3 * File name: ir/ir/iropt.c
4 * Purpose: iropt --- optimizations intertwined with IR construction.
5 * Author: Christian Schaefer
6 * Modified by: Goetz Lindenmaier
9 * Copyright: (c) 1998-2003 Universität Karlsruhe
10 * Licence: This file protected by GPL - GNU GENERAL PUBLIC LICENSE.
17 # include "irnode_t.h"
18 # include "irgraph_t.h"
19 # include "irmode_t.h"
25 # include "dbginfo_t.h"
26 # include "iropt_dbg.h"
27 # include "irflag_t.h"
28 # include "firmstat.h"
30 /* Make types visible to allow most efficient access */
31 # include "entity_t.h"
34 * Trivial INLINEable routine for copy propagation.
35 * Does follow Ids, needed to optimize INLINEd code.
37 static INLINE ir_node *
38 follow_Id (ir_node *n)
40 while (get_irn_op (n) == op_Id) n = get_Id_pred (n);
45 * Returns the tarval of a Const node or tarval_bad for all other nodes.
47 static INLINE tarval *
50 if ((n != NULL) && (get_irn_op(n) == op_Const))
51 return get_Const_tarval(n); /* might return tarval_bad */
56 static tarval *computed_value_Const(ir_node *n)
58 return get_Const_tarval(n);
61 static tarval *computed_value_SymConst(ir_node *n)
63 if ((get_SymConst_kind(n) == size) &&
64 (get_type_state(get_SymConst_type(n))) == layout_fixed)
65 return new_tarval_from_long (get_type_size(get_SymConst_type(n)), mode_Is);
69 static tarval *computed_value_Add(ir_node *n)
71 ir_node *a = get_Add_left(n);
72 ir_node *b = get_Add_right(n);
74 tarval *ta = value_of(a);
75 tarval *tb = value_of(b);
77 if ((ta != tarval_bad) && (tb != tarval_bad)
78 && (get_irn_mode(a) == get_irn_mode(b))
79 && !(get_mode_sort(get_irn_mode(a)) == irms_reference)) {
80 return tarval_add(ta, tb);
85 static tarval *computed_value_Sub(ir_node *n)
87 ir_node *a = get_Sub_left(n);
88 ir_node *b = get_Sub_right(n);
90 tarval *ta = value_of(a);
91 tarval *tb = value_of(b);
93 if ((ta != tarval_bad) && (tb != tarval_bad)
94 && (get_irn_mode(a) == get_irn_mode(b))
95 && !(get_mode_sort(get_irn_mode(a)) == irms_reference)) {
96 return tarval_sub(ta, tb);
101 static tarval *computed_value_Minus(ir_node *n)
103 ir_node *a = get_Minus_op(n);
104 tarval *ta = value_of(a);
106 if ((ta != tarval_bad) && mode_is_signed(get_irn_mode(a)))
107 return tarval_neg(ta);
112 static tarval *computed_value_Mul(ir_node *n)
114 ir_node *a = get_Mul_left(n);
115 ir_node *b = get_Mul_right(n);
117 tarval *ta = value_of(a);
118 tarval *tb = value_of(b);
120 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
121 return tarval_mul(ta, tb);
123 /* a*0 = 0 or 0*b = 0:
124 calls computed_value recursive and returns the 0 with proper
128 if ( ( ((v = ta) != tarval_bad)
129 && (v == get_mode_null(get_tarval_mode(v))) )
130 || ( ((v = tb) != tarval_bad)
131 && (v == get_mode_null(get_tarval_mode(v))) )) {
138 static tarval *computed_value_Quot(ir_node *n)
140 ir_node *a = get_Quot_left(n);
141 ir_node *b = get_Quot_right(n);
143 tarval *ta = value_of(a);
144 tarval *tb = value_of(b);
146 /* This was missing in original implementation. Why? */
147 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
148 if (tb != get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
149 return tarval_quo(ta, tb);
154 static tarval *computed_value_Div(ir_node *n)
156 ir_node *a = get_Div_left(n);
157 ir_node *b = get_Div_right(n);
159 tarval *ta = value_of(a);
160 tarval *tb = value_of(b);
162 /* This was missing in original implementation. Why? */
163 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
164 if (tb != get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
165 return tarval_div(ta, tb);
170 static tarval *computed_value_Mod(ir_node *n)
172 ir_node *a = get_Mod_left(n);
173 ir_node *b = get_Mod_right(n);
175 tarval *ta = value_of(a);
176 tarval *tb = value_of(b);
178 /* This was missing in original implementation. Why? */
179 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
180 if (tb != get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
181 return tarval_mod(ta, tb);
186 static tarval *computed_value_Abs(ir_node *n)
188 ir_node *a = get_Abs_op(n);
189 tarval *ta = value_of(a);
191 if (ta != tarval_bad)
192 return tarval_abs(ta);
197 static tarval *computed_value_And(ir_node *n)
199 ir_node *a = get_And_left(n);
200 ir_node *b = get_And_right(n);
202 tarval *ta = value_of(a);
203 tarval *tb = value_of(b);
205 if ((ta != tarval_bad) && (tb != tarval_bad)) {
206 return tarval_and (ta, tb);
210 if ( (tarval_classify ((v = computed_value (a))) == TV_CLASSIFY_NULL)
211 || (tarval_classify ((v = computed_value (b))) == TV_CLASSIFY_NULL)) {
218 static tarval *computed_value_Or(ir_node *n)
220 ir_node *a = get_Or_left(n);
221 ir_node *b = get_Or_right(n);
223 tarval *ta = value_of(a);
224 tarval *tb = value_of(b);
226 if ((ta != tarval_bad) && (tb != tarval_bad)) {
227 return tarval_or (ta, tb);
230 if ( (tarval_classify ((v = computed_value (a))) == TV_CLASSIFY_ALL_ONE)
231 || (tarval_classify ((v = computed_value (b))) == TV_CLASSIFY_ALL_ONE)) {
238 static tarval *computed_value_Eor(ir_node *n)
240 ir_node *a = get_Eor_left(n);
241 ir_node *b = get_Eor_right(n);
243 tarval *ta = value_of(a);
244 tarval *tb = value_of(b);
246 if ((ta != tarval_bad) && (tb != tarval_bad)) {
247 return tarval_eor (ta, tb);
252 static tarval *computed_value_Not(ir_node *n)
254 ir_node *a = get_Not_op(n);
255 tarval *ta = value_of(a);
257 if (ta != tarval_bad)
