From: notaz <notasas@gmail.com>
Date: Fri, 9 Sep 2022 19:11:30 +0000 (+0300)
Subject: psxinterpreter: reduce the use of globals
X-Git-Tag: r24~316
X-Git-Url: https://notaz.gp2x.de/cgi-bin/gitweb.cgi?a=commitdiff_plain;h=4cc373dd225bc4c0db251b3689571e97a1e4ec5a;p=pcsx_rearmed.git

psxinterpreter: reduce the use of globals

they induce penalties with -fPIC
---

diff --git a/libpcsxcore/gte.c b/libpcsxcore/gte.c
index 6b3b299f..03261ee2 100644
--- a/libpcsxcore/gte.c
+++ b/libpcsxcore/gte.c
@@ -301,8 +301,7 @@ void gteCheckStall(u32 op) {
 	gteCheckStallRaw(gte_cycletab[op], &psxRegs);
 }
 
-static inline u32 MFC2(int reg) {
-	psxCP2Regs *regs = &psxRegs.CP2;
+u32 MFC2(struct psxCP2Regs *regs, int reg) {
 	switch (reg) {
 		case 1:
 		case 3:
@@ -311,7 +310,7 @@ static inline u32 MFC2(int reg) {
 		case 9:
 		case 10:
 		case 11:
-			psxRegs.CP2D.r[reg] = (s32)psxRegs.CP2D.p[reg].sw.l;
+			regs->CP2D.r[reg] = (s32)regs->CP2D.p[reg].sw.l;
 			break;
 
 		case 7:
@@ -319,25 +318,24 @@ static inline u32 MFC2(int reg) {
 		case 17:
 		case 18:
 		case 19:
-			psxRegs.CP2D.r[reg] = (u32)psxRegs.CP2D.p[reg].w.l;
+			regs->CP2D.r[reg] = (u32)regs->CP2D.p[reg].w.l;
 			break;
 
 		case 15:
-			psxRegs.CP2D.r[reg] = gteSXY2;
+			regs->CP2D.r[reg] = gteSXY2;
 			break;
 
 		case 28:
 		case 29:
-			psxRegs.CP2D.r[reg] = LIM(gteIR1 >> 7, 0x1f, 0, 0) |
+			regs->CP2D.r[reg] = LIM(gteIR1 >> 7, 0x1f, 0, 0) |
 									(LIM(gteIR2 >> 7, 0x1f, 0, 0) << 5) |
 									(LIM(gteIR3 >> 7, 0x1f, 0, 0) << 10);
 			break;
 	}
-	return psxRegs.CP2D.r[reg];
+	return regs->CP2D.r[reg];
 }
 
-static inline void MTC2(u32 value, int reg) {
-	psxCP2Regs *regs = &psxRegs.CP2;
+void MTC2(struct psxCP2Regs *regs, u32 value, int reg) {
 	switch (reg) {
 		case 15:
 			gteSXY0 = gteSXY1;
@@ -379,11 +377,11 @@ static inline void MTC2(u32 value, int reg) {
 			return;
 
 		default:
-			psxRegs.CP2D.r[reg] = value;
+			regs->CP2D.r[reg] = value;
 	}
 }
 
-static inline void CTC2(u32 value, int reg) {
+void CTC2(struct psxCP2Regs *regs, u32 value, int reg) {
 	switch (reg) {
 		case 4:
 		case 12:
@@ -401,45 +399,7 @@ static inline void CTC2(u32 value, int reg) {
 			break;
 	}
 
-	psxRegs.CP2C.r[reg] = value;
-}
-
-void gteMFC2() {
-	if (!_Rt_) return;
-	psxRegs.GPR.r[_Rt_] = MFC2(_Rd_);
-}
-
-void gteCFC2() {
-	if (!_Rt_) return;
-	psxRegs.GPR.r[_Rt_] = psxRegs.CP2C.r[_Rd_];
-}
-
-void gteMTC2() {
-	MTC2(psxRegs.GPR.r[_Rt_], _Rd_);
-}
-
-void gteCTC2() {
-	CTC2(psxRegs.GPR.r[_Rt_], _Rd_);
-}
-
-#define _oB_ (psxRegs.GPR.r[_Rs_] + _Imm_)
-
-void gteLWC2() {
-	MTC2(psxMemRead32(_oB_), _Rt_);
-}
-
-void gteSWC2() {
-	psxMemWrite32(_oB_, MFC2(_Rt_));
-}
-
-void gteLWC2_stall() {
-	gteCheckStall(0);
-	gteLWC2();
-}
-
-void gteSWC2_stall() {
-	gteCheckStall(0);
-	gteSWC2();
+	regs->CP2C.r[reg] = value;
 }
 
 #endif // FLAGLESS
diff --git a/libpcsxcore/gte.h b/libpcsxcore/gte.h
index 75e9e5b3..f1dcc66a 100644
--- a/libpcsxcore/gte.h
+++ b/libpcsxcore/gte.h
@@ -72,14 +72,9 @@ extern const unsigned char gte_cycletab[64];
 int  gteCheckStallRaw(u32 op_cycles, psxRegisters *regs);
 void gteCheckStall(u32 op);
 
-void gteMFC2();
-void gteCFC2();
-void gteMTC2();
-void gteCTC2();
-void gteLWC2();
-void gteSWC2();
-void gteLWC2_stall();
-void gteSWC2_stall();
+u32  MFC2(struct psxCP2Regs *regs, int reg);
+void MTC2(struct psxCP2Regs *regs, u32 value, int reg);
+void CTC2(struct psxCP2Regs *regs, u32 value, int reg);
 
 void gteRTPS(struct psxCP2Regs *regs);
 void gteOP(struct psxCP2Regs *regs);
diff --git a/libpcsxcore/new_dynarec/emu_if.c b/libpcsxcore/new_dynarec/emu_if.c
index bf64e0f0..aa093564 100644
--- a/libpcsxcore/new_dynarec/emu_if.c
+++ b/libpcsxcore/new_dynarec/emu_if.c
@@ -105,13 +105,10 @@ void gen_interupt()
 		next_interupt, next_interupt - psxRegs.cycle);
 }
 
-// from interpreter
-extern void MTC0(int reg, u32 val);
-
 void pcsx_mtc0(u32 reg, u32 val)
 {
 	evprintf("MTC0 %d #%x @%08x %u\n", reg, val, psxRegs.pc, psxRegs.cycle);
-	MTC0(reg, val);
+	MTC0(&psxRegs, reg, val);
 	gen_interupt();
 	if (Cause & Status & 0x0300) // possible sw irq
 		pending_exception = 1;
@@ -120,7 +117,7 @@ void pcsx_mtc0(u32 reg, u32 val)
 void pcsx_mtc0_ds(u32 reg, u32 val)
 {
 	evprintf("MTC0 %d #%x @%08x %u\n", reg, val, psxRegs.pc, psxRegs.cycle);
-	MTC0(reg, val);
+	MTC0(&psxRegs, reg, val);
 }
 
 void new_dyna_before_save(void)
@@ -299,15 +296,13 @@ const uint64_t gte_reg_writes[64] = {
 static int ari64_init()
 {
 	static u32 scratch_buf[8*8*2] __attribute__((aligned(64)));
-	extern void (*psxCP2[64])();
-	extern void psxNULL();
 	size_t i;
 
 	new_dynarec_init();
 	new_dyna_pcsx_mem_init();
 
 	for (i = 0; i < ARRAY_SIZE(gte_handlers); i++)
-		if (psxCP2[i] != psxNULL)
+		if (psxCP2[i] != gteNULL)
 			gte_handlers[i] = psxCP2[i];
 
