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"
21 # include "ircons_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 */
57 * return the value of a Constant
59 static tarval *computed_value_Const(ir_node *n)
61 return get_Const_tarval(n);
65 * return the value of a 'sizeof' SymConst
67 static tarval *computed_value_SymConst(ir_node *n)
69 if ((get_SymConst_kind(n) == symconst_size) &&
70 (get_type_state(get_SymConst_type(n))) == layout_fixed)
71 return new_tarval_from_long(get_type_size_bytes(get_SymConst_type(n)), get_irn_mode(n));
75 static tarval *computed_value_Add(ir_node *n)
77 ir_node *a = get_Add_left(n);
78 ir_node *b = get_Add_right(n);
80 tarval *ta = value_of(a);
81 tarval *tb = value_of(b);
83 if ((ta != tarval_bad) && (tb != tarval_bad)
84 && (get_irn_mode(a) == get_irn_mode(b))
85 && !(get_mode_sort(get_irn_mode(a)) == irms_reference)) {
86 return tarval_add(ta, tb);
91 static tarval *computed_value_Sub(ir_node *n)
93 ir_node *a = get_Sub_left(n);
94 ir_node *b = get_Sub_right(n);
96 tarval *ta = value_of(a);
97 tarval *tb = value_of(b);
99 if ((ta != tarval_bad) && (tb != tarval_bad)
100 && (get_irn_mode(a) == get_irn_mode(b))
101 && !(get_mode_sort(get_irn_mode(a)) == irms_reference)) {
102 return tarval_sub(ta, tb);
107 static tarval *computed_value_Minus(ir_node *n)
109 ir_node *a = get_Minus_op(n);
110 tarval *ta = value_of(a);
112 if ((ta != tarval_bad) && mode_is_signed(get_irn_mode(a)))
113 return tarval_neg(ta);
118 static tarval *computed_value_Mul(ir_node *n)
120 ir_node *a = get_Mul_left(n);
121 ir_node *b = get_Mul_right(n);
123 tarval *ta = value_of(a);
124 tarval *tb = value_of(b);
126 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
127 return tarval_mul(ta, tb);
129 /* a*0 = 0 or 0*b = 0:
130 calls computed_value recursive and returns the 0 with proper
134 if ( ( ((v = ta) != tarval_bad)
135 && (v == get_mode_null(get_tarval_mode(v))) )
136 || ( ((v = tb) != tarval_bad)
137 && (v == get_mode_null(get_tarval_mode(v))) )) {
144 static tarval *computed_value_Quot(ir_node *n)
146 ir_node *a = get_Quot_left(n);
147 ir_node *b = get_Quot_right(n);
149 tarval *ta = value_of(a);
150 tarval *tb = value_of(b);
152 /* This was missing in original implementation. Why? */
153 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
154 if (tb != get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
155 return tarval_quo(ta, tb);
160 static tarval *computed_value_Div(ir_node *n)
162 ir_node *a = get_Div_left(n);
163 ir_node *b = get_Div_right(n);
165 tarval *ta = value_of(a);
166 tarval *tb = value_of(b);
168 /* This was missing in original implementation. Why? */
169 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
170 if (tb != get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
171 return tarval_div(ta, tb);
176 static tarval *computed_value_Mod(ir_node *n)
178 ir_node *a = get_Mod_left(n);
179 ir_node *b = get_Mod_right(n);
181 tarval *ta = value_of(a);
182 tarval *tb = value_of(b);
184 /* This was missing in original implementation. Why? */
185 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
186 if (tb != get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
187 return tarval_mod(ta, tb);
192 static tarval *computed_value_Abs(ir_node *n)
194 ir_node *a = get_Abs_op(n);
195 tarval *ta = value_of(a);
197 if (ta != tarval_bad)
198 return tarval_abs(ta);
203 static tarval *computed_value_And(ir_node *n)
205 ir_node *a = get_And_left(n);
206 ir_node *b = get_And_right(n);
208 tarval *ta = value_of(a);
209 tarval *tb = value_of(b);
211 if ((ta != tarval_bad) && (tb != tarval_bad)) {
212 return tarval_and (ta, tb);
216 if ( (tarval_classify ((v = computed_value (a))) == TV_CLASSIFY_NULL)
217 || (tarval_classify ((v = computed_value (b))) == TV_CLASSIFY_NULL)) {
224 static tarval *computed_value_Or(ir_node *n)
226 ir_node *a = get_Or_left(n);
227 ir_node *b = get_Or_right(n);
229 tarval *ta = value_of(a);
230 tarval *tb = value_of(b);
232 if ((ta != tarval_bad) && (tb != tarval_bad)) {
233 return tarval_or (ta, tb);
236 if ( (tarval_classify ((v = computed_value (a))) == TV_CLASSIFY_ALL_ONE)
237 || (tarval_classify ((v = computed_value (b))) == TV_CLASSIFY_ALL_ONE)) {
244 static tarval *computed_value_Eor(ir_node *n)
246 ir_node *a = get_Eor_left(n);
247 ir_node *b = get_Eor_right(n);
249 tarval *ta = value_of(a);
250 tarval *tb = value_of(b);
252 if ((ta != tarval_bad) && (tb != tarval_bad)) {
253 return tarval_eor (ta, tb);
258 static tarval *computed_value_Not(ir_node *n)
260 ir_node *a = get_Not_op(n);
261 tarval *ta = value_of(a);
263 if (ta != tarval_bad)
264 return tarval_not(ta);
269 static tarval *computed_value_Shl(ir_node *n)
271 ir_node *a = get_Shl_left(n);
272 ir_node *b = get_Shl_right(n);
274 tarval *ta = value_of(a);
275 tarval *tb = value_of(b);
277 if ((ta != tarval_bad) && (tb != tarval_bad)) {
278 return tarval_shl (ta, tb);
283 static tarval *computed_value_Shr(ir_node *n)
285 ir_node *a = get_Shr_left(n);
286 ir_node *b = get_Shr_right(n);
288 tarval *ta = value_of(a);
289 tarval *tb = value_of(b);
291 if ((ta != tarval_bad) && (tb != tarval_bad)) {
292 return tarval_shr (ta, tb);
297 static tarval *computed_value_Shrs(ir_node *n)
299 ir_node *a = get_Shrs_left(n);
300 ir_node *b = get_Shrs_right(n);
302 tarval *ta = value_of(a);
303 tarval *tb = value_of(b);
305 if ((ta != tarval_bad) && (tb != tarval_bad)) {
306 return tarval_shrs (ta, tb);
311 static tarval *computed_value_Rot(ir_node *n)
313 ir_node *a = get_Rot_left(n);
314 ir_node *b = get_Rot_right(n);
316 tarval *ta = value_of(a);
317 tarval *tb = value_of(b);
319 if ((ta != tarval_bad) && (tb != tarval_bad)) {
320 return tarval_rot (ta, tb);
325 static tarval *computed_value_Conv(ir_node *n)
327 ir_node *a = get_Conv_op(n);
328 tarval *ta = value_of(a);
330 if (ta != tarval_bad)
331 return tarval_convert_to(ta, get_irn_mode(n));
336 static tarval *computed_value_Proj(ir_node *n)
338 ir_node *a = get_Proj_pred(n), *b;
341 /* Optimize Cmp nodes.
