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)); DBG_OPT_STG;
502 } else if ((get_Block_n_cfgpreds(n) == 2) &&
503 (get_opt_control_flow_weak_simplification())) {
504 /* Test whether Cond jumps twice to this block
505 @@@ we could do this also with two loops finding two preds from several ones. */
506 ir_node *a = get_Block_cfgpred(n, 0);
507 ir_node *b = get_Block_cfgpred(n, 1);
509 if ((get_irn_op(a) == op_Proj) &&
510 (get_irn_op(b) == op_Proj) &&
511 (get_Proj_pred(a) == get_Proj_pred(b)) &&
512 (get_irn_op(get_Proj_pred(a)) == op_Cond) &&
513 (get_irn_mode(get_Cond_selector(get_Proj_pred(a))) == mode_b)) {
514 /* Also a single entry Block following a single exit Block. Phis have
515 twice the same operand and will be optimized away. */
516 n = get_nodes_Block(a); DBG_OPT_IFSIM;
518 } else if (get_opt_unreachable_code() &&
519 (n != current_ir_graph->start_block) &&
520 (n != current_ir_graph->end_block) ) {
522 /* If all inputs are dead, this block is dead too, except if it is
523 the start or end block. This is a step of unreachable code
525 for (i = 0; i < get_Block_n_cfgpreds(n); i++) {
526 if (!is_Bad(get_Block_cfgpred(n, i))) break;
528 if (i == get_Block_n_cfgpreds(n))
535 static ir_node *equivalent_node_Jmp(ir_node *n)
537 /* GL: Why not same for op_Raise?? */
538 /* unreachable code elimination */
539 if (is_Bad(get_nodes_Block(n)))
545 static ir_node *equivalent_node_Cond(ir_node *n)
547 /* We do not evaluate Cond here as we replace it by a new node, a Jmp.
548 See cases for iro_Cond and iro_Proj in transform_node. */
552 static ir_node *equivalent_node_Or(ir_node *n)
556 ir_node *a = get_Or_left(n);
557 ir_node *b = get_Or_right(n);
561 n = a; DBG_OPT_ALGSIM1;
568 * optimize operations that are commutative and have neutral 0.
570 static ir_node *equivalent_node_neutral_zero(ir_node *n)
574 ir_node *a = get_binop_left(n);
575 ir_node *b = get_binop_right(n);
580 /* After running compute_node there is only one constant predecessor.
581 Find this predecessors value and remember the other node: */
582 if ((tv = computed_value (a)) != tarval_bad) {
584 } else if ((tv = computed_value (b)) != tarval_bad) {
589 /* If this predecessors constant value is zero, the operation is
590 unnecessary. Remove it: */
591 if (tarval_classify (tv) == TV_CLASSIFY_NULL) {
592 n = on; DBG_OPT_ALGSIM1;
598 static ir_node *equivalent_node_Add(ir_node *n)
600 return equivalent_node_neutral_zero(n);
603 static ir_node *equivalent_node_Eor(ir_node *n)
605 return equivalent_node_neutral_zero(n);
609 * optimize operations that are not commutative but have neutral 0 on left.
610 * Test only one predecessor.
612 static ir_node *equivalent_node_left_zero(ir_node *n)
616 ir_node *a = get_binop_left(n);
617 ir_node *b = get_binop_right(n);
619 if (tarval_classify (computed_value (b)) == TV_CLASSIFY_NULL) {
620 n = a; DBG_OPT_ALGSIM1;
626 static ir_node *equivalent_node_Sub(ir_node *n)
628 return equivalent_node_left_zero(n);
631 static ir_node *equivalent_node_Shl(ir_node *n)
633 return equivalent_node_left_zero(n);
636 static ir_node *equivalent_node_Shr(ir_node *n)
638 return equivalent_node_left_zero(n);
