remove unnecessary members from be_abi_irg_t structure, cleanup beabi a bit
[libfirm] / ir / be / bespillslots.c
index a7a13ed..359dc7d 100644 (file)
 /*
- * Author:      Matthias Braun
- * Date:               26.7.06
- * Copyright:   (c) Universitaet Karlsruhe
- * Licence:     This file protected by GPL -  GNU GENERAL PUBLIC LICENSE.
+ * Copyright (C) 1995-2008 University of Karlsruhe.  All right reserved.
+ *
+ * This file is part of libFirm.
+ *
+ * This file may be distributed and/or modified under the terms of the
+ * GNU General Public License version 2 as published by the Free Software
+ * Foundation and appearing in the file LICENSE.GPL included in the
+ * packaging of this file.
+ *
+ * Licensees holding valid libFirm Professional Edition licenses may use
+ * this file in accordance with the libFirm Commercial License.
+ * Agreement provided with the Software.
+ *
+ * This file is provided AS IS with NO WARRANTY OF ANY KIND, INCLUDING THE
+ * WARRANTY OF DESIGN, MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE.
+ */
+
+/**
+ * @file
+ * @brief       Spillslot coalescer.
+ * @author      Matthias Braun
+ * @date        26.07.2006
+ * @version     $Id$
  */
-#ifdef HAVE_CONFIG_H
 #include "config.h"
-#endif
 
 #include <stdlib.h>
 
 #include "set.h"
-
+#include "array.h"
 #include "irgwalk.h"
 #include "ircons.h"
 #include "irprintf.h"
 #include "execfreq.h"
 #include "unionfind.h"
-#include "type.h"
 #include "irdump_t.h"
 
-#include "benode_t.h"
+#include "benode.h"
 #include "besched.h"
+#include "bespill.h"
 #include "bespillslots.h"
 #include "bechordal_t.h"
-#include "bejavacoal.h"
-#include "benodesets.h"
 #include "bestatevent.h"
-#include "bespilloptions.h"
+#include "bemodule.h"
+#include "beintlive_t.h"
+#include "beirg.h"
+#include "bearch.h"
 
 #define DBG_COALESCING         1
 #define DBG_INTERFERENCES      2
 
-DEBUG_ONLY(
-static firm_dbg_module_t *dbg = NULL;
-)
+DEBUG_ONLY(static firm_dbg_module_t *dbg = NULL;)
 
 typedef struct _spill_t {
-       ir_node *spill;
-       /** regclass of the spilled value */
-       const arch_register_class_t *cls;
-       /** index into spillslot_unionfind unionfind structure */
-       int spillslot;
+       ir_node       *spill;
+       const ir_mode *mode;      /**< mode of the spilled value */
+       int            alignment; /**< alignment for the spilled value */
+       int            spillslot; /**< index into spillslot_unionfind structure */
 } spill_t;
 
 typedef struct _affinity_edge_t {
        double affinity;
-       int slot1, slot2;
+       int    slot1;
+       int    slot2;
 } affinity_edge_t;
 
-typedef struct _ss_env_t {
-       struct obstack obst;
+struct _be_fec_env_t {
+       struct obstack    obst;
        const arch_env_t *arch_env;
-       const be_chordal_env_t *chordal_env;
-       set *spills;
-       ir_node **reloads;
+       ir_graph         *irg;
+       set              *spills;
+       ir_node         **reloads;
        affinity_edge_t **affinity_edges;
-       set *memperms;
-} ss_env_t;
+       set              *memperms;
+};
 
 /** Compare 2 affinity edges (used in quicksort) */
-static int cmp_affinity(const void *d1, const void *d2) {
+static int cmp_affinity(const void *d1, const void *d2)
+{
        const affinity_edge_t * const *e1 = d1;
        const affinity_edge_t * const *e2 = d2;
 
-       // sort in descending order
+       /* sort in descending order */
        return (*e1)->affinity < (*e2)->affinity ? 1 : -1;
 }
 
-static int cmp_spill(const void* d1, const void* d2, size_t size) {
+static int cmp_spill(const void* d1, const void* d2, size_t size)
+{
        const spill_t* s1 = d1;
        const spill_t* s2 = d2;
+       (void) size;
+
        return s1->spill != s2->spill;
 }
 
-static spill_t *get_spill(ss_env_t *env, ir_node *node) {
+static spill_t *get_spill(be_fec_env_t *env, ir_node *node)
+{
        spill_t spill, *res;
-       int hash = nodeset_hash(node);
+       int hash = hash_irn(node);
 
        spill.spill = node;
        res = set_find(env->spills, &spill, sizeof(spill), hash);
@@ -84,150 +106,124 @@ static spill_t *get_spill(ss_env_t *env, ir_node *node) {
        return res;
 }
 
-/*
- *   ____      _ _           _     ____        _ _ _
- *  / ___|___ | | | ___  ___| |_  / ___| _ __ (_) | |___
- * | |   / _ \| | |/ _ \/ __| __| \___ \| '_ \| | | / __|
- * | |__| (_) | | |  __/ (__| |_   ___) | |_) | | | \__ \
- *  \____\___/|_|_|\___|\___|\__| |____/| .__/|_|_|_|___/
- *                                      |_|
- */
 
-static ir_node *get_memory_edge(const ir_node *node) {
+static inline ir_node *get_memory_edge(const ir_node *node)
+{
        int i, arity;
 
        arity = get_irn_arity(node);
-       for(i = arity - 1; i >= 0; --i) {
+       for (i = arity - 1; i >= 0; --i) {
                ir_node *arg = get_irn_n(node, i);
-               if(get_irn_mode(arg) == mode_M)
+               if (get_irn_mode(arg) == mode_M)
                        return arg;
        }
 
        return NULL;
 }
 
-static spill_t *collect_spill(ss_env_t *env, ir_node *node) {
-       const arch_env_t *arch_env = env->arch_env;
-       const arch_register_class_t *cls;
+static spill_t *collect_spill(be_fec_env_t *env, ir_node *node,
+                                     const ir_mode *mode, int align)
+{
        spill_t spill, *res;
-       int hash = nodeset_hash(node);
-
-       assert(arch_irn_class_is(arch_env, node, spill));
-
-       if(be_is_Spill(node)) {
-               cls = arch_get_irn_reg_class(arch_env, node, be_pos_Spill_val);
-       } else {
-               // TODO add a way to detect the type of the spilled value
-               assert(0);
-       }
+       int     hash = hash_irn(node);
 