 #if defined(__arm__) && !defined(DRC_DBG)
diff --git a/libpcsxcore/new_dynarec/new_dynarec.c b/libpcsxcore/new_dynarec/new_dynarec.c
index 1c708566..2b57e59d 100644
--- a/libpcsxcore/new_dynarec/new_dynarec.c
+++ b/libpcsxcore/new_dynarec/new_dynarec.c
@@ -4152,7 +4152,7 @@ static void syscall_assemble(int i, const struct regstat *i_regs, int ccadj_)
 
 static void hlecall_assemble(int i, const struct regstat *i_regs, int ccadj_)
 {
-  void *hlefunc = psxNULL;
+  void *hlefunc = gteNULL;
   uint32_t hleCode = source[i] & 0x03ffffff;
   if (hleCode < ARRAY_SIZE(psxHLEt))
     hlefunc = psxHLEt[hleCode];
diff --git a/libpcsxcore/psxinterpreter.c b/libpcsxcore/psxinterpreter.c
index e7e32690..ea20cab9 100644
--- a/libpcsxcore/psxinterpreter.c
+++ b/libpcsxcore/psxinterpreter.c
@@ -26,6 +26,7 @@
 #include "gte.h"
 #include "psxhle.h"
 #include "psxinterpreter.h"
+#include <stddef.h>
 #include <assert.h>
 //#include "debug.h"
 #define ProcessDebug()
@@ -42,18 +43,26 @@ static u32 branchPC;
 #define debugI()
 #endif
 
+#ifdef __i386__
+#define INT_ATTR __attribute__((regparm(2)))
+#else
+#define INT_ATTR
+#endif
+#ifndef INVALID_PTR
+#define INVALID_PTR NULL
+#endif
+
 // Subsets
-void (*psxBSC[64])();
-void (*psxSPC[64])();
-void (*psxREG[32])();
-void (*psxCP0[32])();
-void (*psxCP2[64])(struct psxCP2Regs *regs);
-void (*psxCP2BSC[32])();
-
-static u32 fetchNoCache(u32 pc)
+static void (INT_ATTR *psxBSC[64])(psxRegisters *regs_, u32 code);
+static void (INT_ATTR *psxSPC[64])(psxRegisters *regs_, u32 code);
+
+static u32 INT_ATTR fetchNoCache(u8 **memRLUT, u32 pc)
 {
-	u32 *code = (u32 *)PSXM(pc);
-	return ((code == NULL) ? 0 : SWAP32(*code));
+	u8 *base = memRLUT[pc >> 16];
+	if (base == INVALID_PTR)
+		return 0;
+	u32 *code = (u32 *)(base + (pc & 0xfffc));
+	return SWAP32(*code);
 }
 
 /*
@@ -65,7 +74,7 @@ static struct cache_entry {
 	u32 data[4];
 } ICache[256];
 
-static u32 fetchICache(u32 pc)
+static u32 INT_ATTR fetchICache(u8 **memRLUT, u32 pc)
 {
 	// cached?
 	if (pc < 0xa0000000)
@@ -75,9 +84,11 @@ static u32 fetchICache(u32 pc)
 
 		if (((entry->tag ^ pc) & 0xfffffff0) != 0 || pc < entry->tag)
 		{
-			u32 *code = (u32 *)PSXM(pc & ~0x0f);
-			if (!code)
+			const u8 *base = memRLUT[pc >> 16];
+			const u32 *code;
+			if (base == INVALID_PTR)
 				return 0;
+			code = (u32 *)(base + (pc & 0xfff0));
 
 			entry->tag = pc;
 			// treat as 4 words, although other configurations are said to be possible
@@ -92,10 +103,10 @@ static u32 fetchICache(u32 pc)
 		return entry->data[(pc & 0x0f) >> 2];
 	}
 
-	return fetchNoCache(pc);
+	return fetchNoCache(memRLUT, pc);
 }
 
-u32 (*fetch)(u32 pc) = fetchNoCache;
+static u32 (INT_ATTR *fetch)(u8 **memRLUT, u32 pc) = fetchNoCache;
 
 static void delayRead(int reg, u32 bpc) {
 	u32 rold, rnew;
@@ -103,7 +114,7 @@ static void delayRead(int reg, u32 bpc) {
 //	SysPrintf("delayRead at %x!\n", psxRegs.pc);
 
 	rold = psxRegs.GPR.r[reg];
-	psxBSC[psxRegs.code >> 26](); // branch delay load
+	psxBSC[psxRegs.code >> 26](&psxRegs, psxRegs.code); // branch delay load
 	rnew = psxRegs.GPR.r[reg];
 
 	psxRegs.pc = bpc;
@@ -126,7 +137,7 @@ static void delayWrite(int reg, u32 bpc) {
 
 	// no changes from normal behavior
 
-	psxBSC[psxRegs.code >> 26]();
+	psxBSC[psxRegs.code >> 26](&psxRegs, psxRegs.code);
 
 	branch = 0;
 	psxRegs.pc = bpc;
@@ -146,6 +157,50 @@ static void delayReadWrite(int reg, u32 bpc) {
 	psxBranchTest();
 }
 
+/**** R3000A Instruction Macros ****/
+#define _PC_            regs_->pc       // The next PC to be executed
+
+#define _fOp_(code)     ((code >> 26)       )  // The opcode part of the instruction register
+#define _fFunct_(code)  ((code      ) & 0x3F)  // The funct part of the instruction register
+#define _fRd_(code)     ((code >> 11) & 0x1F)  // The rd part of the instruction register
+#define _fRt_(code)     ((code >> 16) & 0x1F)  // The rt part of the instruction register
+#define _fRs_(code)     ((code >> 21) & 0x1F)  // The rs part of the instruction register
+#define _fSa_(code)     ((code >>  6) & 0x1F)  // The sa part of the instruction register
+#define _fIm_(code)     ((u16)code)            // The immediate part of the instruction register
+#define _fTarget_(code) (code & 0x03ffffff)    // The target part of the instruction register
+
+#define _fImm_(code)    ((s16)code)            // sign-extended immediate
+#define _fImmU_(code)   (code&0xffff)          // zero-extended immediate
+
+#define _Op_     _fOp_(code)
+#define _Funct_  _fFunct_(code)
+#define _Rd_     _fRd_(code)
+#define _Rt_     _fRt_(code)
+#define _Rs_     _fRs_(code)
+#define _Sa_     _fSa_(code)
+#define _Im_     _fIm_(code)
+#define _Target_ _fTarget_(code)
+
+#define _Imm_    _fImm_(code)
+#define _ImmU_   _fImmU_(code)
+
+#define _rRs_   regs_->GPR.r[_Rs_]   // Rs register
+#define _rRt_   regs_->GPR.r[_Rt_]   // Rt register
+#define _rRd_   regs_->GPR.r[_Rd_]   // Rd register
+#define _rSa_   regs_->GPR.r[_Sa_]   // Sa register
+#define _rFs_   regs_->CP0.r[_Rd_]   // Fs register
+
+#define _rHi_   regs_->GPR.n.hi   // The HI register
+#define _rLo_   regs_->GPR.n.lo   // The LO register
+
+#define _JumpTarget_    ((_Target_ * 4) + (_PC_ & 0xf0000000))   // Calculates the target during a jump instruction
+#define _BranchTarget_  ((s16)_Im_ * 4 + _PC_)                 // Calculates the target during a branch instruction
+
+#define _SetLink(x)     regs_->GPR.r[x] = _PC_ + 4;       // Sets the return address in the link register
+
+#define OP(name) \
+	static inline INT_ATTR void name(psxRegisters *regs_, u32 code)
+
 // this defines shall be used with the tmp 
 // of the next func (instead of _Funct_...)
 #define _tFunct_  ((tmp      ) & 0x3F)  // The funct part of the instruction register 
@@ -154,7 +209,10 @@ static void delayReadWrite(int reg, u32 bpc) {
 #define _tRs_     ((tmp >> 21) & 0x1F)  // The rs part of the instruction register 
 #define _tSa_     ((tmp >>  6) & 0x1F)  // The sa part of the instruction register
 
-int psxTestLoadDelay(int reg, u32 tmp) {
+#define _i32(x) (s32)(x)
+#define _u32(x) (u32)(x)
+
+static int psxTestLoadDelay(int reg, u32 tmp) {
 	if (tmp == 0) return 0; // NOP
 	switch (tmp >> 26) {
 		case 0x00: // SPECIAL
@@ -310,8 +368,8 @@ int psxTestLoadDelay(int reg, u32 tmp) {
 	return 0;
 }
 
