X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?p=pcsx_rearmed.git;a=blobdiff_plain;f=libpcsxcore%2Fnew_dynarec%2Fnew_dynarec.c;h=a64cec6f0a2de5dc0f5b41e85823f4e1f7770daf;hp=3d6a2fef38a7dc70beebb4c60fdbc8fed2ea64a5;hb=1a4301c419a84729d34f8a2f96112097d80f716c;hpb=1edfcc68047e356a9c57c4734cc3bbe084922ce7

diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c
index 3d6a2fef..a64cec6f 100644
--- a/libpcsxcore/new_dynarec/new_dynarec.c
+++ b/libpcsxcore/new_dynarec/new_dynarec.c
@@ -23,9 +23,25 @@
 #include <assert.h>
 #include <errno.h>
 #include <sys/mman.h>
+#ifdef __MACH__
+#include <libkern/OSCacheControl.h>
+#endif
+#ifdef _3DS
+#include <3ds_utils.h>
+#endif
+#ifdef VITA
+#include <psp2/kernel/sysmem.h>
+static int sceBlock;
+#endif
 
+#include "new_dynarec_config.h"
+#include "../psxhle.h" //emulator interface
 #include "emu_if.h" //emulator interface
 
+#ifndef ARRAY_SIZE
+#define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
+#endif
+
 //#define DISASM
 //#define assem_debug printf
 //#define inv_debug printf
@@ -42,22 +58,27 @@
 #include "assem_arm.h"
 #endif
 
-#ifdef __BLACKBERRY_QNX__
-#undef __clear_cache
-#define __clear_cache(start,end) msync(start, (size_t)((void*)end - (void*)start), MS_SYNC | MS_CACHE_ONLY | MS_INVALIDATE_ICACHE);
-#elif defined(__MACH__)
-#include <libkern/OSCacheControl.h>
-#define __clear_cache mach_clear_cache
-static void __clear_cache(void *start, void *end) {
-  size_t len = (char *)end - (char *)start;
-  sys_dcache_flush(start, len);
-  sys_icache_invalidate(start, len);
-}
-#endif
-
 #define MAXBLOCK 4096
 #define MAX_OUTPUT_BLOCK_SIZE 262144
 
+// stubs
+enum stub_type {
+  CC_STUB = 1,
+  FP_STUB = 2,
+  LOADB_STUB = 3,
+  LOADH_STUB = 4,
+  LOADW_STUB = 5,
+  LOADD_STUB = 6,
+  LOADBU_STUB = 7,
+  LOADHU_STUB = 8,
+  STOREB_STUB = 9,
+  STOREH_STUB = 10,
+  STOREW_STUB = 11,
+  STORED_STUB = 12,
+  STORELR_STUB = 13,
+  INVCODE_STUB = 14,
+};
+
 struct regstat
 {
   signed char regmap_entry[HOST_REGS];
@@ -83,22 +104,47 @@ struct ll_entry
   struct ll_entry *next;
 };
 
-  u_int start;
-  u_int *source;
-  char insn[MAXBLOCK][10];
-  u_char itype[MAXBLOCK];
-  u_char opcode[MAXBLOCK];
-  u_char opcode2[MAXBLOCK];
-  u_char bt[MAXBLOCK];
-  u_char rs1[MAXBLOCK];
-  u_char rs2[MAXBLOCK];
-  u_char rt1[MAXBLOCK];
-  u_char rt2[MAXBLOCK];
-  u_char us1[MAXBLOCK];
-  u_char us2[MAXBLOCK];
-  u_char dep1[MAXBLOCK];
-  u_char dep2[MAXBLOCK];
-  u_char lt1[MAXBLOCK];
+struct ht_entry
+{
+  u_int vaddr[2];
+  void *tcaddr[2];
+};
+
+struct code_stub
+{
+  enum stub_type type;
+  void *addr;
+  void *retaddr;
+  u_int a;
+  uintptr_t b;
+  uintptr_t c;
+  u_int d;
+  u_int e;
+};
+
+  // used by asm:
+  u_char *out;
+  struct ht_entry hash_table[65536]  __attribute__((aligned(16)));
+  struct ll_entry *jump_in[4096] __attribute__((aligned(16)));
+  struct ll_entry *jump_dirty[4096];
+
+  static struct ll_entry *jump_out[4096];
+  static u_int start;
+  static u_int *source;
+  static char insn[MAXBLOCK][10];
+  static u_char itype[MAXBLOCK];
+  static u_char opcode[MAXBLOCK];
+  static u_char opcode2[MAXBLOCK];
+  static u_char bt[MAXBLOCK];
+  static u_char rs1[MAXBLOCK];
+  static u_char rs2[MAXBLOCK];
+  static u_char rt1[MAXBLOCK];
+  static u_char rt2[MAXBLOCK];
+  static u_char us1[MAXBLOCK];
+  static u_char us2[MAXBLOCK];
+  static u_char dep1[MAXBLOCK];
+  static u_char dep2[MAXBLOCK];
+  static u_char lt1[MAXBLOCK];
   static uint64_t gte_rs[MAXBLOCK]; // gte: 32 data and 32 ctl regs
   static uint64_t gte_rt[MAXBLOCK];
   static uint64_t gte_unneeded[MAXBLOCK];
@@ -107,52 +153,47 @@ struct ll_entry
   static u_int smrv_weak; // same, but somewhat less likely
   static u_int smrv_strong_next; // same, but after current insn executes
   static u_int smrv_weak_next;
-  int imm[MAXBLOCK];
-  u_int ba[MAXBLOCK];
-  char likely[MAXBLOCK];
-  char is_ds[MAXBLOCK];
-  char ooo[MAXBLOCK];
-  uint64_t unneeded_reg[MAXBLOCK];
-  uint64_t unneeded_reg_upper[MAXBLOCK];
-  uint64_t branch_unneeded_reg[MAXBLOCK];
-  uint64_t branch_unneeded_reg_upper[MAXBLOCK];
-  uint64_t pr32[MAXBLOCK];
-  signed char regmap_pre[MAXBLOCK][HOST_REGS];
+  static int imm[MAXBLOCK];
+  static u_int ba[MAXBLOCK];
+  static char likely[MAXBLOCK];
+  static char is_ds[MAXBLOCK];
+  static char ooo[MAXBLOCK];
+  static uint64_t unneeded_reg[MAXBLOCK];
+  static uint64_t unneeded_reg_upper[MAXBLOCK];
+  static uint64_t branch_unneeded_reg[MAXBLOCK];
+  static uint64_t branch_unneeded_reg_upper[MAXBLOCK];
+  static signed char regmap_pre[MAXBLOCK][HOST_REGS];
   static uint64_t current_constmap[HOST_REGS];
   static uint64_t constmap[MAXBLOCK][HOST_REGS];
   static struct regstat regs[MAXBLOCK];
   static struct regstat branch_regs[MAXBLOCK];
-  signed char minimum_free_regs[MAXBLOCK];
-  u_int needed_reg[MAXBLOCK];
-  u_int wont_dirty[MAXBLOCK];
-  u_int will_dirty[MAXBLOCK];
-  int ccadj[MAXBLOCK];
-  int slen;
-  u_int instr_addr[MAXBLOCK];
-  u_int link_addr[MAXBLOCK][3];
-  int linkcount;
-  u_int stubs[MAXBLOCK*3][8];
-  int stubcount;
-  u_int literals[1024][2];
-  int literalcount;
-  int is_delayslot;
-  int cop1_usable;
-  u_char *out;
-  struct ll_entry *jump_in[4096] __attribute__((aligned(16)));
-  struct ll_entry *jump_out[4096];
-  struct ll_entry *jump_dirty[4096];
-  u_int hash_table[65536][4]  __attribute__((aligned(16)));
-  char shadow[1048576]  __attribute__((aligned(16)));
-  void *copy;
-  int expirep;
-  int new_dynarec_did_compile;
-  int new_dynarec_hacks;
-  u_int stop_after_jal;
+  static signed char minimum_free_regs[MAXBLOCK];
+  static u_int needed_reg[MAXBLOCK];
+  static u_int wont_dirty[MAXBLOCK];
+  static u_int will_dirty[MAXBLOCK];
+  static int ccadj[MAXBLOCK];
+  static int slen;
+  static void *instr_addr[MAXBLOCK];
+  static u_int link_addr[MAXBLOCK][3];
+  static int linkcount;
+  static struct code_stub stubs[MAXBLOCK*3];
+  static int stubcount;
+  static u_int literals[1024][2];
+  static int literalcount;
+  static int is_delayslot;
+  static int cop1_usable;
+  static char shadow[1048576]  __attribute__((aligned(16)));
+  static void *copy;
+  static int expirep;
+  static u_int stop_after_jal;
 #ifndef RAM_FIXED
   static u_int ram_offset;
 #else
   static const u_int ram_offset=0;
 #endif
+
+  int new_dynarec_hacks;
+  int new_dynarec_did_compile;
   extern u_char restore_candidate[512];
   extern int cycle_count;
 
@@ -186,7 +227,7 @@ struct ll_entry
 #define STORE 2   // Store
 #define LOADLR 3  // Unaligned load
 #define STORELR 4 // Unaligned store
-#define MOV 5     // Move 
+#define MOV 5     // Move
 #define ALU 6     // Arithmetic/logic
 #define MULTDIV 7 // Multiply/divide
 #define SHIFT 8   // Shift by register
@@ -213,22 +254,6 @@ struct ll_entry
 #define C2OP 29   // Coprocessor 2 operation
 #define INTCALL 30// Call interpreter to handle rare corner cases
 
-  /* stubs */
-#define CC_STUB 1
-#define FP_STUB 2
-#define LOADB_STUB 3
-#define LOADH_STUB 4
-#define LOADW_STUB 5
-#define LOADD_STUB 6
-#define LOADBU_STUB 7
-#define LOADHU_STUB 8
-#define STOREB_STUB 9
-#define STOREH_STUB 10
-#define STOREW_STUB 11
-#define STORED_STUB 12
-#define STORELR_STUB 13
-#define INVCODE_STUB 14
-
   /* branch codes */
 #define TAKEN 1
 #define NOTTAKEN 2
@@ -254,15 +279,83 @@ void jump_intcall();
 void new_dyna_leave();
 
 // Needed by assembler
-void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32);
-void wb_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty);
-void wb_needed_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr);
-void load_all_regs(signed char i_regmap[]);
-void load_needed_regs(signed char i_regmap[],signed char next_regmap[]);
-void load_regs_entry(int t);
-void load_all_consts(signed char regmap[],int is32,u_int dirty,int i);
+static void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32);
+static void wb_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty);
+static void wb_needed_dirtys(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr);
+static void load_all_regs(signed char i_regmap[]);
+static void load_needed_regs(signed char i_regmap[],signed char next_regmap[]);
+static void load_regs_entry(int t);
+static void load_all_consts(signed char regmap[],int is32,u_int dirty,int i);
+
+static int verify_dirty(u_int *ptr);
+static int get_final_value(int hr, int i, int *value);
+static void add_stub(enum stub_type type, void *addr, void *retaddr,
+  u_int a, uintptr_t b, uintptr_t c, u_int d, u_int e);
+static void add_stub_r(enum stub_type type, void *addr, void *retaddr,
+  int i, int addr_reg, struct regstat *i_regs, int ccadj, u_int reglist);
+static void add_to_linker(int addr,int target,int ext);
+
+static int tracedebug=0;
+
+static void mprotect_w_x(void *start, void *end, int is_x)
+{
+#ifdef NO_WRITE_EXEC
+  #if defined(VITA)
+  // *Open* enables write on all memory that was
+  // allocated by sceKernelAllocMemBlockForVM()?
+  if (is_x)
+    sceKernelCloseVMDomain();
+  else
+    sceKernelOpenVMDomain();
+  #else
+  u_long mstart = (u_long)start & ~4095ul;
+  u_long mend = (u_long)end;
+  if (mprotect((void *)mstart, mend - mstart,
+               PROT_READ | (is_x ? PROT_EXEC : PROT_WRITE)) != 0)
+    SysPrintf("mprotect(%c) failed: %s\n", is_x ? 'x' : 'w', strerror(errno));
+  #endif
+#endif
+}
+
+static void start_tcache_write(void *start, void *end)
+{
+  mprotect_w_x(start, end, 0);
+}
 
-int tracedebug=0;
+static void end_tcache_write(void *start, void *end)
+{
+#ifdef __arm__
+  size_t len = (char *)end - (char *)start;
+  #if   defined(__BLACKBERRY_QNX__)
+  msync(start, len, MS_SYNC | MS_CACHE_ONLY | MS_INVALIDATE_ICACHE);
+  #elif defined(__MACH__)
+  sys_cache_control(kCacheFunctionPrepareForExecution, start, len);
+  #elif defined(VITA)
+  sceKernelSyncVMDomain(sceBlock, start, len);
+  #elif defined(_3DS)
+  ctr_flush_invalidate_cache();
+  #else
+  __clear_cache(start, end);
+  #endif
+  (void)len;
+#endif
+
+  mprotect_w_x(start, end, 1);
+}
+
+static void *start_block(void)
+{
+  u_char *end = out + MAX_OUTPUT_BLOCK_SIZE;
+  if (end > (u_char *)BASE_ADDR + (1<<TARGET_SIZE_2))
+    end = (u_char *)BASE_ADDR + (1<<TARGET_SIZE_2);
+  start_tcache_write(out, end);
+  return out;
+}
+
+static void end_block(void *start)
+{
+  end_tcache_write(start, out);
+}
 
 //#define DEBUG_CYCLE_COUNT 1
 
@@ -292,6 +385,26 @@ static u_int get_vpage(u_int vaddr)
   return get_page(vaddr);
 }
 
