Merge pull request #6989 from hrydgard/x86-emitter-merge

Merge from Dolphin's x86-64 emitter
This commit is contained in:
Henrik Rydgård 2014-10-12 19:52:59 +02:00
commit 6cb2c9c97d
12 changed files with 1254 additions and 737 deletions

View File

@ -49,6 +49,17 @@ void do_cpuid(u32 regs[4], u32 cpuid_leaf) {
#ifdef _M_SSE
#include <xmmintrin.h>
#define _XCR_XFEATURE_ENABLED_MASK 0
static unsigned long long _xgetbv(unsigned int index)
{
unsigned int eax, edx;
__asm__ __volatile__("xgetbv" : "=a"(eax), "=d"(edx) : "c"(index));
return ((unsigned long long)edx << 32) | eax;
}
#else
#define _XCR_XFEATURE_ENABLED_MASK 0
#endif
#if defined __FreeBSD__
@ -172,6 +183,38 @@ void CPUInfo::Detect() {
bFMA = true;
}
if ((cpu_id[2] >> 25) & 1) bAES = true;
if ((cpu_id[3] >> 24) & 1)
{
// We can use FXSAVE.
bFXSR = true;
}
// AVX support requires 3 separate checks:
// - Is the AVX bit set in CPUID?
// - Is the XSAVE bit set in CPUID?
// - XGETBV result has the XCR bit set.
if (((cpu_id[2] >> 28) & 1) && ((cpu_id[2] >> 27) & 1))
{
if ((_xgetbv(_XCR_XFEATURE_ENABLED_MASK) & 0x6) == 0x6)
{
bAVX = true;
if ((cpu_id[2] >> 12) & 1)
bFMA = true;
}
}
if (max_std_fn >= 7)
{
do_cpuid(cpu_id, 0x00000007);
// careful; we can't enable AVX2 unless the XSAVE/XGETBV checks above passed
if ((cpu_id[1] >> 5) & 1)
bAVX2 = bAVX;
if ((cpu_id[1] >> 3) & 1)
bBMI1 = true;
if ((cpu_id[1] >> 8) & 1)
bBMI2 = true;
}
}
if (max_ex_fn >= 0x80000004) {
// Extract brand string

View File

@ -56,10 +56,15 @@ struct CPUInfo {
bool bLZCNT;
bool bSSE4A;
bool bAVX;
bool bAVX2;
bool bFMA;
bool bAES;
bool bLAHFSAHF64;
bool bLongMode;
bool bBMI1;
bool bBMI2;
bool bMOVBE;
bool bFXSR;
// ARM specific CPUInfo
bool bSwp;

