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
32 #include "irgraph_t.h"
35 #include "iredges_t.h"
38 #include "irprintf_t.h"
45 #include "raw_bitset.h"
56 #include "bessaconstr.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 in registers. */
64 unsigned on_stack : 1; /**< 1: this argument is transmitted on the stack. */
65 unsigned callee : 1; /**< 1: someone called us. 0: We call another function */
68 const arch_register_t *reg;
71 unsigned alignment; /**< stack alignment */
72 unsigned space_before; /**< allocate space before */
73 unsigned space_after; /**< allocate space after */
76 struct be_abi_call_t {
77 be_abi_call_flags_t flags; /**< Flags describing the ABI behavior on calls */
78 int pop; /**< number of bytes the stack frame is shrinked by the callee on return. */
79 const be_abi_callbacks_t *cb;
80 ir_type *between_type;
82 const arch_register_class_t *cls_addr; /**< register class of the call address */
86 * The ABI information for the current graph.
89 survive_dce_t *dce_survivor;
91 be_abi_call_t *call; /**< The ABI call information. */
93 ir_node *init_sp; /**< The node representing the stack pointer
94 at the start of the function. */
96 ir_node *start; /**< The be_Start params node. */
97 pmap *regs; /**< A map of all callee-save and ignore regs to
98 their Projs to the RegParams node. */
100 int start_block_bias; /**< The stack bias at the end of the start block. */
102 void *cb; /**< ABI Callback self pointer. */
104 pmap *keep_map; /**< mapping blocks to keep nodes. */
106 ir_node **calls; /**< flexible array containing all be_Call nodes */
109 static ir_heights_t *ir_heights;
111 /** Flag: if set, try to omit the frame pointer in all routines. */
112 static int be_omit_fp = 1;
115 _ ____ ___ ____ _ _ _ _
116 / \ | __ )_ _| / ___|__ _| | | |__ __ _ ___| | _____
117 / _ \ | _ \| | | | / _` | | | '_ \ / _` |/ __| |/ / __|
118 / ___ \| |_) | | | |__| (_| | | | |_) | (_| | (__| <\__ \
119 /_/ \_\____/___| \____\__,_|_|_|_.__/ \__,_|\___|_|\_\___/
121 These callbacks are used by the backend to set the parameters
122 for a specific call type.
126 * Set compare function: compares two ABI call object arguments.
128 static int cmp_call_arg(const void *a, const void *b, size_t n)
130 const be_abi_call_arg_t *p = (const be_abi_call_arg_t*)a;
131 const be_abi_call_arg_t *q = (const be_abi_call_arg_t*)b;
133 return !(p->is_res == q->is_res && p->pos == q->pos && p->callee == q->callee);
137 * Get an ABI call object argument.
139 * @param call the abi call
140 * @param is_res true for call results, false for call arguments
141 * @param pos position of the argument
142 * @param callee context type - if we are callee or caller
144 static be_abi_call_arg_t *get_call_arg(be_abi_call_t *call, int is_res, int pos, int callee)
146 be_abi_call_arg_t arg;
149 memset(&arg, 0, sizeof(arg));
154 hash = is_res * 128 + pos;
156 return (be_abi_call_arg_t*)set_find(call->params, &arg, sizeof(arg), hash);
160 * Set an ABI call object argument.
162 static void remember_call_arg(be_abi_call_arg_t *arg, be_abi_call_t *call, be_abi_context_t context)
164 unsigned hash = arg->is_res * 128 + arg->pos;
165 if (context & ABI_CONTEXT_CALLEE) {
167 set_insert(call->params, arg, sizeof(*arg), hash);
169 if (context & ABI_CONTEXT_CALLER) {
171 set_insert(call->params, arg, sizeof(*arg), hash);
175 /* Set the flags for a call. */
176 void be_abi_call_set_flags(be_abi_call_t *call, be_abi_call_flags_t flags, const be_abi_callbacks_t *cb)
182 /* Sets the number of bytes the stackframe is shrinked by the callee on return */
183 void be_abi_call_set_pop(be_abi_call_t *call, int pop)
189 /* Set register class for call address */
190 void be_abi_call_set_call_address_reg_class(be_abi_call_t *call, const arch_register_class_t *cls)
192 call->cls_addr = cls;
196 void be_abi_call_param_stack(be_abi_call_t *call, int arg_pos,
197 ir_mode *load_mode, unsigned alignment,
198 unsigned space_before, unsigned space_after,
199 be_abi_context_t context)
201 be_abi_call_arg_t arg;
202 memset(&arg, 0, sizeof(arg));
203 assert(alignment > 0 && "Alignment must be greater than 0");
205 arg.load_mode = load_mode;
206 arg.alignment = alignment;
207 arg.space_before = space_before;
208 arg.space_after = space_after;
212 remember_call_arg(&arg, call, context);
215 void be_abi_call_param_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
217 be_abi_call_arg_t arg;
218 memset(&arg, 0, sizeof(arg));
225 remember_call_arg(&arg, call, context);
228 void be_abi_call_res_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
230 be_abi_call_arg_t arg;
231 memset(&arg, 0, sizeof(arg));
238 remember_call_arg(&arg, call, context);
241 /* Get the flags of a ABI call object. */
242 be_abi_call_flags_t be_abi_call_get_flags(const be_abi_call_t *call)
248 * Constructor for a new ABI call object.
250 * @param cls_addr register class of the call address
252 * @return the new ABI call object
254 static be_abi_call_t *be_abi_call_new(const arch_register_class_t *cls_addr)
256 be_abi_call_t *call = XMALLOCZ(be_abi_call_t);
259 call->params = new_set(cmp_call_arg, 16);
261 call->cls_addr = cls_addr;
263 call->flags.bits.try_omit_fp = be_omit_fp;
269 * Destructor for an ABI call object.
271 static void be_abi_call_free(be_abi_call_t *call)
273 del_set(call->params);
278 * Initializes the frame layout from parts
280 * @param frame the stack layout that will be initialized
281 * @param args the stack argument layout type
282 * @param between the between layout type
283 * @param locals the method frame type
284 * @param stack_dir the stack direction: < 0 decreasing, > 0 increasing addresses
285 * @param param_map an array mapping method argument positions to the stack argument type
287 * @return the initialized stack layout
289 static be_stack_layout_t *stack_frame_init(be_stack_layout_t *frame, ir_type *args,
290 ir_type *between, ir_type *locals, int stack_dir,
291 ir_entity *param_map[])
293 frame->arg_type = args;
294 frame->between_type = between;
295 frame->frame_type = locals;
296 frame->initial_offset = 0;
297 frame->initial_bias = 0;
298 frame->stack_dir = stack_dir;
299 frame->order[1] = between;
300 frame->param_map = param_map;
303 frame->order[0] = args;
304 frame->order[2] = locals;
306 /* typical decreasing stack: locals have the
307 * lowest addresses, arguments the highest */
308 frame->order[0] = locals;
309 frame->order[2] = args;
321 Adjustment of the calls inside a graph.
326 * Transform a call node into a be_Call node.
328 * @param env The ABI environment for the current irg.
329 * @param irn The call node.
330 * @param curr_sp The stack pointer node to use.
331 * @return The stack pointer after the call.
333 static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp)
335 ir_graph *irg = get_irn_irg(irn);
336 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
337 ir_type *call_tp = get_Call_type(irn);
338 ir_node *call_ptr = get_Call_ptr(irn);
339 size_t n_params = get_method_n_params(call_tp);
340 ir_node *curr_mem = get_Call_mem(irn);
341 ir_node *bl = get_nodes_block(irn);
343 int stack_dir = arch_env->stack_dir;
344 const arch_register_t *sp = arch_env->sp;
345 be_abi_call_t *call = be_abi_call_new(sp->reg_class);
346 ir_mode *mach_mode = sp->reg_class->mode;
347 int no_alloc = call->flags.bits.frame_is_setup_on_call;
348 int n_res = get_method_n_ress(call_tp);
349 int do_seq = call->flags.bits.store_args_sequential && !no_alloc;
351 ir_node *res_proj = NULL;
352 int n_reg_params = 0;
353 int n_stack_params = 0;
356 const arch_register_t **states = NEW_ARR_F(const arch_register_t*, 0);
357 const arch_register_t **destroyed_regs = NEW_ARR_F(const arch_register_t*, 0);
361 int n_reg_results = 0;
362 const ir_edge_t *edge;
364 int *stack_param_idx;
365 int i, n, destroy_all_regs;
370 /* Let the isa fill out the abi description for that call node. */
371 arch_env_get_call_abi(arch_env, call_tp, call);
373 /* Insert code to put the stack arguments on the stack. */
374 assert(get_Call_n_params(irn) == n_params);
375 stack_param_idx = ALLOCAN(int, n_params);
376 for (p = 0; p < n_params; ++p) {
377 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
380 int arg_size = get_type_size_bytes(get_method_param_type(call_tp, p));
382 stack_size += round_up2(arg->space_before, arg->alignment);
383 stack_size += round_up2(arg_size, arg->alignment);
384 stack_size += round_up2(arg->space_after, arg->alignment);
386 stack_param_idx[n_stack_params++] = p;
390 /* Collect all arguments which are passed in registers. */
391 reg_param_idxs = ALLOCAN(int, n_params);
392 for (p = 0; p < n_params; ++p) {
393 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
394 if (arg && arg->in_reg) {
395 reg_param_idxs[n_reg_params++] = p;
400 * If the stack is decreasing and we do not want to store sequentially,
401 * or someone else allocated the call frame
402 * we allocate as much space on the stack all parameters need, by
403 * moving the stack pointer along the stack's direction.