+static struct ht_entry *hash_table_get(u_int vaddr)
+{
+  return &hash_table[((vaddr>>16)^vaddr)&0xFFFF];
+}
+
+static void hash_table_add(struct ht_entry *ht_bin, u_int vaddr, void *tcaddr)
+{
+  ht_bin->vaddr[1] = ht_bin->vaddr[0];
+  ht_bin->tcaddr[1] = ht_bin->tcaddr[0];
+  ht_bin->vaddr[0] = vaddr;
+  ht_bin->tcaddr[0] = tcaddr;
+}
+
+// some messy ari64's code, seems to rely on unsigned 32bit overflow
+static int doesnt_expire_soon(void *tcaddr)
+{
+  u_int diff = (u_int)((u_char *)tcaddr - out) << (32-TARGET_SIZE_2);
+  return diff > (u_int)(0x60000000 + (MAX_OUTPUT_BLOCK_SIZE << (32-TARGET_SIZE_2)));
+}
+
 // Get address from virtual address
 // This is called from the recompiled JR/JALR instructions
 void *get_addr(u_int vaddr)
@@ -304,11 +417,7 @@ void *get_addr(u_int vaddr)
   while(head!=NULL) {
     if(head->vaddr==vaddr) {
   //printf("TRACE: count=%d next=%d (get_addr match %x: %x)\n",Count,next_interupt,vaddr,(int)head->addr);
-      int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF];
-      ht_bin[3]=ht_bin[1];
-      ht_bin[2]=ht_bin[0];
-      ht_bin[1]=(int)head->addr;
-      ht_bin[0]=vaddr;
+      hash_table_add(hash_table_get(vaddr), vaddr, head->addr);
       return head->addr;
     }
     head=head->next;
@@ -318,8 +427,8 @@ void *get_addr(u_int vaddr)
     if(head->vaddr==vaddr) {
       //printf("TRACE: count=%d next=%d (get_addr match dirty %x: %x)\n",Count,next_interupt,vaddr,(int)head->addr);
       // Don't restore blocks which are about to expire from the cache
-      if((((u_int)head->addr-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2)))
-      if(verify_dirty(head->addr)) {
+      if (doesnt_expire_soon(head->addr))
+      if (verify_dirty(head->addr)) {
         //printf("restore candidate: %x (%d) d=%d\n",vaddr,page,invalid_code[vaddr>>12]);
         invalid_code[vaddr>>12]=0;
         inv_code_start=inv_code_end=~0;
@@ -327,17 +436,12 @@ void *get_addr(u_int vaddr)
           restore_candidate[vpage>>3]|=1<<(vpage&7);
         }
         else restore_candidate[page>>3]|=1<<(page&7);
-        int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF];
-        if(ht_bin[0]==vaddr) {
-          ht_bin[1]=(int)head->addr; // Replace existing entry
-        }
+        struct ht_entry *ht_bin = hash_table_get(vaddr);
+        if (ht_bin->vaddr[0] == vaddr)
+          ht_bin->tcaddr[0] = head->addr; // Replace existing entry
         else
-        {
-          ht_bin[3]=ht_bin[1];
-          ht_bin[2]=ht_bin[0];
-          ht_bin[1]=(int)head->addr;
-          ht_bin[0]=vaddr;
-        }
+          hash_table_add(ht_bin, vaddr, head->addr);
+
         return head->addr;
       }
     }
@@ -359,9 +463,9 @@ void *get_addr(u_int vaddr)
 void *get_addr_ht(u_int vaddr)
 {
   //printf("TRACE: count=%d next=%d (get_addr_ht %x)\n",Count,next_interupt,vaddr);
-  int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF];
-  if(ht_bin[0]==vaddr) return (void *)ht_bin[1];
-  if(ht_bin[2]==vaddr) return (void *)ht_bin[3];
+  const struct ht_entry *ht_bin = hash_table_get(vaddr);
+  if (ht_bin->vaddr[0] == vaddr) return ht_bin->tcaddr[0];
+  if (ht_bin->vaddr[1] == vaddr) return ht_bin->tcaddr[1];
   return get_addr(vaddr);
 }
 
@@ -421,7 +525,7 @@ static void flush_dirty_uppers(struct regstat *cur)
   for (hr=0;hr<HOST_REGS;hr++) {
     if((cur->dirty>>hr)&1) {
       reg=cur->regmap[hr];
-      if(reg>=64) 
+      if(reg>=64)
         if((cur->is32>>(reg&63))&1) cur->regmap[hr]=-1;
     }
   }
@@ -574,7 +678,7 @@ int needed_again(int r, int i)
   int j;
   int b=-1;
   int rn=10;
-  
+
   if(i>0&&(itype[i-1]==UJUMP||itype[i-1]==RJUMP||(source[i-1]>>16)==0x1000))
   {
     if(ba[i-1]<start || ba[i-1]>start+slen*4-4)
@@ -627,6 +731,7 @@ int needed_again(int r, int i)
     }
   }*/
   if(rn<10) return 1;
+  (void)b;
   return 0;
 }
 
@@ -677,7 +782,7 @@ int loop_reg(int i, int r, int hr)
 void alloc_all(struct regstat *cur,int i)
 {
   int hr;
-  
+
   for(hr=0;hr<HOST_REGS;hr++) {
     if(hr!=EXCLUDE_REG) {
       if(((cur->regmap[hr]&63)!=rs1[i])&&((cur->regmap[hr]&63)!=rs2[i])&&
@@ -729,39 +834,39 @@ void ll_add_flags(struct ll_entry **head,int vaddr,u_int reg_sv_flags,void *addr
 // but don't return addresses which are about to expire from the cache
 void *check_addr(u_int vaddr)
 {
-  u_int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF];
-  if(ht_bin[0]==vaddr) {
-    if(((ht_bin[1]-MAX_OUTPUT_BLOCK_SIZE-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2)))
-      if(isclean(ht_bin[1])) return (void *)ht_bin[1];
-  }
-  if(ht_bin[2]==vaddr) {
-    if(((ht_bin[3]-MAX_OUTPUT_BLOCK_SIZE-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2)))
-      if(isclean(ht_bin[3])) return (void *)ht_bin[3];
+  struct ht_entry *ht_bin = hash_table_get(vaddr);
+  size_t i;
+  for (i = 0; i < ARRAY_SIZE(ht_bin->vaddr); i++) {
+    if (ht_bin->vaddr[i] == vaddr)
+      if (doesnt_expire_soon((u_char *)ht_bin->tcaddr[i] - MAX_OUTPUT_BLOCK_SIZE))
+        if (isclean(ht_bin->tcaddr[i]))
+          return ht_bin->tcaddr[i];
   }
   u_int page=get_page(vaddr);
   struct ll_entry *head;
   head=jump_in[page];
-  while(head!=NULL) {
-    if(head->vaddr==vaddr) {
-      if((((u_int)head->addr-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2))) {
+  while (head != NULL) {
+    if (head->vaddr == vaddr) {
+      if (doesnt_expire_soon(head->addr)) {
         // Update existing entry with current address
-        if(ht_bin[0]==vaddr) {
-          ht_bin[1]=(int)head->addr;
+        if (ht_bin->vaddr[0] == vaddr) {
+          ht_bin->tcaddr[0] = head->addr;
           return head->addr;
         }
-        if(ht_bin[2]==vaddr) {
-          ht_bin[3]=(int)head->addr;
+        if (ht_bin->vaddr[1] == vaddr) {
+          ht_bin->tcaddr[1] = head->addr;
           return head->addr;
         }
         // Insert into hash table with low priority.
         // Don't evict existing entries, as they are probably
         // addresses that are being accessed frequently.
-        if(ht_bin[0]==-1) {
-          ht_bin[1]=(int)head->addr;
-          ht_bin[0]=vaddr;
-        }else if(ht_bin[2]==-1) {
-          ht_bin[3]=(int)head->addr;
-          ht_bin[2]=vaddr;
+        if (ht_bin->vaddr[0] == -1) {
+          ht_bin->vaddr[0] = vaddr;
+          ht_bin->tcaddr[0] = head->addr;
+        }
+        else if (ht_bin->vaddr[1] == -1) {
+          ht_bin->vaddr[1] = vaddr;
+          ht_bin->tcaddr[1] = head->addr;
         }
         return head->addr;
       }
@@ -774,14 +879,16 @@ void *check_addr(u_int vaddr)
 void remove_hash(int vaddr)
 {
   //printf("remove hash: %x\n",vaddr);
-  int *ht_bin=hash_table[(((vaddr)>>16)^vaddr)&0xFFFF];
-  if(ht_bin[2]==vaddr) {
-    ht_bin[2]=ht_bin[3]=-1;
+  struct ht_entry *ht_bin = hash_table_get(vaddr);
+  if (ht_bin->vaddr[1] == vaddr) {
+    ht_bin->vaddr[1] = -1;
+    ht_bin->tcaddr[1] = NULL;
   }
-  if(ht_bin[0]==vaddr) {
-    ht_bin[0]=ht_bin[2];
-    ht_bin[1]=ht_bin[3];
-    ht_bin[2]=ht_bin[3]=-1;
+  if (ht_bin->vaddr[0] == vaddr) {
+    ht_bin->vaddr[0] = ht_bin->vaddr[1];
+    ht_bin->tcaddr[0] = ht_bin->tcaddr[1];
+    ht_bin->vaddr[1] = -1;
+    ht_bin->tcaddr[1] = NULL;
   }
 }
 
@@ -789,7 +896,7 @@ void ll_remove_matching_addrs(struct ll_entry **head,int addr,int shift)
 {
   struct ll_entry *next;
   while(*head) {
-    if(((u_int)((*head)->addr)>>shift)==(addr>>shift) || 
+    if(((u_int)((*head)->addr)>>shift)==(addr>>shift) ||
        ((u_int)((*head)->addr-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(addr>>shift))
     {
       inv_debug("EXP: Remove pointer to %x (%x)\n",(int)(*head)->addr,(*head)->vaddr);
@@ -810,7 +917,7 @@ void ll_clear(struct ll_entry **head)
 {
   struct ll_entry *cur;
   struct ll_entry *next;
-  if(cur=*head) {
+  if((cur=*head)) {
     *head=0;
     while(cur) {
       next=cur->next;
@@ -821,7 +928,7 @@ void ll_clear(struct ll_entry **head)
 }
 
 // Dereference the pointers and remove if it matches
-void ll_kill_pointers(struct ll_entry *head,int addr,int shift)
+static void ll_kill_pointers(struct ll_entry *head,int addr,int shift)
 {
   while(head) {
     int ptr=get_pointer(head->addr);
@@ -830,10 +937,11 @@ void ll_kill_pointers(struct ll_entry *head,int addr,int shift)
        (((ptr-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(addr>>shift)))
     {
       inv_debug("EXP: Kill pointer at %x (%x)\n",(int)head->addr,head->vaddr);
-      u_int host_addr=(u_int)kill_pointer(head->addr);
+      void *host_addr=find_extjump_insn(head->addr);
       #ifdef __arm__
-        needs_clear_cache[(host_addr-(u_int)BASE_ADDR)>>17]|=1<<(((host_addr-(u_int)BASE_ADDR)>>12)&31);
+        mark_clear_cache(host_addr);
       #endif
+      set_jump_target(host_addr, head->addr);
     }
     head=head->next;
   }
@@ -857,10 +965,11 @@ void invalidate_page(u_int page)
   jump_out[page]=0;
   while(head!=NULL) {
     inv_debug("INVALIDATE: kill pointer to %x (%x)\n",head->vaddr,(int)head->addr);
-    u_int host_addr=(u_int)kill_pointer(head->addr);
+    void *host_addr=find_extjump_insn(head->addr);
     #ifdef __arm__
-      needs_clear_cache[(host_addr-(u_int)BASE_ADDR)>>17]|=1<<(((host_addr-(u_int)BASE_ADDR)>>12)&31);
+      mark_clear_cache(host_addr);
     #endif
+    set_jump_target(host_addr, head->addr);
     next=head->next;
     free(head);
     head=next;
@@ -885,7 +994,7 @@ static void invalidate_block_range(u_int block, u_int first, u_int last)
   #ifdef __arm__
     do_clear_cache();
   #endif
-  
+
   // Don't trap writes
   invalid_code[block]=1;
 
@@ -984,7 +1093,7 @@ void invalidate_addr(u_int addr)
 // Anything could have changed, so invalidate everything.
 void invalidate_all_pages()
 {
-  u_int page,n;
+  u_int page;
   for(page=0;page<4096;page++)
     invalidate_page(page);
   for(page=0;page<1048576;page++)
@@ -992,9 +1101,6 @@ void invalidate_all_pages()
       restore_candidate[(page&2047)>>3]|=1<<(page&7);
       restore_candidate[((page&2047)>>3)+256]|=1<<(page&7);
     }
-  #ifdef __arm__
-  __clear_cache((void *)BASE_ADDR,(void *)BASE_ADDR+(1<<TARGET_SIZE_2));
-  #endif
   #ifdef USE_MINI_HT
   memset(mini_ht,-1,sizeof(mini_ht));
   #endif
@@ -1007,6 +1113,7 @@ void add_link(u_int vaddr,void *src)
   inv_debug("add_link: %x -> %x (%d)\n",(int)src,vaddr,page);
   int *ptr=(int *)(src+4);
   assert((*ptr&0x0fff0000)==0x059f0000);
+  (void)ptr;
   ll_add(jump_out+page,vaddr,src);
   //int ptr=get_pointer(src);
   //inv_debug("add_link: Pointer is to %x\n",(int)ptr);
@@ -1024,9 +1131,9 @@ void clean_blocks(u_int page)
   while(head!=NULL) {
     if(!invalid_code[head->vaddr>>12]) {
       // Don't restore blocks which are about to expire from the cache
-      if((((u_int)head->addr-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2))) {
+      if (doesnt_expire_soon(head->addr)) {
         u_int start,end;
-        if(verify_dirty((int)head->addr)) {
+        if(verify_dirty(head->addr)) {
           //printf("Possibly Restore %x (%x)\n",head->vaddr, (int)head->addr);
           u_int i;
           u_int inv=0;
@@ -1040,20 +1147,18 @@ void clean_blocks(u_int page)
             inv=1;
           }
           if(!inv) {
-            void * clean_addr=(void *)get_clean_addr((int)head->addr);
-            if((((u_int)clean_addr-(u_int)out)<<(32-TARGET_SIZE_2))>0x60000000+(MAX_OUTPUT_BLOCK_SIZE<<(32-TARGET_SIZE_2))) {
+            void *clean_addr = get_clean_addr(head->addr);
+            if (doesnt_expire_soon(clean_addr)) {
               u_int ppage=page;
               inv_debug("INV: Restored %x (%x/%x)\n",head->vaddr, (int)head->addr, (int)clean_addr);
               //printf("page=%x, addr=%x\n",page,head->vaddr);
               //assert(head->vaddr>>12==(page|0x80000));
               ll_add_flags(jump_in+ppage,head->vaddr,head->reg_sv_flags,clean_addr);
-              int *ht_bin=hash_table[((head->vaddr>>16)^head->vaddr)&0xFFFF];
-              if(ht_bin[0]==head->vaddr) {
-                ht_bin[1]=(int)clean_addr; // Replace existing entry
-              }
-              if(ht_bin[2]==head->vaddr) {
-                ht_bin[3]=(int)clean_addr; // Replace existing entry
-              }
+              struct ht_entry *ht_bin = hash_table_get(head->vaddr);
+              if (ht_bin->vaddr[0] == head->vaddr)
+                ht_bin->tcaddr[0] = clean_addr; // Replace existing entry
+              if (ht_bin->vaddr[1] == head->vaddr)
+                ht_bin->tcaddr[1] = clean_addr; // Replace existing entry
             }
           }
         }
@@ -1747,19 +1852,27 @@ static void pagespan_alloc(struct regstat *current,int i)
   //else ...
 }
 