-void psxDelayTest(int reg, u32 bpc) {
-	u32 tmp = fetch(bpc);
+static void psxDelayTest(int reg, u32 bpc) {
+	u32 tmp = fetch(psxMemRLUT, bpc);
 	branch = 1;
 
 	switch (psxTestLoadDelay(reg, tmp)) {
@@ -322,7 +380,7 @@ void psxDelayTest(int reg, u32 bpc) {
 		case 3:
 			delayWrite(reg, bpc); return;
 	}
-	psxBSC[psxRegs.code >> 26]();
+	psxBSC[psxRegs.code >> 26](&psxRegs, psxRegs.code);
 
 	branch = 0;
 	psxRegs.pc = bpc;
@@ -330,10 +388,10 @@ void psxDelayTest(int reg, u32 bpc) {
 	psxBranchTest();
 }
 
-static u32 psxBranchNoDelay(void) {
-	u32 temp;
+static u32 psxBranchNoDelay(psxRegisters *regs_) {
+	u32 temp, code;
 
-	psxRegs.code = fetch(psxRegs.pc);
+	regs_->code = code = fetch(psxMemRLUT, regs_->pc);
 	switch (_Op_) {
 		case 0x00: // SPECIAL
 			switch (_Funct_) {
@@ -408,7 +466,7 @@ static int psxDelayBranchExec(u32 tar) {
 static int psxDelayBranchTest(u32 tar1) {
 	u32 tar2, tmp1, tmp2;
 
-	tar2 = psxBranchNoDelay();
+	tar2 = psxBranchNoDelay(&psxRegs);
 	if (tar2 == (u32)-1)
 		return 0;
 
@@ -421,7 +479,7 @@ static int psxDelayBranchTest(u32 tar1) {
 	 *   has no normal delay slot, instruction at tar1 was fetched instead)
 	 */
 	psxRegs.pc = tar1;
-	tmp1 = psxBranchNoDelay();
+	tmp1 = psxBranchNoDelay(&psxRegs);
 	if (tmp1 == (u32)-1) {
 		return psxDelayBranchExec(tar2);
 	}
@@ -434,7 +492,7 @@ static int psxDelayBranchTest(u32 tar1) {
 	 * - jump to target of that branch (tmp1)
 	 */
 	psxRegs.pc = tar2;
-	tmp2 = psxBranchNoDelay();
+	tmp2 = psxBranchNoDelay(&psxRegs);
 	if (tmp2 == (u32)-1) {
 		return psxDelayBranchExec(tmp1);
 	}
@@ -451,7 +509,7 @@ static int psxDelayBranchTest(u32 tar1) {
 }
 
 static void doBranch(u32 tar) {
-	u32 tmp;
+	u32 tmp, code;
 
 	branch2 = branch = 1;
 	branchPC = tar;
@@ -460,7 +518,7 @@ static void doBranch(u32 tar) {
 	if (psxDelayBranchTest(tar))
 		return;
 
-	psxRegs.code = fetch(psxRegs.pc);
+	psxRegs.code = code = fetch(psxMemRLUT, psxRegs.pc);
 
 	debugI();
 
@@ -501,7 +559,7 @@ static void doBranch(u32 tar) {
 			break;
 	}
 
-	psxBSC[psxRegs.code >> 26]();
+	psxBSC[psxRegs.code >> 26](&psxRegs, psxRegs.code);
 
 	branch = 0;
 	psxRegs.pc = branchPC;
@@ -513,107 +571,104 @@ static void doBranch(u32 tar) {
 * Arithmetic with immediate operand                      *
 * Format:  OP rt, rs, immediate                          *
 *********************************************************/
-void psxADDI() 	{ if (!_Rt_) return; _rRt_ = _u32(_rRs_) + _Imm_ ; }		// Rt = Rs + Im 	(Exception on Integer Overflow)
-void psxADDIU() { if (!_Rt_) return; _rRt_ = _u32(_rRs_) + _Imm_ ; }		// Rt = Rs + Im
-void psxANDI() 	{ if (!_Rt_) return; _rRt_ = _u32(_rRs_) & _ImmU_; }		// Rt = Rs And Im
-void psxORI() 	{ if (!_Rt_) return; _rRt_ = _u32(_rRs_) | _ImmU_; }		// Rt = Rs Or  Im
-void psxXORI() 	{ if (!_Rt_) return; _rRt_ = _u32(_rRs_) ^ _ImmU_; }		// Rt = Rs Xor Im
-void psxSLTI() 	{ if (!_Rt_) return; _rRt_ = _i32(_rRs_) < _Imm_ ; }		// Rt = Rs < Im		(Signed)
-void psxSLTIU() { if (!_Rt_) return; _rRt_ = _u32(_rRs_) < ((u32)_Imm_); }		// Rt = Rs < Im		(Unsigned)
+OP(psxADDI)  { if (!_Rt_) return; _rRt_ = _u32(_rRs_) + _Imm_ ; }  // Rt = Rs + Im 	(Exception on Integer Overflow)
+OP(psxADDIU) { if (!_Rt_) return; _rRt_ = _u32(_rRs_) + _Imm_ ; }  // Rt = Rs + Im
+OP(psxANDI)  { if (!_Rt_) return; _rRt_ = _u32(_rRs_) & _ImmU_; }  // Rt = Rs And Im
+OP(psxORI)   { if (!_Rt_) return; _rRt_ = _u32(_rRs_) | _ImmU_; }  // Rt = Rs Or  Im
+OP(psxXORI)  { if (!_Rt_) return; _rRt_ = _u32(_rRs_) ^ _ImmU_; }  // Rt = Rs Xor Im
+OP(psxSLTI)  { if (!_Rt_) return; _rRt_ = _i32(_rRs_) < _Imm_ ; }  // Rt = Rs < Im		(Signed)
+OP(psxSLTIU) { if (!_Rt_) return; _rRt_ = _u32(_rRs_) < ((u32)_Imm_); } // Rt = Rs < Im		(Unsigned)
 
 /*********************************************************
 * Register arithmetic                                    *
 * Format:  OP rd, rs, rt                                 *
 *********************************************************/
-void psxADD()	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) + _u32(_rRt_); }	// Rd = Rs + Rt		(Exception on Integer Overflow)
-void psxADDU() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) + _u32(_rRt_); }	// Rd = Rs + Rt
-void psxSUB() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) - _u32(_rRt_); }	// Rd = Rs - Rt		(Exception on Integer Overflow)
-void psxSUBU() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) - _u32(_rRt_); }	// Rd = Rs - Rt
-void psxAND() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) & _u32(_rRt_); }	// Rd = Rs And Rt
-void psxOR() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) | _u32(_rRt_); }	// Rd = Rs Or  Rt
-void psxXOR() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) ^ _u32(_rRt_); }	// Rd = Rs Xor Rt
-void psxNOR() 	{ if (!_Rd_) return; _rRd_ =~(_u32(_rRs_) | _u32(_rRt_)); }// Rd = Rs Nor Rt
-void psxSLT() 	{ if (!_Rd_) return; _rRd_ = _i32(_rRs_) < _i32(_rRt_); }	// Rd = Rs < Rt		(Signed)
-void psxSLTU() 	{ if (!_Rd_) return; _rRd_ = _u32(_rRs_) < _u32(_rRt_); }	// Rd = Rs < Rt		(Unsigned)
+OP(psxADD)   { if (!_Rd_) return; _rRd_ = _u32(_rRs_) + _u32(_rRt_); }	// Rd = Rs + Rt		(Exception on Integer Overflow)
+OP(psxADDU)  { if (!_Rd_) return; _rRd_ = _u32(_rRs_) + _u32(_rRt_); }	// Rd = Rs + Rt
+OP(psxSUB)   { if (!_Rd_) return; _rRd_ = _u32(_rRs_) - _u32(_rRt_); }	// Rd = Rs - Rt		(Exception on Integer Overflow)
+OP(psxSUBU)  { if (!_Rd_) return; _rRd_ = _u32(_rRs_) - _u32(_rRt_); }	// Rd = Rs - Rt
+OP(psxAND)   { if (!_Rd_) return; _rRd_ = _u32(_rRs_) & _u32(_rRt_); }	// Rd = Rs And Rt
+OP(psxOR)    { if (!_Rd_) return; _rRd_ = _u32(_rRs_) | _u32(_rRt_); }	// Rd = Rs Or  Rt
+OP(psxXOR)   { if (!_Rd_) return; _rRd_ = _u32(_rRs_) ^ _u32(_rRt_); }	// Rd = Rs Xor Rt
+OP(psxNOR)   { if (!_Rd_) return; _rRd_ =~(_u32(_rRs_) | _u32(_rRt_)); }// Rd = Rs Nor Rt
+OP(psxSLT)   { if (!_Rd_) return; _rRd_ = _i32(_rRs_) < _i32(_rRt_); }	// Rd = Rs < Rt		(Signed)
+OP(psxSLTU)  { if (!_Rd_) return; _rRd_ = _u32(_rRs_) < _u32(_rRt_); }	// Rd = Rs < Rt		(Unsigned)
 