641 static ir_node *equivalent_node_Shrs(ir_node *n)
643 return equivalent_node_left_zero(n);
646 static ir_node *equivalent_node_Rot(ir_node *n)
648 return equivalent_node_left_zero(n);
651 static ir_node *equivalent_node_symmetric_unop(ir_node *n)
655 /* optimize symmetric unop */
656 if (get_irn_op(get_unop_op(n)) == get_irn_op(n)) {
657 n = get_unop_op(get_unop_op(n)); DBG_OPT_ALGSIM2;
662 static ir_node *equivalent_node_Not(ir_node *n)
665 return equivalent_node_symmetric_unop(n);
668 static ir_node *equivalent_node_Minus(ir_node *n)
670 /* --x == x */ /* ??? Is this possible or can --x raise an
671 out of bounds exception if min =! max? */
672 return equivalent_node_symmetric_unop(n);
675 static ir_node *equivalent_node_Mul(ir_node *n)
679 ir_node *a = get_Mul_left(n);
680 ir_node *b = get_Mul_right(n);
682 /* Mul is commutative and has again an other neutral element. */
683 if (tarval_classify (computed_value (a)) == TV_CLASSIFY_ONE) {
684 n = b; DBG_OPT_ALGSIM1;
685 } else if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE) {
686 n = a; DBG_OPT_ALGSIM1;
691 static ir_node *equivalent_node_Div(ir_node *n)
693 ir_node *a = get_Div_left(n);
694 ir_node *b = get_Div_right(n);
696 /* Div is not commutative. */
697 if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE) { /* div(x, 1) == x */
698 /* Turn Div into a tuple (mem, bad, a) */
699 ir_node *mem = get_Div_mem(n);
700 turn_into_tuple(n, 3);
701 set_Tuple_pred(n, pn_Div_M, mem);
702 set_Tuple_pred(n, pn_Div_X_except, new_Bad()); /* no exception */
703 set_Tuple_pred(n, pn_Div_res, a);
708 static ir_node *equivalent_node_And(ir_node *n)
712 ir_node *a = get_And_left(n);
713 ir_node *b = get_And_right(n);
716 n = a; /* And has it's own neutral element */
717 } else if (tarval_classify (computed_value (a)) == TV_CLASSIFY_ALL_ONE) {
719 } else if (tarval_classify (computed_value (b)) == TV_CLASSIFY_ALL_ONE) {
722 if (n != oldn) DBG_OPT_ALGSIM1;
726 static ir_node *equivalent_node_Conv(ir_node *n)
729 ir_node *a = get_Conv_op(n);
732 ir_mode *n_mode = get_irn_mode(n);
733 ir_mode *a_mode = get_irn_mode(a);
735 if (n_mode == a_mode) { /* No Conv necessary */
736 n = a; DBG_OPT_ALGSIM3;
737 } else if (get_irn_op(a) == op_Conv) { /* Conv(Conv(b)) */
741 n_mode = get_irn_mode(n);
742 b_mode = get_irn_mode(b);
744 if (n_mode == b_mode) {
745 if (n_mode == mode_b) {
746 n = b; /* Convb(Conv*(xxxb(...))) == xxxb(...) */ DBG_OPT_ALGSIM1;
748 else if (mode_is_int(n_mode) || mode_is_character(n_mode)) {
749 if (smaller_mode(b_mode, a_mode)){
750 n = b; /* ConvS(ConvL(xxxS(...))) == xxxS(...) */ DBG_OPT_ALGSIM1;
758 static ir_node *equivalent_node_Phi(ir_node *n)
760 /* Several optimizations:
761 - no Phi in start block.
762 - remove Id operators that are inputs to Phi
763 - fold Phi-nodes, iff they have only one predecessor except
769 ir_node *block = NULL; /* to shutup gcc */
770 ir_node *first_val = NULL; /* to shutup gcc */
771 ir_node *scnd_val = NULL; /* to shutup gcc */
773 if (!get_opt_normalize()) return n;
775 n_preds = get_Phi_n_preds(n);
777 block = get_nodes_Block(n);
778 /* @@@ fliegt 'raus, sollte aber doch immer wahr sein!!!