-add_stub(int type,int addr,int retaddr,int a,int b,int c,int d,int e)
+static void add_stub(enum stub_type type, void *addr, void *retaddr,
+  u_int a, uintptr_t b, uintptr_t c, u_int d, u_int e)
 {
-  stubs[stubcount][0]=type;
-  stubs[stubcount][1]=addr;
-  stubs[stubcount][2]=retaddr;
-  stubs[stubcount][3]=a;
-  stubs[stubcount][4]=b;
-  stubs[stubcount][5]=c;
-  stubs[stubcount][6]=d;
-  stubs[stubcount][7]=e;
+  assert(a < ARRAY_SIZE(stubs));
+  stubs[stubcount].type = type;
+  stubs[stubcount].addr = addr;
+  stubs[stubcount].retaddr = retaddr;
+  stubs[stubcount].a = a;
+  stubs[stubcount].b = b;
+  stubs[stubcount].c = c;
+  stubs[stubcount].d = d;
+  stubs[stubcount].e = e;
   stubcount++;
 }
 
+static void add_stub_r(enum stub_type type, void *addr, void *retaddr,
+  int i, int addr_reg, struct regstat *i_regs, int ccadj, u_int reglist)
+{
+  add_stub(type, addr, retaddr, i, addr_reg, (uintptr_t)i_regs, ccadj, reglist);
+}
+
 // Write out a single register
 void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32)
 {
@@ -1781,7 +1894,6 @@ void wb_register(signed char r,signed char regmap[],uint64_t dirty,uint64_t is32
 
 int mchecksum()
 {
-  //if(!tracedebug) return 0;
   int i;
   int sum=0;
   for(i=0;i<2097152;i++) {
@@ -1809,37 +1921,6 @@ void rlist()
   printf("\n");
 }
 
-void enabletrace()
-{
-  tracedebug=1;
-}
-
-void memdebug(int i)
-{
-  //printf("TRACE: count=%d next=%d (checksum %x) lo=%8x%8x\n",Count,next_interupt,mchecksum(),(int)(reg[LOREG]>>32),(int)reg[LOREG]);
-  //printf("TRACE: count=%d next=%d (rchecksum %x)\n",Count,next_interupt,rchecksum());
-  //rlist();
-  //if(tracedebug) {
-  //if(Count>=-2084597794) {
-  if((signed int)Count>=-2084597794&&(signed int)Count<0) {
-  //if(0) {
-    printf("TRACE: count=%d next=%d (checksum %x)\n",Count,next_interupt,mchecksum());
-    //printf("TRACE: count=%d next=%d (checksum %x) Status=%x\n",Count,next_interupt,mchecksum(),Status);
-    //printf("TRACE: count=%d next=%d (checksum %x) hi=%8x%8x\n",Count,next_interupt,mchecksum(),(int)(reg[HIREG]>>32),(int)reg[HIREG]);
-    rlist();
-    #ifdef __i386__
-    printf("TRACE: %x\n",(&i)[-1]);
-    #endif
-    #ifdef __arm__
-    int j;
-    printf("TRACE: %x \n",(&j)[10]);
-    printf("TRACE: %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x %x\n",(&j)[1],(&j)[2],(&j)[3],(&j)[4],(&j)[5],(&j)[6],(&j)[7],(&j)[8],(&j)[9],(&j)[10],(&j)[11],(&j)[12],(&j)[13],(&j)[14],(&j)[15],(&j)[16],(&j)[17],(&j)[18],(&j)[19],(&j)[20]);
-    #endif
-    //fflush(stdout);
-  }
-  //printf("TRACE: %x\n",(&i)[-1]);
-}
-
 void alu_assemble(int i,struct regstat *i_regs)
 {
   if(opcode2[i]>=0x20&&opcode2[i]<=0x23) { // ADD/ADDU/SUB/SUBU
@@ -2326,23 +2407,25 @@ void imm16_assemble(int i,struct regstat *i_regs)
                 emit_mov(sh,th);
               }
             }
-            if(opcode[i]==0x0d) //ORI
-            if(sl<0) {
-              emit_orimm(tl,imm[i],tl);
-            }else{
-              if(!((i_regs->wasconst>>sl)&1))
-                emit_orimm(sl,imm[i],tl);
-              else
-                emit_movimm(constmap[i][sl]|imm[i],tl);
+            if(opcode[i]==0x0d) { // ORI
+              if(sl<0) {
+                emit_orimm(tl,imm[i],tl);
+              }else{
+                if(!((i_regs->wasconst>>sl)&1))
+                  emit_orimm(sl,imm[i],tl);
+                else
+                  emit_movimm(constmap[i][sl]|imm[i],tl);
+              }
             }
-            if(opcode[i]==0x0e) //XORI
-            if(sl<0) {
-              emit_xorimm(tl,imm[i],tl);
-            }else{
-              if(!((i_regs->wasconst>>sl)&1))
-                emit_xorimm(sl,imm[i],tl);
-              else
-                emit_movimm(constmap[i][sl]^imm[i],tl);
+            if(opcode[i]==0x0e) { // XORI
+              if(sl<0) {
+                emit_xorimm(tl,imm[i],tl);
+              }else{
+                if(!((i_regs->wasconst>>sl)&1))
+                  emit_xorimm(sl,imm[i],tl);
+                else
+                  emit_movimm(constmap[i][sl]^imm[i],tl);
+              }
             }
           }
           else {
@@ -2505,7 +2588,7 @@ void load_assemble(int i,struct regstat *i_regs)
 {
   int s,th,tl,addr,map=-1;
   int offset;
-  int jaddr=0;
+  void *jaddr=0;
   int memtarget=0,c=0;
   int fastload_reg_override=0;
   u_int hr,reglist=0;
@@ -2526,7 +2609,7 @@ void load_assemble(int i,struct regstat *i_regs)
   //printf("load_assemble: c=%d\n",c);
   //if(c) printf("load_assemble: const=%x\n",(int)constmap[i][s]+offset);
   // FIXME: Even if the load is a NOP, we should check for pagefaults...
-  if(tl<0&&(!c||(((u_int)constmap[i][s]+offset)>>16)==0x1f80)
+  if((tl<0&&(!c||(((u_int)constmap[i][s]+offset)>>16)==0x1f80))
     ||rt1[i]==0) {
       // could be FIFO, must perform the read
       // ||dummy read
@@ -2585,7 +2668,7 @@ void load_assemble(int i,struct regstat *i_regs)
         }
       }
       if(jaddr)
-        add_stub(LOADB_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADB_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else
       inline_readstub(LOADB_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
@@ -2622,7 +2705,7 @@ void load_assemble(int i,struct regstat *i_regs)
         }
       }
       if(jaddr)
-        add_stub(LOADH_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADH_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else
       inline_readstub(LOADH_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
@@ -2641,7 +2724,7 @@ void load_assemble(int i,struct regstat *i_regs)
         emit_readword_indexed_tlb(0,a,map,tl);
       }
       if(jaddr)
-        add_stub(LOADW_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADW_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else
       inline_readstub(LOADW_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
@@ -2670,7 +2753,7 @@ void load_assemble(int i,struct regstat *i_regs)
         }
       }
       if(jaddr)
-        add_stub(LOADBU_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADBU_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else
       inline_readstub(LOADBU_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
@@ -2707,7 +2790,7 @@ void load_assemble(int i,struct regstat *i_regs)
         }
       }
       if(jaddr)
-        add_stub(LOADHU_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADHU_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else
       inline_readstub(LOADHU_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
@@ -2727,7 +2810,7 @@ void load_assemble(int i,struct regstat *i_regs)
         emit_readword_indexed_tlb(0,a,map,tl);
       }
       if(jaddr)
-        add_stub(LOADW_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADW_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else {
       inline_readstub(LOADW_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
@@ -2749,41 +2832,12 @@ void load_assemble(int i,struct regstat *i_regs)
         emit_readdword_indexed_tlb(0,a,map,th,tl);
       }
       if(jaddr)
-        add_stub(LOADD_STUB,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+        add_stub_r(LOADD_STUB,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     }
     else
       inline_readstub(LOADD_STUB,i,constmap[i][s]+offset,i_regs->regmap,rt1[i],ccadj[i],reglist);
   }
  }
-  //emit_storereg(rt1[i],tl); // DEBUG
-  //if(opcode[i]==0x23)
-  //if(opcode[i]==0x24)
-  //if(opcode[i]==0x23||opcode[i]==0x24)
-  /*if(opcode[i]==0x21||opcode[i]==0x23||opcode[i]==0x24)
-  {
-    //emit_pusha();
-    save_regs(0x100f);
-        emit_readword((int)&last_count,ECX);
-        #ifdef __i386__
-        if(get_reg(i_regs->regmap,CCREG)<0)
-          emit_loadreg(CCREG,HOST_CCREG);
-        emit_add(HOST_CCREG,ECX,HOST_CCREG);
-        emit_addimm(HOST_CCREG,2*ccadj[i],HOST_CCREG);
-        emit_writeword(HOST_CCREG,(int)&Count);
-        #endif
-        #ifdef __arm__
-        if(get_reg(i_regs->regmap,CCREG)<0)
-          emit_loadreg(CCREG,0);
-        else
-          emit_mov(HOST_CCREG,0);
-        emit_add(0,ECX,0);
-        emit_addimm(0,2*ccadj[i],0);
-        emit_writeword(0,(int)&Count);
-        #endif
-    emit_call((int)memdebug);
-    //emit_popa();
-    restore_regs(0x100f);
-  }/**/
 }
 
 #ifndef loadlr_assemble
@@ -2799,7 +2853,8 @@ void store_assemble(int i,struct regstat *i_regs)
   int s,th,tl,map=-1;
   int addr,temp;
   int offset;
-  int jaddr=0,jaddr2,type;
+  void *jaddr=0;
+  enum stub_type type;
   int memtarget=0,c=0;
   int agr=AGEN1+(i&1);
   int faststore_reg_override=0;
@@ -2898,7 +2953,7 @@ void store_assemble(int i,struct regstat *i_regs)
   if(jaddr) {
     // PCSX store handlers don't check invcode again
     reglist|=1<<addr;
-    add_stub(type,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+    add_stub_r(type,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
     jaddr=0;
   }
   if(!(i_regs->waswritten&(1<<rs1[i]))&&!(new_dynarec_hacks&NDHACK_NO_SMC_CHECK)) {
@@ -2918,15 +2973,15 @@ void store_assemble(int i,struct regstat *i_regs)
       #if defined(HAVE_CONDITIONAL_CALL) && !defined(DESTRUCTIVE_SHIFT)
       emit_callne(invalidate_addr_reg[addr]);
       #else
-      jaddr2=(int)out;
+      void *jaddr2 = out;
       emit_jne(0);
-      add_stub(INVCODE_STUB,jaddr2,(int)out,reglist|(1<<HOST_CCREG),addr,0,0,0);
+      add_stub(INVCODE_STUB,jaddr2,out,reglist|(1<<HOST_CCREG),addr,0,0,0);
       #endif
     }
   }
   u_int addr_val=constmap[i][s]+offset;
   if(jaddr) {
-    add_stub(type,jaddr,(int)out,i,addr,(int)i_regs,ccadj[i],reglist);
+    add_stub_r(type,jaddr,out,i,addr,i_regs,ccadj[i],reglist);
   } else if(c&&!memtarget) {
     inline_writestub(type,i,addr_val,i_regs->regmap,rs2[i],ccadj[i],reglist);
   }
@@ -2940,57 +2995,20 @@ void store_assemble(int i,struct regstat *i_regs)
       wb_dirtys(regs[i].regmap_entry,regs[i].was32,regs[i].wasdirty);
       emit_movimm(start+i*4+4,0);
       emit_writeword(0,(int)&pcaddr);
-      emit_jmp((int)do_interrupt);
+      emit_jmp(do_interrupt);
     }
   }
-  //if(opcode[i]==0x2B || opcode[i]==0x3F)
-  //if(opcode[i]==0x2B || opcode[i]==0x28)
-  //if(opcode[i]==0x2B || opcode[i]==0x29)
-  //if(opcode[i]==0x2B)
-  /*if(opcode[i]==0x2B || opcode[i]==0x28 || opcode[i]==0x29 || opcode[i]==0x3F)
-  {
-    #ifdef __i386__
-    emit_pusha();
-    #endif
-    #ifdef __arm__
-    save_regs(0x100f);
-    #endif
-        emit_readword((int)&last_count,ECX);
-        #ifdef __i386__
-        if(get_reg(i_regs->regmap,CCREG)<0)
-          emit_loadreg(CCREG,HOST_CCREG);
-        emit_add(HOST_CCREG,ECX,HOST_CCREG);
-        emit_addimm(HOST_CCREG,2*ccadj[i],HOST_CCREG);
-        emit_writeword(HOST_CCREG,(int)&Count);
-        #endif
-        #ifdef __arm__
-        if(get_reg(i_regs->regmap,CCREG)<0)
-          emit_loadreg(CCREG,0);
-        else
-          emit_mov(HOST_CCREG,0);
-        emit_add(0,ECX,0);
-        emit_addimm(0,2*ccadj[i],0);
-        emit_writeword(0,(int)&Count);
-        #endif
-    emit_call((int)memdebug);
-    #ifdef __i386__
-    emit_popa();
-    #endif
-    #ifdef __arm__
-    restore_regs(0x100f);
-    #endif
-  }/**/
 }
 
 void storelr_assemble(int i,struct regstat *i_regs)
 {
   int s,th,tl;
   int temp;
-  int temp2;
+  int temp2=-1;
   int offset;
-  int jaddr=0,jaddr2;
-  int case1,case2,case3;
-  int done0,done1,done2;
+  void *jaddr=0;
+  void *case1, *case2, *case3;
+  void *done0, *done1, *done2;
   int memtarget=0,c=0;
   int agr=AGEN1+(i&1);
   u_int hr,reglist=0;
@@ -3014,13 +3032,13 @@ void storelr_assemble(int i,struct regstat *i_regs)
   if(!c) {
     emit_cmpimm(s<0||offset?temp:s,RAM_SIZE);
     if(!offset&&s!=temp) emit_mov(s,temp);
-    jaddr=(int)out;
+    jaddr=out;
     emit_jno(0);
   }
   else
   {
     if(!memtarget||!rs1[i]) {
-      jaddr=(int)out;
+      jaddr=out;
       emit_jmp(0);
     }
   }
@@ -3028,7 +3046,7 @@ void storelr_assemble(int i,struct regstat *i_regs)
   int map=get_reg(i_regs->regmap,ROREG);
   if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG);
   #else
-  if((u_int)rdram!=0x80000000) 
+  if((u_int)rdram!=0x80000000)
     emit_addimm_no_flags((u_int)rdram-(u_int)0x80000000,temp);
   #endif
 