258 return tarval_not(ta);
263 static tarval *computed_value_Shl(ir_node *n)
265 ir_node *a = get_Shl_left(n);
266 ir_node *b = get_Shl_right(n);
268 tarval *ta = value_of(a);
269 tarval *tb = value_of(b);
271 if ((ta != tarval_bad) && (tb != tarval_bad)) {
272 return tarval_shl (ta, tb);
277 static tarval *computed_value_Shr(ir_node *n)
279 ir_node *a = get_Shr_left(n);
280 ir_node *b = get_Shr_right(n);
282 tarval *ta = value_of(a);
283 tarval *tb = value_of(b);
285 if ((ta != tarval_bad) && (tb != tarval_bad)) {
286 return tarval_shr (ta, tb);
291 static tarval *computed_value_Shrs(ir_node *n)
293 ir_node *a = get_Shrs_left(n);
294 ir_node *b = get_Shrs_right(n);
296 tarval *ta = value_of(a);
297 tarval *tb = value_of(b);
299 if ((ta != tarval_bad) && (tb != tarval_bad)) {
300 return tarval_shrs (ta, tb);
305 static tarval *computed_value_Rot(ir_node *n)
307 ir_node *a = get_Rot_left(n);
308 ir_node *b = get_Rot_right(n);
310 tarval *ta = value_of(a);
311 tarval *tb = value_of(b);
313 if ((ta != tarval_bad) && (tb != tarval_bad)) {
314 /* return tarval_rot (ta, tb); */
319 static tarval *computed_value_Conv(ir_node *n)
321 ir_node *a = get_Conv_op(n);
322 tarval *ta = value_of(a);
324 if (ta != tarval_bad)
325 return tarval_convert_to(ta, get_irn_mode(n));
330 static tarval *computed_value_Proj(ir_node *n)
332 ir_node *a = get_Proj_pred(n), *b;
335 /* Optimize Cmp nodes.
336 This performs a first step of unreachable code elimination.
337 Proj can not be computed, but folding a Cmp above the Proj here is
338 not as wasteful as folding a Cmp into a Tuple of 16 Consts of which
340 There are several case where we can evaluate a Cmp node:
341 1. The nodes compared are both the same. If we compare for
342 equal, greater equal, ... this will return true, else it
343 will return false. This step relies on cse.
344 2. The predecessors of Cmp are target values. We can evaluate
346 3. The predecessors are Allocs or void* constants. Allocs never
347 return NULL, they raise an exception. Therefore we can predict
349 if (get_irn_op(a) == op_Cmp) {
350 aa = get_Cmp_left(a);
351 ab = get_Cmp_right(a);
353 if (aa == ab) { /* 1.: */
354 /* This is a tric with the bits used for encoding the Cmp
355 Proj numbers, the following statement is not the same:
356 return new_tarval_from_long ((get_Proj_proj(n) == Eq), mode_b) */
357 return new_tarval_from_long ((get_Proj_proj(n) & Eq), mode_b);
359 tarval *taa = computed_value (aa);
360 tarval *tab = computed_value (ab);
362 if ((taa != tarval_bad) && (tab != tarval_bad)) { /* 2.: */
363 /* strange checks... */
364 pnc_number flags = tarval_cmp (taa, tab);
365 if (flags != False) {
366 return new_tarval_from_long (get_Proj_proj(n) & flags, mode_b);
368 } else { /* check for 3.: */
369 ir_node *aaa = skip_nop(skip_Proj(aa));
370 ir_node *aba = skip_nop(skip_Proj(ab));
372 if ( ( (/* aa is ProjP and aaa is Alloc */
373 (get_irn_op(aa) == op_Proj)
374 && (mode_is_reference(get_irn_mode(aa)))
375 && (get_irn_op(aaa) == op_Alloc))
376 && ( (/* ab is constant void */
377 (get_irn_op(ab) == op_Const)
378 && (mode_is_reference(get_irn_mode(ab)))
379 && (get_Const_tarval(ab) == get_mode_null(get_irn_mode(ab))))
380 || (/* ab is other Alloc */
381 (get_irn_op(ab) == op_Proj)
382 && (mode_is_reference(get_irn_mode(ab)))
383 && (get_irn_op(aba) == op_Alloc)
385 || (/* aa is void and aba is Alloc */
386 (get_irn_op(aa) == op_Const)
387 && (mode_is_reference(get_irn_mode(aa)))
388 && (get_Const_tarval(aa) == get_mode_null(get_irn_mode(aa)))
389 && (get_irn_op(ab) == op_Proj)
390 && (mode_is_reference(get_irn_mode(ab)))
391 && (get_irn_op(aba) == op_Alloc)))
393 return new_tarval_from_long (get_Proj_proj(n) & Ne, mode_b);
396 } else if (get_irn_op(a) == op_DivMod) {
397 tarval *tb = value_of(b = get_DivMod_right(a));
398 tarval *ta = value_of(a = get_DivMod_left(a));
400 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
401 if (tb == get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
403 if (get_Proj_proj(n)== 0) /* Div */
404 return tarval_div(ta, tb);
406 return tarval_mod(ta, tb);
413 * If the parameter n can be computed, return its value, else tarval_bad.
414 * Performs constant folding.
416 * GL: Only if n is arithmetic operator?
418 tarval *computed_value(ir_node *n)
420 if (n->op->computed_value)
421 return n->op->computed_value(n);
426 * set the default computed_value evaluator
428 static ir_op *firm_set_default_computed_value(ir_op *op)
432 op->computed_value = computed_value_##a; \
457 op->computed_value = NULL;
465 /* returns 1 if the a and b are pointers to different locations. */
467 different_identity (ir_node *a, ir_node *b)
469 assert (mode_is_reference(get_irn_mode (a))
470 && mode_is_reference(get_irn_mode (b)));
472 if (get_irn_op (a) == op_Proj && get_irn_op(b) == op_Proj) {
473 ir_node *a1 = get_Proj_pred (a);
474 ir_node *b1 = get_Proj_pred (b);
475 if (a1 != b1 && get_irn_op (a1) == op_Alloc
476 && get_irn_op (b1) == op_Alloc)
483 static ir_node *equivalent_node_Block(ir_node *n)
487 /* The Block constructor does not call optimize, but mature_block
488 calls the optimization. */
489 assert(get_Block_matured(n));
491 /* Straightening: a single entry Block following a single exit Block
492 can be merged, if it is not the Start block. */
493 /* !!! Beware, all Phi-nodes of n must have been optimized away.