+       /* insert into set of spills if not already there */
        spill.spill = node;
-       res = set_find(env->spills, &spill, sizeof(spill), hash);
+       res         = set_find(env->spills, &spill, sizeof(spill), hash);
 
-       if(res == NULL) {
+       if (res == NULL) {
                spill.spillslot = set_count(env->spills);
-               spill.cls = cls;
-               res = set_insert(env->spills, &spill, sizeof(spill), hash);
+               spill.mode      = mode;
+               spill.alignment = align;
+               res             = set_insert(env->spills, &spill, sizeof(spill), hash);
+               DB((dbg, DBG_COALESCING, "Slot %d: %+F\n", spill.spillslot, node));
+       } else {
+               assert(res->mode == mode);
+               assert(res->alignment == align);
        }
 
        return res;
 }
 
-static spill_t *collect_memphi(ss_env_t *env, ir_node *node) {
+static spill_t *collect_memphi(be_fec_env_t *env, ir_node *node,
+                               const ir_mode *mode, int align)
+{
        int i, arity;
        spill_t spill, *res;
-       int hash = nodeset_hash(node);
+       int hash = hash_irn(node);
+       const ir_exec_freq *exec_freq = be_get_irg_exec_freq(env->irg);
 
        assert(is_Phi(node));
 
        spill.spill = node;
        res = set_find(env->spills, &spill, sizeof(spill), hash);
-       if(res != NULL) {
+       if (res != NULL) {
+               assert(res->mode == mode);
+               assert(res->alignment == align);
                return res;
        }
 
        spill.spillslot = set_count(env->spills);
-       spill.cls = NULL;
-       res = set_insert(env->spills, &spill, sizeof(spill), hash);
+       spill.mode      = mode;
+       spill.alignment = align;
+       DB((dbg, DBG_COALESCING, "Slot %d: %+F\n", spill.spillslot, node));
+       res             = set_insert(env->spills, &spill, sizeof(spill), hash);
 
-       // is 1 of the arguments a spill?
-       for(i = 0, arity = get_irn_arity(node); i < arity; ++i) {
+       /* collect attached spills and mem-phis */
+       arity = get_irn_arity(node);
+       for (i = 0; i < arity; ++i) {
                affinity_edge_t *affinty_edge;
-               ir_node* arg = get_irn_n(node, i);
-               spill_t* arg_spill;
-
-               if(be_is_Spill(arg)) {
-                       arg_spill = collect_spill(env, arg);
-               } else {
-                       // if it wasn't a spill then it must be a Mem-Phi
-                       assert(is_Phi(arg));
-                       arg_spill = collect_memphi(env, arg);
-               }
+               ir_node *arg = get_irn_n(node, i);
+               spill_t *arg_spill;
 
-               if(res->cls == NULL) {
-                       res->cls = arg_spill->cls;
+               if (is_Phi(arg)) {
+                       arg_spill = collect_memphi(env, arg, mode, align);
                } else {
-                       assert(arg_spill->cls == NULL || res->cls == arg_spill->cls);
+                       arg_spill = collect_spill(env, arg, mode, align);
                }
 
-               // add an affinity edge
-               affinty_edge = obstack_alloc(&env->obst, sizeof(affinty_edge[0]));
-               affinty_edge->affinity = get_block_execfreq(env->chordal_env->exec_freq, get_nodes_block(arg));
-               affinty_edge->slot1 = res->spillslot;
-               affinty_edge->slot2 = arg_spill->spillslot;
+               /* add an affinity edge */
+               affinty_edge           = OALLOC(&env->obst, affinity_edge_t);
+               affinty_edge->affinity = get_block_execfreq(exec_freq, get_nodes_block(arg));
+               affinty_edge->slot1    = res->spillslot;
+               affinty_edge->slot2    = arg_spill->spillslot;
                ARR_APP1(affinity_edge_t*, env->affinity_edges, affinty_edge);
        }
 
        return res;
 }
 
-/**
- * This walker function searches for reloads and collects all the spills
- * and memphis attached to them.
- */
-static void collect_spills_walker(ir_node *node, void *data) {
-       ss_env_t *env = data;
-       const arch_env_t *arch_env = env->arch_env;
-
-       // classify returns classification of the irn the proj is attached to
-       if(is_Proj(node))
-               return;
-
-       if(arch_irn_class_is(arch_env, node, reload)) {
-               ir_node *spillnode = get_memory_edge(node);
-               spill_t *spill;
-
-               assert(spillnode != NULL);
+void be_node_needs_frame_entity(be_fec_env_t *env, ir_node *node,
+                                const ir_mode *mode, int align)
+{
+       ir_node *spillnode = get_memory_edge(node);
+       spill_t *spill;
 
-               if(is_Phi(spillnode)) {
-                       spill = collect_memphi(env, spillnode);
-               } else {
-                       spill = collect_spill(env, spillnode);
-               }
+       assert(spillnode != NULL);
 
-               assert(!be_is_Reload(node) || spill->cls == arch_get_irn_reg_class(arch_env, node, -1));
-               ARR_APP1(ir_node*, env->reloads, node);
+       /* walk upwards and collect all phis and spills on this way */
+       if (is_Phi(spillnode)) {
+               spill = collect_memphi(env, spillnode, mode, align);
+       } else {
+               spill = collect_spill(env, spillnode, mode, align);
        }
+
+       ARR_APP1(ir_node *, env->reloads, node);
 }
 
