2 * Copyright (C) 1995-2011 University of Karlsruhe. All right reserved.
4 * This file is part of libFirm.
6 * This file may be distributed and/or modified under the terms of the
7 * GNU General Public License version 2 as published by the Free Software
8 * Foundation and appearing in the file LICENSE.GPL included in the
9 * packaging of this file.
11 * Licensees holding valid libFirm Professional Edition licenses may use
12 * this file in accordance with the libFirm Commercial License.
13 * Agreement provided with the Software.
15 * This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE
16 * WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR
22 * @brief Backend ABI implementation.
23 * @author Sebastian Hack, Michael Beck
31 #include "irgraph_t.h"
34 #include "iredges_t.h"
37 #include "irprintf_t.h"
44 #include "raw_bitset.h"
55 #include "bessaconstr.h"
57 #include "betranshlp.h"
59 DEBUG_ONLY(static firm_dbg_module_t *dbg;)
61 typedef struct be_abi_call_arg_t {
62 unsigned is_res : 1; /**< 1: the call argument is a return value. 0: it's a call parameter. */
63 unsigned in_reg : 1; /**< 1: this argument is transmitted 1: in registers, 0: on stack. */
64 unsigned callee : 1; /**< 1: someone called us. 0: We call another function */
67 const arch_register_t *reg;
70 unsigned alignment; /**< stack alignment */
71 unsigned space_before; /**< allocate space before */
72 unsigned space_after; /**< allocate space after */
75 struct be_abi_call_t {
76 be_abi_call_flags_t flags; /**< Flags describing the ABI behavior on calls */
77 int pop; /**< number of bytes the stack frame is shrinked by the callee on return. */
78 const be_abi_callbacks_t *cb;
80 const arch_register_class_t *cls_addr; /**< register class of the call address */
84 * The ABI information for the current graph.
87 be_abi_call_t *call; /**< The ABI call information. */
89 ir_node *init_sp; /**< The node representing the stack pointer
90 at the start of the function. */
92 pmap *regs; /**< A map of all callee-save and ignore regs to
93 their Projs to the RegParams node. */
94 pmap *keep_map; /**< mapping blocks to keep nodes. */
96 ir_node **calls; /**< flexible array containing all be_Call nodes */
99 static ir_heights_t *ir_heights;
101 static ir_node *be_abi_reg_map_get(pmap *map, const arch_register_t *reg)
103 return pmap_get(ir_node, map, reg);
106 static void be_abi_reg_map_set(pmap *map, const arch_register_t* reg,
109 pmap_insert(map, reg, node);
113 * Check if the given register is callee save, ie. will be saved by the callee.
115 static bool arch_register_is_callee_save(
116 const arch_env_t *arch_env,
117 const arch_register_t *reg)
119 if (arch_env->impl->register_saved_by)
120 return arch_env->impl->register_saved_by(reg, /*callee=*/1);
125 * Check if the given register is caller save, ie. must be saved by the caller.
127 static bool arch_register_is_caller_save(
128 const arch_env_t *arch_env,
129 const arch_register_t *reg)
131 if (arch_env->impl->register_saved_by)
132 return arch_env->impl->register_saved_by(reg, /*callee=*/0);
139 _ ____ ___ ____ _ _ _ _
140 / \ | __ )_ _| / ___|__ _| | | |__ __ _ ___| | _____
141 / _ \ | _ \| | | | / _` | | | '_ \ / _` |/ __| |/ / __|
142 / ___ \| |_) | | | |__| (_| | | | |_) | (_| | (__| <\__ \
143 /_/ \_\____/___| \____\__,_|_|_|_.__/ \__,_|\___|_|\_\___/
145 These callbacks are used by the backend to set the parameters
146 for a specific call type.
150 * Set compare function: compares two ABI call object arguments.
152 static int cmp_call_arg(const void *a, const void *b, size_t n)
154 const be_abi_call_arg_t *p = (const be_abi_call_arg_t*)a;
155 const be_abi_call_arg_t *q = (const be_abi_call_arg_t*)b;
157 return !(p->is_res == q->is_res && p->pos == q->pos && p->callee == q->callee);
161 * Get an ABI call object argument.
163 * @param call the abi call
164 * @param is_res true for call results, false for call arguments
165 * @param pos position of the argument
166 * @param callee context type - if we are callee or caller
168 static be_abi_call_arg_t *get_call_arg(be_abi_call_t *call, int is_res, int pos, int callee)
170 be_abi_call_arg_t arg;
173 memset(&arg, 0, sizeof(arg));
178 hash = is_res * 128 + pos;
180 return set_find(be_abi_call_arg_t, call->params, &arg, sizeof(arg), hash);
184 * Set an ABI call object argument.
186 static void remember_call_arg(be_abi_call_arg_t *arg, be_abi_call_t *call, be_abi_context_t context)
188 unsigned hash = arg->is_res * 128 + arg->pos;
189 if (context & ABI_CONTEXT_CALLEE) {
191 (void)set_insert(be_abi_call_arg_t, call->params, arg, sizeof(*arg), hash);
193 if (context & ABI_CONTEXT_CALLER) {
195 (void)set_insert(be_abi_call_arg_t, call->params, arg, sizeof(*arg), hash);
199 /* Set the flags for a call. */
200 void be_abi_call_set_flags(be_abi_call_t *call, be_abi_call_flags_t flags, const be_abi_callbacks_t *cb)
206 /* Sets the number of bytes the stackframe is shrinked by the callee on return */
207 void be_abi_call_set_pop(be_abi_call_t *call, int pop)
213 void be_abi_call_param_stack(be_abi_call_t *call, int arg_pos,
214 ir_mode *load_mode, unsigned alignment,
215 unsigned space_before, unsigned space_after,
216 be_abi_context_t context)
218 be_abi_call_arg_t arg;
219 memset(&arg, 0, sizeof(arg));
220 assert(alignment > 0 && "Alignment must be greater than 0");
221 arg.load_mode = load_mode;
222 arg.alignment = alignment;
223 arg.space_before = space_before;
224 arg.space_after = space_after;
228 remember_call_arg(&arg, call, context);
231 void be_abi_call_param_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
233 be_abi_call_arg_t arg;
234 memset(&arg, 0, sizeof(arg));
241 remember_call_arg(&arg, call, context);
244 void be_abi_call_res_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
246 be_abi_call_arg_t arg;
247 memset(&arg, 0, sizeof(arg));
254 remember_call_arg(&arg, call, context);
257 /* Get the flags of a ABI call object. */
258 be_abi_call_flags_t be_abi_call_get_flags(const be_abi_call_t *call)
264 * Constructor for a new ABI call object.
266 * @param cls_addr register class of the call address
268 * @return the new ABI call object
270 static be_abi_call_t *be_abi_call_new(const arch_register_class_t *cls_addr)
272 be_abi_call_t *call = XMALLOCZ(be_abi_call_t);
274 call->params = new_set(cmp_call_arg, 16);
276 call->cls_addr = cls_addr;
277 call->flags.try_omit_fp = be_options.omit_fp;
283 * Destructor for an ABI call object.