494 This should be true, as the block is matured before optimize is called.
495 But what about Phi-cycles with the Phi0/Id that could not be resolved?
496 Remaining Phi nodes are just Ids. */
497 if ((get_Block_n_cfgpreds(n) == 1) &&
498 (get_irn_op(get_Block_cfgpred(n, 0)) == op_Jmp) &&
499 (get_opt_control_flow_straightening())) {
500 n = get_nodes_Block(get_Block_cfgpred(n, 0));
502 /* Jmp jumps into the block it is in -- deal self cycle. */
503 n = new_Bad(); DBG_OPT_STG;
505 } else if ((get_Block_n_cfgpreds(n) == 2) &&
506 (get_opt_control_flow_weak_simplification())) {
507 /* Test whether Cond jumps twice to this block
508 @@@ we could do this also with two loops finding two preds from several ones. */
509 ir_node *a = get_Block_cfgpred(n, 0);
510 ir_node *b = get_Block_cfgpred(n, 1);
512 if ((get_irn_op(a) == op_Proj) &&
513 (get_irn_op(b) == op_Proj) &&
514 (get_Proj_pred(a) == get_Proj_pred(b)) &&
515 (get_irn_op(get_Proj_pred(a)) == op_Cond) &&
516 (get_irn_mode(get_Cond_selector(get_Proj_pred(a))) == mode_b)) {
517 /* Also a single entry Block following a single exit Block. Phis have
518 twice the same operand and will be optimized away. */
519 n = get_nodes_Block(a); DBG_OPT_IFSIM;
521 } else if (get_opt_unreachable_code() &&
522 (n != current_ir_graph->start_block) &&
523 (n != current_ir_graph->end_block) ) {
525 /* If all inputs are dead, this block is dead too, except if it is
526 the start or end block. This is a step of unreachable code
528 for (i = 0; i < get_Block_n_cfgpreds(n); i++) {
529 if (!is_Bad(get_Block_cfgpred(n, i))) break;
531 if (i == get_Block_n_cfgpreds(n))
538 static ir_node *equivalent_node_Jmp(ir_node *n)
540 /* GL: Why not same for op_Raise?? */
541 /* unreachable code elimination */
542 if (is_Bad(get_nodes_Block(n)))
548 static ir_node *equivalent_node_Cond(ir_node *n)
550 /* We do not evaluate Cond here as we replace it by a new node, a Jmp.
551 See cases for iro_Cond and iro_Proj in transform_node. */
555 static ir_node *equivalent_node_Or(ir_node *n)
559 ir_node *a = get_Or_left(n);
560 ir_node *b = get_Or_right(n);
564 n = a; DBG_OPT_ALGSIM1;
571 * optimize operations that are commutative and have neutral 0.
573 static ir_node *equivalent_node_neutral_zero(ir_node *n)
577 ir_node *a = get_binop_left(n);
578 ir_node *b = get_binop_right(n);
583 /* After running compute_node there is only one constant predecessor.
584 Find this predecessors value and remember the other node: */
585 if ((tv = computed_value (a)) != tarval_bad) {
587 } else if ((tv = computed_value (b)) != tarval_bad) {
592 /* If this predecessors constant value is zero, the operation is
593 unnecessary. Remove it: */
594 if (tarval_classify (tv) == TV_CLASSIFY_NULL) {
595 n = on; DBG_OPT_ALGSIM1;
601 static ir_node *equivalent_node_Add(ir_node *n)
603 return equivalent_node_neutral_zero(n);
606 static ir_node *equivalent_node_Eor(ir_node *n)
608 return equivalent_node_neutral_zero(n);
612 * optimize operations that are not commutative but have neutral 0 on left.
613 * Test only one predecessor.
615 static ir_node *equivalent_node_left_zero(ir_node *n)
619 ir_node *a = get_binop_left(n);
620 ir_node *b = get_binop_right(n);
622 if (tarval_classify (computed_value (b)) == TV_CLASSIFY_NULL) {
623 n = a; DBG_OPT_ALGSIM1;
629 static ir_node *equivalent_node_Sub(ir_node *n)
631 return equivalent_node_left_zero(n);
634 static ir_node *equivalent_node_Shl(ir_node *n)
636 return equivalent_node_left_zero(n);
639 static ir_node *equivalent_node_Shr(ir_node *n)
641 return equivalent_node_left_zero(n);
644 static ir_node *equivalent_node_Shrs(ir_node *n)
646 return equivalent_node_left_zero(n);
649 static ir_node *equivalent_node_Rot(ir_node *n)
651 return equivalent_node_left_zero(n);
654 static ir_node *equivalent_node_symmetric_unop(ir_node *n)
658 /* optimize symmetric unop */
659 if (get_irn_op(get_unop_op(n)) == get_irn_op(n)) {
660 n = get_unop_op(get_unop_op(n)); DBG_OPT_ALGSIM2;
665 static ir_node *equivalent_node_Not(ir_node *n)
668 return equivalent_node_symmetric_unop(n);
671 static ir_node *equivalent_node_Minus(ir_node *n)
673 /* --x == x */ /* ??? Is this possible or can --x raise an
674 out of bounds exception if min =! max? */
675 return equivalent_node_symmetric_unop(n);
678 static ir_node *equivalent_node_Mul(ir_node *n)
682 ir_node *a = get_Mul_left(n);
683 ir_node *b = get_Mul_right(n);
685 /* Mul is commutative and has again an other neutral element. */
686 if (tarval_classify (computed_value (a)) == TV_CLASSIFY_ONE) {
687 n = b; DBG_OPT_ALGSIM1;
688 } else if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE) {
689 n = a; DBG_OPT_ALGSIM1;
694 static ir_node *equivalent_node_Div(ir_node *n)
696 ir_node *a = get_Div_left(n);
697 ir_node *b = get_Div_right(n);
699 /* Div is not commutative. */
700 if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE) { /* div(x, 1) == x */
701 /* Turn Div into a tuple (mem, bad, a) */
702 ir_node *mem = get_Div_mem(n);
703 turn_into_tuple(n, 3);
704 set_Tuple_pred(n, pn_Div_M, mem);
705 set_Tuple_pred(n, pn_Div_X_except, new_Bad()); /* no exception */
706 set_Tuple_pred(n, pn_Div_res, a);
711 static ir_node *equivalent_node_And(ir_node *n)
715 ir_node *a = get_And_left(n);
716 ir_node *b = get_And_right(n);