File diff suppressed because it is too large Load Diff

View File

@ -22,8 +22,8 @@
#include "Common.h"
#if !defined(_M_IX86) && !defined(_M_X64)
#error "Don't build this on arm."
#ifdef _M_X64
#define _ARCH_64
#endif
#ifdef _M_X64
@ -39,7 +39,7 @@ enum X64Reg
{
EAX = 0, EBX = 3, ECX = 1, EDX = 2,
ESI = 6, EDI = 7, EBP = 5, ESP = 4,
RAX = 0, RBX = 3, RCX = 1, RDX = 2,
RSI = 6, RDI = 7, RBP = 5, RSP = 4,
R8 = 8, R9 = 9, R10 = 10,R11 = 11,
@ -52,9 +52,12 @@ enum X64Reg
AX = 0, BX = 3, CX = 1, DX = 2,
SI = 6, DI = 7, BP = 5, SP = 4,
XMM0=0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
XMM0=0, XMM1, XMM2, XMM3, XMM4, XMM5, XMM6, XMM7,
XMM8, XMM9, XMM10, XMM11, XMM12, XMM13, XMM14, XMM15,
YMM0=0, YMM1, YMM2, YMM3, YMM4, YMM5, YMM6, YMM7,
YMM8, YMM9, YMM10, YMM11, YMM12, YMM13, YMM14, YMM15,
INVALID_REG = 0xFFFFFFFF
};
@ -62,10 +65,10 @@ enum CCFlags
{
CC_O = 0,
CC_NO = 1,
CC_B = 2, CC_C = 2, CC_NAE = 2,
CC_NB = 3, CC_NC = 3, CC_AE = 3,
CC_B = 2, CC_C = 2, CC_NAE = 2,
CC_NB = 3, CC_NC = 3, CC_AE = 3,
CC_Z = 4, CC_E = 4,
CC_NZ = 5, CC_NE = 5,
CC_NZ = 5, CC_NE = 5,
CC_BE = 6, CC_NA = 6,
CC_NBE = 7, CC_A = 7,
CC_S = 8,
@ -117,8 +120,7 @@ enum NormalOp {
nrmXCHG,
};
enum
{
enum {
CMP_EQ = 0,
CMP_LT = 1,
CMP_LE = 2,
@ -129,8 +131,19 @@ enum
CMP_ORD = 7,
};
enum FloatOp {
floatLD = 0,
floatST = 2,
floatSTP = 3,
floatLD80 = 5,
floatSTP80 = 7,
floatINVALID = -1,
};
class XEmitter;
// RIP addressing does not benefit from micro op fusion on Core arch
struct OpArg
{
OpArg() {} // dummy op arg, used for storage
@ -140,11 +153,18 @@ struct OpArg
scale = (u8)_scale;
offsetOrBaseReg = (u16)rmReg;
indexReg = (u16)scaledReg;
//if scale == 0 never mind offseting
//if scale == 0 never mind offsetting
offset = _offset;
}
bool operator==(OpArg b)
{
return operandReg == b.operandReg && scale == b.scale && offsetOrBaseReg == b.offsetOrBaseReg &&
indexReg == b.indexReg && offset == b.offset;
}
void WriteRex(XEmitter *emit, int opBits, int bits, int customOp = -1) const;
void WriteRest(XEmitter *emit, int extraBytes=0, X64Reg operandReg=(X64Reg)0xFF, bool warn_64bit_offset = true) const;
void WriteVex(XEmitter* emit, X64Reg regOp1, X64Reg regOp2, int L, int pp, int mmmmm, int W = 0) const;
void WriteRest(XEmitter *emit, int extraBytes=0, X64Reg operandReg=INVALID_REG, bool warn_64bit_offset = true) const;
void WriteFloatModRM(XEmitter *emit, FloatOp op);
void WriteSingleByteOp(XEmitter *emit, u8 op, X64Reg operandReg, int bits);
// This one is public - must be written to
u64 offset; // use RIP-relative as much as possible - 64-bit immediates are not available.
@ -153,7 +173,10 @@ struct OpArg
void WriteNormalOp(XEmitter *emit, bool toRM, NormalOp op, const OpArg &operand, int bits) const;
bool IsImm() const {return scale == SCALE_IMM8 || scale == SCALE_IMM16 || scale == SCALE_IMM32 || scale == SCALE_IMM64;}
bool IsSimpleReg() const {return scale == SCALE_NONE;}
bool IsSimpleReg(X64Reg reg) const {
bool IsSimpleReg(X64Reg reg) const
{
if (!IsSimpleReg())
return false;
return GetSimpleReg() == reg;
}
@ -192,6 +215,7 @@ struct OpArg
void IncreaseOffset(int sz) {
offset += sz;
}
private:
u8 scale;
u16 offsetOrBaseReg;
@ -203,21 +227,30 @@ template <typename T>
inline OpArg M(const T *ptr) {return OpArg((u64)(const void *)ptr, (int)SCALE_RIP);}
inline OpArg R(X64Reg value) {return OpArg(0, SCALE_NONE, value);}
inline OpArg MatR(X64Reg value) {return OpArg(0, SCALE_ATREG, value);}
inline OpArg MDisp(X64Reg value, int offset) {
inline OpArg MDisp(X64Reg value, int offset)
{
return OpArg((u32)offset, SCALE_ATREG, value);
}
inline OpArg MComplex(X64Reg base, X64Reg scaled, int scale, int offset) {
inline OpArg MComplex(X64Reg base, X64Reg scaled, int scale, int offset)
{
return OpArg(offset, scale, base, scaled);
}
inline OpArg MScaled(X64Reg scaled, int scale, int offset) {
inline OpArg MScaled(X64Reg scaled, int scale, int offset)
{
if (scale == SCALE_1)
return OpArg(offset, SCALE_ATREG, scaled);
else
return OpArg(offset, scale | 0x20, RAX, scaled);
}
inline OpArg MRegSum(X64Reg base, X64Reg offset) {
inline OpArg MRegSum(X64Reg base, X64Reg offset)
{
return MComplex(base, offset, 1, 0);
}
inline OpArg Imm8 (u8 imm) {return OpArg(imm, SCALE_IMM8);}
inline OpArg Imm16(u16 imm) {return OpArg(imm, SCALE_IMM16);} //rarely used
inline OpArg Imm32(u32 imm) {return OpArg(imm, SCALE_IMM32);}
@ -229,19 +262,23 @@ inline OpArg SImmAuto(s32 imm) {
return OpArg(imm, (imm >= 128 || imm < -128) ? SCALE_IMM32 : SCALE_IMM8);
}
#ifdef _M_X64
inline OpArg ImmPtr(const void *imm) {return Imm64((u64)imm);}
#ifdef _ARCH_64
inline OpArg ImmPtr(const void* imm) {return Imm64((u64)imm);}
#else
inline OpArg ImmPtr(const void *imm) {return Imm32((u32)imm);}
inline OpArg ImmPtr(const void* imm) {return Imm32((u32)imm);}
#endif
inline u32 PtrOffset(const void *ptr, const void *base) {
#ifdef _M_X64
inline u32 PtrOffset(const void* ptr, const void* base)
{
#ifdef _ARCH_64
s64 distance = (s64)ptr-(s64)base;
if (distance >= 0x80000000LL ||
distance < -0x80000000LL) {
_assert_msg_(JIT, 0, "pointer offset out of range");
distance < -0x80000000LL)
{
_assert_msg_(DYNA_REC, 0, "pointer offset out of range");
return 0;
}
return (u32)distance;
#else
return (u32)ptr-(u32)base;
@ -259,6 +296,18 @@ struct FixupBranch
int type; //0 = 8bit 1 = 32bit
};
enum SSECompare
{
EQ = 0,
LT,
LE,
UNORD,
NEQ,
NLT,
NLE,
ORD,
};
typedef const u8* JumpTarget;
class XEmitter
@ -266,33 +315,40 @@ class XEmitter
friend struct OpArg; // for Write8 etc
private:
u8 *code;
bool flags_locked;
void CheckFlags();
void Rex(int w, int r, int x, int b);
void WriteSimple1Byte(int bits, u8 byte, X64Reg reg);
void WriteSimple2Byte(int bits, u8 byte1, u8 byte2, X64Reg reg);
void WriteMulDivType(int bits, OpArg src, int ext);
void WriteBitSearchType(int bits, X64Reg dest, OpArg src, u8 byte2);
void WriteBitSearchType(int bits, X64Reg dest, OpArg src, u8 byte2, bool rep = false);
void WriteShift(int bits, OpArg dest, OpArg &shift, int ext);
void WriteBitTest(int bits, OpArg &dest, OpArg &index, int ext);
void WriteMXCSR(OpArg arg, int ext);
void WriteSSEOp(int size, u8 sseOp, bool packed, X64Reg regOp, OpArg arg, int extrabytes = 0);
void WriteSSEOp2(int size, u8 sseOp, bool packed, X64Reg regOp, OpArg arg, int extrabytes = 0);
