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 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 be_abi_call_t *call; /**< The ABI call information. */
91 ir_node *init_sp; /**< The node representing the stack pointer
92 at the start of the function. */
94 ir_node *start; /**< The be_Start params node. */
95 pmap *regs; /**< A map of all callee-save and ignore regs to
96 their Projs to the RegParams node. */
97 int start_block_bias; /**< The stack bias at the end of the start block. */
99 pmap *keep_map; /**< mapping blocks to keep nodes. */
101 ir_node **calls; /**< flexible array containing all be_Call nodes */
104 static ir_heights_t *ir_heights;
106 /** Flag: if set, try to omit the frame pointer in all routines. */
107 static int be_omit_fp = 1;
109 static ir_node *be_abi_reg_map_get(pmap *map, const arch_register_t *reg)
111 return (ir_node*)pmap_get(map, reg);
114 static void be_abi_reg_map_set(pmap *map, const arch_register_t* reg,
117 pmap_insert(map, reg, node);
121 * Check if the given register is callee save, ie. will be saved by the callee.
123 static bool arch_register_is_callee_save(
124 const arch_env_t *arch_env,
125 const arch_register_t *reg)
127 if (arch_env->impl->register_saved_by)
128 return arch_env->impl->register_saved_by(reg, /*callee=*/1);
133 * Check if the given register is caller save, ie. must be saved by the caller.
135 static bool arch_register_is_caller_save(
136 const arch_env_t *arch_env,
137 const arch_register_t *reg)
139 if (arch_env->impl->register_saved_by)
140 return arch_env->impl->register_saved_by(reg, /*callee=*/0);
147 _ ____ ___ ____ _ _ _ _
148 / \ | __ )_ _| / ___|__ _| | | |__ __ _ ___| | _____
149 / _ \ | _ \| | | | / _` | | | '_ \ / _` |/ __| |/ / __|
150 / ___ \| |_) | | | |__| (_| | | | |_) | (_| | (__| <\__ \
151 /_/ \_\____/___| \____\__,_|_|_|_.__/ \__,_|\___|_|\_\___/
153 These callbacks are used by the backend to set the parameters
154 for a specific call type.
158 * Set compare function: compares two ABI call object arguments.
160 static int cmp_call_arg(const void *a, const void *b, size_t n)
162 const be_abi_call_arg_t *p = (const be_abi_call_arg_t*)a;
163 const be_abi_call_arg_t *q = (const be_abi_call_arg_t*)b;
165 return !(p->is_res == q->is_res && p->pos == q->pos && p->callee == q->callee);
169 * Get an ABI call object argument.
171 * @param call the abi call
172 * @param is_res true for call results, false for call arguments
173 * @param pos position of the argument
174 * @param callee context type - if we are callee or caller
176 static be_abi_call_arg_t *get_call_arg(be_abi_call_t *call, int is_res, int pos, int callee)
178 be_abi_call_arg_t arg;
181 memset(&arg, 0, sizeof(arg));
186 hash = is_res * 128 + pos;
188 return (be_abi_call_arg_t*)set_find(call->params, &arg, sizeof(arg), hash);
192 * Set an ABI call object argument.
194 static void remember_call_arg(be_abi_call_arg_t *arg, be_abi_call_t *call, be_abi_context_t context)
196 unsigned hash = arg->is_res * 128 + arg->pos;
197 if (context & ABI_CONTEXT_CALLEE) {
199 set_insert(call->params, arg, sizeof(*arg), hash);
201 if (context & ABI_CONTEXT_CALLER) {
203 set_insert(call->params, arg, sizeof(*arg), hash);
207 /* Set the flags for a call. */
208 void be_abi_call_set_flags(be_abi_call_t *call, be_abi_call_flags_t flags, const be_abi_callbacks_t *cb)
214 /* Sets the number of bytes the stackframe is shrinked by the callee on return */
215 void be_abi_call_set_pop(be_abi_call_t *call, int pop)
221 /* Set register class for call address */
222 void be_abi_call_set_call_address_reg_class(be_abi_call_t *call, const arch_register_class_t *cls)
224 call->cls_addr = cls;
228 void be_abi_call_param_stack(be_abi_call_t *call, int arg_pos,
229 ir_mode *load_mode, unsigned alignment,
230 unsigned space_before, unsigned space_after,
231 be_abi_context_t context)
233 be_abi_call_arg_t arg;
234 memset(&arg, 0, sizeof(arg));
235 assert(alignment > 0 && "Alignment must be greater than 0");
237 arg.load_mode = load_mode;
238 arg.alignment = alignment;
239 arg.space_before = space_before;
240 arg.space_after = space_after;
244 remember_call_arg(&arg, call, context);
247 void be_abi_call_param_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
249 be_abi_call_arg_t arg;
250 memset(&arg, 0, sizeof(arg));
257 remember_call_arg(&arg, call, context);
260 void be_abi_call_res_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
262 be_abi_call_arg_t arg;
263 memset(&arg, 0, sizeof(arg));
270 remember_call_arg(&arg, call, context);
273 /* Get the flags of a ABI call object. */
274 be_abi_call_flags_t be_abi_call_get_flags(const be_abi_call_t *call)
280 * Constructor for a new ABI call object.
282 * @param cls_addr register class of the call address
284 * @return the new ABI call object
286 static be_abi_call_t *be_abi_call_new(const arch_register_class_t *cls_addr)
288 be_abi_call_t *call = XMALLOCZ(be_abi_call_t);
291 call->params = new_set(cmp_call_arg, 16);
293 call->cls_addr = cls_addr;
295 call->flags.bits.try_omit_fp = be_omit_fp;
301 * Destructor for an ABI call object.
303 static void be_abi_call_free(be_abi_call_t *call)
305 del_set(call->params);
310 * Initializes the frame layout from parts
312 * @param frame the stack layout that will be initialized
313 * @param args the stack argument layout type
314 * @param between the between layout type
315 * @param locals the method frame type
316 * @param param_map an array mapping method argument positions to the stack argument type
318 * @return the initialized stack layout
320 static be_stack_layout_t *stack_frame_init(be_stack_layout_t *frame, ir_type *args,
321 ir_type *between, ir_type *locals,
322 ir_entity *param_map[])
324 frame->arg_type = args;
325 frame->between_type = between;
326 frame->frame_type = locals;
327 frame->initial_offset = 0;
328 frame->initial_bias = 0;
329 frame->order[1] = between;
330 frame->param_map = param_map;
332 /* typical decreasing stack: locals have the
333 * lowest addresses, arguments the highest */
334 frame->order[0] = locals;
335 frame->order[2] = args;
346 Adjustment of the calls inside a graph.
351 * Transform a call node into a be_Call node.
353 * @param env The ABI environment for the current irg.
354 * @param irn The call node.
355 * @param curr_sp The stack pointer node to use.
356 * @return The stack pointer after the call.