719 n = a; /* And has it's own neutral element */
720 } else if (tarval_classify (computed_value (a)) == TV_CLASSIFY_ALL_ONE) {
722 } else if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ALL_ONE) {
725 if (n != oldn) DBG_OPT_ALGSIM1;
729 static ir_node *equivalent_node_Conv(ir_node *n)
732 ir_node *a = get_Conv_op(n);
735 ir_mode *n_mode = get_irn_mode(n);
736 ir_mode *a_mode = get_irn_mode(a);
738 if (n_mode == a_mode) { /* No Conv necessary */
739 n = a; DBG_OPT_ALGSIM3;
740 } else if (get_irn_op(a) == op_Conv) { /* Conv(Conv(b)) */
744 n_mode = get_irn_mode(n);
745 b_mode = get_irn_mode(b);
747 if (n_mode == b_mode) {
748 if (n_mode == mode_b) {
749 n = b; /* Convb(Conv*(xxxb(...))) == xxxb(...) */ DBG_OPT_ALGSIM1;
751 else if (mode_is_int(n_mode) || mode_is_character(n_mode)) {
752 if (smaller_mode(b_mode, a_mode)){
753 n = b; /* ConvS(ConvL(xxxS(...))) == xxxS(...) */ DBG_OPT_ALGSIM1;
761 static ir_node *equivalent_node_Phi(ir_node *n)
763 /* Several optimizations:
764 - no Phi in start block.
765 - remove Id operators that are inputs to Phi
766 - fold Phi-nodes, iff they have only one predecessor except
772 ir_node *block = NULL; /* to shutup gcc */
773 ir_node *first_val = NULL; /* to shutup gcc */
774 ir_node *scnd_val = NULL; /* to shutup gcc */
776 if (!get_opt_normalize()) return n;
778 n_preds = get_Phi_n_preds(n);
780 block = get_nodes_Block(n);
781 /* @@@ fliegt 'raus, sollte aber doch immer wahr sein!!!
782 assert(get_irn_arity(block) == n_preds && "phi in wrong block!"); */
783 if ((is_Bad(block)) || /* Control dead */
784 (block == current_ir_graph->start_block)) /* There should be no Phi nodes */
785 return new_Bad(); /* in the Start Block. */
787 if (n_preds == 0) return n; /* Phi of dead Region without predecessors. */
790 /* first we test for a special case: */
791 /* Confirm is a special node fixing additional information for a
792 value that is known at a certain point. This is useful for
793 dataflow analysis. */
795 ir_node *a = follow_Id (get_Phi_pred(n, 0));
796 ir_node *b = follow_Id (get_Phi_pred(n, 1));
797 if ( (get_irn_op(a) == op_Confirm)
798 && (get_irn_op(b) == op_Confirm)
799 && follow_Id (get_irn_n(a, 0) == get_irn_n(b, 0))
800 && (get_irn_n(a, 1) == get_irn_n (b, 1))
801 && (a->data.num == (~b->data.num & irpn_True) )) {
802 return get_irn_n(a, 0);
807 /* Find first non-self-referencing input */
808 for (i = 0; i < n_preds; ++i) {
809 first_val = follow_Id(get_Phi_pred(n, i));
811 set_Phi_pred(n, i, first_val);
812 if ( (first_val != n) /* not self pointer */
813 && (get_irn_op(first_val) != op_Bad) /* value not dead */
814 && !(is_Bad (get_Block_cfgpred(block, i))) ) { /* not dead control flow */
815 break; /* then found first value. */
819 /* A totally Bad or self-referencing Phi (we didn't break the above loop) */
820 if (i >= n_preds) { return new_Bad(); }
824 /* follow_Id () for rest of inputs, determine if any of these
825 are non-self-referencing */
826 while (++i < n_preds) {
827 scnd_val = follow_Id(get_Phi_pred(n, i));
829 set_Phi_pred(n, i, scnd_val);
831 && (scnd_val != first_val)
832 && (get_irn_op(scnd_val) != op_Bad)
833 && !(is_Bad (get_Block_cfgpred(block, i))) ) {
838 /* Fold, if no multiple distinct non-self-referencing inputs */
840 n = first_val; DBG_OPT_PHI;
842 /* skip the remaining Ids. */
843 while (++i < n_preds) {
844 set_Phi_pred(n, i, follow_Id(get_Phi_pred(n, i)));
850 static ir_node *equivalent_node_Load(ir_node *n)
852 #if 0 /* Is an illegal transformation: different nodes can
853 represent the same pointer value!! */
854 ir_node *a = skip_Proj(get_Load_mem(n));
855 ir_node *b = get_Load_ptr(n);
857 if (get_irn_op(a) == op_Store) {
858 if ( different_identity (b, get_Store_ptr(a))) {
859 /* load and store use different pointers, therefore load
860 needs not take store's memory but the state before. */
861 set_Load_mem (n, get_Store_mem(a));
862 } else if (( 0 /* ???didn't get cryptic test that returns 0 */ )) {
869 static ir_node *equivalent_node_Store(ir_node *n)
873 /* remove unnecessary store. */
874 ir_node *a = skip_Proj(get_Store_mem(n));
875 ir_node *b = get_Store_ptr(n);
876 ir_node *c = skip_Proj(get_Store_value(n));
878 if (get_irn_op(a) == op_Store
879 && get_Store_ptr(a) == b
880 && skip_Proj(get_Store_value(a)) == c) {
881 /* We have twice exactly the same store -- a write after write. */
883 } else if (get_irn_op(c) == op_Load
884 && (a == c || skip_Proj(get_Load_mem(c)) == a)
885 && get_Load_ptr(c) == b ) {
886 /* We just loaded the value from the same memory, i.e., the store
887 doesn't change the memory -- a write after read. */
888 a = get_Store_mem(n);
889 turn_into_tuple(n, 2);
890 set_Tuple_pred(n, pn_Store_M, a);
891 set_Tuple_pred(n, pn_Store_X_except, new_Bad()); DBG_OPT_WAR;
896 static ir_node *equivalent_node_Proj(ir_node *n)
900 ir_node *a = get_Proj_pred(n);
902 if ( get_irn_op(a) == op_Tuple) {
903 /* Remove the Tuple/Proj combination. */
904 if ( get_Proj_proj(n) <= get_Tuple_n_preds(a) ) {
905 n = get_Tuple_pred(a, get_Proj_proj(n)); DBG_OPT_TUPLE;
907 assert(0); /* This should not happen! */
910 } else if (get_irn_mode(n) == mode_X &&
911 is_Bad(get_nodes_Block(n))) {
912 /* Remove dead control flow -- early gigo. */
918 static ir_node *equivalent_node_Id(ir_node *n)
922 n = follow_Id (n); DBG_OPT_ID;
