avoid unnecessary passing around of arch_env_t* in backend APIs
[libfirm] / ir / be / beabi.c
index 781a7df..9bc4ca4 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 1995-2007 University of Karlsruhe.  All right reserved.
+ * Copyright (C) 1995-2008 University of Karlsruhe.  All right reserved.
  *
  * This file is part of libFirm.
  *
  * @author      Sebastian Hack, Michael Beck
  * @version     $Id$
  */
-#ifdef HAVE_CONFIG_H
-# include "config.h"
-#endif
+#include "config.h"
 
 #include "obst.h"
-#include "offset.h"
 
 #include "irgopt.h"
 
 #include "irprintf_t.h"
 #include "irgopt.h"
 #include "irbitset.h"
+#include "iropt_t.h"
 #include "height.h"
 #include "pdeq.h"
 #include "irtools.h"
 #include "raw_bitset.h"
 #include "error.h"
+#include "pset_new.h"
 
 #include "be.h"
 #include "beabi.h"
-#include "bearch_t.h"
-#include "benode_t.h"
+#include "bearch.h"
+#include "benode.h"
 #include "belive_t.h"
-#include "besched_t.h"
-#include "beirg_t.h"
+#include "besched.h"
+#include "beirg.h"
 #include "bessaconstr.h"
+#include "bemodule.h"
+
+DEBUG_ONLY(static firm_dbg_module_t *dbg;)
 
 typedef struct _be_abi_call_arg_t {
        unsigned is_res   : 1;  /**< 1: the call argument is a return value. 0: it's a call parameter. */
        unsigned in_reg   : 1;  /**< 1: this argument is transmitted in registers. */
        unsigned on_stack : 1;  /**< 1: this argument is transmitted on the stack. */
+       unsigned callee   : 1;  /**< 1: someone called us. 0: We call another function */
 
-       int pos;
+       int                    pos;
        const arch_register_t *reg;
-       ir_entity *stack_ent;
-       unsigned alignment;     /**< stack alignment */
-       unsigned space_before;  /**< allocate space before */
-       unsigned space_after;   /**< allocate space after */
+       ir_entity             *stack_ent;
+       ir_mode               *load_mode;
+       unsigned               alignment;    /**< stack alignment */
+       unsigned               space_before; /**< allocate space before */
+       unsigned               space_after;  /**< allocate space after */
 } be_abi_call_arg_t;
 
 struct _be_abi_call_t {
-       be_abi_call_flags_t         flags;
+       be_abi_call_flags_t          flags;  /**< Flags describing the ABI behavior on calls */
+       int                          pop;    /**< number of bytes the stack frame is shrinked by the callee on return. */
        const be_abi_callbacks_t    *cb;
        ir_type                     *between_type;
        set                         *params;
-       const arch_register_class_t *cls_addr;
+       const arch_register_class_t *cls_addr; /**< register class of the call address */
 };
 
+/**
+ * The ABI information for the current graph.
+ */
 struct _be_abi_irg_t {
-       struct obstack       obst;
-       be_stack_layout_t    *frame;        /**< The stack frame model. */
-       be_irg_t             *birg;         /**< The back end IRG. */
-       const arch_isa_t     *isa;          /**< The isa. */
        survive_dce_t        *dce_survivor;
 
        be_abi_call_t        *call;         /**< The ABI call information. */
-       ir_type              *method_type;  /**< The type of the method of the IRG. */
 
        ir_node              *init_sp;      /**< The node representing the stack pointer
                                                 at the start of the function. */
 
-       ir_node              *start_barrier; /**< The barrier of the start block */
-
-       ir_node              *reg_params;   /**< The reg params node. */
+       ir_node              *start;        /**< The be_Start params node. */
        pmap                 *regs;         /**< A map of all callee-save and ignore regs to
                                                 their Projs to the RegParams node. */
 
@@ -103,19 +104,19 @@ struct _be_abi_irg_t {
        pmap                 *keep_map;     /**< mapping blocks to keep nodes. */
        pset                 *ignore_regs;  /**< Additional registers which shall be ignored. */
 
-       ir_node             **calls;        /**< flexible array containing all be_Call nodes */
-
-       arch_register_req_t sp_req;
-       arch_register_req_t sp_cls_req;
+       ir_node              **calls;       /**< flexible array containing all be_Call nodes */
 
-       DEBUG_ONLY(firm_dbg_module_t    *dbg;)          /**< The debugging module. */
+       arch_register_req_t  *sp_req;
 };
 
 static heights_t *ir_heights;
 
-/* Flag: if set, try to omit the frame pointer if called by the backend */
+/** Flag: if set, try to omit the frame pointer in all routines. */
 static int be_omit_fp = 1;
 
+/** Flag: if set, try to omit the frame pointer in leaf routines only. */
+static int be_omit_leaf_fp = 1;
+
 /*
      _    ____ ___    ____      _ _ _                _
     / \  | __ )_ _|  / ___|__ _| | | |__   __ _  ___| | _____
@@ -134,18 +135,18 @@ static int cmp_call_arg(const void *a, const void *b, size_t n)
 {
        const be_abi_call_arg_t *p = a, *q = b;
        (void) n;
-       return !(p->is_res == q->is_res && p->pos == q->pos);
+       return !(p->is_res == q->is_res && p->pos == q->pos && p->callee == q->callee);
 }
 
 /**
- * Get or set an ABI call object argument.
+ * Get  an ABI call object argument.
  *
  * @param call      the abi call
  * @param is_res    true for call results, false for call arguments
  * @param pos       position of the argument
- * @param do_insert true if the argument is set, false if it's retrieved
+ * @param callee       context type - if we are callee or caller
  */
-static be_abi_call_arg_t *get_or_set_call_arg(be_abi_call_t *call, int is_res, int pos, int do_insert)
+static be_abi_call_arg_t *get_call_arg(be_abi_call_t *call, int is_res, int pos, int callee)
 {
        be_abi_call_arg_t arg;
        unsigned hash;
@@ -153,24 +154,27 @@ static be_abi_call_arg_t *get_or_set_call_arg(be_abi_call_t *call, int is_res, i
        memset(&arg, 0, sizeof(arg));
        arg.is_res = is_res;
        arg.pos    = pos;
+       arg.callee = callee;
 
        hash = is_res * 128 + pos;
 
-       return do_insert
-               ? set_insert(call->params, &arg, sizeof(arg), hash)
-               : set_find(call->params, &arg, sizeof(arg), hash);
+       return set_find(call->params, &arg, sizeof(arg), hash);
 }
 
 /**
- * Retrieve an ABI call object argument.
- *
- * @param call      the ABI call object
- * @param is_res    true for call results, false for call arguments
- * @param pos       position of the argument
+ * Set an ABI call object argument.
  */
-static INLINE be_abi_call_arg_t *get_call_arg(be_abi_call_t *call, int is_res, int pos)
+static void remember_call_arg(be_abi_call_arg_t *arg, be_abi_call_t *call, be_abi_context_t context)
 {
-       return get_or_set_call_arg(call, is_res, pos, 0);
+       unsigned hash = arg->is_res * 128 + arg->pos;
+       if (context & ABI_CONTEXT_CALLEE) {
+               arg->callee = 1;
+               set_insert(call->params, arg, sizeof(*arg), hash);
+       }
+       if (context & ABI_CONTEXT_CALLER) {
+               arg->callee = 0;
+               set_insert(call->params, arg, sizeof(*arg), hash);
+       }
 }
 
 /* Set the flags for a call. */
@@ -180,6 +184,12 @@ void be_abi_call_set_flags(be_abi_call_t *call, be_abi_call_flags_t flags, const
        call->cb    = cb;
 }
 
+/* Sets the number of bytes the stackframe is shrinked by the callee on return */
+void be_abi_call_set_pop(be_abi_call_t *call, int pop)
+{
+       assert(pop >= 0);
+       call->pop = pop;
+}
 
 /* Set register class for call address */
 void be_abi_call_set_call_address_reg_class(be_abi_call_t *call, const arch_register_class_t *cls)
@@ -188,28 +198,49 @@ void be_abi_call_set_call_address_reg_class(be_abi_call_t *call, const arch_regi
 }
 
 
-void be_abi_call_param_stack(be_abi_call_t *call, int arg_pos, unsigned alignment, unsigned space_before, unsigned space_after)
+void be_abi_call_param_stack(be_abi_call_t *call, int arg_pos,
+                             ir_mode *load_mode, unsigned alignment,
+                             unsigned space_before, unsigned space_after,
+                             be_abi_context_t context)
 {
-       be_abi_call_arg_t *arg = get_or_set_call_arg(call, 0, arg_pos, 1);
-       arg->on_stack     = 1;
-       arg->alignment    = alignment;
-       arg->space_before = space_before;
-       arg->space_after  = space_after;
+       be_abi_call_arg_t arg;
+       memset(&arg, 0, sizeof(arg));
        assert(alignment > 0 && "Alignment must be greater than 0");
+       arg.on_stack     = 1;
+       arg.load_mode    = load_mode;
+       arg.alignment    = alignment;
+       arg.space_before = space_before;
+       arg.space_after  = space_after;
+       arg.is_res       = 0;
+       arg.pos          = arg_pos;
+
+       remember_call_arg(&arg, call, context);
 }
 
-void be_abi_call_param_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg)
+void be_abi_call_param_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
 {
-       be_abi_call_arg_t *arg = get_or_set_call_arg(call, 0, arg_pos, 1);
-       arg->in_reg = 1;
-       arg->reg = reg;
+       be_abi_call_arg_t arg;
+       memset(&arg, 0, sizeof(arg));
+
+       arg.in_reg = 1;
+       arg.reg    = reg;
+       arg.is_res = 0;
+       arg.pos    = arg_pos;
+
+       remember_call_arg(&arg, call, context);
 }
 
-void be_abi_call_res_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg)
+void be_abi_call_res_reg(be_abi_call_t *call, int arg_pos, const arch_register_t *reg, be_abi_context_t context)
 {
-       be_abi_call_arg_t *arg = get_or_set_call_arg(call, 1, arg_pos, 1);
-       arg->in_reg = 1;
-       arg->reg = reg;
+       be_abi_call_arg_t arg;
+       memset(&arg, 0, sizeof(arg));
+
+       arg.in_reg = 1;
+       arg.reg    = reg;
+       arg.is_res = 1;
+       arg.pos    = arg_pos;
+
+       remember_call_arg(&arg, call, context);
 }
 
 /* Get the flags of a ABI call object. */
@@ -221,18 +252,20 @@ be_abi_call_flags_t be_abi_call_get_flags(const be_abi_call_t *call)
 /**
  * Constructor for a new ABI call object.
  *
+ * @param cls_addr  register class of the call address
+ *
  * @return the new ABI call object
  */
 static be_abi_call_t *be_abi_call_new(const arch_register_class_t *cls_addr)
 {
-       be_abi_call_t *call = xmalloc(sizeof(call[0]));
+       be_abi_call_t *call = XMALLOCZ(be_abi_call_t);
 
        call->flags.val  = 0;
        call->params     = new_set(cmp_call_arg, 16);
        call->cb         = NULL;
        call->cls_addr   = cls_addr;
 
-       call->flags.bits.try_omit_fp = be_omit_fp;
+       call->flags.bits.try_omit_fp = be_omit_fp | be_omit_leaf_fp;
 
        return call;
 }
@@ -263,23 +296,22 @@ static void be_abi_call_free(be_abi_call_t *call)
      and the spills.
 */
 
-static int get_stack_entity_offset(be_stack_layout_t *frame, ir_entity *ent, int bias)
+static int get_stack_entity_offset(be_stack_layout_t *frame, ir_entity *ent,
+                                   int bias)
 {
        ir_type *t = get_entity_owner(ent);
        int ofs    = get_entity_offset(ent);
 
-       int i, index;
+       int index;
 
        /* Find the type the entity is contained in. */
-       for(index = 0; index < N_FRAME_TYPES; ++index) {
-               if(frame->order[index] == t)
+       for (index = 0; index < N_FRAME_TYPES; ++index) {
+               if (frame->order[index] == t)
                        break;
+               /* Add the size of all the types below the one of the entity to the entity's offset */
+               ofs += get_type_size_bytes(frame->order[index]);
        }
 
-       /* Add the size of all the types below the one of the entity to the entity's offset */
-       for(i = 0; i < index; ++i)
-               ofs += get_type_size_bytes(frame->order[i]);
-
        /* correct the offset by the initial position of the frame pointer */
        ofs -= frame->initial_offset;
 
@@ -296,9 +328,9 @@ static ir_entity *search_ent_with_offset(ir_type *t, int offset)
 {
        int i, n;
 
-       for(i = 0, n = get_compound_n_members(t); i < n; ++i) {
+       for (i = 0, n = get_compound_n_members(t); i < n; ++i) {
                ir_entity *ent = get_compound_member(t, i);
-               if(get_entity_offset(ent) == offset)
+               if (get_entity_offset(ent) == offset)
                        return ent;
        }
 
@@ -310,7 +342,12 @@ static int stack_frame_compute_initial_offset(be_stack_layout_t *frame)
        ir_type  *base = frame->stack_dir < 0 ? frame->between_type : frame->frame_type;
        ir_entity *ent = search_ent_with_offset(base, 0);
 
-       frame->initial_offset = ent ? get_stack_entity_offset(frame, ent, 0) : 0;
+       if (ent == NULL) {
+               frame->initial_offset
+                       = frame->stack_dir < 0 ? get_type_size_bytes(frame->frame_type) : get_type_size_bytes(frame->between_type);
+       } else {
+               frame->initial_offset = get_stack_entity_offset(frame, ent, 0);
+       }
 
        return frame->initial_offset;
 }
@@ -322,7 +359,7 @@ static int stack_frame_compute_initial_offset(be_stack_layout_t *frame)
  * @param args      the stack argument layout type
  * @param between   the between layout type
  * @param locals    the method frame type
- * @param stack_dir the stack direction
+ * @param stack_dir the stack direction: < 0 decreasing, > 0 increasing addresses
  * @param param_map an array mapping method argument positions to the stack argument type
  *
  * @return the initialized stack layout
@@ -335,50 +372,23 @@ static be_stack_layout_t *stack_frame_init(be_stack_layout_t *frame, ir_type *ar
        frame->between_type   = between;
        frame->frame_type     = locals;
        frame->initial_offset = 0;
+       frame->initial_bias   = 0;
        frame->stack_dir      = stack_dir;
        frame->order[1]       = between;
        frame->param_map      = param_map;
 
-       if(stack_dir > 0) {
+       if (stack_dir > 0) {
                frame->order[0] = args;
                frame->order[2] = locals;
-       }
-       else {
+       } else {
+               /* typical decreasing stack: locals have the
+                * lowest addresses, arguments the highest */
                frame->order[0] = locals;
                frame->order[2] = args;
        }
        return frame;
 }
 
-#if 0
-/** Dumps the stack layout to file. */
-static void stack_layout_dump(FILE *file, be_stack_layout_t *frame)
-{
-       int i, j, n;
-
-       ir_fprintf(file, "initial offset: %d\n", frame->initial_offset);
-       for (j = 0; j < N_FRAME_TYPES; ++j) {
-               ir_type *t = frame->order[j];
-
-               ir_fprintf(file, "type %d: %F size: %d\n", j, t, get_type_size_bytes(t));
-               for (i = 0, n = get_compound_n_members(t); i < n; ++i) {
-                       ir_entity *ent = get_compound_member(t, i);
-                       ir_fprintf(file, "\t%F int ofs: %d glob ofs: %d\n", ent, get_entity_offset_bytes(ent), get_stack_entity_offset(frame, ent, 0));
-               }
-       }
-}
-#endif
-
-/**
- * Returns non-zero if the call argument at given position
- * is transfered on the stack.
- */
-static INLINE int is_on_stack(be_abi_call_t *call, int pos)
-{
-       be_abi_call_arg_t *arg = get_call_arg(call, 0, pos);
-       return arg && !arg->in_reg;
-}
-
 /*
    ____      _ _
   / ___|__ _| | |___
@@ -398,76 +408,93 @@ static INLINE int is_on_stack(be_abi_call_t *call, int pos)
  * @param curr_sp The stack pointer node to use.
  * @return The stack pointer after the call.
  */
-static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp, ir_node *alloca_copy)
+static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp)
 {
-       ir_graph *irg              = env->birg->irg;
-       const arch_env_t *arch_env = env->birg->main_env->arch_env;
-       const arch_isa_t *isa      = arch_env->isa;
-       ir_type *mt                = get_Call_type(irn);
+       ir_graph *irg              = get_irn_irg(irn);
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
+       ir_type *call_tp           = get_Call_type(irn);
        ir_node *call_ptr          = get_Call_ptr(irn);
-       int n_params               = get_method_n_params(mt);
+       int n_params               = get_method_n_params(call_tp);
        ir_node *curr_mem          = get_Call_mem(irn);
        ir_node *bl                = get_nodes_block(irn);
-       pset *results              = pset_new_ptr(8);
-       pset *caller_save          = pset_new_ptr(8);
-       pset *states               = pset_new_ptr(2);
        int stack_size             = 0;
-       int stack_dir              = arch_isa_stack_dir(isa);
-       const arch_register_t *sp  = arch_isa_sp(isa);
+       int stack_dir              = arch_env->stack_dir;
+       const arch_register_t *sp  = arch_env->sp;
        be_abi_call_t *call        = be_abi_call_new(sp->reg_class);
        ir_mode *mach_mode         = sp->reg_class->mode;
-       struct obstack *obst       = &env->obst;
        int no_alloc               = call->flags.bits.frame_is_setup_on_call;
+       int n_res                  = get_method_n_ress(call_tp);
+       int do_seq                 = call->flags.bits.store_args_sequential && !no_alloc;
 
        ir_node *res_proj  = NULL;
-       int curr_res_proj  = pn_Call_max;
        int n_reg_params   = 0;
        int n_stack_params = 0;
        int n_ins;
 