358 static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp)
360 ir_graph *irg = get_irn_irg(irn);
361 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
362 ir_type *call_tp = get_Call_type(irn);
363 ir_node *call_ptr = get_Call_ptr(irn);
364 size_t n_params = get_method_n_params(call_tp);
365 ir_node *curr_mem = get_Call_mem(irn);
366 ir_node *bl = get_nodes_block(irn);
368 const arch_register_t *sp = arch_env->sp;
369 be_abi_call_t *call = be_abi_call_new(sp->reg_class);
370 ir_mode *mach_mode = sp->reg_class->mode;
371 int no_alloc = call->flags.bits.frame_is_setup_on_call;
372 int n_res = get_method_n_ress(call_tp);
373 int do_seq = call->flags.bits.store_args_sequential && !no_alloc;
375 ir_node *res_proj = NULL;
376 int n_reg_params = 0;
377 int n_stack_params = 0;
380 const arch_register_t **states = NEW_ARR_F(const arch_register_t*, 0);
381 const arch_register_t **destroyed_regs = NEW_ARR_F(const arch_register_t*, 0);
385 int n_reg_results = 0;
386 const ir_edge_t *edge;
388 int *stack_param_idx;
390 int throws_exception;
395 /* Let the isa fill out the abi description for that call node. */
396 arch_env_get_call_abi(arch_env, call_tp, call);
398 /* Insert code to put the stack arguments on the stack. */
399 assert(get_Call_n_params(irn) == n_params);
400 stack_param_idx = ALLOCAN(int, n_params);
401 for (p = 0; p < n_params; ++p) {
402 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
405 int arg_size = get_type_size_bytes(get_method_param_type(call_tp, p));
407 stack_size += round_up2(arg->space_before, arg->alignment);
408 stack_size += round_up2(arg_size, arg->alignment);
409 stack_size += round_up2(arg->space_after, arg->alignment);
411 stack_param_idx[n_stack_params++] = p;
415 /* Collect all arguments which are passed in registers. */
416 reg_param_idxs = ALLOCAN(int, n_params);
417 for (p = 0; p < n_params; ++p) {
418 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
419 if (arg && arg->in_reg) {
420 reg_param_idxs[n_reg_params++] = p;
425 * If the stack is decreasing and we do not want to store sequentially,
426 * or someone else allocated the call frame
427 * we allocate as much space on the stack all parameters need, by
428 * moving the stack pointer along the stack's direction.
430 * Note: we also have to do this for stack_size == 0, because we may have
431 * to adjust stack alignment for the call.
433 if (!do_seq && !no_alloc) {
434 curr_sp = be_new_IncSP(sp, bl, curr_sp, stack_size, 1);
437 dbgi = get_irn_dbg_info(irn);
438 /* If there are some parameters which shall be passed on the stack. */
439 if (n_stack_params > 0) {
441 ir_node **in = ALLOCAN(ir_node*, n_stack_params+1);
444 /* push params in reverse direction because stack grows downwards */
446 for (i = 0; i < n_stack_params >> 1; ++i) {
447 int other = n_stack_params - i - 1;
448 int tmp = stack_param_idx[i];
449 stack_param_idx[i] = stack_param_idx[other];
450 stack_param_idx[other] = tmp;
454 curr_mem = get_Call_mem(irn);
456 in[n_in++] = curr_mem;
459 for (i = 0; i < n_stack_params; ++i) {
460 int p = stack_param_idx[i];
461 be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
462 ir_node *param = get_Call_param(irn, p);
463 ir_node *addr = curr_sp;
465 ir_type *param_type = get_method_param_type(call_tp, p);
466 int param_size = get_type_size_bytes(param_type) + arg->space_after;
469 * If we wanted to build the arguments sequentially,
470 * the stack pointer for the next must be incremented,
471 * and the memory value propagated.
475 addr = curr_sp = be_new_IncSP(sp, bl, curr_sp,
476 param_size + arg->space_before, 0);
477 add_irn_dep(curr_sp, curr_mem);
479 curr_ofs += arg->space_before;
480 curr_ofs = round_up2(curr_ofs, arg->alignment);
482 /* Make the expression to compute the argument's offset. */
484 ir_mode *constmode = mach_mode;
485 if (mode_is_reference(mach_mode)) {
488 addr = new_r_Const_long(irg, constmode, curr_ofs);
489 addr = new_r_Add(bl, curr_sp, addr, mach_mode);
493 /* Insert a store for primitive arguments. */
494 if (is_atomic_type(param_type)) {
495 ir_node *nomem = get_irg_no_mem(irg);
496 ir_node *mem_input = do_seq ? curr_mem : nomem;
497 ir_node *store = new_rd_Store(dbgi, bl, mem_input, addr, param, cons_none);
498 mem = new_r_Proj(store, mode_M, pn_Store_M);
500 /* Make a mem copy for compound arguments. */
503 assert(mode_is_reference(get_irn_mode(param)));
504 copy = new_rd_CopyB(dbgi, bl, curr_mem, addr, param, param_type);
505 mem = new_r_Proj(copy, mode_M, pn_CopyB_M);
508 curr_ofs += param_size;
516 /* We need the sync only, if we didn't build the stores sequentially. */
518 if (n_stack_params >= 1) {
519 curr_mem = new_r_Sync(bl, n_in, in);
521 curr_mem = get_Call_mem(irn);
526 /* Put caller save into the destroyed set and state registers in the states
528 for (i = 0, n = arch_env->n_register_classes; i < n; ++i) {
530 const arch_register_class_t *cls = &arch_env->register_classes[i];
531 for (j = 0; j < cls->n_regs; ++j) {
532 const arch_register_t *reg = arch_register_for_index(cls, j);
534 /* even if destroyed all is specified, neither SP nor FP are
535 * destroyed (else bad things will happen) */
536 if (reg == arch_env->sp || reg == arch_env->bp)
539 if (reg->type & arch_register_type_state) {
540 ARR_APP1(const arch_register_t*, destroyed_regs, reg);
541 ARR_APP1(const arch_register_t*, states, reg);
542 /* we're already in the destroyed set so no need for further
546 if (arch_register_is_caller_save(arch_env, reg)) {
547 if (!(reg->type & arch_register_type_ignore)) {
548 ARR_APP1(const arch_register_t*, destroyed_regs, reg);
554 /* search the largest result proj number */
555 res_projs = ALLOCANZ(ir_node*, n_res);
557 foreach_out_edge(irn, edge) {
558 const ir_edge_t *res_edge;
559 ir_node *irn = get_edge_src_irn(edge);
561 if (!is_Proj(irn) || get_Proj_proj(irn) != pn_Call_T_result)
564 foreach_out_edge(irn, res_edge) {
566 ir_node *res = get_edge_src_irn(res_edge);
568 assert(is_Proj(res));
570 proj = get_Proj_proj(res);
571 assert(proj < n_res);
572 assert(res_projs[proj] == NULL);
573 res_projs[proj] = res;
579 /** TODO: this is not correct for cases where return values are passed
580 * on the stack, but no known ABI does this currently...