779 assert(get_irn_arity(block) == n_preds && "phi in wrong block!"); */
780 if ((is_Bad(block)) || /* Control dead */
781 (block == current_ir_graph->start_block)) /* There should be no Phi nodes */
782 return new_Bad(); /* in the Start Block. */
784 if (n_preds == 0) return n; /* Phi of dead Region without predecessors. */
787 /* first we test for a special case: */
788 /* Confirm is a special node fixing additional information for a
789 value that is known at a certain point. This is useful for
790 dataflow analysis. */
792 ir_node *a = follow_Id (get_Phi_pred(n, 0));
793 ir_node *b = follow_Id (get_Phi_pred(n, 1));
794 if ( (get_irn_op(a) == op_Confirm)
795 && (get_irn_op(b) == op_Confirm)
796 && follow_Id (get_irn_n(a, 0) == get_irn_n(b, 0))
797 && (get_irn_n(a, 1) == get_irn_n (b, 1))
798 && (a->data.num == (~b->data.num & irpn_True) )) {
799 return get_irn_n(a, 0);
804 /* Find first non-self-referencing input */
805 for (i = 0; i < n_preds; ++i) {
806 first_val = follow_Id(get_Phi_pred(n, i));
808 set_Phi_pred(n, i, first_val);
809 if ( (first_val != n) /* not self pointer */
810 && (get_irn_op(first_val) != op_Bad) /* value not dead */
811 && !(is_Bad (get_Block_cfgpred(block, i))) ) { /* not dead control flow */
812 break; /* then found first value. */
816 /* A totally Bad or self-referencing Phi (we didn't break the above loop) */
817 if (i >= n_preds) { return new_Bad(); }
821 /* follow_Id () for rest of inputs, determine if any of these
822 are non-self-referencing */
823 while (++i < n_preds) {
824 scnd_val = follow_Id(get_Phi_pred(n, i));
826 set_Phi_pred(n, i, scnd_val);
828 && (scnd_val != first_val)
829 && (get_irn_op(scnd_val) != op_Bad)
830 && !(is_Bad (get_Block_cfgpred(block, i))) ) {
835 /* Fold, if no multiple distinct non-self-referencing inputs */
837 n = first_val; DBG_OPT_PHI;
839 /* skip the remaining Ids. */
840 while (++i < n_preds) {
841 set_Phi_pred(n, i, follow_Id(get_Phi_pred(n, i)));
847 static ir_node *equivalent_node_Load(ir_node *n)
849 #if 0 /* Is an illegal transformation: different nodes can
850 represent the same pointer value!! */
851 ir_node *a = skip_Proj(get_Load_mem(n));
852 ir_node *b = get_Load_ptr(n);
854 if (get_irn_op(a) == op_Store) {
855 if ( different_identity (b, get_Store_ptr(a))) {
856 /* load and store use different pointers, therefore load
857 needs not take store's memory but the state before. */
858 set_Load_mem (n, get_Store_mem(a));
859 } else if (( 0 /* ???didn't get cryptic test that returns 0 */ )) {
866 static ir_node *equivalent_node_Store(ir_node *n)
870 /* remove unnecessary store. */
871 ir_node *a = skip_Proj(get_Store_mem(n));
872 ir_node *b = get_Store_ptr(n);
873 ir_node *c = skip_Proj(get_Store_value(n));
875 if (get_irn_op(a) == op_Store
876 && get_Store_ptr(a) == b
877 && skip_Proj(get_Store_value(a)) == c) {
878 /* We have twice exactly the same store -- a write after write. */
880 } else if (get_irn_op(c) == op_Load
881 && (a == c || skip_Proj(get_Load_mem(c)) == a)
882 && get_Load_ptr(c) == b ) {
883 /* We just loaded the value from the same memory, i.e., the store
884 doesn't change the memory -- a write after read. */
885 a = get_Store_mem(n);
886 turn_into_tuple(n, 2);
887 set_Tuple_pred(n, pn_Store_M, a);
888 set_Tuple_pred(n, pn_Store_X_except, new_Bad()); DBG_OPT_WAR;
893 static ir_node *equivalent_node_Proj(ir_node *n)
897 ir_node *a = get_Proj_pred(n);
899 if ( get_irn_op(a) == op_Tuple) {
900 /* Remove the Tuple/Proj combination. */
901 if ( get_Proj_proj(n) <= get_Tuple_n_preds(a) ) {
902 n = get_Tuple_pred(a, get_Proj_proj(n)); DBG_OPT_TUPLE;
904 assert(0); /* This should not happen! */
907 } else if (get_irn_mode(n) == mode_X &&
908 is_Bad(get_nodes_Block(n))) {
909 /* Remove dead control flow -- early gigo. */
915 static ir_node *equivalent_node_Id(ir_node *n)
919 n = follow_Id (n); DBG_OPT_ID;
924 case iro_Mod, Quot, DivMod
925 DivMod allocates new nodes --> it's treated in transform node.
926 What about Quot, DivMod?
930 * equivalent_node() returns a node equivalent to input n. It skips all nodes that
931 * perform no actual computation, as, e.g., the Id nodes. It does not create
932 * new nodes. It is therefore safe to free n if the node returned is not n.
933 * If a node returns a Tuple we can not just skip it. If the size of the
934 * in array fits, we transform n into a tuple (e.g., Div).
937 equivalent_node (ir_node *n)
939 if (n->op->equivalent_node)
940 return n->op->equivalent_node(n);
945 * set the default equivalent node operation
947 static ir_op *firm_set_default_equivalent_node(ir_op *op)
951 op->equivalent_node = equivalent_node_##a; \
978 op->equivalent_node = NULL;
986 * Do node specific optimizations of nodes predecessors.