void WriteSSEOp(u8 opPrefix, u16 op, X64Reg regOp, OpArg arg, int extrabytes = 0);
void WriteSSSE3Op(u8 opPrefix, u16 op, X64Reg regOp, OpArg arg, int extrabytes = 0);
void WriteSSE41Op(u8 opPrefix, u16 op, X64Reg regOp, OpArg arg, int extrabytes = 0);
void WriteAVXOp(u8 opPrefix, u16 op, X64Reg regOp, OpArg arg, int extrabytes = 0);
void WriteAVXOp(u8 opPrefix, u16 op, X64Reg regOp1, X64Reg regOp2, OpArg arg, int extrabytes = 0);
void WriteVEXOp(int size, u8 opPrefix, u16 op, X64Reg regOp1, X64Reg regOp2, OpArg arg, int extrabytes = 0);
void WriteBMI1Op(int size, u8 opPrefix, u16 op, X64Reg regOp1, X64Reg regOp2, OpArg arg, int extrabytes = 0);
void WriteBMI2Op(int size, u8 opPrefix, u16 op, X64Reg regOp1, X64Reg regOp2, OpArg arg, int extrabytes = 0);
void WriteFloatLoadStore(int bits, FloatOp op, FloatOp op_80b, OpArg arg);
void WriteNormalOp(XEmitter *emit, int bits, NormalOp op, const OpArg &a1, const OpArg &a2);
void ABI_CalculateFrameSize(u32 mask, size_t rsp_alignment, size_t needed_frame_size, size_t* shadowp, size_t* subtractionp, size_t* xmm_offsetp);
protected:
inline void Write8(u8 value) {
//if (value == 0xcc) {
// value = 0xcc; // set breakpoint here to find where mysterious 0xcc are written
//}
*code++ = value;
}
inline void Write8(u8 value) {*code++ = value;}
inline void Write16(u16 value) {*(u16*)code = (value); code += 2;}
inline void Write32(u32 value) {*(u32*)code = (value); code += 4;}
inline void Write64(u64 value) {*(u64*)code = (value); code += 8;}
public:
XEmitter() { code = NULL; }
XEmitter(u8 *code_ptr) { code = code_ptr; }
XEmitter() { code = nullptr; flags_locked = false; }
XEmitter(u8 *code_ptr) { code = code_ptr; flags_locked = false; }
virtual ~XEmitter() {}
void WriteModRM(int mod, int rm, int reg);
@ -306,8 +362,11 @@ public:
const u8 *GetCodePtr() const;
u8 *GetWritableCodePtr();
void LockFlags() { flags_locked = true; }
void UnlockFlags() { flags_locked = false; }
// Looking for one of these? It's BANNED!! Some instructions are slow on modern CPU
// INC, DEC, LOOP, LOOPNE, LOOPE, ENTER, LEAVE, XCHG, XLAT, REP MOVSB/MOVSD, REP SCASD + other string instr.,
// INC, DEC, LOOP, LOOPNE, LOOPE, ENTER, LEAVE, XCHG, XLAT, REP MOVSB/MOVSD, REP SCASD + other string instr.,
// INC and DEC are slow on Intel Core, but not on AMD. They create a
// false flag dependency because they only update a subset of the flags.
// XCHG is SLOW and should be avoided.
@ -316,7 +375,7 @@ public:
void INT3();
// Do nothing
void NOP(int count = 1); //nop padding - TODO: fast nop slides, for amd and intel (check their manuals)
void NOP(size_t count = 1);
// Save energy in wait-loops on P4 only. Probably not too useful.
void PAUSE();
@ -396,7 +455,7 @@ public:
void DIV(int bits, OpArg src);
void IDIV(int bits, OpArg src);
// Shift
// Shift
void ROL(int bits, OpArg dest, OpArg shift);
void ROR(int bits, OpArg dest, OpArg shift);
void RCL(int bits, OpArg dest, OpArg shift);
@ -451,7 +510,15 @@ public:
// Sign/zero extension
void MOVSX(int dbits, int sbits, X64Reg dest, OpArg src); //automatically uses MOVSXD if necessary
void MOVZX(int dbits, int sbits, X64Reg dest, OpArg src);
void MOVZX(int dbits, int sbits, X64Reg dest, OpArg src);
// Available only on Atom or >= Haswell so far. Test with cpu_info.bMOVBE.
void MOVBE(int dbits, const OpArg& dest, const OpArg& src);
// Available only on AMD >= Phenom or Intel >= Haswell
void LZCNT(int bits, X64Reg dest, OpArg src);
// Note: this one is actually part of BMI1
void TZCNT(int bits, X64Reg dest, OpArg src);
// WARNING - These two take 11-13 cycles and are VectorPath! (AMD64)
void STMXCSR(OpArg memloc);
@ -461,29 +528,53 @@ public:
void LOCK();
void REP();
void REPNE();
void FSOverride();
void GSOverride();
// x87
enum x87StatusWordBits {
x87_InvalidOperation = 0x1,
x87_DenormalizedOperand = 0x2,
x87_DivisionByZero = 0x4,
x87_Overflow = 0x8,
x87_Underflow = 0x10,
x87_Precision = 0x20,
x87_StackFault = 0x40,
x87_ErrorSummary = 0x80,
x87_C0 = 0x100,
x87_C1 = 0x200,
x87_C2 = 0x400,
x87_TopOfStack = 0x2000 | 0x1000 | 0x800,
x87_C3 = 0x4000,
x87_FPUBusy = 0x8000,
};
void FLD(int bits, OpArg src);
void FST(int bits, OpArg dest);
void FSTP(int bits, OpArg dest);
void FNSTSW_AX();
void FWAIT();
// SSE/SSE2: Floating point arithmetic
void ADDSS(X64Reg regOp, OpArg arg);
void ADDSD(X64Reg regOp, OpArg arg);
void SUBSS(X64Reg regOp, OpArg arg);
void SUBSD(X64Reg regOp, OpArg arg);
void MULSS(X64Reg regOp, OpArg arg);
void MULSD(X64Reg regOp, OpArg arg);
void DIVSS(X64Reg regOp, OpArg arg);
void DIVSD(X64Reg regOp, OpArg arg);
void MINSS(X64Reg regOp, OpArg arg);
void MINSD(X64Reg regOp, OpArg arg);
void MAXSS(X64Reg regOp, OpArg arg);
void MAXSD(X64Reg regOp, OpArg arg);
void SQRTSS(X64Reg regOp, OpArg arg);
void SQRTSD(X64Reg regOp, OpArg arg);
void ADDSS(X64Reg regOp, OpArg arg);
void ADDSD(X64Reg regOp, OpArg arg);
void SUBSS(X64Reg regOp, OpArg arg);
void SUBSD(X64Reg regOp, OpArg arg);
void MULSS(X64Reg regOp, OpArg arg);
void MULSD(X64Reg regOp, OpArg arg);
void DIVSS(X64Reg regOp, OpArg arg);
void DIVSD(X64Reg regOp, OpArg arg);
void MINSS(X64Reg regOp, OpArg arg);
void MINSD(X64Reg regOp, OpArg arg);
void MAXSS(X64Reg regOp, OpArg arg);
void MAXSD(X64Reg regOp, OpArg arg);
void SQRTSS(X64Reg regOp, OpArg arg);
void SQRTSD(X64Reg regOp, OpArg arg);
void RSQRTSS(X64Reg regOp, OpArg arg);
// SSE/SSE2: Floating point bitwise (yes)
void CMPSS(X64Reg regOp, OpArg arg, u8 compare);
void CMPSD(X64Reg regOp, OpArg arg, u8 compare);
void CMPSS(X64Reg regOp, OpArg arg, u8 compare);
void CMPSD(X64Reg regOp, OpArg arg, u8 compare);
inline void CMPEQSS(X64Reg regOp, OpArg arg) { CMPSS(regOp, arg, CMP_EQ); }
inline void CMPLTSS(X64Reg regOp, OpArg arg) { CMPSS(regOp, arg, CMP_LT); }
@ -493,24 +584,12 @@ public:
inline void CMPNLTSS(X64Reg regOp, OpArg arg) { CMPSS(regOp, arg, CMP_NLT); }
inline void CMPORDSS(X64Reg regOp, OpArg arg) { CMPSS(regOp, arg, CMP_ORD); }
// I don't think these exist
/*
void ANDSD(X64Reg regOp, OpArg arg);
void ANDNSS(X64Reg regOp, OpArg arg);
void ANDNSD(X64Reg regOp, OpArg arg);
void ORSS(X64Reg regOp, OpArg arg);
void ORSD(X64Reg regOp, OpArg arg);
void XORSS(X64Reg regOp, OpArg arg);
void XORSD(X64Reg regOp, OpArg arg);
*/
// SSE/SSE2: Floating point packed arithmetic (x4 for float, x2 for double)