582 n_reg_results = n_res;
585 in = ALLOCAN(ir_node*, n_reg_params + ARR_LEN(states));
587 /* make the back end call node and set its register requirements. */
588 for (i = 0; i < n_reg_params; ++i) {
589 in[n_ins++] = get_Call_param(irn, reg_param_idxs[i]);
592 /* add state registers ins */
593 for (s = 0; s < ARR_LEN(states); ++s) {
594 const arch_register_t *reg = states[s];
595 const arch_register_class_t *cls = reg->reg_class;
596 ir_node *regnode = new_r_Unknown(irg, cls->mode);
597 in[n_ins++] = regnode;
599 assert(n_ins == (int) (n_reg_params + ARR_LEN(states)));
601 /* ins collected, build the call */
602 throws_exception = ir_throws_exception(irn);
603 if (env->call->flags.bits.call_has_imm && is_SymConst(call_ptr)) {
605 low_call = be_new_Call(dbgi, irg, bl, curr_mem, curr_sp, curr_sp,
606 n_reg_results + pn_be_Call_first_res + ARR_LEN(destroyed_regs),
607 n_ins, in, get_Call_type(irn));
608 be_Call_set_entity(low_call, get_SymConst_entity(call_ptr));
611 low_call = be_new_Call(dbgi, irg, bl, curr_mem, curr_sp, call_ptr,
612 n_reg_results + pn_be_Call_first_res + ARR_LEN(destroyed_regs),
613 n_ins, in, get_Call_type(irn));
615 ir_set_throws_exception(low_call, throws_exception);
616 be_Call_set_pop(low_call, call->pop);
618 /* put the call into the list of all calls for later processing */
619 ARR_APP1(ir_node *, env->calls, low_call);
621 /* create new stack pointer */
622 curr_sp = new_r_Proj(low_call, get_irn_mode(curr_sp), pn_be_Call_sp);
623 be_set_constr_single_reg_out(low_call, pn_be_Call_sp, sp,
624 arch_register_req_type_ignore | arch_register_req_type_produces_sp);
625 arch_set_irn_register(curr_sp, sp);
627 /* now handle results */
628 for (i = 0; i < n_res; ++i) {
629 ir_node *proj = res_projs[i];
630 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 0);
631 long pn = i + pn_be_Call_first_res;
633 /* returns values on stack not supported yet */
637 shift the proj number to the right, since we will drop the
638 unspeakable Proj_T from the Call. Therefore, all real argument
639 Proj numbers must be increased by pn_be_Call_first_res
641 pn = i + pn_be_Call_first_res;
644 ir_type *res_type = get_method_res_type(call_tp, i);
645 ir_mode *mode = get_type_mode(res_type);
646 proj = new_r_Proj(low_call, mode, pn);
649 set_Proj_pred(proj, low_call);
650 set_Proj_proj(proj, pn);
654 /* remove register from destroyed regs */
656 size_t n = ARR_LEN(destroyed_regs);
657 for (j = 0; j < n; ++j) {
658 if (destroyed_regs[j] == arg->reg) {
659 destroyed_regs[j] = destroyed_regs[n-1];
660 ARR_SHRINKLEN(destroyed_regs,n-1);
668 Set the register class of the call address to
669 the backend provided class (default: stack pointer class)
671 be_node_set_reg_class_in(low_call, n_be_Call_ptr, call->cls_addr);
673 DBG((dbg, LEVEL_3, "\tcreated backend call %+F\n", low_call));
675 /* Set the register classes and constraints of the Call parameters. */
676 for (i = 0; i < n_reg_params; ++i) {
677 int index = reg_param_idxs[i];
678 be_abi_call_arg_t *arg = get_call_arg(call, 0, index, 0);
679 assert(arg->reg != NULL);
681 be_set_constr_single_reg_in(low_call, n_be_Call_first_arg + i,
682 arg->reg, arch_register_req_type_none);
685 /* Set the register constraints of the results. */
686 for (i = 0; i < n_res; ++i) {
687 ir_node *proj = res_projs[i];
688 const be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 0);
689 int pn = get_Proj_proj(proj);
692 be_set_constr_single_reg_out(low_call, pn, arg->reg,
693 arch_register_req_type_none);
694 arch_set_irn_register(proj, arg->reg);
696 exchange(irn, low_call);
698 /* kill the ProjT node */
699 if (res_proj != NULL) {
703 /* Make additional projs for the caller save registers
704 and the Keep node which keeps them alive. */
710 int curr_res_proj = pn_be_Call_first_res + n_reg_results;
713 n_ins = ARR_LEN(destroyed_regs) + n_reg_results + 1;
714 in = ALLOCAN(ir_node *, n_ins);
716 /* also keep the stack pointer */
717 set_irn_link(curr_sp, (void*) sp);
720 for (d = 0; d < ARR_LEN(destroyed_regs); ++d) {
721 const arch_register_t *reg = destroyed_regs[d];
722 ir_node *proj = new_r_Proj(low_call, reg->reg_class->mode, curr_res_proj);
724 /* memorize the register in the link field. we need afterwards to set the register class of the keep correctly. */
725 be_set_constr_single_reg_out(low_call, curr_res_proj, reg,
726 arch_register_req_type_none);
727 arch_set_irn_register(proj, reg);
729 set_irn_link(proj, (void*) reg);
734 for (i = 0; i < n_reg_results; ++i) {
735 ir_node *proj = res_projs[i];
736 const arch_register_t *reg = arch_get_irn_register(proj);
737 set_irn_link(proj, (void*) reg);
742 /* create the Keep for the caller save registers */
743 keep = be_new_Keep(bl, n, in);
744 for (i = 0; i < n; ++i) {
745 const arch_register_t *reg = (const arch_register_t*)get_irn_link(in[i]);
746 be_node_set_reg_class_in(keep, i, arch_register_get_class(reg));
750 /* Clean up the stack. */
751 assert(stack_size >= call->pop);
752 stack_size -= call->pop;
754 if (stack_size > 0) {
755 ir_node *mem_proj = NULL;
757 foreach_out_edge(low_call, edge) {
758 ir_node *irn = get_edge_src_irn(edge);
759 if (is_Proj(irn) && get_Proj_proj(irn) == pn_Call_M) {
766 mem_proj = new_r_Proj(low_call, mode_M, pn_be_Call_M);
767 keep_alive(mem_proj);
770 /* Clean up the stack frame or revert alignment fixes if we allocated it */
772 curr_sp = be_new_IncSP(sp, bl, curr_sp, -stack_size, 0);
775 be_abi_call_free(call);
778 DEL_ARR_F(destroyed_regs);
784 * Adjust the size of a node representing a stack alloc or free for the minimum stack alignment.
786 * @param alignment the minimum stack alignment
787 * @param size the node containing the non-aligned size
788 * @param block the block where new nodes are allocated on
789 * @param dbg debug info for new nodes
791 * @return a node representing the aligned size
793 static ir_node *adjust_alloc_size(unsigned stack_alignment, ir_node *size,
794 ir_node *block, dbg_info *dbg)
796 if (stack_alignment > 1) {
802 assert(is_po2(stack_alignment));
804 mode = get_irn_mode(size);
805 tv = new_tarval_from_long(stack_alignment-1, mode);
806 irg = get_Block_irg(block);
807 mask = new_r_Const(irg, tv);
808 size = new_rd_Add(dbg, block, size, mask, mode);
810 tv = new_tarval_from_long(-(long)stack_alignment, mode);
811 mask = new_r_Const(irg, tv);
812 size = new_rd_And(dbg, block, size, mask, mode);
818 * The alloca is transformed into a back end alloca node and connected to the stack nodes.
820 static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp)
822 ir_node *block = get_nodes_block(alloc);
823 ir_graph *irg = get_Block_irg(block);
824 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
825 ir_node *alloc_mem = NULL;
826 ir_node *alloc_res = NULL;
827 ir_type *type = get_Alloc_type(alloc);
830 const ir_edge_t *edge;
835 unsigned stack_alignment;
837 /* all non-stack Alloc nodes should already be lowered before the backend */
838 assert(get_Alloc_where(alloc) == stack_alloc);
840 foreach_out_edge(alloc, edge) {
841 ir_node *irn = get_edge_src_irn(edge);
843 assert(is_Proj(irn));
844 switch (get_Proj_proj(irn)) {
856 /* Beware: currently Alloc nodes without a result might happen,
857 only escape analysis kills them and this phase runs only for object
858 oriented source. We kill the Alloc here. */
859 if (alloc_res == NULL && alloc_mem) {
860 exchange(alloc_mem, get_Alloc_mem(alloc));
864 dbg = get_irn_dbg_info(alloc);
865 count = get_Alloc_count(alloc);
867 /* we might need to multiply the count with the element size */
868 if (!is_unknown_type(type) && get_type_size_bytes(type) != 1) {
869 ir_mode *mode = get_irn_mode(count);
870 ir_tarval *tv = new_tarval_from_long(get_type_size_bytes(type),
872 ir_node *cnst = new_rd_Const(dbg, irg, tv);
873 size = new_rd_Mul(dbg, block, count, cnst, mode);
878 /* The stack pointer will be modified in an unknown manner.
