partially revert "(VITA) Some dynarec"

See github issue #74 for details.

diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c
index 1c0ab56..65c8f03 100644 (file)
--- a/libpcsxcore/new_dynarec/new_dynarec.c
+++ b/libpcsxcore/new_dynarec/new_dynarec.c
@@ -56,13 +56,6 @@ static void __clear_cache(void *start, void *end) {
 #elif defined(_3DS)
 #include "3ds_utils.h"
 #define __clear_cache(start,end) svcFlushProcessDataCache(0xFFFF8001, start, (u32)(end)-(u32)(start))
-#elif defined(VITA)
-#define __clear_cache vita_clear_cache
-static void __clear_cache(void *start, void *end) {
-  size_t len = (char *)end - (char *)start;
-  int block = sceKernelFindMemBlockByAddr(start,len);
-  sceKernelSyncVMDomain(block, start, len);
-}
 #endif
 
 #define MAXBLOCK 4096
@@ -203,7 +196,7 @@ 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
@@ -321,18 +314,18 @@ static void tlb_hacks()
   {
     u_int addr;
     int n;
-    switch (ROM_HEADER->Country_code&0xFF)
+    switch (ROM_HEADER->Country_code&0xFF) 
     {
       case 0x45: // U
         addr=0x34b30;
-        break;
-      case 0x4A: // J
-        addr=0x34b70;
-        break;
-      case 0x50: // E
+        break;                   
+      case 0x4A: // J 
+        addr=0x34b70;    
+        break;    
+      case 0x50: // E 
         addr=0x329f0;
-        break;
-      default:
+        break;                        
+      default: 
         // Unknown country code
         addr=0;
         break;
@@ -533,7 +526,7 @@ 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;
     }
   }
@@ -690,7 +683,7 @@ 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)
@@ -793,7 +786,7 @@ 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])&&
@@ -834,7 +827,7 @@ void mult64(uint64_t m1,uint64_t m2)
    unsigned long long int result1, result2, result3, result4;
    unsigned long long int temp1, temp2, temp3, temp4;
    int sign = 0;
-
+   
    if (m1 < 0)
      {
     op2 = -m1;
@@ -847,22 +840,22 @@ void mult64(uint64_t m1,uint64_t m2)
     sign = 1 - sign;
      }
    else op4 = m2;
-
+   
    op1 = op2 & 0xFFFFFFFF;
    op2 = (op2 >> 32) & 0xFFFFFFFF;
    op3 = op4 & 0xFFFFFFFF;
    op4 = (op4 >> 32) & 0xFFFFFFFF;
-
+   
    temp1 = op1 * op3;
    temp2 = (temp1 >> 32) + op1 * op4;
    temp3 = op2 * op3;
    temp4 = (temp3 >> 32) + op2 * op4;
-
+   
    result1 = temp1 & 0xFFFFFFFF;
    result2 = temp2 + (temp3 & 0xFFFFFFFF);
    result3 = (result2 >> 32) + temp4;
    result4 = (result3 >> 32);
-
+   
    lo = result1 | (result2 << 32);
    hi = (result3 & 0xFFFFFFFF) | (result4 << 32);
    if (sign)
@@ -878,25 +871,25 @@ void multu64(uint64_t m1,uint64_t m2)
    unsigned long long int op1, op2, op3, op4;
    unsigned long long int result1, result2, result3, result4;
    unsigned long long int temp1, temp2, temp3, temp4;
-
+   
    op1 = m1 & 0xFFFFFFFF;
    op2 = (m1 >> 32) & 0xFFFFFFFF;
    op3 = m2 & 0xFFFFFFFF;
    op4 = (m2 >> 32) & 0xFFFFFFFF;
-
+   
    temp1 = op1 * op3;
    temp2 = (temp1 >> 32) + op1 * op4;
    temp3 = op2 * op3;
    temp4 = (temp3 >> 32) + op2 * op4;
-
+   
    result1 = temp1 & 0xFFFFFFFF;
    result2 = temp2 + (temp3 & 0xFFFFFFFF);
    result3 = (result2 >> 32) + temp4;
    result4 = (result3 >> 32);
-
+   
    lo = result1 | (result2 << 32);
    hi = (result3 & 0xFFFFFFFF) | (result4 << 32);
-
+   
   //printf("TRACE: dmultu %8x%8x %8x%8x\n",(int)reg[HIREG],(int)(reg[HIREG]>>32)
   //                                      ,(int)reg[LOREG],(int)(reg[LOREG]>>32));
 }
@@ -1018,7 +1011,7 @@ 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);
@@ -1114,7 +1107,7 @@ 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;
 #ifndef DISABLE_TLB
@@ -3393,7 +3386,7 @@ void storelr_assemble(int i,struct regstat *i_regs)
     if(map<0) emit_loadreg(ROREG,map=HOST_TEMPREG);
     gen_tlb_addr_w(temp,map);
     #else
-    if((u_int)rdram!=0x80000000)
+    if((u_int)rdram!=0x80000000) 
       emit_addimm_no_flags((u_int)rdram-(u_int)0x80000000,temp);
     #endif
   }else{ // using tlb
@@ -4166,7 +4159,7 @@ void address_generation(int i,struct regstat *i_regs,signed char entry[])
     int mgr=MGEN1+(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) {
@@ -4764,7 +4757,7 @@ 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)
@@ -5010,7 +5003,7 @@ 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);
@@ -5201,7 +5194,7 @@ void do_ccstub(int n)
     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);
@@ -5235,7 +5228,7 @@ 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++;
 }
 
