X-Git-Url: http://nsz.repo.hu/git/?a=blobdiff_plain;f=ir%2Fbe%2Fbenode.c;h=3383f3883cee69c86360cbac49986423f3c6e093;hb=4764ebb82834c3370640980c9299f0dbb1ac598d;hp=4199e8f568fd9479c10475fbc3f4827209c05421;hpb=be8b3b023b424af04a93c389d1a7672e1717a98e;p=libfirm

diff --git a/ir/be/benode.c b/ir/be/benode.c
index 4199e8f56..3383f3883 100644
--- a/ir/be/benode.c
+++ b/ir/be/benode.c
@@ -27,9 +27,7 @@
  * Backend node support for generic backend nodes.
  * This file provides Perm, Copy, Spill and Reload nodes.
  */
-#ifdef HAVE_CONFIG_H
 #include "config.h"
-#endif
 
 #include <stdlib.h>
 
@@ -39,9 +37,10 @@
 #include "util.h"
 #include "debug.h"
 #include "fourcc.h"
-#include "offset.h"
 #include "bitfiddle.h"
 #include "raw_bitset.h"
+#include "error.h"
+#include "array_t.h"
 
 #include "irop_t.h"
 #include "irmode_t.h"
@@ -53,26 +52,17 @@
 
 #include "be_t.h"
 #include "belive_t.h"
-#include "besched_t.h"
-#include "benode_t.h"
-#include "bearch_t.h"
+#include "besched.h"
+#include "benode.h"
+#include "bearch.h"
 
 #include "beirgmod.h"
 
-#define OUT_POS(x) (-((x) + 1))
-
 #define get_irn_attr(irn) get_irn_generic_attr(irn)
 #define get_irn_attr_const(irn) get_irn_generic_attr_const(irn)
 
 typedef struct {
-	arch_register_req_t req;
-	arch_irn_flags_t    flags;
-} be_req_t;
-
-typedef struct {
-	const arch_register_t *reg;
-	be_req_t               req;
-	be_req_t               in_req;
+	const arch_register_req_t *in_req;
 } be_reg_data_t;
 
 /** The generic be nodes attribute type. */
@@ -129,10 +119,9 @@ ir_op *op_be_Return;
 ir_op *op_be_IncSP;
 ir_op *op_be_AddSP;
 ir_op *op_be_SubSP;
-ir_op *op_be_RegParams;
+ir_op *op_be_Start;
 ir_op *op_be_FrameAddr;
 ir_op *op_be_Barrier;
-ir_op *op_be_Unwind;
 
 static const ir_op_ops be_node_op_ops;
 
@@ -148,63 +137,46 @@ static const ir_op_ops be_node_op_ops;
 #define K   irop_flag_keep
 #define M   irop_flag_uses_memory
 
-static int be_reqs_equal(const be_req_t *req1, const be_req_t *req2)
-{
-	if(!reg_reqs_equal(&req1->req, &req2->req))
-		return 0;
-	if(req1->flags != req2->flags)
-		return 0;
-
-	return 1;
-}
-
 /**
  * Compare two be node attributes.
  *
  * @return zero if both attributes are identically
  */
-static int _node_cmp_attr(const be_node_attr_t *a, const be_node_attr_t *b) {
-	int i, len;
+static int node_cmp_attr(ir_node *a, ir_node *b)
+{
+	const be_node_attr_t *a_attr = get_irn_attr_const(a);
+	const be_node_attr_t *b_attr = get_irn_attr_const(b);
+	int i, len = ARR_LEN(a_attr->reg_data);
 
-	if (ARR_LEN(a->reg_data) != ARR_LEN(b->reg_data))
+	if (len != ARR_LEN(b_attr->reg_data))
 		return 1;
 
-	len = ARR_LEN(a->reg_data);
-	for (i = 0; i < len; ++i) {
-		if (a->reg_data[i].reg != b->reg_data[i].reg ||
-				!be_reqs_equal(&a->reg_data[i].in_req, &b->reg_data[i].in_req) ||
-			    !be_reqs_equal(&a->reg_data[i].req,    &b->reg_data[i].req))
+	if (!be_nodes_equal(a, b))
+		return 1;
+
+	for (i = len - 1; i >= 0; --i) {
+		if (!reg_reqs_equal(a_attr->reg_data[i].in_req,
+		                    b_attr->reg_data[i].in_req))
 			return 1;
 	}
 
 	return 0;
 }
 
-/**
- * Compare the node attributes of two be_node's.
- *
- * @return zero if both nodes have identically attributes
- */
-static int node_cmp_attr(ir_node *a, ir_node *b) {
-	const be_node_attr_t *a_attr = get_irn_attr_const(a);
-	const be_node_attr_t *b_attr = get_irn_attr_const(b);
-
-	return _node_cmp_attr(a_attr, b_attr);
-}
-
 /**
  * Compare the attributes of two be_FrameAddr nodes.
  *
  * @return zero if both nodes have identically attributes
  */
-static int FrameAddr_cmp_attr(ir_node *a, ir_node *b) {
+static int FrameAddr_cmp_attr(ir_node *a, ir_node *b)
+{
 	const be_frame_attr_t *a_attr = get_irn_attr_const(a);
 	const be_frame_attr_t *b_attr = get_irn_attr_const(b);
 
 	if (a_attr->ent != b_attr->ent || a_attr->offset != b_attr->offset)
 		return 1;
 
-	return _node_cmp_attr(&a_attr->node_attr, &b_attr->node_attr);
+	return node_cmp_attr(a, b);
 }
 
 /**
@@ -212,7 +184,8 @@ static int FrameAddr_cmp_attr(ir_node *a, ir_node *b) {
  *
  * @return zero if both nodes have identically attributes
  */
-static int Return_cmp_attr(ir_node *a, ir_node *b) {
+static int Return_cmp_attr(ir_node *a, ir_node *b)
+{
 	const be_return_attr_t *a_attr = get_irn_attr_const(a);
 	const be_return_attr_t *b_attr = get_irn_attr_const(b);
 
@@ -223,7 +196,7 @@ static int Return_cmp_attr(ir_node *a, ir_node *b) {
 	if (a_attr->emit_pop != b_attr->emit_pop)
 		return 1;
 
-	return _node_cmp_attr(&a_attr->node_attr, &b_attr->node_attr);
+	return node_cmp_attr(a, b);
 }
 
 /**
@@ -231,14 +204,15 @@ static int Return_cmp_attr(ir_node *a, ir_node *b) {
  *
  * @return zero if both nodes have identically attributes
  */
-static int IncSP_cmp_attr(ir_node *a, ir_node *b) {
+static int IncSP_cmp_attr(ir_node *a, ir_node *b)
+{
 	const be_incsp_attr_t *a_attr = get_irn_attr_const(a);
 	const be_incsp_attr_t *b_attr = get_irn_attr_const(b);
 
 	if (a_attr->offset != b_attr->offset)
 		return 1;
 
-	return _node_cmp_attr(&a_attr->node_attr, &b_attr->node_attr);
+	return node_cmp_attr(a, b);
 }
 
 /**
@@ -246,158 +220,153 @@ static int IncSP_cmp_attr(ir_node *a, ir_node *b) {
  *
  * @return zero if both nodes have identically attributes
  */
-static int Call_cmp_attr(ir_node *a, ir_node *b) {
+static int Call_cmp_attr(ir_node *a, ir_node *b)
+{
 	const be_call_attr_t *a_attr = get_irn_attr_const(a);
 	const be_call_attr_t *b_attr = get_irn_attr_const(b);
 
 	if (a_attr->ent != b_attr->ent ||
-			a_attr->call_tp != b_attr->call_tp)
+		a_attr->call_tp != b_attr->call_tp)
 		return 1;
 
-	return _node_cmp_attr(&a_attr->node_attr, &b_attr->node_attr);
+	return node_cmp_attr(a, b);
 }
 
-static INLINE be_req_t *get_be_req(const ir_node *node, int pos)
+static arch_register_req_t *allocate_reg_req(const ir_node *node)
 {
-	int idx;
-	const be_node_attr_t *attr;
-	be_reg_data_t *rd;
-
-	assert(is_be_node(node));
-	attr = get_irn_attr_const(node);
+	ir_graph       *irg  = get_irn_irg(node);
+	struct obstack *obst = be_get_be_obst(irg);
 
-	if(pos < 0) {
-		idx = -(pos + 1);
-	} else {
-		idx = pos;
-		assert(idx < get_irn_arity(node));
-	}
-	assert(idx < ARR_LEN(attr->reg_data));
-	rd = &attr->reg_data[idx];
+	arch_register_req_t *req = obstack_alloc(obst, sizeof(*req));
+	memset(req, 0, sizeof(*req));
+	return req;
+}
 
-	return pos < 0 ? &rd->req : &rd->in_req;
+void be_set_constr_in(ir_node *node, int pos, const arch_register_req_t *req)
+{
+	const be_node_attr_t *attr = get_irn_attr_const(node);
+	be_reg_data_t *rd = &attr->reg_data[pos];
+	assert(pos < ARR_LEN(attr->reg_data));
+	rd->in_req = req;
 }
 
-static INLINE arch_register_req_t *get_req(const ir_node *node, int pos)
+void be_set_constr_out(ir_node *node, int pos, const arch_register_req_t *req)
 {
-	be_req_t *bereq = get_be_req(node, pos);
-	return &bereq->req;
+	backend_info_t *info = be_get_info(node);
+	info->out_infos[pos].req = req;
 }
 
 /**
- * Initializes the generic attribute of all be nodes and return ir.
+ * Initializes the generic attribute of all be nodes and return it.
  */
-static void *init_node_attr(ir_node *node, int max_reg_data)
+static void *init_node_attr(ir_node *node, int n_inputs, int n_outputs)
 {
-	ir_graph *irg = get_irn_irg(node);
-	struct obstack *obst = get_irg_obstack(irg);
-	be_node_attr_t *a = get_irn_attr(node);
+	ir_graph       *irg  = get_irn_irg(node);
+	struct obstack *obst = be_get_be_obst(irg);
+	be_node_attr_t *a    = get_irn_attr(node);
+	backend_info_t *info = be_get_info(node);
 
 	memset(a, 0, sizeof(get_op_attr_size(get_irn_op(node))));
 
-	if(max_reg_data >= 0) {
-		a->reg_data = NEW_ARR_D(be_reg_data_t, obst, max_reg_data);
-		memset(a->reg_data, 0, max_reg_data * sizeof(a->reg_data[0]));
+	if (n_inputs >= 0) {
+		int i;
+		a->reg_data = NEW_ARR_D(be_reg_data_t, obst, n_inputs);
+		for (i = 0; i < n_inputs; ++i) {
+			a->reg_data[i].in_req = arch_no_register_req;
+		}
 	} else {
 		a->reg_data = NEW_ARR_F(be_reg_data_t, 0);
 	}
 