285 static void be_abi_call_free(be_abi_call_t *call)
287 del_set(call->params);
292 * Initializes the frame layout from parts
294 * @param frame the stack layout that will be initialized
295 * @param args the stack argument layout type
296 * @param between the between layout type
297 * @param locals the method frame type
299 * @return the initialized stack layout
301 static be_stack_layout_t *stack_frame_init(be_stack_layout_t *frame, ir_type *args,
302 ir_type *between, ir_type *locals)
304 frame->arg_type = args;
305 frame->between_type = between;
306 frame->frame_type = locals;
307 frame->initial_offset = 0;
308 frame->initial_bias = 0;
309 frame->order[1] = between;
311 /* typical decreasing stack: locals have the
312 * lowest addresses, arguments the highest */
313 frame->order[0] = locals;
314 frame->order[2] = args;
325 Adjustment of the calls inside a graph.
330 * Transform a call node into a be_Call node.
332 * @param env The ABI environment for the current irg.
333 * @param irn The call node.
334 * @param curr_sp The stack pointer node to use.
335 * @return The stack pointer after the call.
337 static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp)
339 ir_graph *irg = get_irn_irg(irn);
340 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
341 ir_type *call_tp = get_Call_type(irn);
342 ir_node *call_ptr = get_Call_ptr(irn);
343 size_t n_params = get_method_n_params(call_tp);
344 ir_node *curr_mem = get_Call_mem(irn);
345 ir_node *bl = get_nodes_block(irn);
347 const arch_register_t *sp = arch_env->sp;
348 be_abi_call_t *call = be_abi_call_new(sp->reg_class);
349 ir_mode *mach_mode = sp->reg_class->mode;
350 int n_res = get_method_n_ress(call_tp);
352 ir_node *res_proj = NULL;
353 int n_reg_params = 0;
354 int n_stack_params = 0;
357 const arch_register_t **states = NEW_ARR_F(const arch_register_t*, 0);
358 const arch_register_t **destroyed_regs = NEW_ARR_F(const arch_register_t*, 0);
362 int n_reg_results = 0;
364 int *stack_param_idx;
366 int throws_exception;
371 /* Let the isa fill out the abi description for that call node. */
372 arch_env_get_call_abi(arch_env, call_tp, call);
374 /* Insert code to put the stack arguments on the stack. */
375 assert((size_t)get_Call_n_params(irn) == n_params);
376 stack_param_idx = ALLOCAN(int, n_params);
377 for (p = 0; p < n_params; ++p) {
378 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
381 int arg_size = get_type_size_bytes(get_method_param_type(call_tp, p));
383 stack_size += round_up2(arg->space_before, arg->alignment);
384 stack_size += round_up2(arg_size, arg->alignment);
385 stack_size += round_up2(arg->space_after, arg->alignment);
387 stack_param_idx[n_stack_params++] = p;
391 /* Collect all arguments which are passed in registers. */
392 reg_param_idxs = ALLOCAN(int, n_params);
393 for (p = 0; p < n_params; ++p) {
394 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
395 if (arg && arg->in_reg) {
396 reg_param_idxs[n_reg_params++] = p;
401 * If the stack is decreasing and we do not want to store sequentially,
402 * or someone else allocated the call frame
403 * we allocate as much space on the stack all parameters need, by
404 * moving the stack pointer along the stack's direction.
406 * Note: we also have to do this for stack_size == 0, because we may have
407 * to adjust stack alignment for the call.
409 curr_sp = be_new_IncSP(sp, bl, curr_sp, stack_size, 1);
411 dbgi = get_irn_dbg_info(irn);
412 /* If there are some parameters which shall be passed on the stack. */
413 if (n_stack_params > 0) {
415 ir_node **in = ALLOCAN(ir_node*, n_stack_params+1);
418 curr_mem = get_Call_mem(irn);
419 in[n_in++] = curr_mem;
421 for (i = 0; i < n_stack_params; ++i) {
422 int p = stack_param_idx[i];
423 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
424 ir_node *param = get_Call_param(irn, p);
425 ir_node *addr = curr_sp;
427 ir_type *param_type = get_method_param_type(call_tp, p);
428 int param_size = get_type_size_bytes(param_type) + arg->space_after;
431 * If we wanted to build the arguments sequentially,
432 * the stack pointer for the next must be incremented,
433 * and the memory value propagated.
435 curr_ofs += arg->space_before;
436 curr_ofs = round_up2(curr_ofs, arg->alignment);
438 /* Make the expression to compute the argument's offset. */
440 ir_mode *constmode = mach_mode;
441 if (mode_is_reference(mach_mode)) {
444 addr = new_r_Const_long(irg, constmode, curr_ofs);
445 addr = new_r_Add(bl, curr_sp, addr, mach_mode);
448 /* Insert a store for primitive arguments. */
449 if (is_atomic_type(param_type)) {
450 ir_node *nomem = get_irg_no_mem(irg);
451 ir_node *mem_input = nomem;
452 ir_node *store = new_rd_Store(dbgi, bl, mem_input, addr, param, cons_none);
453 mem = new_r_Proj(store, mode_M, pn_Store_M);
455 /* Make a mem copy for compound arguments. */
458 assert(mode_is_reference(get_irn_mode(param)));
459 copy = new_rd_CopyB(dbgi, bl, curr_mem, addr, param, param_type);
460 mem = new_r_Proj(copy, mode_M, pn_CopyB_M);
463 curr_ofs += param_size;
468 /* We need the sync only, if we didn't build the stores sequentially. */
469 if (n_stack_params >= 1) {
470 curr_mem = new_r_Sync(bl, n_in, in);
472 curr_mem = get_Call_mem(irn);
476 /* Put caller save into the destroyed set and state registers in the states
478 for (i = 0, n = arch_env->n_register_classes; i < n; ++i) {
480 const arch_register_class_t *cls = &arch_env->register_classes[i];
481 for (j = 0; j < cls->n_regs; ++j) {
482 const arch_register_t *reg = arch_register_for_index(cls, j);
484 /* even if destroyed all is specified, neither SP nor FP are
485 * destroyed (else bad things will happen) */
486 if (reg == arch_env->sp || reg == arch_env->bp)
489 if (reg->type & arch_register_type_state) {
490 ARR_APP1(const arch_register_t*, destroyed_regs, reg);
491 ARR_APP1(const arch_register_t*, states, reg);
492 /* we're already in the destroyed set so no need for further
496 if (arch_register_is_caller_save(arch_env, reg)) {
497 if (!(reg->type & arch_register_type_ignore)) {
498 ARR_APP1(const arch_register_t*, destroyed_regs, reg);
504 /* search the largest result proj number */
505 res_projs = ALLOCANZ(ir_node*, n_res);
507 foreach_out_edge(irn, edge) {
508 ir_node *irn = get_edge_src_irn(edge);
510 if (!is_Proj(irn) || get_Proj_proj(irn) != pn_Call_T_result)
513 foreach_out_edge(irn, res_edge) {
515 ir_node *res = get_edge_src_irn(res_edge);
517 assert(is_Proj(res));
519 proj = get_Proj_proj(res);
520 assert(proj < n_res);
521 assert(res_projs[proj] == NULL);
522 res_projs[proj] = res;
528 /** TODO: this is not correct for cases where return values are passed
529 * on the stack, but no known ABI does this currently...