989 optimize_preds(ir_node *n) {
990 ir_node *a = NULL, *b = NULL;
992 /* get the operands we will work on for simple cases. */
994 a = get_binop_left(n);
995 b = get_binop_right(n);
996 } else if (is_unop(n)) {
1000 switch (get_irn_opcode(n)) {
1003 /* We don't want Cast as input to Cmp. */
1004 if (get_irn_op(a) == op_Cast) {
1008 if (get_irn_op(b) == op_Cast) {
1010 set_Cmp_right(n, b);
1018 static ir_node *transform_node_Div(ir_node *n)
1020 tarval *ta = computed_value(n);
1022 if (ta != tarval_bad) {
1023 /* Turn Div into a tuple (mem, bad, value) */
1024 ir_node *mem = get_Div_mem(n);
1026 turn_into_tuple(n, 3);
1027 set_Tuple_pred(n, pn_Div_M, mem);
1028 set_Tuple_pred(n, pn_Div_X_except, new_Bad());
1029 set_Tuple_pred(n, pn_Div_res, new_Const(get_tarval_mode(ta), ta));
1034 static ir_node *transform_node_Mod(ir_node *n)
1036 tarval *ta = computed_value(n);
1038 if (ta != tarval_bad) {
1039 /* Turn Mod into a tuple (mem, bad, value) */
1040 ir_node *mem = get_Mod_mem(n);
1041 turn_into_tuple(n, 3);
1042 set_Tuple_pred(n, pn_Mod_M, mem);
1043 set_Tuple_pred(n, pn_Mod_X_except, new_Bad());
1044 set_Tuple_pred(n, pn_Mod_res, new_Const(get_tarval_mode(ta), ta));
1049 static ir_node *transform_node_DivMod(ir_node *n)
1053 ir_node *a = get_DivMod_left(n);
1054 ir_node *b = get_DivMod_right(n);
1055 ir_mode *mode = get_irn_mode(a);
1057 if (!(mode_is_int(mode) && mode_is_int(get_irn_mode(b))))
1061 a = new_Const(mode, get_mode_one(mode));
1062 b = new_Const(mode, get_mode_null(mode));
1065 tarval *ta = value_of(a);
1066 tarval *tb = value_of(b);
1068 if (tb != tarval_bad) {
1069 if (tb == get_mode_one(get_tarval_mode(tb))) {
1070 b = new_Const (mode, get_mode_null(mode));
1072 } else if (ta != tarval_bad) {
1073 tarval *resa, *resb;
1074 resa = tarval_div (ta, tb);
1075 if (resa == tarval_bad) return n; /* Causes exception!!! Model by replacing through
1076 Jmp for X result!? */
1077 resb = tarval_mod (ta, tb);
1078 if (resb == tarval_bad) return n; /* Causes exception! */
1079 a = new_Const (mode, resa);
1080 b = new_Const (mode, resb);
1083 } else if (ta == get_mode_null(mode)) {
1088 if (evaluated) { /* replace by tuple */
1089 ir_node *mem = get_DivMod_mem(n);
1090 turn_into_tuple(n, 4);
1091 set_Tuple_pred(n, pn_DivMod_M, mem);
1092 set_Tuple_pred(n, pn_DivMod_X_except, new_Bad()); /* no exception */
1093 set_Tuple_pred(n, pn_DivMod_res_div, a);
1094 set_Tuple_pred(n, pn_DivMod_res_mod, b);
1095 assert(get_nodes_Block(n));
1101 static ir_node *transform_node_Cond(ir_node *n)
1103 /* Replace the Cond by a Jmp if it branches on a constant
1106 ir_node *a = get_Cond_selector(n);
1107 tarval *ta = value_of(a);
1109 if ((ta != tarval_bad) &&
1110 (get_irn_mode(a) == mode_b) &&
1111 (get_opt_unreachable_code())) {
1112 /* It's a boolean Cond, branching on a boolean constant.
1113 Replace it by a tuple (Bad, Jmp) or (Jmp, Bad) */
1114 jmp = new_r_Jmp(current_ir_graph, get_nodes_Block(n));
1115 turn_into_tuple(n, 2);
1116 if (ta == tarval_b_true) {
1117 set_Tuple_pred(n, pn_Cond_false, new_Bad());
1118 set_Tuple_pred(n, pn_Cond_true, jmp);
1120 set_Tuple_pred(n, pn_Cond_false, jmp);
1121 set_Tuple_pred(n, pn_Cond_true, new_Bad());
1123 /* We might generate an endless loop, so keep it alive. */
1124 add_End_keepalive(get_irg_end(current_ir_graph), get_nodes_Block(n));
1125 } else if ((ta != tarval_bad) &&
1126 (get_irn_mode(a) == mode_Iu) &&
1127 (get_Cond_kind(n) == dense) &&
1128 (get_opt_unreachable_code())) {
1129 /* I don't want to allow Tuples smaller than the biggest Proj.