void ADDPS(X64Reg regOp, OpArg arg);
void ADDPD(X64Reg regOp, OpArg arg);
void SUBPS(X64Reg regOp, OpArg arg);
void SUBPD(X64Reg regOp, OpArg arg);
void CMPPS(X64Reg regOp, OpArg arg, u8 compare);
void ADDPS(X64Reg regOp, OpArg arg);
void ADDPD(X64Reg regOp, OpArg arg);
void SUBPS(X64Reg regOp, OpArg arg);
void SUBPD(X64Reg regOp, OpArg arg);
void CMPPS(X64Reg regOp, OpArg arg, u8 compare);
void CMPPD(X64Reg regOp, OpArg arg, u8 compare);
void MULPS(X64Reg regOp, OpArg arg);
void MULPD(X64Reg regOp, OpArg arg);
@ -525,8 +604,8 @@ public:
void RSQRTPS(X64Reg regOp, OpArg arg);
// SSE/SSE2: Floating point packed bitwise (x4 for float, x2 for double)
void ANDPS(X64Reg regOp, OpArg arg);
void ANDPD(X64Reg regOp, OpArg arg);
void ANDPS(X64Reg regOp, OpArg arg);
void ANDPD(X64Reg regOp, OpArg arg);
void ANDNPS(X64Reg regOp, OpArg arg);
void ANDNPD(X64Reg regOp, OpArg arg);
void ORPS(X64Reg regOp, OpArg arg);
@ -535,17 +614,14 @@ public:
void XORPD(X64Reg regOp, OpArg arg);
// SSE/SSE2: Shuffle components. These are tricky - see Intel documentation.
void SHUFPS(X64Reg regOp, OpArg arg, u8 shuffle);
void SHUFPD(X64Reg regOp, OpArg arg, u8 shuffle);
void SHUFPS(X64Reg regOp, OpArg arg, u8 shuffle);
void SHUFPD(X64Reg regOp, OpArg arg, u8 shuffle);
// SSE/SSE2: Useful alternative to shuffle in some cases.
void MOVDDUP(X64Reg regOp, OpArg arg);
// THESE TWO ARE NEW AND UNTESTED
void UNPCKLPS(X64Reg dest, OpArg src);
void UNPCKHPS(X64Reg dest, OpArg src);
// These are OK.
void UNPCKLPD(X64Reg dest, OpArg src);
void UNPCKHPD(X64Reg dest, OpArg src);
@ -555,19 +631,17 @@ public:
void UCOMISS(X64Reg regOp, OpArg arg);
void UCOMISD(X64Reg regOp, OpArg arg);
// SSE/SSE2: Moves. Use the right data type for your data to avoid slight penalties on some CPUs.
// Singles
// SSE/SSE2: Moves. Use the right data type for your data, in most cases.
void MOVAPS(X64Reg regOp, OpArg arg);
void MOVAPS(OpArg arg, X64Reg regOp);
void MOVUPS(X64Reg regOp, OpArg arg);
void MOVUPS(OpArg arg, X64Reg regOp);
// Doubles
void MOVAPD(X64Reg regOp, OpArg arg);
void MOVAPS(OpArg arg, X64Reg regOp);
void MOVAPD(OpArg arg, X64Reg regOp);
void MOVUPS(X64Reg regOp, OpArg arg);
void MOVUPD(X64Reg regOp, OpArg arg);
void MOVUPS(OpArg arg, X64Reg regOp);
void MOVUPD(OpArg arg, X64Reg regOp);
// Integers (NOTE: untested - I added these then it turned out I didn't have a use for them after all).
void MOVDQA(X64Reg regOp, OpArg arg);
void MOVDQA(OpArg arg, X64Reg regOp);
void MOVDQU(X64Reg regOp, OpArg arg);
@ -578,6 +652,14 @@ public:
void MOVSS(OpArg arg, X64Reg regOp);
void MOVSD(OpArg arg, X64Reg regOp);
void MOVLPD(X64Reg regOp, OpArg arg);
void MOVHPD(X64Reg regOp, OpArg arg);
void MOVLPD(OpArg arg, X64Reg regOp);
void MOVHPD(OpArg arg, X64Reg regOp);
void MOVHLPS(X64Reg regOp1, X64Reg regOp2);
void MOVLHPS(X64Reg regOp1, X64Reg regOp2);
void MOVD_xmm(X64Reg dest, const OpArg &arg);
void MOVQ_xmm(X64Reg dest, OpArg arg);
void MOVD_xmm(const OpArg &arg, X64Reg src);
@ -595,112 +677,161 @@ public:
void CVTPS2PD(X64Reg dest, OpArg src);
void CVTPD2PS(X64Reg dest, OpArg src);
void CVTSS2SD(X64Reg dest, OpArg src);
void CVTSI2SS(X64Reg dest, OpArg src);
void CVTSD2SS(X64Reg dest, OpArg src);
void CVTSD2SI(X64Reg dest, OpArg src);
void CVTSI2SD(X64Reg dest, OpArg src);
void CVTDQ2PD(X64Reg regOp, OpArg arg);
void CVTPD2DQ(X64Reg regOp, OpArg arg);
void CVTDQ2PS(X64Reg regOp, OpArg arg);
void CVTPS2DQ(X64Reg regOp, OpArg arg);
void CVTSI2SS(X64Reg xregdest, OpArg arg); // Yeah, destination really is a GPR like EAX!
void CVTSS2SI(X64Reg xregdest, OpArg arg); // Yeah, destination really is a GPR like EAX!
void CVTTSS2SI(X64Reg xregdest, OpArg arg); // Yeah, destination really is a GPR like EAX!
void CVTTSD2SI(X64Reg xregdest, OpArg arg); // Yeah, destination really is a GPR like EAX!
void CVTTPS2DQ(X64Reg regOp, OpArg arg);
void CVTTPD2DQ(X64Reg xregdest, OpArg arg);
void CVTTPD2DQ(X64Reg regOp, OpArg arg);
// Destinations are X64 regs (rax, rbx, ...) for these instructions.
void CVTSS2SI(X64Reg xregdest, OpArg src);
void CVTSD2SI(X64Reg xregdest, OpArg src);
void CVTTSS2SI(X64Reg xregdest, OpArg arg);
void CVTTSD2SI(X64Reg xregdest, OpArg arg);
// SSE2: Packed integer instructions
void PACKSSDW(X64Reg dest, OpArg arg);
void PACKSSWB(X64Reg dest, OpArg arg);
//void PACKUSDW(X64Reg dest, OpArg arg);
void PACKUSDW(X64Reg dest, OpArg arg);
void PACKUSWB(X64Reg dest, OpArg arg);
void PUNPCKLBW(X64Reg dest, const OpArg &arg);
void PUNPCKLWD(X64Reg dest, const OpArg &arg);
void PUNPCKLDQ(X64Reg dest, const OpArg &arg);
void PMOVSXBW(X64Reg dest, const OpArg &arg);
void PMOVSXBD(X64Reg dest, const OpArg &arg);
void PMOVSXWD(X64Reg dest, const OpArg &arg);
void PMOVZXBW(X64Reg dest, const OpArg &arg);
void PMOVZXBD(X64Reg dest, const OpArg &arg);
void PMOVZXWD(X64Reg dest, const OpArg &arg);
void PTEST(X64Reg dest, OpArg arg);
void PAND(X64Reg dest, OpArg arg);
void PANDN(X64Reg dest, OpArg arg);
void PXOR(X64Reg dest, OpArg arg);
void POR(X64Reg dest, OpArg arg);
void PANDN(X64Reg dest, OpArg arg);
void PXOR(X64Reg dest, OpArg arg);
void POR(X64Reg dest, OpArg arg);
void PADDB(X64Reg dest, OpArg arg);
void PADDW(X64Reg dest, OpArg arg);
void PADDD(X64Reg dest, OpArg arg);
void PADDQ(X64Reg dest, OpArg arg);
void PADDW(X64Reg dest, OpArg arg);
void PADDD(X64Reg dest, OpArg arg);
void PADDQ(X64Reg dest, OpArg arg);
void PADDSB(X64Reg dest, OpArg arg);
void PADDSW(X64Reg dest, OpArg arg);
void PADDUSB(X64Reg dest, OpArg arg);
void PADDUSW(X64Reg dest, OpArg arg);
void PADDSB(X64Reg dest, OpArg arg);
void PADDSW(X64Reg dest, OpArg arg);
void PADDUSB(X64Reg dest, OpArg arg);
void PADDUSW(X64Reg dest, OpArg arg);
void PSUBB(X64Reg dest, OpArg arg);
void PSUBW(X64Reg dest, OpArg arg);
void PSUBD(X64Reg dest, OpArg arg);
void PSUBQ(X64Reg dest, OpArg arg);
void PSUBB(X64Reg dest, OpArg arg);
void PSUBW(X64Reg dest, OpArg arg);
void PSUBD(X64Reg dest, OpArg arg);
void PSUBQ(X64Reg dest, OpArg arg);
void PSUBSB(X64Reg dest, OpArg arg);
void PSUBSW(X64Reg dest, OpArg arg);
void PSUBUSB(X64Reg dest, OpArg arg);
void PSUBUSW(X64Reg dest, OpArg arg);
void PSUBSB(X64Reg dest, OpArg arg);
void PSUBSW(X64Reg dest, OpArg arg);
void PSUBUSB(X64Reg dest, OpArg arg);