342 This performs a first step of unreachable code elimination.
343 Proj can not be computed, but folding a Cmp above the Proj here is
344 not as wasteful as folding a Cmp into a Tuple of 16 Consts of which
346 There are several case where we can evaluate a Cmp node:
347 1. The nodes compared are both the same. If we compare for
348 equal, greater equal, ... this will return true, else it
349 will return false. This step relies on cse.
350 2. The predecessors of Cmp are target values. We can evaluate
352 3. The predecessors are Allocs or void* constants. Allocs never
353 return NULL, they raise an exception. Therefore we can predict
355 switch (get_irn_opcode(a)) {
357 aa = get_Cmp_left(a);
358 ab = get_Cmp_right(a);
360 if (aa == ab) { /* 1.: */
361 /* This is a trick with the bits used for encoding the Cmp
362 Proj numbers, the following statement is not the same:
363 return new_tarval_from_long ((get_Proj_proj(n) == Eq), mode_b) */
364 return new_tarval_from_long ((get_Proj_proj(n) & Eq), mode_b);
366 tarval *taa = computed_value (aa);
367 tarval *tab = computed_value (ab);
369 if ((taa != tarval_bad) && (tab != tarval_bad)) { /* 2.: */
370 /* strange checks... */
371 pnc_number flags = tarval_cmp (taa, tab);
372 if (flags != False) {
373 return new_tarval_from_long (get_Proj_proj(n) & flags, mode_b);
375 } else { /* check for 3.: */
376 ir_node *aaa = skip_nop(skip_Proj(aa));
377 ir_node *aba = skip_nop(skip_Proj(ab));
379 if ( ( (/* aa is ProjP and aaa is Alloc */
380 (get_irn_op(aa) == op_Proj)
381 && (mode_is_reference(get_irn_mode(aa)))
382 && (get_irn_op(aaa) == op_Alloc))
383 && ( (/* ab is constant void */
384 (get_irn_op(ab) == op_Const)
385 && (mode_is_reference(get_irn_mode(ab)))
386 && (get_Const_tarval(ab) == get_mode_null(get_irn_mode(ab))))
387 || (/* ab is other Alloc */
388 (get_irn_op(ab) == op_Proj)
389 && (mode_is_reference(get_irn_mode(ab)))
390 && (get_irn_op(aba) == op_Alloc)
392 || (/* aa is void and aba is Alloc */
393 (get_irn_op(aa) == op_Const)
394 && (mode_is_reference(get_irn_mode(aa)))
395 && (get_Const_tarval(aa) == get_mode_null(get_irn_mode(aa)))
396 && (get_irn_op(ab) == op_Proj)
397 && (mode_is_reference(get_irn_mode(ab)))
398 && (get_irn_op(aba) == op_Alloc)))
400 return new_tarval_from_long (get_Proj_proj(n) & Ne, mode_b);
407 tarval *tb = value_of(b = get_DivMod_right(a));
408 tarval *ta = value_of(a = get_DivMod_left(a));
410 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
411 if (tb == get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
413 if (get_Proj_proj(n)== pn_DivMod_res_div)
414 return tarval_div(ta, tb);
415 else if (get_Proj_proj(n)== pn_DivMod_res_mod)
416 return tarval_mod(ta, tb);
423 tarval *tb = value_of(b = get_Div_right(a));
424 tarval *ta = value_of(a = get_Div_left(a));
426 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
427 if (tb == get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
429 if (get_Proj_proj(n)== pn_Div_res)
430 return tarval_div(ta, tb);
437 tarval *tb = value_of(b = get_Mod_right(a));
438 tarval *ta = value_of(a = get_Mod_left(a));
440 if ((ta != tarval_bad) && (tb != tarval_bad) && (get_irn_mode(a) == get_irn_mode(b))) {
441 if (tb == get_mode_null(get_tarval_mode(tb))) /* div by zero: return tarval_bad */
443 if (get_Proj_proj(n)== pn_Mod_res)
444 return tarval_mod(ta, tb);
456 * If the parameter n can be computed, return its value, else tarval_bad.
457 * Performs constant folding.
459 * @param n The node this should be evaluated
461 tarval *computed_value(ir_node *n)
463 if (n->op->computed_value)
464 return n->op->computed_value(n);
469 * set the default computed_value evaluator
471 static ir_op *firm_set_default_computed_value(ir_op *op)
475 op->computed_value = computed_value_##a; \
500 op->computed_value = NULL;
508 /* returns 1 if the a and b are pointers to different locations. */
510 different_identity (ir_node *a, ir_node *b)
512 assert (mode_is_reference(get_irn_mode (a))
513 && mode_is_reference(get_irn_mode (b)));
515 if (get_irn_op (a) == op_Proj && get_irn_op(b) == op_Proj) {
516 ir_node *a1 = get_Proj_pred (a);
517 ir_node *b1 = get_Proj_pred (b);
518 if (a1 != b1 && get_irn_op (a1) == op_Alloc
519 && get_irn_op (b1) == op_Alloc)
526 static ir_node *equivalent_node_Block(ir_node *n)
530 /* The Block constructor does not call optimize, but mature_block
531 calls the optimization. */
532 assert(get_Block_matured(n));
534 /* Straightening: a single entry Block following a single exit Block
535 can be merged, if it is not the Start block. */
536 /* !!! Beware, all Phi-nodes of n must have been optimized away.
537 This should be true, as the block is matured before optimize is called.
538 But what about Phi-cycles with the Phi0/Id that could not be resolved?
539 Remaining Phi nodes are just Ids. */
540 if ((get_Block_n_cfgpreds(n) == 1) &&
541 (get_irn_op(get_Block_cfgpred(n, 0)) == op_Jmp)) {
542 ir_node *predblock = get_nodes_Block(get_Block_cfgpred(n, 0));
543 if (predblock == oldn) {
544 /* Jmp jumps into the block it is in -- deal self cycle. */
545 n = new_Bad(); DBG_OPT_DEAD;
546 } else if (get_opt_control_flow_straightening()) {
547 n = predblock; DBG_OPT_STG;
550 else if ((get_Block_n_cfgpreds(n) == 1) &&
551 (get_irn_op(skip_Proj(get_Block_cfgpred(n, 0))) == op_Cond)) {
552 ir_node *predblock = get_nodes_Block(get_Block_cfgpred(n, 0));
553 if (predblock == oldn) {
554 /* Jmp jumps into the block it is in -- deal self cycle. */
555 n = new_Bad(); DBG_OPT_DEAD;
558 else if ((get_Block_n_cfgpreds(n) == 2) &&
559 (get_opt_control_flow_weak_simplification())) {
560 /* Test whether Cond jumps twice to this block
561 @@@ we could do this also with two loops finding two preds from several ones. */
562 ir_node *a = get_Block_cfgpred(n, 0);
563 ir_node *b = get_Block_cfgpred(n, 1);
565 if ((get_irn_op(a) == op_Proj) &&
566 (get_irn_op(b) == op_Proj) &&
567 (get_Proj_pred(a) == get_Proj_pred(b)) &&
568 (get_irn_op(get_Proj_pred(a)) == op_Cond) &&
569 (get_irn_mode(get_Cond_selector(get_Proj_pred(a))) == mode_b)) {
570 /* Also a single entry Block following a single exit Block. Phis have
571 twice the same operand and will be optimized away. */
572 n = get_nodes_Block(a); DBG_OPT_IFSIM;
574 } else if (get_opt_unreachable_code() &&
575 (n != current_ir_graph->start_block) &&
576 (n != current_ir_graph->end_block) ) {
578 /* If all inputs are dead, this block is dead too, except if it is
579 the start or end block. This is a step of unreachable code
581 for (i = 0; i < get_Block_n_cfgpreds(n); i++) {
582 if (!is_Bad(get_Block_cfgpred(n, i))) break;
584 if (i == get_Block_n_cfgpreds(n))
592 * Returns a equivalent node for a Jmp, a Bad :-)
593 * Of course this only happens if the Block of the Jmp is Bad.