531 n_reg_results = n_res;
534 in = ALLOCAN(ir_node*, n_reg_params + ARR_LEN(states));
536 /* make the back end call node and set its register requirements. */
537 for (i = 0; i < n_reg_params; ++i) {
538 in[n_ins++] = get_Call_param(irn, reg_param_idxs[i]);
541 /* add state registers ins */
542 for (s = 0; s < ARR_LEN(states); ++s) {
543 const arch_register_t *reg = states[s];
544 const arch_register_class_t *cls = reg->reg_class;
545 ir_node *regnode = new_r_Unknown(irg, cls->mode);
546 in[n_ins++] = regnode;
548 assert(n_ins == (int) (n_reg_params + ARR_LEN(states)));
550 /* ins collected, build the call */
551 throws_exception = ir_throws_exception(irn);
552 if (env->call->flags.call_has_imm && is_SymConst(call_ptr)) {
554 low_call = be_new_Call(dbgi, irg, bl, curr_mem, sp->single_req, curr_sp,
555 sp->single_req, curr_sp,
556 n_reg_results + pn_be_Call_first_res + ARR_LEN(destroyed_regs),
557 n_ins, in, get_Call_type(irn));
558 be_Call_set_entity(low_call, get_SymConst_entity(call_ptr));
561 low_call = be_new_Call(dbgi, irg, bl, curr_mem, sp->single_req, curr_sp,
562 call->cls_addr->class_req, call_ptr,
563 n_reg_results + pn_be_Call_first_res + ARR_LEN(destroyed_regs),
564 n_ins, in, get_Call_type(irn));
566 ir_set_throws_exception(low_call, throws_exception);
567 be_Call_set_pop(low_call, call->pop);
569 /* put the call into the list of all calls for later processing */
570 ARR_APP1(ir_node *, env->calls, low_call);
572 /* create new stack pointer */
573 curr_sp = new_r_Proj(low_call, get_irn_mode(curr_sp), pn_be_Call_sp);
574 be_set_constr_single_reg_out(low_call, pn_be_Call_sp, sp,
575 arch_register_req_type_ignore | arch_register_req_type_produces_sp);
576 arch_set_irn_register(curr_sp, sp);
578 /* now handle results */
579 for (i = 0; i < n_res; ++i) {
580 ir_node *proj = res_projs[i];
581 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 0);
583 /* returns values on stack not supported yet */
587 shift the proj number to the right, since we will drop the
588 unspeakable Proj_T from the Call. Therefore, all real argument
589 Proj numbers must be increased by pn_be_Call_first_res
591 long pn = i + pn_be_Call_first_res;
594 ir_type *res_type = get_method_res_type(call_tp, i);
595 ir_mode *mode = get_type_mode(res_type);
596 proj = new_r_Proj(low_call, mode, pn);
599 set_Proj_pred(proj, low_call);
600 set_Proj_proj(proj, pn);
604 /* remove register from destroyed regs */
606 size_t n = ARR_LEN(destroyed_regs);
607 for (j = 0; j < n; ++j) {
608 if (destroyed_regs[j] == arg->reg) {
609 destroyed_regs[j] = destroyed_regs[n-1];
610 ARR_SHRINKLEN(destroyed_regs,n-1);
617 DBG((dbg, LEVEL_3, "\tcreated backend call %+F\n", low_call));
619 /* Set the register classes and constraints of the Call parameters. */
620 for (i = 0; i < n_reg_params; ++i) {
621 int index = reg_param_idxs[i];
622 be_abi_call_arg_t *arg = get_call_arg(call, 0, index, 0);
623 assert(arg->reg != NULL);
625 be_set_constr_single_reg_in(low_call, n_be_Call_first_arg + i,
626 arg->reg, arch_register_req_type_none);
629 /* Set the register constraints of the results. */
630 for (i = 0; i < n_res; ++i) {
631 ir_node *proj = res_projs[i];
632 const be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 0);
633 int pn = get_Proj_proj(proj);
636 be_set_constr_single_reg_out(low_call, pn, arg->reg,
637 arch_register_req_type_none);
638 arch_set_irn_register(proj, arg->reg);
640 exchange(irn, low_call);
642 /* kill the ProjT node */
643 if (res_proj != NULL) {
647 /* Make additional projs for the caller save registers
648 and the Keep node which keeps them alive. */
654 int curr_res_proj = pn_be_Call_first_res + n_reg_results;
657 n_ins = ARR_LEN(destroyed_regs) + n_reg_results + 1;
658 in = ALLOCAN(ir_node *, n_ins);
660 /* also keep the stack pointer */
661 set_irn_link(curr_sp, (void*) sp);
664 for (d = 0; d < ARR_LEN(destroyed_regs); ++d) {
665 const arch_register_t *reg = destroyed_regs[d];
666 ir_node *proj = new_r_Proj(low_call, reg->reg_class->mode, curr_res_proj);
668 /* memorize the register in the link field. we need afterwards to set the register class of the keep correctly. */
669 be_set_constr_single_reg_out(low_call, curr_res_proj, reg,
670 arch_register_req_type_none);
671 arch_set_irn_register(proj, reg);
673 set_irn_link(proj, (void*) reg);
678 for (i = 0; i < n_reg_results; ++i) {
679 ir_node *proj = res_projs[i];
680 const arch_register_t *reg = arch_get_irn_register(proj);
681 set_irn_link(proj, (void*) reg);
686 /* create the Keep for the caller save registers */
687 keep = be_new_Keep(bl, n, in);
688 for (i = 0; i < n; ++i) {
689 const arch_register_t *reg = (const arch_register_t*)get_irn_link(in[i]);
690 be_node_set_reg_class_in(keep, i, reg->reg_class);
694 /* Clean up the stack. */
695 assert(stack_size >= call->pop);
696 stack_size -= call->pop;
698 if (stack_size > 0) {
699 ir_node *mem_proj = NULL;
701 foreach_out_edge(low_call, edge) {
702 ir_node *irn = get_edge_src_irn(edge);
703 if (is_Proj(irn) && get_Proj_proj(irn) == pn_Call_M) {
710 mem_proj = new_r_Proj(low_call, mode_M, pn_be_Call_M);
711 keep_alive(mem_proj);
714 /* Clean up the stack frame or revert alignment fixes if we allocated it */
715 curr_sp = be_new_IncSP(sp, bl, curr_sp, -stack_size, 0);
717 be_abi_call_free(call);
720 DEL_ARR_F(destroyed_regs);
726 * Adjust the size of a node representing a stack alloc or free for the minimum stack alignment.
728 * @param alignment the minimum stack alignment
729 * @param size the node containing the non-aligned size
730 * @param block the block where new nodes are allocated on
731 * @param dbg debug info for new nodes
733 * @return a node representing the aligned size
735 static ir_node *adjust_alloc_size(unsigned stack_alignment, ir_node *size,
736 ir_node *block, dbg_info *dbg)
738 if (stack_alignment > 1) {
744 assert(is_po2(stack_alignment));
746 mode = get_irn_mode(size);
747 tv = new_tarval_from_long(stack_alignment-1, mode);
748 irg = get_Block_irg(block);
749 mask = new_r_Const(irg, tv);
750 size = new_rd_Add(dbg, block, size, mask, mode);
752 tv = new_tarval_from_long(-(long)stack_alignment, mode);
753 mask = new_r_Const(irg, tv);
754 size = new_rd_And(dbg, block, size, mask, mode);
760 * The alloca is transformed into a back end alloca node and connected to the stack nodes.
