}
}
+typedef enum tail_rec_variants {
+ TR_DIRECT, /**< direct return value, i.e. return func(). */
+ TR_ADD, /**< additive return value, i.e. return x +/- func() */
+ TR_MUL, /**< multiplicative return value, i.e. return x * func() or return -func() */
+ TR_BAD, /**< any other transformation */
+ TR_UNKNOWN /**< during construction */
+} tail_rec_variants;
+
+typedef struct tr_env {
+ int n_tail_calls; /**< number of tail calls found */
+ int n_ress; /**< number of return values */
+ tail_rec_variants *variants; /**< return value variants */
+ ir_node *rets; /**< list of returns that can be transformed */
+} tr_env;
+
+
/**
* do the graph reconstruction for tail-recursion elimination
*
* @param rets linked list of all rets
* @param n_tail_calls number of tail-recursion calls
*/
-static void do_opt_tail_rec(ir_graph *irg, ir_node *rets, int n_tail_calls) {
+static void do_opt_tail_rec(ir_graph *irg, tr_env *env) {
ir_node *end_block = get_irg_end_block(irg);
ir_node *block, *jmp, *call, *calls;
ir_node **in;
ir_node **phis;
ir_node ***call_params;
ir_node *p, *n;
- int i, j, n_params;
+ int i, j, n_params, n_locs;
collect_t data;
int rem = get_optimize();
ir_entity *ent = get_irg_entity(irg);
ir_type *method_tp = get_entity_type(ent);
+ ir_graph *old = current_ir_graph;
+
+ current_ir_graph = irg;
- assert(n_tail_calls);
+ assert(env->n_tail_calls > 0);
/* we add new nodes, so the outs are inconsistent */
set_irg_outs_inconsistent(irg);
assert((data.proj_data || n_params == 0) && "Could not find Proj(ProjT(Start)) of non-void function");
/* allocate in's for phi and block construction */
- NEW_ARR_A(ir_node *, in, n_tail_calls + 1);
+ NEW_ARR_A(ir_node *, in, env->n_tail_calls + 1);
in[0] = data.proj_X;
/* turn Return's into Jmp's */
- for (i = 1, p = rets; p; p = n) {
+ for (i = 1, p = env->rets; p; p = n) {
ir_node *block = get_nodes_block(p);
n = get_irn_link(p);
in[i++] = new_r_Jmp(irg, block);
- exchange(p, new_r_Bad(irg));
+ // exchange(p, new_r_Bad(irg));
/* we might generate an endless loop, so add
* the block to the keep-alive list */
}
/* create a new block at start */
- block = new_r_Block(irg, n_tail_calls + 1, in);
+ block = new_r_Block(irg, env->n_tail_calls + 1, in);
jmp = new_r_Jmp(irg, block);
/* the old first block is now the second one */
in[i] = get_Call_mem(calls);
++i;
}
- assert(i == n_tail_calls + 1);
+ assert(i == env->n_tail_calls + 1);
- phis[0] = new_r_Phi(irg, block, n_tail_calls + 1, in, mode_M);
+ phis[0] = new_r_Phi(irg, block, env->n_tail_calls + 1, in, mode_M);
/* build the data Phi's */
if (n_params > 0) {
ir_node *args;
ir_node *args_bl;
- NEW_ARR_A(ir_node **, call_params, n_tail_calls);
+ NEW_ARR_A(ir_node **, call_params, env->n_tail_calls);
/* collect all parameters */
for (i = 0, calls = call; calls; calls = get_irn_link(calls)) {
ir_mode *mode = get_type_mode(get_method_param_type(method_tp, i));
in[0] = new_r_Proj(irg, args_bl, args, mode, i);
- for (j = 0; j < n_tail_calls; ++j)
+ for (j = 0; j < env->n_tail_calls; ++j)
in[j + 1] = call_params[j][i];
- phis[i + 1] = new_r_Phi(irg, block, n_tail_calls + 1, in, mode);