879 We cannot omit it. */
880 env->call->flags.bits.try_omit_fp = 0;
882 stack_alignment = 1 << arch_env->stack_alignment;
883 size = adjust_alloc_size(stack_alignment, size, block, dbg);
884 new_alloc = be_new_AddSP(arch_env->sp, block, curr_sp, size);
885 set_irn_dbg_info(new_alloc, dbg);
887 if (alloc_mem != NULL) {
891 addsp_mem = new_r_Proj(new_alloc, mode_M, pn_be_AddSP_M);
893 /* We need to sync the output mem of the AddSP with the input mem
894 edge into the alloc node. */
895 ins[0] = get_Alloc_mem(alloc);
897 sync = new_r_Sync(block, 2, ins);
899 exchange(alloc_mem, sync);
902 exchange(alloc, new_alloc);
904 /* fix projnum of alloca res */
905 set_Proj_proj(alloc_res, pn_be_AddSP_res);
907 curr_sp = new_r_Proj(new_alloc, get_irn_mode(curr_sp), pn_be_AddSP_sp);
914 * The Free is transformed into a back end free node and connected to the stack nodes.
916 static ir_node *adjust_free(be_abi_irg_t *env, ir_node *free, ir_node *curr_sp)
918 ir_node *block = get_nodes_block(free);
919 ir_graph *irg = get_irn_irg(free);
920 ir_type *type = get_Free_type(free);
921 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
922 ir_mode *sp_mode = arch_env->sp->reg_class->mode;
923 dbg_info *dbg = get_irn_dbg_info(free);
924 ir_node *subsp, *mem, *res, *size, *sync;
926 unsigned stack_alignment;
928 /* all non-stack-alloc Free nodes should already be lowered before the
930 assert(get_Free_where(free) == stack_alloc);
932 /* we might need to multiply the size with the element size */
933 if (!is_unknown_type(type) && get_type_size_bytes(type) != 1) {
934 ir_tarval *tv = new_tarval_from_long(get_type_size_bytes(type), mode_Iu);
935 ir_node *cnst = new_rd_Const(dbg, irg, tv);
936 ir_node *mul = new_rd_Mul(dbg, block, get_Free_count(free),
940 size = get_Free_count(free);
943 stack_alignment = 1 << arch_env->stack_alignment;
944 size = adjust_alloc_size(stack_alignment, size, block, dbg);
946 /* The stack pointer will be modified in an unknown manner.
947 We cannot omit it. */
948 env->call->flags.bits.try_omit_fp = 0;
949 subsp = be_new_SubSP(arch_env->sp, block, curr_sp, size);
950 set_irn_dbg_info(subsp, dbg);
952 mem = new_r_Proj(subsp, mode_M, pn_be_SubSP_M);
953 res = new_r_Proj(subsp, sp_mode, pn_be_SubSP_sp);
955 /* we need to sync the memory */
956 in[0] = get_Free_mem(free);
958 sync = new_r_Sync(block, 2, in);
960 /* and make the AddSP dependent on the former memory */
961 add_irn_dep(subsp, get_Free_mem(free));
964 exchange(free, sync);
971 * Check if a node is somehow data dependent on another one.
972 * both nodes must be in the same basic block.
973 * @param n1 The first node.
974 * @param n2 The second node.
975 * @return 1, if n1 is data dependent (transitively) on n2, 0 if not.
977 static int dependent_on(ir_node *n1, ir_node *n2)
979 assert(get_nodes_block(n1) == get_nodes_block(n2));
981 return heights_reachable_in_block(ir_heights, n1, n2);
984 static int cmp_call_dependency(const void *c1, const void *c2)
986 ir_node *n1 = *(ir_node **) c1;
987 ir_node *n2 = *(ir_node **) c2;
991 Classical qsort() comparison function behavior:
992 0 if both elements are equal
993 1 if second is "smaller" that first
994 -1 if first is "smaller" that second
996 if (dependent_on(n1, n2))
999 if (dependent_on(n2, n1))
1002 /* The nodes have no depth order, but we need a total order because qsort()
1005 * Additionally, we need to respect transitive dependencies. Consider a
1006 * Call a depending on Call b and an independent Call c.
1007 * We MUST NOT order c > a and b > c. */
1008 h1 = get_irn_height(ir_heights, n1);
1009 h2 = get_irn_height(ir_heights, n2);
1010 if (h1 < h2) return -1;
1011 if (h1 > h2) return 1;
1012 /* Same height, so use a random (but stable) order */
1013 return get_irn_idx(n1) - get_irn_idx(n2);
1017 * Walker: links all Call/Alloc/Free nodes to the Block they are contained.
1018 * Clears the irg_is_leaf flag if a Call is detected.
1020 static void link_ops_in_block_walker(ir_node *irn, void *data)
1022 be_abi_irg_t *env = (be_abi_irg_t*)data;
1023 unsigned code = get_irn_opcode(irn);
1025 if (code == iro_Call ||
1026 (code == iro_Alloc && get_Alloc_where(irn) == stack_alloc) ||
1027 (code == iro_Free && get_Free_where(irn) == stack_alloc)) {
1028 ir_node *bl = get_nodes_block(irn);
1029 void *save = get_irn_link(bl);
1031 if (code == iro_Call)
1032 env->call->flags.bits.irg_is_leaf = 0;
1034 set_irn_link(irn, save);
1035 set_irn_link(bl, irn);
1038 if (code == iro_Builtin && get_Builtin_kind(irn) == ir_bk_return_address) {
1039 ir_node *param = get_Builtin_param(irn, 0);
1040 ir_tarval *tv = get_Const_tarval(param);
1041 unsigned long value = get_tarval_long(tv);
1042 /* use ebp, so the climbframe algo works... */
1044 env->call->flags.bits.try_omit_fp = 0;
1051 * Process all Call/Alloc/Free nodes inside a basic block.
1052 * Note that the link field of the block must contain a linked list of all
1053 * nodes inside the Block. We first order this list according to data dependency
1054 * and that connect the nodes together.
1056 static void process_ops_in_block(ir_node *bl, void *data)
1058 be_abi_irg_t *env = (be_abi_irg_t*)data;
1059 ir_node *curr_sp = env->init_sp;
1066 for (irn = (ir_node*)get_irn_link(bl); irn != NULL;
1067 irn = (ir_node*)get_irn_link(irn)) {
1071 nodes = ALLOCAN(ir_node*, n_nodes);
1072 for (irn = (ir_node*)get_irn_link(bl), n = 0; irn != NULL;
1073 irn = (ir_node*)get_irn_link(irn), ++n) {
1077 /* If there were call nodes in the block. */
1082 /* order the call nodes according to data dependency */
1083 qsort(nodes, n_nodes, sizeof(nodes[0]), cmp_call_dependency);
1085 for (i = n_nodes - 1; i >= 0; --i) {
1086 ir_node *irn = nodes[i];
1088 DBG((dbg, LEVEL_3, "\tprocessing call %+F\n", irn));
1089 switch (get_irn_opcode(irn)) {
1092 /* The stack pointer will be modified due to a call. */
1093 env->call->flags.bits.try_omit_fp = 0;
1095 curr_sp = adjust_call(env, irn, curr_sp);
1098 if (get_Alloc_where(irn) == stack_alloc)
1099 curr_sp = adjust_alloc(env, irn, curr_sp);
1102 if (get_Free_where(irn) == stack_alloc)
1103 curr_sp = adjust_free(env, irn, curr_sp);
1106 panic("invalid call");
1110 /* Keep the last stack state in the block by tying it to Keep node,
1111 * the proj from calls is already kept */
1112 if (curr_sp != env->init_sp &&
1113 !(is_Proj(curr_sp) && be_is_Call(get_Proj_pred(curr_sp)))) {
1115 keep = be_new_Keep(bl, 1, nodes);
1116 pmap_insert(env->keep_map, bl, keep);
1120 set_irn_link(bl, curr_sp);
1124 * Adjust all call nodes in the graph to the ABI conventions.