405 * Note: we also have to do this for stack_size == 0, because we may have
406 * to adjust stack alignment for the call.
408 if (stack_dir < 0 && !do_seq && !no_alloc) {
409 curr_sp = be_new_IncSP(sp, bl, curr_sp, stack_size, 1);
412 dbgi = get_irn_dbg_info(irn);
413 /* If there are some parameters which shall be passed on the stack. */
414 if (n_stack_params > 0) {
416 ir_node **in = ALLOCAN(ir_node*, n_stack_params+1);
420 * Reverse list of stack parameters if call arguments are from left to right.
421 * We must them reverse again if they are pushed (not stored) and the stack
422 * direction is downwards.
424 if (call->flags.bits.left_to_right ^ (do_seq && stack_dir < 0)) {
425 for (i = 0; i < n_stack_params >> 1; ++i) {
426 int other = n_stack_params - i - 1;
427 int tmp = stack_param_idx[i];
428 stack_param_idx[i] = stack_param_idx[other];
429 stack_param_idx[other] = tmp;
433 curr_mem = get_Call_mem(irn);
435 in[n_in++] = curr_mem;
438 for (i = 0; i < n_stack_params; ++i) {
439 int p = stack_param_idx[i];
440 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
441 ir_node *param = get_Call_param(irn, p);
442 ir_node *addr = curr_sp;
444 ir_type *param_type = get_method_param_type(call_tp, p);
445 int param_size = get_type_size_bytes(param_type) + arg->space_after;
448 * If we wanted to build the arguments sequentially,
449 * the stack pointer for the next must be incremented,
450 * and the memory value propagated.
454 addr = curr_sp = be_new_IncSP(sp, bl, curr_sp,
455 param_size + arg->space_before, 0);
456 add_irn_dep(curr_sp, curr_mem);
458 curr_ofs += arg->space_before;
459 curr_ofs = round_up2(curr_ofs, arg->alignment);
461 /* Make the expression to compute the argument's offset. */
463 ir_mode *constmode = mach_mode;
464 if (mode_is_reference(mach_mode)) {
467 addr = new_r_Const_long(irg, constmode, curr_ofs);
468 addr = new_r_Add(bl, curr_sp, addr, mach_mode);
472 /* Insert a store for primitive arguments. */
473 if (is_atomic_type(param_type)) {
474 ir_node *mem_input = do_seq ? curr_mem : new_r_NoMem(irg);
475 ir_node *store = new_rd_Store(dbgi, bl, mem_input, addr, param, cons_none);
476 mem = new_r_Proj(store, mode_M, pn_Store_M);
478 /* Make a mem copy for compound arguments. */
481 assert(mode_is_reference(get_irn_mode(param)));
482 copy = new_rd_CopyB(dbgi, bl, curr_mem, addr, param, param_type);
483 mem = new_r_Proj(copy, mode_M, pn_CopyB_M);
486 curr_ofs += param_size;
494 /* We need the sync only, if we didn't build the stores sequentially. */
496 if (n_stack_params >= 1) {
497 curr_mem = new_r_Sync(bl, n_in, in);
499 curr_mem = get_Call_mem(irn);
504 /* check for the return_twice property */
505 destroy_all_regs = 0;
506 if (is_SymConst_addr_ent(call_ptr)) {
507 ir_entity *ent = get_SymConst_entity(call_ptr);
509 if (get_entity_additional_properties(ent) & mtp_property_returns_twice)
510 destroy_all_regs = 1;
512 ir_type *call_tp = get_Call_type(irn);
514 if (get_method_additional_properties(call_tp) & mtp_property_returns_twice)
515 destroy_all_regs = 1;
518 /* Put caller save into the destroyed set and state registers in the states
520 for (i = 0, n = arch_env->n_register_classes; i < n; ++i) {
522 const arch_register_class_t *cls = &arch_env->register_classes[i];
523 for (j = 0; j < cls->n_regs; ++j) {
524 const arch_register_t *reg = arch_register_for_index(cls, j);
526 /* even if destroyed all is specified, neither SP nor FP are
527 * destroyed (else bad things will happen) */
528 if (reg == arch_env->sp || reg == arch_env->bp)
531 if (reg->type & arch_register_type_state) {
532 ARR_APP1(const arch_register_t*, destroyed_regs, reg);
533 ARR_APP1(const arch_register_t*, states, reg);
534 /* we're already in the destroyed set so no need for further
538 if (destroy_all_regs || (reg->type & arch_register_type_caller_save)) {
539 if (!(reg->type & arch_register_type_ignore)) {
540 ARR_APP1(const arch_register_t*, destroyed_regs, reg);
546 /* search the largest result proj number */
547 res_projs = ALLOCANZ(ir_node*, n_res);
549 foreach_out_edge(irn, edge) {
550 const ir_edge_t *res_edge;
551 ir_node *irn = get_edge_src_irn(edge);
553 if (!is_Proj(irn) || get_Proj_proj(irn) != pn_Call_T_result)
556 foreach_out_edge(irn, res_edge) {
558 ir_node *res = get_edge_src_irn(res_edge);
560 assert(is_Proj(res));
562 proj = get_Proj_proj(res);
563 assert(proj < n_res);
564 assert(res_projs[proj] == NULL);
565 res_projs[proj] = res;
571 /** TODO: this is not correct for cases where return values are passed
572 * on the stack, but no known ABI does this currently...
574 n_reg_results = n_res;
577 in = ALLOCAN(ir_node*, n_reg_params + ARR_LEN(states));
579 /* make the back end call node and set its register requirements. */
580 for (i = 0; i < n_reg_params; ++i) {
581 in[n_ins++] = get_Call_param(irn, reg_param_idxs[i]);
584 /* add state registers ins */
585 for (s = 0; s < ARR_LEN(states); ++s) {
586 const arch_register_t *reg = states[s];
587 const arch_register_class_t *cls = arch_register_get_class(reg);
589 ir_node *regnode = be_abi_reg_map_get(env->regs, reg);
590 ir_fprintf(stderr, "Adding %+F\n", regnode);
592 ir_node *regnode = new_r_Unknown(irg, arch_register_class_mode(cls));
593 in[n_ins++] = regnode;
595 assert(n_ins == (int) (n_reg_params + ARR_LEN(states)));
597 /* ins collected, build the call */
598 if (env->call->flags.bits.call_has_imm && is_SymConst(call_ptr)) {
600 low_call = be_new_Call(dbgi, irg, bl, curr_mem, curr_sp, curr_sp,
601 n_reg_results + pn_be_Call_first_res + ARR_LEN(destroyed_regs),
602 n_ins, in, get_Call_type(irn));
603 be_Call_set_entity(low_call, get_SymConst_entity(call_ptr));
606 low_call = be_new_Call(dbgi, irg, bl, curr_mem, curr_sp, call_ptr,
607 n_reg_results + pn_be_Call_first_res + ARR_LEN(destroyed_regs),
608 n_ins, in, get_Call_type(irn));
610 be_Call_set_pop(low_call, call->pop);
612 /* put the call into the list of all calls for later processing */
613 ARR_APP1(ir_node *, env->calls, low_call);
615 /* create new stack pointer */
616 curr_sp = new_r_Proj(low_call, get_irn_mode(curr_sp), pn_be_Call_sp);
617 be_set_constr_single_reg_out(low_call, pn_be_Call_sp, sp,
618 arch_register_req_type_ignore | arch_register_req_type_produces_sp);
619 arch_set_irn_register(curr_sp, sp);
621 /* now handle results */
622 for (i = 0; i < n_res; ++i) {
624 ir_node *proj = res_projs[i];
625 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 0);
627 /* returns values on stack not supported yet */
631 shift the proj number to the right, since we will drop the
632 unspeakable Proj_T from the Call. Therefore, all real argument
633 Proj numbers must be increased by pn_be_Call_first_res
635 pn = i + pn_be_Call_first_res;
638 ir_type *res_type = get_method_res_type(call_tp, i);
639 ir_mode *mode = get_type_mode(res_type);
640 proj = new_r_Proj(low_call, mode, pn);
643 set_Proj_pred(proj, low_call);
644 set_Proj_proj(proj, pn);
648 /* remove register from destroyed regs */
650 size_t n = ARR_LEN(destroyed_regs);
651 for (j = 0; j < n; ++j) {
652 if (destroyed_regs[j] == arg->reg) {
653 destroyed_regs[j] = destroyed_regs[n-1];
654 ARR_SHRINKLEN(destroyed_regs,n-1);
662 Set the register class of the call address to
663 the backend provided class (default: stack pointer class)
665 be_node_set_reg_class_in(low_call, be_pos_Call_ptr, call->cls_addr);
667 DBG((dbg, LEVEL_3, "\tcreated backend call %+F\n", low_call));
669 /* Set the register classes and constraints of the Call parameters. */
670 for (i = 0; i < n_reg_params; ++i) {
671 int index = reg_param_idxs[i];
