-diff --git a/libpcsxcore/new_dynarec/linkage_arm.S b/libpcsxcore/new_dynarec/linkage_arm.S
-index bbc52c3..83c5b08 100644
---- a/libpcsxcore/new_dynarec/linkage_arm.S
-+++ b/libpcsxcore/new_dynarec/linkage_arm.S
-@@ -437,7 +437,7 @@ FUNCTION(cc_interrupt):
- str r1, [fp, #LO_pending_exception]
- and r2, r2, r10, lsr #17
- add r3, fp, #LO_restore_candidate
-- str r10, [fp, #LO_cycle] /* PCSX cycles */
-+@@@ str r10, [fp, #LO_cycle] /* PCSX cycles */
- @@ str r10, [fp, #LO_reg_cop0+36] /* Count */
- ldr r4, [r2, r3]
- mov r10, lr
-@@ -527,7 +527,7 @@ FUNCTION(new_dyna_leave):
- ldr r0, [fp, #LO_last_count]
- add r12, fp, #28
- add r10, r0, r10
-- str r10, [fp, #LO_cycle]
-+@@@ str r10, [fp, #LO_cycle]
- ldmfd sp!, {r4, r5, r6, r7, r8, r9, sl, fp, ip, pc}
- .size new_dyna_leave, .-new_dyna_leave
-
-@@ -644,7 +644,7 @@ FUNCTION(new_dyna_start):
- \readop r0, [r1, r3, lsl #\tab_shift]
- .endif
- movcc pc, lr
-- str r2, [fp, #LO_cycle]
-+@@@ str r2, [fp, #LO_cycle]
- bx r1
- .endm
-
-@@ -679,7 +679,7 @@ FUNCTION(jump_handler_read32):
- mov r0, r1
- add r2, r2, r12
- push {r2, lr}
-- str r2, [fp, #LO_cycle]
-+@@@ str r2, [fp, #LO_cycle]
- blx r3
-
- ldr r0, [fp, #LO_next_interupt]
-@@ -707,7 +707,7 @@ FUNCTION(jump_handler_write_h):
- add r2, r2, r12
- mov r0, r1
- push {r2, lr}
-- str r2, [fp, #LO_cycle]
-+@@@ str r2, [fp, #LO_cycle]
- blx r3
-
- ldr r0, [fp, #LO_next_interupt]
-diff --git a/libpcsxcore/new_dynarec/linkage_arm64.S b/libpcsxcore/new_dynarec/linkage_arm64.S
-index 698bd78..798abea 100644
---- a/libpcsxcore/new_dynarec/linkage_arm64.S
-+++ b/libpcsxcore/new_dynarec/linkage_arm64.S
-@@ -119,7 +119,7 @@ FUNCTION(cc_interrupt):
- str wzr, [rFP, #LO_pending_exception]
- and w2, w2, rCC, lsr #17
- add x3, rFP, #LO_restore_candidate
-- str rCC, [rFP, #LO_cycle] /* PCSX cycles */
-+## str rCC, [rFP, #LO_cycle] /* PCSX cycles */
- # str rCC, [rFP, #LO_reg_cop0+36] /* Count */
- ldr w19, [x3, w2, uxtw]
- mov x21, lr
-@@ -227,7 +227,7 @@ FUNCTION(new_dyna_start):
- FUNCTION(new_dyna_leave):
- ldr w0, [rFP, #LO_last_count]
- add rCC, rCC, w0
-- str rCC, [rFP, #LO_cycle]
-+## str rCC, [rFP, #LO_cycle]
- ldp x19, x20, [sp, #16*1]
- ldp x21, x22, [sp, #16*2]
- ldp x23, x24, [sp, #16*3]
-@@ -245,7 +245,7 @@ FUNCTION(new_dyna_leave):
- /* w0 = adddr/data, x1 = rhandler, w2 = cycles, x3 = whandler */
- ldr w4, [rFP, #LO_last_count]
- add w4, w4, w2
-- str w4, [rFP, #LO_cycle]
-+## str w4, [rFP, #LO_cycle]
- .endm
-
- .macro memhandler_post
diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c
-index 1452db3..8200e44 100644
+index b160a4a..0d91999 100644
--- a/libpcsxcore/new_dynarec/new_dynarec.c
+++ b/libpcsxcore/new_dynarec/new_dynarec.c
-@@ -44,10 +44,10 @@ static int sceBlock;
- #define ARRAY_SIZE(x) (sizeof(x) / sizeof(x[0]))
- #endif
+@@ -285,7 +285,7 @@ static struct decoded_insn
+ int new_dynarec_hacks_old;
+ int new_dynarec_did_compile;
--//#define DISASM
--//#define assem_debug printf
-+#define DISASM
-+#define assem_debug printf
- //#define inv_debug printf
--#define assem_debug(...)