-	return a;
-}
-
-static void add_register_req(ir_node *node)
-{
-	be_node_attr_t *a = get_irn_attr(node);
-	be_reg_data_t regreq;
-	memset(&regreq, 0, sizeof(regreq));
-	ARR_APP1(be_reg_data_t, a->reg_data, regreq);
-}
-
-/**
- * Skip Proj nodes and return their Proj numbers.
- *
- * If *node is a Proj or Proj(Proj) node, skip it.
- *
- * @param node  points to the node to be skipped
- *
- * @return 0 if *node was no Proj node, its Proj number else.
- */
-static int redir_proj(const ir_node **node)
-{
-	const ir_node *n = *node;
-
-	if(is_Proj(n)) {
-		ir_node *irn;
-
-		*node = irn = get_Proj_pred(n);
-		if(is_Proj(irn)) {
-			assert(get_irn_mode(irn) == mode_T);
-			*node = get_Proj_pred(irn);
+	if (n_outputs >= 0) {
+		int i;
+		info->out_infos = NEW_ARR_D(reg_out_info_t, obst, n_outputs);
+		memset(info->out_infos, 0, n_outputs * sizeof(info->out_infos[0]));
+		for (i = 0; i < n_outputs; ++i) {
+			info->out_infos[i].req = arch_no_register_req;
 		}
-		return get_Proj_proj(n);
+	} else {
+		info->out_infos = NEW_ARR_F(reg_out_info_t, 0);
 	}
 
-	return 0;
+	return a;
 }
 
-static be_reg_data_t *retrieve_reg_data(const ir_node *node)
+static void add_register_req_out(ir_node *node)
 {
-	const be_node_attr_t *attr;
-	int pos = 0;
-
-	if(is_Proj(node)) {
-		pos = get_Proj_proj(node);
-		node = get_Proj_pred(node);
-	}
-
-	assert(is_be_node(node));
-	attr = get_irn_attr_const(node);
-	assert(pos >= 0 && pos < ARR_LEN(attr->reg_data) && "illegal proj number");
-
-	return &attr->reg_data[pos];
+	backend_info_t *info = be_get_info(node);
+	reg_out_info_t  out_info;
+	memset(&out_info, 0, sizeof(out_info));
+	out_info.req = arch_no_register_req;
+	ARR_APP1(reg_out_info_t, info->out_infos, out_info);
 }
 
-static void
-be_node_set_irn_reg(const void *self, ir_node *irn, const arch_register_t *reg)
+static void add_register_req_in(ir_node *node)
 {
-	be_reg_data_t *r = retrieve_reg_data(irn);
-	(void) self;
-	r->reg = reg;
+	be_node_attr_t *a    = get_irn_attr(node);
+	be_reg_data_t   regreq;
+	memset(&regreq, 0, sizeof(regreq));
+	regreq.in_req = arch_no_register_req;
+	ARR_APP1(be_reg_data_t, a->reg_data, regreq);
 }
 
-ir_node *be_new_Spill(const arch_register_class_t *cls, const arch_register_class_t *cls_frame,
-	ir_graph *irg, ir_node *bl, ir_node *frame, ir_node *to_spill)
+ir_node *be_new_Spill(const arch_register_class_t *cls,
+		const arch_register_class_t *cls_frame, ir_node *bl,
+		ir_node *frame, ir_node *to_spill)
 {
 	be_frame_attr_t *a;
 	ir_node         *in[2];
 	ir_node         *res;
+	ir_graph        *irg = get_Block_irg(bl);
 
 	in[0]     = frame;
 	in[1]     = to_spill;
 	res       = new_ir_node(NULL, irg, bl, op_be_Spill, mode_M, 2, in);
-	a         = init_node_attr(res, 2);
+	a         = init_node_attr(res, 2, 1);
 	a->ent    = NULL;
 	a->offset = 0;
 
-	be_node_set_reg_class(res, be_pos_Spill_frame, cls_frame);
-	be_node_set_reg_class(res, be_pos_Spill_val, cls);
+	be_node_set_reg_class_in(res, be_pos_Spill_frame, cls_frame);
+	be_node_set_reg_class_in(res, be_pos_Spill_val, cls);
+	/*
+	 * For spills and reloads, we return "none" as requirement for frame
+	 * pointer, so every input is ok. Some backends need this (STA).
+	 * Matze: we should investigate if this is really needed, this solution
+	 *        looks very hacky to me
+	 */
+	be_set_constr_in(res, be_pos_Spill_frame, arch_no_register_req);
+
+	arch_set_out_register_req(res, 0, arch_no_register_req);
+
 	return res;
 }
 
-ir_node *be_new_Reload(const arch_register_class_t *cls, const arch_register_class_t *cls_frame,
-	ir_graph *irg, ir_node *bl, ir_node *frame, ir_node *mem, ir_mode *mode)
+ir_node *be_new_Reload(const arch_register_class_t *cls,
+		const arch_register_class_t *cls_frame, ir_node *block,
+		ir_node *frame, ir_node *mem, ir_mode *mode)
 {
-	ir_node *in[2];
-	ir_node *res;
+	ir_node  *in[2];
+	ir_node  *res;
+	ir_graph *irg = get_Block_irg(block);
 
 	in[0] = frame;
 	in[1] = mem;
-	res   = new_ir_node(NULL, irg, bl, op_be_Reload, mode, 2, in);
+	res   = new_ir_node(NULL, irg, block, op_be_Reload, mode, 2, in);
+
+	init_node_attr(res, 2, 1);
+	be_node_set_reg_class_out(res, 0, cls);
+
+	be_node_set_reg_class_in(res, be_pos_Reload_frame, cls_frame);
+	arch_irn_set_flags(res, arch_irn_flags_rematerializable);
+
+	/*
+	 * For spills and reloads, we return "none" as requirement for frame
+	 * pointer, so every input is ok. Some backends need this (e.g. STA).
+	 * Matze: we should investigate if this is really needed, this solution
+	 *        looks very hacky to me
+	 */
+	be_set_constr_in(res, be_pos_Reload_frame, arch_no_register_req);
 
-	init_node_attr(res, 2);
-	be_node_set_reg_class(res, -1, cls);
-	be_node_set_reg_class(res, be_pos_Reload_frame, cls_frame);
-	be_node_set_flags(res, -1, arch_irn_flags_rematerializable);
 	return res;
 }
 
@@ -425,14 +394,17 @@ ir_node *be_get_Spill_frame(const ir_node *irn)
 	return get_irn_n(irn, be_pos_Spill_frame);
 }
 
-ir_node *be_new_Perm(const arch_register_class_t *cls, ir_graph *irg, ir_node *bl, int n, ir_node *in[])
+ir_node *be_new_Perm(const arch_register_class_t *cls, ir_node *block,
+                     int n, ir_node *in[])
 {
-	int i;
-	ir_node *irn = new_ir_node(NULL, irg, bl, op_be_Perm, mode_T, n, in);
-	init_node_attr(irn, n);
-	for(i = 0; i < n; ++i) {
-		be_node_set_reg_class(irn, i, cls);
-		be_node_set_reg_class(irn, OUT_POS(i), cls);
+	int      i;
+	ir_graph *irg = get_Block_irg(block);
+
+	ir_node *irn = new_ir_node(NULL, irg, block, op_be_Perm, mode_T, n, in);
+	init_node_attr(irn, n, n);
+	for (i = 0; i < n; ++i) {
+		be_node_set_reg_class_in(irn, i, cls);
+		be_node_set_reg_class_out(irn, i, cls);
 	}
 
 	return irn;
@@ -440,104 +412,105 @@ ir_node *be_new_Perm(const arch_register_class_t *cls, ir_graph *irg, ir_node *b
 
 void be_Perm_reduce(ir_node *perm, int new_size, int *map)
 {
-	ir_graph *irg           = get_irn_irg(perm);
-	int            arity    = get_irn_arity(perm);
-	be_reg_data_t *old_data = alloca(arity * sizeof(old_data[0]));
-	be_node_attr_t *attr    = get_irn_attr(perm);
-	ir_node **new_in        = NEW_ARR_D(ir_node *, irg->obst, new_size);
+	int            arity      = get_irn_arity(perm);
+	be_reg_data_t  *old_data  = ALLOCAN(be_reg_data_t, arity);
+	reg_out_info_t *old_infos = ALLOCAN(reg_out_info_t, arity);
+	be_node_attr_t *attr      = get_irn_attr(perm);
+	backend_info_t *info      = be_get_info(perm);
+	ir_node        **new_in;
 
 	int i;
 
 	assert(be_is_Perm(perm));
 	assert(new_size <= arity);
 
+	new_in = alloca(new_size * sizeof(*new_in));
+
 	/* save the old register data */
 	memcpy(old_data, attr->reg_data, arity * sizeof(old_data[0]));
+	memcpy(old_infos, info->out_infos, arity * sizeof(old_infos[0]));
 
 	/* compose the new in array and set the new register data directly in place */
 	for (i = 0; i < new_size; ++i) {
 		int idx = map[i];
-		new_in[i]         = get_irn_n(perm, idx);
-		attr->reg_data[i] = old_data[idx];
+		new_in[i]          = get_irn_n(perm, idx);
+		attr->reg_data[i]  = old_data[idx];
+		info->out_infos[i] = old_infos[idx];
 	}
 
 	set_irn_in(perm, new_size, new_in);
 }
 
-ir_node *be_new_MemPerm(const arch_env_t *arch_env, ir_graph *irg, ir_node *bl, int n, ir_node *in[])
+ir_node *be_new_MemPerm(ir_node *block, int n, ir_node *in[])
 {
-	int i;
-	ir_node *frame = get_irg_frame(irg);
-	const arch_register_class_t *cls_frame = arch_get_irn_reg_class(arch_env, frame, -1);
-	ir_node *irn;
-	const arch_register_t *sp = arch_env->isa->sp;
-	be_memperm_attr_t *attr;
-	ir_node **real_in;
+	ir_graph                     *irg       = get_Block_irg(block);
+	const arch_env_t             *arch_env  = be_get_irg_arch_env(irg);
+	ir_node                      *frame     = get_irg_frame(irg);
+	const arch_register_t        *sp        = arch_env->sp;
+	ir_node                      *irn;
+	be_memperm_attr_t            *attr;
+	ir_node                     **real_in;
 
-	real_in = alloca((n+1) * sizeof(real_in[0]));
+	real_in = ALLOCAN(ir_node*, n + 1);
 	real_in[0] = frame;
 	memcpy(&real_in[1], in, n * sizeof(real_in[0]));
 
-	irn =  new_ir_node(NULL, irg, bl, op_be_MemPerm, mode_T, n+1, real_in);
-
-	init_node_attr(irn, n + 1);
-	be_node_set_reg_class(irn, 0, sp->reg_class);
-	for (i = 0; i < n; ++i) {
-		be_node_set_reg_class(irn, i + 1, cls_frame);
-		be_node_set_reg_class(irn, OUT_POS(i), cls_frame);
-	}
+	irn = new_ir_node(NULL, irg, block, op_be_MemPerm, mode_T, n+1, real_in);
 