@@ -3041,10 +3059,10 @@ void storelr_assemble(int i,struct regstat *i_regs)
     emit_xorimm(temp,3,temp);
 #endif
   emit_testimm(temp,2);
-  case2=(int)out;
+  case2=out;
   emit_jne(0);
   emit_testimm(temp,1);
-  case1=(int)out;
+  case1=out;
   emit_jne(0);
   // 0
   if (opcode[i]==0x2A) { // SWL
@@ -3061,10 +3079,10 @@ void storelr_assemble(int i,struct regstat *i_regs)
     emit_writebyte_indexed(tl,3,temp);
     if(rs2[i]) emit_shldimm(th,tl,24,temp2);
   }
-  done0=(int)out;
+  done0=out;
   emit_jmp(0);
   // 1
-  set_jump_target(case1,(int)out);
+  set_jump_target(case1, out);
   if (opcode[i]==0x2A) { // SWL
     // Write 3 msb into three least significant bytes
     if(rs2[i]) emit_rorimm(tl,8,tl);
@@ -3091,12 +3109,12 @@ void storelr_assemble(int i,struct regstat *i_regs)
     // Write two lsb into two most significant bytes
     emit_writehword_indexed(tl,1,temp);
   }
-  done1=(int)out;
+  done1=out;
   emit_jmp(0);
   // 2
-  set_jump_target(case2,(int)out);
+  set_jump_target(case2, out);
   emit_testimm(temp,1);
-  case3=(int)out;
+  case3=out;
   emit_jne(0);
   if (opcode[i]==0x2A) { // SWL
     // Write two msb into two least significant bytes
@@ -3126,10 +3144,10 @@ void storelr_assemble(int i,struct regstat *i_regs)
     emit_writehword_indexed(tl,0,temp);
     if(rs2[i]) emit_rorimm(tl,24,tl);
   }
-  done2=(int)out;
+  done2=out;
   emit_jmp(0);
   // 3
-  set_jump_target(case3,(int)out);
+  set_jump_target(case3, out);
   if (opcode[i]==0x2A) { // SWL
     // Write msb into least significant byte
     if(rs2[i]) emit_rorimm(tl,24,tl);
@@ -3152,27 +3170,27 @@ void storelr_assemble(int i,struct regstat *i_regs)
     // Write entire word
     emit_writeword_indexed(tl,-3,temp);
   }
-  set_jump_target(done0,(int)out);
-  set_jump_target(done1,(int)out);
-  set_jump_target(done2,(int)out);
+  set_jump_target(done0, out);
+  set_jump_target(done1, out);
+  set_jump_target(done2, out);
   if (opcode[i]==0x2C) { // SDL
     emit_testimm(temp,4);
-    done0=(int)out;
+    done0=out;
     emit_jne(0);
     emit_andimm(temp,~3,temp);
     emit_writeword_indexed(temp2,4,temp);
-    set_jump_target(done0,(int)out);
+    set_jump_target(done0, out);
   }
   if (opcode[i]==0x2D) { // SDR
     emit_testimm(temp,4);
-    done0=(int)out;
+    done0=out;
     emit_jeq(0);
     emit_andimm(temp,~3,temp);
     emit_writeword_indexed(temp2,-4,temp);
-    set_jump_target(done0,(int)out);
+    set_jump_target(done0, out);
   }
   if(!c||!memtarget)
-    add_stub(STORELR_STUB,jaddr,(int)out,i,(int)i_regs,temp,ccadj[i],reglist);
+    add_stub_r(STORELR_STUB,jaddr,out,i,temp,i_regs,ccadj[i],reglist);
   if(!(i_regs->waswritten&(1<<rs1[i]))&&!(new_dynarec_hacks&NDHACK_NO_SMC_CHECK)) {
     #ifdef RAM_OFFSET
     int map=get_reg(i_regs->regmap,ROREG);
@@ -3191,24 +3209,11 @@ void storelr_assemble(int i,struct regstat *i_regs)
     #if defined(HAVE_CONDITIONAL_CALL) && !defined(DESTRUCTIVE_SHIFT)
     emit_callne(invalidate_addr_reg[temp]);
     #else
-    jaddr2=(int)out;
+    void *jaddr2 = out;
     emit_jne(0);
-    add_stub(INVCODE_STUB,jaddr2,(int)out,reglist|(1<<HOST_CCREG),temp,0,0,0);
+    add_stub(INVCODE_STUB,jaddr2,out,reglist|(1<<HOST_CCREG),temp,0,0,0);
     #endif
   }
-  /*
-    emit_pusha();
-    //save_regs(0x100f);
-        emit_readword((int)&last_count,ECX);
-        if(get_reg(i_regs->regmap,CCREG)<0)
-          emit_loadreg(CCREG,HOST_CCREG);
-        emit_add(HOST_CCREG,ECX,HOST_CCREG);
-        emit_addimm(HOST_CCREG,2*ccadj[i],HOST_CCREG);
-        emit_writeword(HOST_CCREG,(int)&Count);
-    emit_call((int)memdebug);
-    emit_popa();
-    //restore_regs(0x100f);
-  /**/
 }
 
 void c1ls_assemble(int i,struct regstat *i_regs)
@@ -3222,7 +3227,8 @@ void c2ls_assemble(int i,struct regstat *i_regs)
   int ar;
   int offset;
   int memtarget=0,c=0;
-  int jaddr2=0,jaddr3,type;
+  void *jaddr2=NULL;
+  enum stub_type type;
   int agr=AGEN1+(i&1);
   int fastio_reg_override=0;
   u_int hr,reglist=0;
@@ -3260,7 +3266,7 @@ void c2ls_assemble(int i,struct regstat *i_regs)
     type=LOADW_STUB;
 
   if(c&&!memtarget) {
-    jaddr2=(int)out;
+    jaddr2=out;
     emit_jmp(0); // inline_readstub/inline_writestub?
   }
   else {
@@ -3290,7 +3296,7 @@ void c2ls_assemble(int i,struct regstat *i_regs)
     }
   }
   if(jaddr2)
-    add_stub(type,jaddr2,(int)out,i,ar,(int)i_regs,ccadj[i],reglist);
+    add_stub_r(type,jaddr2,out,i,ar,i_regs,ccadj[i],reglist);
   if(opcode[i]==0x3a) // SWC2
   if(!(i_regs->waswritten&(1<<rs1[i]))&&!(new_dynarec_hacks&NDHACK_NO_SMC_CHECK)) {
 #if defined(HOST_IMM8)
@@ -3303,9 +3309,9 @@ void c2ls_assemble(int i,struct regstat *i_regs)
     #if defined(HAVE_CONDITIONAL_CALL) && !defined(DESTRUCTIVE_SHIFT)
     emit_callne(invalidate_addr_reg[ar]);
     #else
-    jaddr3=(int)out;
+    void *jaddr3 = out;
     emit_jne(0);
-    add_stub(INVCODE_STUB,jaddr3,(int)out,reglist|(1<<HOST_CCREG),ar,0,0,0);
+    add_stub(INVCODE_STUB,jaddr3,out,reglist|(1<<HOST_CCREG),ar,0,0,0);
     #endif
   }
   if (opcode[i]==0x32) { // LWC2
@@ -3364,20 +3370,27 @@ void syscall_assemble(int i,struct regstat *i_regs)
   signed char ccreg=get_reg(i_regs->regmap,CCREG);
   assert(ccreg==HOST_CCREG);
   assert(!is_delayslot);
+  (void)ccreg;
   emit_movimm(start+i*4,EAX); // Get PC
   emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); // CHECK: is this right?  There should probably be an extra cycle...
-  emit_jmp((int)jump_syscall_hle); // XXX
+  emit_jmp(jump_syscall_hle); // XXX
 }
 
 void hlecall_assemble(int i,struct regstat *i_regs)
 {
+  extern void psxNULL();
   signed char ccreg=get_reg(i_regs->regmap,CCREG);
   assert(ccreg==HOST_CCREG);
   assert(!is_delayslot);
+  (void)ccreg;
   emit_movimm(start+i*4+4,0); // Get PC
-  emit_movimm((int)psxHLEt[source[i]&7],1);
+  uint32_t hleCode = source[i] & 0x03ffffff;
+  if (hleCode >= ARRAY_SIZE(psxHLEt))
+    emit_movimm((int)psxNULL,1);
+  else
+    emit_movimm((int)psxHLEt[hleCode],1);
   emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG); // XXX
-  emit_jmp((int)jump_hlecall);
+  emit_jmp(jump_hlecall);
 }
 
 void intcall_assemble(int i,struct regstat *i_regs)
@@ -3385,9 +3398,10 @@ void intcall_assemble(int i,struct regstat *i_regs)
   signed char ccreg=get_reg(i_regs->regmap,CCREG);
   assert(ccreg==HOST_CCREG);
   assert(!is_delayslot);
+  (void)ccreg;
   emit_movimm(start+i*4,0); // Get PC
   emit_addimm(HOST_CCREG,CLOCK_ADJUST(ccadj[i]),HOST_CCREG);
-  emit_jmp((int)jump_intcall);
+  emit_jmp(jump_intcall);
 }
 
 void ds_assemble(int i,struct regstat *i_regs)
@@ -3600,7 +3614,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[])
     int agr=AGEN1+(i&1);
     if(itype[i]==LOAD) {
       ra=get_reg(i_regs->regmap,rt1[i]);
-      if(ra<0) ra=get_reg(i_regs->regmap,-1); 
+      if(ra<0) ra=get_reg(i_regs->regmap,-1);
       assert(ra>=0);
     }
     if(itype[i]==LOADLR) {
@@ -3703,7 +3717,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[])
   }
 }
 
-int get_final_value(int hr, int i, int *value)
+static int get_final_value(int hr, int i, int *value)
 {
   int reg=regs[i].regmap[hr];
   while(i<slen-1) {
@@ -4120,7 +4134,7 @@ int match_bt(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr)
           {
             return 0;
           }
-          else 
+          else
           if((i_dirty>>hr)&1)
           {
             if(i_regmap[hr]<TEMPREG)
@@ -4186,13 +4200,37 @@ int match_bt(signed char i_regmap[],uint64_t i_is32,uint64_t i_dirty,int addr)
   return 1;
 }
 
+#ifdef DRC_DBG
+static void drc_dbg_emit_do_cmp(int i)
+{
+  extern void do_insn_cmp();
+  extern int cycle;
+  u_int hr,reglist=0;
+
+  for(hr=0;hr<HOST_REGS;hr++)
+    if(regs[i].regmap[hr]>=0) reglist|=1<<hr;
+  save_regs(reglist);
+  emit_movimm(start+i*4,0);
+  emit_writeword(0,(int)&pcaddr);
+  emit_call((int)do_insn_cmp);
+  //emit_readword((int)&cycle,0);
+  //emit_addimm(0,2,0);
+  //emit_writeword(0,(int)&cycle);
+  restore_regs(reglist);
+}
+#else
+#define drc_dbg_emit_do_cmp(x)
+#endif
+
 // Used when a branch jumps into the delay slot of another branch
 void ds_assemble_entry(int i)
 {
   int t=(ba[i]-start)>>2;
-  if(!instr_addr[t]) instr_addr[t]=(u_int)out;
+  if (!instr_addr[t])
+    instr_addr[t] = out;
   assem_debug("Assemble delay slot at %x\n",ba[i]);
   assem_debug("<->\n");
+  drc_dbg_emit_do_cmp(t);
   if(regs[t].regmap_entry[HOST_CCREG]==CCREG&&regs[t].regmap[HOST_CCREG]!=CCREG)
     wb_register(CCREG,regs[t].regmap_entry,regs[t].wasdirty,regs[t].was32);
   load_regs(regs[t].regmap_entry,regs[t].regmap,regs[t].was32,rs1[t],rs2[t]);
@@ -4265,8 +4303,8 @@ void ds_assemble_entry(int i)
 void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert)
 {
   int count;
-  int jaddr;
-  int idle=0;
+  void *jaddr;
+  void *idle=NULL;
   int t=0;
   if(itype[i]==RJUMP)
   {
@@ -4287,10 +4325,10 @@ void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert)
   if(taken==TAKEN && i==(ba[i]-start)>>2 && source[i+1]==0) {
     // Idle loop
     if(count&1) emit_addimm_and_set_flags(2*(count+2),HOST_CCREG);
-    idle=(int)out;
+    idle=out;
     //emit_subfrommem(&idlecount,HOST_CCREG); // Count idle cycles
     emit_andimm(HOST_CCREG,3,HOST_CCREG);
-    jaddr=(int)out;
+    jaddr=out;
     emit_jmp(0);
   }
   else if(*adj==0||invert) {
@@ -4302,39 +4340,39 @@ void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert)
         cycles=CLOCK_ADJUST(*adj)+count+2-*adj;
     }
     emit_addimm_and_set_flags(cycles,HOST_CCREG);
-    jaddr=(int)out;
+    jaddr=out;
     emit_jns(0);
   }
   else
   {
     emit_cmpimm(HOST_CCREG,-CLOCK_ADJUST(count+2));
-    jaddr=(int)out;
+    jaddr=out;
     emit_jns(0);
   }
-  add_stub(CC_STUB,jaddr,idle?idle:(int)out,(*adj==0||invert||idle)?0:(count+2),i,addr,taken,0);
+  add_stub(CC_STUB,jaddr,idle?idle:out,(*adj==0||invert||idle)?0:(count+2),i,addr,taken,0);
 }
 
