
- int i;
-
- /* This is too strong, but probably ok. Function calls with a wrong number of
- parameters should not be inlined. */
- if (n_params != get_method_n_params(ctp))
- return 0;
-
- /* check types: for K&R calls, this was not done by the compiler. Again, this is
- too strong, but ok for now. */
- for (i = n_params - 1; i >= 0; --i) {
- ir_type *param_tp = get_method_param_type(mtp, i);
- ir_type *arg_tp = get_method_param_type(ctp, i);
+ mtp = get_entity_type(ent);
+ ctp = get_Call_type(call);
+ if (get_method_n_params(mtp) > get_method_n_params(ctp)) {
+ /* this is a bad feature of C: without a prototype, we can can call a function with less
+ parameters than needed. Currently we don't support this, although it would be
+ to use Unknown than. */
+ return 0;
+ }
- if (param_tp != arg_tp)
+ /* Argh, compiling C has some bad consequences:
+ the call type AND the method type might be different.
+ It is implementation defendant what happens in that case.
+ We support inlining, if the bitsize of the types matches AND
+ the same arithmetic is used. */
+ n_params = get_method_n_params(mtp);
+ for (i = n_params - 1; i >= 0; --i) {
+ ir_type *param_tp = get_method_param_type(mtp, i);
+ ir_type *arg_tp = get_method_param_type(ctp, i);
+
+ if (param_tp != arg_tp) {
+ ir_mode *pmode = get_type_mode(param_tp);
+ ir_mode *amode = get_type_mode(arg_tp);
+
+ if (pmode == NULL || amode == NULL)
+ return 0;
+ if (get_mode_size_bits(pmode) != get_mode_size_bits(amode))
return 0;
+ if (get_mode_arithmetic(pmode) != get_mode_arithmetic(amode))
+ return 0;
+ /* otherwise we can simply "reinterpret" the bits */
}
- DB((dbg, LEVEL_1, "Inlining allowed for variadic function %+F\n", called_graph));
- /* types match, fine: when the frame is access, the inliner stops at can_inline() */
}
- assert(get_method_n_params(get_entity_type(ent)) ==
- get_method_n_params(get_Call_type(call)));
-
irg = get_irn_irg(call);
/*
else { exc_handling = exc_no_handler; } /* !Mproj && !Xproj */
}
+ /* create the argument tuple */
+ NEW_ARR_A(ir_type *, args_in, n_params);
+
+ block = get_nodes_block(call);
+ for (i = n_params - 1; i >= 0; --i) {
+ ir_node *arg = get_Call_param(call, i);
+ ir_type *param_tp = get_method_param_type(mtp, i);
+ ir_mode *mode = get_type_mode(param_tp);
+
+ if (mode != get_irn_mode(arg)) {
+ arg = new_r_Conv(irg, block, arg, mode);
+ }
+ args_in[i] = arg;
+ }
+
/* --
the procedure and later replaces the Start node of the called graph.
Post_call is the old Call node and collects the results of the called
in[pn_Start_M] = get_Call_mem(call);
in[pn_Start_P_frame_base] = get_irg_frame(irg);
in[pn_Start_P_tls] = get_irg_tls(irg);
- in[pn_Start_T_args] = new_Tuple(get_Call_n_params(call), get_Call_param_arr(call));
+ in[pn_Start_T_args] = new_Tuple(n_params, args_in);
/* in[pn_Start_P_value_arg_base] = ??? */
assert(pn_Start_P_value_arg_base == pn_Start_max - 1 && "pn_Start_P_value_arg_base not supported, fix");
pre_call = new_Tuple(pn_Start_max - 1, in);
/* Visited flags in calling irg must be >= flag in called irg.