-/*
- *   ____            _                      ____  _       _
- *  / ___|___   __ _| | ___  ___  ___ ___  / ___|| | ___ | |_ ___
- * | |   / _ \ / _` | |/ _ \/ __|/ __/ _ \ \___ \| |/ _ \| __/ __|
- * | |__| (_) | (_| | |  __/\__ \ (_|  __/  ___) | | (_) | |_\__ \
- *  \____\___/ \__,_|_|\___||___/\___\___| |____/|_|\___/ \__|___/
- */
 
-static int merge_interferences(ss_env_t *env, bitset_t** interferences, int* spillslot_unionfind, int s1, int s2)
+
+static int merge_interferences(be_fec_env_t *env, bitset_t** interferences,
+                               int* spillslot_unionfind, int s1, int s2)
 {
        int res;
        int i;
        int spillcount;
 
-       // merge spillslots and interferences
+       /* merge spillslots and interferences */
        res = uf_union(spillslot_unionfind, s1, s2);
-       // we assume that we always merge s2 to s1 so swap s1, s2 if necessary
-       if(res != 0) {
+       /* we assume that we always merge s2 to s1 so swap s1, s2 if necessary */
+       if (res != s1) {
                int t = s1;
                s1 = s2;
                s2 = t;
@@ -235,93 +231,116 @@ static int merge_interferences(ss_env_t *env, bitset_t** interferences, int* spi
 
        bitset_or(interferences[s1], interferences[s2]);
 
-       // update other interferences
+       /* update other interferences */
        spillcount = set_count(env->spills);
-       for(i = 0; i < spillcount; ++i) {
+       for (i = 0; i < spillcount; ++i) {
                bitset_t *intfs = interferences[i];
-               if(bitset_is_set(intfs, s2))
+               if (bitset_is_set(intfs, s2))
                        bitset_set(intfs, s1);
        }
 
        return res;
 }
 
-#if 0
-
-static void dump_interference_graph(ss_env_t *env, bitset_t **interferences, const char* suffix) {
-       char name[256];
-       int i;
-       int spillcount;
-       spill_t *spill;
-       FILE *f;
-       static int cnt = 0;
-
-       snprintf(name, sizeof(name), "%d-%s-spillslots-%s.vcg", cnt++, get_irg_dump_name(env->chordal_env->birg->irg), suffix);
-
-       f = fopen(name, "w");
-       assert(f != NULL);
-
-       fprintf(f, "graph: {\n");
-
-       spillcount = set_count(env->spills);
-       for(spill = set_first(env->spills), i = 0; spill != NULL; spill = set_next(env->spills), ++i) {
-               int slotid = spill->spillslot;
-               fprintf(f, "\tnode: { title: \"n%d\" label: \"%d\" }\n", i, slotid);
-       }
-
-       for(i = 0; i < ARR_LEN(env->affinity_edges); ++i) {
-               affinity_edge_t *edge = env->affinity_edges[i];
-               fprintf(f, "\tedge: { sourcename: \"n%d\" targetname: \"n%d\" color: green }\n", edge->slot1, edge->slot2);
-       }
-
-       for(i = 0; i < spillcount; ++i) {
-               int i2;
-               for(i2 = 0; i2 < spillcount; ++i2) {
-                       if(bitset_is_set(interferences[i], i2)) {
-                               fprintf(f, "\tedge: { sourcename: \"n%d\" targetname: \"n%d\" color: red }\n", i, i2);
+static int my_values_interfere2(ir_graph *irg, const ir_node *a,
+                                const ir_node *b)
+{
+       be_lv_t *lv = be_get_irg_liveness(irg);
+
+    int a2b = _value_dominates(a, b);
+    int b2a = _value_dominates(b, a);
+
+    /* If there is no dominance relation, they do not interfere. */
+    if ((a2b | b2a) > 0) {
+        const ir_edge_t *edge;
+        ir_node *bb;
+
+        /*
+         * Adjust a and b so, that a dominates b if
+         * a dominates b or vice versa.
+         */
+        if (b2a) {
+            const ir_node *t = a;
+            a = b;
+            b = t;
+        }
+
+        bb = get_nodes_block(b);
+
+        /*
+         * If a is live end in b's block it is
+         * live at b's definition (a dominates b)
+         */
+        if (be_is_live_end(lv, bb, a))
+            return 1;
+
+        /*
+         * Look at all usages of a.
+         * If there's one usage of a in the block of b, then
+         * we check, if this use is dominated by b, if that's true
+         * a and b interfere. Note that b must strictly dominate the user,
+         * since if b is the last user of in the block, b and a do not
+         * interfere.
+         * Uses of a not in b's block can be disobeyed, because the
+         * check for a being live at the end of b's block is already
+         * performed.
+         */
+        foreach_out_edge(a, edge) {
+            const ir_node *user = get_edge_src_irn(edge);
+                       if (is_Sync(user)) {
+                               const ir_edge_t *edge2;
+                               foreach_out_edge(user, edge2) {
+                                       const ir_node *user2 = get_edge_src_irn(edge2);
+                                       assert(!is_Sync(user2));
+                                       if (get_nodes_block(user2) == bb && !is_Phi(user2) &&
+                                          _value_strictly_dominates(b, user2))
+                                               return 1;
+                               }
+                       } else {
+                               if (get_nodes_block(user) == bb && !is_Phi(user) &&
+                                               _value_strictly_dominates(b, user))
+                return 1;
                        }
-               }
-       }
+        }
+    }
 