672 be_abi_call_arg_t *arg = get_call_arg(call, 0, index, 0);
673 assert(arg->reg != NULL);
675 be_set_constr_single_reg_in(low_call, be_pos_Call_first_arg + i,
676 arg->reg, arch_register_req_type_none);
679 /* Set the register constraints of the results. */
680 for (i = 0; i < n_res; ++i) {
681 ir_node *proj = res_projs[i];
682 const be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 0);
683 int pn = get_Proj_proj(proj);
686 be_set_constr_single_reg_out(low_call, pn, arg->reg,
687 arch_register_req_type_none);
688 arch_set_irn_register(proj, arg->reg);
690 exchange(irn, low_call);
692 /* kill the ProjT node */
693 if (res_proj != NULL) {
697 /* Make additional projs for the caller save registers
698 and the Keep node which keeps them alive. */
704 int curr_res_proj = pn_be_Call_first_res + n_reg_results;
707 n_ins = ARR_LEN(destroyed_regs) + n_reg_results + 1;
708 in = ALLOCAN(ir_node *, n_ins);
710 /* also keep the stack pointer */
711 set_irn_link(curr_sp, (void*) sp);
714 for (d = 0; d < ARR_LEN(destroyed_regs); ++d) {
715 const arch_register_t *reg = destroyed_regs[d];
716 ir_node *proj = new_r_Proj(low_call, reg->reg_class->mode, curr_res_proj);
718 /* memorize the register in the link field. we need afterwards to set the register class of the keep correctly. */
719 be_set_constr_single_reg_out(low_call, curr_res_proj, reg,
720 arch_register_req_type_none);
721 arch_set_irn_register(proj, reg);
723 set_irn_link(proj, (void*) reg);
728 for (i = 0; i < n_reg_results; ++i) {
729 ir_node *proj = res_projs[i];
730 const arch_register_t *reg = arch_get_irn_register(proj);
731 set_irn_link(proj, (void*) reg);
736 /* create the Keep for the caller save registers */
737 keep = be_new_Keep(bl, n, in);
738 for (i = 0; i < n; ++i) {
739 const arch_register_t *reg = (const arch_register_t*)get_irn_link(in[i]);
740 be_node_set_reg_class_in(keep, i, reg->reg_class);
744 /* Clean up the stack. */
745 assert(stack_size >= call->pop);
746 stack_size -= call->pop;
748 if (stack_size > 0) {
749 ir_node *mem_proj = NULL;
751 foreach_out_edge(low_call, edge) {
752 ir_node *irn = get_edge_src_irn(edge);
753 if (is_Proj(irn) && get_Proj_proj(irn) == pn_Call_M) {
760 mem_proj = new_r_Proj(low_call, mode_M, pn_be_Call_M_regular);
761 keep_alive(mem_proj);
764 /* Clean up the stack frame or revert alignment fixes if we allocated it */
766 curr_sp = be_new_IncSP(sp, bl, curr_sp, -stack_size, 0);
769 be_abi_call_free(call);
772 DEL_ARR_F(destroyed_regs);
778 * Adjust the size of a node representing a stack alloc or free for the minimum stack alignment.
780 * @param alignment the minimum stack alignment
781 * @param size the node containing the non-aligned size
782 * @param block the block where new nodes are allocated on
783 * @param dbg debug info for new nodes
785 * @return a node representing the aligned size
787 static ir_node *adjust_alloc_size(unsigned stack_alignment, ir_node *size,
788 ir_node *block, dbg_info *dbg)
790 if (stack_alignment > 1) {
796 assert(is_po2(stack_alignment));
798 mode = get_irn_mode(size);
799 tv = new_tarval_from_long(stack_alignment-1, mode);
800 irg = get_Block_irg(block);
801 mask = new_r_Const(irg, tv);
802 size = new_rd_Add(dbg, block, size, mask, mode);
804 tv = new_tarval_from_long(-(long)stack_alignment, mode);
805 mask = new_r_Const(irg, tv);
806 size = new_rd_And(dbg, block, size, mask, mode);
812 * The alloca is transformed into a back end alloca node and connected to the stack nodes.
814 static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp)
816 ir_node *block = get_nodes_block(alloc);
817 ir_graph *irg = get_Block_irg(block);
818 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
819 ir_node *alloc_mem = NULL;
820 ir_node *alloc_res = NULL;
821 ir_type *type = get_Alloc_type(alloc);
824 const ir_edge_t *edge;
829 unsigned stack_alignment;
831 /* all non-stack Alloc nodes should already be lowered before the backend */
832 assert(get_Alloc_where(alloc) == stack_alloc);
834 foreach_out_edge(alloc, edge) {
835 ir_node *irn = get_edge_src_irn(edge);
837 assert(is_Proj(irn));
838 switch (get_Proj_proj(irn)) {
850 /* Beware: currently Alloc nodes without a result might happen,
851 only escape analysis kills them and this phase runs only for object
852 oriented source. We kill the Alloc here. */
853 if (alloc_res == NULL && alloc_mem) {
854 exchange(alloc_mem, get_Alloc_mem(alloc));
858 dbg = get_irn_dbg_info(alloc);
859 count = get_Alloc_count(alloc);
861 /* we might need to multiply the count with the element size */
862 if (type != firm_unknown_type && get_type_size_bytes(type) != 1) {
863 ir_mode *mode = get_irn_mode(count);
864 ir_tarval *tv = new_tarval_from_long(get_type_size_bytes(type),
866 ir_node *cnst = new_rd_Const(dbg, irg, tv);
867 size = new_rd_Mul(dbg, block, count, cnst, mode);
872 /* The stack pointer will be modified in an unknown manner.
873 We cannot omit it. */
874 env->call->flags.bits.try_omit_fp = 0;
876 stack_alignment = 1 << arch_env->stack_alignment;
877 size = adjust_alloc_size(stack_alignment, size, block, dbg);
878 new_alloc = be_new_AddSP(arch_env->sp, block, curr_sp, size);
879 set_irn_dbg_info(new_alloc, dbg);
881 if (alloc_mem != NULL) {
885 addsp_mem = new_r_Proj(new_alloc, mode_M, pn_be_AddSP_M);
887 /* We need to sync the output mem of the AddSP with the input mem
888 edge into the alloc node. */
889 ins[0] = get_Alloc_mem(alloc);
891 sync = new_r_Sync(block, 2, ins);
893 exchange(alloc_mem, sync);
896 exchange(alloc, new_alloc);
898 /* fix projnum of alloca res */
899 set_Proj_proj(alloc_res, pn_be_AddSP_res);
901 curr_sp = new_r_Proj(new_alloc, get_irn_mode(curr_sp), pn_be_AddSP_sp);
908 * The Free is transformed into a back end free node and connected to the stack nodes.
910 static ir_node *adjust_free(be_abi_irg_t *env, ir_node *free, ir_node *curr_sp)
912 ir_node *block = get_nodes_block(free);
913 ir_graph *irg = get_irn_irg(free);
914 ir_type *type = get_Free_type(free);
915 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
916 ir_mode *sp_mode = arch_env->sp->reg_class->mode;
917 dbg_info *dbg = get_irn_dbg_info(free);
918 ir_node *subsp, *mem, *res, *size, *sync;
920 unsigned stack_alignment;
922 /* all non-stack-alloc Free nodes should already be lowered before the
924 assert(get_Free_where(free) == stack_alloc);
926 /* we might need to multiply the size with the element size */
927 if (type != firm_unknown_type && get_type_size_bytes(type) != 1) {
928 ir_tarval *tv = new_tarval_from_long(get_type_size_bytes(type), mode_Iu);
929 ir_node *cnst = new_rd_Const(dbg, irg, tv);
930 ir_node *mul = new_rd_Mul(dbg, block, get_Free_size(free),
934 size = get_Free_size(free);
937 stack_alignment = 1 << arch_env->stack_alignment;
938 size = adjust_alloc_size(stack_alignment, size, block, dbg);
940 /* The stack pointer will be modified in an unknown manner.
941 We cannot omit it. */
942 env->call->flags.bits.try_omit_fp = 0;
943 subsp = be_new_SubSP(arch_env->sp, block, curr_sp, size);
944 set_irn_dbg_info(subsp, dbg);
946 mem = new_r_Proj(subsp, mode_M, pn_be_SubSP_M);
947 res = new_r_Proj(subsp, sp_mode, pn_be_SubSP_sp);
949 /* we need to sync the memory */
950 in[0] = get_Free_mem(free);
952 sync = new_r_Sync(block, 2, in);
954 /* and make the AddSP dependent on the former memory */
955 add_irn_dep(subsp, get_Free_mem(free));
958 exchange(free, sync);
965 * Check if a node is somehow data dependent on another one.
966 * both nodes must be in the same basic block.
967 * @param n1 The first node.
968 * @param n2 The second node.