762 static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp)
764 ir_node *block = get_nodes_block(alloc);
765 ir_graph *irg = get_Block_irg(block);
766 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
767 ir_node *alloc_mem = NULL;
768 ir_node *alloc_res = NULL;
769 ir_type *type = get_Alloc_type(alloc);
776 unsigned stack_alignment;
778 /* all non-stack Alloc nodes should already be lowered before the backend */
779 assert(get_Alloc_where(alloc) == stack_alloc);
781 foreach_out_edge(alloc, edge) {
782 ir_node *irn = get_edge_src_irn(edge);
784 assert(is_Proj(irn));
785 switch (get_Proj_proj(irn)) {
797 /* Beware: currently Alloc nodes without a result might happen,
798 only escape analysis kills them and this phase runs only for object
799 oriented source. We kill the Alloc here. */
800 if (alloc_res == NULL && alloc_mem) {
801 exchange(alloc_mem, get_Alloc_mem(alloc));
805 dbg = get_irn_dbg_info(alloc);
806 count = get_Alloc_count(alloc);
808 /* we might need to multiply the count with the element size */
809 if (!is_unknown_type(type) && get_type_size_bytes(type) != 1) {
810 ir_mode *mode = get_irn_mode(count);
811 ir_tarval *tv = new_tarval_from_long(get_type_size_bytes(type),
813 ir_node *cnst = new_rd_Const(dbg, irg, tv);
814 size = new_rd_Mul(dbg, block, count, cnst, mode);
819 /* The stack pointer will be modified in an unknown manner.
820 We cannot omit it. */
821 env->call->flags.try_omit_fp = 0;
823 stack_alignment = 1 << arch_env->stack_alignment;
824 size = adjust_alloc_size(stack_alignment, size, block, dbg);
825 new_alloc = be_new_AddSP(arch_env->sp, block, curr_sp, size);
826 set_irn_dbg_info(new_alloc, dbg);
828 if (alloc_mem != NULL) {
832 addsp_mem = new_r_Proj(new_alloc, mode_M, pn_be_AddSP_M);
834 /* We need to sync the output mem of the AddSP with the input mem
835 edge into the alloc node. */
836 ins[0] = get_Alloc_mem(alloc);
838 sync = new_r_Sync(block, 2, ins);
840 exchange(alloc_mem, sync);
843 exchange(alloc, new_alloc);
845 /* fix projnum of alloca res */
846 set_Proj_proj(alloc_res, pn_be_AddSP_res);
848 curr_sp = new_r_Proj(new_alloc, get_irn_mode(curr_sp), pn_be_AddSP_sp);
855 * The Free is transformed into a back end free node and connected to the stack nodes.
857 static ir_node *adjust_free(be_abi_irg_t *env, ir_node *free, ir_node *curr_sp)
859 ir_node *block = get_nodes_block(free);
860 ir_graph *irg = get_irn_irg(free);
861 ir_type *type = get_Free_type(free);
862 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
863 ir_mode *sp_mode = arch_env->sp->reg_class->mode;
864 dbg_info *dbg = get_irn_dbg_info(free);
865 ir_node *subsp, *mem, *res, *size, *sync;
867 unsigned stack_alignment;
869 /* all non-stack-alloc Free nodes should already be lowered before the
871 assert(get_Free_where(free) == stack_alloc);
873 /* we might need to multiply the size with the element size */
874 if (!is_unknown_type(type) && get_type_size_bytes(type) != 1) {
875 ir_tarval *tv = new_tarval_from_long(get_type_size_bytes(type), mode_Iu);
876 ir_node *cnst = new_rd_Const(dbg, irg, tv);
877 ir_node *mul = new_rd_Mul(dbg, block, get_Free_count(free),
881 size = get_Free_count(free);
884 stack_alignment = 1 << arch_env->stack_alignment;
885 size = adjust_alloc_size(stack_alignment, size, block, dbg);
887 /* The stack pointer will be modified in an unknown manner.
888 We cannot omit it. */
889 env->call->flags.try_omit_fp = 0;
890 subsp = be_new_SubSP(arch_env->sp, block, curr_sp, size);
891 set_irn_dbg_info(subsp, dbg);
893 mem = new_r_Proj(subsp, mode_M, pn_be_SubSP_M);
894 res = new_r_Proj(subsp, sp_mode, pn_be_SubSP_sp);
896 /* we need to sync the memory */
897 in[0] = get_Free_mem(free);
899 sync = new_r_Sync(block, 2, in);
901 /* and make the AddSP dependent on the former memory */
902 add_irn_dep(subsp, get_Free_mem(free));
905 exchange(free, sync);
912 * Check if a node is somehow data dependent on another one.
913 * both nodes must be in the same basic block.
914 * @param n1 The first node.
915 * @param n2 The second node.
916 * @return 1, if n1 is data dependent (transitively) on n2, 0 if not.
918 static int dependent_on(ir_node *n1, ir_node *n2)
920 assert(get_nodes_block(n1) == get_nodes_block(n2));
922 return heights_reachable_in_block(ir_heights, n1, n2);
925 static int cmp_call_dependency(const void *c1, const void *c2)
927 ir_node *n1 = *(ir_node **) c1;
928 ir_node *n2 = *(ir_node **) c2;
932 Classical qsort() comparison function behavior:
933 0 if both elements are equal
934 1 if second is "smaller" that first
935 -1 if first is "smaller" that second
937 if (dependent_on(n1, n2))
940 if (dependent_on(n2, n1))
943 /* The nodes have no depth order, but we need a total order because qsort()
946 * Additionally, we need to respect transitive dependencies. Consider a
947 * Call a depending on Call b and an independent Call c.
948 * We MUST NOT order c > a and b > c. */
949 h1 = get_irn_height(ir_heights, n1);
950 h2 = get_irn_height(ir_heights, n2);
951 if (h1 < h2) return -1;
952 if (h1 > h2) return 1;
953 /* Same height, so use a random (but stable) order */
954 return get_irn_idx(n1) - get_irn_idx(n2);
958 * Walker: links all Call/Alloc/Free nodes to the Block they are contained.
960 static void link_ops_in_block_walker(ir_node *irn, void *data)
962 be_abi_irg_t *env = (be_abi_irg_t*)data;
963 unsigned code = get_irn_opcode(irn);
965 if (code == iro_Call ||
966 (code == iro_Alloc && get_Alloc_where(irn) == stack_alloc) ||
967 (code == iro_Free && get_Free_where(irn) == stack_alloc)) {
968 ir_node *bl = get_nodes_block(irn);
969 void *save = get_irn_link(bl);
971 set_irn_link(irn, save);
972 set_irn_link(bl, irn);
975 if (code == iro_Builtin && get_Builtin_kind(irn) == ir_bk_return_address) {
976 ir_node *param = get_Builtin_param(irn, 0);
977 ir_tarval *tv = get_Const_tarval(param);
978 unsigned long value = get_tarval_long(tv);
979 /* use ebp, so the climbframe algo works... */
981 env->call->flags.try_omit_fp = 0;
988 * Process all Call/Alloc/Free nodes inside a basic block.
989 * Note that the link field of the block must contain a linked list of all
990 * nodes inside the Block. We first order this list according to data dependency
991 * and that connect the nodes together.