-void do_ccstub(int n)
+static void do_ccstub(int n)
 {
   literal_pool(256);
-  assem_debug("do_ccstub %x\n",start+stubs[n][4]*4);
-  set_jump_target(stubs[n][1],(int)out);
-  int i=stubs[n][4];
-  if(stubs[n][6]==NULLDS) {
+  assem_debug("do_ccstub %x\n",start+stubs[n].b*4);
+  set_jump_target(stubs[n].addr, out);
+  int i=stubs[n].b;
+  if(stubs[n].d==NULLDS) {
     // Delay slot instruction is nullified ("likely" branch)
     wb_dirtys(regs[i].regmap,regs[i].is32,regs[i].dirty);
   }
-  else if(stubs[n][6]!=TAKEN) {
+  else if(stubs[n].d!=TAKEN) {
     wb_dirtys(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty);
   }
   else {
     if(internal_branch(branch_regs[i].is32,ba[i]))
       wb_needed_dirtys(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,ba[i]);
   }
-  if(stubs[n][5]!=-1)
+  if(stubs[n].c!=-1)
   {
     // Save PC as return address
-    emit_movimm(stubs[n][5],EAX);
+    emit_movimm(stubs[n].c,EAX);
     emit_writeword(EAX,(int)&pcaddr);
   }
   else
@@ -4363,7 +4401,7 @@ void do_ccstub(int n)
       if(rs1[i]) {
         if((branch_regs[i].dirty>>s1l)&(branch_regs[i].is32>>rs1[i])&1)
           emit_loadreg(rs1[i],s1l);
-      } 
+      }
       else {
         if((branch_regs[i].dirty>>s1l)&(branch_regs[i].is32>>rs2[i])&1)
           emit_loadreg(rs2[i],s1l);
@@ -4531,10 +4569,10 @@ void do_ccstub(int n)
   }
   // Update cycle count
   assert(branch_regs[i].regmap[HOST_CCREG]==CCREG||branch_regs[i].regmap[HOST_CCREG]==-1);
-  if(stubs[n][3]) emit_addimm(HOST_CCREG,CLOCK_ADJUST((int)stubs[n][3]),HOST_CCREG);
+  if(stubs[n].a) emit_addimm(HOST_CCREG,CLOCK_ADJUST((int)stubs[n].a),HOST_CCREG);
   emit_call((int)cc_interrupt);
-  if(stubs[n][3]) emit_addimm(HOST_CCREG,-CLOCK_ADJUST((int)stubs[n][3]),HOST_CCREG);
-  if(stubs[n][6]==TAKEN) {
+  if(stubs[n].a) emit_addimm(HOST_CCREG,-CLOCK_ADJUST((int)stubs[n].a),HOST_CCREG);
+  if(stubs[n].d==TAKEN) {
     if(internal_branch(branch_regs[i].is32,ba[i]))
       load_needed_regs(branch_regs[i].regmap,regs[(ba[i]-start)>>2].regmap_entry);
     else if(itype[i]==RJUMP) {
@@ -4543,52 +4581,24 @@ void do_ccstub(int n)
       else
         emit_loadreg(rs1[i],get_reg(branch_regs[i].regmap,rs1[i]));
     }
-  }else if(stubs[n][6]==NOTTAKEN) {
+  }else if(stubs[n].d==NOTTAKEN) {
     if(i<slen-2) load_needed_regs(branch_regs[i].regmap,regmap_pre[i+2]);
     else load_all_regs(branch_regs[i].regmap);
-  }else if(stubs[n][6]==NULLDS) {
+  }else if(stubs[n].d==NULLDS) {
     // Delay slot instruction is nullified ("likely" branch)
     if(i<slen-2) load_needed_regs(regs[i].regmap,regmap_pre[i+2]);
     else load_all_regs(regs[i].regmap);
   }else{
     load_all_regs(branch_regs[i].regmap);
   }
-  emit_jmp(stubs[n][2]); // return address
-  
-  /* This works but uses a lot of memory...
-  emit_readword((int)&last_count,ECX);
-  emit_add(HOST_CCREG,ECX,EAX);
-  emit_writeword(EAX,(int)&Count);
-  emit_call((int)gen_interupt);
-  emit_readword((int)&Count,HOST_CCREG);
-  emit_readword((int)&next_interupt,EAX);
-  emit_readword((int)&pending_exception,EBX);
-  emit_writeword(EAX,(int)&last_count);
-  emit_sub(HOST_CCREG,EAX,HOST_CCREG);
-  emit_test(EBX,EBX);
-  int jne_instr=(int)out;
-  emit_jne(0);
-  if(stubs[n][3]) emit_addimm(HOST_CCREG,-2*stubs[n][3],HOST_CCREG);
-  load_all_regs(branch_regs[i].regmap);
-  emit_jmp(stubs[n][2]); // return address
-  set_jump_target(jne_instr,(int)out);
-  emit_readword((int)&pcaddr,EAX);
-  // Call get_addr_ht instead of doing the hash table here.
-  // This code is executed infrequently and takes up a lot of space
-  // so smaller is better.
-  emit_storereg(CCREG,HOST_CCREG);
-  emit_pushreg(EAX);
-  emit_call((int)get_addr_ht);
-  emit_loadreg(CCREG,HOST_CCREG);
-  emit_addimm(ESP,4,ESP);
-  emit_jmpreg(EAX);*/
+  emit_jmp(stubs[n].retaddr);
 }
 
-add_to_linker(int addr,int target,int ext)
+static void add_to_linker(int addr,int target,int ext)
 {
   link_addr[linkcount][0]=addr;
   link_addr[linkcount][1]=target;
-  link_addr[linkcount][2]=ext;  
+  link_addr[linkcount][2]=ext;
   linkcount++;
 }
 
@@ -4614,14 +4624,14 @@ static void ujump_assemble_write_ra(int i)
     #endif
     {
       #ifdef REG_PREFETCH
-      if(temp>=0) 
+      if(temp>=0)
       {
-        if(i_regmap[temp]!=PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp);
+        if(i_regmap[temp]!=PTEMP) emit_movimm((int)hash_table_get(return_address),temp);
       }
       #endif
       emit_movimm(return_address,rt); // PC into link register
       #ifdef IMM_PREFETCH
-      emit_prefetch(hash_table[((return_address>>16)^return_address)&0xFFFF]);
+      emit_prefetch(hash_table_get(return_address));
       #endif
     }
   }
@@ -4629,17 +4639,17 @@ static void ujump_assemble_write_ra(int i)
 
 void ujump_assemble(int i,struct regstat *i_regs)
 {
-  signed char *i_regmap=i_regs->regmap;
   int ra_done=0;
   if(i==(ba[i]-start)>>2) assem_debug("idle loop\n");
   address_generation(i+1,i_regs,regs[i].regmap_entry);
   #ifdef REG_PREFETCH
   int temp=get_reg(branch_regs[i].regmap,PTEMP);
-  if(rt1[i]==31&&temp>=0) 
+  if(rt1[i]==31&&temp>=0)
   {
+    signed char *i_regmap=i_regs->regmap;
     int return_address=start+i*4+8;
-    if(get_reg(branch_regs[i].regmap,31)>0) 
-    if(i_regmap[temp]==PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp);
+    if(get_reg(branch_regs[i].regmap,31)>0)
+    if(i_regmap[temp]==PTEMP) emit_movimm((int)hash_table_get(return_address),temp);
   }
   #endif
   if(rt1[i]==31&&(rt1[i]==rs1[i+1]||rt1[i]==rs2[i+1])) {
@@ -4689,22 +4699,21 @@ static void rjump_assemble_write_ra(int i)
   assert(rt>=0);
   return_address=start+i*4+8;
   #ifdef REG_PREFETCH
-  if(temp>=0) 
+  if(temp>=0)
   {
-    if(i_regmap[temp]!=PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp);
+    if(i_regmap[temp]!=PTEMP) emit_movimm((int)hash_table_get(return_address),temp);
   }
   #endif
   emit_movimm(return_address,rt); // PC into link register
   #ifdef IMM_PREFETCH
-  emit_prefetch(hash_table[((return_address>>16)^return_address)&0xFFFF]);
+  emit_prefetch(hash_table_get(return_address));
   #endif
 }
 
 void rjump_assemble(int i,struct regstat *i_regs)
 {
-  signed char *i_regmap=i_regs->regmap;
   int temp;
-  int rs,cc,adj;
+  int rs,cc;
   int ra_done=0;
   rs=get_reg(branch_regs[i].regmap,rs1[i]);
   assert(rs>=0);
@@ -4718,11 +4727,12 @@ void rjump_assemble(int i,struct regstat *i_regs)
   }
   address_generation(i+1,i_regs,regs[i].regmap_entry);
   #ifdef REG_PREFETCH
-  if(rt1[i]==31) 
+  if(rt1[i]==31)
   {
     if((temp=get_reg(branch_regs[i].regmap,PTEMP))>=0) {
+      signed char *i_regmap=i_regs->regmap;
       int return_address=start+i*4+8;
-      if(i_regmap[temp]==PTEMP) emit_movimm((int)hash_table[((return_address>>16)^return_address)&0xFFFF],temp);
+      if(i_regmap[temp]==PTEMP) emit_movimm((int)hash_table_get(return_address),temp);
     }
   }
   #endif
@@ -4749,6 +4759,7 @@ void rjump_assemble(int i,struct regstat *i_regs)
     rjump_assemble_write_ra(i);
   cc=get_reg(branch_regs[i].regmap,CCREG);
   assert(cc==HOST_CCREG);
+  (void)cc;
   #ifdef USE_MINI_HT
   int rh=get_reg(branch_regs[i].regmap,RHASH);
   int ht=get_reg(branch_regs[i].regmap,RHTBL);
@@ -4778,7 +4789,7 @@ void rjump_assemble(int i,struct regstat *i_regs)
   //if(adj) emit_addimm(cc,2*(ccadj[i]+2-adj),cc); // ??? - Shouldn't happen
   //assert(adj==0);
   emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
-  add_stub(CC_STUB,(int)out,jump_vaddr_reg[rs],0,i,-1,TAKEN,0);
+  add_stub(CC_STUB,out,jump_vaddr_reg[rs],0,i,-1,TAKEN,0);
   if(itype[i+1]==COP0&&(source[i+1]&0x3f)==0x10)
     // special case for RFE
     emit_jmp(0);
@@ -4792,41 +4803,8 @@ void rjump_assemble(int i,struct regstat *i_regs)
   else
   #endif
   {
-    //if(rs!=EAX) emit_mov(rs,EAX);
-    //emit_jmp((int)jump_vaddr_eax);
     emit_jmp(jump_vaddr_reg[rs]);
   }
-  /* Check hash table
-  temp=!rs;
-  emit_mov(rs,temp);
-  emit_shrimm(rs,16,rs);
-  emit_xor(temp,rs,rs);
-  emit_movzwl_reg(rs,rs);
-  emit_shlimm(rs,4,rs);
-  emit_cmpmem_indexed((int)hash_table,rs,temp);
-  emit_jne((int)out+14);
-  emit_readword_indexed((int)hash_table+4,rs,rs);
-  emit_jmpreg(rs);
-  emit_cmpmem_indexed((int)hash_table+8,rs,temp);
-  emit_addimm_no_flags(8,rs);
-  emit_jeq((int)out-17);
-  // No hit on hash table, call compiler
-  emit_pushreg(temp);
-//DEBUG >
-#ifdef DEBUG_CYCLE_COUNT
-  emit_readword((int)&last_count,ECX);
-  emit_add(HOST_CCREG,ECX,HOST_CCREG);
-  emit_readword((int)&next_interupt,ECX);
-  emit_writeword(HOST_CCREG,(int)&Count);
-  emit_sub(HOST_CCREG,ECX,HOST_CCREG);
-  emit_writeword(ECX,(int)&last_count);
-#endif
-//DEBUG <
-  emit_storereg(CCREG,HOST_CCREG);
-  emit_call((int)get_addr);
-  emit_loadreg(CCREG,HOST_CCREG);
-  emit_addimm(ESP,4,ESP);
-  emit_jmpreg(EAX);*/
   #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK
   if(rt1[i]!=31&&i<slen-2&&(((u_int)out)&7)) emit_mov(13,13);
   #endif
@@ -4850,7 +4828,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
   #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK
   if(i>(ba[i]-start)>>2) invert=1;
   #endif
-  
+
   if(ooo[i]) {
     s1l=get_reg(branch_regs[i].regmap,rs1[i]);
     s1h=get_reg(branch_regs[i].regmap,rs1[i]|64);
@@ -4905,7 +4883,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
     load_regs(regs[i].regmap,branch_regs[i].regmap,regs[i].was32,CCREG,CCREG);
     cc=get_reg(branch_regs[i].regmap,CCREG);
     assert(cc==HOST_CCREG);
-    if(unconditional) 
+    if(unconditional)
       store_regs_bt(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,ba[i]);
     //do_cc(i,branch_regs[i].regmap,&adj,unconditional?ba[i]:-1,unconditional);
     //assem_debug("cycle count (adj)\n");
@@ -4932,12 +4910,12 @@ void cjump_assemble(int i,struct regstat *i_regs)
     }
     else if(nop) {
       emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
-      int jaddr=(int)out;
+      void *jaddr=out;
       emit_jns(0);
-      add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,NOTTAKEN,0);
+      add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
     }
     else {
-      int taken=0,nottaken=0,nottaken1=0;
+      void *taken = NULL, *nottaken = NULL, *nottaken1 = NULL;
       do_cc(i,branch_regs[i].regmap,&adj,-1,0,invert);
       if(adj&&!invert) emit_addimm(cc,CLOCK_ADJUST(ccadj[i]+2-adj),cc);
       if(!only32)
@@ -4947,37 +4925,37 @@ void cjump_assemble(int i,struct regstat *i_regs)
         {
           if(s2h>=0) emit_cmp(s1h,s2h);
           else emit_test(s1h,s1h);
-          nottaken1=(int)out;
+          nottaken1=out;
           emit_jne(1);
         }
         if(opcode[i]==5) // BNE
         {
           if(s2h>=0) emit_cmp(s1h,s2h);
           else emit_test(s1h,s1h);
-          if(invert) taken=(int)out;
+          if(invert) taken=out;
           else add_to_linker((int)out,ba[i],internal);
           emit_jne(0);
         }
         if(opcode[i]==6) // BLEZ
         {
           emit_test(s1h,s1h);
-          if(invert) taken=(int)out;
+          if(invert) taken=out;
           else add_to_linker((int)out,ba[i],internal);
           emit_js(0);
-          nottaken1=(int)out;
+          nottaken1=out;
           emit_jne(1);
         }
         if(opcode[i]==7) // BGTZ
         {
           emit_test(s1h,s1h);
-          nottaken1=(int)out;
+          nottaken1=out;
           emit_js(1);
-          if(invert) taken=(int)out;
+          if(invert) taken=out;
           else add_to_linker((int)out,ba[i],internal);
           emit_jne(0);
         }
       } // if(!only32)
-          
+
       //printf("branch(%d): eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d\n",i,branch_regs[i].regmap[0],branch_regs[i].regmap[1],branch_regs[i].regmap[2],branch_regs[i].regmap[3],branch_regs[i].regmap[5],branch_regs[i].regmap[6],branch_regs[i].regmap[7]);
       assert(s1l>=0);
       if(opcode[i]==4) // BEQ
@@ -4985,7 +4963,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
         if(s2l>=0) emit_cmp(s1l,s2l);
         else emit_test(s1l,s1l);
         if(invert){
-          nottaken=(int)out;
+          nottaken=out;
           emit_jne(1);
         }else{
           add_to_linker((int)out,ba[i],internal);
@@ -4997,7 +4975,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
         if(s2l>=0) emit_cmp(s1l,s2l);
         else emit_test(s1l,s1l);
         if(invert){
-          nottaken=(int)out;
+          nottaken=out;
           emit_jeq(1);
         }else{
           add_to_linker((int)out,ba[i],internal);
@@ -5008,7 +4986,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
       {
         emit_cmpimm(s1l,1);
         if(invert){
-          nottaken=(int)out;
+          nottaken=out;
           emit_jge(1);
         }else{
           add_to_linker((int)out,ba[i],internal);
@@ -5019,7 +4997,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
       {
         emit_cmpimm(s1l,1);
         if(invert){
-          nottaken=(int)out;
+          nottaken=out;
           emit_jl(1);
         }else{
           add_to_linker((int)out,ba[i],internal);
@@ -5027,7 +5005,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
         }
       }
       if(invert) {
-        if(taken) set_jump_target(taken,(int)out);
+        if(taken) set_jump_target(taken, out);
         #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK
         if(match&&(!internal||!is_ds[(ba[i]-start)>>2])) {
           if(adj) {
@@ -5056,10 +5034,10 @@ void cjump_assemble(int i,struct regstat *i_regs)
             emit_jmp(0);
           }
         }
-        set_jump_target(nottaken,(int)out);
+        set_jump_target(nottaken, out);
       }
 