927 case iro_Mod, Quot, DivMod
928 DivMod allocates new nodes --> it's treated in transform node.
929 What about Quot, DivMod?
933 * equivalent_node() returns a node equivalent to input n. It skips all nodes that
934 * perform no actual computation, as, e.g., the Id nodes. It does not create
935 * new nodes. It is therefore safe to free n if the node returned is not n.
936 * If a node returns a Tuple we can not just skip it. If the size of the
937 * in array fits, we transform n into a tuple (e.g., Div).
940 equivalent_node (ir_node *n)
942 if (n->op->equivalent_node)
943 return n->op->equivalent_node(n);
948 * set the default equivalent node operation
950 static ir_op *firm_set_default_equivalent_node(ir_op *op)
954 op->equivalent_node = equivalent_node_##a; \
981 op->equivalent_node = NULL;
989 * Do node specific optimizations of nodes predecessors.
992 optimize_preds(ir_node *n) {
993 ir_node *a = NULL, *b = NULL;
995 /* get the operands we will work on for simple cases. */
997 a = get_binop_left(n);
998 b = get_binop_right(n);
999 } else if (is_unop(n)) {
1003 switch (get_irn_opcode(n)) {
1006 /* We don't want Cast as input to Cmp. */
1007 if (get_irn_op(a) == op_Cast) {
1011 if (get_irn_op(b) == op_Cast) {
1013 set_Cmp_right(n, b);
1021 static ir_node *transform_node_Div(ir_node *n)
1023 tarval *ta = computed_value(n);
1025 if (ta != tarval_bad) {
1026 /* Turn Div into a tuple (mem, bad, value) */
1027 ir_node *mem = get_Div_mem(n);
1029 turn_into_tuple(n, 3);
1030 set_Tuple_pred(n, pn_Div_M, mem);
1031 set_Tuple_pred(n, pn_Div_X_except, new_Bad());
1032 set_Tuple_pred(n, pn_Div_res, new_Const(get_tarval_mode(ta), ta));
1037 static ir_node *transform_node_Mod(ir_node *n)
1039 tarval *ta = computed_value(n);
1041 if (ta != tarval_bad) {
1042 /* Turn Mod into a tuple (mem, bad, value) */
1043 ir_node *mem = get_Mod_mem(n);
1044 turn_into_tuple(n, 3);
1045 set_Tuple_pred(n, pn_Mod_M, mem);
1046 set_Tuple_pred(n, pn_Mod_X_except, new_Bad());
1047 set_Tuple_pred(n, pn_Mod_res, new_Const(get_tarval_mode(ta), ta));
1052 static ir_node *transform_node_DivMod(ir_node *n)
1056 ir_node *a = get_DivMod_left(n);
1057 ir_node *b = get_DivMod_right(n);
1058 ir_mode *mode = get_irn_mode(a);
1060 if (!(mode_is_int(mode) && mode_is_int(get_irn_mode(b))))
1064 a = new_Const(mode, get_mode_one(mode));
1065 b = new_Const(mode, get_mode_null(mode));
1068 tarval *ta = value_of(a);
1069 tarval *tb = value_of(b);
1071 if (tb != tarval_bad) {
1072 if (tb == get_mode_one(get_tarval_mode(tb))) {
1073 b = new_Const (mode, get_mode_null(mode));
1075 } else if (ta != tarval_bad) {
1076 tarval *resa, *resb;
1077 resa = tarval_div (ta, tb);
1078 if (resa == tarval_bad) return n; /* Causes exception!!! Model by replacing through
1079 Jmp for X result!? */
1080 resb = tarval_mod (ta, tb);
1081 if (resb == tarval_bad) return n; /* Causes exception! */
1082 a = new_Const (mode, resa);
1083 b = new_Const (mode, resb);
1086 } else if (ta == get_mode_null(mode)) {
1091 if (evaluated) { /* replace by tuple */
1092 ir_node *mem = get_DivMod_mem(n);
1093 turn_into_tuple(n, 4);
1094 set_Tuple_pred(n, pn_DivMod_M, mem);
1095 set_Tuple_pred(n, pn_DivMod_X_except, new_Bad()); /* no exception */
1096 set_Tuple_pred(n, pn_DivMod_res_div, a);
1097 set_Tuple_pred(n, pn_DivMod_res_mod, b);
1098 assert(get_nodes_Block(n));
1104 static ir_node *transform_node_Cond(ir_node *n)
1106 /* Replace the Cond by a Jmp if it branches on a constant
1109 ir_node *a = get_Cond_selector(n);
1110 tarval *ta = value_of(a);
1112 if ((ta != tarval_bad) &&
1113 (get_irn_mode(a) == mode_b) &&
1114 (get_opt_unreachable_code())) {
1115 /* It's a boolean Cond, branching on a boolean constant.