+ phis[i + 1] = new_r_Phi(irg, block, env->n_tail_calls + 1, in, mode);
}
}
set_irg_callee_info_state(irg, irg_callee_info_inconsistent);
set_optimize(rem);
+
+ /* check if we need new values */
+ n_locs = 0;
+ for (i = 0; i < env->n_ress; ++i) {
+ if (env->variants[i] != TR_DIRECT)
+ ++n_locs;
+ }
+
+ if (n_locs > 0) {
+ ir_node *bad, *start_block;
+ ir_node **in;
+ ir_mode **modes;
+
+ NEW_ARR_A(ir_node *, in, n_locs);
+ NEW_ARR_A(ir_mode *, modes, n_locs);
+ ssa_cons_start(irg, n_locs);
+
+ start_block = get_irg_start_block(irg);
+ set_cur_block(start_block);
+
+ for (i = 0; i < env->n_ress; ++i) {
+ ir_type *tp = get_method_res_type(method_tp, i);
+ ir_mode *mode = get_type_mode(tp);
+
+ modes[i] = mode;
+ if (env->variants[i] == TR_ADD) {
+ set_value(i, new_Const(mode, get_mode_null(mode)));
+ } else if (env->variants[i] == TR_MUL) {
+ set_value(i, new_Const(mode, get_mode_one(mode)));
+ }
+ }
+ mature_immBlock(start_block);
+
+ /* no: we can kill all returns */
+ bad = get_irg_bad(irg);
+
+ for (p = env->rets; p; p = n) {
+ ir_node *block = get_nodes_block(p);
+ ir_node *call, *mem, *jmp, *tuple;
+
+ set_cur_block(block);
+ n = get_irn_link(p);
+
+ call = skip_Proj(get_Return_mem(p));
+ assert(is_Call(call));
+
+ mem = get_Call_mem(call);
+
+ /* create a new jump, free of CSE */
+ set_optimize(0);
+ jmp = new_Jmp();
+ set_optimize(rem);
+
+ for (i = 0; i < env->n_ress; ++i) {
+ if (env->variants[i] != TR_DIRECT) {
+ in[i] = get_value(i, modes[i]);
+ } else {
+ in[i] = bad;
+ }
+ }
+ /* create a new tuple for the return values */
+ tuple = new_Tuple(env->n_ress, in);
+
+ turn_into_tuple(call, pn_Call_max);
+ set_Tuple_pred(call, pn_Call_M, mem);
+ set_Tuple_pred(call, pn_Call_X_regular, jmp);
+ set_Tuple_pred(call, pn_Call_X_except, bad);
+ set_Tuple_pred(call, pn_Call_T_result, tuple);
+ set_Tuple_pred(call, pn_Call_M_except, mem);
+ set_Tuple_pred(call, pn_Call_P_value_res_base, bad);
+
+ for (i = 0; i < env->n_ress; ++i) {
+ ir_node *res = get_Return_res(p, i);
+ if (env->variants[i] != TR_DIRECT) {
+ set_value(i, res);
+ }
+ }
+
+ exchange(p, bad);
+ }
+
+ /* finally fix all other returns */
+ end_block = get_irg_end_block(irg);
+ for (i = get_Block_n_cfgpreds(end_block) - 1; i >= 0; --i) {
+ ir_node *ret = get_Block_cfgpred(end_block, i);
+
+ /* search all Returns of a block */
+ if (! is_Return(ret))
+ continue;
+
+ set_cur_block(get_nodes_block(ret));
+ for (j = 0; j < env->n_ress; ++j) {
+ ir_node *pred = get_Return_res(ret, j);
+ ir_node *n;
+
+ switch (env->variants[j]) {
+ case TR_DIRECT:
+ continue;
+
+ case TR_ADD:
+ n = get_value(j, modes[j]);
+ n = new_Add(n, pred, modes[j]);
+ set_Return_res(ret, j, n);
+ break;
+
+ case TR_MUL:
+ n = get_value(j, modes[j]);
+ n = new_Mul(n, pred, modes[j]);
+ set_Return_res(ret, j, n);
+ break;
+
+ default:
+ assert(!"unexpected tail recursion variant");
+ }
+ }
+ }
+ ssa_cons_finish(irg);
+ } else {
+ ir_node *bad = get_irg_bad(irg);
+
+ /* no: we can kill all returns */
+ for (p = env->rets; p; p = n) {
+ n = get_irn_link(p);
+ exchange(p, bad);
+ }
+ }
+ current_ir_graph = old;
}
/**
return 1;
}
+/**
+ * Examine irn and detect the recursion variant.