595 static ir_node *equivalent_node_Jmp(ir_node *n)
597 /* GL: Why not same for op_Raise?? */
598 /* unreachable code elimination */
599 if (is_Bad(get_nodes_Block(n)))
605 static ir_node *equivalent_node_Cond(ir_node *n)
607 /* We do not evaluate Cond here as we replace it by a new node, a Jmp.
608 See cases for iro_Cond and iro_Proj in transform_node. */
613 * Use algebraic simplification a v a = a.
615 static ir_node *equivalent_node_Or(ir_node *n)
619 ir_node *a = get_Or_left(n);
620 ir_node *b = get_Or_right(n);
624 n = a; DBG_OPT_ALGSIM1;
631 * optimize operations that are commutative and have neutral 0,
632 * so a op 0 = 0 op a = a.
634 static ir_node *equivalent_node_neutral_zero(ir_node *n)
638 ir_node *a = get_binop_left(n);
639 ir_node *b = get_binop_right(n);
644 /* After running compute_node there is only one constant predecessor.
645 Find this predecessors value and remember the other node: */
646 if ((tv = computed_value(a)) != tarval_bad) {
648 } else if ((tv = computed_value(b)) != tarval_bad) {
653 /* If this predecessors constant value is zero, the operation is
654 unnecessary. Remove it: */
655 if (tarval_classify (tv) == TV_CLASSIFY_NULL) {
656 n = on; DBG_OPT_ALGSIM1;
662 #define equivalent_node_Add equivalent_node_neutral_zero
663 #define equivalent_node_Eor equivalent_node_neutral_zero
666 * optimize operations that are not commutative but have neutral 0 on left,
669 static ir_node *equivalent_node_left_zero(ir_node *n)
673 ir_node *a = get_binop_left(n);
674 ir_node *b = get_binop_right(n);
676 if (tarval_classify(computed_value(b)) == TV_CLASSIFY_NULL) {
677 n = a; DBG_OPT_ALGSIM1;
683 #define equivalent_node_Sub equivalent_node_left_zero
684 #define equivalent_node_Shl equivalent_node_left_zero
685 #define equivalent_node_Shr equivalent_node_left_zero
686 #define equivalent_node_Shrs equivalent_node_left_zero
687 #define equivalent_node_Rot equivalent_node_left_zero
690 * Er, a "symmetic unop", ie op(op(n)) = n.
692 static ir_node *equivalent_node_symmetric_unop(ir_node *n)
696 /* optimize symmetric unop */
697 if (get_irn_op(get_unop_op(n)) == get_irn_op(n)) {
698 n = get_unop_op(get_unop_op(n)); DBG_OPT_ALGSIM2;
704 #define equivalent_node_Not equivalent_node_symmetric_unop
706 /* --x == x */ /* ??? Is this possible or can --x raise an
707 out of bounds exception if min =! max? */
708 #define equivalent_node_Minus equivalent_node_symmetric_unop
711 * Optimize a * 1 = 1 * a = a.
713 static ir_node *equivalent_node_Mul(ir_node *n)
717 ir_node *a = get_Mul_left(n);
718 ir_node *b = get_Mul_right(n);
720 /* Mul is commutative and has again an other neutral element. */
721 if (tarval_classify (computed_value (a)) == TV_CLASSIFY_ONE) {
722 n = b; DBG_OPT_ALGSIM1;
723 } else if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE) {
724 n = a; DBG_OPT_ALGSIM1;
730 * Optimize a / 1 = a.
732 static ir_node *equivalent_node_Div(ir_node *n)
734 ir_node *a = get_Div_left(n);
735 ir_node *b = get_Div_right(n);
737 /* Div is not commutative. */
738 if (tarval_classify(computed_value(b)) == TV_CLASSIFY_ONE) { /* div(x, 1) == x */
739 /* Turn Div into a tuple (mem, bad, a) */
740 ir_node *mem = get_Div_mem(n);
741 turn_into_tuple(n, 3);
742 set_Tuple_pred(n, pn_Div_M, mem);
743 set_Tuple_pred(n, pn_Div_X_except, new_Bad()); /* no exception */
744 set_Tuple_pred(n, pn_Div_res, a);
750 * Optimize a & 0b1...1 = 0b1...1 & a = a & a = a.
752 static ir_node *equivalent_node_And(ir_node *n)
756 ir_node *a = get_And_left(n);
757 ir_node *b = get_And_right(n);
760 n = a; /* And has it's own neutral element */
761 } else if (tarval_classify(computed_value(a)) == TV_CLASSIFY_ALL_ONE) {
763 } else if (tarval_classify(computed_value(b)) == TV_CLASSIFY_ALL_ONE) {
766 if (n != oldn) DBG_OPT_ALGSIM1;
771 * Try to remove useless conv's:
773 static ir_node *equivalent_node_Conv(ir_node *n)
776 ir_node *a = get_Conv_op(n);
779 ir_mode *n_mode = get_irn_mode(n);
780 ir_mode *a_mode = get_irn_mode(a);
782 if (n_mode == a_mode) { /* No Conv necessary */
783 n = a; DBG_OPT_ALGSIM3;
784 } else if (get_irn_op(a) == op_Conv) { /* Conv(Conv(b)) */
788 n_mode = get_irn_mode(n);
789 b_mode = get_irn_mode(b);
791 if (n_mode == b_mode) {
792 if (n_mode == mode_b) {
793 n = b; /* Convb(Conv*(xxxb(...))) == xxxb(...) */ DBG_OPT_ALGSIM1;
795 else if (mode_is_int(n_mode) || mode_is_character(n_mode)) {
796 if (smaller_mode(b_mode, a_mode)){
797 n = b; /* ConvS(ConvL(xxxS(...))) == xxxS(...) */ DBG_OPT_ALGSIM1;
805 static ir_node *equivalent_node_Phi(ir_node *n)
807 /* Several optimizations:
808 - no Phi in start block.
809 - remove Id operators that are inputs to Phi
810 - fold Phi-nodes, iff they have only one predecessor except
816 ir_node *block = NULL; /* to shutup gcc */
817 ir_node *first_val = NULL; /* to shutup gcc */
818 ir_node *scnd_val = NULL; /* to shutup gcc */
820 if (!get_opt_normalize()) return n;
822 n_preds = get_Phi_n_preds(n);
824 block = get_nodes_block(n);
825 /* @@@ fliegt 'raus, sollte aber doch immer wahr sein!!!