-	attr = get_irn_attr(irn);
+	init_node_attr(irn, n + 1, n);
+	be_node_set_reg_class_in(irn, 0, sp->reg_class);
 
-	attr->in_entities = obstack_alloc(irg->obst, n * sizeof(attr->in_entities[0]));
-	memset(attr->in_entities, 0, n * sizeof(attr->in_entities[0]));
-	attr->out_entities = obstack_alloc(irg->obst, n*sizeof(attr->out_entities[0]));
-	memset(attr->out_entities, 0, n*sizeof(attr->out_entities[0]));
+	attr               = get_irn_attr(irn);
+	attr->in_entities  = OALLOCNZ(irg->obst, ir_entity*, n);
+	attr->out_entities = OALLOCNZ(irg->obst, ir_entity*, n);
 
 	return irn;
 }
 
-
-ir_node *be_new_Copy(const arch_register_class_t *cls, ir_graph *irg, ir_node *bl, ir_node *op)
+ir_node *be_new_Copy(const arch_register_class_t *cls, ir_node *bl, ir_node *op)
 {
 	ir_node *in[1];
 	ir_node *res;
 	arch_register_req_t *req;
+	ir_graph *irg = get_Block_irg(bl);
 
 	in[0] = op;
 	res   = new_ir_node(NULL, irg, bl, op_be_Copy, get_irn_mode(op), 1, in);
-	init_node_attr(res, 1);
-	be_node_set_reg_class(res, 0, cls);
-	be_node_set_reg_class(res, OUT_POS(0), cls);
+	init_node_attr(res, 1, 1);
+	be_node_set_reg_class_in(res, 0, cls);
+	be_node_set_reg_class_out(res, 0, cls);
 
-	req = get_req(res, OUT_POS(0));
-	req->cls = cls;
-	req->type = arch_register_req_type_should_be_same;
+	req = allocate_reg_req(res);
+	req->cls        = cls;
+	req->type       = arch_register_req_type_should_be_same;
 	req->other_same = 1U << 0;
+	be_set_constr_out(res, 0, req);
 
 	return res;
 }
 
-ir_node *be_get_Copy_op(const ir_node *cpy) {
+ir_node *be_get_Copy_op(const ir_node *cpy)
+{
 	return get_irn_n(cpy, be_pos_Copy_op);
 }
 
-void be_set_Copy_op(ir_node *cpy, ir_node *op) {
+void be_set_Copy_op(ir_node *cpy, ir_node *op)
+{
 	set_irn_n(cpy, be_pos_Copy_op, op);
 }
 
-ir_node *be_new_Keep(const arch_register_class_t *cls, ir_graph *irg, ir_node *bl, int n, ir_node *in[])
+ir_node *be_new_Keep(ir_node *block, int n, ir_node *in[])
 {
 	int i;
 	ir_node *res;
+	ir_graph *irg = get_Block_irg(block);
 
-	res = new_ir_node(NULL, irg, bl, op_be_Keep, mode_ANY, -1, NULL);
-	init_node_attr(res, -1);
+	res = new_ir_node(NULL, irg, block, op_be_Keep, mode_ANY, -1, NULL);
+	init_node_attr(res, -1, 1);
 
-	for(i = 0; i < n; ++i) {
+	for (i = 0; i < n; ++i) {
 		add_irn_n(res, in[i]);
-		add_register_req(res);
-		be_node_set_reg_class(res, i, cls);
+		add_register_req_in(res);
 	}
 	keep_alive(res);
 
@@ -550,13 +523,14 @@ void be_Keep_add_node(ir_node *keep, const arch_register_class_t *cls, ir_node *
 
 	assert(be_is_Keep(keep));
 	n = add_irn_n(keep, node);
-	add_register_req(keep);
-	be_node_set_reg_class(keep, n, cls);
+	add_register_req_in(keep);
+	be_node_set_reg_class_in(keep, n, cls);
 }
 
 /* creates a be_Call */
-ir_node *be_new_Call(dbg_info *dbg, ir_graph *irg, ir_node *bl, ir_node *mem, ir_node *sp, ir_node *ptr,
-                     int n_outs, int n, ir_node *in[], ir_type *call_tp)
+ir_node *be_new_Call(dbg_info *dbg, ir_graph *irg, ir_node *bl, ir_node *mem,
+		ir_node *sp, ir_node *ptr, int n_outs, int n, ir_node *in[],
+		ir_type *call_tp)
 {
 	be_call_attr_t *a;
 	int real_n = be_pos_Call_first_arg + n;
@@ -570,7 +544,7 @@ ir_node *be_new_Call(dbg_info *dbg, ir_graph *irg, ir_node *bl, ir_node *mem, ir
 	memcpy(&real_in[be_pos_Call_first_arg], in, n * sizeof(in[0]));
 
 	irn = new_ir_node(dbg, irg, bl, op_be_Call, mode_T, real_n, real_in);
-	a = init_node_attr(irn, (n_outs > real_n ? n_outs : real_n));
+	a = init_node_attr(irn, real_n, n_outs);
 	a->ent     = NULL;
 	a->call_tp = call_tp;
 	a->pop     = 0;
@@ -578,39 +552,45 @@ ir_node *be_new_Call(dbg_info *dbg, ir_graph *irg, ir_node *bl, ir_node *mem, ir
 }
 
 /* Gets the call entity or NULL if this is no static call. */
-ir_entity *be_Call_get_entity(const ir_node *call) {
+ir_entity *be_Call_get_entity(const ir_node *call)
+{
 	const be_call_attr_t *a = get_irn_attr_const(call);
 	assert(be_is_Call(call));
 	return a->ent;
 }
 
 /* Sets the call entity. */
-void be_Call_set_entity(ir_node *call, ir_entity *ent) {
+void be_Call_set_entity(ir_node *call, ir_entity *ent)
+{
 	be_call_attr_t *a = get_irn_attr(call);
 	assert(be_is_Call(call));
 	a->ent = ent;
 }
 
 /* Gets the call type. */
-ir_type *be_Call_get_type(ir_node *call) {
+ir_type *be_Call_get_type(ir_node *call)
+{
 	const be_call_attr_t *a = get_irn_attr_const(call);
 	assert(be_is_Call(call));
 	return a->call_tp;
 }
 
 /* Sets the call type. */
-void be_Call_set_type(ir_node *call, ir_type *call_tp) {
+void be_Call_set_type(ir_node *call, ir_type *call_tp)
+{
 	be_call_attr_t *a = get_irn_attr(call);
 	assert(be_is_Call(call));
 	a->call_tp = call_tp;
 }
 
-void be_Call_set_pop(ir_node *call, unsigned pop) {
+void be_Call_set_pop(ir_node *call, unsigned pop)
+{
 	be_call_attr_t *a = get_irn_attr(call);
 	a->pop = pop;
 }
 
-unsigned be_Call_get_pop(const ir_node *call) {
+unsigned be_Call_get_pop(const ir_node *call)
+{
 	const be_call_attr_t *a = get_irn_attr_const(call);
 	return a->pop;
 }
@@ -624,11 +604,12 @@ ir_node *be_new_Return(dbg_info *dbg, ir_graph *irg, ir_node *block, int n_res,
 	int i;
 
 	res = new_ir_node(dbg, irg, block, op_be_Return, mode_X, -1, NULL);
-	init_node_attr(res, -1);
-	for(i = 0; i < n; ++i) {
+	init_node_attr(res, -1, 1);
+	for (i = 0; i < n; ++i) {
 		add_irn_n(res, in[i]);
-		add_register_req(res);
+		add_register_req_in(res);
 	}
+	be_set_constr_out(res, 0, arch_no_register_req);
 
 	a = get_irn_attr(res);
 	a->num_ret_vals = n_res;
@@ -639,168 +620,148 @@ ir_node *be_new_Return(dbg_info *dbg, ir_graph *irg, ir_node *block, int n_res,
 }
 
 /* Returns the number of real returns values */
-int be_Return_get_n_rets(const ir_node *ret) {
+int be_Return_get_n_rets(const ir_node *ret)
+{
 	const be_return_attr_t *a = get_irn_generic_attr_const(ret);
 	return a->num_ret_vals;
 }
 
 /* return the number of bytes that should be popped from stack when executing the Return. */
-unsigned be_Return_get_pop(const ir_node *ret) {
+unsigned be_Return_get_pop(const ir_node *ret)
+{
 	const be_return_attr_t *a = get_irn_generic_attr_const(ret);
 	return a->pop;
 }
 
 /* return non-zero, if number of popped bytes must be always emitted */
-int be_Return_get_emit_pop(const ir_node *ret) {
+int be_Return_get_emit_pop(const ir_node *ret)
+{
 	const be_return_attr_t *a = get_irn_generic_attr_const(ret);
 	return a->emit_pop;
 }
 
 /* return non-zero, if number of popped bytes must be always emitted */
-void be_Return_set_emit_pop(ir_node *ret, int emit_pop) {
+void be_Return_set_emit_pop(ir_node *ret, int emit_pop)
+{
 	be_return_attr_t *a = get_irn_generic_attr(ret);
 	a->emit_pop = emit_pop;
 }
 
-int be_Return_append_node(ir_node *ret, ir_node *node) {
+int be_Return_append_node(ir_node *ret, ir_node *node)
+{
 	int pos;
 
 	pos = add_irn_n(ret, node);
-	add_register_req(ret);
+	add_register_req_in(ret);
 
 	return pos;
 }
 
-ir_node *be_new_IncSP(const arch_register_t *sp, ir_graph *irg, ir_node *bl,
+ir_node *be_new_IncSP(const arch_register_t *sp, ir_node *bl,
                       ir_node *old_sp, int offset, int align)
 {
 	be_incsp_attr_t *a;
 	ir_node *irn;
 	ir_node *in[1];
+	ir_graph *irg = get_Block_irg(bl);
 
 	in[0]     = old_sp;
 	irn       = new_ir_node(NULL, irg, bl, op_be_IncSP, sp->reg_class->mode,
 	                        sizeof(in) / sizeof(in[0]), in);
-	a         = init_node_attr(irn, 1);
+	a         = init_node_attr(irn, 1, 1);
 	a->offset = offset;
 	a->align  = align;
 