 /*********************************************************
 * Register mult/div & Register trap logic                *
 * Format:  OP rs, rt                                     *
 *********************************************************/
-void psxDIV() {
-    if (!_i32(_rRt_)) {
-        _i32(_rHi_) = _i32(_rRs_);
-        if (_i32(_rRs_) & 0x80000000) {
-            _i32(_rLo_) = 1;
-        } else {
-            _i32(_rLo_) = 0xFFFFFFFF;
-        }
-/*
- * Notaz said that this was "not needed" for ARM platforms and could slow it down so let's disable for ARM. 
- * This fixes a crash issue that can happen when running Amidog's CPU test.
- * (It still stays stuck to a black screen but at least it doesn't crash anymore)
- */
+OP(psxDIV) {
+	if (!_rRt_) {
+		_rHi_ = _rRs_;
+		if (_rRs_ & 0x80000000) {
+			_rLo_ = 1;
+		} else {
+			_rLo_ = 0xFFFFFFFF;
+		}
+	}
 #if !defined(__arm__) && !defined(__aarch64__)
-    } else if (_i32(_rRs_) == 0x80000000 && _i32(_rRt_) == 0xFFFFFFFF) {
-        _i32(_rLo_) = 0x80000000;
-        _i32(_rHi_) = 0;
+	else if (_rRs_ == 0x80000000 && _rRt_ == 0xFFFFFFFF) {
+		_rLo_ = 0x80000000;
+		_rHi_ = 0;
+	}
 #endif
-    } else {
-        _i32(_rLo_) = _i32(_rRs_) / _i32(_rRt_);
-        _i32(_rHi_) = _i32(_rRs_) % _i32(_rRt_);
-    }
+	else {
+		_rLo_ = _i32(_rRs_) / _i32(_rRt_);
+		_rHi_ = _i32(_rRs_) % _i32(_rRt_);
+	}
 }
 
-void psxDIV_stall() {
-	psxRegs.muldivBusyCycle = psxRegs.cycle + 37;
-	psxDIV();
+OP(psxDIV_stall) {
+	regs_->muldivBusyCycle = regs_->cycle + 37;
+	psxDIV(regs_, code);
 }
 
-void psxDIVU() {
+OP(psxDIVU) {
 	if (_rRt_ != 0) {
 		_rLo_ = _rRs_ / _rRt_;
 		_rHi_ = _rRs_ % _rRt_;
 	}
 	else {
-		_i32(_rLo_) = 0xffffffff;
-		_i32(_rHi_) = _i32(_rRs_);
+		_rLo_ = 0xffffffff;
+		_rHi_ = _rRs_;
 	}
 }
 
-void psxDIVU_stall() {
-	psxRegs.muldivBusyCycle = psxRegs.cycle + 37;
-	psxDIVU();
+OP(psxDIVU_stall) {
+	regs_->muldivBusyCycle = regs_->cycle + 37;
+	psxDIVU(regs_, code);
 }
 
-void psxMULT() {
-	u64 res = (s64)((s64)_i32(_rRs_) * (s64)_i32(_rRt_));
+OP(psxMULT) {
+	u64 res = (s64)_i32(_rRs_) * _i32(_rRt_);
 
-	psxRegs.GPR.n.lo = (u32)(res & 0xffffffff);
-	psxRegs.GPR.n.hi = (u32)((res >> 32) & 0xffffffff);
+	regs_->GPR.n.lo = (u32)res;
+	regs_->GPR.n.hi = (u32)(res >> 32);
 }
 
-void psxMULT_stall() {
+OP(psxMULT_stall) {
 	// approximate, but maybe good enough
 	u32 rs = _rRs_;
 	u32 lz = __builtin_clz(((rs ^ ((s32)rs >> 21)) | 1));
 	u32 c = 7 + (2 - (lz / 11)) * 4;
-	psxRegs.muldivBusyCycle = psxRegs.cycle + c;
-	psxMULT();
+	regs_->muldivBusyCycle = regs_->cycle + c;
+	psxMULT(regs_, code);
 }
 
-void psxMULTU() {
-	u64 res = (u64)((u64)_u32(_rRs_) * (u64)_u32(_rRt_));
+OP(psxMULTU) {
+	u64 res = (u64)_u32(_rRs_) * _u32(_rRt_);
 
-	psxRegs.GPR.n.lo = (u32)(res & 0xffffffff);
-	psxRegs.GPR.n.hi = (u32)((res >> 32) & 0xffffffff);
+	regs_->GPR.n.lo = (u32)(res & 0xffffffff);
+	regs_->GPR.n.hi = (u32)((res >> 32) & 0xffffffff);
 }
 
-void psxMULTU_stall() {
+OP(psxMULTU_stall) {
 	// approximate, but maybe good enough
 	u32 lz = __builtin_clz(_rRs_ | 1);
 	u32 c = 7 + (2 - (lz / 11)) * 4;
-	psxRegs.muldivBusyCycle = psxRegs.cycle + c;
-	psxMULTU();
+	regs_->muldivBusyCycle = regs_->cycle + c;
+	psxMULTU(regs_, code);
 }
 
 /*********************************************************
@@ -623,80 +678,88 @@ void psxMULTU_stall() {
 #define RepZBranchi32(op)      if(_i32(_rRs_) op 0) doBranch(_BranchTarget_);
 #define RepZBranchLinki32(op)  { _SetLink(31); if(_i32(_rRs_) op 0) { doBranch(_BranchTarget_); } }
 
-void psxBGEZ()   { RepZBranchi32(>=) }      // Branch if Rs >= 0
-void psxBGEZAL() { RepZBranchLinki32(>=) }  // Branch if Rs >= 0 and link
-void psxBGTZ()   { RepZBranchi32(>) }       // Branch if Rs >  0
-void psxBLEZ()   { RepZBranchi32(<=) }      // Branch if Rs <= 0
-void psxBLTZ()   { RepZBranchi32(<) }       // Branch if Rs <  0
-void psxBLTZAL() { RepZBranchLinki32(<) }   // Branch if Rs <  0 and link
+OP(psxBGEZ)   { RepZBranchi32(>=) }      // Branch if Rs >= 0
+OP(psxBGEZAL) { RepZBranchLinki32(>=) }  // Branch if Rs >= 0 and link
+OP(psxBGTZ)   { RepZBranchi32(>) }       // Branch if Rs >  0
+OP(psxBLEZ)   { RepZBranchi32(<=) }      // Branch if Rs <= 0
+OP(psxBLTZ)   { RepZBranchi32(<) }       // Branch if Rs <  0
+OP(psxBLTZAL) { RepZBranchLinki32(<) }   // Branch if Rs <  0 and link
 
 /*********************************************************
 * Shift arithmetic with constant shift                   *
 * Format:  OP rd, rt, sa                                 *
 *********************************************************/
-void psxSLL() { if (!_Rd_) return; _u32(_rRd_) = _u32(_rRt_) << _Sa_; } // Rd = Rt << sa
-void psxSRA() { if (!_Rd_) return; _i32(_rRd_) = _i32(_rRt_) >> _Sa_; } // Rd = Rt >> sa (arithmetic)
-void psxSRL() { if (!_Rd_) return; _u32(_rRd_) = _u32(_rRt_) >> _Sa_; } // Rd = Rt >> sa (logical)
+OP(psxSLL) { if (!_Rd_) return; _rRd_ = _u32(_rRt_) << _Sa_; } // Rd = Rt << sa
+OP(psxSRA) { if (!_Rd_) return; _rRd_ = _i32(_rRt_) >> _Sa_; } // Rd = Rt >> sa (arithmetic)
+OP(psxSRL) { if (!_Rd_) return; _rRd_ = _u32(_rRt_) >> _Sa_; } // Rd = Rt >> sa (logical)
 