-       ir_node *low_call;
-       ir_node **in;
-       ir_node **res_projs;
-       const arch_register_t *reg;
-       const ir_edge_t *edge;
-       int *reg_param_idxs;
-       int *stack_param_idx;
-       int i, n;
+       pset_new_t              destroyed_regs, states;
+       pset_new_iterator_t     iter;
+       ir_node                *low_call;
+       ir_node               **in;
+       ir_node               **res_projs;
+       int                     n_reg_results = 0;
+       const arch_register_t  *reg;
+       const ir_edge_t        *edge;
+       int                    *reg_param_idxs;
+       int                    *stack_param_idx;
+       int                     i, n, destroy_all_regs;
+       dbg_info               *dbgi;
+
+       pset_new_init(&destroyed_regs);
+       pset_new_init(&states);
 
        /* Let the isa fill out the abi description for that call node. */
-       arch_isa_get_call_abi(isa, mt, call);
+       arch_env_get_call_abi(arch_env, call_tp, call);
 
        /* Insert code to put the stack arguments on the stack. */
        assert(get_Call_n_params(irn) == n_params);
+       stack_param_idx = ALLOCAN(int, n_params);
        for (i = 0; i < n_params; ++i) {
-               be_abi_call_arg_t *arg = get_call_arg(call, 0, i);
+               be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 0);
                assert(arg);
                if (arg->on_stack) {
-                       int arg_size = get_type_size_bytes(get_method_param_type(mt, i));
+                       int arg_size = get_type_size_bytes(get_method_param_type(call_tp, i));
 
                        stack_size += round_up2(arg->space_before, arg->alignment);
                        stack_size += round_up2(arg_size, arg->alignment);
                        stack_size += round_up2(arg->space_after, arg->alignment);
-                       obstack_int_grow(obst, i);
-                       ++n_stack_params;
+
+                       stack_param_idx[n_stack_params++] = i;
                }
        }
-       stack_param_idx = obstack_finish(obst);
 
        /* Collect all arguments which are passed in registers. */
+       reg_param_idxs = ALLOCAN(int, n_params);
        for (i = 0; i < n_params; ++i) {
-               be_abi_call_arg_t *arg = get_call_arg(call, 0, i);
+               be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 0);
                if (arg && arg->in_reg) {
-                       obstack_int_grow(obst, i);
-                       ++n_reg_params;
+                       reg_param_idxs[n_reg_params++] = i;
                }
        }
-       reg_param_idxs = obstack_finish(obst);
 
+       /*
+        * If the stack is decreasing and we do not want to store sequentially,
+        * or someone else allocated the call frame
+        * we allocate as much space on the stack all parameters need, by
+        * moving the stack pointer along the stack's direction.
+        *
+        * Note: we also have to do this for stack_size == 0, because we may have
+        * to adjust stack alignment for the call.
+        */
+       if (stack_dir < 0 && !do_seq && !no_alloc) {
+               curr_sp = be_new_IncSP(sp, bl, curr_sp, stack_size, 1);
+       }
+
+       dbgi = get_irn_dbg_info(irn);
        /* If there are some parameters which shall be passed on the stack. */
        if (n_stack_params > 0) {
-               int curr_ofs      = 0;
-               int do_seq        = call->flags.bits.store_args_sequential && !no_alloc;
+               int       curr_ofs = 0;
+               ir_node **in       = ALLOCAN(ir_node*, n_stack_params+1);
+               unsigned  n_in     = 0;
 
                /*
                 * Reverse list of stack parameters if call arguments are from left to right.
@@ -483,34 +510,18 @@ static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp, i
                        }
                }
 
-               /*
-                * If the stack is decreasing and we do not want to store sequentially,
-                * or someone else allocated the call frame
-                * we allocate as much space on the stack all parameters need, by
-                * moving the stack pointer along the stack's direction.
-                */
-               if (stack_dir < 0 && !do_seq && !no_alloc) {
-                       curr_sp = be_new_IncSP(sp, irg, bl, curr_sp, stack_size);
-                       if (alloca_copy) {
-                               add_irn_dep(curr_sp, alloca_copy);
-                               alloca_copy = NULL;
-                       }
-               }
-
+               curr_mem = get_Call_mem(irn);
                if (! do_seq) {
-                       obstack_ptr_grow(obst, get_Call_mem(irn));
-                       curr_mem = new_NoMem();
-               } else {
-                       curr_mem = get_Call_mem(irn);
+                       in[n_in++] = curr_mem;
                }
 
                for (i = 0; i < n_stack_params; ++i) {
                        int p                  = stack_param_idx[i];
-                       be_abi_call_arg_t *arg = get_call_arg(call, 0, p);
+                       be_abi_call_arg_t *arg = get_call_arg(call, 0, p, 0);
                        ir_node *param         = get_Call_param(irn, p);
                        ir_node *addr          = curr_sp;
                        ir_node *mem           = NULL;
-                       ir_type *param_type    = get_method_param_type(mt, p);
+                       ir_type *param_type    = get_method_param_type(call_tp, p);
                        int param_size         = get_type_size_bytes(param_type) + arg->space_after;
 
                        /*
@@ -520,42 +531,37 @@ static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp, i
                         */
                        if (do_seq) {
                                curr_ofs = 0;
-                               addr = curr_sp = be_new_IncSP(sp, irg, bl, curr_sp, param_size + arg->space_before);
-                               if (alloca_copy) {
-                                       add_irn_dep(curr_sp, alloca_copy);
-                                       alloca_copy = NULL;
-                               }
+                               addr = curr_sp = be_new_IncSP(sp, bl, curr_sp,
+                                                             param_size + arg->space_before, 0);
                                add_irn_dep(curr_sp, curr_mem);
-                       }
-                       else {
+                       } else {
                                curr_ofs += arg->space_before;
                                curr_ofs =  round_up2(curr_ofs, arg->alignment);
 
                                /* Make the expression to compute the argument's offset. */
                                if (curr_ofs > 0) {
                                        ir_mode *constmode = mach_mode;
-                                       if(mode_is_reference(mach_mode)) {
+                                       if (mode_is_reference(mach_mode)) {
                                                constmode = mode_Is;
                                        }
-                                       addr = new_r_Const_long(irg, bl, constmode, curr_ofs);
-                                       addr = new_r_Add(irg, bl, curr_sp, addr, mach_mode);
+                                       addr = new_r_Const_long(irg, constmode, curr_ofs);
+                                       addr = new_r_Add(bl, curr_sp, addr, mach_mode);
                                }
                        }
 
                        /* Insert a store for primitive arguments. */
                        if (is_atomic_type(param_type)) {
                                ir_node *store;
-                               store = new_r_Store(irg, bl, curr_mem, addr, param);
-                               mem = new_r_Proj(irg, bl, store, mode_M, pn_Store_M);
-                       }
-
-                       /* Make a mem copy for compound arguments. */
-                       else {
+                               ir_node *mem_input = do_seq ? curr_mem : new_NoMem();
+                               store = new_rd_Store(dbgi, bl, mem_input, addr, param, 0);
+                               mem   = new_r_Proj(store, mode_M, pn_Store_M);
+                       } else {
+                               /* Make a mem copy for compound arguments. */
                                ir_node *copy;
 
                                assert(mode_is_reference(get_irn_mode(param)));
-                               copy = new_r_CopyB(irg, bl, curr_mem, addr, param, param_type);
-                               mem = new_r_Proj(irg, bl, copy, mode_M, pn_CopyB_M_regular);
+                               copy = new_rd_CopyB(dbgi, bl, curr_mem, addr, param, param_type);
+                               mem = new_r_Proj(copy, mode_M, pn_CopyB_M_regular);
                        }
 
                        curr_ofs += param_size;
@@ -563,237 +569,272 @@ static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp, i
                        if (do_seq)
                                curr_mem = mem;
                        else
-                               obstack_ptr_grow(obst, mem);
+                               in[n_in++] = mem;
                }
 
-               in = (ir_node **) obstack_finish(obst);
-
                /* We need the sync only, if we didn't build the stores sequentially. */
                if (! do_seq) {
                        if (n_stack_params >= 1) {
-                               curr_mem = new_r_Sync(irg, bl, n_stack_params + 1, in);
+                               curr_mem = new_r_Sync(bl, n_in, in);
                        } else {
                                curr_mem = get_Call_mem(irn);
                        }
                }
-               obstack_free(obst, in);
        }
 
-       /* Collect caller save registers */
-       for (i = 0, n = arch_isa_get_n_reg_class(isa); i < n; ++i) {
-               int j;
-               const arch_register_class_t *cls = arch_isa_get_reg_class(isa, i);
+       /* check for the return_twice property */
+       destroy_all_regs = 0;
+       if (is_SymConst_addr_ent(call_ptr)) {
+               ir_entity *ent = get_SymConst_entity(call_ptr);
+
+               if (get_entity_additional_properties(ent) & mtp_property_returns_twice)
+                       destroy_all_regs = 1;
+       } else {
+               ir_type *call_tp = get_Call_type(irn);
+
+               if (get_method_additional_properties(call_tp) & mtp_property_returns_twice)
+                       destroy_all_regs = 1;
+       }
+
+       /* Put caller save into the destroyed set and state registers in the states set */
+       for (i = 0, n = arch_env_get_n_reg_class(arch_env); i < n; ++i) {
+               unsigned j;
+               const arch_register_class_t *cls = arch_env_get_reg_class(arch_env, i);
                for (j = 0; j < cls->n_regs; ++j) {
                        const arch_register_t *reg = arch_register_for_index(cls, j);
-                       if (arch_register_type_is(reg, caller_save)) {
-                               pset_insert_ptr(caller_save, (void *) reg);
+
+                       if (destroy_all_regs || arch_register_type_is(reg, caller_save)) {
+                               if (! arch_register_type_is(reg, ignore))
+                                       pset_new_insert(&destroyed_regs, (void *) reg);
                        }
                        if (arch_register_type_is(reg, state)) {
-                               pset_insert_ptr(caller_save, (void*) reg);
-                               pset_insert_ptr(states, (void*) reg);
+                               pset_new_insert(&destroyed_regs, (void*) reg);
+                               pset_new_insert(&states, (void*) reg);
                        }
                }
        }
 
-       /* search the greatest result proj number */
+       if (destroy_all_regs) {
+               /* even if destroyed all is specified, neither SP nor FP are destroyed (else bad things will happen) */
+               pset_new_remove(&destroyed_regs, arch_env->sp);
+               pset_new_remove(&destroyed_regs, arch_env->bp);
+       }
+
+       /* search the largest result proj number */
+       res_projs = ALLOCANZ(ir_node*, n_res);
 
-       /* TODO: what if the result is NOT used? Currently there is
-        * no way to detect this later, especially there is no way to
-        * see this in the proj numbers.
-        * While this is ok for the register allocator, it is bad for
-        * backends which need to change the be_Call further (x87 simulator
-        * for instance. However for this particular case the call_type is
-        * sufficient.).
-        */
        foreach_out_edge(irn, edge) {
                const ir_edge_t *res_edge;
-               ir_node *irn = get_edge_src_irn(edge);
+               ir_node         *irn = get_edge_src_irn(edge);
 
-               if (is_Proj(irn) && get_Proj_proj(irn) == pn_Call_T_result) {
-                       res_proj = irn;
-                       foreach_out_edge(irn, res_edge) {
-                               int proj;
-                               be_abi_call_arg_t *arg;
-                               ir_node *res = get_edge_src_irn(res_edge);
-
-                               assert(is_Proj(res));
-
-                               proj = get_Proj_proj(res);
-                               arg  = get_call_arg(call, 1, proj);
-
-                               /*
-                                       shift the proj number to the right, since we will drop the
-                                       unspeakable Proj_T from the Call. Therefore, all real argument
-                                       Proj numbers must be increased by pn_be_Call_first_res
-                               */
-                               proj += pn_be_Call_first_res;
-                               set_Proj_proj(res, proj);
-                               obstack_ptr_grow(obst, res);
-
-                               if (proj > curr_res_proj)
-                                       curr_res_proj = proj;
-                               if (arg->in_reg) {
-                                       pset_remove_ptr(caller_save, arg->reg);
-                                       //pmap_insert(arg_regs, arg->reg, INT_TO_PTR(proj + 1))
-                               }
-                       }
+               if (!is_Proj(irn) || get_Proj_proj(irn) != pn_Call_T_result)
+                       continue;
+
+               foreach_out_edge(irn, res_edge) {
+                       int proj;
+                       ir_node *res = get_edge_src_irn(res_edge);
+
+                       assert(is_Proj(res));
+
+                       proj = get_Proj_proj(res);
+                       assert(proj < n_res);
+                       assert(res_projs[proj] == NULL);
+                       res_projs[proj] = res;
                }
+               res_proj = irn;
+               break;
        }
 
-       curr_res_proj++;
-       obstack_ptr_grow(obst, NULL);
-       res_projs = obstack_finish(obst);
+       /** TODO: this is not correct for cases where return values are passed
+        * on the stack, but no known ABI does this currently...
+        */
+       n_reg_results = n_res;
+
+       n_ins = 0;
+       in    = ALLOCAN(ir_node*, n_reg_params + pset_new_size(&states));
 
        /* make the back end call node and set its register requirements. */
        for (i = 0; i < n_reg_params; ++i) {
-               obstack_ptr_grow(obst, get_Call_param(irn, reg_param_idxs[i]));
+               in[n_ins++] = get_Call_param(irn, reg_param_idxs[i]);
        }
-       foreach_pset(states, reg) {
+
+       /* add state registers ins */
+       foreach_pset_new(&states, reg, iter) {
                const arch_register_class_t *cls = arch_register_get_class(reg);
 #if 0
                ir_node *regnode = be_abi_reg_map_get(env->regs, reg);
                ir_fprintf(stderr, "Adding %+F\n", regnode);
 #endif
-               ir_node *regnode = new_rd_Unknown(irg, arch_register_class_mode(cls));
-               obstack_ptr_grow(obst, regnode);
+               ir_node *regnode = new_r_Unknown(irg, arch_register_class_mode(cls));
+               in[n_ins++]      = regnode;
        }
-       n_ins = n_reg_params + pset_count(states);
-
-       in = obstack_finish(obst);
+       assert(n_ins == (int) (n_reg_params + pset_new_size(&states)));
 
+       /* ins collected, build the call */
        if (env->call->flags.bits.call_has_imm && is_SymConst(call_ptr)) {
                /* direct call */
-               low_call = be_new_Call(get_irn_dbg_info(irn), irg, bl, curr_mem,
-                                      curr_sp, curr_sp,
-                                      curr_res_proj + pset_count(caller_save), n_ins,
-                                      in, get_Call_type(irn));
+               low_call = be_new_Call(dbgi, irg, bl, curr_mem, curr_sp, curr_sp,
+                                      n_reg_results + pn_be_Call_first_res + pset_new_size(&destroyed_regs),
+                                      n_ins, in, get_Call_type(irn));
                be_Call_set_entity(low_call, get_SymConst_entity(call_ptr));
        } else {
                /* indirect call */
-               low_call = be_new_Call(get_irn_dbg_info(irn), irg, bl, curr_mem,
-                                      curr_sp, call_ptr,
-                                      curr_res_proj + pset_count(caller_save),
+               low_call = be_new_Call(dbgi, irg, bl, curr_mem, curr_sp, call_ptr,
+                                      n_reg_results + pn_be_Call_first_res + pset_new_size(&destroyed_regs),
                                       n_ins, in, get_Call_type(irn));
        }
+       be_Call_set_pop(low_call, call->pop);
+
+       /* put the call into the list of all calls for later processing */
        ARR_APP1(ir_node *, env->calls, low_call);
 
+       /* create new stack pointer */
+       curr_sp = new_r_Proj(low_call, get_irn_mode(curr_sp), pn_be_Call_sp);
+       be_set_constr_single_reg_out(low_call, pn_be_Call_sp, sp,
+                       arch_register_req_type_ignore | arch_register_req_type_produces_sp);
+       arch_set_irn_register(curr_sp, sp);
+
+       /* now handle results */
+       for (i = 0; i < n_res; ++i) {
+               int pn;
+               ir_node           *proj = res_projs[i];
+               be_abi_call_arg_t *arg  = get_call_arg(call, 1, i, 0);
+
+               /* returns values on stack not supported yet */
+               assert(arg->in_reg);
+
+               /*
+                       shift the proj number to the right, since we will drop the
+                       unspeakable Proj_T from the Call. Therefore, all real argument
+                       Proj numbers must be increased by pn_be_Call_first_res
+               */
+               pn = i + pn_be_Call_first_res;
+
+               if (proj == NULL) {
+                       ir_type *res_type = get_method_res_type(call_tp, i);
+                       ir_mode *mode     = get_type_mode(res_type);
+                       proj              = new_r_Proj(low_call, mode, pn);
+                       res_projs[i]      = proj;
+               } else {
+                       set_Proj_pred(proj, low_call);
+                       set_Proj_proj(proj, pn);
+               }
+
+               if (arg->in_reg) {
+                       pset_new_remove(&destroyed_regs, arg->reg);
+               }
+       }
+
        /*
                Set the register class of the call address to
                the backend provided class (default: stack pointer class)
        */
-       be_node_set_reg_class(low_call, be_pos_Call_ptr, call->cls_addr);
+       be_node_set_reg_class_in(low_call, be_pos_Call_ptr, call->cls_addr);
 
-       DBG((env->dbg, LEVEL_3, "\tcreated backend call %+F\n", low_call));
+       DBG((dbg, LEVEL_3, "\tcreated backend call %+F\n", low_call));
 
        /* Set the register classes and constraints of the Call parameters. */
        for (i = 0; i < n_reg_params; ++i) {
                int index = reg_param_idxs[i];
-               be_abi_call_arg_t *arg = get_call_arg(call, 0, index);
+               be_abi_call_arg_t *arg = get_call_arg(call, 0, index, 0);
                assert(arg->reg != NULL);
 