1116 Replace it by a tuple (Bad, Jmp) or (Jmp, Bad) */
1117 jmp = new_r_Jmp(current_ir_graph, get_nodes_Block(n));
1118 turn_into_tuple(n, 2);
1119 if (ta == tarval_b_true) {
1120 set_Tuple_pred(n, pn_Cond_false, new_Bad());
1121 set_Tuple_pred(n, pn_Cond_true, jmp);
1123 set_Tuple_pred(n, pn_Cond_false, jmp);
1124 set_Tuple_pred(n, pn_Cond_true, new_Bad());
1126 /* We might generate an endless loop, so keep it alive. */
1127 add_End_keepalive(get_irg_end(current_ir_graph), get_nodes_Block(n));
1128 } else if ((ta != tarval_bad) &&
1129 (get_irn_mode(a) == mode_Iu) &&
1130 (get_Cond_kind(n) == dense) &&
1131 (get_opt_unreachable_code())) {
1132 /* I don't want to allow Tuples smaller than the biggest Proj.
1133 Also this tuple might get really big...
1134 I generate the Jmp here, and remember it in link. Link is used
1135 when optimizing Proj. */
1136 set_irn_link(n, new_r_Jmp(current_ir_graph, get_nodes_Block(n)));
1137 /* We might generate an endless loop, so keep it alive. */
1138 add_End_keepalive(get_irg_end(current_ir_graph), get_nodes_Block(n));
1139 } else if ((get_irn_op(a) == op_Eor)
1140 && (get_irn_mode(a) == mode_b)
1141 && (tarval_classify(computed_value(get_Eor_right(a))) == TV_CLASSIFY_ONE)) {
1142 /* The Eor is a negate. Generate a new Cond without the negate,
1143 simulate the negate by exchanging the results. */
1144 set_irn_link(n, new_r_Cond(current_ir_graph, get_nodes_Block(n),
1146 } else if ((get_irn_op(a) == op_Not)
1147 && (get_irn_mode(a) == mode_b)) {
1148 /* A Not before the Cond. Generate a new Cond without the Not,
1149 simulate the Not by exchanging the results. */
1150 set_irn_link(n, new_r_Cond(current_ir_graph, get_nodes_Block(n),
1156 static ir_node *transform_node_Eor(ir_node *n)
1158 ir_node *a = get_Eor_left(n);
1159 ir_node *b = get_Eor_right(n);
1161 if ((get_irn_mode(n) == mode_b)
1162 && (get_irn_op(a) == op_Proj)
1163 && (get_irn_mode(a) == mode_b)
1164 && (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE)
1165 && (get_irn_op(get_Proj_pred(a)) == op_Cmp))
1166 /* The Eor negates a Cmp. The Cmp has the negated result anyways! */
1167 n = new_r_Proj(current_ir_graph, get_nodes_Block(n), get_Proj_pred(a),
1168 mode_b, get_negated_pnc(get_Proj_proj(a)));
1169 else if ((get_irn_mode(n) == mode_b)
1170 && (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE))
1171 /* The Eor is a Not. Replace it by a Not. */
1172 /* ????!!!Extend to bitfield 1111111. */
1173 n = new_r_Not(current_ir_graph, get_nodes_Block(n), a, mode_b);
1178 static ir_node *transform_node_Not(ir_node *n)
1180 ir_node *a = get_Not_op(n);
1182 if ( (get_irn_mode(n) == mode_b)
1183 && (get_irn_op(a) == op_Proj)
1184 && (get_irn_mode(a) == mode_b)
1185 && (get_irn_op(get_Proj_pred(a)) == op_Cmp))
1186 /* We negate a Cmp. The Cmp has the negated result anyways! */
1187 n = new_r_Proj(current_ir_graph, get_nodes_Block(n), get_Proj_pred(a),
1188 mode_b, get_negated_pnc(get_Proj_proj(a)));
1195 * Tries several [inplace] [optimizing] transformations and returns an
1196 * equivalent node. The difference to equivalent_node() is that these
1197 * transformations _do_ generate new nodes, and thus the old node must
1198 * not be freed even if the equivalent node isn't the old one.
1200 static ir_node *transform_node(ir_node *n)
1202 if (n->op->transform_node)
1203 n = n->op->transform_node(n);
1208 * set the default transform node operation
1210 static ir_op *firm_set_default_transform_node(ir_op *op)
1214 op->transform_node = transform_node_##a; \
1225 op->transform_node = NULL;
1233 /* **************** Common Subexpression Elimination **************** */
1235 /** The size of the hash table used, should estimate the number of nodes
1237 #define N_IR_NODES 512
1239 static int node_cmp_attr_Const(ir_node *a, ir_node *b)
1241 return (get_Const_tarval(a) != get_Const_tarval(b))
1242 || (get_Const_type(a) != get_Const_type(b));
1245 static int node_cmp_attr_Proj(ir_node *a, ir_node *b)
1247 return get_irn_proj_attr (a) != get_irn_proj_attr (b);
1250 static int node_cmp_attr_Filter(ir_node *a, ir_node *b)
1252 return get_Filter_proj(a) != get_Filter_proj(b);
1255 static int node_cmp_attr_Alloc(ir_node *a, ir_node *b)
1257 return (get_irn_alloc_attr(a).where != get_irn_alloc_attr(b).where)
1258 || (get_irn_alloc_attr(a).type != get_irn_alloc_attr(b).type);
1261 static int node_cmp_attr_Free(ir_node *a, ir_node *b)
1263 return (get_irn_free_attr(a) != get_irn_free_attr(b));
1266 static int node_cmp_attr_SymConst(ir_node *a, ir_node *b)
1268 return (get_irn_symconst_attr(a).num != get_irn_symconst_attr(b).num)
1269 || (get_irn_symconst_attr(a).tori.typ != get_irn_symconst_attr(b).tori.typ);
1272 static int node_cmp_attr_Call(ir_node *a, ir_node *b)
1274 return (get_irn_call_attr(a) != get_irn_call_attr(b));
1277 static int node_cmp_attr_FuncCall(ir_node *a, ir_node *b)
1279 return (get_irn_funccall_attr(a) != get_irn_funccall_attr(b));
1282 static int node_cmp_attr_Sel(ir_node *a, ir_node *b)
1284 return (get_irn_sel_attr(a).ent->kind != get_irn_sel_attr(b).ent->kind)
1285 || (get_irn_sel_attr(a).ent->name != get_irn_sel_attr(b).ent->name)
1286 || (get_irn_sel_attr(a).ent->owner != get_irn_sel_attr(b).ent->owner)
1287 || (get_irn_sel_attr(a).ent->ld_name != get_irn_sel_attr(b).ent->ld_name)
1288 || (get_irn_sel_attr(a).ent->type != get_irn_sel_attr(b).ent->type);
1291 static int node_cmp_attr_Phi(ir_node *a, ir_node *b)
1293 return get_irn_phi_attr (a) != get_irn_phi_attr (b);
1296 static int node_cmp_attr_Cast(ir_node *a, ir_node *b)
1298 return get_Cast_type(a) != get_Cast_type(b);
1302 * set the default node attribute compare operation
1304 static ir_op *firm_set_default_node_cmp_attr(ir_op *op)
1308 op->node_cmp_attr = node_cmp_attr_##a; \
1324 op->node_cmp_attr = NULL;
1332 * Compare function for two nodes in the hash table. Gets two
1333 * nodes as parameters. Returns 0 if the nodes are a cse.