 /*********************************************************
 * Shift arithmetic with variant register shift           *
 * Format:  OP rd, rt, rs                                 *
 *********************************************************/
-void psxSLLV() { if (!_Rd_) return; _u32(_rRd_) = _u32(_rRt_) << (_u32(_rRs_) & 0x1F); } // Rd = Rt << rs
-void psxSRAV() { if (!_Rd_) return; _i32(_rRd_) = _i32(_rRt_) >> (_u32(_rRs_) & 0x1F); } // Rd = Rt >> rs (arithmetic)
-void psxSRLV() { if (!_Rd_) return; _u32(_rRd_) = _u32(_rRt_) >> (_u32(_rRs_) & 0x1F); } // Rd = Rt >> rs (logical)
+OP(psxSLLV) { if (!_Rd_) return; _rRd_ = _u32(_rRt_) << (_u32(_rRs_) & 0x1F); } // Rd = Rt << rs
+OP(psxSRAV) { if (!_Rd_) return; _rRd_ = _i32(_rRt_) >> (_u32(_rRs_) & 0x1F); } // Rd = Rt >> rs (arithmetic)
+OP(psxSRLV) { if (!_Rd_) return; _rRd_ = _u32(_rRt_) >> (_u32(_rRs_) & 0x1F); } // Rd = Rt >> rs (logical)
 
 /*********************************************************
 * Load higher 16 bits of the first word in GPR with imm  *
 * Format:  OP rt, immediate                              *
 *********************************************************/
-void psxLUI() { if (!_Rt_) return; _u32(_rRt_) = psxRegs.code << 16; } // Upper halfword of Rt = Im
+OP(psxLUI) { if (!_Rt_) return; _rRt_ = code << 16; } // Upper halfword of Rt = Im
 
 /*********************************************************
 * Move from HI/LO to GPR                                 *
 * Format:  OP rd                                         *
 *********************************************************/
-void psxMFHI() { if (!_Rd_) return; _rRd_ = _rHi_; } // Rd = Hi
-void psxMFLO() { if (!_Rd_) return; _rRd_ = _rLo_; } // Rd = Lo
+OP(psxMFHI) { if (!_Rd_) return; _rRd_ = _rHi_; } // Rd = Hi
+OP(psxMFLO) { if (!_Rd_) return; _rRd_ = _rLo_; } // Rd = Lo
 
-static void mflohiCheckStall(void)
+static void mflohiCheckStall(psxRegisters *regs_)
 {
-	u32 left = psxRegs.muldivBusyCycle - psxRegs.cycle;
+	u32 left = regs_->muldivBusyCycle - regs_->cycle;
 	if (left <= 37) {
 		//printf("muldiv stall %u\n", left);
-		psxRegs.cycle = psxRegs.muldivBusyCycle;
+		regs_->cycle = regs_->muldivBusyCycle;
 	}
 }
 
-void psxMFHI_stall() { mflohiCheckStall(); psxMFHI(); }
-void psxMFLO_stall() { mflohiCheckStall(); psxMFLO(); }
+OP(psxMFHI_stall) { mflohiCheckStall(regs_); psxMFHI(regs_, code); }
+OP(psxMFLO_stall) { mflohiCheckStall(regs_); psxMFLO(regs_, code); }
 
 /*********************************************************
 * Move to GPR to HI/LO & Register jump                   *
 * Format:  OP rs                                         *
 *********************************************************/
-void psxMTHI() { _rHi_ = _rRs_; } // Hi = Rs
-void psxMTLO() { _rLo_ = _rRs_; } // Lo = Rs
+OP(psxMTHI) { _rHi_ = _rRs_; } // Hi = Rs
+OP(psxMTLO) { _rLo_ = _rRs_; } // Lo = Rs
 
 /*********************************************************
 * Special purpose instructions                           *
 * Format:  OP                                            *
 *********************************************************/
-void psxBREAK() {
-	psxRegs.pc -= 4;
+OP(psxBREAK) {
+	regs_->pc -= 4;
 	psxException(0x24, branch);
 }
 
-void psxSYSCALL() {
-	psxRegs.pc -= 4;
+OP(psxSYSCALL) {
+	regs_->pc -= 4;
 	psxException(0x20, branch);
 }
 
-void psxRFE() {
+static inline void psxTestSWInts(psxRegisters *regs_) {
+	if (regs_->CP0.n.Cause & regs_->CP0.n.Status & 0x0300 &&
+	   regs_->CP0.n.Status & 0x1) {
+		regs_->CP0.n.Cause &= ~0x7c;
+		psxException(regs_->CP0.n.Cause, branch);
+	}
+}
+
+OP(psxRFE) {
 //	SysPrintf("psxRFE\n");
-	psxRegs.CP0.n.Status = (psxRegs.CP0.n.Status & 0xfffffff0) |
-						  ((psxRegs.CP0.n.Status & 0x3c) >> 2);
-	psxTestSWInts();
+	regs_->CP0.n.Status = (regs_->CP0.n.Status & 0xfffffff0) |
+	                      ((regs_->CP0.n.Status & 0x3c) >> 2);
+	psxTestSWInts(regs_);
 }
 
 /*********************************************************
@@ -705,26 +768,26 @@ void psxRFE() {
 *********************************************************/
 #define RepBranchi32(op)      if(_i32(_rRs_) op _i32(_rRt_)) doBranch(_BranchTarget_);
 
-void psxBEQ() {	RepBranchi32(==) }  // Branch if Rs == Rt
-void psxBNE() {	RepBranchi32(!=) }  // Branch if Rs != Rt
+OP(psxBEQ) { RepBranchi32(==) }  // Branch if Rs == Rt
+OP(psxBNE) { RepBranchi32(!=) }  // Branch if Rs != Rt
 
 /*********************************************************
 * Jump to target                                         *
 * Format:  OP target                                     *
 *********************************************************/
-void psxJ()   {               doBranch(_JumpTarget_); }
-void psxJAL() {	_SetLink(31); doBranch(_JumpTarget_); }
+OP(psxJ)   {               doBranch(_JumpTarget_); }
+OP(psxJAL) { _SetLink(31); doBranch(_JumpTarget_); }
 
 /*********************************************************
 * Register jump                                          *
 * Format:  OP rs, rd                                     *
 *********************************************************/
-void psxJR()   {
+OP(psxJR) {
 	doBranch(_rRs_ & ~3);
 	psxJumpTest();
 }
 
-void psxJALR() {
+OP(psxJALR) {
 	u32 temp = _u32(_rRs_);
 	if (_Rd_) { _SetLink(_Rd_); }
 	doBranch(temp & ~3);
@@ -735,59 +798,23 @@ void psxJALR() {
 * Format:  OP rt, offset(base)                           *
 *********************************************************/
 
-#define _oB_ (_u32(_rRs_) + _Imm_)
-
-void psxLB() {
-	if (_Rt_) {
-		_i32(_rRt_) = (signed char)psxMemRead8(_oB_); 
-	} else {
-		psxMemRead8(_oB_); 
-	}
-}
-
-void psxLBU() {
-	if (_Rt_) {
-		_u32(_rRt_) = psxMemRead8(_oB_);
-	} else {
-		psxMemRead8(_oB_); 
-	}
-}
-
-void psxLH() {
-	if (_Rt_) {
-		_i32(_rRt_) = (short)psxMemRead16(_oB_);
-	} else {
-		psxMemRead16(_oB_);
-	}
-}
-
-void psxLHU() {
-	if (_Rt_) {
-		_u32(_rRt_) = psxMemRead16(_oB_);
-	} else {
-		psxMemRead16(_oB_);
-	}
-}
+#define _oB_ (regs_->GPR.r[_Rs_] + _Imm_)
 