-+//#define assem_debug(...)
- #define inv_debug(...)
+- #define HACK_ENABLED(x) ((new_dynarec_hacks | new_dynarec_hacks_pergame) & (x))
++ #define HACK_ENABLED(x) ((NDHACK_NO_STALLS) & (x))
- #ifdef __i386__
-@@ -423,6 +423,9 @@ static int doesnt_expire_soon(void *tcaddr)
+ extern int cycle_count; // ... until end of the timeslice, counts -N -> 0
+ extern int last_count; // last absolute target, often = next_interupt
+@@ -532,6 +532,7 @@ static int cycle_multiplier_active;
+
+ static int CLOCK_ADJUST(int x)
+ {
++ return x * 2;
+ int m = cycle_multiplier_active;
+ int s = (x >> 31) | 1;
+ return (x * m + s * 50) / 100;
+@@ -662,6 +663,9 @@ static void *try_restore_block(u_int vaddr, u_int start_page, u_int end_page)
// This is called from the recompiled JR/JALR instructions
- void noinline *get_addr(u_int vaddr)
+ static void noinline *get_addr(u_int vaddr, int can_compile)
{
+#ifdef DRC_DBG
+printf("get_addr %08x, pc=%08x\n", vaddr, psxRegs.pc);
+#endif
- u_int page=get_page(vaddr);
- u_int vpage=get_vpage(vaddr);
- struct ll_entry *head;
-@@ -4393,13 +4396,15 @@ void do_cc(int i,signed char i_regmap[],int *adj,int addr,int taken,int invert)
- }
- emit_addimm_and_set_flags(cycles,HOST_CCREG);
- jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- }
- else
- {
- emit_cmpimm(HOST_CCREG,-CLOCK_ADJUST(count+2));
- jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- }
- add_stub(CC_STUB,jaddr,idle?idle:out,(*adj==0||invert||idle)?0:(count+2),i,addr,taken,0);
- }
-@@ -4807,7 +4812,8 @@ static void rjump_assemble(int i,struct regstat *i_regs)
- // special case for RFE
- emit_jmp(0);
- else
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- //load_regs_bt(branch_regs[i].regmap,branch_regs[i].dirty,-1);
- #ifdef USE_MINI_HT
- if(rs1[i]==31) {
-@@ -4912,7 +4918,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
- else if(nop) {
- emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
- void *jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
- }
- else {
-@@ -5099,7 +5106,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
- emit_loadreg(CCREG,HOST_CCREG);
- emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
- void *jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
- emit_storereg(CCREG,HOST_CCREG);
- }
-@@ -5108,7 +5116,8 @@ static void cjump_assemble(int i,struct regstat *i_regs)
- assert(cc==HOST_CCREG);
- emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
- void *jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
- }
- }
-@@ -5210,7 +5219,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
- else if(nevertaken) {
- emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
- void *jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
- }
- else {
-@@ -5366,7 +5376,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
- emit_loadreg(CCREG,HOST_CCREG);
- emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),HOST_CCREG);
- void *jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,NOTTAKEN,0);
- emit_storereg(CCREG,HOST_CCREG);
- }
-@@ -5375,7 +5386,8 @@ static void sjump_assemble(int i,struct regstat *i_regs)
- assert(cc==HOST_CCREG);
- emit_addimm_and_set_flags(CLOCK_ADJUST(ccadj[i]+2),cc);
- void *jaddr=out;
-- emit_jns(0);
-+ //emit_jns(0);
-+ emit_jmp(0);
- add_stub(CC_STUB,jaddr,out,0,i,start+i*4+8,likely[i]?NULLDS:NOTTAKEN,0);
- }
- }
-@@ -5863,7 +5875,7 @@ void unneeded_registers(int istart,int iend,int r)
+ u_int start_page = get_page_prev(vaddr);
+ u_int i, page, end_page = get_page(vaddr);
+ void *found_clean = NULL;
+@@ -7046,7 +7050,7 @@ static noinline void pass2_unneeded_regs(int istart,int iend,int r)
// R0 is always unneeded
u|=1;
// Save it
gte_unneeded[i]=gte_u;
/*
printf("ur (%d,%d) %x: ",istart,iend,start+i*4);
-@@ -8387,6 +8399,7 @@ int new_recompile_block(int addr)
-
- // This allocates registers (if possible) one instruction prior
- // to use, which can avoid a load-use penalty on certain CPUs.