-	be_node_set_flags(irn, -1, arch_irn_flags_ignore | arch_irn_flags_modify_sp);
-
 	/* Set output constraint to stack register. */
-	be_node_set_reg_class(irn, 0, sp->reg_class);
-	be_set_constr_single_reg(irn, BE_OUT_POS(0), sp);
-	be_node_set_irn_reg(NULL, irn, sp);
+	be_node_set_reg_class_in(irn, 0, sp->reg_class);
+	be_set_constr_single_reg_out(irn, 0, sp, arch_register_req_type_produces_sp);
 
 	return irn;
 }
 
-ir_node *be_new_AddSP(const arch_register_t *sp, ir_graph *irg, ir_node *bl, ir_node *old_sp, ir_node *sz)
+ir_node *be_new_AddSP(const arch_register_t *sp, ir_node *bl, ir_node *old_sp,
+		ir_node *sz)
 {
 	be_node_attr_t *a;
 	ir_node *irn;
 	ir_node *in[be_pos_AddSP_last];
-	const arch_register_class_t *class;
+	const arch_register_class_t *cls;
+	ir_graph *irg;
 
 	in[be_pos_AddSP_old_sp] = old_sp;
 	in[be_pos_AddSP_size]   = sz;
 
+	irg = get_Block_irg(bl);
 	irn = new_ir_node(NULL, irg, bl, op_be_AddSP, mode_T, be_pos_AddSP_last, in);
-	a   = init_node_attr(irn, be_pos_AddSP_last);
-
-	be_node_set_flags(irn, OUT_POS(pn_be_AddSP_sp),
-	                  arch_irn_flags_ignore | arch_irn_flags_modify_sp);
+	a   = init_node_attr(irn, be_pos_AddSP_last, pn_be_AddSP_last);
 
 	/* Set output constraint to stack register. */
-	be_set_constr_single_reg(irn, be_pos_AddSP_old_sp, sp);
-	be_node_set_reg_class(irn, be_pos_AddSP_size, arch_register_get_class(sp));
-	be_set_constr_single_reg(irn, OUT_POS(pn_be_AddSP_sp), sp);
-	a->reg_data[pn_be_AddSP_sp].reg = sp;
+	be_set_constr_single_reg_in(irn, be_pos_AddSP_old_sp, sp, 0);
+	be_node_set_reg_class_in(irn, be_pos_AddSP_size, arch_register_get_class(sp));
+	be_set_constr_single_reg_out(irn, pn_be_AddSP_sp, sp, arch_register_req_type_produces_sp);
 
-	class = arch_register_get_class(sp);
-	be_node_set_reg_class(irn, OUT_POS(pn_be_AddSP_res), class);
+	cls = arch_register_get_class(sp);
 
 	return irn;
 }
 
-ir_node *be_new_SubSP(const arch_register_t *sp, ir_graph *irg, ir_node *bl, ir_node *old_sp, ir_node *sz)
+ir_node *be_new_SubSP(const arch_register_t *sp, ir_node *bl, ir_node *old_sp, ir_node *sz)
 {
 	be_node_attr_t *a;
 	ir_node *irn;
 	ir_node *in[be_pos_SubSP_last];
+	ir_graph *irg;
 
 	in[be_pos_SubSP_old_sp] = old_sp;
 	in[be_pos_SubSP_size]   = sz;
 
+	irg = get_Block_irg(bl);
 	irn = new_ir_node(NULL, irg, bl, op_be_SubSP, mode_T, be_pos_SubSP_last, in);
-	a   = init_node_attr(irn, be_pos_SubSP_last);
-
-	be_node_set_flags(irn, OUT_POS(pn_be_SubSP_sp),
-	                  arch_irn_flags_ignore | arch_irn_flags_modify_sp);
+	a   = init_node_attr(irn, be_pos_SubSP_last, pn_be_SubSP_last);
 
 	/* Set output constraint to stack register. */
-	be_set_constr_single_reg(irn, be_pos_SubSP_old_sp, sp);
-	be_node_set_reg_class(irn, be_pos_SubSP_size, arch_register_get_class(sp));
-	be_set_constr_single_reg(irn, OUT_POS(pn_be_SubSP_sp), sp);
-	a->reg_data[pn_be_SubSP_sp].reg = sp;
+	be_set_constr_single_reg_in(irn, be_pos_SubSP_old_sp, sp, 0);
+	be_node_set_reg_class_in(irn, be_pos_SubSP_size, arch_register_get_class(sp));
+	be_set_constr_single_reg_out(irn, pn_be_SubSP_sp, sp, arch_register_req_type_produces_sp);
 
 	return irn;
 }
 
-ir_node *be_new_RegParams(ir_graph *irg, ir_node *bl, int n_outs)
+ir_node *be_new_Start(dbg_info *dbgi, ir_node *bl, int n_outs)
 {
 	ir_node *res;
 	int i;
+	ir_graph *irg = get_Block_irg(bl);
 
-	res = new_ir_node(NULL, irg, bl, op_be_RegParams, mode_T, 0, NULL);
-	init_node_attr(res, -1);
-	for(i = 0; i < n_outs; ++i)
-		add_register_req(res);
+	res = new_ir_node(dbgi, irg, bl, op_be_Start, mode_T, 0, NULL);
+	init_node_attr(res, 0, -1);
+	for (i = 0; i < n_outs; ++i) {
+		add_register_req_out(res);
+	}
 
 	return res;
 }
 
-ir_node *be_RegParams_append_out_reg(ir_node *regparams,
-                                     const arch_env_t *arch_env,
-                                     const arch_register_t *reg)
-{
-	ir_graph *irg = get_irn_irg(regparams);
-	ir_node *block = get_nodes_block(regparams);
-	be_node_attr_t *attr = get_irn_attr(regparams);
-	const arch_register_class_t *cls = arch_register_get_class(reg);
-	ir_mode *mode = arch_register_class_mode(cls);
-	int n = ARR_LEN(attr->reg_data);
-	ir_node *proj;
-
-	assert(be_is_RegParams(regparams));
-	proj = new_r_Proj(irg, block, regparams, mode, n);
-	add_register_req(regparams);
-	be_set_constr_single_reg(regparams, BE_OUT_POS(n), reg);
-	arch_set_irn_register(arch_env, proj, reg);
-
-	/* TODO decide, whether we need to set ignore/modify sp flags here? */
-
-	return proj;
-}
-
-ir_node *be_new_FrameAddr(const arch_register_class_t *cls_frame, ir_graph *irg, ir_node *bl, ir_node *frame, ir_entity *ent)
+ir_node *be_new_FrameAddr(const arch_register_class_t *cls_frame, ir_node *bl, ir_node *frame, ir_entity *ent)
 {
 	be_frame_attr_t *a;
 	ir_node *irn;
 	ir_node *in[1];
+	ir_graph *irg = get_Block_irg(bl);
 
 	in[0]  = frame;
 	irn    = new_ir_node(NULL, irg, bl, op_be_FrameAddr, get_irn_mode(frame), 1, in);
-	a      = init_node_attr(irn, 1);
+	a      = init_node_attr(irn, 1, 1);
 	a->ent = ent;
 	a->offset = 0;
-	be_node_set_reg_class(irn, 0, cls_frame);
-	be_node_set_reg_class(irn, OUT_POS(0), cls_frame);
+	be_node_set_reg_class_in(irn, 0, cls_frame);
+	be_node_set_reg_class_out(irn, 0, cls_frame);
 
 	return optimize_node(irn);
 }
 
-ir_node *be_get_FrameAddr_frame(const ir_node *node) {
+ir_node *be_get_FrameAddr_frame(const ir_node *node)
+{
 	assert(be_is_FrameAddr(node));
 	return get_irn_n(node, be_pos_FrameAddr_ptr);
 }
@@ -811,44 +772,49 @@ ir_entity *be_get_FrameAddr_entity(const ir_node *node)
 	return attr->ent;
 }
 
-ir_node *be_new_CopyKeep(const arch_register_class_t *cls, ir_graph *irg, ir_node *bl, ir_node *src, int n, ir_node *in_keep[], ir_mode *mode)
+ir_node *be_new_CopyKeep(const arch_register_class_t *cls, ir_node *bl, ir_node *src, int n, ir_node *in_keep[], ir_mode *mode)
 {
-	ir_node *irn;
-	ir_node **in = (ir_node **) alloca((n + 1) * sizeof(in[0]));
+	ir_node  *irn;
+	ir_node **in = ALLOCAN(ir_node*, n + 1);
+	ir_graph *irg = get_Block_irg(bl);
 
 	in[0] = src;
 	memcpy(&in[1], in_keep, n * sizeof(in[0]));
 	irn   = new_ir_node(NULL, irg, bl, op_be_CopyKeep, mode, n + 1, in);
-	init_node_attr(irn, n + 1);
-	be_node_set_reg_class(irn, OUT_POS(0), cls);
-	be_node_set_reg_class(irn, 0, cls);
+	init_node_attr(irn, n + 1, 1);
+	be_node_set_reg_class_in(irn, 0, cls);
+	be_node_set_reg_class_out(irn, 0, cls);
 
 	return irn;
 }
 
-ir_node *be_new_CopyKeep_single(const arch_register_class_t *cls, ir_graph *irg, ir_node *bl, ir_node *src, ir_node *keep, ir_mode *mode)
+ir_node *be_new_CopyKeep_single(const arch_register_class_t *cls, ir_node *bl, ir_node *src, ir_node *keep, ir_mode *mode)
 {
-	return be_new_CopyKeep(cls, irg, bl, src, 1, &keep, mode);
+	return be_new_CopyKeep(cls, bl, src, 1, &keep, mode);
 }
 
-ir_node *be_get_CopyKeep_op(const ir_node *cpy) {
+ir_node *be_get_CopyKeep_op(const ir_node *cpy)
+{
 	return get_irn_n(cpy, be_pos_CopyKeep_op);
 }
 
-void be_set_CopyKeep_op(ir_node *cpy, ir_node *op) {
+void be_set_CopyKeep_op(ir_node *cpy, ir_node *op)
+{
 	set_irn_n(cpy, be_pos_CopyKeep_op, op);
 }
 
-ir_node *be_new_Barrier(ir_graph *irg, ir_node *bl, int n, ir_node *in[])
+ir_node *be_new_Barrier(ir_node *bl, int n, ir_node *in[])
 {
 	ir_node *res;
 	int i;
+	ir_graph *irg = get_Block_irg(bl);
 
 	res = new_ir_node(NULL, irg, bl, op_be_Barrier, mode_T, -1, NULL);
-	init_node_attr(res, -1);
-	for(i = 0; i < n; ++i) {
+	init_node_attr(res, -1, -1);
+	for (i = 0; i < n; ++i) {
 		add_irn_n(res, in[i]);
-		add_register_req(res);
+		add_register_req_in(res);
+		add_register_req_out(res);
 	}
 
 	return res;
@@ -856,41 +822,25 @@ ir_node *be_new_Barrier(ir_graph *irg, ir_node *bl, int n, ir_node *in[])
 
 ir_node *be_Barrier_append_node(ir_node *barrier, ir_node *node)
 {
-	ir_graph *irg = get_irn_irg(barrier);
-	ir_node *block = get_nodes_block(barrier);
 	ir_mode *mode = get_irn_mode(node);
 	int n = add_irn_n(barrier, node);
 