-void psxLW() {
-	if (_Rt_) {
-		_u32(_rRt_) = psxMemRead32(_oB_);
-	} else {
-		psxMemRead32(_oB_);
-	}
-}
-
-u32 LWL_MASK[4] = { 0xffffff, 0xffff, 0xff, 0 };
-u32 LWL_SHIFT[4] = { 24, 16, 8, 0 };
+OP(psxLB)  { u32 v =  (s8)psxMemRead8(_oB_);  if (_Rt_) _rRt_ = v; }
+OP(psxLBU) { u32 v =      psxMemRead8(_oB_);  if (_Rt_) _rRt_ = v; }
+OP(psxLH)  { u32 v = (s16)psxMemRead16(_oB_); if (_Rt_) _rRt_ = v; }
+OP(psxLHU) { u32 v =      psxMemRead16(_oB_); if (_Rt_) _rRt_ = v; }
+OP(psxLW)  { u32 v =      psxMemRead32(_oB_); if (_Rt_) _rRt_ = v; }
 
-void psxLWL() {
+OP(psxLWL) {
+	static const u32 LWL_MASK[4] = { 0xffffff, 0xffff, 0xff, 0 };
+	static const u32 LWL_SHIFT[4] = { 24, 16, 8, 0 };
 	u32 addr = _oB_;
 	u32 shift = addr & 3;
 	u32 mem = psxMemRead32(addr & ~3);
 
 	if (!_Rt_) return;
-	_u32(_rRt_) =	( _u32(_rRt_) & LWL_MASK[shift]) | 
-					( mem << LWL_SHIFT[shift]);
+	_rRt_ = (_u32(_rRt_) & LWL_MASK[shift]) | (mem << LWL_SHIFT[shift]);
 
 	/*
 	Mem = 1234.  Reg = abcd
@@ -799,17 +826,15 @@ void psxLWL() {
 	*/
 }
 
-u32 LWR_MASK[4] = { 0, 0xff000000, 0xffff0000, 0xffffff00 };
-u32 LWR_SHIFT[4] = { 0, 8, 16, 24 };
-
-void psxLWR() {
+OP(psxLWR) {
+	static const u32 LWR_MASK[4] = { 0, 0xff000000, 0xffff0000, 0xffffff00 };
+	static const u32 LWR_SHIFT[4] = { 0, 8, 16, 24 };
 	u32 addr = _oB_;
 	u32 shift = addr & 3;
 	u32 mem = psxMemRead32(addr & ~3);
 
 	if (!_Rt_) return;
-	_u32(_rRt_) =	( _u32(_rRt_) & LWR_MASK[shift]) | 
-					( mem >> LWR_SHIFT[shift]);
+	_rRt_ = (_u32(_rRt_) & LWR_MASK[shift]) | (mem >> LWR_SHIFT[shift]);
 
 	/*
 	Mem = 1234.  Reg = abcd
@@ -821,14 +846,13 @@ void psxLWR() {
 	*/
 }
 
-void psxSB() { psxMemWrite8 (_oB_, _rRt_ &   0xff); }
-void psxSH() { psxMemWrite16(_oB_, _rRt_ & 0xffff); }
-void psxSW() { psxMemWrite32(_oB_, _rRt_); }
-
-u32 SWL_MASK[4] = { 0xffffff00, 0xffff0000, 0xff000000, 0 };
-u32 SWL_SHIFT[4] = { 24, 16, 8, 0 };
+OP(psxSB) { psxMemWrite8 (_oB_, _rRt_ &   0xff); }
+OP(psxSH) { psxMemWrite16(_oB_, _rRt_ & 0xffff); }
+OP(psxSW) { psxMemWrite32(_oB_, _rRt_); }
 
-void psxSWL() {
+OP(psxSWL) {
+	static const u32 SWL_MASK[4] = { 0xffffff00, 0xffff0000, 0xff000000, 0 };
+	static const u32 SWL_SHIFT[4] = { 24, 16, 8, 0 };
 	u32 addr = _oB_;
 	u32 shift = addr & 3;
 	u32 mem = psxMemRead32(addr & ~3);
@@ -845,10 +869,9 @@ void psxSWL() {
 	*/
 }
 
-u32 SWR_MASK[4] = { 0, 0xff, 0xffff, 0xffffff };
-u32 SWR_SHIFT[4] = { 0, 8, 16, 24 };
-
-void psxSWR() {
+OP(psxSWR) {
+	static const u32 SWR_MASK[4] = { 0, 0xff, 0xffff, 0xffffff };
+	static const u32 SWR_SHIFT[4] = { 0, 8, 16, 24 };
 	u32 addr = _oB_;
 	u32 shift = addr & 3;
 	u32 mem = psxMemRead32(addr & ~3);
@@ -870,88 +893,139 @@ void psxSWR() {
 * Moves between GPR and COPx                             *
 * Format:  OP rt, fs                                     *
 *********************************************************/
-void psxMFC0() { if (!_Rt_) return; _i32(_rRt_) = (int)_rFs_; }
-void psxCFC0() { if (!_Rt_) return; _i32(_rRt_) = (int)_rFs_; }
-
-void psxTestSWInts() {
-	if (psxRegs.CP0.n.Cause & psxRegs.CP0.n.Status & 0x0300 &&
-	   psxRegs.CP0.n.Status & 0x1) {
-		psxRegs.CP0.n.Cause &= ~0x7c;
-		psxException(psxRegs.CP0.n.Cause, branch);
-	}
-}
+OP(psxMFC0) { if (!_Rt_) return; _rRt_ = _rFs_; }
+OP(psxCFC0) { if (!_Rt_) return; _rRt_ = _rFs_; }
 
-void MTC0(int reg, u32 val) {
+void MTC0(psxRegisters *regs_, int reg, u32 val) {
 //	SysPrintf("MTC0 %d: %x\n", reg, val);
 	switch (reg) {
 		case 12: // Status
-			psxRegs.CP0.r[12] = val;
-			psxTestSWInts();
+			regs_->CP0.r[12] = val;
+			psxTestSWInts(regs_);
 			break;
 
 		case 13: // Cause
-			psxRegs.CP0.n.Cause &= ~0x0300;
-			psxRegs.CP0.n.Cause |= val & 0x0300;
-			psxTestSWInts();
+			regs_->CP0.n.Cause &= ~0x0300;
+			regs_->CP0.n.Cause |= val & 0x0300;
+			psxTestSWInts(regs_);
 			break;
 
 		default:
-			psxRegs.CP0.r[reg] = val;
+			regs_->CP0.r[reg] = val;
 			break;
 	}
 }
 
-void psxMTC0() { MTC0(_Rd_, _u32(_rRt_)); }
-void psxCTC0() { MTC0(_Rd_, _u32(_rRt_)); }
+OP(psxMTC0) { MTC0(regs_, _Rd_, _u32(_rRt_)); }
+OP(psxCTC0) { MTC0(regs_, _Rd_, _u32(_rRt_)); }
 
 /*********************************************************
 * Unknow instruction (would generate an exception)       *
 * Format:  ?                                             *
 *********************************************************/
-void psxNULL() { 
+static inline void psxNULL_(void) {
 #ifdef PSXCPU_LOG
 	PSXCPU_LOG("psx: Unimplemented op %x\n", psxRegs.code);
 #endif
 }
 
-void psxSPECIAL() {
-	psxSPC[_Funct_]();
-}
+OP(psxNULL) { psxNULL_(); }
+void gteNULL(struct psxCP2Regs *regs) { psxNULL_(); }
 
-void psxREGIMM() {
-	psxREG[_Rt_]();
+OP(psxSPECIAL) {
+	psxSPC[_Funct_](regs_, code);
 }
 
-void psxCOP0() {
-	psxCP0[_Rs_]();
+OP(psxCOP0) {
+	switch (_Rs_) {
+		case 0x00: psxMFC0(regs_, code); break;
+		case 0x02: psxCFC0(regs_, code); break;
+		case 0x04: psxMTC0(regs_, code); break;
+		case 0x06: psxCTC0(regs_, code); break;
+		case 0x10: psxRFE(regs_, code);  break;
+		default:   psxNULL_();           break;
+	}
 }
 