-               be_set_constr_single_reg(low_call, be_pos_Call_first_arg + i, arg->reg);
+               be_set_constr_single_reg_in(low_call, be_pos_Call_first_arg + i,
+                                           arg->reg, 0);
        }
 
        /* Set the register constraints of the results. */
-       for (i = 0; res_projs[i]; ++i) {
-               int pn = get_Proj_proj(res_projs[i]);
-
-               /* Correct Proj number since it has been adjusted! (see above) */
-               const be_abi_call_arg_t *arg = get_call_arg(call, 1, pn - pn_Call_max);
-
-               /* Matze: we need the information about the real mode for later
-                * transforms (signed/unsigend compares, stores...), so leave the fixup
-                * for the backend transform phase... */
-#if 0
-               /* correct mode */
-               const arch_register_class_t *cls = arch_register_get_class(arg->reg);
-               ir_mode *mode = arch_register_class_mode(cls);
-               set_irn_mode(irn, mode);
-#endif
+       for (i = 0; i < n_res; ++i) {
+               ir_node                 *proj = res_projs[i];
+               const be_abi_call_arg_t *arg  = get_call_arg(call, 1, i, 0);
+               int                      pn   = get_Proj_proj(proj);
 
                assert(arg->in_reg);
-               be_set_constr_single_reg(low_call, BE_OUT_POS(pn), arg->reg);
-               arch_set_irn_register(arch_env, res_projs[i], arg->reg);
+               be_set_constr_single_reg_out(low_call, pn, arg->reg, 0);
+               arch_set_irn_register(proj, arg->reg);
        }
-       obstack_free(obst, in);
        exchange(irn, low_call);
 
-       /* redirect the result projs to the lowered call instead of the Proj_T */
-       for (i = 0; res_projs[i]; ++i)
-               set_Proj_pred(res_projs[i], low_call);
-
-       /* set the now unnecessary projT to bad */
+       /* kill the ProjT node */
        if (res_proj != NULL) {
-               be_kill_node(res_proj);
+               kill_node(res_proj);
        }
 
        /* Make additional projs for the caller save registers
           and the Keep node which keeps them alive. */
-       if (pset_count(caller_save) > 0) {
+       {
                const arch_register_t *reg;
                ir_node               **in, *keep;
-               int                   i, n;
+               int                   i;
+               int                   n = 0;
+               int                   curr_res_proj = pn_be_Call_first_res + n_reg_results;
+               pset_new_iterator_t   iter;
+               int                   n_ins;
+
+               n_ins = (int)pset_new_size(&destroyed_regs) + n_reg_results + 1;
+               in    = ALLOCAN(ir_node *, n_ins);
+
+               /* also keep the stack pointer */
+               set_irn_link(curr_sp, (void*) sp);
+               in[n++] = curr_sp;
 
-               for (reg = pset_first(caller_save), n = 0; reg; reg = pset_next(caller_save), ++n) {
-                       ir_node *proj = new_r_Proj(irg, bl, low_call, reg->reg_class->mode, curr_res_proj);
+               foreach_pset_new(&destroyed_regs, reg, iter) {
+                       ir_node *proj = new_r_Proj(low_call, reg->reg_class->mode, curr_res_proj);
 
                        /* memorize the register in the link field. we need afterwards to set the register class of the keep correctly. */
-                       be_set_constr_single_reg(low_call, BE_OUT_POS(curr_res_proj), reg);
+                       be_set_constr_single_reg_out(low_call, curr_res_proj, reg, 0);
+                       arch_set_irn_register(proj, reg);
 
-                       /* a call can produce ignore registers, in this case set the flag and register for the Proj */
-                       if (arch_register_type_is(reg, ignore)) {
-                               arch_set_irn_register(env->birg->main_env->arch_env, proj, reg);
-                               be_node_set_flags(low_call, BE_OUT_POS(curr_res_proj), arch_irn_flags_ignore);
-                       }
+                       set_irn_link(proj, (void*) reg);
+                       in[n++] = proj;
+                       ++curr_res_proj;
+               }
 
-                       set_irn_link(proj, (void *) reg);
-                       obstack_ptr_grow(obst, proj);
-                       curr_res_proj++;
+               for (i = 0; i < n_reg_results; ++i) {
+                       ir_node *proj = res_projs[i];
+                       const arch_register_t *reg = arch_get_irn_register(proj);
+                       set_irn_link(proj, (void*) reg);
+                       in[n++] = proj;
                }
+               assert(n <= n_ins);
 
                /* create the Keep for the caller save registers */
-               in   = (ir_node **) obstack_finish(obst);
-               keep = be_new_Keep(NULL, irg, bl, n, in);
+               keep = be_new_Keep(bl, n, in);
                for (i = 0; i < n; ++i) {
                        const arch_register_t *reg = get_irn_link(in[i]);
-                       be_node_set_reg_class(keep, i, reg->reg_class);
+                       be_node_set_reg_class_in(keep, i, reg->reg_class);
                }
-               obstack_free(obst, in);
        }
 
        /* Clean up the stack. */
+       assert(stack_size >= call->pop);
+       stack_size -= call->pop;
+
        if (stack_size > 0) {
                ir_node *mem_proj = NULL;
 
                foreach_out_edge(low_call, edge) {
                        ir_node *irn = get_edge_src_irn(edge);
-                       if(is_Proj(irn) && get_Proj_proj(irn) == pn_Call_M) {
+                       if (is_Proj(irn) && get_Proj_proj(irn) == pn_Call_M) {
                                mem_proj = irn;
                                break;
                        }
                }
 
                if (! mem_proj) {
-                       mem_proj = new_r_Proj(irg, bl, low_call, mode_M, pn_Call_M);
+                       mem_proj = new_r_Proj(low_call, mode_M, pn_be_Call_M_regular);
                        keep_alive(mem_proj);
                }
-
-                /* Clean up the stack frame if we allocated it */
-               if (! no_alloc) {
-                       curr_sp = be_new_IncSP(sp, irg, bl, curr_sp, -stack_size);
-                       add_irn_dep(curr_sp, mem_proj);
-                       if (alloca_copy) {
-                               add_irn_dep(curr_sp, alloca_copy);
-                               alloca_copy = NULL;
-                       }
-               }
+       }
+       /* Clean up the stack frame or revert alignment fixes if we allocated it */
+       if (! no_alloc) {
+               curr_sp = be_new_IncSP(sp, bl, curr_sp, -stack_size, 0);
        }
 
        be_abi_call_free(call);
-       obstack_free(obst, stack_param_idx);
-       del_pset(results);
-       del_pset(states);
-       del_pset(caller_save);
+
+       pset_new_destroy(&states);
+       pset_new_destroy(&destroyed_regs);
 
        return curr_sp;
 }
@@ -803,23 +844,31 @@ static ir_node *adjust_call(be_abi_irg_t *env, ir_node *irn, ir_node *curr_sp, i
  *
  * @param alignment  the minimum stack alignment
  * @param size       the node containing the non-aligned size
- * @param irg        the irg where new nodes are allocated on
- * @param irg        the block where new nodes are allocated on
+ * @param block      the block where new nodes are allocated on
  * @param dbg        debug info for new nodes
  *
  * @return a node representing the aligned size
  */
-static ir_node *adjust_alloc_size(unsigned stack_alignment, ir_node *size, ir_graph *irg, ir_node *block, dbg_info *dbg) {
+static ir_node *adjust_alloc_size(unsigned stack_alignment, ir_node *size,
+                                  ir_node *block, dbg_info *dbg)
+{
        if (stack_alignment > 1) {
-               ir_mode *mode = get_irn_mode(size);
-               tarval *tv = new_tarval_from_long(stack_alignment-1, mode);
-               ir_node *mask = new_r_Const(irg, block, mode, tv);
+               ir_mode  *mode;
+               tarval   *tv;
+               ir_node  *mask;
+               ir_graph *irg;
+
+               assert(is_po2(stack_alignment));
 
-               size = new_rd_Add(dbg, irg, block, size, mask, mode);
+               mode = get_irn_mode(size);
+               tv   = new_tarval_from_long(stack_alignment-1, mode);
+               irg  = get_Block_irg(block);
+               mask = new_r_Const(irg, tv);
+               size = new_rd_Add(dbg, block, size, mask, mode);
 
                tv   = new_tarval_from_long(-(long)stack_alignment, mode);
-               mask = new_r_Const(irg, block, mode, tv);
-               size = new_rd_And(dbg, irg, block, size, mask, mode);
+               mask = new_r_Const(irg, tv);
+               size = new_rd_And(dbg, block, size, mask, mode);
        }
        return size;
 }
@@ -827,35 +876,31 @@ static ir_node *adjust_alloc_size(unsigned stack_alignment, ir_node *size, ir_gr
  * Adjust an alloca.
  * The alloca is transformed into a back end alloca node and connected to the stack nodes.
  */
-static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp, ir_node **result_copy)
+static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp)
 {
-       ir_node *block;
-       ir_graph *irg;
-       ir_node *alloc_mem;
-       ir_node *alloc_res;
-       ir_type *type;
-       dbg_info *dbg;
+       ir_node          *block     = get_nodes_block(alloc);
+       ir_graph         *irg       = get_Block_irg(block);
+       const arch_env_t *arch_env  = be_get_irg_arch_env(irg);
+       ir_node          *alloc_mem = NULL;
+       ir_node          *alloc_res = NULL;
+       ir_type          *type      = get_Alloc_type(alloc);
+       dbg_info         *dbg;
 
        const ir_edge_t *edge;
-       ir_node *new_alloc, *size, *addr, *copy, *ins[2];
+       ir_node *new_alloc;
+       ir_node *count;
+       ir_node *size;
+       ir_node *ins[2];
        unsigned stack_alignment;
 
-       if (get_Alloc_where(alloc) != stack_alloc) {
-               assert(0);
-               return alloc;
-       }
-
-       block = get_nodes_block(alloc);
-       irg = get_irn_irg(block);
-       alloc_mem = NULL;
-       alloc_res = NULL;
-       type = get_Alloc_type(alloc);
+       /* all non-stack Alloc nodes should already be lowered before the backend */
+       assert(get_Alloc_where(alloc) == stack_alloc);
 
        foreach_out_edge(alloc, edge) {
                ir_node *irn = get_edge_src_irn(edge);
 
                assert(is_Proj(irn));
-               switch(get_Proj_proj(irn)) {
+               switch (get_Proj_proj(irn)) {
                case pn_Alloc_M:
                        alloc_mem = irn;
                        break;
@@ -875,41 +920,40 @@ static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp
                return curr_sp;
        }
 
-       dbg = get_irn_dbg_info(alloc);
+       dbg   = get_irn_dbg_info(alloc);
+       count = get_Alloc_count(alloc);
 
-       /* we might need to multiply the size with the element size */
-       if(type != get_unknown_type() && get_type_size_bytes(type) != 1) {
-               tarval *tv = new_tarval_from_long(get_type_size_bytes(type), mode_Iu);
-               ir_node *cnst = new_rd_Const(dbg, irg, block, mode_Iu, tv);
-               ir_node *mul = new_rd_Mul(dbg, irg, block, get_Alloc_size(alloc),
-                                         cnst, mode_Iu);
-               size = mul;
+       /* we might need to multiply the count with the element size */
+       if (type != firm_unknown_type && get_type_size_bytes(type) != 1) {
+               ir_mode *mode = get_irn_mode(count);
+               tarval *tv    = new_tarval_from_long(get_type_size_bytes(type),
+                                                    mode);
+               ir_node *cnst = new_rd_Const(dbg, irg, tv);
+               size          = new_rd_Mul(dbg, block, count, cnst, mode);
        } else {
-               size = get_Alloc_size(alloc);
+               size = count;
        }
 
        /* The stack pointer will be modified in an unknown manner.
           We cannot omit it. */
        env->call->flags.bits.try_omit_fp = 0;
 
-       /* FIXME: size must be here round up for the stack alignment, but
-          this must be transmitted from the backend. */
-       stack_alignment = 4;
-       size = adjust_alloc_size(stack_alignment, size, irg, block, dbg);
-       new_alloc = be_new_AddSP(env->isa->sp, irg, block, curr_sp, size);
+       stack_alignment = 1 << arch_env->stack_alignment;
+       size            = adjust_alloc_size(stack_alignment, size, block, dbg);
+       new_alloc       = be_new_AddSP(arch_env->sp, block, curr_sp, size);
        set_irn_dbg_info(new_alloc, dbg);
 
-       if(alloc_mem != NULL) {
+       if (alloc_mem != NULL) {
                ir_node *addsp_mem;
                ir_node *sync;
 
-               addsp_mem = new_r_Proj(irg, block, new_alloc, mode_M, pn_be_AddSP_M);
+               addsp_mem = new_r_Proj(new_alloc, mode_M, pn_be_AddSP_M);
 
                /* We need to sync the output mem of the AddSP with the input mem
                   edge into the alloc node. */
                ins[0] = get_Alloc_mem(alloc);
                ins[1] = addsp_mem;
-               sync = new_r_Sync(irg, block, 2, ins);
+               sync = new_r_Sync(block, 2, ins);
 
                exchange(alloc_mem, sync);
        }
@@ -919,22 +963,10 @@ static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp
        /* fix projnum of alloca res */
        set_Proj_proj(alloc_res, pn_be_AddSP_res);
 
-       addr = env->isa->stack_dir < 0 ? alloc_res : curr_sp;
-
-       /* copy the address away, since it could be used after further stack pointer modifications. */
-       /* Let it point curr_sp just for the moment, I'll reroute it in a second. */
-       *result_copy = copy = be_new_Copy(env->isa->sp->reg_class, irg, block, curr_sp);
-
-       /* Let all users of the Alloc() result now point to the copy. */
-       edges_reroute(alloc_res, copy, irg);
-
-       /* Rewire the copy appropriately. */
-       set_irn_n(copy, be_pos_Copy_op, addr);
-
-       curr_sp = alloc_res;
+       curr_sp = new_r_Proj(new_alloc,  get_irn_mode(curr_sp), pn_be_AddSP_sp);
 
        return curr_sp;
-}  /* adjust_alloc */
+}
 
 /**
  * Adjust a Free.
@@ -942,55 +974,47 @@ static ir_node *adjust_alloc(be_abi_irg_t *env, ir_node *alloc, ir_node *curr_sp
  */
 static ir_node *adjust_free(be_abi_irg_t *env, ir_node *free, ir_node *curr_sp)
 {
-       ir_node *block;
-       ir_graph *irg;
-       ir_node *subsp, *mem, *res, *size, *sync;
-       ir_type *type;
+       ir_node          *block    = get_nodes_block(free);
+       ir_graph         *irg      = get_irn_irg(free);
+       ir_type          *type     = get_Free_type(free);
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
+       ir_mode          *sp_mode  = arch_env->sp->reg_class->mode;
+       dbg_info         *dbg      = get_irn_dbg_info(free);
+       ir_node  *subsp, *mem, *res, *size, *sync;
        ir_node *in[2];
-       ir_mode *sp_mode;
        unsigned stack_alignment;
-       dbg_info *dbg;
 
-       if (get_Free_where(free) != stack_alloc) {
-               assert(0);
-               return free;
-       }
-
-       block = get_nodes_block(free);
-       irg = get_irn_irg(block);
-       type = get_Free_type(free);
-       sp_mode = env->isa->sp->reg_class->mode;
-       dbg = get_irn_dbg_info(free);
+       /* all non-stack-alloc Free nodes should already be lowered before the
+        * backend phase */
+       assert(get_Free_where(free) == stack_alloc);
 
        /* we might need to multiply the size with the element size */
-       if(type != get_unknown_type() && get_type_size_bytes(type) != 1) {
+       if (type != firm_unknown_type && get_type_size_bytes(type) != 1) {
                tarval *tv = new_tarval_from_long(get_type_size_bytes(type), mode_Iu);
-               ir_node *cnst = new_rd_Const(dbg, irg, block, mode_Iu, tv);
-               ir_node *mul = new_rd_Mul(dbg, irg, block, get_Free_size(free),
+               ir_node *cnst = new_rd_Const(dbg, irg, tv);
+               ir_node *mul = new_rd_Mul(dbg, block, get_Free_size(free),
                                          cnst, mode_Iu);
                size = mul;
        } else {
                size = get_Free_size(free);
        }
 
-       /* FIXME: size must be here round up for the stack alignment, but
-          this must be transmitted from the backend. */
-       stack_alignment = 4;
-       size = adjust_alloc_size(stack_alignment, size, irg, block, dbg);
+       stack_alignment = 1 << arch_env->stack_alignment;
+       size            = adjust_alloc_size(stack_alignment, size, block, dbg);
 
        /* The stack pointer will be modified in an unknown manner.
           We cannot omit it. */
        env->call->flags.bits.try_omit_fp = 0;
-       subsp = be_new_SubSP(env->isa->sp, irg, block, curr_sp, size);
+       subsp = be_new_SubSP(arch_env->sp, block, curr_sp, size);
        set_irn_dbg_info(subsp, dbg);
 
-       mem = new_r_Proj(irg, block, subsp, mode_M, pn_be_SubSP_M);
-       res = new_r_Proj(irg, block, subsp, sp_mode, pn_be_SubSP_res);
+       mem = new_r_Proj(subsp, mode_M, pn_be_SubSP_M);
+       res = new_r_Proj(subsp, sp_mode, pn_be_SubSP_sp);
 
        /* we need to sync the memory */
        in[0] = get_Free_mem(free);
        in[1] = mem;
-       sync = new_r_Sync(irg, block, 2, in);
+       sync = new_r_Sync(block, 2, in);
 