-      if(nottaken1) set_jump_target(nottaken1,(int)out);
+      if(nottaken1) set_jump_target(nottaken1, out);
       if(adj) {
         if(!invert) emit_addimm(cc,CLOCK_ADJUST(adj),cc);
       }
@@ -5071,7 +5049,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
     //if(likely[i]) printf("IOL\n");
     //else
     //printf("IOE\n");
-    int taken=0,nottaken=0,nottaken1=0;
+    void *taken = NULL, *nottaken = NULL, *nottaken1 = NULL;
     if(!unconditional&&!nop) {
       if(!only32)
       {
@@ -5080,60 +5058,60 @@ void cjump_assemble(int i,struct regstat *i_regs)
         {
           if(s2h>=0) emit_cmp(s1h,s2h);
           else emit_test(s1h,s1h);
-          nottaken1=(int)out;
+          nottaken1=out;
           emit_jne(2);
         }
         if((opcode[i]&0x2f)==5) // BNE
         {
           if(s2h>=0) emit_cmp(s1h,s2h);
           else emit_test(s1h,s1h);
-          taken=(int)out;
+          taken=out;
           emit_jne(1);
         }
         if((opcode[i]&0x2f)==6) // BLEZ
         {
           emit_test(s1h,s1h);
-          taken=(int)out;
+          taken=out;
           emit_js(1);
-          nottaken1=(int)out;
+          nottaken1=out;
           emit_jne(2);
         }
         if((opcode[i]&0x2f)==7) // BGTZ
         {
           emit_test(s1h,s1h);
-          nottaken1=(int)out;
+          nottaken1=out;
           emit_js(2);
-          taken=(int)out;
+          taken=out;
           emit_jne(1);
         }
       } // if(!only32)
-          
+
       //printf("branch(%d): eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d\n",i,branch_regs[i].regmap[0],branch_regs[i].regmap[1],branch_regs[i].regmap[2],branch_regs[i].regmap[3],branch_regs[i].regmap[5],branch_regs[i].regmap[6],branch_regs[i].regmap[7]);
       assert(s1l>=0);
       if((opcode[i]&0x2f)==4) // BEQ
       {
         if(s2l>=0) emit_cmp(s1l,s2l);
         else emit_test(s1l,s1l);
-        nottaken=(int)out;
+        nottaken=out;
         emit_jne(2);
       }
       if((opcode[i]&0x2f)==5) // BNE
       {
         if(s2l>=0) emit_cmp(s1l,s2l);
         else emit_test(s1l,s1l);
-        nottaken=(int)out;
+        nottaken=out;
         emit_jeq(2);
       }
       if((opcode[i]&0x2f)==6) // BLEZ
       {
         emit_cmpimm(s1l,1);
-        nottaken=(int)out;
+        nottaken=out;
         emit_jge(2);
       }
       if((opcode[i]&0x2f)==7) // BGTZ
       {
         emit_cmpimm(s1l,1);
-        nottaken=(int)out;
+        nottaken=out;
         emit_jl(2);
       }
     } // if(!unconditional)
@@ -5147,7 +5125,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
     ds_unneeded_upper|=1;
     // branch taken
     if(!nop) {
-      if(taken) set_jump_target(taken,(int)out);
+      if(taken) set_jump_target(taken, out);
       assem_debug("1:\n");
       wb_invalidate(regs[i].regmap,branch_regs[i].regmap,regs[i].dirty,regs[i].is32,
                     ds_unneeded,ds_unneeded_upper);
@@ -5182,8 +5160,8 @@ void cjump_assemble(int i,struct regstat *i_regs)
     // branch not taken
     cop1_usable=prev_cop1_usable;
     if(!unconditional) {
-      if(nottaken1) set_jump_target(nottaken1,(int)out);
-      set_jump_target(nottaken,(int)out);
+      if(nottaken1) set_jump_target(nottaken1, out);
+      set_jump_target(nottaken, out);
       assem_debug("2:\n");
       if(!likely[i]) {
         wb_invalidate(regs[i].regmap,branch_regs[i].regmap,regs[i].dirty,regs[i].is32,
@@ -5198,18 +5176,18 @@ void cjump_assemble(int i,struct regstat *i_regs)
         // Cycle count isn't in a register, temporarily load it then write it out
         emit_loadreg(CCREG,HOST_CCREG);
         emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
-        int jaddr=(int)out;
+        void *jaddr=out;
         emit_jns(0);
-        add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,NOTTAKEN,0);
+        add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
         emit_storereg(CCREG,HOST_CCREG);
       }
       else{
         cc=get_reg(i_regmap,CCREG);
         assert(cc==HOST_CCREG);
         emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
-        int jaddr=(int)out;
+        void *jaddr=out;
         emit_jns(0);
-        add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
+        add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
       }
     }
   }
@@ -5284,13 +5262,13 @@ void sjump_assemble(int i,struct regstat *i_regs)
         return_address=start+i*4+8;
         emit_movimm(return_address,rt); // PC into link register
         #ifdef IMM_PREFETCH
-        if(!nevertaken) emit_prefetch(hash_table[((return_address>>16)^return_address)&0xFFFF]);
+        if(!nevertaken) emit_prefetch(hash_table_get(return_address));
         #endif
       }
     }
     cc=get_reg(branch_regs[i].regmap,CCREG);
     assert(cc==HOST_CCREG);
-    if(unconditional) 
+    if(unconditional)
       store_regs_bt(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,ba[i]);
     //do_cc(i,branch_regs[i].regmap,&adj,unconditional?ba[i]:-1,unconditional);
     assem_debug("cycle count (adj)\n");
@@ -5317,12 +5295,12 @@ void sjump_assemble(int i,struct regstat *i_regs)
     }
     else if(nevertaken) {
       emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
-      int jaddr=(int)out;
+      void *jaddr=out;
       emit_jns(0);
-      add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,NOTTAKEN,0);
+      add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
     }
     else {
-      int nottaken=0;
+      void *nottaken = NULL;
       do_cc(i,branch_regs[i].regmap,&adj,-1,0,invert);
       if(adj&&!invert) emit_addimm(cc,CLOCK_ADJUST(ccadj[i]+2-adj),cc);
       if(!only32)
@@ -5332,7 +5310,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
         {
           emit_test(s1h,s1h);
           if(invert){
-            nottaken=(int)out;
+            nottaken=out;
             emit_jns(1);
           }else{
             add_to_linker((int)out,ba[i],internal);
@@ -5343,7 +5321,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
         {
           emit_test(s1h,s1h);
           if(invert){
-            nottaken=(int)out;
+            nottaken=out;
             emit_js(1);
           }else{
             add_to_linker((int)out,ba[i],internal);
@@ -5358,7 +5336,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
         {
           emit_test(s1l,s1l);
           if(invert){
-            nottaken=(int)out;
+            nottaken=out;
             emit_jns(1);
           }else{
             add_to_linker((int)out,ba[i],internal);
@@ -5369,7 +5347,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
         {
           emit_test(s1l,s1l);
           if(invert){
-            nottaken=(int)out;
+            nottaken=out;
             emit_js(1);
           }else{
             add_to_linker((int)out,ba[i],internal);
@@ -5377,7 +5355,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
           }
         }
       } // if(!only32)
-          
+
       if(invert) {
         #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK
         if(match&&(!internal||!is_ds[(ba[i]-start)>>2])) {
@@ -5407,7 +5385,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
             emit_jmp(0);
           }
         }
-        set_jump_target(nottaken,(int)out);
+        set_jump_target(nottaken, out);
       }
 
       if(adj) {
@@ -5419,7 +5397,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
   {
     // In-order execution (branch first)
     //printf("IOE\n");
-    int nottaken=0;
+    void *nottaken = NULL;
     if(rt1[i]==31) {
       int rt,return_address;
       rt=get_reg(branch_regs[i].regmap,31);
@@ -5428,7 +5406,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
         return_address=start+i*4+8;
         emit_movimm(return_address,rt); // PC into link register
         #ifdef IMM_PREFETCH
-        emit_prefetch(hash_table[((return_address>>16)^return_address)&0xFFFF]);
+        emit_prefetch(hash_table_get(return_address));
         #endif
       }
     }
@@ -5440,13 +5418,13 @@ void sjump_assemble(int i,struct regstat *i_regs)
         if((opcode2[i]&0x0d)==0) // BLTZ/BLTZL/BLTZAL/BLTZALL
         {
           emit_test(s1h,s1h);
-          nottaken=(int)out;
+          nottaken=out;
           emit_jns(1);
         }
         if((opcode2[i]&0x0d)==1) // BGEZ/BGEZL/BGEZAL/BGEZALL
         {
           emit_test(s1h,s1h);
-          nottaken=(int)out;
+          nottaken=out;
           emit_js(1);
         }
       } // if(!only32)
@@ -5456,13 +5434,13 @@ void sjump_assemble(int i,struct regstat *i_regs)
         if((opcode2[i]&0x0d)==0) // BLTZ/BLTZL/BLTZAL/BLTZALL
         {
           emit_test(s1l,s1l);
-          nottaken=(int)out;
+          nottaken=out;
           emit_jns(1);
         }
         if((opcode2[i]&0x0d)==1) // BGEZ/BGEZL/BGEZAL/BGEZALL
         {
           emit_test(s1l,s1l);
-          nottaken=(int)out;
+          nottaken=out;
           emit_js(1);
         }
       }
@@ -5511,7 +5489,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
     // branch not taken
     cop1_usable=prev_cop1_usable;
     if(!unconditional) {
-      set_jump_target(nottaken,(int)out);
+      set_jump_target(nottaken, out);
       assem_debug("1:\n");
       if(!likely[i]) {
         wb_invalidate(regs[i].regmap,branch_regs[i].regmap,regs[i].dirty,regs[i].is32,
@@ -5526,18 +5504,18 @@ void sjump_assemble(int i,struct regstat *i_regs)
         // Cycle count isn't in a register, temporarily load it then write it out
         emit_loadreg(CCREG,HOST_CCREG);
         emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
-        int jaddr=(int)out;
+        void *jaddr=out;
         emit_jns(0);
-        add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,NOTTAKEN,0);
+        add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
         emit_storereg(CCREG,HOST_CCREG);
       }
       else{
         cc=get_reg(i_regmap,CCREG);
         assert(cc==HOST_CCREG);
         emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
-        int jaddr=(int)out;
+        void *jaddr=out;
         emit_jns(0);
-        add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
+        add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
       }
     }
   }
@@ -5551,7 +5529,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
   match=match_bt(branch_regs[i].regmap,branch_regs[i].is32,branch_regs[i].dirty,ba[i]);
   assem_debug("fmatch=%d\n",match);
   int fs,cs;
-  int eaddr;
+  void *eaddr;
   int invert=0;
   int internal=internal_branch(branch_regs[i].is32,ba[i]);
   if(i==(ba[i]-start)>>2) assem_debug("idle loop\n");
@@ -5573,9 +5551,9 @@ void fjump_assemble(int i,struct regstat *i_regs)
     cs=get_reg(i_regmap,CSREG);
     assert(cs>=0);
     emit_testimm(cs,0x20000000);
-    eaddr=(int)out;
+    eaddr=out;
     emit_jeq(0);
-    add_stub(FP_STUB,eaddr,(int)out,i,cs,(int)i_regs,0,0);
+    add_stub_r(FP_STUB,eaddr,out,i,cs,i_regs,0,0);
     cop1_usable=1;
   }
 
@@ -5599,7 +5577,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
     do_cc(i,branch_regs[i].regmap,&adj,-1,0,invert);
     assem_debug("cycle count (adj)\n");
     if(1) {
-      int nottaken=0;
+      void *nottaken = NULL;
       if(adj&&!invert) emit_addimm(cc,CLOCK_ADJUST(ccadj[i]+2-adj),cc);
       if(1) {
         assert(fs>=0);
@@ -5607,7 +5585,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
         if(source[i]&0x10000) // BC1T
         {
           if(invert){
-            nottaken=(int)out;
+            nottaken=out;
             emit_jeq(1);
           }else{
             add_to_linker((int)out,ba[i],internal);
@@ -5616,7 +5594,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
         }
         else // BC1F
           if(invert){
-            nottaken=(int)out;
+            nottaken=out;
             emit_jne(1);
           }else{
             add_to_linker((int)out,ba[i],internal);
@@ -5625,7 +5603,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
         {
         }
       } // if(!only32)
-          
+
       if(invert) {
         if(adj) emit_addimm(cc,-CLOCK_ADJUST(adj),cc);
         #ifdef CORTEX_A8_BRANCH_PREDICTION_HACK
@@ -5644,7 +5622,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
           add_to_linker((int)out,ba[i],internal);
           emit_jmp(0);
         }
-        set_jump_target(nottaken,(int)out);
+        set_jump_target(nottaken, out);
       }
 
       if(adj) {
@@ -5656,7 +5634,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
   {
     // In-order execution (branch first)
     //printf("IOE\n");
-    int nottaken=0;
+    void *nottaken = NULL;
     if(1) {
       //printf("branch(%d): eax=%d ecx=%d edx=%d ebx=%d ebp=%d esi=%d edi=%d\n",i,branch_regs[i].regmap[0],branch_regs[i].regmap[1],branch_regs[i].regmap[2],branch_regs[i].regmap[3],branch_regs[i].regmap[5],branch_regs[i].regmap[6],branch_regs[i].regmap[7]);
       if(1) {
@@ -5664,12 +5642,12 @@ void fjump_assemble(int i,struct regstat *i_regs)
         emit_testimm(fs,0x800000);
         if(source[i]&0x10000) // BC1T
         {
-          nottaken=(int)out;
+          nottaken=out;
           emit_jeq(1);
         }
         else // BC1F
         {
-          nottaken=(int)out;
+          nottaken=out;
           emit_jne(1);
         }
       }
@@ -5716,7 +5694,7 @@ void fjump_assemble(int i,struct regstat *i_regs)
 