-	ir_node *proj = new_r_Proj(irg, block, barrier, mode, n);
-	add_register_req(barrier);
+	ir_node *proj = new_r_Proj(barrier, mode, n);
+	add_register_req_in(barrier);
+	add_register_req_out(barrier);
 
 	return proj;
 }
 
-/* Construct a new be_Unwind. */
-ir_node *be_new_Unwind(dbg_info *dbg, ir_graph *irg, ir_node *block,
-					   ir_node *mem, ir_node *sp)
-{
-	ir_node *res;
-	ir_node *in[2];
-
-	in[be_pos_Unwind_mem] = mem;
-	in[be_pos_Unwind_sp]  = sp;
-	res = new_ir_node(dbg, irg, block, op_be_Unwind, mode_X, 2, in);
-	init_node_attr(res, -1);
-
-	return res;
-}
-
-int be_has_frame_entity(const ir_node *irn)
+static bool be_has_frame_entity(const ir_node *irn)
 {
 	switch (get_irn_opcode(irn)) {
 	case beo_Spill:
 	case beo_Reload:
 	case beo_FrameAddr:
-		return 1;
+		return true;
 	default:
-		return 0;
+		return false;
 	}
 }
 
@@ -958,75 +908,79 @@ int be_get_MemPerm_entity_arity(const ir_node *irn)
 	return get_irn_arity(irn) - 1;
 }
 
-void be_set_constr_single_reg(ir_node *node, int pos, const arch_register_t *reg)
+const arch_register_req_t *be_create_reg_req(struct obstack *obst,
+		const arch_register_t *reg, arch_register_req_type_t additional_types)
 {
-	arch_register_req_t *req = get_req(node, pos);
+	arch_register_req_t         *req = obstack_alloc(obst, sizeof(*req));
 	const arch_register_class_t *cls = arch_register_get_class(reg);
-	ir_graph *irg = get_irn_irg(node);
-	struct obstack *obst = get_irg_obstack(irg);
-	unsigned *limited_bitset;
-
-	assert(req->cls == NULL || req->cls == cls);
-	assert(! (req->type & arch_register_req_type_limited));
-	assert(req->limited == NULL);
+	unsigned                    *limited_bitset;
 
 	limited_bitset = rbitset_obstack_alloc(obst, arch_register_class_n_regs(cls));
 	rbitset_set(limited_bitset, arch_register_get_index(reg));
 
-	req->cls = cls;
-	req->type |= arch_register_req_type_limited;
+	req->type    = arch_register_req_type_limited | additional_types;
+	req->cls     = cls;
 	req->limited = limited_bitset;
+	return req;
 }
 
-void be_set_constr_limited(ir_node *node, int pos, const arch_register_req_t *req)
+void be_set_constr_single_reg_in(ir_node *node, int pos,
+		const arch_register_t *reg, arch_register_req_type_t additional_types)
 {
-	ir_graph *irg = get_irn_irg(node);
-	struct obstack *obst = get_irg_obstack(irg);
-	arch_register_req_t *r = get_req(node, pos);
+	const arch_register_req_t *req;
 
-	assert(arch_register_req_is(req, limited));
-	assert(! (req->type & (arch_register_req_type_should_be_same | arch_register_req_type_should_be_different)));
-	memcpy(r, req, sizeof(r[0]));
-	r->limited = rbitset_duplicate_obstack_alloc(obst, req->limited, req->cls->n_regs);
+	if (additional_types == 0) {
+		req = reg->single_req;
+	} else {
+		ir_graph       *irg  = get_irn_irg(node);
+		struct obstack *obst = be_get_be_obst(irg);
+		req = be_create_reg_req(obst, reg, additional_types);
+	}
+	be_set_constr_in(node, pos, req);
 }
 
-void be_node_set_flags(ir_node *irn, int pos, arch_irn_flags_t flags)
+void be_set_constr_single_reg_out(ir_node *node, int pos,
+		const arch_register_t *reg, arch_register_req_type_t additional_types)
 {
-	be_req_t *bereq = get_be_req(irn, pos);
-	bereq->flags = flags;
-}
+	const arch_register_req_t *req;
 
-void be_node_add_flags(ir_node *irn, int pos, arch_irn_flags_t flags)
-{
-	be_req_t *bereq = get_be_req(irn, pos);
-	bereq->flags |= flags;
-}
+	/* if we have an ignore register, add ignore flag and just assign it */
+	if (reg->type & arch_register_type_ignore) {
+		additional_types |= arch_register_req_type_ignore;
+	}
 
-void be_node_set_reg_class(ir_node *irn, int pos, const arch_register_class_t *cls)
-{
-	arch_register_req_t *req = get_req(irn, pos);
+	if (additional_types == 0) {
+		req = reg->single_req;
+	} else {
+		ir_graph       *irg  = get_irn_irg(node);
+		struct obstack *obst = be_get_be_obst(irg);
+		req = be_create_reg_req(obst, reg, additional_types);
+	}
 
-	req->cls = cls;
+	arch_irn_set_register(node, pos, reg);
+	be_set_constr_out(node, pos, req);
+}
 
-	if (cls == NULL) {
-		req->type = arch_register_req_type_none;
-	} else if (req->type == arch_register_req_type_none) {
-		req->type = arch_register_req_type_normal;
-	}
+void be_node_set_reg_class_in(ir_node *irn, int pos,
+                              const arch_register_class_t *cls)
+{
+	be_set_constr_in(irn, pos, cls->class_req);
 }
 
-void be_node_set_req_type(ir_node *irn, int pos, arch_register_req_type_t type)
+void be_node_set_reg_class_out(ir_node *irn, int pos,
+                               const arch_register_class_t *cls)
 {
-	arch_register_req_t *req = get_req(irn, pos);
-	req->type = type;
+	be_set_constr_out(irn, pos, cls->class_req);
 }
 
-ir_node *be_get_IncSP_pred(ir_node *irn) {
+ir_node *be_get_IncSP_pred(ir_node *irn)
+{
 	assert(be_is_IncSP(irn));
 	return get_irn_n(irn, 0);
 }
 
-void be_set_IncSP_pred(ir_node *incsp, ir_node *pred) {
+void be_set_IncSP_pred(ir_node *incsp, ir_node *pred)
+{
 	assert(be_is_IncSP(incsp));
 	set_irn_n(incsp, 0, pred);
 }
@@ -1052,32 +1006,32 @@ int be_get_IncSP_align(const ir_node *irn)
 	return a->align;
 }
 
-ir_node *be_spill(const arch_env_t *arch_env, ir_node *block, ir_node *irn)
+ir_node *be_spill(ir_node *block, ir_node *irn)
 {
-	ir_graph                    *irg       = get_irn_irg(block);
+	ir_graph                    *irg       = get_Block_irg(block);
 	ir_node                     *frame     = get_irg_frame(irg);
-	const arch_register_class_t *cls       = arch_get_irn_reg_class(arch_env, irn, -1);
-	const arch_register_class_t *cls_frame = arch_get_irn_reg_class(arch_env, frame, -1);
+	const arch_register_class_t *cls       = arch_get_irn_reg_class_out(irn);
+	const arch_register_class_t *cls_frame = arch_get_irn_reg_class_out(frame);
 	ir_node                     *spill;
 
-	spill = be_new_Spill(cls, cls_frame, irg, block, frame, irn);
+	spill = be_new_Spill(cls, cls_frame, block, frame, irn);
 	return spill;
 }
 
-ir_node *be_reload(const arch_env_t *arch_env, const arch_register_class_t *cls, ir_node *insert, ir_mode *mode, ir_node *spill)
+ir_node *be_reload(const arch_register_class_t *cls, ir_node *insert, ir_mode *mode, ir_node *spill)
 {
 	ir_node  *reload;
 	ir_node  *bl    = is_Block(insert) ? insert : get_nodes_block(insert);
-	ir_graph *irg   = get_irn_irg(bl);
+	ir_graph *irg   = get_Block_irg(bl);
 	ir_node  *frame = get_irg_frame(irg);
-	const arch_register_class_t *cls_frame = arch_get_irn_reg_class(arch_env, frame, -1);
+	const arch_register_class_t *cls_frame = arch_get_irn_reg_class_out(frame);
 
 	assert(be_is_Spill(spill) || (is_Phi(spill) && get_irn_mode(spill) == mode_M));
 
-	reload = be_new_Reload(cls, cls_frame, irg, bl, frame, spill, mode);
+	reload = be_new_Reload(cls, cls_frame, bl, frame, spill, mode);
 
 	if (is_Block(insert)) {
-		insert = sched_skip(insert, 0, sched_skip_cf_predicator, (void *) arch_env);
+		insert = sched_skip(insert, 0, sched_skip_cf_predicator, NULL);
 		sched_add_after(insert, reload);
 	} else {
 		sched_add_before(insert, reload);
@@ -1097,123 +1051,37 @@ ir_node *be_reload(const arch_env_t *arch_env, const arch_register_class_t *cls,
 */
 
 
-static const
-arch_register_req_t *get_out_reg_req(const ir_node *irn, int out_pos)
-{
-	const be_node_attr_t *a = get_irn_attr_const(irn);
-
-	if(out_pos >= ARR_LEN(a->reg_data)) {
-		return arch_no_register_req;
-	}
-
-	return &a->reg_data[out_pos].req.req;
-}
-
-static const
-arch_register_req_t *get_in_reg_req(const ir_node *irn, int pos)
+static const arch_register_req_t *be_node_get_in_reg_req(
+		const ir_node *irn, int pos)
 {
 	const be_node_attr_t *a = get_irn_attr_const(irn);
 
-	if(pos >= get_irn_arity(irn) || pos >= ARR_LEN(a->reg_data))
+	assert(pos >= 0);
+	if (pos >= get_irn_arity(irn) || pos >= ARR_LEN(a->reg_data))
 		return arch_no_register_req;
 