        /* and make the AddSP dependent on the former memory */
        add_irn_dep(subsp, get_Free_mem(free));
@@ -1000,37 +1024,7 @@ static ir_node *adjust_free(be_abi_irg_t *env, ir_node *free, ir_node *curr_sp)
        curr_sp = res;
 
        return curr_sp;
-}  /* adjust_free */
-
-/* the following function is replaced by the usage of the heights module */
-#if 0
-/**
- * Walker for dependent_on().
- * This function searches a node tgt recursively from a given node
- * but is restricted to the given block.
- * @return 1 if tgt was reachable from curr, 0 if not.
- */
-static int check_dependence(ir_node *curr, ir_node *tgt, ir_node *bl)
-{
-       int n, i;
-
-       if (get_nodes_block(curr) != bl)
-               return 0;
-
-       if (curr == tgt)
-               return 1;
-
-       /* Phi functions stop the recursion inside a basic block */
-       if (! is_Phi(curr)) {
-               for(i = 0, n = get_irn_arity(curr); i < n; ++i) {
-                       if (check_dependence(get_irn_n(curr, i), tgt, bl))
-                               return 1;
-               }
-       }
-
-       return 0;
 }
-#endif /* if 0 */
 
 /**
  * Check if a node is somehow data dependent on another one.
@@ -1063,20 +1057,23 @@ static int cmp_call_dependency(const void *c1, const void *c2)
        if (dependent_on(n2, n1))
                return 1;
 
-       return 0;
+       /* The nodes have no depth order, but we need a total order because qsort()
+        * is not stable. */
+       return get_irn_idx(n1) - get_irn_idx(n2);
 }
 
 /**
- * Walker: links all Call/alloc/Free nodes to the Block they are contained.
+ * Walker: links all Call/Alloc/Free nodes to the Block they are contained.
+ * Clears the irg_is_leaf flag if a Call is detected.
  */
-static void link_calls_in_block_walker(ir_node *irn, void *data)
+static void link_ops_in_block_walker(ir_node *irn, void *data)
 {
-       ir_opcode code = get_irn_opcode(irn);
+       be_abi_irg_t *env  = data;
+       ir_opcode     code = get_irn_opcode(irn);
 
        if (code == iro_Call ||
-               (code == iro_Alloc && get_Alloc_where(irn) == stack_alloc) ||
-               (code == iro_Free && get_Free_where(irn) == stack_alloc)) {
-               be_abi_irg_t *env = data;
+          (code == iro_Alloc && get_Alloc_where(irn) == stack_alloc) ||
+          (code == iro_Free && get_Free_where(irn) == stack_alloc)) {
                ir_node *bl       = get_nodes_block(irn);
                void *save        = get_irn_link(bl);
 
@@ -1086,211 +1083,153 @@ static void link_calls_in_block_walker(ir_node *irn, void *data)
                set_irn_link(irn, save);
                set_irn_link(bl, irn);
        }
+
+       if (code == iro_Builtin && get_Builtin_kind(irn) == ir_bk_return_address) {
+               ir_node       *param = get_Builtin_param(irn, 0);
+               tarval        *tv    = get_Const_tarval(param);
+               unsigned long  value = get_tarval_long(tv);
+               /* use ebp, so the climbframe algo works... */
+               if (value > 0) {
+                       env->call->flags.bits.try_omit_fp = 0;
+               }
+       }
 }
 
 /**
  * Block-walker:
- * Process all Call nodes inside a basic block.
+ * Process all Call/Alloc/Free nodes inside a basic block.
  * Note that the link field of the block must contain a linked list of all
  * Call nodes inside the Block. We first order this list according to data dependency
  * and that connect the calls together.
  */
-static void process_calls_in_block(ir_node *bl, void *data)
+static void process_ops_in_block(ir_node *bl, void *data)
 {
-       be_abi_irg_t *env = data;
-       ir_node *curr_sp  = env->init_sp;
-       ir_node *irn;
-       int n;
+       be_abi_irg_t   *env     = data;
+       ir_node        *curr_sp = env->init_sp;
+       ir_node        *irn;
+       ir_node       **nodes;
+       int             n;
+       int             n_nodes;
 
-       for(irn = get_irn_link(bl), n = 0; irn; irn = get_irn_link(irn), ++n)
-               obstack_ptr_grow(&env->obst, irn);
+       n_nodes = 0;
+       for (irn = get_irn_link(bl); irn != NULL; irn = get_irn_link(irn)) {
+               ++n_nodes;
+       }
+
+       nodes = ALLOCAN(ir_node*, n_nodes);
+       for (irn = get_irn_link(bl), n = 0; irn; irn = get_irn_link(irn), ++n) {
+               nodes[n] = irn;
+       }
 
        /* If there were call nodes in the block. */
-       if(n > 0) {
+       if (n > 0) {
                ir_node *keep;
-               ir_node **nodes;
-               ir_node *copy = NULL;
                int i;
 
-               nodes = obstack_finish(&env->obst);
-
                /* order the call nodes according to data dependency */
-               qsort(nodes, n, sizeof(nodes[0]), cmp_call_dependency);
+               qsort(nodes, n_nodes, sizeof(nodes[0]), cmp_call_dependency);
 
-               for(i = n - 1; i >= 0; --i) {
+               for (i = n_nodes - 1; i >= 0; --i) {
                        ir_node *irn = nodes[i];
 
-                       DBG((env->dbg, LEVEL_3, "\tprocessing call %+F\n", irn));
-                       switch(get_irn_opcode(irn)) {
+                       DBG((dbg, LEVEL_3, "\tprocessing call %+F\n", irn));
+                       switch (get_irn_opcode(irn)) {
                        case iro_Call:
-                               curr_sp = adjust_call(env, irn, curr_sp, copy);
+                               if (! be_omit_fp) {
+                                       /* The stack pointer will be modified due to a call. */
+                                       env->call->flags.bits.try_omit_fp = 0;
+                               }
+                               curr_sp = adjust_call(env, irn, curr_sp);
                                break;
                        case iro_Alloc:
-                               curr_sp = adjust_alloc(env, irn, curr_sp, &copy);
+                               if (get_Alloc_where(irn) == stack_alloc)
+                                       curr_sp = adjust_alloc(env, irn, curr_sp);
                                break;
                        case iro_Free:
-                               curr_sp = adjust_free(env, irn, curr_sp);
+                               if (get_Free_where(irn) == stack_alloc)
+                                       curr_sp = adjust_free(env, irn, curr_sp);
                                break;
                        default:
                                panic("invalid call");
-                               break;
                        }
                }
 
-               obstack_free(&env->obst, nodes);
-
-               /* Keep the last stack state in the block by tying it to Keep node */
-               if(curr_sp != env->init_sp) {
+               /* Keep the last stack state in the block by tying it to Keep node,
+                * the proj from calls is already kept */
+               if (curr_sp != env->init_sp &&
+                   !(is_Proj(curr_sp) && be_is_Call(get_Proj_pred(curr_sp)))) {
                        nodes[0] = curr_sp;
-                       keep     = be_new_Keep(env->isa->sp->reg_class, get_irn_irg(bl),
-                                              bl, 1, nodes);
+                       keep     = be_new_Keep(bl, 1, nodes);
                        pmap_insert(env->keep_map, bl, keep);
                }
        }
 
        set_irn_link(bl, curr_sp);
-}  /* process_calls_in_block */
+}
 
 /**
  * Adjust all call nodes in the graph to the ABI conventions.
  */
-static void process_calls(be_abi_irg_t *env)
+static void process_calls(ir_graph *irg)
 {
-       ir_graph *irg = env->birg->irg;
+       be_abi_irg_t *abi = be_get_irg_abi(irg);
 
-       env->call->flags.bits.irg_is_leaf = 1;
-       irg_walk_graph(irg, firm_clear_link, link_calls_in_block_walker, env);
+       abi->call->flags.bits.irg_is_leaf = 1;
+       irg_walk_graph(irg, firm_clear_link, link_ops_in_block_walker, abi);
 
-       ir_heights = heights_new(env->birg->irg);
-       irg_block_walk_graph(irg, NULL, process_calls_in_block, env);
+       ir_heights = heights_new(irg);
+       irg_block_walk_graph(irg, NULL, process_ops_in_block, abi);
        heights_free(ir_heights);
 }
 
-#if 0 /*
-static ir_node *setup_frame(be_abi_irg_t *env)
-{
-       const arch_isa_t *isa = env->birg->main_env->arch_env->isa;
-       const arch_register_t *sp = isa->sp;
-       const arch_register_t *bp = isa->bp;
-       be_abi_call_flags_bits_t flags = env->call->flags.bits;
-       ir_graph *irg      = env->birg->irg;
-       ir_node *bl        = get_irg_start_block(irg);
-       ir_node *no_mem    = get_irg_no_mem(irg);
-       ir_node *old_frame = get_irg_frame(irg);
-       ir_node *stack     = pmap_get(env->regs, (void *) sp);
-       ir_node *frame     = pmap_get(env->regs, (void *) bp);
-
-       int stack_nr       = get_Proj_proj(stack);
-
-       if(flags.try_omit_fp) {
-               stack = be_new_IncSP(sp, irg, bl, stack, no_mem, BE_STACK_FRAME_SIZE_EXPAND);
-               frame = stack;
-       }
-
-       else {
-               frame = be_new_Copy(bp->reg_class, irg, bl, stack);
-
-               be_node_set_flags(frame, -1, arch_irn_flags_dont_spill);
-               if(!flags.fp_free) {
-                       be_set_constr_single_reg(frame, -1, bp);
-                       be_node_set_flags(frame, -1, arch_irn_flags_ignore);
-                       arch_set_irn_register(env->birg->main_env->arch_env, frame, bp);
-               }
-
-               stack = be_new_IncSP(sp, irg, bl, stack, frame, BE_STACK_FRAME_SIZE_EXPAND);
-       }
-
-       be_node_set_flags(env->reg_params, -(stack_nr + 1), arch_irn_flags_ignore);
-       env->init_sp = stack;
-       set_irg_frame(irg, frame);
-       edges_reroute(old_frame, frame, irg);
-
-       return frame;
-}
-
-static void clearup_frame(be_abi_irg_t *env, ir_node *ret, pmap *reg_map, struct obstack *obst)
-{
-       const arch_isa_t *isa = env->birg->main_env->arch_env->isa;
-       const arch_register_t *sp = isa->sp;
-       const arch_register_t *bp = isa->bp;
-       ir_graph *irg      = env->birg->irg;
-       ir_node *ret_mem   = get_Return_mem(ret);
-       ir_node *frame     = get_irg_frame(irg);
-       ir_node *bl        = get_nodes_block(ret);
-       ir_node *stack     = get_irn_link(bl);
-
-       pmap_entry *ent;
-
-       if(env->call->flags.bits.try_omit_fp) {
-               stack = be_new_IncSP(sp, irg, bl, stack, ret_mem, -BE_STACK_FRAME_SIZE_SHRINK);
-       }
-
-       else {
-               stack = be_new_SetSP(sp, irg, bl, stack, frame, ret_mem);
-               be_set_constr_single_reg(stack, -1, sp);
-               be_node_set_flags(stack, -1, arch_irn_flags_ignore);
-       }
-
-       pmap_foreach(env->regs, ent) {
-               const arch_register_t *reg = ent->key;
-               ir_node *irn               = ent->value;
-
-               if(reg == sp)
-                       obstack_ptr_grow(&env->obst, stack);
-               else if(reg == bp)
-                       obstack_ptr_grow(&env->obst, frame);
-               else if(arch_register_type_is(reg, callee_save) || arch_register_type_is(reg, ignore))
-                       obstack_ptr_grow(obst, irn);
-       }
-}
-*/
-#endif
-
 /**
  * Computes the stack argument layout type.
  * Changes a possibly allocated value param type by moving
  * entities to the stack layout type.
  *
- * @param env          the ABI environment
- * @param call         the current call ABI
- * @param method_type  the method type
- * @param param_map    an array mapping method arguments to the stack layout type
+ * @param env           the ABI environment
+ * @param call          the current call ABI
+ * @param method_type   the method type
+ * @param val_param_tp  the value parameter type, will be destroyed
+ * @param param_map     an array mapping method arguments to the stack layout type
  *
  * @return the stack argument layout type
  */
-static ir_type *compute_arg_type(be_abi_irg_t *env, be_abi_call_t *call, ir_type *method_type, ir_entity ***param_map)
+static ir_type *compute_arg_type(be_abi_irg_t *env, ir_graph *irg,
+                                 be_abi_call_t *call,
+                                                                ir_type *method_type, ir_type *val_param_tp,
+                                                                ir_entity ***param_map)
 {
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
        int dir  = env->call->flags.bits.left_to_right ? 1 : -1;
-       int inc  = env->birg->main_env->arch_env->isa->stack_dir * dir;
+       int inc  = arch_env->stack_dir * dir;
        int n    = get_method_n_params(method_type);
        int curr = inc > 0 ? 0 : n - 1;
+       struct obstack *obst = be_get_be_obst(irg);
        int ofs  = 0;
 
        char buf[128];
        ir_type *res;
        int i;
-       ir_type *val_param_tp = get_method_value_param_type(method_type);
-       ident *id = get_entity_ident(get_irg_entity(env->birg->irg));
+       ident *id = get_entity_ident(get_irg_entity(irg));
        ir_entity **map;
 
-       *param_map = map = obstack_alloc(&env->obst, n * sizeof(ir_entity *));
-       res = new_type_struct(mangle_u(id, new_id_from_chars("arg_type", 8)));
+       *param_map = map = OALLOCN(obst, ir_entity*, n);
+       res = new_type_struct(id_mangle_u(id, new_id_from_chars("arg_type", 8)));
        for (i = 0; i < n; ++i, curr += inc) {
                ir_type *param_type    = get_method_param_type(method_type, curr);
-               be_abi_call_arg_t *arg = get_call_arg(call, 0, curr);
+               be_abi_call_arg_t *arg = get_call_arg(call, 0, curr, 1);
 
                map[i] = NULL;
                if (arg->on_stack) {
-                       if (val_param_tp) {
-                               /* the entity was already created, move it to the param type */
-                               arg->stack_ent = get_method_value_param_ent(method_type, i);
-                               remove_struct_member(val_param_tp, arg->stack_ent);
-                               set_entity_owner(arg->stack_ent, res);
-                               add_struct_member(res, arg->stack_ent);
-                               /* must be automatic to set a fixed layout */
-                               set_entity_allocation(arg->stack_ent, allocation_automatic);
-                       }
-                       else {
+                       if (val_param_tp != NULL) {
+                               /* the entity was already created, create a copy in the param type */
+                               ir_entity *val_ent = get_method_value_param_ent(method_type, i);
+                               arg->stack_ent = copy_entity_own(val_ent, res);
+                               set_entity_link(val_ent, arg->stack_ent);
+                               set_entity_link(arg->stack_ent, NULL);
+                       } else {
+                               /* create a new entity */
                                snprintf(buf, sizeof(buf), "param_%d", i);
                                arg->stack_ent = new_entity(res, new_id_from_str(buf), param_type);
                        }
@@ -1307,50 +1246,6 @@ static ir_type *compute_arg_type(be_abi_irg_t *env, be_abi_call_t *call, ir_type
        return res;
 }
 
-#if 0
-static void create_register_perms(const arch_isa_t *isa, ir_graph *irg, ir_node *bl, pmap *regs)
-{
-       int i, j, n;
-       struct obstack obst;
-
-       obstack_init(&obst);
-
-       /* Create a Perm after the RegParams node to delimit it. */
-       for(i = 0, n = arch_isa_get_n_reg_class(isa); i < n; ++i) {
-               const arch_register_class_t *cls = arch_isa_get_reg_class(isa, i);
-               ir_node *perm;
-               ir_node **in;
-               int n_regs;
-
-               for(n_regs = 0, j = 0; j < cls->n_regs; ++j) {
-                       const arch_register_t *reg = &cls->regs[j];
-                       ir_node *irn = pmap_get(regs, (void *) reg);
-
-                       if(irn && !arch_register_type_is(reg, ignore)) {
-                               n_regs++;
-                               obstack_ptr_grow(&obst, irn);
-                               set_irn_link(irn, (void *) reg);
-                       }
-               }
-
-               obstack_ptr_grow(&obst, NULL);
-               in = obstack_finish(&obst);
-               if(n_regs > 0) {
-                       perm = be_new_Perm(cls, irg, bl, n_regs, in);
-                       for(j = 0; j < n_regs; ++j) {
-                               ir_node *arg = in[j];
-                               arch_register_t *reg = get_irn_link(arg);
-                               pmap_insert(regs, reg, arg);
-                               be_set_constr_single_reg(perm, BE_OUT_POS(j), reg);
-                       }
-               }
-               obstack_free(&obst, in);
-       }
-
-       obstack_free(&obst, NULL);
-}
-#endif
-
 typedef struct {
        const arch_register_t *reg;
        ir_node *irn;
@@ -1361,86 +1256,84 @@ static int cmp_regs(const void *a, const void *b)
        const reg_node_map_t *p = a;
        const reg_node_map_t *q = b;
 
-       if(p->reg->reg_class == q->reg->reg_class)
+       if (p->reg->reg_class == q->reg->reg_class)
                return p->reg->index - q->reg->index;
        else
                return p->reg->reg_class - q->reg->reg_class;
 }
 
-static reg_node_map_t *reg_map_to_arr(struct obstack *obst, pmap *reg_map)
+static void reg_map_to_arr(reg_node_map_t *res, pmap *reg_map)
 {
        pmap_entry *ent;
        int n = pmap_count(reg_map);
        int i = 0;
-       reg_node_map_t *res = obstack_alloc(obst, n * sizeof(res[0]));
 
-       pmap_foreach(reg_map, ent) {
+       foreach_pmap(reg_map, ent) {
                res[i].reg = ent->key;
                res[i].irn = ent->value;
                i++;
        }
 
        qsort(res, n, sizeof(res[0]), cmp_regs);
-       return res;
 }
 
 /**
  * Creates a barrier.
  */
-static ir_node *create_barrier(be_abi_irg_t *env, ir_node *bl, ir_node **mem, pmap *regs, int in_req)
+static ir_node *create_barrier(ir_node *bl, ir_node **mem, pmap *regs,
+                               int in_req)
 {
-       ir_graph *irg = env->birg->irg;
-       int n_regs    = pmap_count(regs);
-       int n;
-       ir_node *irn;
-       ir_node **in;
+       int             n_regs = pmap_count(regs);
+       int             n;
+       ir_node        *irn;
+       ir_node       **in;
        reg_node_map_t *rm;
 