     // branch not taken
     if(1) { // <- FIXME (don't need this)
-      set_jump_target(nottaken,(int)out);
+      set_jump_target(nottaken, out);
       assem_debug("1:\n");
       if(!likely[i]) {
         wb_invalidate(regs[i].regmap,branch_regs[i].regmap,regs[i].dirty,regs[i].is32,
@@ -5731,18 +5709,18 @@ void fjump_assemble(int i,struct regstat *i_regs)
         // Cycle count isn't in a register, temporarily load it then write it out
         emit_loadreg(CCREG,HOST_CCREG);
         emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
-        int jaddr=(int)out;
+        void *jaddr=out;
         emit_jns(0);
-        add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,NOTTAKEN,0);
+        add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
         emit_storereg(CCREG,HOST_CCREG);
       }
       else{
         cc=get_reg(i_regmap,CCREG);
         assert(cc==HOST_CCREG);
         emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
-        int jaddr=(int)out;
+        void *jaddr=out;
         emit_jns(0);
-        add_stub(CC_STUB,jaddr,(int)out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
+        add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
       }
     }
   }
@@ -5754,9 +5732,8 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
   int s1h=get_reg(i_regs->regmap,rs1[i]|64);
   int s2l=get_reg(i_regs->regmap,rs2[i]);
   int s2h=get_reg(i_regs->regmap,rs2[i]|64);
-  void *nt_branch=NULL;
-  int taken=0;
-  int nottaken=0;
+  void *taken = NULL;
+  void *nottaken = NULL;
   int unconditional=0;
   if(rs1[i]==0)
   {
@@ -5771,7 +5748,7 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
     s1h=s2h=-1;
   }
   int hr=0;
-  int addr,alt,ntaddr;
+  int addr=-1,alt=-1,ntaddr=-1;
   if(i_regs->regmap[HOST_BTREG]<0) {addr=HOST_BTREG;}
   else {
     while(hr<HOST_REGS)
@@ -5889,13 +5866,13 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
     if(s1h>=0) {
       if(s2h>=0) emit_cmp(s1h,s2h);
       else emit_test(s1h,s1h);
-      nottaken=(int)out;
+      nottaken=out;
       emit_jne(0);
     }
     if(s2l>=0) emit_cmp(s1l,s2l);
     else emit_test(s1l,s1l);
-    if(nottaken) set_jump_target(nottaken,(int)out);
-    nottaken=(int)out;
+    if(nottaken) set_jump_target(nottaken, out);
+    nottaken=out;
     emit_jne(0);
   }
   if((opcode[i]&0x3f)==0x15) // BNEL
@@ -5903,14 +5880,14 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
     if(s1h>=0) {
       if(s2h>=0) emit_cmp(s1h,s2h);
       else emit_test(s1h,s1h);
-      taken=(int)out;
+      taken=out;
       emit_jne(0);
     }
     if(s2l>=0) emit_cmp(s1l,s2l);
     else emit_test(s1l,s1l);
-    nottaken=(int)out;
+    nottaken=out;
     emit_jeq(0);
-    if(taken) set_jump_target(taken,(int)out);
+    if(taken) set_jump_target(taken, out);
   }
   if((opcode[i]&0x3f)==6) // BLEZ
   {
@@ -5963,13 +5940,13 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
     if((source[i]&0x30000)==0x20000) // BC1FL
     {
       emit_testimm(s1l,0x800000);
-      nottaken=(int)out;
+      nottaken=out;
       emit_jne(0);
     }
     if((source[i]&0x30000)==0x30000) // BC1TL
     {
       emit_testimm(s1l,0x800000);
-      nottaken=(int)out;
+      nottaken=out;
       emit_jeq(0);
     }
   }
@@ -5991,13 +5968,13 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
   void *compiled_target_addr=check_addr(target_addr);
   emit_extjump_ds((int)branch_addr,target_addr);
   if(compiled_target_addr) {
-    set_jump_target((int)branch_addr,(int)compiled_target_addr);
+    set_jump_target(branch_addr, compiled_target_addr);
     add_link(target_addr,stub);
   }
-  else set_jump_target((int)branch_addr,(int)stub);
+  else set_jump_target(branch_addr, stub);
   if(likely[i]) {
     // Not-taken path
-    set_jump_target((int)nottaken,(int)out);
+    set_jump_target(nottaken, out);
     wb_dirtys(regs[i].regmap,regs[i].is32,regs[i].dirty);
     void *branch_addr=out;
     emit_jmp(0);
@@ -6006,10 +5983,10 @@ static void pagespan_assemble(int i,struct regstat *i_regs)
     void *compiled_target_addr=check_addr(target_addr);
     emit_extjump_ds((int)branch_addr,target_addr);
     if(compiled_target_addr) {
-      set_jump_target((int)branch_addr,(int)compiled_target_addr);
+      set_jump_target(branch_addr, compiled_target_addr);
       add_link(target_addr,stub);
     }
-    else set_jump_target((int)branch_addr,(int)stub);
+    else set_jump_target(branch_addr, stub);
   }
 }
 
@@ -6097,11 +6074,11 @@ static void pagespan_ds()
 #else
   emit_cmpimm(btaddr,start+4);
 #endif
-  int branch=(int)out;
+  void *branch = out;
   emit_jeq(0);
   store_regs_bt(regs[0].regmap,regs[0].is32,regs[0].dirty,-1);
   emit_jmp(jump_vaddr_reg[btaddr]);
-  set_jump_target(branch,(int)out);
+  set_jump_target(branch, out);
   store_regs_bt(regs[0].regmap,regs[0].is32,regs[0].dirty,start+4);
   load_regs_bt(regs[0].regmap,regs[0].is32,regs[0].dirty,start+4);
 }
@@ -6133,14 +6110,14 @@ void unneeded_registers(int istart,int iend,int r)
     {
       // If subroutine call, flag return address as a possible branch target
       if(rt1[i]==31 && i<slen-2) bt[i+2]=1;
-      
+
       if(ba[i]<start || ba[i]>=(start+slen*4))
       {
         // Branch out of this block, flush all regs
         u=1;
         uu=1;
         gte_u=gte_u_unknown;
-        /* Hexagon hack 
+        /* Hexagon hack
         if(itype[i]==UJUMP&&rt1[i]==31)
         {
           uu=u=0x300C00F; // Discard at, v0-v1, t6-t9
@@ -6730,7 +6707,7 @@ void clean_registers(int istart,int iend,int wr)
         if((regs[i].regmap[r]&63)==rt2[i]) wont_dirty_i|=1<<r;
         if(regs[i].regmap[r]==CCREG) wont_dirty_i|=1<<r;
         if(i>istart) {
-          if(itype[i]!=RJUMP&&itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=FJUMP) 
+          if(itype[i]!=RJUMP&&itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=FJUMP)
           {
             // Don't store a register immediately after writing it,
             // may prevent dual-issue.
@@ -6764,7 +6741,7 @@ void clean_registers(int istart,int iend,int wr)
               if(r!=EXCLUDE_REG) {
                 if(regs[i].regmap[r]==regmap_pre[i+2][r]) {
                   regs[i+2].wasdirty&=wont_dirty_i|~(1<<r);
-                }else {/*printf("i: %x (%d) mismatch(+2): %d\n",start+i*4,i,r);/*assert(!((wont_dirty_i>>r)&1));*/}
+                }else {/*printf("i: %x (%d) mismatch(+2): %d\n",start+i*4,i,r);assert(!((wont_dirty_i>>r)&1));*/}
               }
             }
           }
@@ -6776,7 +6753,7 @@ void clean_registers(int istart,int iend,int wr)
               if(r!=EXCLUDE_REG) {
                 if(regs[i].regmap[r]==regmap_pre[i+1][r]) {
                   regs[i+1].wasdirty&=wont_dirty_i|~(1<<r);
-                }else {/*printf("i: %x (%d) mismatch(+1): %d\n",start+i*4,i,r);/*assert(!((wont_dirty_i>>r)&1));*/}
+                }else {/*printf("i: %x (%d) mismatch(+1): %d\n",start+i*4,i,r);assert(!((wont_dirty_i>>r)&1));*/}
               }
             }
           }
@@ -6819,7 +6796,7 @@ void clean_registers(int istart,int iend,int wr)
             wont_dirty_i|=((unneeded_reg[i]>>(regmap_pre[i][r]&63))&1)<<r;
           } else {
             wont_dirty_i|=1<<r;
-            /*printf("i: %x (%d) mismatch: %d\n",start+i*4,i,r);/*assert(!((will_dirty>>r)&1));*/
+            /*printf("i: %x (%d) mismatch: %d\n",start+i*4,i,r);assert(!((will_dirty>>r)&1));*/
           }
         }
       }
@@ -6925,13 +6902,14 @@ static void disassemble_inst(int i) {}
 static int new_dynarec_test(void)
 {
   int (*testfunc)(void) = (void *)out;
+  void *beginning;
   int ret;
+
+  beginning = start_block();
   emit_movimm(DRC_TEST_VAL,0); // test
   emit_jmpreg(14);
   literal_pool(0);
-#ifdef __arm__
-  __clear_cache((void *)testfunc, out);
-#endif
+  end_block(beginning);
   SysPrintf("testing if we can run recompiled code..\n");
   ret = testfunc();
   if (ret == DRC_TEST_VAL)
@@ -6968,20 +6946,43 @@ void new_dynarec_clear_full()
 void new_dynarec_init()
 {
   SysPrintf("Init new dynarec\n");
-  out=(u_char *)BASE_ADDR;
-#if BASE_ADDR_FIXED
-  if (mmap (out, 1<<TARGET_SIZE_2,
+
+  // allocate/prepare a buffer for translation cache
+  // see assem_arm.h for some explanation
+#if   defined(BASE_ADDR_FIXED)
+  if (mmap (translation_cache, 1 << TARGET_SIZE_2,
+            PROT_READ | PROT_WRITE | PROT_EXEC,
+            MAP_PRIVATE | MAP_ANONYMOUS,
+            -1, 0) != translation_cache) {
+    SysPrintf("mmap() failed: %s\n", strerror(errno));
+    SysPrintf("disable BASE_ADDR_FIXED and recompile\n");
+    abort();
+  }
+#elif defined(BASE_ADDR_DYNAMIC)
+  #ifdef VITA
+  sceBlock = sceKernelAllocMemBlockForVM("code", 1 << TARGET_SIZE_2);
+  if (sceBlock < 0)
+    SysPrintf("sceKernelAllocMemBlockForVM failed\n");
+  int ret = sceKernelGetMemBlockBase(sceBlock, (void **)&translation_cache);
+  if (ret < 0)
+    SysPrintf("sceKernelGetMemBlockBase failed\n");
+  #else
+  translation_cache = mmap (NULL, 1 << TARGET_SIZE_2,
             PROT_READ | PROT_WRITE | PROT_EXEC,
-            MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS,
-            -1, 0) <= 0) {
+            MAP_PRIVATE | MAP_ANONYMOUS, -1, 0);
+  if (translation_cache == MAP_FAILED) {
     SysPrintf("mmap() failed: %s\n", strerror(errno));
+    abort();
   }
+  #endif
 #else
+  #ifndef NO_WRITE_EXEC
   // not all systems allow execute in data segment by default
-  if (mprotect(out, 1<<TARGET_SIZE_2, PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
+  if (mprotect((void *)BASE_ADDR, 1<<TARGET_SIZE_2, PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
     SysPrintf("mprotect() failed: %s\n", strerror(errno));
+  #endif
 #endif
-  int n;
+  out=(u_char *)BASE_ADDR;
   cycle_multiplier=200;
   new_dynarec_clear_full();
 #ifdef HOST_IMM8
@@ -7000,9 +7001,15 @@ void new_dynarec_init()
 void new_dynarec_cleanup()
 {
   int n;
-  #if BASE_ADDR_FIXED
-  if (munmap ((void *)BASE_ADDR, 1<<TARGET_SIZE_2) < 0) {SysPrintf("munmap() failed\n");}
+#if defined(BASE_ADDR_FIXED) || defined(BASE_ADDR_DYNAMIC)
+  #ifdef VITA
+  sceKernelFreeMemBlock(sceBlock);
+  sceBlock = -1;
+  #else
+  if (munmap ((void *)BASE_ADDR, 1<<TARGET_SIZE_2) < 0)
+    SysPrintf("munmap() failed\n");
   #endif
+#endif
   for(n=0;n<4096;n++) ll_clear(jump_in+n);
   for(n=0;n<4096;n++) ll_clear(jump_out+n);
   for(n=0;n<4096;n++) ll_clear(jump_dirty+n);
@@ -7030,6 +7037,7 @@ static u_int *get_source_start(u_int addr, u_int *limit)
     *limit = (addr & 0x80600000) + 0x00200000;
     return (u_int *)((u_int)rdram + (addr&0x1fffff));
   }
+  return NULL;
 }
 
 static u_int scan_for_ret(u_int addr)
@@ -7047,6 +7055,7 @@ static u_int scan_for_ret(u_int addr)
     if (*mem == 0x03e00008) // jr $ra
       return addr + 8;
   }
+  return addr;
 }
 
 struct savestate_block {
@@ -7070,7 +7079,7 @@ int new_dynarec_save_blocks(void *save, int size)
   u_int addr;
 
   o = 0;
-  for (p = 0; p < sizeof(jump_in) / sizeof(jump_in[0]); p++) {
+  for (p = 0; p < ARRAY_SIZE(jump_in); p++) {
     bcnt = 0;
     for (head = jump_in[p]; head != NULL; head = head->next) {
       tmp_blocks[bcnt].addr = head->vaddr;
@@ -7137,16 +7146,11 @@ int new_recompile_block(int addr)
   u_int state_rflags = 0;
   int i;
 
-  assem_debug("NOTCOMPILED: addr = %x -> %x\n", (int)addr, (int)out);
-  //printf("NOTCOMPILED: addr = %x -> %x\n", (int)addr, (int)out);
+  assem_debug("NOTCOMPILED: addr = %x -> %p\n", addr, out);
   //printf("TRACE: count=%d next=%d (compile %x)\n",Count,next_interupt,addr);
-  //if(debug) 
+  //if(debug)
   //printf("TRACE: count=%d next=%d (checksum %x)\n",Count,next_interupt,mchecksum());
   //printf("fpu mapping=%x enabled=%x\n",(Status & 0x04000000)>>26,(Status & 0x20000000)>>29);
-  /*if(Count>=312978186) {
-    rlist();
-  }*/
-  //rlist();
 
   // this is just for speculation
   for (i = 1; i < 32; i++) {
@@ -7160,16 +7164,15 @@ int new_recompile_block(int addr)
   if (Config.HLE && start == 0x80001000) // hlecall
   {
     // XXX: is this enough? Maybe check hleSoftCall?
-    u_int beginning=(u_int)out;
+    void *beginning=start_block();
     u_int page=get_page(start);
+
     invalid_code[start>>12]=0;
     emit_movimm(start,0);
     emit_writeword(0,(int)&pcaddr);
-    emit_jmp((int)new_dyna_leave);
+    emit_jmp(new_dyna_leave);
     literal_pool(0);
-#ifdef __arm__
-    __clear_cache((void *)beginning,out);
-#endif
+    end_block(beginning);
     ll_add_flags(jump_in+page,start,state_rflags,(void *)beginning);
     return 0;
   }
@@ -7196,7 +7199,7 @@ int new_recompile_block(int addr)
   unsigned int type,op,op2;
 