993 static void process_ops_in_block(ir_node *bl, void *data)
995 be_abi_irg_t *env = (be_abi_irg_t*)data;
996 ir_node *curr_sp = env->init_sp;
1003 for (irn = (ir_node*)get_irn_link(bl); irn != NULL;
1004 irn = (ir_node*)get_irn_link(irn)) {
1008 nodes = ALLOCAN(ir_node*, n_nodes);
1009 for (irn = (ir_node*)get_irn_link(bl), n = 0; irn != NULL;
1010 irn = (ir_node*)get_irn_link(irn), ++n) {
1014 /* If there were call nodes in the block. */
1019 /* order the call nodes according to data dependency */
1020 qsort(nodes, n_nodes, sizeof(nodes[0]), cmp_call_dependency);
1022 for (i = n_nodes - 1; i >= 0; --i) {
1023 ir_node *irn = nodes[i];
1025 DBG((dbg, LEVEL_3, "\tprocessing call %+F\n", irn));
1026 switch (get_irn_opcode(irn)) {
1028 curr_sp = adjust_call(env, irn, curr_sp);
1031 if (get_Alloc_where(irn) == stack_alloc)
1032 curr_sp = adjust_alloc(env, irn, curr_sp);
1035 if (get_Free_where(irn) == stack_alloc)
1036 curr_sp = adjust_free(env, irn, curr_sp);
1039 panic("invalid call");
1043 /* Keep the last stack state in the block by tying it to Keep node,
1044 * the proj from calls is already kept */
1045 if (curr_sp != env->init_sp &&
1046 !(is_Proj(curr_sp) && be_is_Call(get_Proj_pred(curr_sp)))) {
1048 keep = be_new_Keep(bl, 1, nodes);
1049 pmap_insert(env->keep_map, bl, keep);
1053 set_irn_link(bl, curr_sp);
1057 * Adjust all call nodes in the graph to the ABI conventions.
1059 static void process_calls(ir_graph *const irg, be_abi_irg_t *const abi)
1061 irg_walk_graph(irg, firm_clear_link, link_ops_in_block_walker, abi);
1063 ir_heights = heights_new(irg);
1064 irg_block_walk_graph(irg, NULL, process_ops_in_block, abi);
1065 heights_free(ir_heights);
1069 * Computes the stack argument layout type.
1070 * Changes a possibly allocated value param type by moving
1071 * entities to the stack layout type.
1073 * @param call the current call ABI
1074 * @param method_type the method type
1076 * @return the stack argument layout type
1078 static ir_type *compute_arg_type(ir_graph *irg, be_abi_call_t *call,
1079 ir_type *method_type)
1081 struct obstack *obst = be_get_be_obst(irg);
1082 ir_type *frame_type = get_irg_frame_type(irg);
1083 size_t n_params = get_method_n_params(method_type);
1084 size_t n_frame_members = get_compound_n_members(frame_type);
1085 ir_entity *va_start_entity = NULL;
1091 ir_entity **map = OALLOCNZ(obst, ir_entity*, n_params);
1092 res = new_type_struct(new_id_from_chars("arg_type", 8));
1094 /* collect existing entities for value_param_types */
1095 for (f = n_frame_members; f > 0; ) {
1096 ir_entity *entity = get_compound_member(frame_type, --f);
1099 set_entity_link(entity, NULL);
1100 if (!is_parameter_entity(entity))
1102 num = get_entity_parameter_number(entity);
1103 if (num == IR_VA_START_PARAMETER_NUMBER) {
1104 /* move entity to new arg_type */
1105 set_entity_owner(entity, res);
1106 va_start_entity = entity;
1109 assert(num < n_params);
1110 if (map[num] != NULL)
1111 panic("multiple entities for parameter %u in %+F found", f, irg);
1113 if (num != n_params && get_call_arg(call, 0, num, 1)->in_reg) {
1114 /* don't move this entity */
1119 /* move entity to new arg_type */
1120 set_entity_owner(entity, res);
1123 for (i = 0; i < n_params; ++i) {
1124 be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 1);
1125 ir_type *param_type = get_method_param_type(method_type, i);
1131 if (entity == NULL) {
1132 /* create a new entity */
1133 entity = new_parameter_entity(res, i, param_type);
1135 ofs += arg->space_before;
1136 ofs = round_up2(ofs, arg->alignment);
1137 set_entity_offset(entity, ofs);
1138 ofs += arg->space_after;
1139 ofs += get_type_size_bytes(param_type);
1140 arg->stack_ent = entity;
1142 if (va_start_entity != NULL) {
1143 set_entity_offset(va_start_entity, ofs);
1145 set_type_size_bytes(res, ofs);
1146 set_type_state(res, layout_fixed);
1152 const arch_register_t *reg;
1156 static int cmp_regs(const void *a, const void *b)
1158 const reg_node_map_t *p = (const reg_node_map_t*)a;
1159 const reg_node_map_t *q = (const reg_node_map_t*)b;
1161 if (p->reg->reg_class == q->reg->reg_class)
1162 return p->reg->index - q->reg->index;
1164 return p->reg->reg_class < q->reg->reg_class ? -1 : +1;
1167 static void reg_map_to_arr(reg_node_map_t *res, pmap *reg_map)
1170 size_t n = pmap_count(reg_map);
1173 foreach_pmap(reg_map, ent) {
1174 res[i].reg = (const arch_register_t*)ent->key;
1175 res[i].irn = (ir_node*)ent->value;
1179 qsort(res, n, sizeof(res[0]), cmp_regs);
1183 * Creates a be_Return for a Return node.
1185 * @param @env the abi environment
1186 * @param irn the Return node or NULL if there was none
1187 * @param bl the block where the be_Retun should be placed
1188 * @param mem the current memory
1189 * @param n_res number of return results
1191 static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl,
1192 ir_node *mem, int n_res)
1194 be_abi_call_t *call = env->call;
1195 ir_graph *irg = get_Block_irg(bl);
1196 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1198 pmap *reg_map = pmap_create();
1199 ir_node *keep = pmap_get(ir_node, env->keep_map, bl);
1206 const arch_register_t **regs;
1210 get the valid stack node in this block.
1211 If we had a call in that block there is a Keep constructed by process_calls()
1212 which points to the last stack modification in that block. we'll use
1213 it then. Else we use the stack from the start block and let
1214 the ssa construction fix the usage.