-       rm = reg_map_to_arr(&env->obst, regs);
-
-       for(n = 0; n < n_regs; ++n)
-               obstack_ptr_grow(&env->obst, rm[n].irn);
-
-       if(mem) {
-               obstack_ptr_grow(&env->obst, *mem);
-               n++;
+       in = ALLOCAN(ir_node*, n_regs+1);
+       rm = ALLOCAN(reg_node_map_t, n_regs);
+       reg_map_to_arr(rm, regs);
+       for (n = 0; n < n_regs; ++n) {
+               in[n] = rm[n].irn;
        }
 
-       in = (ir_node **) obstack_finish(&env->obst);
-       irn = be_new_Barrier(irg, bl, n, in);
-       obstack_free(&env->obst, in);
-
-       for(n = 0; n < n_regs; ++n) {
-               const arch_register_t *reg = rm[n].reg;
-               int flags                  = 0;
-               int pos                    = BE_OUT_POS(n);
-               ir_node *proj;
+       if (mem) {
+               in[n++] = *mem;
+       }
 
-               proj = new_r_Proj(irg, bl, irn, get_irn_mode(rm[n].irn), n);
-               be_node_set_reg_class(irn, n, reg->reg_class);
-               if(in_req)
-                       be_set_constr_single_reg(irn, n, reg);
-               be_set_constr_single_reg(irn, pos, reg);
-               be_node_set_reg_class(irn, pos, reg->reg_class);
-               arch_set_irn_register(env->birg->main_env->arch_env, proj, reg);
+       irn = be_new_Barrier(bl, n, in);
 
-               /* if the proj projects a ignore register or a node which is set to ignore, propagate this property. */
-               if(arch_register_type_is(reg, ignore) || arch_irn_is(env->birg->main_env->arch_env, in[n], ignore))
-                       flags |= arch_irn_flags_ignore;
+       for (n = 0; n < n_regs; ++n) {
+               ir_node               *pred     = rm[n].irn;
+               const arch_register_t *reg      = rm[n].reg;
+               arch_register_type_t   add_type = 0;
+               ir_node               *proj;
+               const backend_info_t  *info;
 
-               if(arch_irn_is(env->birg->main_env->arch_env, in[n], modify_sp))
-                       flags |= arch_irn_flags_modify_sp;
+               /* stupid workaround for now... as not all nodes report register
+                * requirements. */
+               info = be_get_info(skip_Proj(pred));
+               if (info != NULL && info->out_infos != NULL) {
+                       const arch_register_req_t *ireq = arch_get_register_req_out(pred);
+                       if (ireq->type & arch_register_req_type_ignore)
+                               add_type |= arch_register_req_type_ignore;
+                       if (ireq->type & arch_register_req_type_produces_sp)
+                               add_type |= arch_register_req_type_produces_sp;
+               }
 
-               be_node_set_flags(irn, pos, flags);
+               proj = new_r_Proj(irn, get_irn_mode(pred), n);
+               be_node_set_reg_class_in(irn, n, reg->reg_class);
+               if (in_req)
+                       be_set_constr_single_reg_in(irn, n, reg, 0);
+               be_set_constr_single_reg_out(irn, n, reg, add_type);
+               arch_set_irn_register(proj, reg);
 
                pmap_insert(regs, (void *) reg, proj);
        }
 
-       if(mem) {
-               *mem = new_r_Proj(irg, bl, irn, mode_M, n);
+       if (mem) {
+               *mem = new_r_Proj(irn, mode_M, n);
        }
 
-       obstack_free(&env->obst, rm);
        return irn;
 }
 
@@ -1453,19 +1346,23 @@ static ir_node *create_barrier(be_abi_irg_t *env, ir_node *bl, ir_node **mem, pm
  * @param mem     the current memory
  * @param n_res   number of return results
  */
-static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl, ir_node *mem, int n_res) {
-       be_abi_call_t *call = env->call;
-       const arch_isa_t *isa = env->birg->main_env->arch_env->isa;
-
+static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl,
+               ir_node *mem, int n_res)
+{
+       be_abi_call_t    *call     = env->call;
+       ir_graph         *irg      = get_Block_irg(bl);
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
+       dbg_info *dbgi;
        pmap *reg_map  = pmap_create();
        ir_node *keep  = pmap_get(env->keep_map, bl);
        int in_max;
        ir_node *ret;
        int i, n;
+       unsigned pop;
        ir_node **in;
        ir_node *stack;
        const arch_register_t **regs;
-       pmap_entry *ent ;
+       pmap_entry *ent;
 
        /*
                get the valid stack node in this block.
@@ -1474,32 +1371,32 @@ static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl, i
                it then. Else we use the stack from the start block and let
                the ssa construction fix the usage.
        */
-       stack = be_abi_reg_map_get(env->regs, isa->sp);
+       stack = be_abi_reg_map_get(env->regs, arch_env->sp);
        if (keep) {
                stack = get_irn_n(keep, 0);
-               be_kill_node(keep);
-               remove_End_keepalive(get_irg_end(env->birg->irg), keep);
+               kill_node(keep);
+               remove_End_keepalive(get_irg_end(irg), keep);
        }
 
        /* Insert results for Return into the register map. */
-       for(i = 0; i < n_res; ++i) {
+       for (i = 0; i < n_res; ++i) {
                ir_node *res           = get_Return_res(irn, i);
-               be_abi_call_arg_t *arg = get_call_arg(call, 1, i);
+               be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
                assert(arg->in_reg && "return value must be passed in register");
                pmap_insert(reg_map, (void *) arg->reg, res);
        }
 
        /* Add uses of the callee save registers. */
-       pmap_foreach(env->regs, ent) {
+       foreach_pmap(env->regs, ent) {
                const arch_register_t *reg = ent->key;
-               if(arch_register_type_is(reg, callee_save) || arch_register_type_is(reg, ignore))
+               if (arch_register_type_is(reg, callee_save) || arch_register_type_is(reg, ignore))
                        pmap_insert(reg_map, ent->key, ent->value);
        }
 
-       be_abi_reg_map_set(reg_map, isa->sp, stack);
+       be_abi_reg_map_set(reg_map, arch_env->sp, stack);
 
        /* Make the Epilogue node and call the arch's epilogue maker. */
-       create_barrier(env, bl, &mem, reg_map, 1);
+       create_barrier(bl, &mem, reg_map, 1);
        call->cb->epilogue(env->cb, bl, &mem, reg_map);
 
        /*
@@ -1508,19 +1405,19 @@ static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl, i
        */
        in_max = pmap_count(reg_map) + n_res + 2;
 
-       in   = obstack_alloc(&env->obst, in_max * sizeof(in[0]));
-       regs = obstack_alloc(&env->obst, in_max * sizeof(regs[0]));
+       in   = ALLOCAN(ir_node*,               in_max);
+       regs = ALLOCAN(arch_register_t const*, in_max);
 
        in[0]   = mem;
-       in[1]   = be_abi_reg_map_get(reg_map, isa->sp);
+       in[1]   = be_abi_reg_map_get(reg_map, arch_env->sp);
        regs[0] = NULL;
-       regs[1] = isa->sp;
+       regs[1] = arch_env->sp;
        n       = 2;
 
        /* clear SP entry, since it has already been grown. */
-       pmap_insert(reg_map, (void *) isa->sp, NULL);
-       for(i = 0; i < n_res; ++i) {
-               be_abi_call_arg_t *arg = get_call_arg(call, 1, i);
+       pmap_insert(reg_map, (void *) arch_env->sp, NULL);
+       for (i = 0; i < n_res; ++i) {
+               be_abi_call_arg_t *arg = get_call_arg(call, 1, i, 1);
 
                in[n]     = be_abi_reg_map_get(reg_map, arg->reg);
                regs[n++] = arg->reg;
@@ -1530,60 +1427,128 @@ static ir_node *create_be_return(be_abi_irg_t *env, ir_node *irn, ir_node *bl, i
        }
 
        /* grow the rest of the stuff. */
-       pmap_foreach(reg_map, ent) {
-               if(ent->value) {
+       foreach_pmap(reg_map, ent) {
+               if (ent->value) {
                        in[n]     = ent->value;
                        regs[n++] = ent->key;
                }
        }
 
        /* The in array for the new back end return is now ready. */
-       ret = be_new_Return(irn ? get_irn_dbg_info(irn) : NULL, env->birg->irg, bl, n_res, n, in);
+       if (irn != NULL) {
+               dbgi = get_irn_dbg_info(irn);
+       } else {
+               dbgi = NULL;
+       }
+       /* we have to pop the shadow parameter in in case of struct returns */
+       pop = call->pop;
+       ret = be_new_Return(dbgi, irg, bl, n_res, pop, n, in);
 
        /* Set the register classes of the return's parameter accordingly. */
-       for(i = 0; i < n; ++i)
-               if(regs[i])
-                       be_node_set_reg_class(ret, i, regs[i]->reg_class);
+       for (i = 0; i < n; ++i) {
+               if (regs[i] == NULL)
+                       continue;
+
+               be_node_set_reg_class_in(ret, i, regs[i]->reg_class);
+       }
 
        /* Free the space of the Epilog's in array and the register <-> proj map. */
-       obstack_free(&env->obst, in);
        pmap_destroy(reg_map);
 
        return ret;
 }
 
+typedef struct ent_pos_pair ent_pos_pair;
+struct ent_pos_pair {
+       ir_entity    *ent;   /**< a value param entity */
+       int          pos;    /**< its parameter number */
+       ent_pos_pair *next;  /**< for linking */
+};
+
 typedef struct lower_frame_sels_env_t {
-       be_abi_irg_t *env;
-       ir_entity    *value_param_list;  /**< the list of all value param entities */
+       ent_pos_pair *value_param_list;          /**< the list of all value param entities */
+       ir_node      *frame;                     /**< the current frame */
+       const arch_register_class_t *sp_class;   /**< register class of the stack pointer */
+       const arch_register_class_t *link_class; /**< register class of the link pointer */
+       ir_type      *value_tp;                  /**< the value type if any */
+       ir_type      *frame_tp;                  /**< the frame type */
+       int          static_link_pos;            /**< argument number of the hidden static link */
 } lower_frame_sels_env_t;
 
+/**
+ * Return an entity from the backend for an value param entity.
+ *
+ * @param ent  an value param type entity
+ * @param ctx  context
+ */
+static ir_entity *get_argument_entity(ir_entity *ent, lower_frame_sels_env_t *ctx)
+{
+       ir_entity *argument_ent = get_entity_link(ent);
+
+       if (argument_ent == NULL) {
+               /* we have NO argument entity yet: This is bad, as we will
+               * need one for backing store.
+               * Create one here.
+               */
+               ir_type *frame_tp = ctx->frame_tp;
+               unsigned offset   = get_type_size_bytes(frame_tp);
+               ir_type  *tp      = get_entity_type(ent);
+               unsigned align    = get_type_alignment_bytes(tp);
+
+               offset += align - 1;
+               offset &= ~(align - 1);
+
+               argument_ent = copy_entity_own(ent, frame_tp);
+
+               /* must be automatic to set a fixed layout */
+               set_entity_offset(argument_ent, offset);
+               offset += get_type_size_bytes(tp);
+
+               set_type_size_bytes(frame_tp, offset);
+               set_entity_link(ent, argument_ent);
+       }
+       return argument_ent;
+}
 /**
  * Walker: Replaces Sels of frame type and
  * value param type entities by FrameAddress.
  * Links all used entities.
  */
-static void lower_frame_sels_walker(ir_node *irn, void *data) {
+static void lower_frame_sels_walker(ir_node *irn, void *data)
+{
        lower_frame_sels_env_t *ctx = data;
 
        if (is_Sel(irn)) {
-               ir_graph *irg        = current_ir_graph;
-               ir_node  *frame      = get_irg_frame(irg);
-               ir_node  *param_base = get_irg_value_param_base(irg);
-               ir_node  *ptr        = get_Sel_ptr(irn);
+               ir_node *ptr = get_Sel_ptr(irn);
 
-               if (ptr == frame || ptr == param_base) {
-                       be_abi_irg_t *env = ctx->env;
+               if (ptr == ctx->frame) {
                        ir_entity    *ent = get_Sel_entity(irn);
                        ir_node      *bl  = get_nodes_block(irn);
                        ir_node      *nw;
+                       int          pos = 0;
+                       int          is_value_param = 0;
+
+                       if (get_entity_owner(ent) == ctx->value_tp) {
+                               is_value_param = 1;
 
-                       nw = be_new_FrameAddr(env->isa->sp->reg_class, irg, bl, frame, ent);
+                               /* replace by its copy from the argument type */
+                               pos = get_struct_member_index(ctx->value_tp, ent);
+                               ent = get_argument_entity(ent, ctx);
+                       }
+
+                       nw = be_new_FrameAddr(ctx->sp_class, bl, ctx->frame, ent);
                        exchange(irn, nw);
 
-                       /* check, if it's a param sel and if have not seen this entity before */
-                       if (ptr == param_base && get_entity_link(ent) == NULL) {
+                       /* check, if it's a param Sel and if have not seen this entity before */
+                       if (is_value_param && get_entity_link(ent) == NULL) {
+                               ent_pos_pair pair;
+
+                               pair.ent  = ent;
+                               pair.pos  = pos;
+                               pair.next = NULL;
+                               ARR_APP1(ent_pos_pair, ctx->value_param_list, pair);
+                               /* just a mark */
                                set_entity_link(ent, ctx->value_param_list);
-                               ctx->value_param_list = ent;
                        }
                }
        }
@@ -1600,42 +1565,35 @@ static void lower_frame_sels_walker(ir_node *irn, void *data) {
  * In the default case we move the entity to the frame type and create
  * a backing store into the first block.
  */
-static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_entity *value_param_list) {
-       be_abi_call_t *call = env->call;
-       ir_graph *irg       = env->birg->irg;
-       ir_entity *ent, *next_ent, *new_list;
-       ir_type *frame_tp;
-       DEBUG_ONLY(firm_dbg_module_t *dbg = env->dbg;)
+static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_graph *irg,
+                                            ent_pos_pair *value_param_list)
+{
+       be_abi_call_t    *call     = env->call;
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
+       ent_pos_pair  *entry, *new_list;
+       ir_type       *frame_tp;
+       int           i, n = ARR_LEN(value_param_list);
 
        new_list = NULL;
-       for (ent = value_param_list; ent; ent = next_ent) {
-               int i = get_struct_member_index(get_entity_owner(ent), ent);
-               be_abi_call_arg_t *arg = get_call_arg(call, 0, i);
+       for (i = 0; i < n; ++i) {
+               int               pos  = value_param_list[i].pos;
+               be_abi_call_arg_t *arg = get_call_arg(call, 0, pos, 1);
 
-               next_ent = get_entity_link(ent);
                if (arg->in_reg) {
-                       DBG((dbg, LEVEL_2, "\targ #%d need backing store\n", i));
-                       set_entity_link(ent, new_list);
-                       new_list = ent;
+                       DBG((dbg, LEVEL_2, "\targ #%d need backing store\n", pos));
+                       value_param_list[i].next = new_list;
+                       new_list = &value_param_list[i];
                }
        }
-       if (new_list) {
+       if (new_list != NULL) {
                /* ok, change the graph */
                ir_node *start_bl = get_irg_start_block(irg);
-               ir_node *first_bl = NULL;
-               ir_node *frame, *imem, *nmem, *store, *mem, *args, *args_bl;
-               const ir_edge_t *edge;
+               ir_node *first_bl = get_first_block_succ(start_bl);
+               ir_node *frame, *imem, *nmem, *store, *mem, *args;
                optimization_state_t state;
-               int offset;
+               unsigned offset;
 
-               foreach_block_succ(start_bl, edge) {
-                       ir_node *succ = get_edge_src_irn(edge);
-                       if (start_bl != succ) {
-                               first_bl = succ;
-                               break;
-                       }
-               }
-               assert(first_bl);
+               assert(first_bl && first_bl != start_bl);
                /* we had already removed critical edges, so the following
                   assertion should be always true. */
                assert(get_Block_n_cfgpreds(first_bl) == 1);
@@ -1646,7 +1604,7 @@ static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_entity *value_
 
                save_optimization_state(&state);
                set_optimize(0);
-               nmem = new_r_Proj(irg, first_bl, get_irg_start(irg), mode_M, pn_Start_M);
+               nmem = new_r_Proj(get_irg_start(irg), mode_M, pn_Start_M);
                restore_optimization_state(&state);
 
                /* reroute all edges to the new memory source */
@@ -1655,22 +1613,21 @@ static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_entity *value_
                store   = NULL;
                mem     = imem;
                args    = get_irg_args(irg);
-               args_bl = get_nodes_block(args);
-               for (ent = new_list; ent; ent = get_entity_link(ent)) {
-                       int     i     = get_struct_member_index(get_entity_owner(ent), ent);
-                       ir_type *tp   = get_entity_type(ent);
+               for (entry = new_list; entry != NULL; entry = entry->next) {
+                       int     i     = entry->pos;
+                       ir_type *tp   = get_entity_type(entry->ent);
                        ir_mode *mode = get_type_mode(tp);
                        ir_node *addr;
 
                        /* address for the backing store */
-                       addr = be_new_FrameAddr(env->isa->sp->reg_class, irg, first_bl, frame, ent);
+                       addr = be_new_FrameAddr(arch_env->sp->reg_class, first_bl, frame, entry->ent);
 
                        if (store)
-                               mem = new_r_Proj(irg, first_bl, store, mode_M, pn_Store_M);
+                               mem = new_r_Proj(store, mode_M, pn_Store_M);
 