-void psxCOP2() {
-	psxCP2[_Funct_]((struct psxCP2Regs *)&psxRegs.CP2D);
+OP(psxCOP2) {
+	psxCP2[_Funct_](&regs_->CP2);
 }
 
-void psxCOP2_stall() {
+OP(psxCOP2_stall) {
 	u32 f = _Funct_;
 	gteCheckStall(f);
-	psxCP2[f]((struct psxCP2Regs *)&psxRegs.CP2D);
+	psxCP2[f](&regs_->CP2);
+}
+
+OP(gteMFC2) {
+	if (!_Rt_) return;
+	regs_->GPR.r[_Rt_] = MFC2(&regs_->CP2, _Rd_);
+}
+
+OP(gteCFC2) {
+	if (!_Rt_) return;
+	regs_->GPR.r[_Rt_] = regs_->CP2C.r[_Rd_];
+}
+
+OP(gteMTC2) {
+	MTC2(&regs_->CP2, regs_->GPR.r[_Rt_], _Rd_);
+}
+
+OP(gteCTC2) {
+	CTC2(&regs_->CP2, regs_->GPR.r[_Rt_], _Rd_);
+}
+
+OP(gteLWC2) {
+	MTC2(&regs_->CP2, psxMemRead32(_oB_), _Rt_);
+}
+
+OP(gteSWC2) {
+	psxMemWrite32(_oB_, MFC2(&regs_->CP2, _Rt_));
+}
+
+OP(gteLWC2_stall) {
+	gteCheckStall(0);
+	gteLWC2(regs_, code);
 }
 
-void psxBASIC(struct psxCP2Regs *regs) {
-	psxCP2BSC[_Rs_]();
+OP(gteSWC2_stall) {
+	gteCheckStall(0);
+	gteSWC2(regs_, code);
+}
+
+static void psxBASIC(struct psxCP2Regs *cp2regs) {
+	psxRegisters *regs_ = (void *)((char *)cp2regs - offsetof(psxRegisters, CP2));
+	u32 code = regs_->code;
+	assert(regs_ == &psxRegs);
+	switch (_Rs_) {
+		case 0x00: gteMFC2(regs_, code); break;
+		case 0x02: gteCFC2(regs_, code); break;
+		case 0x04: gteMTC2(regs_, code); break;
+		case 0x06: gteCTC2(regs_, code); break;
+		default:   psxNULL_();           break;
+	}
+}
+
+OP(psxREGIMM) {
+	switch (_Rt_) {
+		case 0x00: psxBLTZ(regs_, code);   break;
+		case 0x01: psxBGEZ(regs_, code);   break;
+		case 0x10: psxBLTZAL(regs_, code); break;
+		case 0x11: psxBGEZAL(regs_, code); break;
+		default:   psxNULL_();             break;
+	}
 }
 
-void psxHLE() {
-//	psxHLEt[psxRegs.code & 0xffff]();
-//	psxHLEt[psxRegs.code & 0x07]();		// HDHOSHY experimental patch
-    uint32_t hleCode = psxRegs.code & 0x03ffffff;
+OP(psxHLE) {
+    uint32_t hleCode = code & 0x03ffffff;
     if (hleCode >= (sizeof(psxHLEt) / sizeof(psxHLEt[0]))) {
-        psxNULL();
+        psxNULL_();
     } else {
         psxHLEt[hleCode]();
     }
 }
 
-void (*psxBSC[64])() = {
+static void (INT_ATTR *psxBSC[64])(psxRegisters *regs_, u32 code) = {
 	psxSPECIAL, psxREGIMM, psxJ   , psxJAL  , psxBEQ , psxBNE , psxBLEZ, psxBGTZ,
 	psxADDI   , psxADDIU , psxSLTI, psxSLTIU, psxANDI, psxORI , psxXORI, psxLUI ,
 	psxCOP0   , psxNULL  , psxCOP2, psxNULL , psxNULL, psxNULL, psxNULL, psxNULL,
@@ -962,8 +1036,7 @@ void (*psxBSC[64])() = {
 	psxNULL   , psxNULL  , gteSWC2, psxHLE  , psxNULL, psxNULL, psxNULL, psxNULL 
 };
 
-
-void (*psxSPC[64])() = {
+static void (INT_ATTR *psxSPC[64])(psxRegisters *regs_, u32 code) = {
 	psxSLL , psxNULL , psxSRL , psxSRA , psxSLLV   , psxNULL , psxSRLV, psxSRAV,
 	psxJR  , psxJALR , psxNULL, psxNULL, psxSYSCALL, psxBREAK, psxNULL, psxNULL,
 	psxMFHI, psxMTHI , psxMFLO, psxMTLO, psxNULL   , psxNULL , psxNULL, psxNULL,
@@ -974,39 +1047,17 @@ void (*psxSPC[64])() = {
 	psxNULL, psxNULL , psxNULL, psxNULL, psxNULL   , psxNULL , psxNULL, psxNULL
 };
 
-void (*psxREG[32])() = {
-	psxBLTZ  , psxBGEZ  , psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxNULL  , psxNULL  , psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxBLTZAL, psxBGEZAL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxNULL  , psxNULL  , psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL
-};
-
-void (*psxCP0[32])() = {
-	psxMFC0, psxNULL, psxCFC0, psxNULL, psxMTC0, psxNULL, psxCTC0, psxNULL,
-	psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxRFE , psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL
-};
-
 void (*psxCP2[64])(struct psxCP2Regs *regs) = {
-	psxBASIC, gteRTPS , psxNULL , psxNULL, psxNULL, psxNULL , gteNCLIP, psxNULL, // 00
-	psxNULL , psxNULL , psxNULL , psxNULL, gteOP  , psxNULL , psxNULL , psxNULL, // 08
-	gteDPCS , gteINTPL, gteMVMVA, gteNCDS, gteCDP , psxNULL , gteNCDT , psxNULL, // 10
-	psxNULL , psxNULL , psxNULL , gteNCCS, gteCC  , psxNULL , gteNCS  , psxNULL, // 18
-	gteNCT  , psxNULL , psxNULL , psxNULL, psxNULL, psxNULL , psxNULL , psxNULL, // 20
-	gteSQR  , gteDCPL , gteDPCT , psxNULL, psxNULL, gteAVSZ3, gteAVSZ4, psxNULL, // 28 
-	gteRTPT , psxNULL , psxNULL , psxNULL, psxNULL, psxNULL , psxNULL , psxNULL, // 30
-	psxNULL , psxNULL , psxNULL , psxNULL, psxNULL, gteGPF  , gteGPL  , gteNCCT  // 38
-};
-
-void (*psxCP2BSC[32])() = {
-	gteMFC2, psxNULL, gteCFC2, psxNULL, gteMTC2, psxNULL, gteCTC2, psxNULL,
-	psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL,
-	psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL, psxNULL
+	psxBASIC, gteRTPS , gteNULL , gteNULL, gteNULL, gteNULL , gteNCLIP, gteNULL, // 00
+	gteNULL , gteNULL , gteNULL , gteNULL, gteOP  , gteNULL , gteNULL , gteNULL, // 08
+	gteDPCS , gteINTPL, gteMVMVA, gteNCDS, gteCDP , gteNULL , gteNCDT , gteNULL, // 10
+	gteNULL , gteNULL , gteNULL , gteNCCS, gteCC  , gteNULL , gteNCS  , gteNULL, // 18
+	gteNCT  , gteNULL , gteNULL , gteNULL, gteNULL, gteNULL , gteNULL , gteNULL, // 20
+	gteSQR  , gteDCPL , gteDPCT , gteNULL, gteNULL, gteAVSZ3, gteAVSZ4, gteNULL, // 28
+	gteRTPT , gteNULL , gteNULL , gteNULL, gteNULL, gteNULL , gteNULL , gteNULL, // 30
+	gteNULL , gteNULL , gteNULL , gteNULL, gteNULL, gteGPF  , gteGPL  , gteNCCT  // 38
 };
 
-
 ///////////////////////////////////////////
 
 static int intInit() {
@@ -1017,15 +1068,35 @@ static void intReset() {
 	memset(&ICache, 0xff, sizeof(ICache));
 }
 