1126 static void process_calls(ir_graph *irg)
1128 be_abi_irg_t *abi = be_get_irg_abi(irg);
1130 abi->call->flags.bits.irg_is_leaf = 1;
1131 irg_walk_graph(irg, firm_clear_link, link_ops_in_block_walker, abi);
1133 ir_heights = heights_new(irg);
1134 irg_block_walk_graph(irg, NULL, process_ops_in_block, abi);
1135 heights_free(ir_heights);
1139 * Computes the stack argument layout type.
1140 * Changes a possibly allocated value param type by moving
1141 * entities to the stack layout type.
1143 * @param call the current call ABI
1144 * @param method_type the method type
1145 * @param param_map an array mapping method arguments to the stack layout
1148 * @return the stack argument layout type
1150 static ir_type *compute_arg_type(ir_graph *irg, be_abi_call_t *call,
1151 ir_type *method_type, ir_entity ***param_map)
1153 struct obstack *obst = be_get_be_obst(irg);
1154 ir_type *frame_type = get_irg_frame_type(irg);
1155 size_t n_params = get_method_n_params(method_type);
1156 size_t n_frame_members = get_compound_n_members(frame_type);
1157 ir_entity *va_start_entity = NULL;
1165 *param_map = map = OALLOCNZ(obst, ir_entity*, n_params);
1166 res = new_type_struct(new_id_from_chars("arg_type", 8));
1168 /* collect existing entities for value_param_types */
1169 for (f = n_frame_members; f > 0; ) {
1170 ir_entity *entity = get_compound_member(frame_type, --f);
1173 set_entity_link(entity, NULL);
1174 if (!is_parameter_entity(entity))
1176 num = get_entity_parameter_number(entity);
1177 if (num == IR_VA_START_PARAMETER_NUMBER) {
1178 /* move entity to new arg_type */
1179 set_entity_owner(entity, res);
1180 va_start_entity = entity;
1183 assert(num < n_params);
1184 if (map[num] != NULL)
1185 panic("multiple entities for parameter %u in %+F found", f, irg);
1187 if (num != n_params && !get_call_arg(call, 0, num, 1)->on_stack) {
1188 /* don't move this entity */
1193 /* move entity to new arg_type */
1194 set_entity_owner(entity, res);
1197 for (i = 0; i < n_params; ++i) {
1198 be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 1);
1199 ir_type *param_type = get_method_param_type(method_type, i);
1202 if (!arg->on_stack) {
1206 if (entity == NULL) {
1207 /* create a new entity */
1208 entity = new_parameter_entity(res, i, param_type);
1210 ofs += arg->space_before;
1211 ofs = round_up2(ofs, arg->alignment);
1212 set_entity_offset(entity, ofs);
1213 ofs += arg->space_after;
1214 ofs += get_type_size_bytes(param_type);
1215 arg->stack_ent = entity;
1217 if (va_start_entity != NULL) {
1218 set_entity_offset(va_start_entity, ofs);
1220 set_type_size_bytes(res, ofs);
1221 set_type_state(res, layout_fixed);
1227 const arch_register_t *reg;
1231 static int cmp_regs(const void *a, const void *b)
1233 const reg_node_map_t *p = (const reg_node_map_t*)a;
1234 const reg_node_map_t *q = (const reg_node_map_t*)b;
1236 if (p->reg->reg_class == q->reg->reg_class)
1237 return p->reg->index - q->reg->index;
1239 return p->reg->reg_class < q->reg->reg_class ? -1 : +1;
1242 static void reg_map_to_arr(reg_node_map_t *res, pmap *reg_map)
1245 size_t n = pmap_count(reg_map);
1248 foreach_pmap(reg_map, ent) {
1249 res[i].reg = (const arch_register_t*)ent->key;
1250 res[i].irn = (ir_node*)ent->value;
1254 qsort(res, n, sizeof(res[0]), cmp_regs);
1258 * Creates a be_Return for a Return node.
1260 * @param @env the abi environment
1261 * @param irn the Return node or NULL if there was none
1262 * @param bl the block where the be_Retun should be placed
1263 * @param mem the current memory
1264 * @param n_res number of return results
1266 static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl,
1267 ir_node *mem, int n_res)
1269 be_abi_call_t *call = env->call;
1270 ir_graph *irg = get_Block_irg(bl);
1271 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1273 pmap *reg_map = pmap_create();
1274 ir_node *keep = (ir_node*)pmap_get(env->keep_map, bl);
1281 const arch_register_t **regs;
1285 get the valid stack node in this block.
1286 If we had a call in that block there is a Keep constructed by process_calls()
1287 which points to the last stack modification in that block. we'll use
1288 it then. Else we use the stack from the start block and let
1289 the ssa construction fix the usage.
1291 stack = be_abi_reg_map_get(env->regs, arch_env->sp);
1293 stack = get_irn_n(keep, 0);
1295 remove_End_keepalive(get_irg_end(irg), keep);
1298 /* Insert results for Return into the register map. */
1299 for (i = 0; i < n_res; ++i) {
1300 ir_node *res = get_Return_res(irn, i);
1301 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
1302 assert(arg->in_reg && "return value must be passed in register");
1303 pmap_insert(reg_map, (void *) arg->reg, res);
1306 /* Add uses of the callee save registers. */
1307 foreach_pmap(env->regs, ent) {
1308 const arch_register_t *reg = (const arch_register_t*)ent->key;
1309 if ((reg->type & arch_register_type_ignore) || arch_register_is_callee_save(arch_env, reg))
1310 pmap_insert(reg_map, ent->key, ent->value);
1313 be_abi_reg_map_set(reg_map, arch_env->sp, stack);
1316 Maximum size of the in array for Return nodes is
1317 return args + callee save/ignore registers + memory + stack pointer
1319 in_max = pmap_count(reg_map) + n_res + 2;
1321 in = ALLOCAN(ir_node*, in_max);
1322 regs = ALLOCAN(arch_register_t const*, in_max);
1325 in[1] = be_abi_reg_map_get(reg_map, arch_env->sp);
1327 regs[1] = arch_env->sp;
1330 /* clear SP entry, since it has already been grown. */
1331 pmap_insert(reg_map, (void *) arch_env->sp, NULL);
1332 for (i = 0; i < n_res; ++i) {
1333 be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
1335 in[n] = be_abi_reg_map_get(reg_map, arg->reg);
1336 regs[n++] = arg->reg;
1338 /* Clear the map entry to mark the register as processed. */
1339 be_abi_reg_map_set(reg_map, arg->reg, NULL);
1342 /* grow the rest of the stuff. */
1343 foreach_pmap(reg_map, ent) {
1345 in[n] = (ir_node*)ent->value;
1346 regs[n++] = (const arch_register_t*)ent->key;
1350 /* The in array for the new back end return is now ready. */
1352 dbgi = get_irn_dbg_info(irn);
1356 /* we have to pop the shadow parameter in in case of struct returns */
1358 ret = be_new_Return(dbgi, irg, bl, n_res, pop, n, in);
1360 /* Set the register classes of the return's parameter accordingly. */
1361 for (i = 0; i < n; ++i) {
1362 if (regs[i] == NULL)
1365 be_set_constr_single_reg_in(ret, i, regs[i], arch_register_req_type_none);
1368 /* Free the space of the Epilog's in array and the register <-> proj map. */
1369 pmap_destroy(reg_map);
1374 typedef struct lower_frame_sels_env_t {
1375 ir_node *frame; /**< the current frame */
1376 const arch_register_class_t *sp_class; /**< register class of the stack pointer */
1377 const arch_register_class_t *link_class; /**< register class of the link pointer */
1378 ir_type *frame_tp; /**< the frame type */
1379 int static_link_pos; /**< argument number of the hidden static link */
1380 } lower_frame_sels_env_t;
1383 * Walker: Replaces Sels of frame type and
1384 * value param type entities by FrameAddress.