+ */
+static tail_rec_variants find_variant(ir_node *irn, ir_node *call) {
+ ir_node *a, *b;
+ tail_rec_variants va, vb, res;
+
+ if (skip_Proj(skip_Proj(irn)) == call) {
+ /* found it */
+ return TR_DIRECT;
+ }
+ switch (get_irn_opcode(irn)) {
+ case iro_Add:
+ /* try additive */
+ a = get_Add_left(irn);
+ if (get_irn_MacroBlock(a) != get_irn_MacroBlock(call)) {
+ /* we are outside, ignore */
+ va = TR_UNKNOWN;
+ } else {
+ va = find_variant(a, call);
+ if (va == TR_BAD)
+ return TR_BAD;
+ }
+ b = get_Add_right(irn);
+ if (get_irn_MacroBlock(b) != get_irn_MacroBlock(call)) {
+ /* we are outside, ignore */
+ vb = TR_UNKNOWN;
+ } else {
+ vb = find_variant(b, call);
+ if (vb == TR_BAD)
+ return TR_BAD;
+ }
+ if (va == vb) {
+ res = va;
+ }
+ else if (va == TR_UNKNOWN)
+ res = vb;
+ else if (vb == TR_UNKNOWN)
+ res = va;
+ else {
+ /* they are different but none is TR_UNKNOWN -> incompatible */
+ return TR_BAD;
+ }
+ if (res == TR_DIRECT || res == TR_ADD)
+ return TR_ADD;
+ /* not compatible */
+ return TR_BAD;
+
+ case iro_Sub:
+ /* try additive, but return value mut be left */
+ a = get_Sub_left(irn);
+ if (get_irn_MacroBlock(a) != get_irn_MacroBlock(call)) {
+ /* we are outside, ignore */
+ va = TR_UNKNOWN;
+ } else {
+ va = find_variant(a, call);
+ if (va == TR_BAD)
+ return TR_BAD;
+ }
+ b = get_Sub_right(irn);
+ if (get_irn_MacroBlock(b) != get_irn_MacroBlock(call)) {
+ /* we are outside, ignore */
+ vb = TR_UNKNOWN;
+ } else {
+ vb = find_variant(b, call);
+ if (vb != TR_UNKNOWN)
+ return TR_BAD;
+ }
+ res = va;
+ if (res == TR_DIRECT || res == TR_ADD)
+ return res;
+ /* not compatible */
+ return TR_BAD;
+
+ case iro_Mul:
+ /* try multiplicative */
+ a = get_Mul_left(irn);
+ if (get_irn_MacroBlock(a) != get_irn_MacroBlock(call)) {
+ /* we are outside, ignore */
+ va = TR_UNKNOWN;
+ } else {
+ va = find_variant(a, call);
+ if (va == TR_BAD)
+ return TR_BAD;
+ }
+ b = get_Mul_right(irn);
+ if (get_irn_MacroBlock(b) != get_irn_MacroBlock(call)) {
+ /* we are outside, ignore */
+ vb = TR_UNKNOWN;
+ } else {
+ vb = find_variant(b, call);
+ if (vb == TR_BAD)
+ return TR_BAD;
+ }
+ if (va == vb) {
+ res = va;
+ }
+ else if (va == TR_UNKNOWN)
+ res = vb;
+ else if (vb == TR_UNKNOWN)
+ res = va;
+ else {
+ /* they are different but none is TR_UNKNOWN -> incompatible */
+ return TR_BAD;
+ }
+ if (res == TR_DIRECT || res == TR_MUL)
+ return TR_MUL;
+ /* not compatible */
+ return TR_BAD;
+
+ case iro_Minus:
+ /* try multiplicative */
+ a = get_Minus_op(irn);
+ res = find_variant(a, call);
+ if (res == TR_DIRECT)
+ return TR_MUL;