@@ -5261,7 +5254,7 @@ 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);
       }
@@ -5282,10 +5275,10 @@ void ujump_assemble(int i,struct regstat *i_regs)
   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) 
   {
     int return_address=start+i*4+8;
-    if(get_reg(branch_regs[i].regmap,31)>0)
+    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);
   }
   #endif
@@ -5336,7 +5329,7 @@ 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);
   }
@@ -5365,7 +5358,7 @@ 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) {
       int return_address=start+i*4+8;
@@ -5499,7 +5492,7 @@ 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);
@@ -5554,7 +5547,7 @@ 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");
@@ -5626,7 +5619,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
           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
@@ -5756,7 +5749,7 @@ void cjump_assemble(int i,struct regstat *i_regs)
           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
@@ -5939,7 +5932,7 @@ void sjump_assemble(int i,struct regstat *i_regs)
     }
     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");
@@ -6026,7 +6019,7 @@ 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])) {
@@ -6274,7 +6267,7 @@ 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
@@ -6782,14 +6775,14 @@ 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
@@ -7052,7 +7045,7 @@ static void provisional_32bit()
   int i,j;
   uint64_t is32=1;
   uint64_t lastbranch=1;
-
+  
   for(i=0;i<slen;i++)
   {
     if(i>0) {
@@ -7089,13 +7082,13 @@ static void provisional_32bit()
       uint64_t temp_is32=is32;
       for(j=i-1;j>=0;j--)
       {
-        if(ba[j]==start+i*4)
+        if(ba[j]==start+i*4) 
           //temp_is32&=branch_regs[j].is32;
           temp_is32&=p32[j];
       }
       for(j=i;j<slen;j++)
       {
-        if(ba[j]==start+i*4)
+        if(ba[j]==start+i*4) 
           temp_is32=1;
       }
       is32=temp_is32;
@@ -7282,7 +7275,7 @@ static void provisional_r32()
 {
   u_int r32=0;
   int i;
-
+  
   for (i=slen-1;i>=0;i--)
   {
     int hr;
@@ -7380,7 +7373,7 @@ static void provisional_r32()
     }
     //requires_32bit[i]=r32;
     pr32[i]=r32;
-
+    
     // Dirty registers which are 32-bit, require 32-bit input
     // as they will be written as 32-bit values
     for(hr=0;hr<HOST_REGS;hr++)
@@ -7733,7 +7726,7 @@ 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.
@@ -7981,28 +7974,17 @@ void new_dynarec_init()
 {
   SysPrintf("Init new dynarec\n");
   out=(u_char *)BASE_ADDR;
-#if defined(VITA)
-
+#if BASE_ADDR_FIXED
   if (mmap (out, 1<<TARGET_SIZE_2,
-            0,
-            0,
+            PROT_READ | PROT_WRITE | PROT_EXEC,
+            MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS,
             -1, 0) <= 0) {
     SysPrintf("mmap() failed: %s\n", strerror(errno));
   }
-
 #else
-  #if BASE_ADDR_FIXED
-    if (mmap (out, 1<<TARGET_SIZE_2,
-              PROT_READ | PROT_WRITE | PROT_EXEC,
-              MAP_FIXED | MAP_PRIVATE | MAP_ANONYMOUS,
-              -1, 0) <= 0) {
-      SysPrintf("mmap() failed: %s\n", strerror(errno));
-    }
-  #else
-    // not all systems allow execute in data segment by default
-    if (mprotect(out, 1<<TARGET_SIZE_2, PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
-      SysPrintf("mprotect() failed: %s\n", strerror(errno));
-  #endif
+  // not all systems allow execute in data segment by default
+  if (mprotect(out, 1<<TARGET_SIZE_2, PROT_READ | PROT_WRITE | PROT_EXEC) != 0)
+    SysPrintf("mprotect() failed: %s\n", strerror(errno));
 #endif
 #ifdef MUPEN64
   rdword=&readmem_dword;
@@ -8208,7 +8190,7 @@ int new_recompile_block(int addr)
   assem_debug("NOTCOMPILED: addr = %x -> %x\n", (int)addr, (int)out);
   //printf("NOTCOMPILED: addr = %x -> %x\n", (int)addr, (int)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) {
@@ -8264,7 +8246,7 @@ 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++) {
@@ -8924,7 +8906,7 @@ 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
@@ -8954,7 +8936,7 @@ 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])
@@ -8995,12 +8977,12 @@ int new_recompile_block(int addr)
       uint64_t temp_is32=current.is32;
       for(j=i-1;j>=0;j--)
       {
-        if(ba[j]==start+i*4)