-       fprintf(f, "}\n");
-       fclose(f);
+       return 0;
 }
 
-static void show_stats(ss_env_t *env) {
-       int spillcount;
-       int slotcount;
-       int *slotused;
-       spill_t *spill;
-
-       spillcount = set_count(env->spills);
-       fprintf(stderr, "%s: Collected %d spills\n", get_irg_dump_name(env->chordal_env->birg->irg), spillcount);
-
-       slotused = alloca(spillcount * sizeof(slotused[0]));
-       memset(slotused, 0, spillcount * sizeof(slotused[0]));
-
-       slotcount = 0;
-       for(spill = set_first(env->spills); spill != NULL; spill = set_next(env->spills)) {
-               int slot = spill->spillslot;
-               if(slotused[slot] == 0) {
-                       slotused[slot] = 1;
-                       slotcount++;
+/**
+ * same as values_interfere but with special handling for Syncs
+ */
+static int my_values_interfere(ir_graph *irg, ir_node *a, ir_node *b)
+{
+       if (is_Sync(a)) {
+               int i, arity = get_irn_arity(a);
+               for (i = 0; i < arity; ++i) {
+                       ir_node *in = get_irn_n(a, i);
+                       if (my_values_interfere(irg, in, b))
+                               return 1;
+               }
+               return 0;
+       } else if (is_Sync(b)) {
+               int i, arity = get_irn_arity(b);
+               for (i = 0; i < arity; ++i) {
+                       ir_node *in = get_irn_n(b, i);
+                       /* a is not a sync, so no need for my_values_interfere */
+                       if (my_values_interfere2(irg, a, in))
+                               return 1;
                }
+               return 0;
        }
 
-       fprintf(stderr, "%s: Coalesced to %d spillslots\n", get_irg_dump_name(env->chordal_env->birg->irg), slotcount);
+       return my_values_interfere2(irg, a, b);
 }
 
-#endif
-
-static void assign_spillslots(ss_env_t *env);
-
 /**
  * A greedy coalescing algorithm for spillslots:
  *  1. Sort the list of affinity edges
  *  2. Try to merge slots with affinity edges (most expensive slots first)
  *  3. Try to merge everything else that is possible
  */
-static void do_greedy_coalescing(ss_env_t *env)
+static void do_greedy_coalescing(be_fec_env_t *env)
 {
        int spillcount;
        spill_t **spilllist;
@@ -330,183 +349,139 @@ static void do_greedy_coalescing(ss_env_t *env)
        int affinity_edge_count;
        bitset_t **interferences;
        int* spillslot_unionfind;
+       struct obstack data;
 
        spillcount = set_count(env->spills);
-       if(spillcount == 0)
+       if (spillcount == 0)
                return;
 
-       DBG((dbg, DBG_COALESCING, "Coalescing %d spillslots\n", spillcount));
+       obstack_init(&data);
+
+       DB((dbg, DBG_COALESCING, "Coalescing %d spillslots\n", spillcount));
 
-       interferences = alloca(spillcount * sizeof(interferences[0]));
-       spillslot_unionfind = alloca(spillcount * sizeof(spillslot_unionfind[0]));
-       spilllist = alloca(spillcount * sizeof(spilllist[0]));
+       interferences       = OALLOCN(&data, bitset_t*, spillcount);
+       spillslot_unionfind = OALLOCN(&data, int,       spillcount);
+       spilllist           = OALLOCN(&data, spill_t*,  spillcount);
 
-       uf_init(spillslot_unionfind, 0, spillcount);
+       uf_init(spillslot_unionfind, spillcount);
 
        DEBUG_ONLY(
                memset(spilllist, 0, spillcount * sizeof(spilllist[0]));
        );
 
-       for(spill = set_first(env->spills), i = 0; spill != NULL; spill = set_next(env->spills), ++i) {
+       for (spill = set_first(env->spills), i = 0; spill != NULL;
+           spill = set_next(env->spills), ++i) {
                assert(spill->spillslot < spillcount);
                spilllist[spill->spillslot] = spill;
        }
 
-       for(i = 0; i < spillcount; ++i) {
-               interferences[i] = bitset_alloca(spillcount);
+       for (i = 0; i < spillcount; ++i) {
+               interferences[i] = bitset_obstack_alloc(&data, spillcount);
        }
 
-       // construct interferences
-       for(i = 0; i < spillcount; ++i) {
-               for(i2 = i+1; i2 < spillcount; ++i2) {
-                       if(values_interfere(env->chordal_env->lv, spilllist[i]->spill, spilllist[i2]->spill)) {
-                               DBG((dbg, DBG_INTERFERENCES, "Slot %d and %d interfere\n", i, i2));
+       /* construct interferences */
+       for (i = 0; i < spillcount; ++i) {
+               ir_node *spill1 = spilllist[i]->spill;
+
+               if (is_NoMem(spill1))
+                       continue;
+
+               for (i2 = i+1; i2 < spillcount; ++i2) {
+                       ir_node *spill2 = spilllist[i2]->spill;
+
+                       if (is_NoMem(spill2))
+                               continue;
+
+                       if (my_values_interfere(env->irg, spill1, spill2)) {
+                               DB((dbg, DBG_INTERFERENCES,
+                                    "Slot %d and %d interfere\n", i, i2));
+
                                bitset_set(interferences[i], i2);
                                bitset_set(interferences[i2], i);
                        }
                }
        }
 
-       // sort affinity edges
+       /* sort affinity edges */
        affinity_edge_count = ARR_LEN(env->affinity_edges);
-       qsort(env->affinity_edges, affinity_edge_count, sizeof(env->affinity_edges[0]), cmp_affinity);
+       qsort(env->affinity_edges, affinity_edge_count,
+             sizeof(env->affinity_edges[0]), cmp_affinity);
 
-       //dump_interference_graph(env, interferences, "before");
+       /*dump_interference_graph(env, interferences, "before"); */
 
-       // try to merge affine nodes
-       for(i = 0; i < affinity_edge_count; ++i) {
+       /* try to merge affine nodes */
+       for (i = 0; i < affinity_edge_count; ++i) {
                const affinity_edge_t *edge = env->affinity_edges[i];
                int s1 = uf_find(spillslot_unionfind, edge->slot1);
                int s2 = uf_find(spillslot_unionfind, edge->slot2);
 