1216 stack = be_abi_reg_map_get(env->regs, arch_env->sp);
1218 stack = get_irn_n(keep, 0);
1220 remove_End_keepalive(get_irg_end(irg), keep);
1223 /* Insert results for Return into the register map. */
1224 for (i = 0; i < n_res; ++i) {
1225 ir_node *res = get_Return_res(irn, i);
1226 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
1227 assert(arg->in_reg && "return value must be passed in register");
1228 pmap_insert(reg_map, (void *) arg->reg, res);
1231 /* Add uses of the callee save registers. */
1232 foreach_pmap(env->regs, ent) {
1233 const arch_register_t *reg = (const arch_register_t*)ent->key;
1234 if ((reg->type & arch_register_type_ignore) || arch_register_is_callee_save(arch_env, reg))
1235 pmap_insert(reg_map, ent->key, ent->value);
1238 be_abi_reg_map_set(reg_map, arch_env->sp, stack);
1241 Maximum size of the in array for Return nodes is
1242 return args + callee save/ignore registers + memory + stack pointer
1244 in_max = pmap_count(reg_map) + n_res + 2;
1246 in = ALLOCAN(ir_node*, in_max);
1247 regs = ALLOCAN(arch_register_t const*, in_max);
1250 in[1] = be_abi_reg_map_get(reg_map, arch_env->sp);
1252 regs[1] = arch_env->sp;
1255 /* clear SP entry, since it has already been grown. */
1256 pmap_insert(reg_map, (void *) arch_env->sp, NULL);
1257 for (i = 0; i < n_res; ++i) {
1258 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
1260 in[n] = be_abi_reg_map_get(reg_map, arg->reg);
1261 regs[n++] = arg->reg;
1263 /* Clear the map entry to mark the register as processed. */
1264 be_abi_reg_map_set(reg_map, arg->reg, NULL);
1267 /* grow the rest of the stuff. */
1268 foreach_pmap(reg_map, ent) {
1270 in[n] = (ir_node*)ent->value;
1271 regs[n++] = (const arch_register_t*)ent->key;
1275 /* The in array for the new back end return is now ready. */
1277 dbgi = get_irn_dbg_info(irn);
1281 /* we have to pop the shadow parameter in in case of struct returns */
1283 ret = be_new_Return(dbgi, irg, bl, n_res, pop, n, in);
1285 /* Set the register classes of the return's parameter accordingly. */
1286 for (i = 0; i < n; ++i) {
1287 if (regs[i] == NULL)
1290 be_set_constr_single_reg_in(ret, i, regs[i], arch_register_req_type_none);
1293 /* Free the space of the Epilog's in array and the register <-> proj map. */
1294 pmap_destroy(reg_map);
1299 typedef struct lower_frame_sels_env_t {
1300 ir_node *frame; /**< the current frame */
1301 const arch_register_class_t *sp_class; /**< register class of the stack pointer */
1302 } lower_frame_sels_env_t;
1305 * Walker: Replaces Sels of frame type and
1306 * value param type entities by FrameAddress.
1307 * Links all used entities.
1309 static void lower_frame_sels_walker(ir_node *irn, void *data)
1311 lower_frame_sels_env_t *ctx = (lower_frame_sels_env_t*)data;
1314 ir_node *ptr = get_Sel_ptr(irn);
1316 if (ptr == ctx->frame) {
1317 ir_entity *ent = get_Sel_entity(irn);
1318 ir_node *bl = get_nodes_block(irn);
1321 nw = be_new_FrameAddr(ctx->sp_class, bl, ctx->frame, ent);
1328 * The start block has no jump, instead it has an initial exec Proj.
1329 * The backend wants to handle all blocks the same way, so we replace
1330 * the out cfg edge with a real jump.
1332 static void fix_start_block(ir_graph *irg)
1334 ir_node *initial_X = get_irg_initial_exec(irg);
1335 ir_node *start_block = get_irg_start_block(irg);
1336 ir_node *jmp = new_r_Jmp(start_block);
1338 assert(is_Proj(initial_X));
1339 exchange(initial_X, jmp);
1340 set_irg_initial_exec(irg, new_r_Bad(irg, mode_X));
1342 /* merge start block with successor if possible */
1344 foreach_out_edge(jmp, edge) {
1345 ir_node *succ = get_edge_src_irn(edge);
1346 if (!is_Block(succ))
1349 if (get_irn_arity(succ) == 1) {
1350 exchange(succ, start_block);
1358 * Modify the irg itself and the frame type.
1360 static void modify_irg(ir_graph *const irg, be_abi_irg_t *const env)
1362 be_abi_call_t *call = env->call;
1363 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1364 const arch_register_t *sp = arch_env->sp;
1365 ir_type *method_type = get_entity_type(get_irg_entity(irg));
1366 be_irg_t *birg = be_birg_from_irg(irg);
1367 struct obstack *obst = be_get_be_obst(irg);
1368 be_stack_layout_t *stack_layout = be_get_irg_stack_layout(irg);
1371 ir_node *new_mem_proj;
1379 const arch_register_t *fp_reg;
1380 ir_node *frame_pointer;
1384 ir_type *arg_type, *bet_type;
1385 lower_frame_sels_env_t ctx;
1387 DBG((dbg, LEVEL_1, "introducing abi on %+F\n", irg));
1389 old_mem = get_irg_initial_mem(irg);
1391 irp_reserve_resources(irp, IRP_RESOURCE_ENTITY_LINK);
1393 arg_type = compute_arg_type(irg, call, method_type);
1395 /* Convert the Sel nodes in the irg to frame addr nodes: */
1396 ctx.frame = get_irg_frame(irg);
1397 ctx.sp_class = arch_env->sp->reg_class;
1399 ir_type *const frame_tp = get_irg_frame_type(irg);
1400 /* layout the stackframe now */
1401 if (get_type_state(frame_tp) == layout_undefined) {
1402 default_layout_compound_type(frame_tp);
1405 /* align stackframe */
1406 unsigned const alignment = 1U << arch_env->stack_alignment;
1407 unsigned const frame_size = round_up2(get_type_size_bytes(frame_tp), alignment);
1408 set_type_size_bytes(frame_tp, frame_size);
1410 env->regs = pmap_create();
1412 n_params = get_method_n_params(method_type);
1413 args = OALLOCNZ(obst, ir_node*, n_params);
1415 be_add_parameter_entity_stores(irg);
1417 irg_walk_graph(irg, lower_frame_sels_walker, NULL, &ctx);
1419 irp_free_resources(irp, IRP_RESOURCE_ENTITY_LINK);
1421 /* Fill the argument vector */
1422 arg_tuple = get_irg_args(irg);
1423 foreach_out_edge(arg_tuple, edge) {
1424 ir_node *irn = get_edge_src_irn(edge);
1425 if (! is_Anchor(irn)) {
1426 int nr = get_Proj_proj(irn);
1428 DBG((dbg, LEVEL_2, "\treading arg: %d -> %+F\n", nr, irn));
1432 stack_layout->sp_relative = call->flags.try_omit_fp;
1433 bet_type = call->cb->get_between_type(irg);
1434 stack_frame_init(stack_layout, arg_type, bet_type,
1435 get_irg_frame_type(irg));
1437 /* Count the register params and add them to the number of Projs for the RegParams node */
1438 for (i = 0; i < n_params; ++i) {
1439 be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 1);
1440 if (arg->in_reg && args[i]) {
1441 assert(arg->reg != sp && "cannot use stack pointer as parameter register");
1442 assert(i == get_Proj_proj(args[i]));
1444 /* For now, associate the register with the old Proj from Start representing that argument. */
1445 pmap_insert(env->regs, (void *) arg->reg, args[i]);
1446 DBG((dbg, LEVEL_2, "\targ #%d -> reg %s\n", i, arg->reg->name));
1450 /* Collect all callee-save registers */
1451 for (i = 0, n = arch_env->n_register_classes; i < n; ++i) {
1452 const arch_register_class_t *cls = &arch_env->register_classes[i];
1453 for (j = 0; j < cls->n_regs; ++j) {
1454 const arch_register_t *reg = &cls->regs[j];
1455 if ((reg->type & arch_register_type_state) || arch_register_is_callee_save(arch_env, reg)) {
1456 pmap_insert(env->regs, (void *) reg, NULL);
1461 fp_reg = call->flags.try_omit_fp ? arch_env->sp : arch_env->bp;
1462 rbitset_clear(birg->allocatable_regs, fp_reg->global_index);
1464 /* handle start block here (place a jump in the block) */
1465 fix_start_block(irg);
1467 pmap_insert(env->regs, (void *) sp, NULL);
1468 pmap_insert(env->regs, (void *) arch_env->bp, NULL);
1469 start_bl = get_irg_start_block(irg);
1470 ir_node *const start = be_new_Start(NULL, start_bl, pmap_count(env->regs) + 1);
1471 set_irg_start(irg, start);
1474 * make proj nodes for the callee save registers.