969 * @return 1, if n1 is data dependent (transitively) on n2, 0 if not.
971 static int dependent_on(ir_node *n1, ir_node *n2)
973 assert(get_nodes_block(n1) == get_nodes_block(n2));
975 return heights_reachable_in_block(ir_heights, n1, n2);
978 static int cmp_call_dependency(const void *c1, const void *c2)
980 ir_node *n1 = *(ir_node **) c1;
981 ir_node *n2 = *(ir_node **) c2;
984 Classical qsort() comparison function behavior:
985 0 if both elements are equal
986 1 if second is "smaller" that first
987 -1 if first is "smaller" that second
989 if (dependent_on(n1, n2))
992 if (dependent_on(n2, n1))
995 /* The nodes have no depth order, but we need a total order because qsort()
997 return get_irn_idx(n1) - get_irn_idx(n2);
1001 * Walker: links all Call/Alloc/Free nodes to the Block they are contained.
1002 * Clears the irg_is_leaf flag if a Call is detected.
1004 static void link_ops_in_block_walker(ir_node *irn, void *data)
1006 be_abi_irg_t *env = (be_abi_irg_t*)data;
1007 unsigned code = get_irn_opcode(irn);
1009 if (code == iro_Call ||
1010 (code == iro_Alloc && get_Alloc_where(irn) == stack_alloc) ||
1011 (code == iro_Free && get_Free_where(irn) == stack_alloc)) {
1012 ir_node *bl = get_nodes_block(irn);
1013 void *save = get_irn_link(bl);
1015 if (code == iro_Call)
1016 env->call->flags.bits.irg_is_leaf = 0;
1018 set_irn_link(irn, save);
1019 set_irn_link(bl, irn);
1022 if (code == iro_Builtin && get_Builtin_kind(irn) == ir_bk_return_address) {
1023 ir_node *param = get_Builtin_param(irn, 0);
1024 ir_tarval *tv = get_Const_tarval(param);
1025 unsigned long value = get_tarval_long(tv);
1026 /* use ebp, so the climbframe algo works... */
1028 env->call->flags.bits.try_omit_fp = 0;
1035 * Process all Call/Alloc/Free nodes inside a basic block.
1036 * Note that the link field of the block must contain a linked list of all
1037 * nodes inside the Block. We first order this list according to data dependency
1038 * and that connect the nodes together.
1040 static void process_ops_in_block(ir_node *bl, void *data)
1042 be_abi_irg_t *env = (be_abi_irg_t*)data;
1043 ir_node *curr_sp = env->init_sp;
1050 for (irn = (ir_node*)get_irn_link(bl); irn != NULL;
1051 irn = (ir_node*)get_irn_link(irn)) {
1055 nodes = ALLOCAN(ir_node*, n_nodes);
1056 for (irn = (ir_node*)get_irn_link(bl), n = 0; irn != NULL;
1057 irn = (ir_node*)get_irn_link(irn), ++n) {
1061 /* If there were call nodes in the block. */
1066 /* order the call nodes according to data dependency */
1067 qsort(nodes, n_nodes, sizeof(nodes[0]), cmp_call_dependency);
1069 for (i = n_nodes - 1; i >= 0; --i) {
1070 ir_node *irn = nodes[i];
1072 DBG((dbg, LEVEL_3, "\tprocessing call %+F\n", irn));
1073 switch (get_irn_opcode(irn)) {
1076 /* The stack pointer will be modified due to a call. */
1077 env->call->flags.bits.try_omit_fp = 0;
1079 curr_sp = adjust_call(env, irn, curr_sp);
1082 if (get_Alloc_where(irn) == stack_alloc)
1083 curr_sp = adjust_alloc(env, irn, curr_sp);
1086 if (get_Free_where(irn) == stack_alloc)
1087 curr_sp = adjust_free(env, irn, curr_sp);
1090 panic("invalid call");
1094 /* Keep the last stack state in the block by tying it to Keep node,
1095 * the proj from calls is already kept */
1096 if (curr_sp != env->init_sp &&
1097 !(is_Proj(curr_sp) && be_is_Call(get_Proj_pred(curr_sp)))) {
1099 keep = be_new_Keep(bl, 1, nodes);
1100 pmap_insert(env->keep_map, bl, keep);
1104 set_irn_link(bl, curr_sp);
1108 * Adjust all call nodes in the graph to the ABI conventions.
1110 static void process_calls(ir_graph *irg)
1112 be_abi_irg_t *abi = be_get_irg_abi(irg);
1114 abi->call->flags.bits.irg_is_leaf = 1;
1115 irg_walk_graph(irg, firm_clear_link, link_ops_in_block_walker, abi);
1117 ir_heights = heights_new(irg);
1118 irg_block_walk_graph(irg, NULL, process_ops_in_block, abi);
1119 heights_free(ir_heights);
1123 * Computes the stack argument layout type.
1124 * Changes a possibly allocated value param type by moving
1125 * entities to the stack layout type.
1127 * @param env the ABI environment
1128 * @param call the current call ABI
1129 * @param method_type the method type
1130 * @param val_param_tp the value parameter type, will be destroyed
1131 * @param param_map an array mapping method arguments to the stack layout type
1133 * @return the stack argument layout type
1135 static ir_type *compute_arg_type(be_abi_irg_t *env, ir_graph *irg,
1136 be_abi_call_t *call,
1137 ir_type *method_type, ir_type *val_param_tp,
1138 ir_entity ***param_map)
1140 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1141 int dir = env->call->flags.bits.left_to_right ? 1 : -1;
1142 int inc = arch_env->stack_dir * dir;
1143 int n = get_method_n_params(method_type);
1144 int curr = inc > 0 ? 0 : n - 1;
1145 struct obstack *obst = be_get_be_obst(irg);
1151 ident *id = get_entity_ident(get_irg_entity(irg));
1154 *param_map = map = OALLOCN(obst, ir_entity*, n);
1155 res = new_type_struct(id_mangle_u(id, new_id_from_chars("arg_type", 8)));
1156 for (i = 0; i < n; ++i, curr += inc) {
1157 ir_type *param_type = get_method_param_type(method_type, curr);
1158 be_abi_call_arg_t *arg = get_call_arg(call, 0, curr, 1);
1161 if (arg->on_stack) {
1162 if (val_param_tp != NULL) {
1163 /* the entity was already created, create a copy in the param type */
1164 ir_entity *val_ent = get_method_value_param_ent(method_type, i);
1165 arg->stack_ent = copy_entity_own(val_ent, res);
1166 set_entity_link(val_ent, arg->stack_ent);
1167 set_entity_link(arg->stack_ent, NULL);
1169 /* create a new entity */
1170 snprintf(buf, sizeof(buf), "param_%d", i);
1171 arg->stack_ent = new_entity(res, new_id_from_str(buf), param_type);
1173 ofs += arg->space_before;
1174 ofs = round_up2(ofs, arg->alignment);
1175 set_entity_offset(arg->stack_ent, ofs);
1176 ofs += arg->space_after;
1177 ofs += get_type_size_bytes(param_type);
1178 map[i] = arg->stack_ent;
1181 set_type_size_bytes(res, ofs);
1182 set_type_state(res, layout_fixed);
1187 const arch_register_t *reg;
1191 static int cmp_regs(const void *a, const void *b)
1193 const reg_node_map_t *p = (const reg_node_map_t*)a;
1194 const reg_node_map_t *q = (const reg_node_map_t*)b;
1196 if (p->reg->reg_class == q->reg->reg_class)
1197 return p->reg->index - q->reg->index;
1199 return p->reg->reg_class < q->reg->reg_class ? -1 : +1;
1202 static void reg_map_to_arr(reg_node_map_t *res, pmap *reg_map)
1205 size_t n = pmap_count(reg_map);
1208 foreach_pmap(reg_map, ent) {
1209 res[i].reg = (const arch_register_t*)ent->key;
1210 res[i].irn = (ir_node*)ent->value;
1214 qsort(res, n, sizeof(res[0]), cmp_regs);
1218 * Creates a barrier.
1220 static ir_node *create_barrier(ir_node *bl, ir_node **mem, pmap *regs,
1223 int n_regs = pmap_count(regs);
1229 in = ALLOCAN(ir_node*, n_regs+1);
1230 rm = ALLOCAN(reg_node_map_t, n_regs);
1231 reg_map_to_arr(rm, regs);
1232 for (n = 0; n < n_regs; ++n) {
1240 irn = be_new_Barrier(bl, n, in);
1242 for (n = 0; n < n_regs; ++n) {
1243 ir_node *pred = rm[n].irn;
1244 const arch_register_t *reg = rm[n].reg;
1245 arch_register_req_type_t add_type = arch_register_req_type_none;
1247 const backend_info_t *info;
1249 /* stupid workaround for now... as not all nodes report register
1251 info = be_get_info(skip_Proj(pred));
1252 if (info != NULL && info->out_infos != NULL) {
1253 const arch_register_req_t *ireq = arch_get_register_req_out(pred);
1254 if (ireq->type & arch_register_req_type_ignore)
1255 add_type |= arch_register_req_type_ignore;
1256 if (ireq->type & arch_register_req_type_produces_sp)
1257 add_type |= arch_register_req_type_produces_sp;
1260 proj = new_r_Proj(irn, get_irn_mode(pred), n);
1261 be_node_set_reg_class_in(irn, n, reg->reg_class);
1263 be_set_constr_single_reg_in(irn, n, reg,
1264 arch_register_req_type_none);
1266 be_set_constr_single_reg_out(irn, n, reg, add_type);
1267 arch_set_irn_register(proj, reg);
1269 pmap_insert(regs, (void *) reg, proj);
1273 *mem = new_r_Proj(irn, mode_M, n);
1280 * Creates a be_Return for a Return node.