void PSUBUSW(X64Reg dest, OpArg arg);
void PAVGB(X64Reg dest, OpArg arg);
void PAVGW(X64Reg dest, OpArg arg);
void PAVGB(X64Reg dest, OpArg arg);
void PAVGW(X64Reg dest, OpArg arg);
void PCMPEQB(X64Reg dest, OpArg arg);
void PCMPEQW(X64Reg dest, OpArg arg);
void PCMPEQD(X64Reg dest, OpArg arg);
void PCMPEQB(X64Reg dest, OpArg arg);
void PCMPEQW(X64Reg dest, OpArg arg);
void PCMPEQD(X64Reg dest, OpArg arg);
void PCMPGTB(X64Reg dest, OpArg arg);
void PCMPGTW(X64Reg dest, OpArg arg);
void PCMPGTD(X64Reg dest, OpArg arg);
void PCMPGTB(X64Reg dest, OpArg arg);
void PCMPGTW(X64Reg dest, OpArg arg);
void PCMPGTD(X64Reg dest, OpArg arg);
void PEXTRW(X64Reg dest, OpArg arg, u8 subreg);
void PINSRW(X64Reg dest, OpArg arg, u8 subreg);
void PMADDWD(X64Reg dest, OpArg arg);
void PSADBW(X64Reg dest, OpArg arg);
void PMADDWD(X64Reg dest, OpArg arg);
void PSADBW(X64Reg dest, OpArg arg);
void PMAXSW(X64Reg dest, OpArg arg);
void PMAXUB(X64Reg dest, OpArg arg);
void PMINSW(X64Reg dest, OpArg arg);
void PMINUB(X64Reg dest, OpArg arg);
void PMAXSW(X64Reg dest, OpArg arg);
void PMAXUB(X64Reg dest, OpArg arg);
void PMINSW(X64Reg dest, OpArg arg);
void PMINUB(X64Reg dest, OpArg arg);
// SSE4 has PMAXSB and PMINSB and PMAXUW and PMINUW too if we need them.
void PMOVMSKB(X64Reg dest, OpArg arg);
void PSHUFD(X64Reg dest, OpArg arg, u8 shuffle);
void PSHUFB(X64Reg dest, OpArg arg);
void PSHUFD(X64Reg dest, OpArg arg, u8 shuffle);
void PSHUFLW(X64Reg dest, OpArg arg, u8 shuffle);
void PSHUFHW(X64Reg dest, OpArg arg, u8 shuffle);
void PSRLW(X64Reg reg, int shift);
void PSRLD(X64Reg reg, int shift);
void PSRLQ(X64Reg reg, int shift);
void PSRLQ(X64Reg reg, OpArg arg);
void PSRLDQ(X64Reg reg, int shift);
void PSLLW(X64Reg reg, int shift);
void PSLLD(X64Reg reg, int shift);
void PSLLQ(X64Reg reg, int shift);
void PSRLDQ(X64Reg reg, int shift);
void PSLLDQ(X64Reg reg, int shift);
void PSRAW(X64Reg reg, int shift);
void PSRAD(X64Reg reg, int shift);
void RTDSC();
// SSE4: data type conversions
void PMOVSXBW(X64Reg dest, OpArg arg);
void PMOVSXBD(X64Reg dest, OpArg arg);
void PMOVSXBQ(X64Reg dest, OpArg arg);
void PMOVSXWD(X64Reg dest, OpArg arg);
void PMOVSXWQ(X64Reg dest, OpArg arg);
void PMOVSXDQ(X64Reg dest, OpArg arg);
void PMOVZXBW(X64Reg dest, OpArg arg);
void PMOVZXBD(X64Reg dest, OpArg arg);
void PMOVZXBQ(X64Reg dest, OpArg arg);
void PMOVZXWD(X64Reg dest, OpArg arg);
void PMOVZXWQ(X64Reg dest, OpArg arg);
void PMOVZXDQ(X64Reg dest, OpArg arg);
// SSE4: variable blend instructions (xmm0 implicit argument)
void PBLENDVB(X64Reg dest, OpArg arg);
void BLENDVPS(X64Reg dest, OpArg arg);
void BLENDVPD(X64Reg dest, OpArg arg);
// AVX
void VADDSD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VSUBSD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VMULSD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VDIVSD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VADDPD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VSUBPD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VMULPD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VDIVPD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VSQRTSD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VPAND(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VPANDN(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VPOR(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VPXOR(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VSHUFPD(X64Reg regOp1, X64Reg regOp2, OpArg arg, u8 shuffle);
void VUNPCKLPD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
void VUNPCKHPD(X64Reg regOp1, X64Reg regOp2, OpArg arg);
// VEX GPR instructions
void SARX(int bits, X64Reg regOp1, OpArg arg, X64Reg regOp2);
void SHLX(int bits, X64Reg regOp1, OpArg arg, X64Reg regOp2);
void SHRX(int bits, X64Reg regOp1, OpArg arg, X64Reg regOp2);
void RORX(int bits, X64Reg regOp, OpArg arg, u8 rotate);
void PEXT(int bits, X64Reg regOp1, X64Reg regOp2, OpArg arg);
void PDEP(int bits, X64Reg regOp1, X64Reg regOp2, OpArg arg);
void MULX(int bits, X64Reg regOp1, X64Reg regOp2, OpArg arg);
void BZHI(int bits, X64Reg regOp1, OpArg arg, X64Reg regOp2);
void BLSR(int bits, X64Reg regOp, OpArg arg);
void BLSMSK(int bits, X64Reg regOp, OpArg arg);
void BLSI(int bits, X64Reg regOp, OpArg arg);
void BEXTR(int bits, X64Reg regOp1, OpArg arg, X64Reg regOp2);
void ANDN(int bits, X64Reg regOp1, X64Reg regOp2, OpArg arg);
void RDTSC();
// Utility functions
// The difference between this and CALL is that this aligns the stack
// where appropriate.
void ABI_CallFunction(const void *func);
template <typename T>
void ABI_CallFunction(T (*func)()) {
ABI_CallFunction((const void *)func);
@ -709,10 +840,10 @@ public:
void ABI_CallFunction(const u8 *func) {
ABI_CallFunction((const void *)func);
}
void ABI_CallFunctionC16(const void *func, u16 param1);
void ABI_CallFunctionCC16(const void *func, u32 param1, u16 param2);
// These only support u32 parameters, but that's enough for a lot of uses.
// These will destroy the 1 or 2 first "parameter regs".
void ABI_CallFunctionC(const void *func, u32 param1);
@ -730,8 +861,8 @@ public:
void ABI_CallFunctionAA(const void *func, const Gen::OpArg &arg1, const Gen::OpArg &arg2);
// Pass a register as a parameter.
void ABI_CallFunctionR(const void *func, Gen::X64Reg reg1);
void ABI_CallFunctionRR(const void *func, Gen::X64Reg reg1, Gen::X64Reg reg2);
void ABI_CallFunctionR(const void *func, X64Reg reg1);
void ABI_CallFunctionRR(const void *func, X64Reg reg1, X64Reg reg2);
template <typename Tr, typename T1>
void ABI_CallFunctionC(Tr (*func)(T1), u32 param1) {
@ -789,8 +920,7 @@ public:
// Call this when shutting down. Don't rely on the destructor, even though it'll do the job.
void FreeCodeSpace();
bool IsInSpace(const u8 *ptr) const
{
bool IsInSpace(const u8 *ptr) const {
return ptr >= region && ptr < region + region_size;
}
@ -798,13 +928,11 @@ public:
// Start over if you need to change the code (call FreeCodeSpace(), AllocCodeSpace()).
void WriteProtect();
void ResetCodePtr()
{
void ResetCodePtr() {
SetCodePtr(region);
}
size_t GetSpaceLeft() const
{
size_t GetSpaceLeft() const {
return region_size - (GetCodePtr() - region);
}
@ -819,4 +947,4 @@ public:
} // namespace
#endif // _DOLPHIN_INTEL_CODEGEN_
#endif