-void intExecute() {
+static inline void execI_(u8 **memRLUT, psxRegisters *regs_) {
+	regs_->code = fetch(memRLUT, regs_->pc);
+
+	debugI();
+
+	if (Config.Debug) ProcessDebug();
+
+	regs_->pc += 4;
+	regs_->cycle += BIAS;
+
+	psxBSC[regs_->code >> 26](regs_, regs_->code);
+}
+
+static void intExecute() {
+	psxRegisters *regs_ = &psxRegs;
+	u8 **memRLUT = psxMemRLUT;
 	extern int stop;
-	for (;!stop;) 
-		execI();
+
+	while (!stop)
+		execI_(memRLUT, regs_);
 }
 
-void intExecuteBlock() {
+static void intExecuteBlock() {
+	psxRegisters *regs_ = &psxRegs;
+	u8 **memRLUT = psxMemRLUT;
+
 	branch2 = 0;
-	while (!branch2) execI();
+	while (!branch2)
+		execI_(memRLUT, regs_);
 }
 
 static void intClear(u32 Addr, u32 Size) {
@@ -1083,18 +1154,9 @@ void intApplyConfig() {
 static void intShutdown() {
 }
 
-// interpreter execution
+// single step (may do several ops in case of a branch)
 void execI() {
-	psxRegs.code = fetch(psxRegs.pc);
-
-	debugI();
-
-	if (Config.Debug) ProcessDebug();
-
-	psxRegs.pc += 4;
-	psxRegs.cycle += BIAS;
-
-	psxBSC[psxRegs.code >> 26]();
+	execI_(psxMemRLUT, &psxRegs);
 }
 
 R3000Acpu psxInt = {
diff --git a/libpcsxcore/psxinterpreter.h b/libpcsxcore/psxinterpreter.h
index 89dd7ea1..87c7a324 100644
--- a/libpcsxcore/psxinterpreter.h
+++ b/libpcsxcore/psxinterpreter.h
@@ -1,7 +1,11 @@
-
-extern u32 (*fetch)(u32 pc);
+#ifndef __PSXINTERPRETER_H__
+#define __PSXINTERPRETER_H__
 
 // called by "new_dynarec"
 void execI();
-void psxNULL();
 void intApplyConfig();
+void MTC0(psxRegisters *regs_, int reg, u32 val);
+void gteNULL(struct psxCP2Regs *regs);
+void (*psxCP2[64])(struct psxCP2Regs *regs);
+
+#endif // __PSXINTERPRETER_H__
diff --git a/libpcsxcore/r3000a.c b/libpcsxcore/r3000a.c
index a9b800e3..ba22d45f 100644
--- a/libpcsxcore/r3000a.c
+++ b/libpcsxcore/r3000a.c
@@ -83,15 +83,14 @@ void psxShutdown() {
 }
 
 void psxException(u32 code, u32 bd) {
-	psxRegs.code = fetch(psxRegs.pc);
+	psxRegs.code = PSXMu32(psxRegs.pc);
 	
 	if (!Config.HLE && ((((psxRegs.code) >> 24) & 0xfe) == 0x4a)) {
 		// "hokuto no ken" / "Crash Bandicot 2" ...
 		// BIOS does not allow to return to GTE instructions
 		// (just skips it, supposedly because it's scheduled already)
 		// so we execute it here
-		extern void (*psxCP2[64])(void *cp2regs);
-		psxCP2[psxRegs.code & 0x3f](&psxRegs.CP2D);
+		psxCP2[psxRegs.code & 0x3f](&psxRegs.CP2);
 	}
 
 	// Set the Cause
diff --git a/libpcsxcore/r3000a.h b/libpcsxcore/r3000a.h
index 49afcb1c..f99e03ba 100644
--- a/libpcsxcore/r3000a.h
+++ b/libpcsxcore/r3000a.h
@@ -219,85 +219,12 @@ void new_dyna_freeze(void *f, int mode);
 #define new_dyna_set_event(e, c) \
 	new_dyna_set_event_abs(e, psxRegs.cycle + (c))
 
-#if __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
-
-#define _i32(x) *(s32 *)&x
-#define _u32(x) x
-
-#define _i16(x) (((short *)&x)[1])
-#define _u16(x) (((unsigned short *)&x)[1])
-
-#define _i8(x) (((char *)&x)[3])
-#define _u8(x) (((unsigned char *)&x)[3])
-
-#else
-
-#define _i32(x) *(s32 *)&x
-#define _u32(x) x
-
-#define _i16(x) *(short *)&x
-#define _u16(x) *(unsigned short *)&x
-
-#define _i8(x) *(char *)&x
-#define _u8(x) *(unsigned char *)&x
-
-#endif
-
-/**** R3000A Instruction Macros ****/
-#define _PC_       psxRegs.pc       // The next PC to be executed
-
-#define _fOp_(code)		((code >> 26)       )  // The opcode part of the instruction register 
-#define _fFunct_(code)	((code      ) & 0x3F)  // The funct part of the instruction register 
-#define _fRd_(code)		((code >> 11) & 0x1F)  // The rd part of the instruction register 
-#define _fRt_(code)		((code >> 16) & 0x1F)  // The rt part of the instruction register 
-#define _fRs_(code)		((code >> 21) & 0x1F)  // The rs part of the instruction register 
-#define _fSa_(code)		((code >>  6) & 0x1F)  // The sa part of the instruction register
-#define _fIm_(code)		((u16)code)            // The immediate part of the instruction register
-#define _fTarget_(code)	(code & 0x03ffffff)    // The target part of the instruction register
-
-#define _fImm_(code)	((s16)code)            // sign-extended immediate
-#define _fImmU_(code)	(code&0xffff)          // zero-extended immediate
-
-#define _Op_     _fOp_(psxRegs.code)
-#define _Funct_  _fFunct_(psxRegs.code)
-#define _Rd_     _fRd_(psxRegs.code)
-#define _Rt_     _fRt_(psxRegs.code)
-#define _Rs_     _fRs_(psxRegs.code)
-#define _Sa_     _fSa_(psxRegs.code)
-#define _Im_     _fIm_(psxRegs.code)
-#define _Target_ _fTarget_(psxRegs.code)
-
-#define _Imm_	 _fImm_(psxRegs.code)
-#define _ImmU_	 _fImmU_(psxRegs.code)
-
-#define _rRs_   psxRegs.GPR.r[_Rs_]   // Rs register
-#define _rRt_   psxRegs.GPR.r[_Rt_]   // Rt register
-#define _rRd_   psxRegs.GPR.r[_Rd_]   // Rd register
-#define _rSa_   psxRegs.GPR.r[_Sa_]   // Sa register
-#define _rFs_   psxRegs.CP0.r[_Rd_]   // Fs register
-
-#define _c2dRs_ psxRegs.CP2D.r[_Rs_]  // Rs cop2 data register
-#define _c2dRt_ psxRegs.CP2D.r[_Rt_]  // Rt cop2 data register
-#define _c2dRd_ psxRegs.CP2D.r[_Rd_]  // Rd cop2 data register
-#define _c2dSa_ psxRegs.CP2D.r[_Sa_]  // Sa cop2 data register
-
-#define _rHi_   psxRegs.GPR.n.hi   // The HI register
-#define _rLo_   psxRegs.GPR.n.lo   // The LO register
-
-#define _JumpTarget_    ((_Target_ * 4) + (_PC_ & 0xf0000000))   // Calculates the target during a jump instruction
-#define _BranchTarget_  ((s16)_Im_ * 4 + _PC_)                 // Calculates the target during a branch instruction
-
-#define _SetLink(x)     psxRegs.GPR.r[x] = _PC_ + 4;       // Sets the return address in the link register
-
 int  psxInit();
 void psxReset();
 void psxShutdown();
 void psxException(u32 code, u32 bd);
 void psxBranchTest();
 void psxExecuteBios();
-int  psxTestLoadDelay(int reg, u32 tmp);
-void psxDelayTest(int reg, u32 bpc);
-void psxTestSWInts();
 void psxJumpTest();
 
 #ifdef __cplusplus