                /* test if values interfere */
-               if(bitset_is_set(interferences[s1], s2)) {
+               if (bitset_is_set(interferences[s1], s2)) {
                        assert(bitset_is_set(interferences[s2], s1));
                        continue;
                }
 
-               DBG((dbg, DBG_COALESCING, "Merging %d and %d because of affinity edge\n", s1, s2));
+               DB((dbg, DBG_COALESCING,
+                    "Merging %d and %d because of affinity edge\n", s1, s2));
 
                merge_interferences(env, interferences, spillslot_unionfind, s1, s2);
        }
 
-       // try to merge as much remaining spillslots as possible
-       for(i = 0; i < spillcount; ++i) {
+       /* try to merge as much remaining spillslots as possible */
+       for (i = 0; i < spillcount; ++i) {
                int s1 = uf_find(spillslot_unionfind, i);
-               if(s1 != i)
+               if (s1 != i)
                        continue;
 
-               for(i2 = i+1; i2 < spillcount; ++i2) {
+               for (i2 = i+1; i2 < spillcount; ++i2) {
                        int s2 = uf_find(spillslot_unionfind, i2);
-                       if(s2 != i2)
+                       if (s2 != i2)
                                continue;
 
                        /* test if values interfere
                         * we have to test n1-n2 and n2-n1, because only 1 side gets updated
                         * when node merging occurs
                         */
-                       if(bitset_is_set(interferences[s1], s2)) {
+                       if (bitset_is_set(interferences[s1], s2)) {
                                assert(bitset_is_set(interferences[s2], s1));
                                continue;
                        }
 
-                       DBG((dbg, DBG_COALESCING, "Merging %d and %d because it is possible\n", s1, s2));
+                       DB((dbg, DBG_COALESCING,
+                            "Merging %d and %d because it is possible\n", s1, s2));
 
-                       if(merge_interferences(env, interferences, spillslot_unionfind, s1, s2) != 0) {
-                               // we can break the loop here, because s2 is the new supernode now
-                               // and we'll test s2 again later anyway
+                       if (merge_interferences(env, interferences, spillslot_unionfind, s1, s2) != 0) {
+                               /* we can break the loop here, because s2 is the new supernode
+                                * now and we'll test s2 again later anyway */
                                break;
                        }
                }
        }
 
-       // assign spillslots to spills
-       for(i = 0; i < spillcount; ++i) {
+       /* assign spillslots to spills */
+       for (i = 0; i < spillcount; ++i) {
                spill_t *spill = spilllist[i];
 
                spill->spillslot = uf_find(spillslot_unionfind, i);
        }
 
-       //dump_interference_graph(env, interferences, "after");
+       /*dump_interference_graph(env, interferences, "after");*/
+       obstack_free(&data, 0);
 }
 
-#if 0
-static void do_java_coalescing(ss_env_t *env)
-{
-       int spillcount;
-       spill_t **spilllist;
-       spill_t *spill;
-       int i, i2;
-       be_java_coal_t *coal;
-
-       spillcount = set_count(env->spills);
-       if(spillcount == 0)
-               return;
 
-       spilllist = alloca(spillcount * sizeof(spilllist[0]));
-
-       DEBUG_ONLY(
-               memset(spilllist, 0, spillcount * sizeof(spilllist[0]));
-       );
-
-       coal = be_java_coal_init("spillslot coalescing", spillcount, spillcount, 1);
-
-       for(spill = set_first(env->spills), i = 0; spill != NULL; spill = set_next(env->spills), ++i) {
-               assert(spill->spillslot < spillcount);
-               DEBUG_ONLY(assert(spilllist[spill->spillslot] == NULL));
-               spilllist[spill->spillslot] = spill;
-
-               be_java_coal_set_color(coal, spill->spillslot, spill->spillslot);
-       }
-
-       // construct interferences
-       for(i = 0; i < spillcount; ++i) {
-               for(i2 = i+1; i2 < spillcount; ++i2) {
-                       if(values_interfere(env->chordal_env->lv, spilllist[i]->spill, spilllist[i2]->spill)) {
-                               be_java_coal_add_int_edge(coal, i, i2);
-                       }
-               }
-       }
-
-       for(i = 0; i < ARR_LEN(env->affinity_edges); ++i) {
-               const affinity_edge_t *edge = env->affinity_edges[i];
-               int n = edge->slot1;
-               int m = edge->slot2;
-               int costs = (int) (edge->affinity * 10000);
-               be_java_coal_add_aff_edge(coal, n, m, costs);
-       }
-
-       be_java_coal_coalesce(coal);
-
-       // construct spillslots
-       for(i = 0; i < spillcount; ++i) {
-               spill_t *spill = spilllist[i];
-               spill->spillslot = be_java_coal_get_color(coal, i);
-       }
-       be_java_coal_destroy(coal);
-}
-#endif
-
-/*
- *     _            _               _____       _   _ _   _
- *    / \   ___ ___(_) __ _ _ __   | ____|_ __ | |_(_) |_(_) ___  ___
- *   / _ \ / __/ __| |/ _` | '_ \  |  _| | '_ \| __| | __| |/ _ \/ __|
- *  / ___ \\__ \__ \ | (_| | | | | | |___| | | | |_| | |_| |  __/\__ \
- * /_/   \_\___/___/_|\__, |_| |_| |_____|_| |_|\__|_|\__|_|\___||___/
- *                    |___/
- */
 
 typedef struct _spill_slot_t {
        int size;
        int align;
-       entity   *entity;
+       ir_entity *entity;
 } spill_slot_t;
 
 typedef struct _memperm_entry_t {
        ir_node* node;
        int pos;
-       entity *in;
-       entity *out;
+       ir_entity *in;
+       ir_entity *out;
        struct _memperm_entry_t *next;
 } memperm_entry_t;
 