View File

@ -772,7 +772,7 @@ bool DisassemblyMacro::disassemble(u32 address, DisassemblyLineInfo& dest, bool
dest.params = buffer;
dest.info.hasRelevantAddress = true;
dest.info.releventAddress = immediate;
dest.info.relevantAddress = immediate;
break;
case MACRO_MEMORYIMM:
dest.name = name;
@ -792,7 +792,7 @@ bool DisassemblyMacro::disassemble(u32 address, DisassemblyLineInfo& dest, bool
dest.info.dataSize = dataSize;
dest.info.hasRelevantAddress = true;
dest.info.releventAddress = immediate;
dest.info.relevantAddress = immediate;
break;
default:
return false;

View File

@ -26,8 +26,7 @@ class PointerWrap;
typedef Memory::Opcode MIPSOpcode;
enum MIPSGPReg
{
enum MIPSGPReg {
MIPS_REG_ZERO=0,
MIPS_REG_COMPILER_SCRATCH=1,
@ -65,17 +64,16 @@ enum MIPSGPReg
MIPS_REG_FP=30,
MIPS_REG_RA=31,
MIPS_REG_INVALID=-1,
// Not real regs, just for convenience/jit mapping.
MIPS_REG_HI = 32,
MIPS_REG_LO = 33,
MIPS_REG_FPCOND = 34,
MIPS_REG_VFPUCC = 35,
MIPS_REG_INVALID=-1,
};
enum
{
enum {
VFPU_CTRL_SPREFIX,
VFPU_CTRL_TPREFIX,
VFPU_CTRL_DPREFIX,

View File

@ -1204,19 +1204,19 @@ skip:
case 0x20: // add
case 0x21: // addu
info.hasRelevantAddress = true;
info.releventAddress = cpu->GetRegValue(0,MIPS_GET_RS(op))+cpu->GetRegValue(0,MIPS_GET_RT(op));
info.relevantAddress = cpu->GetRegValue(0,MIPS_GET_RS(op))+cpu->GetRegValue(0,MIPS_GET_RT(op));
break;
case 0x22: // sub
case 0x23: // subu
info.hasRelevantAddress = true;
info.releventAddress = cpu->GetRegValue(0,MIPS_GET_RS(op))-cpu->GetRegValue(0,MIPS_GET_RT(op));
info.relevantAddress = cpu->GetRegValue(0,MIPS_GET_RS(op))-cpu->GetRegValue(0,MIPS_GET_RT(op));
break;
}
break;
case 0x08: // addi
case 0x09: // adiu
info.hasRelevantAddress = true;
info.releventAddress = cpu->GetRegValue(0,MIPS_GET_RS(op))+((s16)(op & 0xFFFF));
info.relevantAddress = cpu->GetRegValue(0,MIPS_GET_RS(op))+((s16)(op & 0xFFFF));
break;
}
@ -1323,7 +1323,7 @@ skip:
info.dataAddress = rs + imm16;
info.hasRelevantAddress = true;
info.releventAddress = info.dataAddress;
info.relevantAddress = info.dataAddress;
}
return info;

View File

@ -154,7 +154,7 @@ namespace MIPSAnalyst
u32 dataAddress;
bool hasRelevantAddress;
u32 releventAddress;
u32 relevantAddress;
} MipsOpcodeInfo;
MipsOpcodeInfo GetOpcodeInfo(DebugInterface* cpu, u32 address);

View File

@ -74,29 +74,13 @@ int MIPS_SingleStep()
#else
MIPSOpcode op = Memory::Read_Opcode_JIT(mipsr4k.pc);
#endif
/*
// Choke on VFPU
MIPSInfo info = MIPSGetInfo(op);
if (info & IS_VFPU)
{
if (!Core_IsStepping() && !GetAsyncKeyState(VK_LSHIFT))
{
Core_EnableStepping(true);
return;
}
}*/
if (mipsr4k.inDelaySlot)
{
if (mipsr4k.inDelaySlot) {
MIPSInterpret(op);
if (mipsr4k.inDelaySlot)
{
if (mipsr4k.inDelaySlot) {
mipsr4k.pc = mipsr4k.nextPC;
mipsr4k.inDelaySlot = false;
}
}
else
{
} else {
MIPSInterpret(op);
}
return 1;
@ -872,14 +856,12 @@ namespace MIPSInt
int pos = _POS;
// Don't change $zr.
if (rt == 0)
{
if (rt == 0) {
PC += 4;
return;
}
switch (op & 0x3f)
{
switch (op & 0x3f) {
case 0x0: //ext
{
int size = _SIZE + 1;
@ -1025,10 +1007,10 @@ namespace MIPSInt
switch (op & 0x3f)
{
case 0: F(fd) = F(fs) + F(ft); break; //add
case 1: F(fd) = F(fs) - F(ft); break; //sub
case 2: F(fd) = F(fs) * F(ft); break; //mul
case 3: F(fd) = F(fs) / F(ft); break; //div
case 0: F(fd) = F(fs) + F(ft); break; // add.s
case 1: F(fd) = F(fs) - F(ft); break; // sub.s
case 2: F(fd) = F(fs) * F(ft); break; // mul.s
case 3: F(fd) = F(fs) / F(ft); break; // div.s
default:
_dbg_assert_msg_(CPU,0,"Trying to interpret FPU3Op instruction that can't be interpreted");
break;