1130 Also this tuple might get really big...
1131 I generate the Jmp here, and remember it in link. Link is used
1132 when optimizing Proj. */
1133 set_irn_link(n, new_r_Jmp(current_ir_graph, get_nodes_Block(n)));
1134 /* We might generate an endless loop, so keep it alive. */
1135 add_End_keepalive(get_irg_end(current_ir_graph), get_nodes_Block(n));
1136 } else if ((get_irn_op(a) == op_Eor)
1137 && (get_irn_mode(a) == mode_b)
1138 && (tarval_classify(computed_value(get_Eor_right(a))) == TV_CLASSIFY_ONE)) {
1139 /* The Eor is a negate. Generate a new Cond without the negate,
1140 simulate the negate by exchanging the results. */
1141 set_irn_link(n, new_r_Cond(current_ir_graph, get_nodes_Block(n),
1143 } else if ((get_irn_op(a) == op_Not)
1144 && (get_irn_mode(a) == mode_b)) {
1145 /* A Not before the Cond. Generate a new Cond without the Not,
1146 simulate the Not by exchanging the results. */
1147 set_irn_link(n, new_r_Cond(current_ir_graph, get_nodes_Block(n),
1153 static ir_node *transform_node_Eor(ir_node *n)
1155 ir_node *a = get_Eor_left(n);
1156 ir_node *b = get_Eor_right(n);
1158 if ((get_irn_mode(n) == mode_b)
1159 && (get_irn_op(a) == op_Proj)
1160 && (get_irn_mode(a) == mode_b)
1161 && (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE)
1162 && (get_irn_op(get_Proj_pred(a)) == op_Cmp))
1163 /* The Eor negates a Cmp. The Cmp has the negated result anyways! */
1164 n = new_r_Proj(current_ir_graph, get_nodes_Block(n), get_Proj_pred(a),
1165 mode_b, get_negated_pnc(get_Proj_proj(a)));
1166 else if ((get_irn_mode(n) == mode_b)
1167 && (tarval_classify (computed_value (b)) == TV_CLASSIFY_ONE))
1168 /* The Eor is a Not. Replace it by a Not. */
1169 /* ????!!!Extend to bitfield 1111111. */
1170 n = new_r_Not(current_ir_graph, get_nodes_Block(n), a, mode_b);
1175 static ir_node *transform_node_Not(ir_node *n)
1177 ir_node *a = get_Not_op(n);
1179 if ( (get_irn_mode(n) == mode_b)
1180 && (get_irn_op(a) == op_Proj)
1181 && (get_irn_mode(a) == mode_b)
1182 && (get_irn_op(get_Proj_pred(a)) == op_Cmp))
1183 /* We negate a Cmp. The Cmp has the negated result anyways! */
1184 n = new_r_Proj(current_ir_graph, get_nodes_Block(n), get_Proj_pred(a),
1185 mode_b, get_negated_pnc(get_Proj_proj(a)));
1192 * Tries several [inplace] [optimizing] transformations and returns an
1193 * equivalent node. The difference to equivalent_node() is that these
1194 * transformations _do_ generate new nodes, and thus the old node must
1195 * not be freed even if the equivalent node isn't the old one.