1475 * memorize them, since Return nodes get those as inputs.
1477 * Note, that if a register corresponds to an argument, the regs map
1478 * contains the old Proj from start for that argument.
1480 rm = ALLOCAN(reg_node_map_t, pmap_count(env->regs));
1481 reg_map_to_arr(rm, env->regs);
1482 for (i = 0, n = pmap_count(env->regs); i < n; ++i) {
1483 const arch_register_t *reg = rm[i].reg;
1484 ir_mode *mode = reg->reg_class->mode;
1486 arch_register_req_type_t add_type = arch_register_req_type_none;
1490 add_type |= arch_register_req_type_produces_sp;
1491 if (!rbitset_is_set(birg->allocatable_regs, reg->global_index)) {
1492 add_type |= arch_register_req_type_ignore;
1496 proj = new_r_Proj(start, mode, nr + 1);
1497 pmap_insert(env->regs, (void *) reg, proj);
1498 be_set_constr_single_reg_out(start, nr + 1, reg, add_type);
1499 arch_set_irn_register(proj, reg);
1501 DBG((dbg, LEVEL_2, "\tregister save proj #%d -> reg %s\n", nr, reg->name));
1504 /* create a new initial memory proj */
1505 assert(is_Proj(old_mem));
1506 arch_set_irn_register_req_out(start, 0, arch_no_register_req);
1507 new_mem_proj = new_r_Proj(start, mode_M, 0);
1509 set_irg_initial_mem(irg, mem);
1511 env->init_sp = be_abi_reg_map_get(env->regs, sp);
1513 /* set new frame_pointer */
1514 frame_pointer = be_abi_reg_map_get(env->regs, fp_reg);
1515 set_irg_frame(irg, frame_pointer);
1517 /* rewire old mem users to new mem */
1518 exchange(old_mem, mem);
1520 /* keep the mem (for functions with an endless loop = no return) */
1523 set_irg_initial_mem(irg, mem);
1525 /* Now, introduce stack param nodes for all parameters passed on the stack */
1526 for (i = 0; i < n_params; ++i) {
1527 ir_node *arg_proj = args[i];
1528 ir_node *repl = NULL;
1530 if (arg_proj != NULL) {
1531 be_abi_call_arg_t *arg;
1532 ir_type *param_type;
1533 int nr = get_Proj_proj(arg_proj);
1536 nr = MIN(nr, n_params);
1537 arg = get_call_arg(call, 0, nr, 1);
1538 param_type = get_method_param_type(method_type, nr);
1541 repl = pmap_get(ir_node, env->regs, arg->reg);
1543 ir_node *addr = be_new_FrameAddr(sp->reg_class, start_bl, frame_pointer, arg->stack_ent);
1545 /* For atomic parameters which are actually used, we create a Load node. */
1546 if (is_atomic_type(param_type) && get_irn_n_edges(args[i]) > 0) {
1547 ir_mode *mode = get_type_mode(param_type);
1548 ir_mode *load_mode = arg->load_mode;
1549 ir_node *nomem = get_irg_no_mem(irg);
1551 ir_node *load = new_r_Load(start_bl, nomem, addr, load_mode, cons_floats);
1552 repl = new_r_Proj(load, load_mode, pn_Load_res);
1554 if (mode != load_mode) {
1555 repl = new_r_Conv(start_bl, repl, mode);
1558 /* The stack parameter is not primitive (it is a struct or array),
1559 * we thus will create a node representing the parameter's address
1565 assert(repl != NULL);
1567 /* Beware: the mode of the register parameters is always the mode of the register class
1568 which may be wrong. Add Conv's then. */
1569 mode = get_irn_mode(args[i]);
1570 if (mode != get_irn_mode(repl)) {
1571 repl = new_r_Conv(get_nodes_block(repl), repl, mode);
1573 exchange(args[i], repl);
1577 /* the arg proj is not needed anymore now and should be only used by the anchor */
1578 assert(get_irn_n_edges(arg_tuple) == 1);
1579 kill_node(arg_tuple);
1580 set_irg_args(irg, new_r_Bad(irg, mode_T));
1582 /* All Return nodes hang on the End node, so look for them there. */
1583 end = get_irg_end_block(irg);
1584 for (i = 0, n = get_Block_n_cfgpreds(end); i < n; ++i) {
1585 ir_node *irn = get_Block_cfgpred(end, i);
1587 if (is_Return(irn)) {
1588 ir_node *blk = get_nodes_block(irn);
1589 ir_node *mem = get_Return_mem(irn);
1590 ir_node *ret = create_be_return(env, irn, blk, mem, get_Return_n_ress(irn));
1595 /* if we have endless loops here, n might be <= 0. Do NOT create a be_Return then,
1596 the code is dead and will never be executed. */
1599 /** Fix the state inputs of calls that still hang on unknowns */
1600 static void fix_call_state_inputs(ir_graph *const irg, be_abi_irg_t *const env)
1602 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1604 arch_register_t **stateregs = NEW_ARR_F(arch_register_t*, 0);
1606 /* Collect caller save registers */
1607 n = arch_env->n_register_classes;
1608 for (i = 0; i < n; ++i) {
1610 const arch_register_class_t *cls = &arch_env->register_classes[i];
1611 for (j = 0; j < cls->n_regs; ++j) {
1612 const arch_register_t *reg = arch_register_for_index(cls, j);
1613 if (reg->type & arch_register_type_state) {
1614 ARR_APP1(arch_register_t*, stateregs, (arch_register_t *)reg);
1619 n = ARR_LEN(env->calls);
1620 n_states = ARR_LEN(stateregs);
1621 for (i = 0; i < n; ++i) {
1623 ir_node *call = env->calls[i];
1625 arity = get_irn_arity(call);
1627 /* the state reg inputs are the last n inputs of the calls */
1628 for (s = 0; s < n_states; ++s) {
1629 int inp = arity - n_states + s;
1630 const arch_register_t *reg = stateregs[s];
1631 ir_node *regnode = be_abi_reg_map_get(env->regs, reg);
1633 set_irn_n(call, inp, regnode);
1637 DEL_ARR_F(stateregs);
1641 * Create a trampoline entity for the given method.
1643 static ir_entity *create_trampoline(be_main_env_t *be, ir_entity *method)
1645 ir_type *type = get_entity_type(method);
1646 ident *old_id = get_entity_ld_ident(method);
1647 ident *id = id_mangle3("", old_id, "$stub");
1648 ir_type *parent = be->pic_trampolines_type;
1649 ir_entity *ent = new_entity(parent, old_id, type);
1650 set_entity_ld_ident(ent, id);
1651 set_entity_visibility(ent, ir_visibility_private);
1657 * Returns the trampoline entity for the given method.