826 assert(get_irn_arity(block) == n_preds && "phi in wrong block!"); */
827 if ((is_Bad(block)) || /* Control dead */
828 (block == current_ir_graph->start_block)) /* There should be no Phi nodes */
829 return new_Bad(); /* in the Start Block. */
831 if (n_preds == 0) return n; /* Phi of dead Region without predecessors. */
834 /* first we test for a special case: */
835 /* Confirm is a special node fixing additional information for a
836 value that is known at a certain point. This is useful for
837 dataflow analysis. */
839 ir_node *a = get_Phi_pred(n, 0);
840 ir_node *b = get_Phi_pred(n, 1);
841 if ( (get_irn_op(a) == op_Confirm)
842 && (get_irn_op(b) == op_Confirm)
843 && follow_Id (get_irn_n(a, 0) == get_irn_n(b, 0))
844 && (get_irn_n(a, 1) == get_irn_n (b, 1))
845 && (a->data.num == (~b->data.num & irpn_True) )) {
846 return get_irn_n(a, 0);
851 /* If the Block has a Bad pred, we also have one. */
852 for (i = 0; i < n_preds; ++i)
853 if (is_Bad (get_Block_cfgpred(block, i)))
854 set_Phi_pred(n, i, new_Bad());
856 /* Find first non-self-referencing input */
857 for (i = 0; i < n_preds; ++i) {
858 first_val = get_Phi_pred(n, i);
859 if ( (first_val != n) /* not self pointer */
861 && (get_irn_op(first_val) != op_Bad)
863 ) { /* value not dead */
864 break; /* then found first value. */
868 /* A totally Bad or self-referencing Phi (we didn't break the above loop) */
869 if (i >= n_preds) { return new_Bad(); }
873 /* follow_Id () for rest of inputs, determine if any of these
874 are non-self-referencing */
875 while (++i < n_preds) {
876 scnd_val = get_Phi_pred(n, i);
878 && (scnd_val != first_val)
880 && (get_irn_op(scnd_val) != op_Bad)
887 /* Fold, if no multiple distinct non-self-referencing inputs */
889 n = first_val; DBG_OPT_PHI;
891 /* skip the remaining Ids (done in get_Phi_pred). */
892 /* superfluous, since we walk all to propagate Block's Bads.
893 while (++i < n_preds) get_Phi_pred(n, i); */
899 * Optimize Loads after Store.
901 * @todo FAILS for volatile entities
903 static ir_node *equivalent_node_Load(ir_node *n)
907 /* remove unnecessary Load. */
908 ir_node *a = skip_Proj(get_Load_mem(n));
909 ir_node *b = get_Load_ptr(n);
912 /* TODO: check for volatile */
913 if (get_irn_op(a) == op_Store && get_Store_ptr(a) == b) {
914 /* We load immediately after a store -- a read after write. */
915 ir_node *mem = get_Load_mem(n);
917 c = get_Store_value(a);
918 turn_into_tuple(n, 3);
919 set_Tuple_pred(n, pn_Load_M, mem);
920 set_Tuple_pred(n, pn_Load_res, c);
921 set_Tuple_pred(n, pn_Load_X_except, new_Bad()); DBG_OPT_RAW;
927 * Optimize store after store and load atfter store.
929 * @todo FAILS for volatile entities
931 static ir_node *equivalent_node_Store(ir_node *n)
935 /* remove unnecessary store. */
936 ir_node *a = skip_Proj(get_Store_mem(n));
937 ir_node *b = get_Store_ptr(n);
938 ir_node *c = skip_Proj(get_Store_value(n));
940 if (get_irn_op(a) == op_Store
941 && get_Store_ptr(a) == b
942 && skip_Proj(get_Store_value(a)) == c) {
943 /* We have twice exactly the same store -- a write after write. */
945 } else if (get_irn_op(c) == op_Load
946 && (a == c || skip_Proj(get_Load_mem(c)) == a)
947 && get_Load_ptr(c) == b ) {
948 /* We just loaded the value from the same memory, i.e., the store
949 doesn't change the memory -- a write after read. */
950 a = get_Store_mem(n);
951 turn_into_tuple(n, 2);
952 set_Tuple_pred(n, pn_Store_M, a);
953 set_Tuple_pred(n, pn_Store_X_except, new_Bad()); DBG_OPT_WAR;
958 static ir_node *equivalent_node_Proj(ir_node *n)
962 ir_node *a = get_Proj_pred(n);
964 if ( get_irn_op(a) == op_Tuple) {
965 /* Remove the Tuple/Proj combination. */
966 if ( get_Proj_proj(n) <= get_Tuple_n_preds(a) ) {
967 n = get_Tuple_pred(a, get_Proj_proj(n)); DBG_OPT_TUPLE;
969 assert(0); /* This should not happen! */
972 } else if (get_irn_mode(n) == mode_X &&
973 is_Bad(get_nodes_Block(n))) {
974 /* Remove dead control flow -- early gigo. */
983 static ir_node *equivalent_node_Id(ir_node *n)
987 n = follow_Id(n); DBG_OPT_ID;
992 case iro_Mod, Quot, DivMod
993 DivMod allocates new nodes --> it's treated in transform node.
994 What about Quot, DivMod?
998 * equivalent_node() returns a node equivalent to input n. It skips all nodes that
999 * perform no actual computation, as, e.g., the Id nodes. It does not create
1000 * new nodes. It is therefore safe to free n if the node returned is not n.
1001 * If a node returns a Tuple we can not just skip it. If the size of the
1002 * in array fits, we transform n into a tuple (e.g., Div).
1005 equivalent_node(ir_node *n)
1007 if (n->op->equivalent_node)
1008 return n->op->equivalent_node(n);
1013 * set the default equivalent node operation
1015 static ir_op *firm_set_default_equivalent_node(ir_op *op)
1019 op->equivalent_node = equivalent_node_##a; \
1041 // CASE(Load); /* dangerous */
1042 // CASE(Store); /* dangerous, see todo */
1046 op->equivalent_node = NULL;
1054 * Do node specific optimizations of nodes predecessors.
1057 optimize_preds(ir_node *n) {
1058 ir_node *a = NULL, *b = NULL;
1060 /* get the operands we will work on for simple cases. */
1062 a = get_binop_left(n);
1063 b = get_binop_right(n);
1064 } else if (is_unop(n)) {
1068 switch (get_irn_opcode(n)) {
1071 /* We don't want Cast as input to Cmp. */
1072 if (get_irn_op(a) == op_Cast) {
1076 if (get_irn_op(b) == op_Cast) {
1078 set_Cmp_right(n, b);
1086 static ir_node *transform_node_Div(ir_node *n)
1088 tarval *tv = computed_value(n);
1090 /* BEWARE: it is NOT possible to optimize a/a to 1, as this may cause a exception */
1092 if (tv != tarval_bad) {
1093 /* Turn Div into a tuple (mem, bad, value) */
1094 ir_node *mem = get_Div_mem(n);
1096 turn_into_tuple(n, 3);
1097 set_Tuple_pred(n, pn_Div_M, mem);
1098 set_Tuple_pred(n, pn_Div_X_except, new_Bad());
1099 set_Tuple_pred(n, pn_Div_res, new_Const(get_tarval_mode(tv), tv));
1104 static ir_node *transform_node_Mod(ir_node *n)
1106 tarval *tv = computed_value(n);
1108 /* BEWARE: it is NOT possible to optimize a%a to 0, as this may cause a exception */
1110 if (tv != tarval_bad) {
1111 /* Turn Mod into a tuple (mem, bad, value) */
1112 ir_node *mem = get_Mod_mem(n);
1113 turn_into_tuple(n, 3);
1114 set_Tuple_pred(n, pn_Mod_M, mem);
1115 set_Tuple_pred(n, pn_Mod_X_except, new_Bad());
1116 set_Tuple_pred(n, pn_Mod_res, new_Const(get_tarval_mode(tv), tv));
1121 static ir_node *transform_node_DivMod(ir_node *n)
1125 ir_node *a = get_DivMod_left(n);
1126 ir_node *b = get_DivMod_right(n);
1127 ir_mode *mode = get_irn_mode(a);
1128 tarval *ta = value_of(a);
1129 tarval *tb = value_of(b);
1131 if (!(mode_is_int(mode) && mode_is_int(get_irn_mode(b))))
1134 /* BEWARE: it is NOT possible to optimize a/a to 1, as this may cause a exception */
1136 if (tb != tarval_bad) {
1137 if (tb == get_mode_one(get_tarval_mode(tb))) {
1138 b = new_Const (mode, get_mode_null(mode));
1140 } else if (ta != tarval_bad) {
1141 tarval *resa, *resb;
1142 resa = tarval_div (ta, tb);
1143 if (resa == tarval_bad) return n; /* Causes exception!!! Model by replacing through
1144 Jmp for X result!? */
1145 resb = tarval_mod (ta, tb);
1146 if (resb == tarval_bad) return n; /* Causes exception! */
1147 a = new_Const (mode, resa);
1148 b = new_Const (mode, resb);
1151 } else if (ta == get_mode_null(mode)) {
1155 if (evaluated) { /* replace by tuple */
1156 ir_node *mem = get_DivMod_mem(n);
1157 turn_into_tuple(n, 4);
1158 set_Tuple_pred(n, pn_DivMod_M, mem);
1159 set_Tuple_pred(n, pn_DivMod_X_except, new_Bad()); /* no exception */
1160 set_Tuple_pred(n, pn_DivMod_res_div, a);
1161 set_Tuple_pred(n, pn_DivMod_res_mod, b);
1162 assert(get_nodes_Block(n));
1168 static ir_node *transform_node_Cond(ir_node *n)
1170 /* Replace the Cond by a Jmp if it branches on a constant
1173 ir_node *a = get_Cond_selector(n);
1174 tarval *ta = value_of(a);
1176 if ((ta != tarval_bad) &&
1177 (get_irn_mode(a) == mode_b) &&
1178 (get_opt_unreachable_code())) {
1179 /* It's a boolean Cond, branching on a boolean constant.