1282 * @param @env the abi environment
1283 * @param irn the Return node or NULL if there was none
1284 * @param bl the block where the be_Retun should be placed
1285 * @param mem the current memory
1286 * @param n_res number of return results
1288 static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl,
1289 ir_node *mem, int n_res)
1291 be_abi_call_t *call = env->call;
1292 ir_graph *irg = get_Block_irg(bl);
1293 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1295 pmap *reg_map = pmap_create();
1296 ir_node *keep = (ir_node*)pmap_get(env->keep_map, bl);
1303 const arch_register_t **regs;
1307 get the valid stack node in this block.
1308 If we had a call in that block there is a Keep constructed by process_calls()
1309 which points to the last stack modification in that block. we'll use
1310 it then. Else we use the stack from the start block and let
1311 the ssa construction fix the usage.
1313 stack = be_abi_reg_map_get(env->regs, arch_env->sp);
1315 stack = get_irn_n(keep, 0);
1317 remove_End_keepalive(get_irg_end(irg), keep);
1320 /* Insert results for Return into the register map. */
1321 for (i = 0; i < n_res; ++i) {
1322 ir_node *res = get_Return_res(irn, i);
1323 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
1324 assert(arg->in_reg && "return value must be passed in register");
1325 pmap_insert(reg_map, (void *) arg->reg, res);
1328 /* Add uses of the callee save registers. */
1329 foreach_pmap(env->regs, ent) {
1330 const arch_register_t *reg = (const arch_register_t*)ent->key;
1331 if (reg->type & (arch_register_type_callee_save | arch_register_type_ignore))
1332 pmap_insert(reg_map, ent->key, ent->value);
1335 be_abi_reg_map_set(reg_map, arch_env->sp, stack);
1337 /* Make the Epilogue node and call the arch's epilogue maker. */
1338 create_barrier(bl, &mem, reg_map, 1);
1339 call->cb->epilogue(env->cb, bl, &mem, reg_map);
1342 Maximum size of the in array for Return nodes is
1343 return args + callee save/ignore registers + memory + stack pointer
1345 in_max = pmap_count(reg_map) + n_res + 2;
1347 in = ALLOCAN(ir_node*, in_max);
1348 regs = ALLOCAN(arch_register_t const*, in_max);
1351 in[1] = be_abi_reg_map_get(reg_map, arch_env->sp);
1353 regs[1] = arch_env->sp;
1356 /* clear SP entry, since it has already been grown. */
1357 pmap_insert(reg_map, (void *) arch_env->sp, NULL);
1358 for (i = 0; i < n_res; ++i) {
1359 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
1361 in[n] = be_abi_reg_map_get(reg_map, arg->reg);
1362 regs[n++] = arg->reg;
1364 /* Clear the map entry to mark the register as processed. */
1365 be_abi_reg_map_set(reg_map, arg->reg, NULL);
1368 /* grow the rest of the stuff. */
1369 foreach_pmap(reg_map, ent) {
1371 in[n] = (ir_node*)ent->value;
1372 regs[n++] = (const arch_register_t*)ent->key;
1376 /* The in array for the new back end return is now ready. */
1378 dbgi = get_irn_dbg_info(irn);
1382 /* we have to pop the shadow parameter in in case of struct returns */
1384 ret = be_new_Return(dbgi, irg, bl, n_res, pop, n, in);
1386 /* Set the register classes of the return's parameter accordingly. */
1387 for (i = 0; i < n; ++i) {
1388 if (regs[i] == NULL)
1391 be_node_set_reg_class_in(ret, i, regs[i]->reg_class);
1394 /* Free the space of the Epilog's in array and the register <-> proj map. */
1395 pmap_destroy(reg_map);
1400 typedef struct ent_pos_pair ent_pos_pair;
1401 struct ent_pos_pair {
1402 ir_entity *ent; /**< a value param entity */
1403 int pos; /**< its parameter number */
1404 ent_pos_pair *next; /**< for linking */
1407 typedef struct lower_frame_sels_env_t {
1408 ent_pos_pair *value_param_list; /**< the list of all value param entities */
1409 ir_node *frame; /**< the current frame */
1410 const arch_register_class_t *sp_class; /**< register class of the stack pointer */
1411 const arch_register_class_t *link_class; /**< register class of the link pointer */
1412 ir_type *value_tp; /**< the value type if any */
1413 ir_type *frame_tp; /**< the frame type */
1414 int static_link_pos; /**< argument number of the hidden static link */
1415 } lower_frame_sels_env_t;
1418 * Return an entity from the backend for an value param entity.
1420 * @param ent an value param type entity
1421 * @param ctx context
1423 static ir_entity *get_argument_entity(ir_entity *ent, lower_frame_sels_env_t *ctx)
1425 ir_entity *argument_ent = (ir_entity*)get_entity_link(ent);
1427 if (argument_ent == NULL) {
1428 /* we have NO argument entity yet: This is bad, as we will
1429 * need one for backing store.
1432 ir_type *frame_tp = ctx->frame_tp;
1433 unsigned offset = get_type_size_bytes(frame_tp);
1434 ir_type *tp = get_entity_type(ent);
1435 unsigned align = get_type_alignment_bytes(tp);
1437 offset += align - 1;
1438 offset &= ~(align - 1);
1440 argument_ent = copy_entity_own(ent, frame_tp);
1442 /* must be automatic to set a fixed layout */
1443 set_entity_offset(argument_ent, offset);
1444 offset += get_type_size_bytes(tp);
1446 set_type_size_bytes(frame_tp, offset);
1447 set_entity_link(ent, argument_ent);
1449 return argument_ent;
1452 * Walker: Replaces Sels of frame type and
1453 * value param type entities by FrameAddress.
1454 * Links all used entities.
1456 static void lower_frame_sels_walker(ir_node *irn, void *data)
1458 lower_frame_sels_env_t *ctx = (lower_frame_sels_env_t*)data;
1461 ir_node *ptr = get_Sel_ptr(irn);
1463 if (ptr == ctx->frame) {
1464 ir_entity *ent = get_Sel_entity(irn);
1465 ir_node *bl = get_nodes_block(irn);
1468 int is_value_param = 0;
1470 if (get_entity_owner(ent) == ctx->value_tp) {
1473 /* replace by its copy from the argument type */
1474 pos = get_struct_member_index(ctx->value_tp, ent);
1475 ent = get_argument_entity(ent, ctx);
1478 nw = be_new_FrameAddr(ctx->sp_class, bl, ctx->frame, ent);
1481 /* check, if it's a param Sel and if have not seen this entity before */
1482 if (is_value_param && get_entity_link(ent) == NULL) {
1488 ARR_APP1(ent_pos_pair, ctx->value_param_list, pair);
1490 set_entity_link(ent, ctx->value_param_list);
1497 * Check if a value parameter is transmitted as a register.
1498 * This might happen if the address of an parameter is taken which is
1499 * transmitted in registers.
1501 * Note that on some architectures this case must be handled specially
1502 * because the place of the backing store is determined by their ABI.
1504 * In the default case we move the entity to the frame type and create
1505 * a backing store into the first block.