1659 static ir_entity *get_trampoline(be_main_env_t *env, ir_entity *method)
1661 ir_entity *result = pmap_get(ir_entity, env->ent_trampoline_map, method);
1662 if (result == NULL) {
1663 result = create_trampoline(env, method);
1664 pmap_insert(env->ent_trampoline_map, method, result);
1670 static ir_entity *create_pic_symbol(be_main_env_t *be, ir_entity *entity)
1672 ident *old_id = get_entity_ld_ident(entity);
1673 ident *id = id_mangle3("", old_id, "$non_lazy_ptr");
1674 ir_type *e_type = get_entity_type(entity);
1675 ir_type *type = new_type_pointer(e_type);
1676 ir_type *parent = be->pic_symbols_type;
1677 ir_entity *ent = new_entity(parent, old_id, type);
1678 set_entity_ld_ident(ent, id);
1679 set_entity_visibility(ent, ir_visibility_private);
1684 static ir_entity *get_pic_symbol(be_main_env_t *env, ir_entity *entity)
1686 ir_entity *result = pmap_get(ir_entity, env->ent_pic_symbol_map, entity);
1687 if (result == NULL) {
1688 result = create_pic_symbol(env, entity);
1689 pmap_insert(env->ent_pic_symbol_map, entity, result);
1698 * Returns non-zero if a given entity can be accessed using a relative address.
1700 static int can_address_relative(ir_entity *entity)
1702 return entity_has_definition(entity) && !(get_entity_linkage(entity) & IR_LINKAGE_MERGE);
1705 static ir_node *get_pic_base(ir_graph *irg)
1707 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1708 if (arch_env->impl->get_pic_base == NULL)
1710 return arch_env->impl->get_pic_base(irg);
1713 /** patches SymConsts to work in position independent code */
1714 static void fix_pic_symconsts(ir_node *node, void *data)
1716 ir_graph *irg = get_irn_irg(node);
1717 be_main_env_t *be = be_get_irg_main_env(irg);
1727 arity = get_irn_arity(node);
1728 for (i = 0; i < arity; ++i) {
1730 ir_node *pred = get_irn_n(node, i);
1732 ir_entity *pic_symbol;
1733 ir_node *pic_symconst;
1735 if (!is_SymConst(pred))
1738 entity = get_SymConst_entity(pred);
1739 block = get_nodes_block(pred);
1741 /* calls can jump to relative addresses, so we can directly jump to
1742 the (relatively) known call address or the trampoline */
1743 if (i == 1 && is_Call(node)) {
1744 ir_entity *trampoline;
1745 ir_node *trampoline_const;
1747 if (can_address_relative(entity))
1750 dbgi = get_irn_dbg_info(pred);
1751 trampoline = get_trampoline(be, entity);
1752 trampoline_const = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
1754 set_irn_n(node, i, trampoline_const);
1758 /* everything else is accessed relative to EIP */
1759 mode = get_irn_mode(pred);
1760 pic_base = get_pic_base(irg);
1762 /* all ok now for locally constructed stuff */
1763 if (can_address_relative(entity)) {
1764 ir_node *add = new_r_Add(block, pic_base, pred, mode);
1766 /* make sure the walker doesn't visit this add again */
1767 mark_irn_visited(add);
1768 set_irn_n(node, i, add);
1772 /* get entry from pic symbol segment */
1773 dbgi = get_irn_dbg_info(pred);
1774 pic_symbol = get_pic_symbol(be, entity);
1775 pic_symconst = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
1777 add = new_r_Add(block, pic_base, pic_symconst, mode);
1778 mark_irn_visited(add);
1780 /* we need an extra indirection for global data outside our current
1781 module. The loads are always safe and can therefore float
1782 and need no memory input */
1783 load = new_r_Load(block, get_irg_no_mem(irg), add, mode, cons_floats);
1784 load_res = new_r_Proj(load, mode, pn_Load_res);
1786 set_irn_n(node, i, load_res);
1790 void be_abi_introduce(ir_graph *irg)
1792 ir_node *old_frame = get_irg_frame(irg);
1793 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1794 ir_entity *entity = get_irg_entity(irg);
1795 ir_type *method_type = get_entity_type(entity);
1796 be_irg_t *birg = be_birg_from_irg(irg);
1797 struct obstack *obst = &birg->obst;
1798 ir_node *dummy = new_r_Dummy(irg,
1799 arch_env->sp->reg_class->mode);
1802 /* determine allocatable registers */
1803 assert(birg->allocatable_regs == NULL);
1804 birg->allocatable_regs = rbitset_obstack_alloc(obst, arch_env->n_registers);
1805 for (r = 0; r < arch_env->n_registers; ++r) {
1806 const arch_register_t *reg = &arch_env->registers[r];
1807 if ( !(reg->type & arch_register_type_ignore)) {
1808 rbitset_set(birg->allocatable_regs, r);
1812 /* Break here if backend provides a custom API. */
1815 env.keep_map = pmap_create();
1816 env.call = be_abi_call_new(arch_env->sp->reg_class);
1817 arch_env_get_call_abi(arch_env, method_type, env.call);
1819 env.init_sp = dummy;
1820 env.calls = NEW_ARR_F(ir_node*, 0);
1824 if (be_options.pic) {
1825 irg_walk_graph(irg, fix_pic_symconsts, NULL, NULL);
1828 /* Lower all call nodes in the IRG. */
1829 process_calls(irg, &env);
1831 /* Process the IRG */
1832 modify_irg(irg, &env);
1834 /* fix call inputs for state registers */
1835 fix_call_state_inputs(irg, &env);
1837 be_abi_call_free(env.call);
1839 /* We don't need the keep map anymore. */
1840 pmap_destroy(env.keep_map);
1842 /* calls array is not needed anymore */
1843 DEL_ARR_F(env.calls);
1845 /* reroute the stack origin of the calls to the true stack origin. */
1846 exchange(dummy, env.init_sp);
1847 exchange(old_frame, get_irg_frame(irg));
1849 pmap_destroy(env.regs);
1852 void be_put_allocatable_regs(const ir_graph *irg,
1853 const arch_register_class_t *cls, bitset_t *bs)
1855 be_irg_t *birg = be_birg_from_irg(irg);
1856 unsigned *allocatable_regs = birg->allocatable_regs;
1859 assert(bitset_size(bs) == cls->n_regs);
1860 bitset_clear_all(bs);
1861 for (i = 0; i < cls->n_regs; ++i) {
1862 const arch_register_t *reg = &cls->regs[i];
1863 if (rbitset_is_set(allocatable_regs, reg->global_index))
1868 unsigned be_get_n_allocatable_regs(const ir_graph *irg,
1869 const arch_register_class_t *cls)
1871 bitset_t *bs = bitset_alloca(cls->n_regs);
1872 be_put_allocatable_regs(irg, cls, bs);
1873 return bitset_popcount(bs);
1876 void be_set_allocatable_regs(const ir_graph *irg,
1877 const arch_register_class_t *cls,
1878 unsigned *raw_bitset)
1880 be_irg_t *birg = be_birg_from_irg(irg);
1881 unsigned *allocatable_regs = birg->allocatable_regs;
1884 rbitset_clear_all(raw_bitset, cls->n_regs);
1885 for (i = 0; i < cls->n_regs; ++i) {
1886 const arch_register_t *reg = &cls->regs[i];
1887 if (rbitset_is_set(allocatable_regs, reg->global_index))
1888 rbitset_set(raw_bitset, i);
1892 BE_REGISTER_MODULE_CONSTRUCTOR(be_init_abi)
1893 void be_init_abi(void)
1895 FIRM_DBG_REGISTER(dbg, "firm.be.abi");