@@ -516,22 +491,26 @@ typedef struct _memperm_t {
        memperm_entry_t *entries;
 } memperm_t;
 
-static int cmp_memperm(const void* d1, const void* d2, size_t size) {
+static int cmp_memperm(const void* d1, const void* d2, size_t size)
+{
        const memperm_t* e1 = d1;
        const memperm_t* e2 = d2;
+       (void) size;
+
        return e1->block != e2->block;
 }
 
-static memperm_t *get_memperm(ss_env_t *env, ir_node *block) {
+static memperm_t *get_memperm(be_fec_env_t *env, ir_node *block)
+{
        memperm_t entry, *res;
        int hash;
 
        entry.block = block;
-       hash = nodeset_hash(block);
+       hash        = hash_irn(block);
 
        res = set_find(env->memperms, &entry, sizeof(entry), hash);
 
-       if(res == NULL) {
+       if (res == NULL) {
                entry.entrycount = 0;
                entry.entries = NULL;
                res = set_insert(env->memperms, &entry, sizeof(entry), hash);
@@ -540,11 +519,16 @@ static memperm_t *get_memperm(ss_env_t *env, ir_node *block) {
        return res;
 }
 
-static entity* create_stack_entity(ss_env_t *env, spill_slot_t *slot) {
-       ir_type* frame = get_irg_frame_type(env->chordal_env->irg);
-       entity* res = frame_alloc_area(frame, slot->size, slot->align, 0);
-
-       // adjust size of the entity type...
+static ir_entity* create_stack_entity(be_fec_env_t *env, spill_slot_t *slot)
+{
+       ir_graph *irg  = env->irg;
+       ir_type *frame = get_irg_frame_type(irg);
+       /* TODO: backend should be able to specify wether we want spill slots
+        * at begin or end of frame */
+       int        at_start = 1;
+       ir_entity *res = frame_alloc_area(frame, slot->size, slot->align, at_start);
+
+       /* adjust size of the entity type... */
        ir_type *enttype = get_entity_type(res);
        set_type_size_bytes(enttype, slot->size);
 
@@ -553,60 +537,72 @@ static entity* create_stack_entity(ss_env_t *env, spill_slot_t *slot) {
        return res;
 }
 
-static int get_spillslotsize_for_spill(ss_env_t *env, spill_t *spill) {
-       const ir_mode *mode = arch_register_class_mode(spill->cls);
-
-       return get_mode_size_bytes(mode);
-}
-
-static int get_spillslotalign_for_spill(ss_env_t *env, spill_t *spill) {
-       const arch_isa_t *isa = env->chordal_env->birg->main_env->arch_env->isa;
-
-       return arch_isa_get_reg_class_alignment(isa, spill->cls);
-}
-
 /**
  * Enlarges a spillslot (if necessary) so that it can carry a value of size
  * @p othersize and alignment @p otheralign.
  */
-static void enlarge_spillslot(spill_slot_t *slot, int otheralign, int othersize) {
-       if(othersize > slot->size) {
+static void enlarge_spillslot(spill_slot_t *slot, int otheralign, int othersize)
+{
+       if (othersize > slot->size) {
                slot->size = othersize;
        }
-       if(otheralign > slot->align) {
-               if(otheralign % slot->align != 0)
+       if (otheralign > slot->align) {
+               if (otheralign % slot->align != 0)
                        slot->align *= otheralign;
                else
                        slot->align = otheralign;
-       } else if(slot->align % otheralign != 0) {
+       } else if (slot->align % otheralign != 0) {
                slot->align *= otheralign;
        }
 }
 
+
+static void assign_spill_entity(ir_node *node, ir_entity *entity)
+{
+       if (is_NoMem(node))
+               return;
+       if (is_Sync(node)) {
+               int i, arity;
+
+               arity = get_irn_arity(node);
+               for (i = 0; i < arity; ++i) {
+                       ir_node *in = get_irn_n(node, i);
+                       assert(!is_Phi(in));
+
+                       assign_spill_entity(in, entity);
+               }
+               return;
+       }
+
+       /* beware: we might have Stores with Memory Proj's, ia32 fisttp for
+          instance */
+       node = skip_Proj(node);
+       assert(arch_get_frame_entity(node) == NULL);
+       arch_set_frame_entity(node, entity);
+}
+
 /**
  * Create stack entities for the spillslots and assign them to the spill and
  * reload nodes.
  */
-static void assign_spillslots(ss_env_t *env) {
-       const arch_env_t *arch_env = env->arch_env;
-       int i;
-       int spillcount;
-       spill_t *spill;
-       spill_slot_t* spillslots;
-
-       spillcount = set_count(env->spills);
-       spillslots = alloca(spillcount * sizeof(spillslots[0]));
+static void assign_spillslots(be_fec_env_t *env)
+{
+       int           spillcount = set_count(env->spills);
+       spill_slot_t *spillslots = ALLOCANZ(spill_slot_t, spillcount);
+       spill_t      *spill;
+       int           i;
 
-       memset(spillslots, 0, spillcount * sizeof(spillslots[0]));
+       /* construct spillslots */
+       for (spill = set_first(env->spills); spill != NULL;
+               spill = set_next(env->spills)) {
 
-       // construct spillslots
-       for(spill = set_first(env->spills); spill != NULL; spill = set_next(env->spills)) {
                int slotid = spill->spillslot;
+               const ir_mode *mode = spill->mode;
                spill_slot_t *slot = & (spillslots[slotid]);
-               int align = get_spillslotalign_for_spill(env, spill);
-               int size = get_spillslotsize_for_spill(env, spill);
+               int size = get_mode_size_bytes(mode);
+               int align = spill->alignment;
 
-               if(slot->align == 0 && slot->size == 0) {
+               if (slot->align == 0 && slot->size == 0) {
                        slot->align = align;
                        slot->size = size;
                } else {
@@ -614,24 +610,26 @@ static void assign_spillslots(ss_env_t *env) {
                }
        }
 