1507 static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_graph *irg,
1508 ent_pos_pair *value_param_list)
1510 be_abi_call_t *call = env->call;
1511 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1512 ent_pos_pair *entry, *new_list;
1514 int i, n = ARR_LEN(value_param_list);
1517 for (i = 0; i < n; ++i) {
1518 int pos = value_param_list[i].pos;
1519 be_abi_call_arg_t *arg = get_call_arg(call, 0, pos, 1);
1522 DBG((dbg, LEVEL_2, "\targ #%d need backing store\n", pos));
1523 value_param_list[i].next = new_list;
1524 new_list = &value_param_list[i];
1527 if (new_list != NULL) {
1528 /* ok, change the graph */
1529 ir_node *start_bl = get_irg_start_block(irg);
1530 ir_node *first_bl = get_first_block_succ(start_bl);
1531 ir_node *frame, *imem, *nmem, *store, *mem, *args;
1532 optimization_state_t state;
1535 assert(first_bl && first_bl != start_bl);
1536 /* we had already removed critical edges, so the following
1537 assertion should be always true. */
1538 assert(get_Block_n_cfgpreds(first_bl) == 1);
1540 /* now create backing stores */
1541 frame = get_irg_frame(irg);
1542 imem = get_irg_initial_mem(irg);
1544 save_optimization_state(&state);
1546 nmem = new_r_Proj(get_irg_start(irg), mode_M, pn_Start_M);
1547 restore_optimization_state(&state);
1549 /* reroute all edges to the new memory source */
1550 edges_reroute(imem, nmem, irg);
1554 args = get_irg_args(irg);
1555 for (entry = new_list; entry != NULL; entry = entry->next) {
1557 ir_type *tp = get_entity_type(entry->ent);
1558 ir_mode *mode = get_type_mode(tp);
1561 /* address for the backing store */
1562 addr = be_new_FrameAddr(arch_env->sp->reg_class, first_bl, frame, entry->ent);
1565 mem = new_r_Proj(store, mode_M, pn_Store_M);
1567 /* the backing store itself */
1568 store = new_r_Store(first_bl, mem, addr,
1569 new_r_Proj(args, mode, i), cons_none);
1571 /* the new memory Proj gets the last Proj from store */
1572 set_Proj_pred(nmem, store);
1573 set_Proj_proj(nmem, pn_Store_M);
1575 /* move all entities to the frame type */
1576 frame_tp = get_irg_frame_type(irg);
1577 offset = get_type_size_bytes(frame_tp);
1579 /* we will add new entities: set the layout to undefined */
1580 assert(get_type_state(frame_tp) == layout_fixed);
1581 set_type_state(frame_tp, layout_undefined);
1582 for (entry = new_list; entry != NULL; entry = entry->next) {
1583 ir_entity *ent = entry->ent;
1585 /* If the entity is still on the argument type, move it to the
1587 * This happens if the value_param type was build due to compound
1589 if (get_entity_owner(ent) != frame_tp) {
1590 ir_type *tp = get_entity_type(ent);
1591 unsigned align = get_type_alignment_bytes(tp);
1593 offset += align - 1;
1594 offset &= ~(align - 1);
1595 set_entity_owner(ent, frame_tp);
1596 /* must be automatic to set a fixed layout */
1597 set_entity_offset(ent, offset);
1598 offset += get_type_size_bytes(tp);
1601 set_type_size_bytes(frame_tp, offset);
1602 /* fix the layout again */
1603 set_type_state(frame_tp, layout_fixed);
1608 * The start block has no jump, instead it has an initial exec Proj.
1609 * The backend wants to handle all blocks the same way, so we replace
1610 * the out cfg edge with a real jump.
1612 static void fix_start_block(ir_graph *irg)
1614 ir_node *initial_X = get_irg_initial_exec(irg);
1615 ir_node *start_block = get_irg_start_block(irg);
1616 const ir_edge_t *edge;
1618 assert(is_Proj(initial_X));
1620 foreach_out_edge(initial_X, edge) {
1621 ir_node *block = get_edge_src_irn(edge);
1623 if (is_Anchor(block))
1625 if (block != start_block) {
1626 ir_node *jmp = new_r_Jmp(start_block);
1627 set_Block_cfgpred(block, get_edge_src_pos(edge), jmp);
1628 set_irg_initial_exec(irg, jmp);
1632 panic("Initial exec has no follow block in %+F", irg);
1636 * Update the entity of Sels to the outer value parameters.
1638 static void update_outer_frame_sels(ir_node *irn, void *env)
1640 lower_frame_sels_env_t *ctx = (lower_frame_sels_env_t*)env;
1647 ptr = get_Sel_ptr(irn);
1648 if (! is_arg_Proj(ptr))
1650 if (get_Proj_proj(ptr) != ctx->static_link_pos)
1652 ent = get_Sel_entity(irn);
1654 if (get_entity_owner(ent) == ctx->value_tp) {
1655 /* replace by its copy from the argument type */
1656 pos = get_struct_member_index(ctx->value_tp, ent);
1657 ent = get_argument_entity(ent, ctx);
1658 set_Sel_entity(irn, ent);
1660 /* check, if we have not seen this entity before */
1661 if (get_entity_link(ent) == NULL) {
1667 ARR_APP1(ent_pos_pair, ctx->value_param_list, pair);
1669 set_entity_link(ent, ctx->value_param_list);
1675 * Fix access to outer local variables.
1677 static void fix_outer_variable_access(be_abi_irg_t *env,
1678 lower_frame_sels_env_t *ctx)
1684 for (i = get_class_n_members(ctx->frame_tp) - 1; i >= 0; --i) {
1685 ir_entity *ent = get_class_member(ctx->frame_tp, i);
1687 if (! is_method_entity(ent))
1690 irg = get_entity_irg(ent);
1695 * FIXME: find the number of the static link parameter
1696 * for now we assume 0 here
1698 ctx->static_link_pos = 0;
1700 irg_walk_graph(irg, NULL, update_outer_frame_sels, ctx);
1705 * Modify the irg itself and the frame type.
1707 static void modify_irg(ir_graph *irg)
1709 be_abi_irg_t *env = be_get_irg_abi(irg);
1710 be_abi_call_t *call = env->call;
1711 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1712 const arch_register_t *sp = arch_env->sp;
1713 ir_type *method_type = get_entity_type(get_irg_entity(irg));
1714 be_irg_t *birg = be_birg_from_irg(irg);
1715 struct obstack *obst = be_get_be_obst(irg);
1716 be_stack_layout_t *stack_layout = be_get_irg_stack_layout(irg);
1719 ir_node *new_mem_proj;
1725 unsigned frame_size;
1728 const arch_register_t *fp_reg;
1729 ir_node *frame_pointer;
1733 const ir_edge_t *edge;
1734 ir_type *arg_type, *bet_type, *tp;
1735 lower_frame_sels_env_t ctx;
1736 ir_entity **param_map;
1738 DBG((dbg, LEVEL_1, "introducing abi on %+F\n", irg));
1740 /* Must fetch memory here, otherwise the start Barrier gets the wrong
1741 * memory, which leads to loops in the DAG. */
1742 old_mem = get_irg_initial_mem(irg);
1744 irp_reserve_resources(irp, IR_RESOURCE_ENTITY_LINK);
1746 /* set the links of all frame entities to NULL, we use it
1747 to detect if an entity is already linked in the value_param_list */
1748 tp = get_method_value_param_type(method_type);
1751 /* clear the links of the clone type, let the
1752 original entities point to its clones */
1753 for (i = get_struct_n_members(tp) - 1; i >= 0; --i) {
1754 ir_entity *mem = get_struct_member(tp, i);
1755 set_entity_link(mem, NULL);
1759 arg_type = compute_arg_type(env, irg, call, method_type, tp, ¶m_map);
1761 /* Convert the Sel nodes in the irg to frame addr nodes: */
1762 ctx.value_param_list = NEW_ARR_F(ent_pos_pair, 0);
1763 ctx.frame = get_irg_frame(irg);
1764 ctx.sp_class = arch_env->sp->reg_class;
1765 ctx.link_class = arch_env->link_class;
1766 ctx.frame_tp = get_irg_frame_type(irg);
1768 /* layout the stackframe now */
1769 if (get_type_state(ctx.frame_tp) == layout_undefined) {
1770 default_layout_compound_type(ctx.frame_tp);
1773 /* we will possible add new entities to the frame: set the layout to undefined */
1774 assert(get_type_state(ctx.frame_tp) == layout_fixed);
1775 set_type_state(ctx.frame_tp, layout_undefined);
1777 irg_walk_graph(irg, lower_frame_sels_walker, NULL, &ctx);
1779 /* fix the frame type layout again */
1780 set_type_state(ctx.frame_tp, layout_fixed);
1781 /* align stackframe to 4 byte */
1782 frame_size = get_type_size_bytes(ctx.frame_tp);
1783 if (frame_size % 4 != 0) {
1784 set_type_size_bytes(ctx.frame_tp, frame_size + 4 - (frame_size % 4));
1787 env->regs = pmap_create();
1789 n_params = get_method_n_params(method_type);
1790 args = OALLOCNZ(obst, ir_node*, n_params);
1793 * for inner function we must now fix access to outer frame entities.
1795 fix_outer_variable_access(env, &ctx);
1797 /* Check if a value parameter is transmitted as a register.
1798 * This might happen if the address of an parameter is taken which is
1799 * transmitted in registers.
1801 * Note that on some architectures this case must be handled specially
1802 * because the place of the backing store is determined by their ABI.
1804 * In the default case we move the entity to the frame type and create
1805 * a backing store into the first block.