View File

@ -31,8 +31,7 @@
#include "JitCommon/JitCommon.h"
enum MipsEncoding
{
enum MipsEncoding {
Imme,
Spec,
Spe2,
@ -66,8 +65,7 @@ enum MipsEncoding
Inval = -2,
};
struct MIPSInstruction
{
struct MIPSInstruction {
MipsEncoding altEncoding;
const char *name;
MIPSComp::MIPSCompileFunc compile;
@ -152,7 +150,7 @@ const MIPSInstruction tableImmediate[64] = // xxxxxx ..... ..... ...............
INVALID,
INVALID,
INSTR("swr", &Jit::Comp_ITypeMem, Dis_ITypeMem, Int_ITypeMem, IN_IMM16|IN_RS_ADDR|IN_RT|OUT_MEM|MEMTYPE_WORD),
INSTR("cache", &Jit::Comp_Cache, Dis_Cache, Int_Cache, IN_MEM|IN_IMM16|IN_RS_ADDR|IN_OTHER|OUT_OTHER),
INSTR("cache", &Jit::Comp_Cache, Dis_Cache, Int_Cache, IN_MEM|IN_IMM16|IN_RS_ADDR),
//48
INSTR("ll", &Jit::Comp_Generic, Dis_Generic, Int_StoreSync, IN_MEM|IN_IMM16|IN_RS_ADDR|OUT_RT|OUT_OTHER|MEMTYPE_WORD),
INSTR("lwc1", &Jit::Comp_FPULS, Dis_FPULS, Int_FPULS, IN_MEM|IN_IMM16|IN_RS_ADDR|OUT_OTHER|MEMTYPE_FLOAT),
@ -198,22 +196,22 @@ const MIPSInstruction tableSpecial[64] = // 000000 ..... ..... ..... ..... xxxxx
INSTR("sync", &Jit::Comp_DoNothing, Dis_Generic, Int_Sync, 0),
//16
INSTR("mfhi", &Jit::Comp_MulDivType, Dis_FromHiloTransfer, Int_MulDivType, OUT_RD|IN_OTHER),
INSTR("mthi", &Jit::Comp_MulDivType, Dis_ToHiloTransfer, Int_MulDivType, IN_RS|OUT_OTHER),
INSTR("mflo", &Jit::Comp_MulDivType, Dis_FromHiloTransfer, Int_MulDivType, OUT_RD|IN_OTHER),
INSTR("mtlo", &Jit::Comp_MulDivType, Dis_ToHiloTransfer, Int_MulDivType, IN_RS|OUT_OTHER),
INSTR("mfhi", &Jit::Comp_MulDivType, Dis_FromHiloTransfer, Int_MulDivType, OUT_RD|IN_HI),
INSTR("mthi", &Jit::Comp_MulDivType, Dis_ToHiloTransfer, Int_MulDivType, IN_RS|OUT_HI),
INSTR("mflo", &Jit::Comp_MulDivType, Dis_FromHiloTransfer, Int_MulDivType, OUT_RD|IN_LO),
INSTR("mtlo", &Jit::Comp_MulDivType, Dis_ToHiloTransfer, Int_MulDivType, IN_RS|OUT_LO),
INVALID,
INVALID,
INSTR("clz", &Jit::Comp_RType2, Dis_RType2, Int_RType2, OUT_RD|IN_RS),
INSTR("clo", &Jit::Comp_RType2, Dis_RType2, Int_RType2, OUT_RD|IN_RS),
//24
INSTR("mult", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_OTHER),
INSTR("multu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_OTHER),
INSTR("div", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_OTHER),
INSTR("divu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_OTHER),
INSTR("madd", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_OTHER|OUT_OTHER),
INSTR("maddu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_OTHER|OUT_OTHER),
INSTR("mult", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_HI|OUT_LO),
INSTR("multu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_HI|OUT_LO),
INSTR("div", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_HI|OUT_LO),
INSTR("divu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|OUT_HI|OUT_LO),
INSTR("madd", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_HI|IN_LO|OUT_HI|OUT_LO),
INSTR("maddu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_HI|IN_LO|OUT_HI|OUT_LO),
INVALID,
INVALID,
@ -234,8 +232,8 @@ const MIPSInstruction tableSpecial[64] = // 000000 ..... ..... ..... ..... xxxxx
INSTR("sltu", &Jit::Comp_RType3, Dis_RType3, Int_RType3, IN_RS|IN_RT|OUT_RD),
INSTR("max", &Jit::Comp_RType3, Dis_RType3, Int_RType3, IN_RS|IN_RT|OUT_RD),
INSTR("min", &Jit::Comp_RType3, Dis_RType3, Int_RType3, IN_RS|IN_RT|OUT_RD),
INSTR("msub", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_OTHER|OUT_OTHER),
INSTR("msubu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_OTHER|OUT_OTHER),
INSTR("msub", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_HI|IN_LO|OUT_HI|OUT_LO),
INSTR("msubu", &Jit::Comp_MulDivType, Dis_MulDivType, Int_MulDivType, IN_RS|IN_RT|IN_HI|IN_LO|OUT_HI|OUT_LO),
//48
INSTR("tge", &Jit::Comp_Generic, Dis_RType3, 0, 0),
@ -262,9 +260,9 @@ const MIPSInstruction tableSpecial2[64] = // 011100 ..... ..... ..... ..... xxxx
INVALID_X_8,
//32
INVALID, INVALID, INVALID, INVALID,
INSTR("mfic", &Jit::Comp_Generic, Dis_Generic, Int_Special2, 0),
INSTR("mfic", &Jit::Comp_Generic, Dis_Generic, Int_Special2, OUT_OTHER),
INVALID,
INSTR("mtic", &Jit::Comp_Generic, Dis_Generic, Int_Special2, 0),
INSTR("mtic", &Jit::Comp_Generic, Dis_Generic, Int_Special2, OUT_OTHER),
INVALID,
//40
INVALID_X_8,
@ -369,11 +367,11 @@ const MIPSInstruction tableCop2BC2[4] = // 010010 01000 ...xx ................
const MIPSInstruction tableCop0[32] = // 010000 xxxxx ..... ................
{
INSTR("mfc0", &Jit::Comp_Generic, Dis_Generic, 0, OUT_RT),
INSTR("mfc0", &Jit::Comp_Generic, Dis_Generic, 0, OUT_RT), // unused
INVALID,
INVALID,
INVALID,
INSTR("mtc0", &Jit::Comp_Generic, Dis_Generic, 0, IN_RT),
INSTR("mtc0", &Jit::Comp_Generic, Dis_Generic, 0, IN_RT), // unused
INVALID,
INVALID,
INVALID,
@ -423,11 +421,11 @@ const MIPSInstruction tableCop0CO[64] = // 010000 1.... ..... ..... ..... xxxxxx
const MIPSInstruction tableCop1[32] = // 010001 xxxxx ..... ..... ...........
{
INSTR("mfc1", &Jit::Comp_mxc1, Dis_mxc1, Int_mxc1, IN_OTHER|OUT_RT),
INSTR("mfc1", &Jit::Comp_mxc1, Dis_mxc1, Int_mxc1, IN_FS|OUT_RT),
INVALID,
INSTR("cfc1", &Jit::Comp_mxc1, Dis_mxc1, Int_mxc1, IN_OTHER|IN_FPUFLAG|OUT_RT),
INVALID,
INSTR("mtc1", &Jit::Comp_mxc1, Dis_mxc1, Int_mxc1, IN_RT|OUT_OTHER),
INSTR("mtc1", &Jit::Comp_mxc1, Dis_mxc1, Int_mxc1, IN_RT|OUT_FS),
INVALID,
INSTR("ctc1", &Jit::Comp_mxc1, Dis_mxc1, Int_mxc1, IN_RT|OUT_FPUFLAG|OUT_OTHER),
INVALID,
@ -455,20 +453,20 @@ const MIPSInstruction tableCop1BC[32] = // 010001 01000 xxxxx ................
const MIPSInstruction tableCop1S[64] = // 010001 10000 ..... ..... ..... xxxxxx
{
INSTR("add.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, IN_OTHER|OUT_OTHER),
INSTR("sub.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, IN_OTHER|OUT_OTHER),
INSTR("mul.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, IN_OTHER|OUT_OTHER),
INSTR("div.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, IN_OTHER|OUT_OTHER),
INSTR("sqrt.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("abs.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("mov.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("neg.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("add.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, OUT_FD|IN_FS|IN_FT),
INSTR("sub.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, OUT_FD|IN_FS|IN_FT),
INSTR("mul.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, OUT_FD|IN_FS|IN_FT),
INSTR("div.s", &Jit::Comp_FPU3op, Dis_FPU3op, Int_FPU3op, OUT_FD|IN_FS|IN_FT),
INSTR("sqrt.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INSTR("abs.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INSTR("mov.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INSTR("neg.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
//8
INVALID, INVALID, INVALID, INVALID,
INSTR("round.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("trunc.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("ceil.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("floor.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("round.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INSTR("trunc.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INSTR("ceil.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INSTR("floor.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
//16
INVALID_X_8,
//24
@ -476,29 +474,29 @@ const MIPSInstruction tableCop1S[64] = // 010001 10000 ..... ..... ..... xxxxxx
//32
INVALID, INVALID, INVALID, INVALID,
//36
INSTR("cvt.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("cvt.w.s", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INVALID,
INSTR("dis.int", &Jit::Comp_Generic, Dis_Generic, Int_Interrupt, 0),
INVALID,
//40
INVALID_X_8,
//48 - 010001 10000 ..... ..... ..... 11xxxx
INSTR("c.f", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.un", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.eq", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ueq", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.olt", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ult", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ole", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ule", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.sf", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ngle",&Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.seq", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ngl", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.lt", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.nge", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.le", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.ngt", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_OTHER|OUT_FPUFLAG),
INSTR("c.f", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, OUT_FPUFLAG),
INSTR("c.un", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.eq", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.ueq", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.olt", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.ult", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.ole", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.ule", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.sf", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, OUT_FPUFLAG),
INSTR("c.ngle",&Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.seq", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.ngl", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.lt", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.nge", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.le", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
INSTR("c.ngt", &Jit::Comp_FPUComp, Dis_FPUComp, Int_FPUComp, IN_FS|IN_FT|OUT_FPUFLAG),
};
const MIPSInstruction tableCop1W[64] = // 010001 10100 ..... ..... ..... xxxxxx
@ -511,7 +509,7 @@ const MIPSInstruction tableCop1W[64] = // 010001 10100 ..... ..... ..... xxxxxx
//24
INVALID_X_8,
//32
INSTR("cvt.s.w", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, IN_OTHER|OUT_OTHER),
INSTR("cvt.s.w", &Jit::Comp_FPU2op, Dis_FPU2op, Int_FPU2op, OUT_FD|IN_FS),
INVALID, INVALID, INVALID,
//36
INVALID,
@ -890,8 +888,6 @@ const MIPSInstruction *mipsTables[NumEncodings] =
0,
};
//arm encoding table
//const MIPSInstruction mipsinstructions[] =
//{