-       for(spill = set_first(env->spills); spill != NULL; spill = set_next(env->spills)) {
+       for (spill = set_first(env->spills); spill != NULL;
+           spill = set_next(env->spills)) {
+
+               ir_node      *node   = spill->spill;
+               int           slotid = spill->spillslot;
                spill_slot_t *slot;
-               ir_node *node = spill->spill;
-               int slotid = spill->spillslot;
 
                slot = &spillslots[slotid];
-               if(slot->entity == NULL) {
+               if (slot->entity == NULL) {
                        create_stack_entity(env, slot);
                }
 
-               if(is_Phi(node)) {
+               if (is_Phi(node)) {
                        int i, arity;
                        ir_node *block = get_nodes_block(node);
 
-                       // should be a PhiM
+                       /* should be a PhiM */
                        assert(is_Phi(node));
 
-                       for(i = 0, arity = get_irn_arity(node); i < arity; ++i) {
+                       for (i = 0, arity = get_irn_arity(node); i < arity; ++i) {
                                ir_node *arg = get_irn_n(node, i);
                                ir_node *predblock = get_Block_cfgpred_block(block, i);
                                spill_t *argspill;
@@ -641,17 +639,17 @@ static void assign_spillslots(ss_env_t *env) {
                                assert(argspill != NULL);
 
                                argslotid = argspill->spillslot;
-                               if(slotid != argslotid) {
+                               if (slotid != argslotid) {
                                        memperm_t *memperm;
                                        memperm_entry_t *entry;
                                        spill_slot_t *argslot = &spillslots[argslotid];
-                                       if(argslot->entity == NULL) {
+                                       if (argslot->entity == NULL) {
                                                create_stack_entity(env, argslot);
                                        }
 
                                        memperm = get_memperm(env, predblock);
 
-                                       entry = obstack_alloc(&env->obst, sizeof(entry[0]));
+                                       entry = OALLOC(&env->obst, memperm_entry_t);
                                        entry->node = node;
                                        entry->pos = i;
                                        entry->in = argslot->entity;
@@ -662,20 +660,19 @@ static void assign_spillslots(ss_env_t *env) {
                                }
                        }
                } else {
-                       assert(arch_irn_class_is(arch_env, node, spill));
-                       arch_set_frame_entity(arch_env, node, slot->entity);
+                       assign_spill_entity(node, slot->entity);
                }
        }
 
-       for(i = 0; i < ARR_LEN(env->reloads); ++i) {
-               ir_node* reload = env->reloads[i];
-               ir_nodespillnode = get_memory_edge(reload);
-               spill_t *spill = get_spill(env, spillnode);
-               const spill_slot_t *slot = & spillslots[spill->spillslot];
+       for (i = 0; i < ARR_LEN(env->reloads); ++i) {
+               ir_node            *reload    = env->reloads[i];
+               ir_node            *spillnode = get_memory_edge(reload);
+               spill_t            *spill     = get_spill(env, spillnode);
+               const spill_slot_t *slot      = & spillslots[spill->spillslot];
 
                assert(slot->entity != NULL);
 
-               arch_set_frame_entity(arch_env, reload, slot->entity);
+               arch_set_frame_entity(reload, slot->entity);
        }
 }
 
@@ -685,15 +682,15 @@ static void assign_spillslots(ss_env_t *env) {
 static ir_node *get_end_of_block_insertion_point(ir_node* block)
 {
        ir_node* ins = sched_last(block);
-       while(is_Proj(ins) && get_irn_mode(ins) == mode_X) {
+       while (is_Proj(ins) && get_irn_mode(ins) == mode_X) {
                ins = sched_prev(ins);
                assert(ins != NULL);
        }
 
-       if(is_cfop(ins)) {
-               while(1) {
+       if (is_cfop(ins)) {
+               for (;;) {
                        ir_node *prev = sched_prev(ins);
-                       if(!is_cfop(prev))
+                       if (!is_cfop(prev))
                                break;
                        ins = prev;
                }
@@ -702,57 +699,61 @@ static ir_node *get_end_of_block_insertion_point(ir_node* block)
        return ins;
 }
 
-static void create_memperms(ss_env_t *env) {
-       memperm_t *memperm;
+static void create_memperms(be_fec_env_t *env)
+{
+       const arch_env_t *arch_env = env->arch_env;
+       ir_graph         *irg      = env->irg;
+       memperm_t        *memperm;
 
-       for(memperm = set_first(env->memperms); memperm != NULL; memperm = set_next(env->memperms)) {
-               int i;
-               memperm_entry_t *entry;
-               ir_node *blockend;
-               ir_node** nodes = alloca(memperm->entrycount * sizeof(nodes[0]));
-               ir_node* mempermnode;
+       for (memperm = set_first(env->memperms); memperm != NULL; memperm = set_next(env->memperms)) {
+               ir_node         **nodes = ALLOCAN(ir_node*, memperm->entrycount);
+               memperm_entry_t  *entry;
+               ir_node          *blockend;
+               ir_node          *mempermnode;
+               int               i;
 
                assert(memperm->entrycount > 0);
 
-               for(entry = memperm->entries, i = 0; entry != NULL; entry = entry->next, ++i) {
+               for (entry = memperm->entries, i = 0; entry != NULL; entry = entry->next, ++i) {
                        ir_node* arg = get_irn_n(entry->node, entry->pos);
                        nodes[i] = arg;
                }
 
-               mempermnode = be_new_MemPerm(env->chordal_env->birg->main_env->arch_env, env->chordal_env->irg, memperm->block,
-                       memperm->entrycount, nodes);
+               mempermnode = be_new_MemPerm(arch_env, memperm->block,
+                                            memperm->entrycount, nodes);
 
-               // insert node into schedule
+               /* insert node into schedule */
                blockend = get_end_of_block_insertion_point(memperm->block);
                sched_add_before(blockend, mempermnode);
-               be_stat_ev("mem_perm", memperm->entrycount);
+               stat_ev_dbl("mem_perm", memperm->entrycount);
 