1197 static ir_node *transform_node(ir_node *n)
1199 if (n->op->transform_node)
1200 n = n->op->transform_node(n);
1205 * set the default transform node operation
1207 static ir_op *firm_set_default_transform_node(ir_op *op)
1211 op->transform_node = transform_node_##a; \
1222 op->transform_node = NULL;
1230 /* **************** Common Subexpression Elimination **************** */
1232 /** The size of the hash table used, should estimate the number of nodes
1234 #define N_IR_NODES 512
1236 static int node_cmp_attr_Const(ir_node *a, ir_node *b)
1238 return (get_Const_tarval(a) != get_Const_tarval(b))
1239 || (get_Const_type(a) != get_Const_type(b));
1242 static int node_cmp_attr_Proj(ir_node *a, ir_node *b)
1244 return get_irn_proj_attr (a) != get_irn_proj_attr (b);
1247 static int node_cmp_attr_Filter(ir_node *a, ir_node *b)
1249 return get_Filter_proj(a) != get_Filter_proj(b);
1252 static int node_cmp_attr_Alloc(ir_node *a, ir_node *b)
1254 return (get_irn_alloc_attr(a).where != get_irn_alloc_attr(b).where)
1255 || (get_irn_alloc_attr(a).type != get_irn_alloc_attr(b).type);
1258 static int node_cmp_attr_Free(ir_node *a, ir_node *b)
1260 return (get_irn_free_attr(a) != get_irn_free_attr(b));
1263 static int node_cmp_attr_SymConst(ir_node *a, ir_node *b)
1265 return (get_irn_symconst_attr(a).num != get_irn_symconst_attr(b).num)
1266 || (get_irn_symconst_attr(a).tori.typ != get_irn_symconst_attr(b).tori.typ);
1269 static int node_cmp_attr_Call(ir_node *a, ir_node *b)
1271 return (get_irn_call_attr(a) != get_irn_call_attr(b));
1274 static int node_cmp_attr_FuncCall(ir_node *a, ir_node *b)
1276 return (get_irn_funccall_attr(a) != get_irn_funccall_attr(b));
1279 static int node_cmp_attr_Sel(ir_node *a, ir_node *b)
1281 return (get_irn_sel_attr(a).ent->kind != get_irn_sel_attr(b).ent->kind)
1282 || (get_irn_sel_attr(a).ent->name != get_irn_sel_attr(b).ent->name)
1283 || (get_irn_sel_attr(a).ent->owner != get_irn_sel_attr(b).ent->owner)
1284 || (get_irn_sel_attr(a).ent->ld_name != get_irn_sel_attr(b).ent->ld_name)
1285 || (get_irn_sel_attr(a).ent->type != get_irn_sel_attr(b).ent->type);
1288 static int node_cmp_attr_Phi(ir_node *a, ir_node *b)
1290 return get_irn_phi_attr (a) != get_irn_phi_attr (b);
1293 static int node_cmp_attr_Cast(ir_node *a, ir_node *b)
1295 return get_Cast_type(a) != get_Cast_type(b);
1299 * set the default node attribute compare operation
1301 static ir_op *firm_set_default_node_cmp_attr(ir_op *op)
1305 op->node_cmp_attr = node_cmp_attr_##a; \
1321 op->node_cmp_attr = NULL;
1329 * Compare function for two nodes in the hash table. Gets two
1330 * nodes as parameters. Returns 0 if the nodes are a cse.
1333 vt_cmp (const void *elt, const void *key)
1341 if (a == b) return 0;
1343 if ((get_irn_op(a) != get_irn_op(b)) ||
1344 (get_irn_mode(a) != get_irn_mode(b))) return 1;
1346 /* compare if a's in and b's in are equal */
1347 irn_arity_a = get_irn_arity (a);
1348 if (irn_arity_a != get_irn_arity(b))
1351 /* for block-local cse and pinned nodes: */
1352 if (!get_opt_global_cse() || (get_op_pinned(get_irn_op(a)) == pinned)) {
1353 if (get_irn_n(a, -1) != get_irn_n(b, -1))
1357 /* compare a->in[0..ins] with b->in[0..ins] */
1358 for (i = 0; i < irn_arity_a; i++)
1359 if (get_irn_n(a, i) != get_irn_n(b, i))
1363 * here, we already now that the nodes are identical except their
1366 if (a->op->node_cmp_attr)
1367 return a->op->node_cmp_attr(a, b);
1373 * Calculate a hash value of a node.
1376 ir_node_hash (ir_node *node)
1381 /* hash table value = 9*(9*(9*(9*(9*arity+in[0])+in[1])+ ...)+mode)+code */
1382 h = irn_arity = get_irn_arity(node);
1384 /* consider all in nodes... except the block. */
1385 for (i = 0; i < irn_arity; i++) {
1386 h = 9*h + (unsigned long)get_irn_n(node, i);
1390 h = 9*h + (unsigned long) get_irn_mode (node);
1392 h = 9*h + (unsigned long) get_irn_op (node);
1398 new_identities (void)
1400 return new_pset (vt_cmp, N_IR_NODES);
1404 del_identities (pset *value_table)
1406 del_pset (value_table);
1410 * Return the canonical node computing the same value as n.
1411 * Looks up the node in a hash table.