+ if (res == TR_MUL || res == TR_UNKNOWN)
+ return res;
+ /* not compatible */
+ return TR_BAD;
+
+ default:
+ return TR_UNKNOWN;
+ }
+}
+
+
/*
* convert simple tail-calls into loops
*/
int opt_tail_rec_irg(ir_graph *irg) {
- ir_node *end_block;
- int i, n_tail_calls = 0;
- ir_node *rets = NULL;
- ir_type *mtd_type, *call_type;
+ tr_env env;
+ ir_node *end_block;
+ int i, n_ress, n_tail_calls = 0;
+ ir_node *rets = NULL;
+ ir_type *mtd_type, *call_type;
+ ir_entity *ent;
assure_irg_outs(irg);
if (! check_lifetime_of_locals(irg))
return 0;
+
+ ent = get_irg_entity(irg);
+ mtd_type = get_entity_type(ent);
+ n_ress = get_method_n_ress(mtd_type);
+
+ env.variants = NULL;
+ env.n_ress = n_ress;
+
+ if (n_ress > 0) {
+ NEW_ARR_A(tail_rec_variants, env.variants, n_ress);
+
+ for (i = 0; i < n_ress; ++i)
+ env.variants[i] = TR_DIRECT;
+ }
+
/*
* This tail recursion optimization works best
* if the Returns are normalized.
for (i = get_Block_n_cfgpreds(end_block) - 1; i >= 0; --i) {
ir_node *ret = get_Block_cfgpred(end_block, i);
ir_node *call, *call_ptr;
- ir_entity *ent;
int j;
ir_node **ress;
if (!ent || get_entity_irg(ent) != irg)
continue;
- /* ok, mem is routed to a recursive call, check return args */
- ress = get_Return_res_arr(ret);
- for (j = get_Return_n_ress(ret) - 1; j >= 0; --j) {
- ir_node *irn = skip_Proj(skip_Proj(ress[j]));
-
- if (irn != call) {
- /* not routed to a call */
- break;
- }
- }
- if (j >= 0)
- continue;
-
/*
* Check, that the types match. At least in C
* this might fail.
continue;
}
+ /* ok, mem is routed to a recursive call, check return args */
+ ress = get_Return_res_arr(ret);
+ for (j = get_Return_n_ress(ret) - 1; j >= 0; --j) {
+ tail_rec_variants var = find_variant(ress[j], call);
+
+ if (var >= TR_BAD) {
+ /* cannot be transformed */
+ break;
+ }
+ if (var == TR_DIRECT)
+ var = env.variants[j];
+ else if (env.variants[j] == TR_DIRECT)
+ env.variants[j] = var;
+ if (env.variants[j] != var) {
+ /* not compatible */
+ break;
+ }
+ }
+ if (j >= 0)
+ continue;
+
/* here, we have found a call */
set_irn_link(call, get_irn_link(end_block));
set_irn_link(end_block, call);
}
/* now, end_block->link contains the list of all tail calls */
- if (! n_tail_calls)
+ if (n_tail_calls <= 0)
return 0;
DB((dbg, LEVEL_2, " Performing tail recursion for graph %s and %d Calls\n",
get_entity_ld_name(get_irg_entity(irg)), n_tail_calls));
hook_tail_rec(irg, n_tail_calls);
- do_opt_tail_rec(irg, rets, n_tail_calls);
+
+ env.n_tail_calls = n_tail_calls;
+ env.rets = rets;
+ do_opt_tail_rec(irg, &env);
return n_tail_calls;
}
projects / libfirm / commitdiff