1336 vt_cmp (const void *elt, const void *key)
1344 if (a == b) return 0;
1346 if ((get_irn_op(a) != get_irn_op(b)) ||
1347 (get_irn_mode(a) != get_irn_mode(b))) return 1;
1349 /* compare if a's in and b's in are equal */
1350 irn_arity_a = get_irn_arity (a);
1351 if (irn_arity_a != get_irn_arity(b))
1354 /* for block-local cse and pinned nodes: */
1355 if (!get_opt_global_cse() || (get_op_pinned(get_irn_op(a)) == pinned)) {
1356 if (get_irn_n(a, -1) != get_irn_n(b, -1))
1360 /* compare a->in[0..ins] with b->in[0..ins] */
1361 for (i = 0; i < irn_arity_a; i++)
1362 if (get_irn_n(a, i) != get_irn_n(b, i))
1366 * here, we already now that the nodes are identical except their
1369 if (a->op->node_cmp_attr)
1370 return a->op->node_cmp_attr(a, b);
1376 * Calculate a hash value of a node.
1379 ir_node_hash (ir_node *node)
1384 /* hash table value = 9*(9*(9*(9*(9*arity+in[0])+in[1])+ ...)+mode)+code */
1385 h = irn_arity = get_irn_arity(node);
1387 /* consider all in nodes... except the block. */
1388 for (i = 0; i < irn_arity; i++) {
1389 h = 9*h + (unsigned long)get_irn_n(node, i);
1393 h = 9*h + (unsigned long) get_irn_mode (node);
1395 h = 9*h + (unsigned long) get_irn_op (node);
1401 new_identities (void)
1403 return new_pset (vt_cmp, N_IR_NODES);
1407 del_identities (pset *value_table)
1409 del_pset (value_table);
1413 * Return the canonical node computing the same value as n.
1414 * Looks up the node in a hash table.
1416 static INLINE ir_node *
1417 identify (pset *value_table, ir_node *n)
1421 if (!value_table) return n;
1423 /* TODO: use a generic commutative attribute */
1424 if (get_opt_reassociation()) {
1425 if (is_op_commutative(get_irn_op(n))) {
1426 /* for commutative operators perform a OP b == b OP a */
1427 if (get_binop_left(n) > get_binop_right(n)) {
1428 ir_node *h = get_binop_left(n);
1429 set_binop_left(n, get_binop_right(n));
1430 set_binop_right(n, h);
1435 o = pset_find (value_table, n, ir_node_hash (n));
1442 * During construction we set the pinned flag in the graph right when the
1443 * optimizatin is performed. The flag turning on procedure global cse could
1444 * be changed between two allocations. This way we are safe.
1446 static INLINE ir_node *
1447 identify_cons (pset *value_table, ir_node *n) {
1449 n = identify(value_table, n);
1450 if (get_irn_n(old, -1) != get_irn_n(n, -1))
1451 set_irg_pinned(current_ir_graph, floats);
1456 * Return the canonical node computing the same value as n.
1457 * Looks up the node in a hash table, enters it in the table
1458 * if it isn't there yet.
1461 identify_remember (pset *value_table, ir_node *node)
1465 if (!value_table) return node;
1467 /* lookup or insert in hash table with given hash key. */
1468 o = pset_insert (value_table, node, ir_node_hash (node));
1470 if (o == node) return node;
1476 add_identities (pset *value_table, ir_node *node) {
1477 identify_remember (value_table, node);
1481 * garbage in, garbage out. If a node has a dead input, i.e., the
1482 * Bad node is input to the node, return the Bad node.
1484 static INLINE ir_node *
1485 gigo (ir_node *node)
1488 ir_op* op = get_irn_op(node);
1490 /* remove garbage blocks by looking at control flow that leaves the block
1491 and replacing the control flow by Bad. */
1492 if (get_irn_mode(node) == mode_X) {
1493 ir_node *block = get_nodes_block(node);
1494 if (op == op_End) return node; /* Don't optimize End, may have Bads. */
1495 if (get_irn_op(block) == op_Block && get_Block_matured(block)) {
1496 irn_arity = get_irn_arity(block);
1497 for (i = 0; i < irn_arity; i++) {
1498 if (!is_Bad(get_irn_n(block, i))) break;
1500 if (i == irn_arity) return new_Bad();
1504 /* Blocks, Phis and Tuples may have dead inputs, e.g., if one of the
1505 blocks predecessors is dead. */
1506 if ( op != op_Block && op != op_Phi && op != op_Tuple) {
1507 irn_arity = get_irn_arity(node);
1508 for (i = -1; i < irn_arity; i++) {
1509 if (is_Bad(get_irn_n(node, i))) {
1515 /* With this code we violate the agreement that local_optimize
1516 only leaves Bads in Block, Phi and Tuple nodes. */
1517 /* If Block has only Bads as predecessors it's garbage. */
1518 /* If Phi has only Bads as predecessors it's garbage. */
1519 if ((op == op_Block && get_Block_matured(node)) || op == op_Phi) {
1520 irn_arity = get_irn_arity(node);
1521 for (i = 0; i < irn_arity; i++) {
1522 if (!is_Bad(get_irn_n(node, i))) break;
1524 if (i == irn_arity) node = new_Bad();
1532 * These optimizations deallocate nodes from the obstack.