1385 * Links all used entities.
1387 static void lower_frame_sels_walker(ir_node *irn, void *data)
1389 lower_frame_sels_env_t *ctx = (lower_frame_sels_env_t*)data;
1392 ir_node *ptr = get_Sel_ptr(irn);
1394 if (ptr == ctx->frame) {
1395 ir_entity *ent = get_Sel_entity(irn);
1396 ir_node *bl = get_nodes_block(irn);
1399 nw = be_new_FrameAddr(ctx->sp_class, bl, ctx->frame, ent);
1406 * The start block has no jump, instead it has an initial exec Proj.
1407 * The backend wants to handle all blocks the same way, so we replace
1408 * the out cfg edge with a real jump.
1410 static void fix_start_block(ir_graph *irg)
1412 ir_node *initial_X = get_irg_initial_exec(irg);
1413 ir_node *start_block = get_irg_start_block(irg);
1414 ir_node *jmp = new_r_Jmp(start_block);
1416 assert(is_Proj(initial_X));
1417 exchange(initial_X, jmp);
1418 set_irg_initial_exec(irg, new_r_Bad(irg, mode_X));
1420 /* merge start block with successor if possible */
1422 const ir_edge_t *edge;
1423 foreach_out_edge(jmp, edge) {
1424 ir_node *succ = get_edge_src_irn(edge);
1425 if (!is_Block(succ))
1428 if (get_irn_arity(succ) == 1) {
1429 exchange(succ, start_block);
1437 * Modify the irg itself and the frame type.
1439 static void modify_irg(ir_graph *irg)
1441 be_abi_irg_t *env = be_get_irg_abi(irg);
1442 be_abi_call_t *call = env->call;
1443 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1444 const arch_register_t *sp = arch_env->sp;
1445 ir_type *method_type = get_entity_type(get_irg_entity(irg));
1446 be_irg_t *birg = be_birg_from_irg(irg);
1447 struct obstack *obst = be_get_be_obst(irg);
1448 be_stack_layout_t *stack_layout = be_get_irg_stack_layout(irg);
1451 ir_node *new_mem_proj;
1457 unsigned frame_size;
1460 const arch_register_t *fp_reg;
1461 ir_node *frame_pointer;
1465 const ir_edge_t *edge;
1466 ir_type *arg_type, *bet_type;
1467 lower_frame_sels_env_t ctx;
1468 ir_entity **param_map;
1470 DBG((dbg, LEVEL_1, "introducing abi on %+F\n", irg));
1472 old_mem = get_irg_initial_mem(irg);
1474 irp_reserve_resources(irp, IRP_RESOURCE_ENTITY_LINK);
1476 arg_type = compute_arg_type(irg, call, method_type, ¶m_map);
1478 /* Convert the Sel nodes in the irg to frame addr nodes: */
1479 ctx.frame = get_irg_frame(irg);
1480 ctx.sp_class = arch_env->sp->reg_class;
1481 ctx.link_class = arch_env->link_class;
1482 ctx.frame_tp = get_irg_frame_type(irg);
1484 /* layout the stackframe now */
1485 if (get_type_state(ctx.frame_tp) == layout_undefined) {
1486 default_layout_compound_type(ctx.frame_tp);
1489 /* align stackframe to 4 byte */
1490 frame_size = get_type_size_bytes(ctx.frame_tp);
1491 if (frame_size % 4 != 0) {
1492 set_type_size_bytes(ctx.frame_tp, frame_size + 4 - (frame_size % 4));
1495 env->regs = pmap_create();
1497 n_params = get_method_n_params(method_type);
1498 args = OALLOCNZ(obst, ir_node*, n_params);
1500 be_add_parameter_entity_stores(irg);
1502 irg_walk_graph(irg, lower_frame_sels_walker, NULL, &ctx);
1504 irp_free_resources(irp, IRP_RESOURCE_ENTITY_LINK);
1506 /* Fill the argument vector */
1507 arg_tuple = get_irg_args(irg);
1508 foreach_out_edge(arg_tuple, edge) {
1509 ir_node *irn = get_edge_src_irn(edge);
1510 if (! is_Anchor(irn)) {
1511 int nr = get_Proj_proj(irn);
1513 DBG((dbg, LEVEL_2, "\treading arg: %d -> %+F\n", nr, irn));
1517 stack_layout->sp_relative = call->flags.bits.try_omit_fp;
1518 bet_type = call->cb->get_between_type(irg);
1519 stack_frame_init(stack_layout, arg_type, bet_type,
1520 get_irg_frame_type(irg), param_map);
1522 /* Count the register params and add them to the number of Projs for the RegParams node */
1523 for (i = 0; i < n_params; ++i) {
1524 be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 1);
1525 if (arg->in_reg && args[i]) {
1526 assert(arg->reg != sp && "cannot use stack pointer as parameter register");
1527 assert(i == get_Proj_proj(args[i]));
1529 /* For now, associate the register with the old Proj from Start representing that argument. */
1530 pmap_insert(env->regs, (void *) arg->reg, args[i]);
1531 DBG((dbg, LEVEL_2, "\targ #%d -> reg %s\n", i, arg->reg->name));
1535 /* Collect all callee-save registers */
1536 for (i = 0, n = arch_env->n_register_classes; i < n; ++i) {
1537 const arch_register_class_t *cls = &arch_env->register_classes[i];
1538 for (j = 0; j < cls->n_regs; ++j) {
1539 const arch_register_t *reg = &cls->regs[j];
1540 if ((reg->type & arch_register_type_state) || arch_register_is_callee_save(arch_env, reg)) {
1541 pmap_insert(env->regs, (void *) reg, NULL);
1546 fp_reg = call->flags.bits.try_omit_fp ? arch_env->sp : arch_env->bp;
1547 rbitset_clear(birg->allocatable_regs, fp_reg->global_index);
1549 /* handle start block here (place a jump in the block) */
1550 fix_start_block(irg);
1552 pmap_insert(env->regs, (void *) sp, NULL);
1553 pmap_insert(env->regs, (void *) arch_env->bp, NULL);
1554 start_bl = get_irg_start_block(irg);
1555 env->start = be_new_Start(NULL, start_bl, pmap_count(env->regs) + 1);
1556 set_irg_start(irg, env->start);
1559 * make proj nodes for the callee save registers.