-	return &a->reg_data[pos].in_req.req;
+	return a->reg_data[pos].in_req;
 }
 
-static const arch_register_req_t *
-be_node_get_irn_reg_req(const void *self, const ir_node *irn, int pos)
+static arch_irn_class_t be_node_classify(const ir_node *irn)
 {
-	int out_pos = pos;
-
-	(void) self;
-	if (pos < 0) {
-		if (get_irn_mode(irn) == mode_T)
-			return arch_no_register_req;
-
-		out_pos = redir_proj((const ir_node **)&irn);
-		assert(is_be_node(irn));
-		return get_out_reg_req(irn, out_pos);
-	} else if (is_be_node(irn)) {
-		/*
-   		 * For spills and reloads, we return "none" as requirement for frame
-		 * pointer, so every input is ok. Some backends need this (e.g. STA).
-		 */
-		if ((be_is_Spill(irn)  && pos == be_pos_Spill_frame) ||
-				(be_is_Reload(irn) && pos == be_pos_Reload_frame))
-			return arch_no_register_req;
-
-		return get_in_reg_req(irn, pos);
-	}
-
-	return arch_no_register_req;
-}
-
-const arch_register_t *
-be_node_get_irn_reg(const void *self, const ir_node *irn)
-{
-	be_reg_data_t *r;
-
-	(void) self;
-	if (get_irn_mode(irn) == mode_T)
-		return NULL;
-	r = retrieve_reg_data(irn);
-	return r->reg;
-}
-
-static arch_irn_class_t be_node_classify(const void *self, const ir_node *irn)
-{
-	(void) self;
-
-restart:
 	switch (get_irn_opcode(irn)) {
-#define XXX(a,b) case a: return b
-		XXX(beo_Spill, arch_irn_class_spill);
-		XXX(beo_Reload, arch_irn_class_reload);
-		XXX(beo_Perm, arch_irn_class_perm);
-		XXX(beo_Copy, arch_irn_class_copy);
-		XXX(beo_Return, arch_irn_class_branch);
-#undef XXX
-		case iro_Proj:
-			irn = get_Proj_pred(irn);
-			if (is_Proj(irn)) {
-				assert(get_irn_mode(irn) == mode_T);
-				irn = get_Proj_pred(irn);
-			}
-			goto restart;
-			break;
-		default:
-			return arch_irn_class_normal;
+		case beo_Spill:  return arch_irn_class_spill;
+		case beo_Reload: return arch_irn_class_reload;
+		case beo_Perm:   return arch_irn_class_perm;
+		case beo_Copy:   return arch_irn_class_copy;
+		default:         return 0;
 	}
-
-	return 0;
-}
-
-static arch_irn_flags_t be_node_get_flags(const void *self, const ir_node *node)
-{
-	be_req_t *bereq;
-	int pos = -1;
-	(void) self;
-
-	if(is_Proj(node)) {
-		pos = OUT_POS(get_Proj_proj(node));
-		node = skip_Proj_const(node);
-	}
-
-	bereq = get_be_req(node, pos);
-
-	return bereq->flags;
 }
 
-static ir_entity *be_node_get_frame_entity(const void *self, const ir_node *irn)
+static ir_entity *be_node_get_frame_entity(const ir_node *irn)
 {
-	(void) self;
 	return be_get_frame_entity(irn);
 }
 
-static void be_node_set_frame_entity(const void *self, ir_node *irn, ir_entity *ent)
+void be_node_set_frame_entity(ir_node *irn, ir_entity *ent)
 {
 	be_frame_attr_t *a;
-	(void) self;
 
 	assert(be_has_frame_entity(irn));
 
@@ -1221,21 +1089,22 @@ static void be_node_set_frame_entity(const void *self, ir_node *irn, ir_entity *
 	a->ent = ent;
 }
 
-static void be_node_set_frame_offset(const void *self, ir_node *irn, int offset)
+static void be_node_set_frame_offset(ir_node *irn, int offset)
 {
-	(void) self;
-	if(be_has_frame_entity(irn)) {
-		be_frame_attr_t *a = get_irn_attr(irn);
-		a->offset = offset;
-	}
+	be_frame_attr_t *a;
+
+	if (!be_has_frame_entity(irn))
+		return;
+
+	a = get_irn_attr(irn);
+	a->offset = offset;
 }
 
-static int be_node_get_sp_bias(const void *self, const ir_node *irn)
+static int be_node_get_sp_bias(const ir_node *irn)
 {
-	(void) self;
-	if(be_is_IncSP(irn))
+	if (be_is_IncSP(irn))
 		return be_get_IncSP_offset(irn);
-	if(be_is_Call(irn))
+	if (be_is_Call(irn))
 		return -(int)be_Call_get_pop(irn);
 
 	return 0;
@@ -1250,14 +1119,11 @@ static int be_node_get_sp_bias(const void *self, const ir_node *irn)
 
 */
 
+/* for be nodes */
 static const arch_irn_ops_t be_node_irn_ops = {
-	be_node_get_irn_reg_req,
-	be_node_set_irn_reg,
-	be_node_get_irn_reg,
+	be_node_get_in_reg_req,
 	be_node_classify,
-	be_node_get_flags,
 	be_node_get_frame_entity,
-	be_node_set_frame_entity,
 	be_node_set_frame_offset,
 	be_node_get_sp_bias,
 	NULL,    /* get_inverse             */
@@ -1266,6 +1132,52 @@ static const arch_irn_ops_t be_node_irn_ops = {
 	NULL,    /* perform_memory_operand  */
 };
 
+static const arch_register_req_t *dummy_reg_req(
+		const ir_node *node, int pos)
+{
+	(void) node;
+	(void) pos;
+	return arch_no_register_req;
+}
+
+static arch_irn_class_t dummy_classify(const ir_node *node)
+{
+	(void) node;
+	return 0;
+}
+
+static ir_entity* dummy_get_frame_entity(const ir_node *node)
+{
+	(void) node;
+	return NULL;
+}
+
+static void dummy_set_frame_offset(ir_node *node, int bias)
+{
+	(void) node;
+	(void) bias;
+	panic("dummy_set_frame_offset() should not be called");
+}
+
+static int dummy_get_sp_bias(const ir_node *node)
+{
+	(void) node;
+	return 0;
+}
+
+/* for "middleend" nodes */
+static const arch_irn_ops_t dummy_be_irn_ops = {
+	dummy_reg_req,
+	dummy_classify,
+	dummy_get_frame_entity,
+	dummy_set_frame_offset,
+	dummy_get_sp_bias,
+	NULL,      /* get_inverse           */
+	NULL,      /* get_op_estimated_cost */
+	NULL,      /* possible_memory_operand */
+	NULL,      /* perform_memory_operand */
+};
+
 /*
   ____  _     _   ___ ____  _   _   _   _                 _ _
  |  _ \| |__ (_) |_ _|  _ \| \ | | | | | | __ _ _ __   __| | | ___ _ __
@@ -1275,232 +1187,137 @@ static const arch_irn_ops_t be_node_irn_ops = {
 
 */
 
-typedef struct {
-	const arch_register_t *reg;
-	arch_register_req_t    req;
-	arch_irn_flags_t       flags;
-} phi_attr_t;
-
-struct {
-	arch_env_t  *arch_env;
-	pmap        *phi_attrs;
-} phi_handler;
-
-#define get_phi_handler_from_ops(h)      container_of(h, phi_handler_t, irn_ops)
-
-static INLINE
-phi_attr_t *get_Phi_attr(const ir_node *phi)
-{
-	phi_attr_t *attr = pmap_get(phi_handler.phi_attrs, (void*) phi);
-	if(attr == NULL) {
-		ir_graph *irg = get_irn_irg(phi);
-		struct obstack *obst = get_irg_obstack(irg);
-		attr = obstack_alloc(obst, sizeof(attr[0]));
-		memset(attr, 0, sizeof(attr[0]));
-		pmap_insert(phi_handler.phi_attrs, phi, attr);
+ir_node *be_new_Phi(ir_node *block, int n_ins, ir_node **ins, ir_mode *mode,
+                    const arch_register_class_t *cls)
+{
+	struct obstack *obst = be_get_be_obst(get_irn_irg(block));
+	backend_info_t *info;
+
+	ir_node *phi = new_r_Phi(block, n_ins, ins, mode);
+	info = be_get_info(phi);
+	info->out_infos = NEW_ARR_D(reg_out_info_t, obst, 1);
+	memset(info->out_infos, 0, 1 * sizeof(info->out_infos[0]));
+	if (cls == NULL) {
+		info->out_infos[0].req = arch_no_register_req;
+	} else {
+		info->out_infos[0].req = cls->class_req;
 	}
 
-	return attr;
+	return phi;
 }
 
 /**
- * Get register class of a Phi.
+ * Guess correct register class of a phi node by looking at its arguments
  */
-static
-const arch_register_req_t *get_Phi_reg_req_recursive(const ir_node *phi,
-                                                     pset **visited)
+static const arch_register_req_t *get_Phi_reg_req_recursive(const ir_node *phi,
+                                                            pset **visited)
 {
 	int n = get_irn_arity(phi);
 	ir_node *op;
 	int i;
 
-	if(*visited && pset_find_ptr(*visited, phi))
+	if (*visited && pset_find_ptr(*visited, phi))
 		return NULL;
 
-	for(i = 0; i < n; ++i) {
+	for (i = 0; i < n; ++i) {
 		op = get_irn_n(phi, i);
 		/* Matze: don't we unnecessary constraint our phis with this?
 		 * we only need to take the regclass IMO*/
-		if(!is_Phi(op))
-			return arch_get_register_req(phi_handler.arch_env, op, BE_OUT_POS(0));
+		if (!is_Phi(op))
+			return arch_get_register_req_out(op);
 	}
 
 	/*
 	 * The operands of that Phi were all Phis themselves.
 	 * We have to start a DFS for a non-Phi argument now.
 	 */
-	if(!*visited)
+	if (!*visited)
 		*visited = pset_new_ptr(16);
 
 	pset_insert_ptr(*visited, phi);
 
-	for(i = 0; i < n; ++i) {
+	for (i = 0; i < n; ++i) {
 		const arch_register_req_t *req;
 		op = get_irn_n(phi, i);
 		req = get_Phi_reg_req_recursive(op, visited);
-		if(req != NULL)
+		if (req != NULL)
 			return req;
 	}
 
 	return NULL;
 }
 
-static
-const arch_register_req_t *phi_get_irn_reg_req(const void *self,
-                                               const ir_node *irn, int pos)
+static const arch_register_req_t *phi_get_irn_reg_req(const ir_node *node,
+                                                      int pos)
 {
-	phi_attr_t *attr;
-	(void) self;
+	backend_info_t            *info = be_get_info(node);
+	const arch_register_req_t *req  = info->out_infos[0].req;
 	(void) pos;
 
-	if(!mode_is_datab(get_irn_mode(irn)))
-		return arch_no_register_req;
-
-	attr = get_Phi_attr(irn);
+	if (req == NULL) {
+		if (!mode_is_datab(get_irn_mode(node))) {
+			req = arch_no_register_req;
+		} else {
+			pset *visited = NULL;
 
-	if(attr->req.type == arch_register_req_type_none) {
-		pset *visited = NULL;
-		const arch_register_req_t *req;
-		req = get_Phi_reg_req_recursive(irn, &visited);
+			req = get_Phi_reg_req_recursive(node, &visited);
+			assert(req->cls != NULL);
+			req = req->cls->class_req;
 
-		memcpy(&attr->req, req, sizeof(req[0]));
-		assert(attr->req.cls != NULL);
-		attr->req.type = arch_register_req_type_normal;
-
-		if(visited != NULL)
-			del_pset(visited);
+			if (visited != NULL)
+				del_pset(visited);
+		}
+		info->out_infos[0].req = req;
 	}
 