1533 * It can only be called if it is guaranteed that no other nodes
1534 * reference this one, i.e., right after construction of a node.
1537 optimize_node (ir_node *n)
1541 opcode iro = get_irn_opcode(n);
1543 /* Allways optimize Phi nodes: part of the construction. */
1544 if ((!get_opt_optimize()) && (iro != iro_Phi)) return n;
1546 /* constant expression evaluation / constant folding */
1547 if (get_opt_constant_folding()) {
1548 /* constants can not be evaluated */
1549 if (iro != iro_Const) {
1550 /* try to evaluate */
1551 tv = computed_value (n);
1552 if ((get_irn_mode(n) != mode_T) && (tv != tarval_bad)) {
1554 * we MUST copy the node here temparary, because it's still needed
1555 * for DBG_OPT_ALGSIM0
1559 /* evaluation was successful -- replace the node. */
1560 obstack_free (current_ir_graph->obst, n);
1561 n = new_Const (get_tarval_mode (tv), tv);
1568 /* remove unnecessary nodes */
1569 if (get_opt_constant_folding() ||
1570 (iro == iro_Phi) || /* always optimize these nodes. */
1572 (iro == iro_Proj) ||
1573 (iro == iro_Block) ) /* Flags tested local. */
1574 n = equivalent_node (n);
1576 optimize_preds(n); /* do node specific optimizations of nodes predecessors. */
1578 /** common subexpression elimination **/
1579 /* Checks whether n is already available. */
1580 /* The block input is used to distinguish different subexpressions. Right
1581 now all nodes are pinned to blocks, i.e., the cse only finds common
1582 subexpressions within a block. */
1584 n = identify_cons (current_ir_graph->value_table, n);
1587 /* We found an existing, better node, so we can deallocate the old node. */
1588 obstack_free (current_ir_graph->obst, oldn);
1593 /* Some more constant expression evaluation that does not allow to
1595 iro = get_irn_opcode(n);
1596 if (get_opt_constant_folding() ||
1597 (iro == iro_Cond) ||
1598 (iro == iro_Proj)) /* Flags tested local. */
1599 n = transform_node (n);
1601 /* Remove nodes with dead (Bad) input.
1602 Run always for transformation induced Bads. */
1605 /* Now we have a legal, useful node. Enter it in hash table for cse */
1606 if (get_opt_cse() && (get_irn_opcode(n) != iro_Block)) {
1607 n = identify_remember (current_ir_graph->value_table, n);
1615 * These optimizations never deallocate nodes. This can cause dead
1616 * nodes lying on the obstack. Remove these by a dead node elimination,
1617 * i.e., a copying garbage collection.
1620 optimize_in_place_2 (ir_node *n)
1624 opcode iro = get_irn_opcode(n);
1626 if (!get_opt_optimize() && (get_irn_op(n) != op_Phi)) return n;
1628 /* if not optimize return n */
1631 /* Here this is possible. Why? */
1636 /* constant expression evaluation / constant folding */
1637 if (get_opt_constant_folding()) {
1638 /* constants can not be evaluated */
1639 if (iro != iro_Const) {
1640 /* try to evaluate */
1641 tv = computed_value (n);
1642 if ((get_irn_mode(n) != mode_T) && (tv != tarval_bad)) {
1643 /* evaluation was successful -- replace the node. */
1644 n = new_Const (get_tarval_mode (tv), tv);
1651 /* remove unnecessary nodes */
1652 /*if (get_opt_constant_folding()) */
1653 if (get_opt_constant_folding() ||
1654 (iro == iro_Phi) || /* always optimize these nodes. */
1655 (iro == iro_Id) || /* ... */
1656 (iro == iro_Proj) || /* ... */
1657 (iro == iro_Block) ) /* Flags tested local. */
1658 n = equivalent_node (n);
1660 optimize_preds(n); /* do node specific optimizations of nodes predecessors. */
1662 /** common subexpression elimination **/
1663 /* Checks whether n is already available. */
1664 /* The block input is used to distinguish different subexpressions. Right
1665 now all nodes are pinned to blocks, i.e., the cse only finds common
1666 subexpressions within a block. */
1667 if (get_opt_cse()) {
1668 n = identify (current_ir_graph->value_table, n);
1671 /* Some more constant expression evaluation. */
1672 iro = get_irn_opcode(n);
1673 if (get_opt_constant_folding() ||
1674 (iro == iro_Cond) ||
1675 (iro == iro_Proj)) /* Flags tested local. */
1676 n = transform_node (n);
1678 /* Remove nodes with dead (Bad) input.
1679 Run always for transformation induced Bads. */
1682 /* Now we can verify the node, as it has no dead inputs any more. */
1685 /* Now we have a legal, useful node. Enter it in hash table for cse.
1686 Blocks should be unique anyways. (Except the successor of start:
1687 is cse with the start block!) */
1688 if (get_opt_cse() && (get_irn_opcode(n) != iro_Block))
1689 n = identify_remember (current_ir_graph->value_table, n);
1695 * Wrapper for external use, set proper status bits after optimization.
1698 optimize_in_place (ir_node *n)
1700 /* Handle graph state */
1701 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1703 if (get_opt_global_cse())
1704 set_irg_pinned(current_ir_graph, floats);
1705 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1706 set_irg_outs_inconsistent(current_ir_graph);
1707 /* Maybe we could also test whether optimizing the node can
1708 change the control graph. */
1709 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1710 set_irg_dom_inconsistent(current_ir_graph);
1711 return optimize_in_place_2 (n);
1715 * set the default ir op operations
1717 ir_op *firm_set_default_operations(ir_op *op)
1719 op = firm_set_default_computed_value(op);
1720 op = firm_set_default_equivalent_node(op);
1721 op = firm_set_default_transform_node(op);
1722 op = firm_set_default_node_cmp_attr(op);