   //printf("addr = %x source = %x %x\n", addr,source,source[0]);
-  
+
   /* Pass 1 disassembly */
 
   for(i=0;!done;i++) {
@@ -7849,7 +7852,7 @@ int new_recompile_block(int addr)
   /* Pass 2 - Register dependencies and branch targets */
 
   unneeded_registers(0,slen-1,0);
-  
+
   /* Pass 3 - Register allocation */
 
   struct regstat current; // Current register allocations/status
@@ -7876,7 +7879,7 @@ int new_recompile_block(int addr)
     unneeded_reg_upper[0]=1;
     current.regmap[HOST_BTREG]=BTREG;
   }
-  
+
   for(i=0;i<slen;i++)
   {
     if(bt[i])
@@ -7999,7 +8002,7 @@ int new_recompile_block(int addr)
           }
         } else {
           // First instruction expects CCREG to be allocated
-          if(i==0&&hr==HOST_CCREG) 
+          if(i==0&&hr==HOST_CCREG)
             regs[i].regmap_entry[hr]=CCREG;
           else
             regs[i].regmap_entry[hr]=-1;
@@ -8334,7 +8337,7 @@ int new_recompile_block(int addr)
           pagespan_alloc(&current,i);
           break;
       }
-      
+
       // Drop the upper half of registers that have become 32-bit
       current.uu|=current.is32&((1LL<<rt1[i])|(1LL<<rt2[i]));
       if(itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=RJUMP&&itype[i]!=FJUMP) {
@@ -8352,7 +8355,7 @@ int new_recompile_block(int addr)
       // Create entry (branch target) regmap
       for(hr=0;hr<HOST_REGS;hr++)
       {
-        int r,or,er;
+        int r,or;
         r=current.regmap[hr];
         if(r>=0) {
           if(r!=regmap_pre[i][hr]) {
@@ -8395,7 +8398,7 @@ int new_recompile_block(int addr)
           }
         } else {
           // Branches expect CCREG to be allocated at the target
-          if(regmap_pre[i][hr]==CCREG) 
+          if(regmap_pre[i][hr]==CCREG)
             regs[i].regmap_entry[hr]=CCREG;
           else
             regs[i].regmap_entry[hr]=-1;
@@ -8746,11 +8749,11 @@ int new_recompile_block(int addr)
     if(current.regmap[HOST_BTREG]==BTREG) current.regmap[HOST_BTREG]=-1;
     regs[i].waswritten=current.waswritten;
   }
-  
+
   /* Pass 4 - Cull unused host registers */
-  
+
   uint64_t nr=0;
-  
+
   for (i=slen-1;i>=0;i--)
   {
     int hr;
@@ -8892,7 +8895,7 @@ int new_recompile_block(int addr)
     }
     // Save it
     needed_reg[i]=nr;
-    
+
     // Deallocate unneeded registers
     for(hr=0;hr<HOST_REGS;hr++)
     {
@@ -9010,20 +9013,20 @@ int new_recompile_block(int addr)
       }
     }
   }
-  
+
   /* Pass 5 - Pre-allocate registers */
-  
+
   // If a register is allocated during a loop, try to allocate it for the
   // entire loop, if possible.  This avoids loading/storing registers
   // inside of the loop.
-  
+
   signed char f_regmap[HOST_REGS];
   clear_all_regs(f_regmap);
   for(i=0;i<slen-1;i++)
   {
     if(itype[i]==UJUMP||itype[i]==CJUMP||itype[i]==SJUMP||itype[i]==FJUMP)
     {
-      if(ba[i]>=start && ba[i]<(start+i*4)) 
+      if(ba[i]>=start && ba[i]<(start+i*4))
       if(itype[i+1]==NOP||itype[i+1]==MOV||itype[i+1]==ALU
       ||itype[i+1]==SHIFTIMM||itype[i+1]==IMM16||itype[i+1]==LOAD
       ||itype[i+1]==STORE||itype[i+1]==STORELR||itype[i+1]==C1LS
@@ -9071,10 +9074,10 @@ int new_recompile_block(int addr)
             }
           }
           if(ooo[i]) {
-            if(count_free_regs(regs[i].regmap)<=minimum_free_regs[i+1]) 
+            if(count_free_regs(regs[i].regmap)<=minimum_free_regs[i+1])
               f_regmap[hr]=branch_regs[i].regmap[hr];
           }else{
-            if(count_free_regs(branch_regs[i].regmap)<=minimum_free_regs[i+1]) 
+            if(count_free_regs(branch_regs[i].regmap)<=minimum_free_regs[i+1])
               f_regmap[hr]=branch_regs[i].regmap[hr];
           }
           // Avoid dirty->clean transition
@@ -9244,10 +9247,10 @@ int new_recompile_block(int addr)
                 if(itype[j]==CJUMP||itype[j]==SJUMP||itype[j]==FJUMP)
                 {
                   if(ooo[j]) {
-                    if(count_free_regs(regs[j].regmap)<=minimum_free_regs[j+1]) 
+                    if(count_free_regs(regs[j].regmap)<=minimum_free_regs[j+1])
                       break;
                   }else{
-                    if(count_free_regs(branch_regs[j].regmap)<=minimum_free_regs[j+1]) 
+                    if(count_free_regs(branch_regs[j].regmap)<=minimum_free_regs[j+1])
                       break;
                   }
                   if(get_reg(branch_regs[j].regmap,f_regmap[hr])>=0) {
@@ -9320,7 +9323,7 @@ int new_recompile_block(int addr)
             regs[k].isconst&=~(1<<HOST_CCREG);
             k++;
           }
-          regs[j].regmap_entry[HOST_CCREG]=CCREG;          
+          regs[j].regmap_entry[HOST_CCREG]=CCREG;
         }
         // Work backwards from the branch target
         if(j>i&&f_regmap[HOST_CCREG]==CCREG)
@@ -9362,7 +9365,7 @@ int new_recompile_block(int addr)
       }
     }
   }
-  
+
   // Cache memory offset or tlb map pointer if a register is available
   #ifndef HOST_IMM_ADDR32
   #ifndef RAM_OFFSET
@@ -9542,7 +9545,7 @@ int new_recompile_block(int addr)
     }
   }
   #endif
-  
+
   // This allocates registers (if possible) one instruction prior
   // to use, which can avoid a load-use penalty on certain CPUs.
   for(i=0;i<slen-1;i++)
@@ -9603,7 +9606,7 @@ int new_recompile_block(int addr)
               }
             }
           }
-          // Load source into target register 
+          // Load source into target register
           if(lt1[i+1]&&get_reg(regs[i+1].regmap,rs1[i+1])<0) {
             if((hr=get_reg(regs[i+1].regmap,rt1[i+1]))>=0)
             {
@@ -9677,7 +9680,7 @@ int new_recompile_block(int addr)
             }
           }
           if(itype[i+1]==LOAD||itype[i+1]==LOADLR||itype[i+1]==STORE||itype[i+1]==STORELR/*||itype[i+1]==C1LS||||itype[i+1]==C2LS*/) {
-            if(itype[i+1]==LOAD) 
+            if(itype[i+1]==LOAD)
               hr=get_reg(regs[i+1].regmap,rt1[i+1]);
             if(itype[i+1]==LOADLR||(opcode[i+1]&0x3b)==0x31||(opcode[i+1]&0x3b)==0x32) // LWC1/LDC1, LWC2/LDC2
               hr=get_reg(regs[i+1].regmap,FTEMP);
@@ -9701,10 +9704,10 @@ int new_recompile_block(int addr)
       }
     }
   }
-  
+
   /* Pass 6 - Optimize clean/dirty state */
   clean_registers(0,slen-1,1);
-  
+
   /* Pass 7 - Identify 32-bit registers */
   for (i=slen-1;i>=0;i--)
   {
@@ -9871,17 +9874,17 @@ int new_recompile_block(int addr)
   cop1_usable=0;
   uint64_t is32_pre=0;
   u_int dirty_pre=0;
-  u_int beginning=(u_int)out;
+  void *beginning=start_block();
   if((u_int)addr&1) {
     ds=1;
     pagespan_ds();
   }
-  u_int instr_addr0_override=0;
+  void *instr_addr0_override = NULL;
 
   if (start == 0x80030000) {
     // nasty hack for fastbios thing
     // override block entry to this code
-    instr_addr0_override=(u_int)out;
+    instr_addr0_override = out;
     emit_movimm(start,0);
     // abuse io address var as a flag that we
     // have already returned here once
@@ -9898,7 +9901,7 @@ int new_recompile_block(int addr)
     if(ds) {
       ds=0; // Skip delay slot
       if(bt[i]) assem_debug("OOPS - branch into delay slot\n");
-      instr_addr[i]=0;
+      instr_addr[i] = NULL;
     } else {
       speculate_register_values(i);
       #ifndef DESTRUCTIVE_WRITEBACK
@@ -9923,8 +9926,10 @@ int new_recompile_block(int addr)
         loop_preload(regmap_pre[i],regs[i].regmap_entry);
       }
       // branch target entry point
-      instr_addr[i]=(u_int)out;
+      instr_addr[i] = out;
       assem_debug("<->\n");
+      drc_dbg_emit_do_cmp(i);
+
       // load regs
       if(regs[i].regmap_entry[HOST_CCREG]==CCREG&&regs[i].regmap[HOST_CCREG]!=CCREG)
         wb_register(CCREG,regs[i].regmap_entry,regs[i].wasdirty,regs[i].was32);
@@ -10065,7 +10070,7 @@ int new_recompile_block(int addr)
   // Stubs
   for(i=0;i<stubcount;i++)
   {
-    switch(stubs[i][0])
+    switch(stubs[i].type)
     {
       case LOADB_STUB:
       case LOADH_STUB:
@@ -10104,10 +10109,10 @@ int new_recompile_block(int addr)
       void *addr=check_addr(link_addr[i][1]);
       emit_extjump(link_addr[i][0],link_addr[i][1]);
       if(addr) {
-        set_jump_target(link_addr[i][0],(int)addr);
+        set_jump_target(link_addr[i][0], addr);
         add_link(link_addr[i][1],stub);
       }
-      else set_jump_target(link_addr[i][0],(int)stub);
+      else set_jump_target(link_addr[i][0], stub);
     }
     else
     {
@@ -10135,22 +10140,20 @@ int new_recompile_block(int addr)
         u_int vpage=get_vpage(vaddr);
         literal_pool(256);
         {
-          assem_debug("%8x (%d) <- %8x\n",instr_addr[i],i,start+i*4);
+          assem_debug("%p (%d) <- %8x\n",instr_addr[i],i,start+i*4);
           assem_debug("jump_in: %x\n",start+i*4);
-          ll_add(jump_dirty+vpage,vaddr,(void *)out);
-          int entry_point=do_dirty_stub(i);
-          ll_add_flags(jump_in+page,vaddr,state_rflags,(void *)entry_point);
+          ll_add(jump_dirty+vpage,vaddr,out);
+          void *entry_point = do_dirty_stub(i);
+          ll_add_flags(jump_in+page,vaddr,state_rflags,entry_point);
           // If there was an existing entry in the hash table,
           // replace it with the new address.
           // Don't add new entries.  We'll insert the
           // ones that actually get used in check_addr().
-          int *ht_bin=hash_table[((vaddr>>16)^vaddr)&0xFFFF];
-          if(ht_bin[0]==vaddr) {
-            ht_bin[1]=entry_point;
-          }
-          if(ht_bin[2]==vaddr) {
-            ht_bin[3]=entry_point;
-          }
+          struct ht_entry *ht_bin = hash_table_get(vaddr);
+          if (ht_bin->vaddr[0] == vaddr)
+            ht_bin->tcaddr[0] = entry_point;
+          if (ht_bin->vaddr[1] == vaddr)
+            ht_bin->tcaddr[1] = entry_point;
         }
       }
     }
@@ -10161,19 +10164,17 @@ int new_recompile_block(int addr)
   // Align code
   if(((u_int)out)&7) emit_addnop(13);
   #endif
-  assert((u_int)out-beginning<MAX_OUTPUT_BLOCK_SIZE);
+  assert((u_int)out-(u_int)beginning<MAX_OUTPUT_BLOCK_SIZE);
   //printf("shadow buffer: %x-%x\n",(int)copy,(int)copy+slen*4);
   memcpy(copy,source,slen*4);
   copy+=slen*4;
-  
-  #ifdef __arm__
-  __clear_cache((void *)beginning,out);
-  #endif
-  
+
+  end_block(beginning);
+
   // If we're within 256K of the end of the buffer,
   // start over from the beginning. (Is 256K enough?)
   if((u_int)out>(u_int)BASE_ADDR+(1<<TARGET_SIZE_2)-MAX_OUTPUT_BLOCK_SIZE) out=(u_char *)BASE_ADDR;
-  
+
   // Trap writes to any of the pages we compiled
   for(i=start>>12;i<=(start+slen*4)>>12;i++) {
     invalid_code[i]=0;
@@ -10186,14 +10187,14 @@ int new_recompile_block(int addr)
       invalid_code[((u_int)0x00000000>>12)|(i&0x1ff)]=
       invalid_code[((u_int)0x80000000>>12)|(i&0x1ff)]=
       invalid_code[((u_int)0xa0000000>>12)|(i&0x1ff)]=0;
-  
+
   /* Pass 10 - Free memory by expiring oldest blocks */
-  
+
   int end=((((int)out-(int)BASE_ADDR)>>(TARGET_SIZE_2-16))+16384)&65535;
   while(expirep!=end)
   {
     int shift=TARGET_SIZE_2-3; // Divide into 8 blocks
-    int base=(int)BASE_ADDR+((expirep>>13)<<shift); // Base address of this block
+    uintptr_t base=(uintptr_t)BASE_ADDR+((expirep>>13)<<shift); // Base address of this block
     inv_debug("EXP: Phase %d\n",expirep);
     switch((expirep>>11)&3)
     {
@@ -10212,25 +10213,27 @@ int new_recompile_block(int addr)
       case 2:
         // Clear hash table
         for(i=0;i<32;i++) {
-          int *ht_bin=hash_table[((expirep&2047)<<5)+i];
-          if((ht_bin[3]>>shift)==(base>>shift) ||
-             ((ht_bin[3]-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(base>>shift)) {
-            inv_debug("EXP: Remove hash %x -> %x\n",ht_bin[2],ht_bin[3]);
-            ht_bin[2]=ht_bin[3]=-1;
-          }
-          if((ht_bin[1]>>shift)==(base>>shift) ||
-             ((ht_bin[1]-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(base>>shift)) {
-            inv_debug("EXP: Remove hash %x -> %x\n",ht_bin[0],ht_bin[1]);
-            ht_bin[0]=ht_bin[2];
-            ht_bin[1]=ht_bin[3];
-            ht_bin[2]=ht_bin[3]=-1;
+          struct ht_entry *ht_bin = &hash_table[((expirep&2047)<<5)+i];
+          if (((uintptr_t)ht_bin->tcaddr[1]>>shift) == (base>>shift) ||
+             (((uintptr_t)ht_bin->tcaddr[1]-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(base>>shift)) {
+            inv_debug("EXP: Remove hash %x -> %p\n",ht_bin->vaddr[1],ht_bin->tcaddr[1]);
+            ht_bin->vaddr[1] = -1;
+            ht_bin->tcaddr[1] = NULL;
+          }
+          if (((uintptr_t)ht_bin->tcaddr[0]>>shift) == (base>>shift) ||
+             (((uintptr_t)ht_bin->tcaddr[0]-MAX_OUTPUT_BLOCK_SIZE)>>shift)==(base>>shift)) {
+            inv_debug("EXP: Remove hash %x -> %p\n",ht_bin->vaddr[0],ht_bin->tcaddr[0]);
+            ht_bin->vaddr[0] = ht_bin->vaddr[1];
+            ht_bin->tcaddr[0] = ht_bin->tcaddr[1];
+            ht_bin->vaddr[1] = -1;
+            ht_bin->tcaddr[1] = NULL;
           }
         }
         break;
       case 3:
         // Clear jump_out
         #ifdef __arm__
-        if((expirep&2047)==0) 
+        if((expirep&2047)==0)
           do_clear_cache();
         #endif
         ll_remove_matching_addrs(jump_out+(expirep&2047),base,shift);