1560 * memorize them, since Return nodes get those as inputs.
1562 * Note, that if a register corresponds to an argument, the regs map
1563 * contains the old Proj from start for that argument.
1565 rm = ALLOCAN(reg_node_map_t, pmap_count(env->regs));
1566 reg_map_to_arr(rm, env->regs);
1567 for (i = 0, n = pmap_count(env->regs); i < n; ++i) {
1568 const arch_register_t *reg = rm[i].reg;
1569 ir_mode *mode = reg->reg_class->mode;
1571 arch_register_req_type_t add_type = arch_register_req_type_none;
1575 add_type |= arch_register_req_type_produces_sp;
1576 if (!rbitset_is_set(birg->allocatable_regs, reg->global_index)) {
1577 add_type |= arch_register_req_type_ignore;
1581 proj = new_r_Proj(env->start, mode, nr + 1);
1582 pmap_insert(env->regs, (void *) reg, proj);
1583 be_set_constr_single_reg_out(env->start, nr + 1, reg, add_type);
1584 arch_set_irn_register(proj, reg);
1586 DBG((dbg, LEVEL_2, "\tregister save proj #%d -> reg %s\n", nr, reg->name));
1589 /* create a new initial memory proj */
1590 assert(is_Proj(old_mem));
1591 arch_set_irn_register_req_out(env->start, 0, arch_no_register_req);
1592 new_mem_proj = new_r_Proj(env->start, mode_M, 0);
1594 set_irg_initial_mem(irg, mem);
1596 env->init_sp = be_abi_reg_map_get(env->regs, sp);
1598 /* set new frame_pointer */
1599 frame_pointer = be_abi_reg_map_get(env->regs, fp_reg);
1600 set_irg_frame(irg, frame_pointer);
1602 /* rewire old mem users to new mem */
1603 exchange(old_mem, mem);
1605 /* keep the mem (for functions with an endless loop = no return) */
1608 set_irg_initial_mem(irg, mem);
1610 /* Now, introduce stack param nodes for all parameters passed on the stack */
1611 for (i = 0; i < n_params; ++i) {
1612 ir_node *arg_proj = args[i];
1613 ir_node *repl = NULL;
1615 if (arg_proj != NULL) {
1616 be_abi_call_arg_t *arg;
1617 ir_type *param_type;
1618 int nr = get_Proj_proj(arg_proj);
1621 nr = MIN(nr, n_params);
1622 arg = get_call_arg(call, 0, nr, 1);
1623 param_type = get_method_param_type(method_type, nr);
1626 repl = (ir_node*)pmap_get(env->regs, arg->reg);
1627 } else if (arg->on_stack) {
1628 ir_node *addr = be_new_FrameAddr(sp->reg_class, start_bl, frame_pointer, arg->stack_ent);
1630 /* For atomic parameters which are actually used, we create a Load node. */
1631 if (is_atomic_type(param_type) && get_irn_n_edges(args[i]) > 0) {
1632 ir_mode *mode = get_type_mode(param_type);
1633 ir_mode *load_mode = arg->load_mode;
1634 ir_node *nomem = get_irg_no_mem(irg);
1636 ir_node *load = new_r_Load(start_bl, nomem, addr, load_mode, cons_floats);
1637 repl = new_r_Proj(load, load_mode, pn_Load_res);
1639 if (mode != load_mode) {
1640 repl = new_r_Conv(start_bl, repl, mode);
1643 /* The stack parameter is not primitive (it is a struct or array),
1644 * we thus will create a node representing the parameter's address
1650 assert(repl != NULL);
1652 /* Beware: the mode of the register parameters is always the mode of the register class
1653 which may be wrong. Add Conv's then. */
1654 mode = get_irn_mode(args[i]);
1655 if (mode != get_irn_mode(repl)) {
1656 repl = new_r_Conv(get_nodes_block(repl), repl, mode);
1658 exchange(args[i], repl);
1662 /* the arg proj is not needed anymore now and should be only used by the anchor */
1663 assert(get_irn_n_edges(arg_tuple) == 1);
1664 kill_node(arg_tuple);
1665 set_irg_args(irg, new_r_Bad(irg, mode_T));
1667 /* All Return nodes hang on the End node, so look for them there. */
1668 end = get_irg_end_block(irg);
1669 for (i = 0, n = get_Block_n_cfgpreds(end); i < n; ++i) {
1670 ir_node *irn = get_Block_cfgpred(end, i);
1672 if (is_Return(irn)) {
1673 ir_node *blk = get_nodes_block(irn);
1674 ir_node *mem = get_Return_mem(irn);
1675 ir_node *ret = create_be_return(env, irn, blk, mem, get_Return_n_ress(irn));
1680 /* if we have endless loops here, n might be <= 0. Do NOT create a be_Return then,
1681 the code is dead and will never be executed. */
1684 /** Fix the state inputs of calls that still hang on unknowns */
1685 static void fix_call_state_inputs(ir_graph *irg)
1687 be_abi_irg_t *env = be_get_irg_abi(irg);
1688 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1690 arch_register_t **stateregs = NEW_ARR_F(arch_register_t*, 0);
1692 /* Collect caller save registers */
1693 n = arch_env->n_register_classes;
1694 for (i = 0; i < n; ++i) {
1696 const arch_register_class_t *cls = &arch_env->register_classes[i];
1697 for (j = 0; j < cls->n_regs; ++j) {
1698 const arch_register_t *reg = arch_register_for_index(cls, j);
1699 if (reg->type & arch_register_type_state) {
1700 ARR_APP1(arch_register_t*, stateregs, (arch_register_t *)reg);
1705 n = ARR_LEN(env->calls);
1706 n_states = ARR_LEN(stateregs);
1707 for (i = 0; i < n; ++i) {
1709 ir_node *call = env->calls[i];
1711 arity = get_irn_arity(call);
1713 /* the state reg inputs are the last n inputs of the calls */
1714 for (s = 0; s < n_states; ++s) {
1715 int inp = arity - n_states + s;
1716 const arch_register_t *reg = stateregs[s];
1717 ir_node *regnode = be_abi_reg_map_get(env->regs, reg);
1719 set_irn_n(call, inp, regnode);
1723 DEL_ARR_F(stateregs);
1727 * Create a trampoline entity for the given method.
1729 static ir_entity *create_trampoline(be_main_env_t *be, ir_entity *method)
1731 ir_type *type = get_entity_type(method);
1732 ident *old_id = get_entity_ld_ident(method);
1733 ident *id = id_mangle3("", old_id, "$stub");
1734 ir_type *parent = be->pic_trampolines_type;
1735 ir_entity *ent = new_entity(parent, old_id, type);
1736 set_entity_ld_ident(ent, id);
1737 set_entity_visibility(ent, ir_visibility_private);
1743 * Returns the trampoline entity for the given method.
1745 static ir_entity *get_trampoline(be_main_env_t *env, ir_entity *method)
1747 ir_entity *result = (ir_entity*)pmap_get(env->ent_trampoline_map, method);
1748 if (result == NULL) {
1749 result = create_trampoline(env, method);
1750 pmap_insert(env->ent_trampoline_map, method, result);
1756 static ir_entity *create_pic_symbol(be_main_env_t *be, ir_entity *entity)
1758 ident *old_id = get_entity_ld_ident(entity);
1759 ident *id = id_mangle3("", old_id, "$non_lazy_ptr");
1760 ir_type *e_type = get_entity_type(entity);
1761 ir_type *type = new_type_pointer(e_type);
1762 ir_type *parent = be->pic_symbols_type;
1763 ir_entity *ent = new_entity(parent, old_id, type);
1764 set_entity_ld_ident(ent, id);
1765 set_entity_visibility(ent, ir_visibility_private);
1770 static ir_entity *get_pic_symbol(be_main_env_t *env, ir_entity *entity)
1772 ir_entity *result = (ir_entity*)pmap_get(env->ent_pic_symbol_map, entity);
1773 if (result == NULL) {
1774 result = create_pic_symbol(env, entity);
1775 pmap_insert(env->ent_pic_symbol_map, entity, result);
1784 * Returns non-zero if a given entity can be accessed using a relative address.