-+#if 0
+@@ -8236,6 +8240,7 @@ static noinline void pass5a_preallocate1(void)
+ static noinline void pass5b_preallocate2(void)
+ {
+ int i, hr;
++ return;
for(i=0;i<slen-1;i++)
{
- if(!i||(itype[i-1]!=UJUMP&&itype[i-1]!=CJUMP&&itype[i-1]!=SJUMP&&itype[i-1]!=RJUMP))
-@@ -8543,6 +8556,7 @@ int new_recompile_block(int addr)
- }
- }
- }
-+#endif
+ if (!i || !dops[i-1].is_jump)
+@@ -9045,6 +9050,11 @@ static int new_recompile_block(u_int addr)
+ load_reg(regs[i].regmap_entry,regs[i].regmap,INVCP);
- /* Pass 6 - Optimize clean/dirty state */
- clean_registers(0,slen-1,1);
-@@ -8842,6 +8856,11 @@ int new_recompile_block(int addr)
- case SPAN:
- pagespan_assemble(i,®s[i]);break;
- }
-+
+ ds = assemble(i, ®s[i], ccadj[i]);
+#ifdef DRC_DBG
-+ if(itype[i]!=UJUMP&&itype[i]!=CJUMP&&itype[i]!=SJUMP&&itype[i]!=RJUMP)
-+ wb_dirtys(regs[i].regmap,regs[i].dirty);
++ // write-out non-consts, consts are likely different because of get_final_value()
++ if (!dops[i].is_jump)
++ wb_dirtys(regs[i].regmap,regs[i].dirty&~regs[i].loadedconst);
+#endif
- if(itype[i]==UJUMP||itype[i]==RJUMP||(source[i]>>16)==0x1000)
+
+ if (dops[i].is_ujump)
literal_pool(1024);
- else
-@@ -8950,7 +8969,7 @@ int new_recompile_block(int addr)
- }
- }
- // External Branch Targets (jump_in)
-- if(copy+slen*4>(void *)shadow+sizeof(shadow)) copy=shadow;
-+ if(copy+slen*4>(void *)shadow+sizeof(shadow)) {copy=shadow;printf("shadow overflow\n");}
- for(i=0;i<slen;i++)
- {
- if(bt[i]||i==0)
-@@ -9065,6 +9084,10 @@ int new_recompile_block(int addr)
- }
- expirep=(expirep+1)&65535;
- }
+@@ -9236,6 +9246,10 @@ static int new_recompile_block(u_int addr)
+
+ #ifdef ASSEM_PRINT
+ fflush(stdout);
++#endif
+#ifdef DRC_DBG
+printf("new_recompile_block done\n");
+fflush(stdout);
-+#endif
+ #endif
+ stat_inc(stat_bc_direct);
return 0;
- }
+diff --git a/libpcsxcore/new_dynarec/pcsxmem.c b/libpcsxcore/new_dynarec/pcsxmem.c
+index bb471b6..8f68a3b 100644
+--- a/libpcsxcore/new_dynarec/pcsxmem.c
++++ b/libpcsxcore/new_dynarec/pcsxmem.c
+@@ -272,6 +272,8 @@ static void write_biu(u32 value)
+ if (address != 0xfffe0130)
+ return;
+
++extern u32 handler_cycle;
++handler_cycle = psxRegs.cycle;
+ switch (value) {
+ case 0x800: case 0x804:
+ unmap_ram_write();
+diff --git a/libpcsxcore/psxcounters.c b/libpcsxcore/psxcounters.c
+index ff0efbc..4459644 100644
+--- a/libpcsxcore/psxcounters.c
++++ b/libpcsxcore/psxcounters.c
+@@ -379,9 +379,12 @@ void psxRcntUpdate()
+
+ /******************************************************************************/
+
++extern u32 handler_cycle;
++
+ void psxRcntWcount( u32 index, u32 value )
+ {
+ verboseLog( 2, "[RCNT %i] wcount: %x\n", index, value );
++handler_cycle = psxRegs.cycle;
+
+ _psxRcntWcount( index, value );
+ psxRcntSet();
+@@ -390,6 +393,7 @@ void psxRcntWcount( u32 index, u32 value )
+ void psxRcntWmode( u32 index, u32 value )
+ {
+ verboseLog( 1, "[RCNT %i] wmode: %x\n", index, value );
++handler_cycle = psxRegs.cycle;
+
+ _psxRcntWmode( index, value );
+ _psxRcntWcount( index, 0 );
+@@ -401,6 +405,7 @@ void psxRcntWmode( u32 index, u32 value )
+ void psxRcntWtarget( u32 index, u32 value )
+ {
+ verboseLog( 1, "[RCNT %i] wtarget: %x\n", index, value );
++handler_cycle = psxRegs.cycle;
+
+ rcnts[index].target = value;
+
+@@ -413,6 +418,7 @@ void psxRcntWtarget( u32 index, u32 value )
+ u32 psxRcntRcount( u32 index )
+ {
+ u32 count;
++handler_cycle = psxRegs.cycle;
+
+ count = _psxRcntRcount( index );
+