1180 Replace it by a tuple (Bad, Jmp) or (Jmp, Bad) */
1181 jmp = new_r_Jmp(current_ir_graph, get_nodes_Block(n));
1182 turn_into_tuple(n, 2);
1183 if (ta == tarval_b_true) {
1184 set_Tuple_pred(n, pn_Cond_false, new_Bad());
1185 set_Tuple_pred(n, pn_Cond_true, jmp);
1187 set_Tuple_pred(n, pn_Cond_false, jmp);
1188 set_Tuple_pred(n, pn_Cond_true, new_Bad());
1190 /* We might generate an endless loop, so keep it alive. */
1191 add_End_keepalive(get_irg_end(current_ir_graph), get_nodes_Block(n));
1192 } else if ((ta != tarval_bad) &&
1193 (get_irn_mode(a) == mode_Iu) &&
1194 (get_Cond_kind(n) == dense) &&
1195 (get_opt_unreachable_code())) {
1196 /* I don't want to allow Tuples smaller than the biggest Proj.
1197 Also this tuple might get really big...
1198 I generate the Jmp here, and remember it in link. Link is used
1199 when optimizing Proj. */
1200 set_irn_link(n, new_r_Jmp(current_ir_graph, get_nodes_Block(n)));
1201 /* We might generate an endless loop, so keep it alive. */
1202 add_End_keepalive(get_irg_end(current_ir_graph), get_nodes_Block(n));
1203 } else if ((get_irn_op(a) == op_Eor)
1204 && (get_irn_mode(a) == mode_b)
1205 && (tarval_classify(computed_value(get_Eor_right(a))) == TV_CLASSIFY_ONE)) {
1206 /* The Eor is a negate. Generate a new Cond without the negate,
1207 simulate the negate by exchanging the results. */
1208 set_irn_link(n, new_r_Cond(current_ir_graph, get_nodes_Block(n),
1210 } else if ((get_irn_op(a) == op_Not)
1211 && (get_irn_mode(a) == mode_b)) {
1212 /* A Not before the Cond. Generate a new Cond without the Not,
1213 simulate the Not by exchanging the results. */
1214 set_irn_link(n, new_r_Cond(current_ir_graph, get_nodes_Block(n),
1220 static ir_node *transform_node_Eor(ir_node *n)
1222 ir_node *a = get_Eor_left(n);
1223 ir_node *b = get_Eor_right(n);
1225 if ((get_irn_mode(n) == mode_b)
1226 && (get_irn_op(a) == op_Proj)
1227 && (get_irn_mode(a) == mode_b)
1228 && (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE)
1229 && (get_irn_op(get_Proj_pred(a)) == op_Cmp))
1230 /* The Eor negates a Cmp. The Cmp has the negated result anyways! */
1231 n = new_r_Proj(current_ir_graph, get_nodes_Block(n), get_Proj_pred(a),
1232 mode_b, get_negated_pnc(get_Proj_proj(a)));
1233 else if ((get_irn_mode(n) == mode_b)
1234 && (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE))
1235 /* The Eor is a Not. Replace it by a Not. */
1236 /* ????!!!Extend to bitfield 1111111. */
1237 n = new_r_Not(current_ir_graph, get_nodes_Block(n), a, mode_b);
1243 * Transfor a boolean Not.
1245 static ir_node *transform_node_Not(ir_node *n)
1247 ir_node *a = get_Not_op(n);
1249 if ( (get_irn_mode(n) == mode_b)
1250 && (get_irn_op(a) == op_Proj)
1251 && (get_irn_mode(a) == mode_b)
1252 && (get_irn_op(get_Proj_pred(a)) == op_Cmp))
1253 /* We negate a Cmp. The Cmp has the negated result anyways! */
1254 n = new_r_Proj(current_ir_graph, get_nodes_Block(n), get_Proj_pred(a),
1255 mode_b, get_negated_pnc(get_Proj_proj(a)));
1261 * Transform a Div/Mod/DivMod with a non-zero constant. Must be
1262 * done here to avoid that this optimization runs more than once...