1807 fix_address_of_parameter_access(env, irg, ctx.value_param_list);
1809 DEL_ARR_F(ctx.value_param_list);
1810 irp_free_resources(irp, IR_RESOURCE_ENTITY_LINK);
1812 /* Fill the argument vector */
1813 arg_tuple = get_irg_args(irg);
1814 foreach_out_edge(arg_tuple, edge) {
1815 ir_node *irn = get_edge_src_irn(edge);
1816 if (! is_Anchor(irn)) {
1817 int nr = get_Proj_proj(irn);
1819 DBG((dbg, LEVEL_2, "\treading arg: %d -> %+F\n", nr, irn));
1823 bet_type = call->cb->get_between_type(env->cb);
1824 stack_frame_init(stack_layout, arg_type, bet_type,
1825 get_irg_frame_type(irg), arch_env->stack_dir, param_map);
1826 stack_layout->sp_relative = call->flags.bits.try_omit_fp;
1828 /* Count the register params and add them to the number of Projs for the RegParams node */
1829 for (i = 0; i < n_params; ++i) {
1830 be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 1);
1831 if (arg->in_reg && args[i]) {
1832 assert(arg->reg != sp && "cannot use stack pointer as parameter register");
1833 assert(i == get_Proj_proj(args[i]));
1835 /* For now, associate the register with the old Proj from Start representing that argument. */
1836 pmap_insert(env->regs, (void *) arg->reg, args[i]);
1837 DBG((dbg, LEVEL_2, "\targ #%d -> reg %s\n", i, arg->reg->name));
1841 /* Collect all callee-save registers */
1842 for (i = 0, n = arch_env->n_register_classes; i < n; ++i) {
1843 const arch_register_class_t *cls = &arch_env->register_classes[i];
1844 for (j = 0; j < cls->n_regs; ++j) {
1845 const arch_register_t *reg = &cls->regs[j];
1846 if (reg->type & (arch_register_type_callee_save | arch_register_type_state)) {
1847 pmap_insert(env->regs, (void *) reg, NULL);
1852 /* handle start block here (place a jump in the block) */
1853 fix_start_block(irg);
1855 pmap_insert(env->regs, (void *) sp, NULL);
1856 pmap_insert(env->regs, (void *) arch_env->bp, NULL);
1857 start_bl = get_irg_start_block(irg);
1858 env->start = be_new_Start(NULL, start_bl, pmap_count(env->regs) + 1);
1861 * make proj nodes for the callee save registers.
1862 * memorize them, since Return nodes get those as inputs.
1864 * Note, that if a register corresponds to an argument, the regs map contains
1865 * the old Proj from start for that argument.
1868 rm = ALLOCAN(reg_node_map_t, pmap_count(env->regs));
1869 reg_map_to_arr(rm, env->regs);
1870 for (i = 0, n = pmap_count(env->regs); i < n; ++i) {
1871 const arch_register_t *reg = rm[i].reg;
1872 ir_mode *mode = reg->reg_class->mode;
1874 arch_register_req_type_t add_type = arch_register_req_type_none;
1878 add_type |= arch_register_req_type_produces_sp | arch_register_req_type_ignore;
1881 proj = new_r_Proj(env->start, mode, nr + 1);
1882 pmap_insert(env->regs, (void *) reg, proj);
1883 be_set_constr_single_reg_out(env->start, nr + 1, reg, add_type);
1884 arch_set_irn_register(proj, reg);
1886 DBG((dbg, LEVEL_2, "\tregister save proj #%d -> reg %s\n", nr, reg->name));
1889 /* create a new initial memory proj */
1890 assert(is_Proj(old_mem));
1891 arch_set_out_register_req(env->start, 0, arch_no_register_req);
1892 new_mem_proj = new_r_Proj(env->start, mode_M, 0);
1894 set_irg_initial_mem(irg, mem);
1896 /* Generate the Prologue */
1897 fp_reg = call->cb->prologue(env->cb, &mem, env->regs, &stack_layout->initial_bias);
1899 /* do the stack allocation BEFORE the barrier, or spill code
1900 might be added before it */
1901 env->init_sp = be_abi_reg_map_get(env->regs, sp);
1902 env->init_sp = be_new_IncSP(sp, start_bl, env->init_sp, BE_STACK_FRAME_SIZE_EXPAND, 0);
1903 be_abi_reg_map_set(env->regs, sp, env->init_sp);
1905 create_barrier(start_bl, &mem, env->regs, 0);
1907 env->init_sp = be_abi_reg_map_get(env->regs, sp);
1908 arch_set_irn_register(env->init_sp, sp);
1910 frame_pointer = be_abi_reg_map_get(env->regs, fp_reg);
1911 set_irg_frame(irg, frame_pointer);
1912 rbitset_clear(birg->allocatable_regs, fp_reg->global_index);
1914 /* rewire old mem users to new mem */
1915 exchange(old_mem, mem);
1917 /* keep the mem (for functions with an endless loop = no return) */
1920 set_irg_initial_mem(irg, mem);
1922 /* Now, introduce stack param nodes for all parameters passed on the stack */
1923 for (i = 0; i < n_params; ++i) {
1924 ir_node *arg_proj = args[i];
1925 ir_node *repl = NULL;
1927 if (arg_proj != NULL) {
1928 be_abi_call_arg_t *arg;
1929 ir_type *param_type;
1930 int nr = get_Proj_proj(arg_proj);
1933 nr = MIN(nr, n_params);
1934 arg = get_call_arg(call, 0, nr, 1);
1935 param_type = get_method_param_type(method_type, nr);
1938 repl = (ir_node*)pmap_get(env->regs, arg->reg);
1939 } else if (arg->on_stack) {
1940 ir_node *addr = be_new_FrameAddr(sp->reg_class, start_bl, frame_pointer, arg->stack_ent);
1942 /* For atomic parameters which are actually used, we create a Load node. */
1943 if (is_atomic_type(param_type) && get_irn_n_edges(args[i]) > 0) {
1944 ir_mode *mode = get_type_mode(param_type);
1945 ir_mode *load_mode = arg->load_mode;
1947 ir_node *load = new_r_Load(start_bl, new_r_NoMem(irg), addr, load_mode, cons_floats);
1948 repl = new_r_Proj(load, load_mode, pn_Load_res);
1950 if (mode != load_mode) {
1951 repl = new_r_Conv(start_bl, repl, mode);
1954 /* The stack parameter is not primitive (it is a struct or array),
1955 * we thus will create a node representing the parameter's address
1961 assert(repl != NULL);
1963 /* Beware: the mode of the register parameters is always the mode of the register class
1964 which may be wrong. Add Conv's then. */
1965 mode = get_irn_mode(args[i]);
1966 if (mode != get_irn_mode(repl)) {
1967 repl = new_r_Conv(get_nodes_block(repl), repl, mode);
1969 exchange(args[i], repl);
1973 /* the arg proj is not needed anymore now and should be only used by the anchor */
1974 assert(get_irn_n_edges(arg_tuple) == 1);
1975 kill_node(arg_tuple);
1976 set_irg_args(irg, new_r_Bad(irg));
1978 /* All Return nodes hang on the End node, so look for them there. */
1979 end = get_irg_end_block(irg);
1980 for (i = 0, n = get_Block_n_cfgpreds(end); i < n; ++i) {
1981 ir_node *irn = get_Block_cfgpred(end, i);
1983 if (is_Return(irn)) {
1984 ir_node *blk = get_nodes_block(irn);
1985 ir_node *mem = get_Return_mem(irn);
1986 ir_node *ret = create_be_return(env, irn, blk, mem, get_Return_n_ress(irn));
1991 /* if we have endless loops here, n might be <= 0. Do NOT create a be_Return then,
1992 the code is dead and will never be executed. */
1995 /** Fix the state inputs of calls that still hang on unknowns */
1996 static void fix_call_state_inputs(ir_graph *irg)
1998 be_abi_irg_t *env = be_get_irg_abi(irg);
1999 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
2001 arch_register_t **stateregs = NEW_ARR_F(arch_register_t*, 0);
2003 /* Collect caller save registers */
2004 n = arch_env->n_register_classes;
2005 for (i = 0; i < n; ++i) {
2007 const arch_register_class_t *cls = &arch_env->register_classes[i];
2008 for (j = 0; j < cls->n_regs; ++j) {
2009 const arch_register_t *reg = arch_register_for_index(cls, j);
2010 if (reg->type & arch_register_type_state) {
2011 ARR_APP1(arch_register_t*, stateregs, (arch_register_t *)reg);
2016 n = ARR_LEN(env->calls);
2017 n_states = ARR_LEN(stateregs);
2018 for (i = 0; i < n; ++i) {
2020 ir_node *call = env->calls[i];
2022 arity = get_irn_arity(call);
2024 /* the state reg inputs are the last n inputs of the calls */
2025 for (s = 0; s < n_states; ++s) {
2026 int inp = arity - n_states + s;
2027 const arch_register_t *reg = stateregs[s];
2028 ir_node *regnode = be_abi_reg_map_get(env->regs, reg);
2030 set_irn_n(call, inp, regnode);
2034 DEL_ARR_F(stateregs);
2038 * Create a trampoline entity for the given method.
2040 static ir_entity *create_trampoline(be_main_env_t *be, ir_entity *method)
2042 ir_type *type = get_entity_type(method);
2043 ident *old_id = get_entity_ld_ident(method);
2044 ident *id = id_mangle3("", old_id, "$stub");
2045 ir_type *parent = be->pic_trampolines_type;
2046 ir_entity *ent = new_entity(parent, old_id, type);
2047 set_entity_ld_ident(ent, id);
2048 set_entity_visibility(ent, ir_visibility_private);
2054 * Returns the trampoline entity for the given method.