-	return &attr->req;
-}
-
-void be_set_phi_reg_req(const arch_env_t *arch_env, ir_node *node,
-                        const arch_register_req_t *req)
-{
-	phi_attr_t *attr;
-	(void) arch_env;
-
-	assert(mode_is_datab(get_irn_mode(node)));
-
-	attr = get_Phi_attr(node);
-	memcpy(&attr->req, req, sizeof(req[0]));
+	return req;
 }
 
-void be_set_phi_flags(const arch_env_t *arch_env, ir_node *node,
-                      arch_irn_flags_t flags)
+void be_set_phi_reg_req(ir_node *node, const arch_register_req_t *req)
 {
-	phi_attr_t *attr;
-	(void) arch_env;
+	backend_info_t *info = be_get_info(node);
+	info->out_infos[0].req = req;
 
 	assert(mode_is_datab(get_irn_mode(node)));
-
-	attr = get_Phi_attr(node);
-	attr->flags = flags;
-}
-
-static
-void phi_set_irn_reg(const void *self, ir_node *irn, const arch_register_t *reg)
-{
-	phi_attr_t *attr = get_Phi_attr(irn);
-	(void) self;
-	attr->reg = reg;
-}
-
-static
-const arch_register_t *phi_get_irn_reg(const void *self, const ir_node *irn)
-{
-	phi_attr_t *attr = get_Phi_attr(irn);
-	(void) self;
-	return attr->reg;
 }
 
-static
-arch_irn_class_t phi_classify(const void *self, const ir_node *irn)
-{
-	(void) self;
-	(void) irn;
-	return arch_irn_class_normal;
-}
-
-static
-arch_irn_flags_t phi_get_flags(const void *self, const ir_node *irn)
-{
-	phi_attr_t *attr = get_Phi_attr(irn);
-	(void) self;
-	return attr->flags;
-}
-
-static
-ir_entity *phi_get_frame_entity(const void *self, const ir_node *irn)
-{
-	(void) self;
-	(void) irn;
-	return NULL;
-}
-
-static
-void phi_set_frame_entity(const void *self, ir_node *irn, ir_entity *ent)
-{
-	(void) self;
-	(void) irn;
-	(void) ent;
-	assert(0);
-}
-
-static
-void phi_set_frame_offset(const void *self, ir_node *irn, int bias)
-{
-	(void) self;
-	(void) irn;
-	(void) bias;
-	assert(0);
-}
+void be_dump_phi_reg_reqs(FILE *F, ir_node *node, dump_reason_t reason)
+{
+	switch (reason) {
+	case dump_node_opcode_txt:
+		fputs(get_op_name(get_irn_op(node)), F);
+		break;
+	case dump_node_mode_txt:
+		fprintf(F, "%s", get_mode_name(get_irn_mode(node)));
+		break;
+	case dump_node_nodeattr_txt:
+		break;
+	case dump_node_info_txt:
+	{
+		backend_info_t *info = be_get_info(node);
+		if (info != NULL && info->out_infos[0].req != NULL) {
+			arch_dump_reqs_and_registers(F, node);
+		}
+		break;
+	}
 
-static
-int phi_get_sp_bias(const void* self, const ir_node *irn)
-{
-	(void) self;
-	(void) irn;
-	return 0;
+	default:
+		break;
+	}
 }
 
 static const arch_irn_ops_t phi_irn_ops = {
 	phi_get_irn_reg_req,
-	phi_set_irn_reg,
-	phi_get_irn_reg,
-	phi_classify,
-	phi_get_flags,
-	phi_get_frame_entity,
-	phi_set_frame_entity,
-	phi_set_frame_offset,
-	phi_get_sp_bias,
+	dummy_classify,
+	dummy_get_frame_entity,
+	dummy_set_frame_offset,
+	dummy_get_sp_bias,
 	NULL,    /* get_inverse             */
 	NULL,    /* get_op_estimated_cost   */
 	NULL,    /* possible_memory_operand */
 	NULL,    /* perform_memory_operand  */
 };
 
-void be_phi_handler_new(be_main_env_t *env)
-{
-	phi_handler.arch_env  = &env->arch_env;
-	phi_handler.phi_attrs = pmap_create();
-	op_Phi->ops.be_ops    = &phi_irn_ops;
-}
-
-void be_phi_handler_free(void)
-{
-	pmap_destroy(phi_handler.phi_attrs);
-	phi_handler.phi_attrs = NULL;
-	op_Phi->ops.be_ops    = NULL;
-}
-
-void be_phi_handler_reset(void)
-{
-	if(phi_handler.phi_attrs)
-		pmap_destroy(phi_handler.phi_attrs);
-	phi_handler.phi_attrs = pmap_create();
-}
-
 /*
   _   _           _        ____                        _
  | \ | | ___   __| | ___  |  _ \ _   _ _ __ ___  _ __ (_)_ __   __ _
@@ -1510,87 +1327,33 @@ void be_phi_handler_reset(void)
                                                 |_|            |___/
 */
 
-/**
- * Dumps a register requirement to a file.
- */
-static void dump_node_req(FILE *f, int idx, const arch_register_req_t *req,
-                          const ir_node *node)
-{
-	int did_something = 0;
-	char buf[16];
-	const char *prefix = buf;
-
-	snprintf(buf, sizeof(buf), "#%d ", idx);
-	buf[sizeof(buf) - 1] = '\0';
-
-	if(req->cls != 0) {
-		char tmp[256];
-		fprintf(f, prefix);
-		arch_register_req_format(tmp, sizeof(tmp), req, node);
-		fprintf(f, "%s", tmp);
-		did_something = 1;
-	}
-
-	if(did_something)
-		fprintf(f, "\n");
-}
-
-/**
- * Dumps node register requirements to a file.
- */
-static void dump_node_reqs(FILE *f, ir_node *node)
-{
-	int i;
-	be_node_attr_t *a = get_irn_attr(node);
-	int len = ARR_LEN(a->reg_data);
-
-	fprintf(f, "registers: \n");
-	for(i = 0; i < len; ++i) {
-		be_reg_data_t *rd = &a->reg_data[i];
-		if(rd->reg)
-			fprintf(f, "#%d: %s\n", i, rd->reg->name);
-	}
-
-	fprintf(f, "in requirements:\n");
-	for(i = 0; i < len; ++i) {
-		dump_node_req(f, i, &a->reg_data[i].in_req.req, node);
-	}
-
-	fprintf(f, "\nout requirements:\n");
-	for(i = 0; i < len; ++i) {
-		dump_node_req(f, i, &a->reg_data[i].req.req, node);
-	}
-}
-
 /**
  * ir_op-Operation: dump a be node to file
  */
-static int dump_node(ir_node *irn, FILE *f, dump_reason_t reason)
+static void dump_node(FILE *f, ir_node *irn, dump_reason_t reason)
 {
-	be_node_attr_t *at = get_irn_attr(irn);
-
 	assert(is_be_node(irn));
 
-	switch(reason) {
+	switch (reason) {
 		case dump_node_opcode_txt:
-			fprintf(f, get_op_name(get_irn_op(irn)));
+			fputs(get_op_name(get_irn_op(irn)), f);
 			break;
 		case dump_node_mode_txt:
-			if(be_is_Perm(irn) || be_is_Copy(irn) || be_is_CopyKeep(irn)) {
-				fprintf(f, " %s", get_mode_name(get_irn_mode(irn)));
+			if (be_is_Copy(irn) || be_is_CopyKeep(irn)) {
+				fprintf(f, "%s", get_mode_name(get_irn_mode(irn)));
 			}
 			break;
 		case dump_node_nodeattr_txt:
-			if(be_is_Call(irn)) {
-				be_call_attr_t *a = (be_call_attr_t *) at;
+			if (be_is_Call(irn)) {
+				const be_call_attr_t *a = get_irn_generic_attr_const(irn);
 				if (a->ent)
 					fprintf(f, " [%s] ", get_entity_name(a->ent));
 			}
-			if(be_is_IncSP(irn)) {
+			if (be_is_IncSP(irn)) {
 				const be_incsp_attr_t *attr = get_irn_generic_attr_const(irn);
-				if(attr->offset == BE_STACK_FRAME_SIZE_EXPAND) {
+				if (attr->offset == BE_STACK_FRAME_SIZE_EXPAND) {
 					fprintf(f, " [Setup Stackframe] ");
-				} else if(attr->offset == BE_STACK_FRAME_SIZE_SHRINK) {
+				} else if (attr->offset == BE_STACK_FRAME_SIZE_SHRINK) {
 					fprintf(f, " [Destroy Stackframe] ");
 				} else {
 					fprintf(f, " [%d] ", attr->offset);
@@ -1598,10 +1361,10 @@ static int dump_node(ir_node *irn, FILE *f, dump_reason_t reason)
 			}
 			break;
 		case dump_node_info_txt:
-			dump_node_reqs(f, irn);
+			arch_dump_reqs_and_registers(f, irn);
 
-			if(be_has_frame_entity(irn)) {
-				be_frame_attr_t *a = (be_frame_attr_t *) at;
+			if (be_has_frame_entity(irn)) {
+				const be_frame_attr_t *a = get_irn_generic_attr_const(irn);
 				if (a->ent) {
 					unsigned size = get_type_size_bytes(get_entity_type(a->ent));
 					ir_fprintf(f, "frame entity: %+F, offset 0x%x (%d), size 0x%x (%d) bytes\n",
@@ -1611,101 +1374,98 @@ static int dump_node(ir_node *irn, FILE *f, dump_reason_t reason)
 			}
 
 			switch (get_irn_opcode(irn)) {
-			case beo_IncSP:
-				{
-					be_incsp_attr_t *a = (be_incsp_attr_t *) at;
-					if (a->offset == BE_STACK_FRAME_SIZE_EXPAND)
-						fprintf(f, "offset: FRAME_SIZE\n");
-					else if(a->offset == BE_STACK_FRAME_SIZE_SHRINK)
-						fprintf(f, "offset: -FRAME SIZE\n");
-					else
-						fprintf(f, "offset: %u\n", a->offset);
-				}
+			case beo_IncSP: {
+				const be_incsp_attr_t *a = get_irn_generic_attr_const(irn);
+				fprintf(f, "align: %d\n", a->align);
+				if (a->offset == BE_STACK_FRAME_SIZE_EXPAND)
+					fprintf(f, "offset: FRAME_SIZE\n");
+				else if (a->offset == BE_STACK_FRAME_SIZE_SHRINK)
+					fprintf(f, "offset: -FRAME SIZE\n");
+				else
+					fprintf(f, "offset: %d\n", a->offset);
 				break;
-			case beo_Call:
-				{
-					be_call_attr_t *a = (be_call_attr_t *) at;
+			}
+			case beo_Call: {
+				const be_call_attr_t *a = get_irn_generic_attr_const(irn);
 