1264 static ir_node *transform_node_Proj(ir_node *proj)
1266 ir_node *n = get_Proj_pred(proj);
1270 switch (get_irn_opcode(n)) {
1272 b = get_Div_right(n);
1273 tb = computed_value(b);
1275 if (tb != tarval_bad && tarval_classify(tb) != TV_CLASSIFY_NULL) { /* div(x, c) && c != 0 */
1276 ir_node *div, *proj;
1277 ir_node *a = get_Div_left(n);
1278 ir_node *mem = get_Div_mem(n);
1279 int rem = get_optimize();
1283 div = new_rd_Div(get_irn_dbg_info(n), current_ir_graph,
1284 get_nodes_Block(n), get_irg_initial_mem(current_ir_graph), a, b);
1286 proj = new_r_Proj(current_ir_graph, get_nodes_Block(n), div, get_irn_mode(a), pn_Div_res);
1290 turn_into_tuple(n, 3);
1291 set_Tuple_pred(n, pn_Mod_M, mem);
1292 set_Tuple_pred(n, pn_Mod_X_except, new_Bad());
1293 set_Tuple_pred(n, pn_Mod_res, proj);
1297 b = get_Mod_right(n);
1298 tb = computed_value(b);
1300 if (tb != tarval_bad && tarval_classify(tb) != TV_CLASSIFY_NULL) { /* mod(x, c) && c != 0 */
1301 ir_node *mod, *proj;
1302 ir_node *a = get_Mod_left(n);
1303 ir_node *mem = get_Mod_mem(n);
1304 int rem = get_optimize();
1308 mod = new_rd_Mod(get_irn_dbg_info(n), current_ir_graph,
1309 get_nodes_Block(n), get_irg_initial_mem(current_ir_graph), a, b);
1311 proj = new_r_Proj(current_ir_graph, get_nodes_Block(n), mod, get_irn_mode(a), pn_Mod_res);
1315 turn_into_tuple(n, 3);
1316 set_Tuple_pred(n, pn_Mod_M, mem);
1317 set_Tuple_pred(n, pn_Mod_X_except, new_Bad());
1318 set_Tuple_pred(n, pn_Mod_res, proj);
1322 b = get_DivMod_right(n);
1323 tb = computed_value(b);
1325 if (tb != tarval_bad && tarval_classify(tb) != TV_CLASSIFY_NULL) { /* DivMod(x, c) && c != 0 */
1326 ir_node *div_mod, *proj_div, *proj_mod;
1327 ir_node *a = get_Mod_left(n);
1328 ir_node *mem = get_Mod_mem(n);
1329 int rem = get_optimize();
1333 div_mod = new_rd_DivMod(get_irn_dbg_info(n), current_ir_graph,
1334 get_nodes_Block(n), get_irg_initial_mem(current_ir_graph), a, b);
1336 proj_div = new_r_Proj(current_ir_graph, get_nodes_Block(n), div_mod, get_irn_mode(a), pn_DivMod_res_div);
1337 proj_mod = new_r_Proj(current_ir_graph, get_nodes_Block(n), div_mod, get_irn_mode(a), pn_DivMod_res_mod);
1341 turn_into_tuple(n, 4);
1342 set_Tuple_pred(n, pn_DivMod_M, mem);
1343 set_Tuple_pred(n, pn_DivMod_X_except, new_Bad());
1344 set_Tuple_pred(n, pn_DivMod_res_div, proj_div);
1345 set_Tuple_pred(n, pn_DivMod_res_mod, proj_mod);
1350 if (get_opt_unreachable_code()) {
1351 b = get_Cond_selector(n);
1352 tb = computed_value(b);
1354 if (tb != tarval_bad && mode_is_int(get_tarval_mode(tb))) {
1355 /* we have a constant switch */
1356 long num = get_Proj_proj(proj);
1358 if (num != get_Cond_defaultProj(n)) { /* we cannot optimize default Proj's yet */
1359 if (get_tarval_long(tb) == num) {
1360 /* Do NOT create a jump here, or we will have 2 control flow ops
1361 * in a block. This case is optimized away in optimize_cf(). */
1376 /* we have added a Tuple, optimize it for the current Proj away */
1377 return equivalent_node_Proj(proj);
1381 * Transform a Store before a Store to the same address...
1382 * Both nodes must be in the same block.
1384 * @todo Check for volatile! Moreover, what if the first store
1385 * has a exception handler while the other has not?
1387 static ir_node *transform_node_Store(ir_node *store)
1389 ir_node *pred = skip_Proj(get_Store_mem(store));
1390 ir_node *ptr = get_Store_ptr(store);
1392 if (get_irn_op(pred) == op_Store &&
1393 get_Store_ptr(pred) == ptr &&
1394 get_nodes_block(pred) == get_nodes_block(store)) {
1395 /* the Store n is useless, as it is overwritten by the store store */
1396 ir_node *mem = get_Store_mem(pred);
1398 turn_into_tuple(pred, 2);
1399 set_Tuple_pred(pred, pn_Store_M, mem);
1400 set_Tuple_pred(pred, pn_Store_X_except, new_Bad());
1406 * returns the operands of a commutative bin-op, if one operand is
1407 * a const, it is returned as the second one.
1409 static void get_comm_Binop_Ops(ir_node *binop, ir_node **a, ir_node **c)
1411 ir_node *op_a = get_binop_left(binop);
1412 ir_node *op_b = get_binop_right(binop);
1414 assert(is_op_commutative(get_irn_op(binop)));
1416 if (get_irn_op(op_a) == op_Const) {
1427 * Optimize a Or(And(Or(And(v,c4),c3),c2),c1) pattern if possible.
1428 * Such pattern may arise in bitfield stores.
1430 * value c4 value c4 & c2
1431 * AND c3 AND c1 | c3
1436 static ir_node *transform_node_Or(ir_node *or)
1440 ir_node *and_l, *c3;
1441 ir_node *value, *c4;
1442 ir_node *new_and, *new_const, *block;
1443 ir_mode *mode = get_irn_mode(or);
1445 tarval *tv1, *tv2, *tv3, *tv4, *tv, *n_tv4, *n_tv2;
1447 get_comm_Binop_Ops(or, &and, &c1);
1448 if ((get_irn_op(c1) != op_Const) || (get_irn_op(and) != op_And))
1451 get_comm_Binop_Ops(and, &or_l, &c2);
1452 if ((get_irn_op(c2) != op_Const) || (get_irn_op(or_l) != op_Or))
1455 get_comm_Binop_Ops(or_l, &and_l, &c3);
1456 if ((get_irn_op(c3) != op_Const) || (get_irn_op(and_l) != op_And))
1459 get_comm_Binop_Ops(and_l, &value, &c4);
1460 if (get_irn_op(c4) != op_Const)
1463 /* ok, found the pattern, check for conditions */
1464 assert(mode == get_irn_mode(and));
1465 assert(mode == get_irn_mode(or_l));
1466 assert(mode == get_irn_mode(and_l));
1468 tv1 = get_Const_tarval(c1);
1469 tv2 = get_Const_tarval(c2);
1470 tv3 = get_Const_tarval(c3);
1471 tv4 = get_Const_tarval(c4);
1473 tv = tarval_or(tv4, tv2);
1474 if (tarval_classify(tv) != TV_CLASSIFY_ALL_ONE) {
1475 /* have at least one 0 at the same bit position */
1479 n_tv4 = tarval_not(tv4);
1480 if (tv3 != tarval_and(tv3, n_tv4)) {
1481 /* bit in the or_mask is outside the and_mask */
1485 n_tv2 = tarval_not(tv2);
1486 if (tv1 != tarval_and(tv1, n_tv2)) {
1487 /* bit in the or_mask is outside the and_mask */
1491 /* ok, all conditions met */
1492 block = get_nodes_block(or);
1494 new_and = new_r_And(current_ir_graph, block,
1495 value, new_r_Const(current_ir_graph, block, mode, tarval_and(tv4, tv2)), mode);
1497 new_const = new_r_Const(current_ir_graph, block, mode, tarval_or(tv3, tv1));
1499 set_Or_left(or, new_and);
1500 set_Or_right(or, new_const);
1502 /* check for more */
1503 return transform_node_Or(or);
1507 * Tries several [inplace] [optimizing] transformations and returns an
1508 * equivalent node. The difference to equivalent_node() is that these
1509 * transformations _do_ generate new nodes, and thus the old node must
1510 * not be freed even if the equivalent node isn't the old one.