1413 static INLINE ir_node *
1414 identify (pset *value_table, ir_node *n)
1418 if (!value_table) return n;
1420 /* TODO: use a generic commutative attribute */
1421 if (get_opt_reassociation()) {
1422 if (is_op_commutative(get_irn_op(n))) {
1423 /* for commutative operators perform a OP b == b OP a */
1424 if (get_binop_left(n) > get_binop_right(n)) {
1425 ir_node *h = get_binop_left(n);
1426 set_binop_left(n, get_binop_right(n));
1427 set_binop_right(n, h);
1432 o = pset_find (value_table, n, ir_node_hash (n));
1439 * During construction we set the pinned flag in the graph right when the
1440 * optimizatin is performed. The flag turning on procedure global cse could
1441 * be changed between two allocations. This way we are safe.
1443 static INLINE ir_node *
1444 identify_cons (pset *value_table, ir_node *n) {
1446 n = identify(value_table, n);
1447 if (get_irn_n(old, -1) != get_irn_n(n, -1))
1448 set_irg_pinned(current_ir_graph, floats);
1453 * Return the canonical node computing the same value as n.
1454 * Looks up the node in a hash table, enters it in the table
1455 * if it isn't there yet.
1458 identify_remember (pset *value_table, ir_node *node)
1462 if (!value_table) return node;
1464 /* lookup or insert in hash table with given hash key. */
1465 o = pset_insert (value_table, node, ir_node_hash (node));
1467 if (o == node) return node;
1473 add_identities (pset *value_table, ir_node *node) {
1474 identify_remember (value_table, node);
1478 * garbage in, garbage out. If a node has a dead input, i.e., the
1479 * Bad node is input to the node, return the Bad node.
1481 static INLINE ir_node *
1482 gigo (ir_node *node)
1485 ir_op* op = get_irn_op(node);
1487 /* remove garbage blocks by looking at control flow that leaves the block
1488 and replacing the control flow by Bad. */
1489 if (get_irn_mode(node) == mode_X) {
1490 ir_node *block = get_nodes_block(node);
1491 if (op == op_End) return node; /* Don't optimize End, may have Bads. */
1492 if (get_irn_op(block) == op_Block && get_Block_matured(block)) {
1493 irn_arity = get_irn_arity(block);
1494 for (i = 0; i < irn_arity; i++) {
1495 if (!is_Bad(get_irn_n(block, i))) break;
1497 if (i == irn_arity) return new_Bad();
1501 /* Blocks, Phis and Tuples may have dead inputs, e.g., if one of the
1502 blocks predecessors is dead. */
1503 if ( op != op_Block && op != op_Phi && op != op_Tuple) {
1504 irn_arity = get_irn_arity(node);
1505 for (i = -1; i < irn_arity; i++) {
1506 if (is_Bad(get_irn_n(node, i))) {
1512 /* With this code we violate the agreement that local_optimize
1513 only leaves Bads in Block, Phi and Tuple nodes. */
1514 /* If Block has only Bads as predecessors it's garbage. */
1515 /* If Phi has only Bads as predecessors it's garbage. */
1516 if ((op == op_Block && get_Block_matured(node)) || op == op_Phi) {
1517 irn_arity = get_irn_arity(node);
1518 for (i = 0; i < irn_arity; i++) {
1519 if (!is_Bad(get_irn_n(node, i))) break;
1521 if (i == irn_arity) node = new_Bad();
1529 * These optimizations deallocate nodes from the obstack.
1530 * It can only be called if it is guaranteed that no other nodes
1531 * reference this one, i.e., right after construction of a node.
1534 optimize_node (ir_node *n)
1538 opcode iro = get_irn_opcode(n);
1540 /* Allways optimize Phi nodes: part of the construction. */
1541 if ((!get_opt_optimize()) && (iro != iro_Phi)) return n;
1543 /* constant expression evaluation / constant folding */
1544 if (get_opt_constant_folding()) {
1545 /* constants can not be evaluated */
1546 if (get_irn_op(n) != op_Const) {
1547 /* try to evaluate */
1548 tv = computed_value (n);
1549 if ((get_irn_mode(n) != mode_T) && (tv != tarval_bad)) {
1551 * we MUST copy the node here temparary, because it's still needed
1552 * for DBG_OPT_ALGSIM0
1556 /* evaluation was succesful -- replace the node. */
1557 obstack_free (current_ir_graph->obst, n);
1558 n = new_Const (get_tarval_mode (tv), tv);
1565 /* remove unnecessary nodes */
1566 if (get_opt_constant_folding() ||
1567 (iro == iro_Phi) || /* always optimize these nodes. */
1569 (iro == iro_Proj) ||
1570 (iro == iro_Block) ) /* Flags tested local. */
1571 n = equivalent_node (n);
1573 optimize_preds(n); /* do node specific optimizations of nodes predecessors. */
1575 /** common subexpression elimination **/
1576 /* Checks whether n is already available. */
1577 /* The block input is used to distinguish different subexpressions. Right
1578 now all nodes are pinned to blocks, i.e., the cse only finds common
1579 subexpressions within a block. */
1581 n = identify_cons (current_ir_graph->value_table, n);
1584 /* We found an existing, better node, so we can deallocate the old node. */
1585 obstack_free (current_ir_graph->obst, oldn);
1590 /* Some more constant expression evaluation that does not allow to
1592 iro = get_irn_opcode(n);
1593 if (get_opt_constant_folding() ||
1594 (iro == iro_Cond) ||
1595 (iro == iro_Proj)) /* Flags tested local. */
1596 n = transform_node (n);
1598 /* Remove nodes with dead (Bad) input.