Else walker and arity computation will not work. */
if (get_irg_visited(irg) <= get_irg_visited(called_graph))
- set_irg_visited(irg, get_irg_visited(called_graph)+1);
+ set_irg_visited(irg, get_irg_visited(called_graph) + 1);
if (get_irg_block_visited(irg) < get_irg_block_visited(called_graph))
set_irg_block_visited(irg, get_irg_block_visited(called_graph));
/* Set pre_call as new Start node in link field of the start node of
/* -- Performing dead node elimination inlines the graph -- */
/* Copies the nodes to the obstack of current_ir_graph. Updates links to new
entities. */
- irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds,
+ irg_walk(get_irg_end(called_graph), copy_node_inline, copy_preds_inline,
get_irg_frame_type(called_graph));
/* Repair called_graph */
int n_callers_orig; /**< for statistics */
unsigned got_inline:1; /**< Set, if at least one call inside this graph was inlined. */
unsigned local_vars:1; /**< Set, if a inlined function gets the address of an inlined variable. */
+ unsigned recursive:1; /**< Set, if this function is self recursive. */
call_entry *call_head; /**< The head of the list of all call nodes in this graph. */
call_entry *call_tail; /**< The tail of the list of all call nodes in this graph .*/
unsigned *local_weights; /**< Once allocated, the beneficial weight for transmitting local addresses. */
env->n_callers_orig = 0;
env->got_inline = 0;
env->local_vars = 0;
+ env->recursive = 0;
env->local_weights = NULL;
return env;
}
++callee_env->n_callers;
++callee_env->n_callers_orig;
}
+ if (callee == current_ir_graph)
+ x->recursive = 1;
/* link it in the list of possible inlinable entries */
entry = obstack_alloc(&temp_obst, sizeof(*entry));
}
/**
- * calculate a benefice value for inlining the given call.
+ * Calculate a benefice value for inlining the given call.
+ *
+ * @param call the call node we have to inspect
+ * @param callee the called graph
+ * @param local_adr set after return if an address of a local variable is
+ * transmitted as a parameter
*/
static int calc_inline_benefice(ir_node *call, ir_graph *callee, unsigned *local_adr) {
ir_entity *ent = get_irg_entity(callee);
ir_node *frame_ptr;
ir_type *mtp;
int weight = 0;
- int i, n_params;
+ int i, n_params, all_const;
unsigned cc, v;
inline_irg_env *curr_env, *callee_env;
/* constant parameters improve the benefice */
frame_ptr = get_irg_frame(current_ir_graph);
+ all_const = 1;
for (i = 0; i < n_params; ++i) {
ir_node *param = get_Call_param(call, i);
- if (is_Const(param) || is_SymConst(param))
+ if (is_Const(param)) {
weight += get_method_param_weight(ent, i);
- else if (is_Sel(param) && get_Sel_ptr(param) == frame_ptr) {
- /*
- * An address of a local variable is transmitted. After inlining,
- * scalar_replacement might be able to remove the local variable,
- * so honor this.
- */
- v = get_method_local_adress_weight(callee, i);
- weight += v;
- if (v > 0)
- *local_adr = 1;
+ } else {
+ all_const = 0;
+ if (is_SymConst(param))
+ weight += get_method_param_weight(ent, i);
+ else if (is_Sel(param) && get_Sel_ptr(param) == frame_ptr) {
+ /*
+ * An address of a local variable is transmitted. After inlining,
+ * scalar_replacement might be able to remove the local variable,
+ * so honor this.
+ */
+ v = get_method_local_adress_weight(callee, i);
+ weight += v;
+ if (v > 0)
+ *local_adr = 1;
+ }
}
}
/* reduce the benefice if the current function is already big */
curr_env = get_irg_link(current_ir_graph);
- weight -= curr_env->n_nodes / 100;
+ weight -= curr_env->n_nodes / 50;
/* give a bonus for functions with one block */
if (callee_env->n_blocks == 1)
weight = weight * 3 / 2;
- /* and one for small functions: we want them to be inlined in mostly every case */
- else if (callee_env->n_nodes < 20)
+ /* and one for small non-recursive functions: we want them to be inlined in mostly every case */
+ else if (callee_env->n_nodes < 20 && !callee_env->recursive)
weight += 5000;
/* and finally for leaves: they do not increase the register pressure
else if (callee_env->n_call_nodes == 0)
weight += 25;
+ /*
+ * Reduce the weight for recursive function IFF not all arguments are const.
+ * inlining recursive functions is rarely good.
+ */
+ if (callee_env->recursive && !all_const)
+ weight -= 500;
+
+ /*
+ * All arguments constant is probably a good sign, give an extra bonus
+ */
+ if (all_const)
+ weight += 100;
+
return weight;
}