2056 static ir_entity *get_trampoline(be_main_env_t *env, ir_entity *method)
2058 ir_entity *result = (ir_entity*)pmap_get(env->ent_trampoline_map, method);
2059 if (result == NULL) {
2060 result = create_trampoline(env, method);
2061 pmap_insert(env->ent_trampoline_map, method, result);
2067 static ir_entity *create_pic_symbol(be_main_env_t *be, ir_entity *entity)
2069 ident *old_id = get_entity_ld_ident(entity);
2070 ident *id = id_mangle3("", old_id, "$non_lazy_ptr");
2071 ir_type *e_type = get_entity_type(entity);
2072 ir_type *type = new_type_pointer(e_type);
2073 ir_type *parent = be->pic_symbols_type;
2074 ir_entity *ent = new_entity(parent, old_id, type);
2075 set_entity_ld_ident(ent, id);
2076 set_entity_visibility(ent, ir_visibility_private);
2081 static ir_entity *get_pic_symbol(be_main_env_t *env, ir_entity *entity)
2083 ir_entity *result = (ir_entity*)pmap_get(env->ent_pic_symbol_map, entity);
2084 if (result == NULL) {
2085 result = create_pic_symbol(env, entity);
2086 pmap_insert(env->ent_pic_symbol_map, entity, result);
2095 * Returns non-zero if a given entity can be accessed using a relative address.
2097 static int can_address_relative(ir_entity *entity)
2099 return get_entity_visibility(entity) != ir_visibility_external
2100 && !(get_entity_linkage(entity) & IR_LINKAGE_MERGE);
2103 static ir_node *get_pic_base(ir_graph *irg)
2105 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
2106 if (arch_env->impl->get_pic_base == NULL)
2108 return arch_env->impl->get_pic_base(irg);
2111 /** patches SymConsts to work in position independent code */
2112 static void fix_pic_symconsts(ir_node *node, void *data)
2114 ir_graph *irg = get_irn_irg(node);
2115 be_main_env_t *be = be_get_irg_main_env(irg);
2125 arity = get_irn_arity(node);
2126 for (i = 0; i < arity; ++i) {
2128 ir_node *pred = get_irn_n(node, i);
2130 ir_entity *pic_symbol;
2131 ir_node *pic_symconst;
2133 if (!is_SymConst(pred))
2136 entity = get_SymConst_entity(pred);
2137 block = get_nodes_block(pred);
2139 /* calls can jump to relative addresses, so we can directly jump to
2140 the (relatively) known call address or the trampoline */
2141 if (i == 1 && is_Call(node)) {
2142 ir_entity *trampoline;
2143 ir_node *trampoline_const;
2145 if (can_address_relative(entity))
2148 dbgi = get_irn_dbg_info(pred);
2149 trampoline = get_trampoline(be, entity);
2150 trampoline_const = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
2152 set_irn_n(node, i, trampoline_const);
2156 /* everything else is accessed relative to EIP */
2157 mode = get_irn_mode(pred);
2158 pic_base = get_pic_base(irg);
2160 /* all ok now for locally constructed stuff */
2161 if (can_address_relative(entity)) {
2162 ir_node *add = new_r_Add(block, pic_base, pred, mode);
2164 /* make sure the walker doesn't visit this add again */
2165 mark_irn_visited(add);
2166 set_irn_n(node, i, add);
2170 /* get entry from pic symbol segment */
2171 dbgi = get_irn_dbg_info(pred);
2172 pic_symbol = get_pic_symbol(be, entity);
2173 pic_symconst = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
2175 add = new_r_Add(block, pic_base, pic_symconst, mode);
2176 mark_irn_visited(add);
2178 /* we need an extra indirection for global data outside our current
2179 module. The loads are always safe and can therefore float
2180 and need no memory input */
2181 load = new_r_Load(block, new_r_NoMem(irg), add, mode, cons_floats);
2182 load_res = new_r_Proj(load, mode, pn_Load_res);
2184 set_irn_n(node, i, load_res);
2188 be_abi_irg_t *be_abi_introduce(ir_graph *irg)
2190 be_abi_irg_t *env = XMALLOCZ(be_abi_irg_t);
2191 ir_node *old_frame = get_irg_frame(irg);
2192 be_options_t *options = be_get_irg_options(irg);
2193 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
2194 ir_entity *entity = get_irg_entity(irg);
2195 ir_type *method_type = get_entity_type(entity);
2196 be_irg_t *birg = be_birg_from_irg(irg);
2197 struct obstack *obst = &birg->obst;
2203 /* determine allocatable registers */
2204 assert(birg->allocatable_regs == NULL);
2205 birg->allocatable_regs = rbitset_obstack_alloc(obst, arch_env->n_registers);
2206 for (r = 0; r < arch_env->n_registers; ++r) {
2207 const arch_register_t *reg = &arch_env->registers[r];
2208 if ( !(reg->type & arch_register_type_ignore)) {
2209 rbitset_set(birg->allocatable_regs, r);
2213 /* break here if backend provides a custom API.
2214 * Note: we shouldn't have to setup any be_abi_irg_t* stuff at all,
2215 * but need more cleanup to make this work
2217 be_set_irg_abi(irg, env);
2219 be_omit_fp = options->omit_fp;
2221 env->dce_survivor = new_survive_dce();
2222 env->keep_map = pmap_create();
2223 env->call = be_abi_call_new(arch_env->sp->reg_class);
2224 arch_env_get_call_abi(arch_env, method_type, env->call);
2226 env->init_sp = dummy = new_r_Dummy(irg, arch_env->sp->reg_class->mode);
2227 env->calls = NEW_ARR_F(ir_node*, 0);
2230 irg_walk_graph(irg, fix_pic_symconsts, NULL, env);
2233 /* Lower all call nodes in the IRG. */
2237 Beware: init backend abi call object after processing calls,
2238 otherwise some information might be not yet available.
2240 env->cb = env->call->cb->init(env->call, irg);
2242 /* Process the IRG */
2245 /* fix call inputs for state registers */
2246 fix_call_state_inputs(irg);
2248 /* We don't need the keep map anymore. */
2249 pmap_destroy(env->keep_map);
2250 env->keep_map = NULL;
2252 /* calls array is not needed anymore */
2253 DEL_ARR_F(env->calls);
2256 /* reroute the stack origin of the calls to the true stack origin. */
2257 exchange(dummy, env->init_sp);
2258 exchange(old_frame, get_irg_frame(irg));
2260 /* Make some important node pointers survive the dead node elimination. */
2261 survive_dce_register_irn(env->dce_survivor, &env->init_sp);
2262 foreach_pmap(env->regs, ent) {
2263 survive_dce_register_irn(env->dce_survivor, (ir_node **) &ent->value);
2266 env->call->cb->done(env->cb);
2271 void be_abi_free(ir_graph *irg)
2273 be_abi_irg_t *env = be_get_irg_abi(irg);
2275 if (env->call != NULL)
2276 be_abi_call_free(env->call);
2277 if (env->dce_survivor != NULL)
2278 free_survive_dce(env->dce_survivor);
2279 if (env->regs != NULL)
2280 pmap_destroy(env->regs);
2283 be_set_irg_abi(irg, NULL);
2286 void be_put_allocatable_regs(const ir_graph *irg,
2287 const arch_register_class_t *cls, bitset_t *bs)
2289 be_irg_t *birg = be_birg_from_irg(irg);
2290 unsigned *allocatable_regs = birg->allocatable_regs;
2293 assert(bitset_size(bs) == cls->n_regs);
2294 bitset_clear_all(bs);
2295 for (i = 0; i < cls->n_regs; ++i) {
2296 const arch_register_t *reg = &cls->regs[i];
2297 if (rbitset_is_set(allocatable_regs, reg->global_index))
2302 unsigned be_get_n_allocatable_regs(const ir_graph *irg,
2303 const arch_register_class_t *cls)
2305 bitset_t *bs = bitset_alloca(cls->n_regs);
2306 be_put_allocatable_regs(irg, cls, bs);
2307 return bitset_popcount(bs);
2310 void be_set_allocatable_regs(const ir_graph *irg,
2311 const arch_register_class_t *cls,
2312 unsigned *raw_bitset)
2314 be_irg_t *birg = be_birg_from_irg(irg);
2315 unsigned *allocatable_regs = birg->allocatable_regs;
2318 rbitset_clear_all(raw_bitset, cls->n_regs);
2319 for (i = 0; i < cls->n_regs; ++i) {
2320 const arch_register_t *reg = &cls->regs[i];
2321 if (rbitset_is_set(allocatable_regs, reg->global_index))
2322 rbitset_set(raw_bitset, i);
2326 ir_node *be_abi_get_callee_save_irn(be_abi_irg_t *abi, const arch_register_t *reg)
2328 assert(reg->type & arch_register_type_callee_save);
2329 assert(pmap_contains(abi->regs, (void *) reg));
2330 return (ir_node*)pmap_get(abi->regs, (void *) reg);
2333 ir_node *be_abi_get_ignore_irn(be_abi_irg_t *abi, const arch_register_t *reg)
2335 assert(reg->type & arch_register_type_ignore);
2336 assert(pmap_contains(abi->regs, (void *) reg));
2337 return (ir_node*)pmap_get(abi->regs, (void *) reg);
2340 BE_REGISTER_MODULE_CONSTRUCTOR(be_init_abi);
2341 void be_init_abi(void)
2343 FIRM_DBG_REGISTER(dbg, "firm.be.abi");