-               for(entry = memperm->entries, i = 0; entry != NULL; entry = entry->next, ++i) {
+               i = 0;
+               for (entry = memperm->entries; entry != NULL; entry = entry->next, ++i) {
                        ir_node *proj;
                        ir_node* arg = get_irn_n(entry->node, entry->pos);
 
                        be_set_MemPerm_in_entity(mempermnode, i, entry->in);
                        be_set_MemPerm_out_entity(mempermnode, i, entry->out);
-                       set_irg_current_block(env->chordal_env->irg, memperm->block);
+                       set_irg_current_block(irg, memperm->block);
                        proj = new_Proj(mempermnode, get_irn_mode(arg), i);
-                       sched_add_before(blockend, proj);
 
                        set_irn_n(entry->node, entry->pos, proj);
                }
        }
 }
 
-static int count_spillslots(const ss_env_t *env) {
+static int count_spillslots(const be_fec_env_t *env)
+{
        const spill_t *spill;
        int spillcount = set_count(env->spills);
        bitset_t *counted = bitset_alloca(spillcount);
        int slotcount;
 
        slotcount = 0;
-       for(spill = set_first(env->spills); spill != NULL;
+       for (spill = set_first(env->spills); spill != NULL;
            spill = set_next(env->spills)) {
                int spillslot = spill->spillslot;
-               if(!bitset_is_set(counted, spillslot)) {
+               if (!bitset_is_set(counted, spillslot)) {
                        slotcount++;
                        bitset_set(counted, spillslot);
                }
@@ -761,40 +762,85 @@ static int count_spillslots(const ss_env_t *env) {
        return slotcount;
 }
 
-void be_coalesce_spillslots(const be_chordal_env_t *chordal_env) {
-       ss_env_t env;
+be_fec_env_t *be_new_frame_entity_coalescer(ir_graph *irg)
+{
+       const arch_env_t *arch_env = be_get_irg_arch_env(irg);
+       be_fec_env_t     *env      = XMALLOC(be_fec_env_t);
 
-       obstack_init(&env.obst);
-       env.arch_env = chordal_env->birg->main_env->arch_env;
-       env.chordal_env = chordal_env;
-       env.spills = new_set(cmp_spill, 10);
-       env.reloads = NEW_ARR_F(ir_node*, 0);
-       env.affinity_edges = NEW_ARR_F(affinity_edge_t*, 0);
-       env.memperms = new_set(cmp_memperm, 10);
-       FIRM_DBG_REGISTER(dbg, "firm.be.spillslots");
-       //firm_dbg_set_mask(dbg, DBG_COALESCING);
+       be_liveness_assure_chk(be_assure_liveness(irg));
 
-       /* Get initial spill slots */
-       irg_walk_graph(chordal_env->irg, NULL, collect_spills_walker, &env);
+       obstack_init(&env->obst);
+       env->arch_env       = arch_env;
+       env->irg            = irg;
+       env->spills         = new_set(cmp_spill, 10);
+       env->reloads        = NEW_ARR_F(ir_node*, 0);
+       env->affinity_edges = NEW_ARR_F(affinity_edge_t*, 0);
+       env->memperms       = new_set(cmp_memperm, 10);
 
-       be_stat_ev("spillslots", set_count(env.spills));
+       return env;
+}
 
-       if(be_coalesce_spill_slots) {
-               do_greedy_coalescing(&env);
-               if(be_stat_ev_is_active()) {
-                       be_stat_ev("spillslots_after_coalescing", count_spillslots(&env));
-               }
+void be_free_frame_entity_coalescer(be_fec_env_t *env)
+{
+       del_set(env->memperms);
+       DEL_ARR_F(env->reloads);
+       DEL_ARR_F(env->affinity_edges);
+       del_set(env->spills);
+       obstack_free(&env->obst, NULL);
+
+       free(env);
+}
+
+void be_assign_entities(be_fec_env_t *env)
+{
+       stat_ev_dbl("spillslots", set_count(env->spills));
+
+       if (be_coalesce_spill_slots) {
+               do_greedy_coalescing(env);
        }
 
-       assign_spillslots(&env);
+       stat_ev_dbl("spillslots_after_coalescing", count_spillslots(env));
+
+       assign_spillslots(env);
+
+       create_memperms(env);
+}
 
-       create_memperms(&env);
+/**
+ * This walker function searches for reloads and collects all the spills
+ * and memphis attached to them.
+ */
+static void collect_spills_walker(ir_node *node, void *data)
+{
+       be_fec_env_t *env = data;
+       const ir_mode *mode;
+       const arch_register_class_t *cls;
+       int align;
 
-       //show_stats(&env);
+       if (! (arch_irn_classify(node) & arch_irn_class_reload))
+               return;
 
-       del_set(env.memperms);
-       DEL_ARR_F(env.reloads);
-       DEL_ARR_F(env.affinity_edges);
-       del_set(env.spills);
-       obstack_free(&env.obst, NULL);
+       mode  = get_irn_mode(node);
+       cls   = arch_get_irn_reg_class_out(node);
+       align = arch_env_get_reg_class_alignment(env->arch_env, cls);
+
+       be_node_needs_frame_entity(env, node, mode, align);
+}
+
+void be_coalesce_spillslots(ir_graph *irg)
+{
+       be_fec_env_t *env = be_new_frame_entity_coalescer(irg);
+
+       /* collect reloads */
+       irg_walk_graph(irg, NULL, collect_spills_walker, env);
+
+       be_assign_entities(env);
+
+       be_free_frame_entity_coalescer(env);
+}
+
+BE_REGISTER_MODULE_CONSTRUCTOR(be_init_spillslots);
+void be_init_spillslots(void)
+{
+       FIRM_DBG_REGISTER(dbg, "firm.be.spillslots");
 }