View File

@ -25,14 +25,14 @@ struct MIPSInfo {
value = 0;
}
explicit MIPSInfo(u32 v) : value(v) {
explicit MIPSInfo(u64 v) : value(v) {
}
u32 operator & (const u32 &arg) const {
u64 operator & (const u32 &arg) const {
return value & arg;
}
u32 value;
u64 value;
};
#define CONDTYPE_MASK 0x00000007
@ -49,44 +49,59 @@ struct MIPSInfo {
// as long as the other flags are checked,
// there is no way to misinterpret these
// as CONDTYPE_X
#define MEMTYPE_MASK 0x00000007
#define MEMTYPE_BYTE 0x00000001
#define MEMTYPE_HWORD 0x00000002
#define MEMTYPE_WORD 0x00000003
#define MEMTYPE_FLOAT 0x00000004
#define MEMTYPE_VQUAD 0x00000005
#define MEMTYPE_MASK 0x00000007ULL
#define MEMTYPE_BYTE 0x00000001ULL
#define MEMTYPE_HWORD 0x00000002ULL
#define MEMTYPE_WORD 0x00000003ULL
#define MEMTYPE_FLOAT 0x00000004ULL
#define MEMTYPE_VQUAD 0x00000005ULL
#define IS_CONDMOVE 0x00000008
#define DELAYSLOT 0x00000010
#define BAD_INSTRUCTION 0x00000020
#define LIKELY 0x00000040
#define IS_CONDBRANCH 0x00000080
#define IS_JUMP 0x00000100
#define IS_CONDMOVE 0x00000008ULL
#define DELAYSLOT 0x00000010ULL
#define BAD_INSTRUCTION 0x00000020ULL
#define LIKELY 0x00000040ULL
#define IS_CONDBRANCH 0x00000080ULL
#define IS_JUMP 0x00000100ULL
#define IN_RS 0x00000200
#define IN_RS_ADDR (0x00000400 | IN_RS)
#define IN_RS_SHIFT (0x00000800 | IN_RS)
#define IN_RT 0x00001000
#define IN_SA 0x00002000
#define IN_IMM16 0x00004000
#define IN_IMM26 0x00008000
#define IN_MEM 0x00010000
#define IN_OTHER 0x00020000
#define IN_FPUFLAG 0x00040000
#define IN_VFPU_CC 0x00080000
#define IN_RS 0x00000200ULL
#define IN_RS_ADDR (0x00000400ULL | IN_RS)
#define IN_RS_SHIFT (0x00000800ULL | IN_RS)
#define IN_RT 0x00001000ULL
#define IN_SA 0x00002000ULL
#define IN_IMM16 0x00004000ULL
#define IN_IMM26 0x00008000ULL
#define IN_MEM 0x00010000ULL
#define IN_OTHER 0x00020000ULL
#define IN_FPUFLAG 0x00040000ULL
#define IN_VFPU_CC 0x00080000ULL
#define OUT_RT 0x00100000
#define OUT_RD 0x00200000
#define OUT_RA 0x00400000
#define OUT_MEM 0x00800000
#define OUT_OTHER 0x01000000
#define OUT_FPUFLAG 0x02000000
#define OUT_VFPU_CC 0x04000000
#define OUT_EAT_PREFIX 0x08000000
#define OUT_RT 0x00100000ULL
#define OUT_RD 0x00200000ULL
#define OUT_RA 0x00400000ULL
#define OUT_MEM 0x00800000ULL
#define OUT_OTHER 0x01000000ULL
#define OUT_FPUFLAG 0x02000000ULL
#define OUT_VFPU_CC 0x04000000ULL
#define OUT_EAT_PREFIX 0x08000000ULL
#define VFPU_NO_PREFIX 0x10000000
#define IS_VFPU 0x20000000
#define IS_FPU 0x40000000
#define VFPU_NO_PREFIX 0x10000000ULL
#define IS_VFPU 0x20000000ULL
#define IS_FPU 0x40000000ULL
#define IN_FS 0x000100000000ULL
#define IN_FT 0x000200000000ULL
#define IN_LO 0x000400000000ULL
#define IN_HI 0x000800000000ULL
#define OUT_FD 0x001000000000ULL
#define OUT_FS 0x002000000000ULL
#define OUT_LO 0x004000000000ULL
#define OUT_HI 0x008000000000ULL
#define IN_VS 0x010000000000ULL
#define IN_VT 0x020000000000ULL
#define OUT_VD 0x100000000000ULL
#ifndef CDECL
#define CDECL

View File

@ -643,7 +643,7 @@ void CtrlDisAsmView::followBranch()
} else if (line.info.hasRelevantAddress)
{
// well, not exactly a branch, but we can do something anyway
SendMessage(GetParent(wnd),WM_DEB_GOTOHEXEDIT,line.info.releventAddress,0);
SendMessage(GetParent(wnd),WM_DEB_GOTOHEXEDIT,line.info.relevantAddress,0);
SetFocus(wnd);
}
} else if (line.type == DISTYPE_DATA)