                        /* the backing store itself */
-                       store = new_r_Store(irg, first_bl, mem, addr,
-                                           new_r_Proj(irg, args_bl, args, mode, i));
+                       store = new_r_Store(first_bl, mem, addr,
+                                           new_r_Proj(args, mode, i), 0);
                }
                /* the new memory Proj gets the last Proj from store */
                set_Proj_pred(nmem, store);
@@ -1679,20 +1636,32 @@ static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_entity *value_
                /* move all entities to the frame type */
                frame_tp = get_irg_frame_type(irg);
                offset   = get_type_size_bytes(frame_tp);
-               for (ent = new_list; ent; ent = get_entity_link(ent)) {
-                       ir_type *tp = get_entity_type(ent);
-                       int align = get_type_alignment_bytes(tp);
-
-                       offset += align - 1;
-                       offset &= -align;
-                       set_entity_owner(ent, frame_tp);
-                       add_class_member(frame_tp, ent);
-                       /* must be automatic to set a fixed layout */
-                       set_entity_allocation(ent, allocation_automatic);
-                       set_entity_offset(ent, offset);
-                       offset += get_type_size_bytes(tp);
+
+               /* we will add new entities: set the layout to undefined */
+               assert(get_type_state(frame_tp) == layout_fixed);
+               set_type_state(frame_tp, layout_undefined);
+               for (entry = new_list; entry != NULL; entry = entry->next) {
+                       ir_entity *ent = entry->ent;
+
+                       /* If the entity is still on the argument type, move it to the
+                        * frame type.
+                        * This happens if the value_param type was build due to compound
+                        * params. */
+                       if (get_entity_owner(ent) != frame_tp) {
+                               ir_type  *tp   = get_entity_type(ent);
+                               unsigned align = get_type_alignment_bytes(tp);
+
+                               offset += align - 1;
+                               offset &= ~(align - 1);
+                               set_entity_owner(ent, frame_tp);
+                               /* must be automatic to set a fixed layout */
+                               set_entity_offset(ent, offset);
+                               offset += get_type_size_bytes(tp);
+                       }
                }
                set_type_size_bytes(frame_tp, offset);
+               /* fix the layout again */
+               set_type_state(frame_tp, layout_fixed);
        }
 }
 
@@ -1701,100 +1670,189 @@ static void fix_address_of_parameter_access(be_abi_irg_t *env, ir_entity *value_
  * The backend wants to handle all blocks the same way, so we replace
  * the out cfg edge with a real jump.
  */
-static void fix_start_block(ir_node *block, void *env) {
-       int      *done = env;
+static void fix_start_block(ir_graph *irg)
+{
+       ir_node         *initial_X   = get_irg_initial_exec(irg);
+       ir_node         *start_block = get_irg_start_block(irg);
+       const ir_edge_t *edge;
+
+       assert(is_Proj(initial_X));
+
+       foreach_out_edge(initial_X, edge) {
+               ir_node *block = get_edge_src_irn(edge);
+
+               if (is_Anchor(block))
+                       continue;
+               if (block != start_block) {
+                       ir_node *jmp = new_r_Jmp(start_block);
+                       set_Block_cfgpred(block, get_edge_src_pos(edge), jmp);
+                       set_irg_initial_exec(irg, jmp);
+                       return;
+               }
+       }
+       panic("Initial exec has no follow block in %+F", irg);
+}
+
+/**
+ * Update the entity of Sels to the outer value parameters.
+ */
+static void update_outer_frame_sels(ir_node *irn, void *env)
+{
+       lower_frame_sels_env_t *ctx = env;
+       ir_node                *ptr;
+       ir_entity              *ent;
+       int                    pos = 0;
+
+       if (! is_Sel(irn))
+               return;
+       ptr = get_Sel_ptr(irn);
+       if (! is_arg_Proj(ptr))
+               return;
+       if (get_Proj_proj(ptr) != ctx->static_link_pos)
+               return;
+       ent   = get_Sel_entity(irn);
+
+       if (get_entity_owner(ent) == ctx->value_tp) {
+               /* replace by its copy from the argument type */
+               pos = get_struct_member_index(ctx->value_tp, ent);
+               ent = get_argument_entity(ent, ctx);
+               set_Sel_entity(irn, ent);
+
+               /* check, if we have not seen this entity before */
+               if (get_entity_link(ent) == NULL) {
+                       ent_pos_pair pair;
+
+                       pair.ent  = ent;
+                       pair.pos  = pos;
+                       pair.next = NULL;
+                       ARR_APP1(ent_pos_pair, ctx->value_param_list, pair);
+                       /* just a mark */
+                       set_entity_link(ent, ctx->value_param_list);
+               }
+       }
+}
+
+/**
+ * Fix access to outer local variables.
+ */
+static void fix_outer_variable_access(be_abi_irg_t *env,
+                                      lower_frame_sels_env_t *ctx)
+{
        int      i;
-       ir_node  *start_block;
        ir_graph *irg;
+       (void) env;
 
-       /* we processed the start block, return */
-       if (*done)
-               return;
+       for (i = get_class_n_members(ctx->frame_tp) - 1; i >= 0; --i) {
+               ir_entity *ent = get_class_member(ctx->frame_tp, i);
 
-       irg         = get_irn_irg(block);
-       start_block = get_irg_start_block(irg);
+               if (! is_method_entity(ent))
+                       continue;
 
-       for (i = get_Block_n_cfgpreds(block) - 1; i >= 0; --i) {
-               ir_node *pred       = get_Block_cfgpred(block, i);
-               ir_node *pred_block = get_nodes_block(pred);
+               irg = get_entity_irg(ent);
+               if (irg == NULL)
+                       continue;
 
-               /* ok, we are in the block, having start as cfg predecessor */
-               if (pred_block == start_block) {
-                       ir_node *jump = new_r_Jmp(irg, pred_block);
-                       set_Block_cfgpred(block, i, jump);
-                       *done = 1;
-               }
+               /*
+                * FIXME: find the number of the static link parameter
+                * for now we assume 0 here
+                */
+               ctx->static_link_pos = 0;
+
+               irg_walk_graph(irg, NULL, update_outer_frame_sels, ctx);
        }
 }
 
 /**
  * Modify the irg itself and the frame type.
  */
-static void modify_irg(be_abi_irg_t *env)
+static void modify_irg(ir_graph *irg)
 {
-       be_abi_call_t *call       = env->call;
-       const arch_isa_t *isa     = env->birg->main_env->arch_env->isa;
-       const arch_register_t *sp = arch_isa_sp(isa);
-       ir_graph *irg             = env->birg->irg;
-       ir_node *bl               = get_irg_start_block(irg);
-       ir_node *end              = get_irg_end_block(irg);
-       ir_node *old_mem          = get_irg_initial_mem(irg);
+       be_abi_irg_t          *env          = be_get_irg_abi(irg);
+       be_abi_call_t         *call         = env->call;
+       const arch_env_t      *arch_env     = be_get_irg_arch_env(irg);
+       const arch_register_t *sp           = arch_env->sp;
+       ir_type               *method_type  = get_entity_type(get_irg_entity(irg));
+       struct obstack        *obst         = be_get_be_obst(irg);
+       be_stack_layout_t     *stack_layout = be_get_irg_stack_layout(irg);
+       ir_node *end;
+       ir_node *old_mem;
        ir_node *new_mem_proj;
        ir_node *mem;
-       ir_type *method_type      = get_entity_type(get_irg_entity(irg));
-       pset *dont_save           = pset_new_ptr(8);
 
        int n_params;
-       int i, j, n, temp;
+       int i, n;
+       unsigned j;
+       unsigned frame_size;
 
        reg_node_map_t *rm;
        const arch_register_t *fp_reg;
        ir_node *frame_pointer;
-       ir_node *barrier;
-       ir_node *reg_params_bl;
+       ir_node *start_bl;
        ir_node **args;
        ir_node *arg_tuple;
-       ir_node *value_param_base;
        const ir_edge_t *edge;
        ir_type *arg_type, *bet_type, *tp;
        lower_frame_sels_env_t ctx;
        ir_entity **param_map;
 
-       bitset_t *used_proj_nr;
-       DEBUG_ONLY(firm_dbg_module_t *dbg = env->dbg;)
-
        DBG((dbg, LEVEL_1, "introducing abi on %+F\n", irg));
 
+       /* Must fetch memory here, otherwise the start Barrier gets the wrong
+        * memory, which leads to loops in the DAG. */
+       old_mem = get_irg_initial_mem(irg);
+
+       irp_reserve_resources(irp, IR_RESOURCE_ENTITY_LINK);
+
        /* set the links of all frame entities to NULL, we use it
           to detect if an entity is already linked in the value_param_list */
-       tp = get_irg_frame_type(irg);
-       for (i = get_class_n_members(tp) - 1; i >= 0; --i)
-               set_entity_link(get_class_member(tp, i), NULL);
-
-       /* same for the value_param type */
        tp = get_method_value_param_type(method_type);
+       ctx.value_tp = tp;
        if (tp != NULL) {
-               for (i = get_struct_n_members(tp) - 1; i >= 0; --i)
-                       set_entity_link(get_struct_member(tp, i), NULL);
+               /* clear the links of the clone type, let the
+                  original entities point to its clones */
+               for (i = get_struct_n_members(tp) - 1; i >= 0; --i) {
+                       ir_entity *mem  = get_struct_member(tp, i);
+                       set_entity_link(mem, NULL);
+               }
+       }
+
+       arg_type = compute_arg_type(env, irg, call, method_type, tp, &param_map);
+
+       /* Convert the Sel nodes in the irg to frame addr nodes: */
+       ctx.value_param_list = NEW_ARR_F(ent_pos_pair, 0);
+       ctx.frame            = get_irg_frame(irg);
+       ctx.sp_class         = arch_env->sp->reg_class;
+       ctx.link_class       = arch_env->link_class;
+       ctx.frame_tp         = get_irg_frame_type(irg);
+
+       /* layout the stackframe now */
+       if (get_type_state(ctx.frame_tp) == layout_undefined) {
+               default_layout_compound_type(ctx.frame_tp);
        }
 
-       /* Convert the Sel nodes in the irg to frame load/store/addr nodes. */
-       ctx.env              = env;
-       ctx.value_param_list = NULL;
+       /* we will possible add new entities to the frame: set the layout to undefined */
+       assert(get_type_state(ctx.frame_tp) == layout_fixed);
+       set_type_state(ctx.frame_tp, layout_undefined);
+
        irg_walk_graph(irg, lower_frame_sels_walker, NULL, &ctx);
 
-       /* value_param_base anchor is not needed anymore now */
-       value_param_base = get_irg_value_param_base(irg);
-       be_kill_node(value_param_base);
-       set_irg_value_param_base(irg, new_r_Bad(irg));
+       /* fix the frame type layout again */
+       set_type_state(ctx.frame_tp, layout_fixed);
+       /* align stackframe to 4 byte */
+       frame_size = get_type_size_bytes(ctx.frame_tp);
+       if (frame_size % 4 != 0) {
+               set_type_size_bytes(ctx.frame_tp, frame_size + 4 - (frame_size % 4));
+       }
 
-       env->frame = obstack_alloc(&env->obst, sizeof(env->frame[0]));
        env->regs  = pmap_create();
 
-       used_proj_nr = bitset_alloca(1024);
-       n_params     = get_method_n_params(method_type);
-       args         = obstack_alloc(&env->obst, n_params * sizeof(args[0]));
-       memset(args, 0, n_params * sizeof(args[0]));
+       n_params = get_method_n_params(method_type);
+       args     = OALLOCNZ(obst, ir_node*, n_params);
+
+       /*
+        * for inner function we must now fix access to outer frame entities.
+        */
+       fix_outer_variable_access(env, &ctx);
 
        /* Check if a value parameter is transmitted as a register.
         * This might happen if the address of an parameter is taken which is
@@ -1806,52 +1864,59 @@ static void modify_irg(be_abi_irg_t *env)
         * In the default case we move the entity to the frame type and create
         * a backing store into the first block.
         */
-       fix_address_of_parameter_access(env, ctx.value_param_list);
+       fix_address_of_parameter_access(env, irg, ctx.value_param_list);
+
+       DEL_ARR_F(ctx.value_param_list);
+       irp_free_resources(irp, IR_RESOURCE_ENTITY_LINK);
 
        /* Fill the argument vector */
        arg_tuple = get_irg_args(irg);
        foreach_out_edge(arg_tuple, edge) {
                ir_node *irn = get_edge_src_irn(edge);
-               int nr       = get_Proj_proj(irn);
-               args[nr]     = irn;
-               DBG((dbg, LEVEL_2, "\treading arg: %d -> %+F\n", nr, irn));
+               if (! is_Anchor(irn)) {
+                       int nr       = get_Proj_proj(irn);
+                       args[nr]     = irn;
+                       DBG((dbg, LEVEL_2, "\treading arg: %d -> %+F\n", nr, irn));
+               }
        }
 
-       arg_type = compute_arg_type(env, call, method_type, &param_map);
        bet_type = call->cb->get_between_type(env->cb);
-       stack_frame_init(env->frame, arg_type, bet_type, get_irg_frame_type(irg), isa->stack_dir, param_map);
+       stack_frame_init(stack_layout, arg_type, bet_type,
+                        get_irg_frame_type(irg), arch_env->stack_dir, param_map);
+       stack_layout->sp_relative = call->flags.bits.try_omit_fp;
 
        /* Count the register params and add them to the number of Projs for the RegParams node */
-       for(i = 0; i < n_params; ++i) {
-               be_abi_call_arg_t *arg = get_call_arg(call, 0, i);
-               if(arg->in_reg && args[i]) {
+       for (i = 0; i < n_params; ++i) {
+               be_abi_call_arg_t *arg = get_call_arg(call, 0, i, 1);
+               if (arg->in_reg && args[i]) {
                        assert(arg->reg != sp && "cannot use stack pointer as parameter register");
                        assert(i == get_Proj_proj(args[i]));
 
                        /* For now, associate the register with the old Proj from Start representing that argument. */
                        pmap_insert(env->regs, (void *) arg->reg, args[i]);
-                       bitset_set(used_proj_nr, i);
                        DBG((dbg, LEVEL_2, "\targ #%d -> reg %s\n", i, arg->reg->name));
                }
        }
 
        /* Collect all callee-save registers */
-       for(i = 0, n = arch_isa_get_n_reg_class(isa); i < n; ++i) {
-               const arch_register_class_t *cls = arch_isa_get_reg_class(isa, i);
-               for(j = 0; j < cls->n_regs; ++j) {
+       for (i = 0, n = arch_env_get_n_reg_class(arch_env); i < n; ++i) {
+               const arch_register_class_t *cls = arch_env_get_reg_class(arch_env, i);
+               for (j = 0; j < cls->n_regs; ++j) {
                        const arch_register_t *reg = &cls->regs[j];
-                       if(arch_register_type_is(reg, callee_save) ||
+                       if (arch_register_type_is(reg, callee_save) ||
                                        arch_register_type_is(reg, state)) {
                                pmap_insert(env->regs, (void *) reg, NULL);
                        }
                }
        }
 
+       /* handle start block here (place a jump in the block) */
+       fix_start_block(irg);
+
        pmap_insert(env->regs, (void *) sp, NULL);
-       pmap_insert(env->regs, (void *) isa->bp, NULL);
-       reg_params_bl   = get_irg_start_block(irg);
-       env->reg_params = be_new_RegParams(irg, reg_params_bl, pmap_count(env->regs));
-       add_irn_dep(env->reg_params, get_irg_start(irg));
+       pmap_insert(env->regs, (void *) arch_env->bp, NULL);
+       start_bl   = get_irg_start_block(irg);
+       env->start = be_new_Start(NULL, start_bl, pmap_count(env->regs) + 1);
 
        /*
         * make proj nodes for the callee save registers.
@@ -1861,102 +1926,96 @@ static void modify_irg(be_abi_irg_t *env)
         * the old Proj from start for that argument.
         */
 
-       rm = reg_map_to_arr(&env->obst, env->regs);
-       for(i = 0, n = pmap_count(env->regs); i < n; ++i) {
-               arch_register_t *reg = (void *) rm[i].reg;
-               ir_mode *mode        = reg->reg_class->mode;
-               long nr              = i;
-               int pos              = BE_OUT_POS((int) nr);
-               int flags            = 0;
+       rm = ALLOCAN(reg_node_map_t, pmap_count(env->regs));
+       reg_map_to_arr(rm, env->regs);
+       for (i = 0, n = pmap_count(env->regs); i < n; ++i) {
+               arch_register_t          *reg      = (void *) rm[i].reg;
+               ir_mode                  *mode     = reg->reg_class->mode;
+               long                      nr       = i;
+               arch_register_req_type_t  add_type = 0;
+               ir_node                  *proj;
 
-               ir_node *proj;
+               if (reg == sp)
+                       add_type |= arch_register_req_type_produces_sp | arch_register_req_type_ignore;
 
                assert(nr >= 0);
-               bitset_set(used_proj_nr, nr);
-               proj = new_r_Proj(irg, reg_params_bl, env->reg_params, mode, nr);
+               proj = new_r_Proj(env->start, mode, nr + 1);
                pmap_insert(env->regs, (void *) reg, proj);
-               be_set_constr_single_reg(env->reg_params, pos, reg);
-               arch_set_irn_register(env->birg->main_env->arch_env, proj, reg);
-
-               /*
-                * If the register is an ignore register,
-                * The Proj for that register shall also be ignored during register allocation.
-                */
-               if(arch_register_type_is(reg, ignore))
-                       flags |= arch_irn_flags_ignore;
-
-               if(reg == sp)
-                       flags |= arch_irn_flags_modify_sp;
-
-               be_node_set_flags(env->reg_params, pos, flags);
+               be_set_constr_single_reg_out(env->start, nr + 1, reg, add_type);
+               arch_set_irn_register(proj, reg);
 