1512 static ir_node *transform_node(ir_node *n)
1514 if (n->op->transform_node)
1515 n = n->op->transform_node(n);
1520 * set the default transform node operation
1522 static ir_op *firm_set_default_transform_node(ir_op *op)
1526 op->transform_node = transform_node_##a; \
1537 // CASE(Store); /* dangerous, see todo */
1540 op->transform_node = NULL;
1548 /* **************** Common Subexpression Elimination **************** */
1550 /** The size of the hash table used, should estimate the number of nodes
1552 #define N_IR_NODES 512
1554 static int node_cmp_attr_Const(ir_node *a, ir_node *b)
1556 return (get_Const_tarval(a) != get_Const_tarval(b))
1557 || (get_Const_type(a) != get_Const_type(b));
1560 static int node_cmp_attr_Proj(ir_node *a, ir_node *b)
1562 return get_irn_proj_attr (a) != get_irn_proj_attr (b);
1565 static int node_cmp_attr_Filter(ir_node *a, ir_node *b)
1567 return get_Filter_proj(a) != get_Filter_proj(b);
1570 static int node_cmp_attr_Alloc(ir_node *a, ir_node *b)
1572 return (get_irn_alloc_attr(a).where != get_irn_alloc_attr(b).where)
1573 || (get_irn_alloc_attr(a).type != get_irn_alloc_attr(b).type);
1576 static int node_cmp_attr_Free(ir_node *a, ir_node *b)
1578 return (get_irn_free_attr(a) != get_irn_free_attr(b));
1581 static int node_cmp_attr_SymConst(ir_node *a, ir_node *b)
1583 return (get_irn_symconst_attr(a).num != get_irn_symconst_attr(b).num)
1584 || (get_irn_symconst_attr(a).sym.type_p != get_irn_symconst_attr(b).sym.type_p);
1587 static int node_cmp_attr_Call(ir_node *a, ir_node *b)
1589 return (get_irn_call_attr(a) != get_irn_call_attr(b));
1592 static int node_cmp_attr_FuncCall(ir_node *a, ir_node *b)
1594 return (get_irn_funccall_attr(a) != get_irn_funccall_attr(b));
1597 static int node_cmp_attr_Sel(ir_node *a, ir_node *b)
1599 return (get_irn_sel_attr(a).ent->kind != get_irn_sel_attr(b).ent->kind)
1600 || (get_irn_sel_attr(a).ent->name != get_irn_sel_attr(b).ent->name)
1601 || (get_irn_sel_attr(a).ent->owner != get_irn_sel_attr(b).ent->owner)
1602 || (get_irn_sel_attr(a).ent->ld_name != get_irn_sel_attr(b).ent->ld_name)
1603 || (get_irn_sel_attr(a).ent->type != get_irn_sel_attr(b).ent->type);
1606 static int node_cmp_attr_Phi(ir_node *a, ir_node *b)
1608 return get_irn_phi_attr (a) != get_irn_phi_attr (b);
1611 static int node_cmp_attr_Cast(ir_node *a, ir_node *b)
1613 return get_Cast_type(a) != get_Cast_type(b);
1617 * set the default node attribute compare operation
1619 static ir_op *firm_set_default_node_cmp_attr(ir_op *op)
1623 op->node_cmp_attr = node_cmp_attr_##a; \
1639 op->node_cmp_attr = NULL;
1647 * Compare function for two nodes in the hash table. Gets two
1648 * nodes as parameters. Returns 0 if the nodes are a cse.
1651 vt_cmp (const void *elt, const void *key)
1659 if (a == b) return 0;
1661 if ((get_irn_op(a) != get_irn_op(b)) ||
1662 (get_irn_mode(a) != get_irn_mode(b))) return 1;
1664 /* compare if a's in and b's in are equal */
1665 irn_arity_a = get_irn_arity (a);
1666 if (irn_arity_a != get_irn_arity(b))
1669 /* for block-local cse and pinned nodes: */
1670 if (!get_opt_global_cse() || (get_op_pinned(get_irn_op(a)) == pinned)) {
1671 if (get_irn_n(a, -1) != get_irn_n(b, -1))
1675 /* compare a->in[0..ins] with b->in[0..ins] */
1676 for (i = 0; i < irn_arity_a; i++)
1677 if (get_irn_n(a, i) != get_irn_n(b, i))
1681 * here, we already now that the nodes are identical except their
1684 if (a->op->node_cmp_attr)
1685 return a->op->node_cmp_attr(a, b);
1691 * Calculate a hash value of a node.
1694 ir_node_hash (ir_node *node)
1699 if (node->op == op_Const) {
1700 /* special value for const, as they only differ in their tarval. */
1701 h = ((unsigned) node->attr.con.tv)>>3 ;
1702 h = 9*h + (unsigned)get_irn_mode(node);
1703 } else if (node->op == op_SymConst) {
1704 /* special value for const, as they only differ in their symbol. */
1705 h = ((unsigned) node->attr.i.sym.type_p)>>3 ;
1706 h = 9*h + (unsigned)get_irn_mode(node);
1709 /* hash table value = 9*(9*(9*(9*(9*arity+in[0])+in[1])+ ...)+mode)+code */
1710 h = irn_arity = get_irn_arity(node);
1712 /* consider all in nodes... except the block. */
1713 for (i = 0; i < irn_arity; i++) {
1714 h = 9*h + (unsigned)get_irn_n(node, i);
1718 h = 9*h + (unsigned) get_irn_mode (node);
1720 h = 9*h + (unsigned) get_irn_op (node);
1727 new_identities (void)
1729 return new_pset (vt_cmp, N_IR_NODES);
1733 del_identities (pset *value_table)
1735 del_pset (value_table);
1739 * Return the canonical node computing the same value as n.
1740 * Looks up the node in a hash table.
1742 * For Const nodes this is performed in the constructor, too. Const
1743 * nodes are extremely time critical because of their frequent use in
1744 * constant string arrays.
1746 static INLINE ir_node *
1747 identify (pset *value_table, ir_node *n)
1751 if (!value_table) return n;
1753 /* TODO: use a generic commutative attribute */
1754 if (get_opt_reassociation()) {
1755 if (is_op_commutative(get_irn_op(n))) {
1756 /* for commutative operators perform a OP b == b OP a */
1757 if (get_binop_left(n) > get_binop_right(n)) {
1758 ir_node *h = get_binop_left(n);
1759 set_binop_left(n, get_binop_right(n));
1760 set_binop_right(n, h);
1765 o = pset_find (value_table, n, ir_node_hash (n));
1772 * During construction we set the pinned flag in the graph right when the
1773 * optimization is performed. The flag turning on procedure global cse could
1774 * be changed between two allocations. This way we are safe.
1776 static INLINE ir_node *
1777 identify_cons (pset *value_table, ir_node *n) {
1779 n = identify(value_table, n);
1780 if (get_irn_n(old, -1) != get_irn_n(n, -1))
1781 set_irg_pinned(current_ir_graph, floats);
1786 * Return the canonical node computing the same value as n.
1787 * Looks up the node in a hash table, enters it in the table
1788 * if it isn't there yet.
1791 identify_remember (pset *value_table, ir_node *node)
1795 if (!value_table) return node;
1797 /* lookup or insert in hash table with given hash key. */
1798 o = pset_insert (value_table, node, ir_node_hash (node));
1800 if (o == node) return node;
1806 add_identities (pset *value_table, ir_node *node) {
1807 identify_remember (value_table, node);
1811 * garbage in, garbage out. If a node has a dead input, i.e., the
1812 * Bad node is input to the node, return the Bad node.