1599 Run always for transformation induced Bads. */
1602 /* Now we have a legal, useful node. Enter it in hash table for cse */
1603 if (get_opt_cse() && (get_irn_opcode(n) != iro_Block)) {
1604 n = identify_remember (current_ir_graph->value_table, n);
1612 * These optimizations never deallocate nodes. This can cause dead
1613 * nodes lying on the obstack. Remove these by a dead node elimination,
1614 * i.e., a copying garbage collection.
1617 optimize_in_place_2 (ir_node *n)
1621 opcode iro = get_irn_opcode(n);
1623 if (!get_opt_optimize() && (get_irn_op(n) != op_Phi)) return n;
1625 /* if not optimize return n */
1628 /* Here this is possible. Why? */
1633 /* constant expression evaluation / constant folding */
1634 if (get_opt_constant_folding()) {
1635 /* constants can not be evaluated */
1636 if (iro != iro_Const) {
1637 /* try to evaluate */
1638 tv = computed_value (n);
1639 if ((get_irn_mode(n) != mode_T) && (tv != tarval_bad)) {
1640 /* evaluation was succesful -- replace the node. */
1641 n = new_Const (get_tarval_mode (tv), tv);
1648 /* remove unnecessary nodes */
1649 /*if (get_opt_constant_folding()) */
1650 if (get_opt_constant_folding() ||
1651 (iro == iro_Phi) || /* always optimize these nodes. */
1652 (iro == iro_Id) || /* ... */
1653 (iro == iro_Proj) || /* ... */
1654 (iro == iro_Block) ) /* Flags tested local. */
1655 n = equivalent_node (n);
1657 optimize_preds(n); /* do node specific optimizations of nodes predecessors. */
1659 /** common subexpression elimination **/
1660 /* Checks whether n is already available. */
1661 /* The block input is used to distinguish different subexpressions. Right
1662 now all nodes are pinned to blocks, i.e., the cse only finds common
1663 subexpressions within a block. */
1664 if (get_opt_cse()) {
1665 n = identify (current_ir_graph->value_table, n);
1668 /* Some more constant expression evaluation. */
1669 iro = get_irn_opcode(n);
1670 if (get_opt_constant_folding() ||
1671 (iro == iro_Cond) ||
1672 (iro == iro_Proj)) /* Flags tested local. */
1673 n = transform_node (n);
1675 /* Remove nodes with dead (Bad) input.
1676 Run always for transformation induced Bads. */
1679 /* Now we can verify the node, as it has no dead inputs any more. */
1682 /* Now we have a legal, useful node. Enter it in hash table for cse.
1683 Blocks should be unique anyways. (Except the successor of start:
1684 is cse with the start block!) */
1685 if (get_opt_cse() && (get_irn_opcode(n) != iro_Block))
1686 n = identify_remember (current_ir_graph->value_table, n);
1692 * Wrapper for external use, set proper status bits after optimization.
1695 optimize_in_place (ir_node *n)
1697 /* Handle graph state */
1698 assert(get_irg_phase_state(current_ir_graph) != phase_building);
1700 if (get_opt_global_cse())
1701 set_irg_pinned(current_ir_graph, floats);
1702 if (get_irg_outs_state(current_ir_graph) == outs_consistent)
1703 set_irg_outs_inconsistent(current_ir_graph);
1704 /* Maybe we could also test whether optimizing the node can
1705 change the control graph. */
1706 if (get_irg_dom_state(current_ir_graph) == dom_consistent)
1707 set_irg_dom_inconsistent(current_ir_graph);
1708 return optimize_in_place_2 (n);
1712 * set the default ir op operations
1714 ir_op *firm_set_default_operations(ir_op *op)
1716 op = firm_set_default_computed_value(op);
1717 op = firm_set_default_equivalent_node(op);
1718 op = firm_set_default_transform_node(op);
1719 op = firm_set_default_node_cmp_attr(op);