-					if (a->ent)
-						fprintf(f, "\ncalling: %s\n", get_entity_name(a->ent));
-				}
+				if (a->ent)
+					fprintf(f, "\ncalling: %s\n", get_entity_name(a->ent));
 				break;
-			case beo_MemPerm:
-				{
-					int i;
-					for(i = 0; i < be_get_MemPerm_entity_arity(irn); ++i) {
-						ir_entity *in, *out;
-						in = be_get_MemPerm_in_entity(irn, i);
-						out = be_get_MemPerm_out_entity(irn, i);
-						if(in) {
-							fprintf(f, "\nin[%d]: %s\n", i, get_entity_name(in));
-						}
-						if(out) {
-							fprintf(f, "\nout[%d]: %s\n", i, get_entity_name(out));
-						}
+			}
+			case beo_MemPerm: {
+				int i;
+				for (i = 0; i < be_get_MemPerm_entity_arity(irn); ++i) {
+					ir_entity *in, *out;
+					in = be_get_MemPerm_in_entity(irn, i);
+					out = be_get_MemPerm_out_entity(irn, i);
+					if (in) {
+						fprintf(f, "\nin[%d]: %s\n", i, get_entity_name(in));
+					}
+					if (out) {
+						fprintf(f, "\nout[%d]: %s\n", i, get_entity_name(out));
 					}
 				}
 				break;
+			}
 
 			default:
 				break;
 			}
 	}
-
-	return 0;
 }
 
 /**
  * ir_op-Operation:
  * Copies the backend specific attributes from old node to new node.
  */
-static void copy_attr(const ir_node *old_node, ir_node *new_node)
+static void copy_attr(ir_graph *irg, const ir_node *old_node, ir_node *new_node)
 {
 	const be_node_attr_t *old_attr = get_irn_attr_const(old_node);
 	be_node_attr_t *new_attr = get_irn_attr(new_node);
-	struct obstack *obst = get_irg_obstack(get_irn_irg(new_node));
-	unsigned i, len;
+	struct obstack *obst     = be_get_be_obst(irg);
+	backend_info_t *old_info = be_get_info(old_node);
+	backend_info_t *new_info = be_get_info(new_node);
 
 	assert(is_be_node(old_node));
 	assert(is_be_node(new_node));
 
 	memcpy(new_attr, old_attr, get_op_attr_size(get_irn_op(old_node)));
-	new_attr->reg_data = NULL;
 
-	if(old_attr->reg_data != NULL)
-		len = ARR_LEN(old_attr->reg_data);
-	else
-		len = 0;
-
-	if(get_irn_op(old_node)->opar == oparity_dynamic
-			|| be_is_RegParams(old_node)) {
-		new_attr->reg_data = NEW_ARR_F(be_reg_data_t, len);
+	if (old_info->out_infos != NULL) {
+		unsigned n_outs = ARR_LEN(old_info->out_infos);
+		/* need dyanmic out infos? */
+		if (be_is_Barrier(new_node)	|| be_is_Perm(new_node)) {
+			new_info->out_infos = NEW_ARR_F(reg_out_info_t, n_outs);
+		} else {
+			new_info->out_infos = NEW_ARR_D(reg_out_info_t, obst, n_outs);
+		}
+		memcpy(new_info->out_infos, old_info->out_infos,
+			   n_outs * sizeof(new_info->out_infos[0]));
 	} else {
-		new_attr->reg_data = NEW_ARR_D(be_reg_data_t, obst, len);
+		new_info->out_infos = NULL;
 	}
 
-	if(len > 0) {
-		memcpy(new_attr->reg_data, old_attr->reg_data, len * sizeof(be_reg_data_t));
-		for(i = 0; i < len; ++i) {
-			const be_reg_data_t *rd = &old_attr->reg_data[i];
-			be_reg_data_t *newrd = &new_attr->reg_data[i];
-			if(arch_register_req_is(&rd->req.req, limited)) {
-				const arch_register_req_t *req = &rd->req.req;
-				arch_register_req_t *new_req = &newrd->req.req;
-				new_req->limited
-					= rbitset_duplicate_obstack_alloc(obst, req->limited, req->cls->n_regs);
-			}
-			if(arch_register_req_is(&rd->in_req.req, limited)) {
-				const arch_register_req_t *req = &rd->in_req.req;
-				arch_register_req_t *new_req = &newrd->in_req.req;
-				new_req->limited
-					= rbitset_duplicate_obstack_alloc(obst, req->limited, req->cls->n_regs);
-			}
+	/* input infos */
+	if (old_attr->reg_data != NULL) {
+		unsigned n_ins = ARR_LEN(old_attr->reg_data);
+		/* need dynamic in infos? */
+		if (get_irn_op(old_node)->opar == oparity_dynamic) {
+			new_attr->reg_data = NEW_ARR_F(be_reg_data_t, n_ins);
+		} else {
+			new_attr->reg_data = NEW_ARR_D(be_reg_data_t, obst, n_ins);
 		}
+		memcpy(new_attr->reg_data, old_attr->reg_data,
+		       n_ins * sizeof(be_reg_data_t));
+	} else {
+		new_attr->reg_data = NULL;
 	}
 }
 
 static const ir_op_ops be_node_op_ops = {
+	firm_default_hash,
+	NULL,
+	NULL,
+	NULL,
 	NULL,
 	NULL,
 	NULL,
@@ -1727,13 +1487,9 @@ int is_be_node(const ir_node *irn)
 	return get_op_ops(get_irn_op(irn))->be_ops == &be_node_irn_ops;
 }
 
-void be_node_init(void) {
-	static int inited = 0;
-
-	if(inited)
-		return;
-
-	inited = 1;
+void be_init_op(void)
+{
+	ir_opcode opc;
 
 	/* Acquire all needed opcodes. */
 	op_be_Spill      = new_ir_op(beo_Spill,     "be_Spill",     op_pin_state_pinned, N,   oparity_unary,    0, sizeof(be_frame_attr_t),   &be_node_op_ops);
@@ -1741,17 +1497,16 @@ void be_node_init(void) {
 	op_be_Perm       = new_ir_op(beo_Perm,      "be_Perm",      op_pin_state_pinned, N,   oparity_variable, 0, sizeof(be_node_attr_t),    &be_node_op_ops);
 	op_be_MemPerm    = new_ir_op(beo_MemPerm,   "be_MemPerm",   op_pin_state_pinned, N,   oparity_variable, 0, sizeof(be_memperm_attr_t), &be_node_op_ops);
 	op_be_Copy       = new_ir_op(beo_Copy,      "be_Copy",      op_pin_state_floats, N,   oparity_unary,    0, sizeof(be_node_attr_t),    &be_node_op_ops);
-	op_be_Keep       = new_ir_op(beo_Keep,      "be_Keep",      op_pin_state_pinned, K,   oparity_dynamic,  0, sizeof(be_node_attr_t),    &be_node_op_ops);
-	op_be_CopyKeep   = new_ir_op(beo_CopyKeep,  "be_CopyKeep",  op_pin_state_pinned, K,   oparity_variable, 0, sizeof(be_node_attr_t),    &be_node_op_ops);
+	op_be_Keep       = new_ir_op(beo_Keep,      "be_Keep",      op_pin_state_floats, K,   oparity_dynamic,  0, sizeof(be_node_attr_t),    &be_node_op_ops);
+	op_be_CopyKeep   = new_ir_op(beo_CopyKeep,  "be_CopyKeep",  op_pin_state_floats, K,   oparity_variable, 0, sizeof(be_node_attr_t),    &be_node_op_ops);
 	op_be_Call       = new_ir_op(beo_Call,      "be_Call",      op_pin_state_pinned, F|M, oparity_variable, 0, sizeof(be_call_attr_t),    &be_node_op_ops);
 	op_be_Return     = new_ir_op(beo_Return,    "be_Return",    op_pin_state_pinned, X,   oparity_dynamic,  0, sizeof(be_return_attr_t),  &be_node_op_ops);
 	op_be_AddSP      = new_ir_op(beo_AddSP,     "be_AddSP",     op_pin_state_pinned, N,   oparity_unary,    0, sizeof(be_node_attr_t),    &be_node_op_ops);
 	op_be_SubSP      = new_ir_op(beo_SubSP,     "be_SubSP",     op_pin_state_pinned, N,   oparity_unary,    0, sizeof(be_node_attr_t),    &be_node_op_ops);
 	op_be_IncSP      = new_ir_op(beo_IncSP,     "be_IncSP",     op_pin_state_pinned, N,   oparity_unary,    0, sizeof(be_incsp_attr_t),   &be_node_op_ops);
-	op_be_RegParams  = new_ir_op(beo_RegParams, "be_RegParams", op_pin_state_pinned, N,   oparity_zero,     0, sizeof(be_node_attr_t),    &be_node_op_ops);
+	op_be_Start      = new_ir_op(beo_Start,     "be_Start",     op_pin_state_pinned, N,   oparity_zero,     0, sizeof(be_node_attr_t),    &be_node_op_ops);
 	op_be_FrameAddr  = new_ir_op(beo_FrameAddr, "be_FrameAddr", op_pin_state_floats, N,   oparity_unary,    0, sizeof(be_frame_attr_t),   &be_node_op_ops);
 	op_be_Barrier    = new_ir_op(beo_Barrier,   "be_Barrier",   op_pin_state_pinned, N,   oparity_dynamic,  0, sizeof(be_node_attr_t),    &be_node_op_ops);
-	op_be_Unwind     = new_ir_op(beo_Unwind,    "be_Unwind",    op_pin_state_pinned, X,   oparity_zero,     0, sizeof(be_node_attr_t),    &be_node_op_ops);
 
 	op_be_Spill->ops.node_cmp_attr     = FrameAddr_cmp_attr;
 	op_be_Reload->ops.node_cmp_attr    = FrameAddr_cmp_attr;
@@ -1765,8 +1520,16 @@ void be_node_init(void) {
 	op_be_AddSP->ops.node_cmp_attr     = node_cmp_attr;
 	op_be_SubSP->ops.node_cmp_attr     = node_cmp_attr;
 	op_be_IncSP->ops.node_cmp_attr     = IncSP_cmp_attr;
-	op_be_RegParams->ops.node_cmp_attr = node_cmp_attr;
+	op_be_Start->ops.node_cmp_attr     = node_cmp_attr;
 	op_be_FrameAddr->ops.node_cmp_attr = FrameAddr_cmp_attr;
 	op_be_Barrier->ops.node_cmp_attr   = node_cmp_attr;
-	op_be_Unwind->ops.node_cmp_attr    = node_cmp_attr;
+
+	/* attach out dummy_ops to middle end nodes */
+	for (opc = iro_First; opc <= iro_Last; ++opc) {
+		ir_op *op = get_irp_opcode(opc);
+		assert(op->ops.be_ops == NULL);
+		op->ops.be_ops = &dummy_be_irn_ops;
+	}
+
+	op_Phi->ops.be_ops = &phi_irn_ops;
 }