                DBG((dbg, LEVEL_2, "\tregister save proj #%d -> reg %s\n", nr, reg->name));
        }
-       obstack_free(&env->obst, rm);
 
        /* create a new initial memory proj */
        assert(is_Proj(old_mem));
-       new_mem_proj = new_r_Proj(irg, get_nodes_block(old_mem),
-                                 new_r_Unknown(irg, mode_T), mode_M,
-                                 get_Proj_proj(old_mem));
+       arch_set_out_register_req(env->start, 0, arch_no_register_req);
+       new_mem_proj = new_r_Proj(env->start, mode_M, 0);
        mem = new_mem_proj;
+       set_irg_initial_mem(irg, mem);
 
        /* Generate the Prologue */
-       fp_reg  = call->cb->prologue(env->cb, &mem, env->regs);
+       fp_reg = call->cb->prologue(env->cb, &mem, env->regs, &stack_layout->initial_bias);
 
        /* do the stack allocation BEFORE the barrier, or spill code
           might be added before it */
        env->init_sp = be_abi_reg_map_get(env->regs, sp);
-       env->init_sp = be_new_IncSP(sp, irg, bl, env->init_sp, BE_STACK_FRAME_SIZE_EXPAND);
+       env->init_sp = be_new_IncSP(sp, start_bl, env->init_sp, BE_STACK_FRAME_SIZE_EXPAND, 0);
        be_abi_reg_map_set(env->regs, sp, env->init_sp);
 
-       env->start_barrier = barrier = create_barrier(env, bl, &mem, env->regs, 0);
+       create_barrier(start_bl, &mem, env->regs, 0);
 
        env->init_sp = be_abi_reg_map_get(env->regs, sp);
-       arch_set_irn_register(env->birg->main_env->arch_env, env->init_sp, sp);
+       arch_set_irn_register(env->init_sp, sp);
 
        frame_pointer = be_abi_reg_map_get(env->regs, fp_reg);
        set_irg_frame(irg, frame_pointer);
        pset_insert_ptr(env->ignore_regs, fp_reg);
 
        /* rewire old mem users to new mem */
-       set_Proj_pred(new_mem_proj, get_Proj_pred(old_mem));
        exchange(old_mem, mem);
 
+       /* keep the mem (for functions with an endless loop = no return) */
+       keep_alive(mem);
+
        set_irg_initial_mem(irg, mem);
 
        /* Now, introduce stack param nodes for all parameters passed on the stack */
-       for(i = 0; i < n_params; ++i) {
+       for (i = 0; i < n_params; ++i) {
                ir_node *arg_proj = args[i];
                ir_node *repl     = NULL;
 
-               if(arg_proj != NULL) {
+               if (arg_proj != NULL) {
                        be_abi_call_arg_t *arg;
                        ir_type *param_type;
                        int     nr = get_Proj_proj(arg_proj);
                        ir_mode *mode;
 
                        nr         = MIN(nr, n_params);
-                       arg        = get_call_arg(call, 0, nr);
+                       arg        = get_call_arg(call, 0, nr, 1);
                        param_type = get_method_param_type(method_type, nr);
 
                        if (arg->in_reg) {
                                repl = pmap_get(env->regs, (void *) arg->reg);
-                       }
+                       } else if (arg->on_stack) {
+                               ir_node *addr = be_new_FrameAddr(sp->reg_class, start_bl, frame_pointer, arg->stack_ent);
 
-                       else if(arg->on_stack) {
-                               /* For atomic parameters which are actually used, we create a StackParam node. */
-                               if(is_atomic_type(param_type) && get_irn_n_edges(args[i]) > 0) {
-                                       ir_mode *mode                    = get_type_mode(param_type);
-                                       const arch_register_class_t *cls = arch_isa_get_reg_class_for_mode(isa, mode);
-                                       repl = be_new_StackParam(cls, isa->bp->reg_class, irg, reg_params_bl, mode, frame_pointer, arg->stack_ent);
-                               }
+                               /* For atomic parameters which are actually used, we create a Load node. */
+                               if (is_atomic_type(param_type) && get_irn_n_edges(args[i]) > 0) {
+                                       ir_mode *mode      = get_type_mode(param_type);
+                                       ir_mode *load_mode = arg->load_mode;
+
+                                       ir_node *load = new_r_Load(start_bl, new_NoMem(), addr, load_mode, cons_floats);
+                                       repl = new_r_Proj(load, load_mode, pn_Load_res);
 
-                               /* The stack parameter is not primitive (it is a struct or array),
-                                  we thus will create a node representing the parameter's address
-                                  on the stack. */
-                               else {
-                                       repl = be_new_FrameAddr(sp->reg_class, irg, reg_params_bl, frame_pointer, arg->stack_ent);
+                                       if (mode != load_mode) {
+                                               repl = new_r_Conv(start_bl, repl, mode);
+                                       }
+                               } else {
+                                       /* The stack parameter is not primitive (it is a struct or array),
+                                        * we thus will create a node representing the parameter's address
+                                        * on the stack. */
+                                       repl = addr;
                                }
                        }
 
@@ -1966,53 +2025,50 @@ static void modify_irg(be_abi_irg_t *env)
                           which may be wrong. Add Conv's then. */
                        mode = get_irn_mode(args[i]);
                        if (mode != get_irn_mode(repl)) {
-                               repl = new_r_Conv(irg, get_irn_n(repl, -1), repl, mode);
+                               repl = new_r_Conv(get_nodes_block(repl), repl, mode);
                        }
                        exchange(args[i], repl);
                }
        }
 
-       /* the arg proj is not needed anymore now */
-       assert(get_irn_n_edges(arg_tuple) == 0);
-       be_kill_node(arg_tuple);
-       set_irg_args(irg, new_rd_Bad(irg));
+       /* the arg proj is not needed anymore now and should be only used by the anchor */
+       assert(get_irn_n_edges(arg_tuple) == 1);
+       kill_node(arg_tuple);
+       set_irg_args(irg, new_r_Bad(irg));
 
        /* All Return nodes hang on the End node, so look for them there. */
+       end = get_irg_end_block(irg);
        for (i = 0, n = get_Block_n_cfgpreds(end); i < n; ++i) {
                ir_node *irn = get_Block_cfgpred(end, i);
 
                if (is_Return(irn)) {
-                       ir_node *ret = create_be_return(env, irn, get_nodes_block(irn), get_Return_mem(irn), get_Return_n_ress(irn));
+                       ir_node *blk = get_nodes_block(irn);
+                       ir_node *mem = get_Return_mem(irn);
+                       ir_node *ret = create_be_return(env, irn, blk, mem, get_Return_n_ress(irn));
                        exchange(irn, ret);
                }
        }
+
        /* if we have endless loops here, n might be <= 0. Do NOT create a be_Return then,
           the code is dead and will never be executed. */
-
-       del_pset(dont_save);
-       obstack_free(&env->obst, args);
-
-       /* handle start block here (place a jump in the block) */
-       temp = 0;
-       irg_block_walk_graph(irg, fix_start_block, NULL, &temp);
 }
 
 /** Fix the state inputs of calls that still hang on unknowns */
-static
-void fix_call_state_inputs(be_abi_irg_t *env)
+static void fix_call_state_inputs(ir_graph *irg)
 {
-       const arch_isa_t *isa = env->isa;
+       be_abi_irg_t     *env      = be_get_irg_abi(irg);
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
        int i, n, n_states;
        arch_register_t **stateregs = NEW_ARR_F(arch_register_t*, 0);
 
        /* Collect caller save registers */
-       n = arch_isa_get_n_reg_class(isa);
-       for(i = 0; i < n; ++i) {
-               int j;
-               const arch_register_class_t *cls = arch_isa_get_reg_class(isa, i);
-               for(j = 0; j < cls->n_regs; ++j) {
+       n = arch_env_get_n_reg_class(arch_env);
+       for (i = 0; i < n; ++i) {
+               unsigned j;
+               const arch_register_class_t *cls = arch_env_get_reg_class(arch_env, i);
+               for (j = 0; j < cls->n_regs; ++j) {
                        const arch_register_t *reg = arch_register_for_index(cls, j);
-                       if(arch_register_type_is(reg, state)) {
+                       if (arch_register_type_is(reg, state)) {
                                ARR_APP1(arch_register_t*, stateregs, (arch_register_t *)reg);
                        }
                }
@@ -2020,14 +2076,14 @@ void fix_call_state_inputs(be_abi_irg_t *env)
 
        n = ARR_LEN(env->calls);
        n_states = ARR_LEN(stateregs);
-       for(i = 0; i < n; ++i) {
+       for (i = 0; i < n; ++i) {
                int s, arity;
                ir_node *call = env->calls[i];
 
                arity = get_irn_arity(call);
 
-               /* the statereg inputs are the last n inputs of the calls */
-               for(s = 0; s < n_states; ++s) {
+               /* the state reg inputs are the last n inputs of the calls */
+               for (s = 0; s < n_states; ++s) {
                        int inp = arity - n_states + s;
                        const arch_register_t *reg = stateregs[s];
                        ir_node *regnode = be_abi_reg_map_get(env->regs, reg);
@@ -2035,72 +2091,231 @@ void fix_call_state_inputs(be_abi_irg_t *env)
                        set_irn_n(call, inp, regnode);
                }
        }
+
+       DEL_ARR_F(stateregs);
+}
+
+/**
+ * Create a trampoline entity for the given method.
+ */
+static ir_entity *create_trampoline(be_main_env_t *be, ir_entity *method)
+{
+       ir_type   *type   = get_entity_type(method);
+       ident     *old_id = get_entity_ld_ident(method);
+       ident     *id     = id_mangle3("", old_id, "$stub");
+       ir_type   *parent = be->pic_trampolines_type;
+       ir_entity *ent    = new_entity(parent, old_id, type);
+       set_entity_ld_ident(ent, id);
+       set_entity_visibility(ent, ir_visibility_private);
+
+       return ent;
+}
+
+/**
+ * Returns the trampoline entity for the given method.
+ */
+static ir_entity *get_trampoline(be_main_env_t *env, ir_entity *method)
+{
+       ir_entity *result = pmap_get(env->ent_trampoline_map, method);
+       if (result == NULL) {
+               result = create_trampoline(env, method);
+               pmap_insert(env->ent_trampoline_map, method, result);
+       }
+
+       return result;
+}
+
+static ir_entity *create_pic_symbol(be_main_env_t *be, ir_entity *entity)
+{
+       ident     *old_id = get_entity_ld_ident(entity);
+       ident     *id     = id_mangle3("", old_id, "$non_lazy_ptr");
+       ir_type   *e_type = get_entity_type(entity);
+       ir_type   *type   = new_type_pointer(e_type);
+       ir_type   *parent = be->pic_symbols_type;
+       ir_entity *ent    = new_entity(parent, old_id, type);
+       set_entity_ld_ident(ent, id);
+       set_entity_visibility(ent, ir_visibility_private);
+
+       return ent;
+}
+
+static ir_entity *get_pic_symbol(be_main_env_t *env, ir_entity *entity)
+{
+       ir_entity *result = pmap_get(env->ent_pic_symbol_map, entity);
+       if (result == NULL) {
+               result = create_pic_symbol(env, entity);
+               pmap_insert(env->ent_pic_symbol_map, entity, result);
+       }
+
+       return result;
+}
+
+
+
+/**
+ * Returns non-zero if a given entity can be accessed using a relative address.
+ */
+static int can_address_relative(ir_entity *entity)
+{
+       return get_entity_visibility(entity) != ir_visibility_external
+               && !(get_entity_linkage(entity) & IR_LINKAGE_MERGE);
+}
+
+/** patches SymConsts to work in position independent code */
+static void fix_pic_symconsts(ir_node *node, void *data)
+{
+       ir_node      *pic_base;
+       ir_node      *add;
+       ir_node      *block;
+       ir_mode      *mode;
+       ir_node      *load;
+       ir_node      *load_res;
+       ir_graph     *irg = get_irn_irg(node);
+       int           arity, i;
+       be_main_env_t *be = be_get_irg_main_env(irg);
+       (void) data;
+
+       arity = get_irn_arity(node);
+       for (i = 0; i < arity; ++i) {
+               dbg_info  *dbgi;
+               ir_node   *pred = get_irn_n(node, i);
+               ir_entity *entity;
+               ir_entity *pic_symbol;
+               ir_node   *pic_symconst;
+
+               if (!is_SymConst(pred))
+                       continue;
+
+               entity = get_SymConst_entity(pred);
+               block  = get_nodes_block(pred);
+
+               /* calls can jump to relative addresses, so we can directly jump to
+                  the (relatively) known call address or the trampoline */
+               if (i == 1 && is_Call(node)) {
+                       ir_entity *trampoline;
+                       ir_node   *trampoline_const;
+
+                       if (can_address_relative(entity))
+                               continue;
+
+                       dbgi             = get_irn_dbg_info(pred);
+                       trampoline       = get_trampoline(be, entity);
+                       trampoline_const = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
+                                                                   trampoline, NULL);
+                       set_irn_n(node, i, trampoline_const);
+                       continue;
+               }
+
+               /* everything else is accessed relative to EIP */
+               mode     = get_irn_mode(pred);
+               pic_base = arch_code_generator_get_pic_base(be_get_irg_cg(irg));
+
+               /* all ok now for locally constructed stuff */
+               if (can_address_relative(entity)) {
+                       ir_node *add = new_r_Add(block, pic_base, pred, mode);
+
+                       /* make sure the walker doesn't visit this add again */
+                       mark_irn_visited(add);
+                       set_irn_n(node, i, add);
+                       continue;
+               }
+
+               /* get entry from pic symbol segment */
+               dbgi         = get_irn_dbg_info(pred);
+               pic_symbol   = get_pic_symbol(be, entity);
+               pic_symconst = new_rd_SymConst_addr_ent(dbgi, irg, mode_P_code,
+                                                       pic_symbol, NULL);
+               add = new_r_Add(block, pic_base, pic_symconst, mode);
+               mark_irn_visited(add);
+
+               /* we need an extra indirection for global data outside our current
+                  module. The loads are always safe and can therefore float
+                  and need no memory input */
+               load     = new_r_Load(block, new_NoMem(), add, mode, cons_floats);
+               load_res = new_r_Proj(load, mode, pn_Load_res);
+
+               set_irn_n(node, i, load_res);
+       }
 }
 
-be_abi_irg_t *be_abi_introduce(be_irg_t *birg)
+be_abi_irg_t *be_abi_introduce(ir_graph *irg)
 {
-       be_abi_irg_t *env  = xmalloc(sizeof(env[0]));
-       ir_node *old_frame = get_irg_frame(birg->irg);
-       ir_graph *irg      = birg->irg;
+       be_abi_irg_t     *env         = XMALLOCZ(be_abi_irg_t);
+       ir_node          *old_frame   = get_irg_frame(irg);
+       struct obstack   *obst        = be_get_be_obst(irg);
+       be_options_t     *options     = be_get_irg_options(irg);
+       const arch_env_t *arch_env    = be_get_irg_arch_env(irg);
+       ir_entity        *entity      = get_irg_entity(irg);
+       ir_type          *method_type = get_entity_type(entity);
 
        pmap_entry *ent;
        ir_node *dummy;
-       optimization_state_t state;
        unsigned *limited_bitset;
+       arch_register_req_t *sp_req;
 
-       be_omit_fp = birg->main_env->options->omit_fp;
-
-       obstack_init(&env->obst);
+       be_omit_fp      = options->omit_fp;
+       be_omit_leaf_fp = options->omit_leaf_fp;
 
-       env->isa         = birg->main_env->arch_env->isa;
-       env->method_type = get_entity_type(get_irg_entity(irg));
-       env->call        = be_abi_call_new(env->isa->sp->reg_class);
-       arch_isa_get_call_abi(env->isa, env->method_type, env->call);
+       obstack_init(obst);
 
        env->ignore_regs  = pset_new_ptr_default();
        env->keep_map     = pmap_create();
        env->dce_survivor = new_survive_dce();
-       env->birg         = birg;
 
-       env->sp_req.type    = arch_register_req_type_limited;
-       env->sp_req.cls     = arch_register_get_class(env->isa->sp);
-       limited_bitset      = rbitset_obstack_alloc(&env->obst, env->sp_req.cls->n_regs);
-       rbitset_set(limited_bitset, arch_register_get_index(env->isa->sp));
-       env->sp_req.limited = limited_bitset;
+       sp_req = OALLOCZ(obst, arch_register_req_t);
+       env->sp_req = sp_req;
+
+       sp_req->type = arch_register_req_type_limited
+                    | arch_register_req_type_produces_sp;
+       sp_req->cls  = arch_register_get_class(arch_env->sp);
+
+       limited_bitset = rbitset_obstack_alloc(obst, sp_req->cls->n_regs);
+       rbitset_set(limited_bitset, arch_register_get_index(arch_env->sp));
+       sp_req->limited = limited_bitset;
+       if (arch_env->sp->type & arch_register_type_ignore) {
+               sp_req->type |= arch_register_req_type_ignore;
+       }
+
+       /* break here if backend provides a custom API.
+        * Note: we shouldn't have to setup any be_abi_irg_t* stuff at all,
+        * but need more cleanup to make this work
+        */
+       be_set_irg_abi(irg, env);
+       if (arch_env->custom_abi)
+               return env;
 
-       env->sp_cls_req.type  = arch_register_req_type_normal;
-       env->sp_cls_req.cls   = arch_register_get_class(env->isa->sp);
+       env->call        = be_abi_call_new(arch_env->sp->reg_class);
+       arch_env_get_call_abi(arch_env, method_type, env->call);
 
-       /* Beware: later we replace this node by the real one, ensure it is not CSE'd
-          to another Unknown or the stack pointer gets used */
-       save_optimization_state(&state);
-       set_optimize(0);
-       env->init_sp = dummy  = new_r_Unknown(irg, env->isa->sp->reg_class->mode);
-       restore_optimization_state(&state);
-       FIRM_DBG_REGISTER(env->dbg, "firm.be.abi");
+       env->init_sp = dummy = new_r_Dummy(irg, arch_env->sp->reg_class->mode);
+       env->calls   = NEW_ARR_F(ir_node*, 0);
 