+        if(ba[j]==start+i*4) 
           temp_is32&=branch_regs[j].is32;
       }
       for(j=i;j<slen;j++)
       {
-        if(ba[j]==start+i*4)
+        if(ba[j]==start+i*4) 
           //temp_is32=1;
           temp_is32&=p32[j];
       }
@@ -9040,12 +9022,12 @@ int new_recompile_block(int addr)
       uint64_t temp_is32=current.is32;
       for(j=i-1;j>=0;j--)
       {
-        if(ba[j]==start+i*4+4)
+        if(ba[j]==start+i*4+4) 
           temp_is32&=branch_regs[j].is32;
       }
       for(j=i;j<slen;j++)
       {
-        if(ba[j]==start+i*4+4)
+        if(ba[j]==start+i*4+4) 
           //temp_is32=1;
           temp_is32&=p32[j];
       }
@@ -9063,7 +9045,7 @@ int new_recompile_block(int addr)
                 {
                   //printf("dump %d/r%d\n",hr,r);
                   current.regmap[hr]=-1;
-                  if(get_reg(current.regmap,r|64)>=0)
+                  if(get_reg(current.regmap,r|64)>=0) 
                     current.regmap[get_reg(current.regmap,r|64)]=-1;
                 }
               }
@@ -9077,12 +9059,12 @@ int new_recompile_block(int addr)
       uint64_t temp_is32=current.is32;
       for(j=i-1;j>=0;j--)
       {
-        if(ba[j]==start+i*4+8)
+        if(ba[j]==start+i*4+8) 
           temp_is32&=branch_regs[j].is32;
       }
       for(j=i;j<slen;j++)
       {
-        if(ba[j]==start+i*4+8)
+        if(ba[j]==start+i*4+8) 
           //temp_is32=1;
           temp_is32&=p32[j];
       }
@@ -9098,7 +9080,7 @@ int new_recompile_block(int addr)
               {
                 //printf("dump %d/r%d\n",hr,r);
                 current.regmap[hr]=-1;
-                if(get_reg(current.regmap,r|64)>=0)
+                if(get_reg(current.regmap,r|64)>=0) 
                   current.regmap[get_reg(current.regmap,r|64)]=-1;
               }
             }
@@ -9190,7 +9172,7 @@ 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;
@@ -9525,7 +9507,7 @@ 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) {
@@ -9586,7 +9568,7 @@ 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;
@@ -9937,11 +9919,11 @@ 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;
@@ -10083,7 +10065,7 @@ int new_recompile_block(int addr)
     }
     // Save it
     needed_reg[i]=nr;
-
+    
     // Deallocate unneeded registers
     for(hr=0;hr<HOST_REGS;hr++)
     {
@@ -10212,20 +10194,20 @@ 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
@@ -10273,10 +10255,10 @@ 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
@@ -10446,10 +10428,10 @@ 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) {
@@ -10522,7 +10504,7 @@ 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)
@@ -10564,7 +10546,7 @@ 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
@@ -10744,7 +10726,7 @@ 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++)
@@ -10805,7 +10787,7 @@ 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)
             {
@@ -10919,7 +10901,7 @@ 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);
@@ -10943,16 +10925,16 @@ int new_recompile_block(int addr)
       }
     }
   }
-
+  
   /* Pass 6 - Optimize clean/dirty state */
   clean_registers(0,slen-1,1);
-
+  
   /* Pass 7 - Identify 32-bit registers */
 #ifndef FORCE32
   provisional_r32();
 
   u_int r32=0;
-
+  
   for (i=slen-1;i>=0;i--)
   {
     int hr;
@@ -11048,7 +11030,7 @@ int new_recompile_block(int addr)
       if((regs[i].was32>>dep2[i])&1) r32|=1LL<<dep2[i];
     }
     requires_32bit[i]=r32;
-
+    
     // Dirty registers which are 32-bit, require 32-bit input
     // as they will be written as 32-bit values
     for(hr=0;hr<HOST_REGS;hr++)
@@ -11604,15 +11586,15 @@ int new_recompile_block(int addr)
   //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
-
+  
   // 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;
@@ -11636,9 +11618,9 @@ int new_recompile_block(int addr)
       invalid_code[((u_int)0x80000000>>12)|(i&0x1ff)]=
       invalid_code[((u_int)0xa0000000>>12)|(i&0x1ff)]=0;
 #endif
-
+  
   /* Pass 10 - Free memory by expiring oldest blocks */
-
+  
   int end=((((int)out-(int)BASE_ADDR)>>(TARGET_SIZE_2-16))+16384)&65535;
   while(expirep!=end)
   {
@@ -11680,7 +11662,7 @@ int new_recompile_block(int addr)
       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);