1786 static int can_address_relative(ir_entity *entity)
1788 return get_entity_visibility(entity) != ir_visibility_external
1789 && !(get_entity_linkage(entity) & IR_LINKAGE_MERGE);
1792 static ir_node *get_pic_base(ir_graph *irg)
1794 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1795 if (arch_env->impl->get_pic_base == NULL)
1797 return arch_env->impl->get_pic_base(irg);
1800 /** patches SymConsts to work in position independent code */
1801 static void fix_pic_symconsts(ir_node *node, void *data)
1803 ir_graph *irg = get_irn_irg(node);
1804 be_main_env_t *be = be_get_irg_main_env(irg);
1814 arity = get_irn_arity(node);
1815 for (i = 0; i < arity; ++i) {
1817 ir_node *pred = get_irn_n(node, i);
1819 ir_entity *pic_symbol;
1820 ir_node *pic_symconst;
1822 if (!is_SymConst(pred))
1825 entity = get_SymConst_entity(pred);
1826 block = get_nodes_block(pred);
1828 /* calls can jump to relative addresses, so we can directly jump to
1829 the (relatively) known call address or the trampoline */
1830 if (i == 1 && is_Call(node)) {
1831 ir_entity *trampoline;
1832 ir_node *trampoline_const;
1834 if (can_address_relative(entity))
1837 dbgi = get_irn_dbg_info(pred);
1838 trampoline = get_trampoline(be, entity);
1839 trampoline_const = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
1841 set_irn_n(node, i, trampoline_const);
1845 /* everything else is accessed relative to EIP */
1846 mode = get_irn_mode(pred);
1847 pic_base = get_pic_base(irg);
1849 /* all ok now for locally constructed stuff */
1850 if (can_address_relative(entity)) {
1851 ir_node *add = new_r_Add(block, pic_base, pred, mode);
1853 /* make sure the walker doesn't visit this add again */
1854 mark_irn_visited(add);
1855 set_irn_n(node, i, add);
1859 /* get entry from pic symbol segment */
1860 dbgi = get_irn_dbg_info(pred);
1861 pic_symbol = get_pic_symbol(be, entity);
1862 pic_symconst = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
1864 add = new_r_Add(block, pic_base, pic_symconst, mode);
1865 mark_irn_visited(add);
1867 /* we need an extra indirection for global data outside our current
1868 module. The loads are always safe and can therefore float
1869 and need no memory input */
1870 load = new_r_Load(block, get_irg_no_mem(irg), add, mode, cons_floats);
1871 load_res = new_r_Proj(load, mode, pn_Load_res);
1873 set_irn_n(node, i, load_res);
1877 void be_abi_introduce(ir_graph *irg)
1879 be_abi_irg_t *env = XMALLOCZ(be_abi_irg_t);
1880 ir_node *old_frame = get_irg_frame(irg);
1881 be_options_t *options = be_get_irg_options(irg);
1882 const arch_env_t *arch_env = be_get_irg_arch_env(irg);
1883 ir_entity *entity = get_irg_entity(irg);
1884 ir_type *method_type = get_entity_type(entity);
1885 be_irg_t *birg = be_birg_from_irg(irg);
1886 struct obstack *obst = &birg->obst;
1887 ir_node *dummy = new_r_Dummy(irg,
1888 arch_env->sp->reg_class->mode);
1891 /* determine allocatable registers */
1892 assert(birg->allocatable_regs == NULL);
1893 birg->allocatable_regs = rbitset_obstack_alloc(obst, arch_env->n_registers);
1894 for (r = 0; r < arch_env->n_registers; ++r) {
1895 const arch_register_t *reg = &arch_env->registers[r];
1896 if ( !(reg->type & arch_register_type_ignore)) {
1897 rbitset_set(birg->allocatable_regs, r);
1901 /* break here if backend provides a custom API.
1902 * Note: we shouldn't have to setup any be_abi_irg_t* stuff at all,
1903 * but need more cleanup to make this work
1905 be_set_irg_abi(irg, env);
1907 be_omit_fp = options->omit_fp;
1909 env->keep_map = pmap_create();
1910 env->call = be_abi_call_new(arch_env->sp->reg_class);
1911 arch_env_get_call_abi(arch_env, method_type, env->call);
1913 env->init_sp = dummy;
1914 env->calls = NEW_ARR_F(ir_node*, 0);
1919 irg_walk_graph(irg, fix_pic_symconsts, NULL, env);
1922 /* Lower all call nodes in the IRG. */
1925 /* Process the IRG */
1928 /* fix call inputs for state registers */
1929 fix_call_state_inputs(irg);
1931 /* We don't need the keep map anymore. */
1932 pmap_destroy(env->keep_map);
1933 env->keep_map = NULL;
1935 /* calls array is not needed anymore */
1936 DEL_ARR_F(env->calls);
1939 /* reroute the stack origin of the calls to the true stack origin. */
1940 exchange(dummy, env->init_sp);
1941 exchange(old_frame, get_irg_frame(irg));
1943 pmap_destroy(env->regs);
1947 void be_abi_free(ir_graph *irg)
1949 be_abi_irg_t *env = be_get_irg_abi(irg);
1951 if (env->call != NULL)
1952 be_abi_call_free(env->call);
1953 assert(env->regs == NULL);
1956 be_set_irg_abi(irg, NULL);
1959 void be_put_allocatable_regs(const ir_graph *irg,
1960 const arch_register_class_t *cls, bitset_t *bs)
1962 be_irg_t *birg = be_birg_from_irg(irg);
1963 unsigned *allocatable_regs = birg->allocatable_regs;
1966 assert(bitset_size(bs) == cls->n_regs);
1967 bitset_clear_all(bs);
1968 for (i = 0; i < cls->n_regs; ++i) {
1969 const arch_register_t *reg = &cls->regs[i];
1970 if (rbitset_is_set(allocatable_regs, reg->global_index))
1975 unsigned be_get_n_allocatable_regs(const ir_graph *irg,
1976 const arch_register_class_t *cls)
1978 bitset_t *bs = bitset_alloca(cls->n_regs);
1979 be_put_allocatable_regs(irg, cls, bs);
1980 return bitset_popcount(bs);
1983 void be_set_allocatable_regs(const ir_graph *irg,
1984 const arch_register_class_t *cls,
1985 unsigned *raw_bitset)
1987 be_irg_t *birg = be_birg_from_irg(irg);
1988 unsigned *allocatable_regs = birg->allocatable_regs;
1991 rbitset_clear_all(raw_bitset, cls->n_regs);
1992 for (i = 0; i < cls->n_regs; ++i) {
1993 const arch_register_t *reg = &cls->regs[i];
1994 if (rbitset_is_set(allocatable_regs, reg->global_index))
1995 rbitset_set(raw_bitset, i);
1999 BE_REGISTER_MODULE_CONSTRUCTOR(be_init_abi)
2000 void be_init_abi(void)
2002 FIRM_DBG_REGISTER(dbg, "firm.be.abi");