-       env->calls = NEW_ARR_F(ir_node*, 0);
+       if (options->pic) {
+               irg_walk_graph(irg, fix_pic_symconsts, NULL, env);
+       }
 
        /* Lower all call nodes in the IRG. */
-       process_calls(env);
+       process_calls(irg);
 
        /*
                Beware: init backend abi call object after processing calls,
                otherwise some information might be not yet available.
        */
-       env->cb = env->call->cb->init(env->call, birg->main_env->arch_env, irg);
+       env->cb = env->call->cb->init(env->call, irg);
 
        /* Process the IRG */
-       modify_irg(env);
+       modify_irg(irg);
 
        /* fix call inputs for state registers */
-       fix_call_state_inputs(env);
+       fix_call_state_inputs(irg);
 
        /* We don't need the keep map anymore. */
        pmap_destroy(env->keep_map);
+       env->keep_map = NULL;
 
        /* calls array is not needed anymore */
        DEL_ARR_F(env->calls);
+       env->calls = NULL;
 
        /* reroute the stack origin of the calls to the true stack origin. */
        exchange(dummy, env->init_sp);
@@ -2108,7 +2323,7 @@ be_abi_irg_t *be_abi_introduce(be_irg_t *birg)
 
        /* Make some important node pointers survive the dead node elimination. */
        survive_dce_register_irn(env->dce_survivor, &env->init_sp);
-       pmap_foreach(env->regs, ent) {
+       foreach_pmap(env->regs, ent) {
                survive_dce_register_irn(env->dce_survivor, (ir_node **) &ent->value);
        }
 
@@ -2117,32 +2332,53 @@ be_abi_irg_t *be_abi_introduce(be_irg_t *birg)
        return env;
 }
 
-void be_abi_free(be_abi_irg_t *env)
+void be_abi_free(ir_graph *irg)
 {
-       be_abi_call_free(env->call);
+       be_abi_irg_t *env = be_get_irg_abi(irg);
+
+       if (env->call != NULL)
+               be_abi_call_free(env->call);
        free_survive_dce(env->dce_survivor);
-       del_pset(env->ignore_regs);
-       pmap_destroy(env->regs);
-       obstack_free(&env->obst, NULL);
+       if (env->ignore_regs != NULL)
+               del_pset(env->ignore_regs);
+       if (env->regs != NULL)
+               pmap_destroy(env->regs);
        free(env);
+
+       be_set_irg_abi(irg, NULL);
 }
 
 void be_abi_put_ignore_regs(be_abi_irg_t *abi, const arch_register_class_t *cls, bitset_t *bs)
 {
        arch_register_t *reg;
 
-       for(reg = pset_first(abi->ignore_regs); reg; reg = pset_next(abi->ignore_regs))
-               if(reg->reg_class == cls)
+       for (reg = pset_first(abi->ignore_regs); reg; reg = pset_next(abi->ignore_regs))
+               if (reg->reg_class == cls)
                        bitset_set(bs, reg->index);
 }
 
-/* Returns the stack layout from a abi environment. */
-const be_stack_layout_t *be_abi_get_stack_layout(const be_abi_irg_t *abi) {
-       return abi->frame;
+void be_abi_set_non_ignore_regs(be_abi_irg_t *abi, const arch_register_class_t *cls, unsigned *raw_bitset)
+{
+       unsigned         i;
+       arch_register_t *reg;
+
+       for (i = 0; i < cls->n_regs; ++i) {
+               if (arch_register_type_is(&cls->regs[i], ignore))
+                       continue;
+
+               rbitset_set(raw_bitset, i);
+       }
+
+       for (reg = pset_first(abi->ignore_regs); reg != NULL;
+            reg = pset_next(abi->ignore_regs)) {
+               if (reg->reg_class != cls)
+                       continue;
+
+               rbitset_clear(raw_bitset, reg->index);
+       }
 }
 
 /*
-
   _____ _        ____  _             _
  |  ___(_)_  __ / ___|| |_ __ _  ___| | __
  | |_  | \ \/ / \___ \| __/ _` |/ __| |/ /
@@ -2155,7 +2391,6 @@ typedef ir_node **node_array;
 
 typedef struct fix_stack_walker_env_t {
        node_array sp_nodes;
-       const arch_env_t *arch_env;
 } fix_stack_walker_env_t;
 
 /**
@@ -2163,49 +2398,59 @@ typedef struct fix_stack_walker_env_t {
  */
 static void collect_stack_nodes_walker(ir_node *node, void *data)
 {
-       fix_stack_walker_env_t *env = data;
+       ir_node                   *insn = node;
+       fix_stack_walker_env_t    *env = data;
+       const arch_register_req_t *req;
 
-       if (arch_irn_is(env->arch_env, node, modify_sp)) {
-               assert(get_irn_mode(node) != mode_M && get_irn_mode(node) != mode_T);
-               ARR_APP1(ir_node*, env->sp_nodes, node);
+       if (is_Proj(node)) {
+               insn = get_Proj_pred(node);
        }
+
+       if (arch_irn_get_n_outs(insn) == 0)
+               return;
+       if (get_irn_mode(node) == mode_T)
+               return;
+
+       req = arch_get_register_req_out(node);
+       if (! (req->type & arch_register_req_type_produces_sp))
+               return;
+
+       ARR_APP1(ir_node*, env->sp_nodes, node);
 }
 
-void be_abi_fix_stack_nodes(be_abi_irg_t *env)
+void be_abi_fix_stack_nodes(ir_graph *irg)
 {
+       be_abi_irg_t     *abi      = be_get_irg_abi(irg);
+       be_lv_t          *lv       = be_get_irg_liveness(irg);
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
        be_ssa_construction_env_t senv;
        int i, len;
        ir_node **phis;
-       be_irg_t *birg = env->birg;
-       be_lv_t *lv = be_get_birg_liveness(birg);
        fix_stack_walker_env_t walker_env;
-       arch_isa_t *isa;
 
        walker_env.sp_nodes = NEW_ARR_F(ir_node*, 0);
-       walker_env.arch_env = birg->main_env->arch_env;
-       isa = walker_env.arch_env->isa;
 
-       irg_walk_graph(birg->irg, collect_stack_nodes_walker, NULL, &walker_env);
+       irg_walk_graph(irg, collect_stack_nodes_walker, NULL, &walker_env);
 
        /* nothing to be done if we didn't find any node, in fact we mustn't
         * continue, as for endless loops incsp might have had no users and is bad
         * now.
         */
        len = ARR_LEN(walker_env.sp_nodes);
-       if(len == 0) {
+       if (len == 0) {
                DEL_ARR_F(walker_env.sp_nodes);
                return;
        }
 
-       be_ssa_construction_init(&senv, birg);
+       be_ssa_construction_init(&senv, irg);
        be_ssa_construction_add_copies(&senv, walker_env.sp_nodes,
                                    ARR_LEN(walker_env.sp_nodes));
        be_ssa_construction_fix_users_array(&senv, walker_env.sp_nodes,
-                                     ARR_LEN(walker_env.sp_nodes));
+                                           ARR_LEN(walker_env.sp_nodes));
 
-       if(lv != NULL) {
+       if (lv != NULL) {
                len = ARR_LEN(walker_env.sp_nodes);
-               for(i = 0; i < len; ++i) {
+               for (i = 0; i < len; ++i) {
                        be_liveness_update(lv, walker_env.sp_nodes[i]);
                }
                be_ssa_construction_update_liveness_phis(&senv, lv);
@@ -2215,97 +2460,188 @@ void be_abi_fix_stack_nodes(be_abi_irg_t *env)
 
        /* set register requirements for stack phis */
        len = ARR_LEN(phis);
-       for(i = 0; i < len; ++i) {
+       for (i = 0; i < len; ++i) {
                ir_node *phi = phis[i];
-               be_set_phi_reg_req(walker_env.arch_env, phi, &env->sp_req);
-               be_set_phi_flags(walker_env.arch_env, phi, arch_irn_flags_ignore | arch_irn_flags_modify_sp);
-               arch_set_irn_register(walker_env.arch_env, phi, env->isa->sp);
+               be_set_phi_reg_req(phi, abi->sp_req);
+               arch_set_irn_register(phi, arch_env->sp);
        }
        be_ssa_construction_destroy(&senv);
 
        DEL_ARR_F(walker_env.sp_nodes);
 }
 
-static int process_stack_bias(be_abi_irg_t *env, ir_node *bl, int bias)
+/**
+ * Fix all stack accessing operations in the block bl.
+ *
+ * @param bl         the block to process
+ * @param real_bias  the bias value
+ *
+ * @return the bias at the end of this block
+ */
+static int process_stack_bias(ir_node *bl, int real_bias)
 {
-       const arch_env_t *arch_env = env->birg->main_env->arch_env;
-       int omit_fp            = env->call->flags.bits.try_omit_fp;
-       ir_node *irn;
+       int                wanted_bias = real_bias;
+       ir_graph          *irg         = get_Block_irg(bl);
+       be_stack_layout_t *layout      = be_get_irg_stack_layout(irg);
+       bool               sp_relative = layout->sp_relative;
+       const arch_env_t  *arch_env    = be_get_irg_arch_env(irg);
+       ir_node           *irn;
 
        sched_foreach(bl, irn) {
+               int ofs;
 
                /*
                   Check, if the node relates to an entity on the stack frame.
                   If so, set the true offset (including the bias) for that
                   node.
                 */
-               ir_entity *ent = arch_get_frame_entity(arch_env, irn);
-               if(ent) {
-                       int offset = get_stack_entity_offset(env->frame, ent, bias);
-                       arch_set_frame_offset(arch_env, irn, offset);
-                       DBG((env->dbg, LEVEL_2, "%F has offset %d (including bias %d)\n", ent, offset, bias));
+               ir_entity *ent = arch_get_frame_entity(irn);
+               if (ent != NULL) {
+                       int bias   = sp_relative ? real_bias : 0;
+                       int offset = get_stack_entity_offset(layout, ent, bias);
+                       arch_set_frame_offset(irn, offset);
+                       DBG((dbg, LEVEL_2, "%F has offset %d (including bias %d)\n",
+                            ent, offset, bias));
                }
 
                /*
-                  If the node modifies the stack pointer by a constant offset,
-                  record that in the bias.
+                * If the node modifies the stack pointer by a constant offset,
+                * record that in the bias.
                 */
-               if(arch_irn_is(arch_env, irn, modify_sp)) {
-                       int ofs = arch_get_sp_bias(arch_env, irn);
-
-                       if(be_is_IncSP(irn)) {
-                               if(ofs == BE_STACK_FRAME_SIZE_EXPAND) {
-                                       ofs = get_type_size_bytes(get_irg_frame_type(env->birg->irg));
-                                       be_set_IncSP_offset(irn, ofs);
-                               } else if(ofs == BE_STACK_FRAME_SIZE_SHRINK) {
-                                       ofs = - get_type_size_bytes(get_irg_frame_type(env->birg->irg));
-                                       be_set_IncSP_offset(irn, ofs);
+               ofs = arch_get_sp_bias(irn);
+
+               if (be_is_IncSP(irn)) {
+                       /* fill in real stack frame size */
+                       if (ofs == BE_STACK_FRAME_SIZE_EXPAND) {
+                               ir_type *frame_type = get_irg_frame_type(irg);
+                               ofs = (int) get_type_size_bytes(frame_type);
+                               be_set_IncSP_offset(irn, ofs);
+                       } else if (ofs == BE_STACK_FRAME_SIZE_SHRINK) {
+                               ir_type *frame_type = get_irg_frame_type(irg);
+                               ofs = - (int)get_type_size_bytes(frame_type);
+                               be_set_IncSP_offset(irn, ofs);
+                       } else {
+                               if (be_get_IncSP_align(irn)) {
+                                       /* patch IncSP to produce an aligned stack pointer */
+                                       ir_type *between_type = layout->between_type;
+                                       int      between_size = get_type_size_bytes(between_type);
+                                       int      alignment    = 1 << arch_env->stack_alignment;
+                                       int      delta        = (real_bias + ofs + between_size) & (alignment - 1);
+                                       assert(ofs >= 0);
+                                       if (delta > 0) {
+                                               be_set_IncSP_offset(irn, ofs + alignment - delta);
+                                               real_bias += alignment - delta;
+                                       }
+                               } else {
+                                       /* adjust so real_bias corresponds with wanted_bias */
+                                       int delta = wanted_bias - real_bias;
+                                       assert(delta <= 0);
+                                       if (delta != 0) {
+                                               be_set_IncSP_offset(irn, ofs + delta);
+                                               real_bias += delta;
+                                       }
                                }
                        }
-
-                       if(omit_fp)
-                               bias += ofs;
                }
+
+               real_bias   += ofs;
+               wanted_bias += ofs;
        }
 
-       return bias;
+       assert(real_bias == wanted_bias);
+       return real_bias;
 }
 
 /**
  * A helper struct for the bias walker.
  */
 struct bias_walk {
-       be_abi_irg_t *env;     /**< The ABI irg environment. */
-       int start_block_bias;  /**< The bias at the end of the start block. */
-       ir_node *start_block;  /**< The start block of the current graph. */
+       int           start_block_bias;  /**< The bias at the end of the start block. */
+       int           between_size;
+       ir_node      *start_block;  /**< The start block of the current graph. */
 };
 
 /**
- * Block-Walker: fix all stack offsets
+ * Block-Walker: fix all stack offsets for all blocks
+ * except the start block
  */
 static void stack_bias_walker(ir_node *bl, void *data)
 {
        struct bias_walk *bw = data;
        if (bl != bw->start_block) {
-               process_stack_bias(bw->env, bl, bw->start_block_bias);
+               process_stack_bias(bl, bw->start_block_bias);
        }
 }
 
-void be_abi_fix_stack_bias(be_abi_irg_t *env)
+/**
+ * Walker: finally lower all Sels of outer frame or parameter
+ * entities.
+ */
+static void lower_outer_frame_sels(ir_node *sel, void *ctx)
 {
-       ir_graph *irg  = env->birg->irg;
-       struct bias_walk bw;
+       ir_node           *ptr;
+       ir_entity         *ent;
+       ir_type           *owner;
+       be_stack_layout_t *layout;
+       ir_graph          *irg;
+       (void) ctx;
+
+       if (! is_Sel(sel))
+               return;
+
+       ent    = get_Sel_entity(sel);
+       owner  = get_entity_owner(ent);
+       ptr    = get_Sel_ptr(sel);
+       irg    = get_irn_irg(sel);
+       layout = be_get_irg_stack_layout(irg);
+
+       if (owner == layout->frame_type || owner == layout->arg_type) {
+               /* found access to outer frame or arguments */
+               int offset = get_stack_entity_offset(layout, ent, 0);
 
-       stack_frame_compute_initial_offset(env->frame);
-       // stack_layout_dump(stdout, env->frame);
+               if (offset != 0) {
+                       ir_node  *bl   = get_nodes_block(sel);
+                       dbg_info *dbgi = get_irn_dbg_info(sel);
+                       ir_mode  *mode = get_irn_mode(sel);
+                       ir_mode  *mode_UInt = get_reference_mode_unsigned_eq(mode);
+                       ir_node  *cnst = new_r_Const_long(current_ir_graph, mode_UInt, offset);
+
+                       ptr = new_rd_Add(dbgi, bl, ptr, cnst, mode);
+               }
+               exchange(sel, ptr);
+       }
+}
+
+void be_abi_fix_stack_bias(ir_graph *irg)
+{
+       be_stack_layout_t *stack_layout = be_get_irg_stack_layout(irg);
+       ir_type           *frame_tp;
+       int               i;
+       struct bias_walk  bw;
+
+       stack_frame_compute_initial_offset(stack_layout);
+       // stack_layout_dump(stdout, stack_layout);
 
        /* Determine the stack bias at the end of the start block. */
-       bw.start_block_bias = process_stack_bias(env, get_irg_start_block(irg), 0);
+       bw.start_block_bias = process_stack_bias(get_irg_start_block(irg),
+                                                stack_layout->initial_bias);
+       bw.between_size     = get_type_size_bytes(stack_layout->between_type);
 
        /* fix the bias is all other blocks */
-       bw.env = env;
        bw.start_block = get_irg_start_block(irg);
        irg_block_walk_graph(irg, stack_bias_walker, NULL, &bw);
+
+       /* fix now inner functions: these still have Sel node to outer
+          frame and parameter entities */
+       frame_tp = get_irg_frame_type(irg);
+       for (i = get_class_n_members(frame_tp) - 1; i >= 0; --i) {
+               ir_entity *ent = get_class_member(frame_tp, i);
+               ir_graph  *irg = get_entity_irg(ent);
+
+               if (irg != NULL) {
+                       irg_walk_graph(irg, NULL, lower_outer_frame_sels, NULL);
+               }
+       }
 }
 
 ir_node *be_abi_get_callee_save_irn(be_abi_irg_t *abi, const arch_register_t *reg)
@@ -2322,15 +2658,8 @@ ir_node *be_abi_get_ignore_irn(be_abi_irg_t *abi, const arch_register_t *reg)
        return pmap_get(abi->regs, (void *) reg);
 }
 
-ir_node *be_abi_get_start_barrier(be_abi_irg_t *abi)
+BE_REGISTER_MODULE_CONSTRUCTOR(be_init_abi);
+void be_init_abi(void)
 {
-       return abi->start_barrier;
-}
-
-/**
- * Returns non-zero if the ABI has omitted the frame pointer in
- * the current graph.
- */
-int be_abi_omit_fp(const be_abi_irg_t *abi) {
-       return abi->call->flags.bits.try_omit_fp;
+       FIRM_DBG_REGISTER(dbg, "firm.be.abi");
 }