1814 static INLINE ir_node *
1815 gigo (ir_node *node)
1818 ir_op* op = get_irn_op(node);
1820 /* remove garbage blocks by looking at control flow that leaves the block
1821 and replacing the control flow by Bad. */
1822 if (get_irn_mode(node) == mode_X) {
1823 ir_node *block = get_nodes_block(node);
1824 if (op == op_End) return node; /* Don't optimize End, may have Bads. */
1825 if (get_irn_op(block) == op_Block && get_Block_matured(block)) {
1826 irn_arity = get_irn_arity(block);
1827 for (i = 0; i < irn_arity; i++) {
1828 if (!is_Bad(get_irn_n(block, i))) break;
1830 if (i == irn_arity) return new_Bad();
1834 /* Blocks, Phis and Tuples may have dead inputs, e.g., if one of the
1835 blocks predecessors is dead. */
1836 if ( op != op_Block && op != op_Phi && op != op_Tuple) {
1837 irn_arity = get_irn_arity(node);
1838 for (i = -1; i < irn_arity; i++) {
1839 if (is_Bad(get_irn_n(node, i))) {
1845 /* With this code we violate the agreement that local_optimize
1846 only leaves Bads in Block, Phi and Tuple nodes. */
1847 /* If Block has only Bads as predecessors it's garbage. */
1848 /* If Phi has only Bads as predecessors it's garbage. */
1849 if ((op == op_Block && get_Block_matured(node)) || op == op_Phi) {
1850 irn_arity = get_irn_arity(node);
1851 for (i = 0; i < irn_arity; i++) {
1852 if (!is_Bad(get_irn_n(node, i))) break;
1854 if (i == irn_arity) node = new_Bad();
1862 * These optimizations deallocate nodes from the obstack.
1863 * It can only be called if it is guaranteed that no other nodes
1864 * reference this one, i.e., right after construction of a node.
1867 optimize_node (ir_node *n)
1871 opcode iro = get_irn_opcode(n);
1873 /* Allways optimize Phi nodes: part of the construction. */
1874 if ((!get_opt_optimize()) && (iro != iro_Phi)) return n;
1876 /* constant expression evaluation / constant folding */
1877 if (get_opt_constant_folding()) {
1878 /* constants can not be evaluated */
1879 if (iro != iro_Const) {
1880 /* try to evaluate */
1881 tv = computed_value (n);
1882 if ((get_irn_mode(n) != mode_T) && (tv != tarval_bad)) {
1884 * we MUST copy the node here temparary, because it's still needed
1885 * for DBG_OPT_ALGSIM0
1887 int node_size = offsetof(ir_node, attr) + n->op->attr_size;
1888 ir_node *x = alloca(node_size);
1890 memcpy(x, n, node_size);
1893 /* evaluation was successful -- replace the node. */
1894 obstack_free (current_ir_graph->obst, n);
1895 n = new_Const (get_tarval_mode (tv), tv);
1902 /* remove unnecessary nodes */
1903 if (get_opt_constant_folding() ||
1904 (iro == iro_Phi) || /* always optimize these nodes. */
1906 (iro == iro_Proj) ||
1907 (iro == iro_Block) ) /* Flags tested local. */
1908 n = equivalent_node (n);
1910 optimize_preds(n); /* do node specific optimizations of nodes predecessors. */
1912 /** common subexpression elimination **/
1913 /* Checks whether n is already available. */
1914 /* The block input is used to distinguish different subexpressions. Right
1915 now all nodes are pinned to blocks, i.e., the cse only finds common
1916 subexpressions within a block. */
1918 n = identify_cons (current_ir_graph->value_table, n);
1921 /* We found an existing, better node, so we can deallocate the old node. */
1922 obstack_free (current_ir_graph->obst, oldn);
1927 /* Some more constant expression evaluation that does not allow to
1929 iro = get_irn_opcode(n);
1930 if (get_opt_constant_folding() ||
1931 (iro == iro_Cond) ||
1932 (iro == iro_Proj)) /* Flags tested local. */
1933 n = transform_node (n);
1935 /* Remove nodes with dead (Bad) input.
1936 Run always for transformation induced Bads. */
1939 /* Now we have a legal, useful node. Enter it in hash table for cse */
1940 if (get_opt_cse() && (get_irn_opcode(n) != iro_Block)) {
1941 n = identify_remember (current_ir_graph->value_table, n);
1949 * These optimizations never deallocate nodes. This can cause dead
1950 * nodes lying on the obstack. Remove these by a dead node elimination,
1951 * i.e., a copying garbage collection.
1954 optimize_in_place_2 (ir_node *n)
1958 opcode iro = get_irn_opcode(n);
1960 if (!get_opt_optimize() && (get_irn_op(n) != op_Phi)) return n;
1962 /* if not optimize return n */
1965 /* Here this is possible. Why? */
1970 /* constant expression evaluation / constant folding */
1971 if (get_opt_constant_folding()) {
1972 /* constants can not be evaluated */
1973 if (iro != iro_Const) {
1974 /* try to evaluate */
1975 tv = computed_value (n);
1976 if ((get_irn_mode(n) != mode_T) && (tv != tarval_bad)) {
1977 /* evaluation was successful -- replace the node. */
1978 n = new_Const (get_tarval_mode (tv), tv);
1985 /* remove unnecessary nodes */
1986 /*if (get_opt_constant_folding()) */
1987 if (get_opt_constant_folding() ||
1988 (iro == iro_Phi) || /* always optimize these nodes. */
1989 (iro == iro_Id) || /* ... */
1990 (iro == iro_Proj) || /* ... */
1991 (iro == iro_Block) ) /* Flags tested local. */
1992 n = equivalent_node (n);
1994 optimize_preds(n); /* do node specific optimizations of nodes predecessors. */
1996 /** common subexpression elimination **/
1997 /* Checks whether n is already available. */
1998 /* The block input is used to distinguish different subexpressions. Right
1999 now all nodes are pinned to blocks, i.e., the cse only finds common
2000 subexpressions within a block. */
2001 if (get_opt_cse()) {
2002 n = identify (current_ir_graph->value_table, n);
2005 /* Some more constant expression evaluation. */
2006 iro = get_irn_opcode(n);
2007 if (get_opt_constant_folding() ||
2008 (iro == iro_Cond) ||
2009 (iro == iro_Proj)) /* Flags tested local. */
2010 n = transform_node (n);
2012 /* Remove nodes with dead (Bad) input.
2013 Run always for transformation induced Bads. */
2016 /* Now we can verify the node, as it has no dead inputs any more. */
2019 /* Now we have a legal, useful node. Enter it in hash table for cse.
2020 Blocks should be unique anyways. (Except the successor of start:
2021 is cse with the start block!) */
2022 if (get_opt_cse() && (get_irn_opcode(n) != iro_Block))
2023 n = identify_remember (current_ir_graph->value_table, n);
2029 * Wrapper for external use, set proper status bits after optimization.
2032 optimize_in_place (ir_node *n)
2034 /* Handle graph state */
2035 assert(get_irg_phase_state(current_ir_graph) != phase_building);
2037 if (get_opt_global_cse())
2038 set_irg_pinned(current_ir_graph, floats);
2039 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
2040 set_irg_outs_inconsistent(current_ir_graph);
2041 /* Maybe we could also test whether optimizing the node can
2042 change the control graph. */
2043 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
2044 set_irg_dom_inconsistent(current_ir_graph);
2045 return optimize_in_place_2 (n);
2049 * set the default ir op operations
2051 ir_op *firm_set_default_operations(ir_op *op)
2053 op = firm_set_default_computed_value(op);
2054 op = firm_set_default_equivalent_node(op);
2055 op = firm_set_default_transform_node(op);
2056 op = firm_set_default_node_cmp_attr(op);