public static void Vmad(ShaderIrBlock Block, long OpCode)
{
ShaderIrNode OperA = GetOperGpr8(OpCode);
ShaderIrNode OperB;
if (((OpCode >> 50) & 1) != 0)
{
OperB = GetOperGpr20(OpCode);
}
else
{
OperB = GetOperImm19_20(OpCode);
}
ShaderIrOperGpr OperC = GetOperGpr39(OpCode);
ShaderIrNode Tmp = new ShaderIrOp(ShaderIrInst.Mul, OperA, OperB);
ShaderIrNode Final = new ShaderIrOp(ShaderIrInst.Add, Tmp, OperC);
int Shr = (int)((OpCode >> 51) & 3);
if (Shr != 0)
{
int Shift = (Shr == 2) ? 15 : 7;
Final = new ShaderIrOp(ShaderIrInst.Lsr, Final, new ShaderIrOperImm(Shift));
}
Block.AddNode(new ShaderIrCmnt("Stubbed. Instruction is reduced to a * b + c"));
Block.AddNode(GetPredNode(new ShaderIrAsg(GetOperGpr0(OpCode), Final), OpCode));
}
public static void Ld_C(ShaderIrBlock Block, long OpCode)
{
int Type = (int)(OpCode >> 48) & 7;
if (Type > 5)
{
throw new InvalidOperationException();
}
int Count = Type == 5 ? 2 : 1;
for (int Index = 0; Index < Count; Index++)
{
ShaderIrOperCbuf OperA = GetOperCbuf36(OpCode);
ShaderIrOperGpr OperD = GetOperGpr0(OpCode);
OperA.Pos += Index;
OperD.Index += Index;
ShaderIrNode Node = OperA;
if (Type < 4)
{
//This is a 8 or 16 bits type.
bool Signed = (Type & 1) != 0;
int Size = 8 << (Type >> 1);
Node = ExtendTo32(Node, Signed, Size);
}
Block.AddNode(GetPredNode(new ShaderIrAsg(OperD, Node), OpCode));
}
}
private static void EmitBinaryHalfOp(ShaderIrBlock Block, long OpCode, ShaderIrInst Inst)
{
bool AbsB = OpCode.Read(30);
bool NegB = OpCode.Read(31);
bool Sat = OpCode.Read(32);
bool AbsA = OpCode.Read(44);
ShaderIrOperGpr[] VecA = OpCode.GprHalfVec8();
ShaderIrOperGpr[] VecB = OpCode.GprHalfVec20();
HalfOutputType OutputType = (HalfOutputType)OpCode.Read(49, 3);
int Elems = OutputType == HalfOutputType.PackedFp16 ? 2 : 1;
int First = OutputType == HalfOutputType.MergeH1 ? 1 : 0;
for (int Index = First; Index < Elems; Index++)
{
ShaderIrNode OperA = GetAluFabs(VecA[Index], AbsA);
ShaderIrNode OperB = GetAluFabsFneg(VecB[Index], AbsB, NegB);
ShaderIrNode Op = new ShaderIrOp(Inst, OperA, OperB);
ShaderIrOperGpr Dst = GetHalfDst(OpCode, OutputType, Index);
Block.AddNode(OpCode.PredNode(new ShaderIrAsg(Dst, GetAluFsat(Op, Sat))));
}
}
public static void Vmad(ShaderIrBlock Block, long OpCode, int Position)
{
ShaderIrNode OperA = OpCode.Gpr8();
ShaderIrNode OperB;
if (OpCode.Read(50))
{
OperB = OpCode.Gpr20();
}
else
{
OperB = OpCode.Imm19_20();
}
ShaderIrOperGpr OperC = OpCode.Gpr39();
ShaderIrNode Tmp = new ShaderIrOp(ShaderIrInst.Mul, OperA, OperB);
ShaderIrNode Final = new ShaderIrOp(ShaderIrInst.Add, Tmp, OperC);
int Shr = OpCode.Read(51, 3);
if (Shr != 0)
{
int Shift = (Shr == 2) ? 15 : 7;
Final = new ShaderIrOp(ShaderIrInst.Lsr, Final, new ShaderIrOperImm(Shift));
}
Block.AddNode(new ShaderIrCmnt("Stubbed. Instruction is reduced to a * b + c"));
Block.AddNode(OpCode.PredNode(new ShaderIrAsg(OpCode.Gpr0(), Final)));
}
public static void Vmad(ShaderIrBlock block, long opCode, int position)
{
ShaderIrNode operA = opCode.Gpr8();
ShaderIrNode operB;
if (opCode.Read(50))
{
operB = opCode.Gpr20();
}
else
{
operB = opCode.Imm19_20();
}
ShaderIrOperGpr operC = opCode.Gpr39();
ShaderIrNode tmp = new ShaderIrOp(ShaderIrInst.Mul, operA, operB);
ShaderIrNode final = new ShaderIrOp(ShaderIrInst.Add, tmp, operC);
int shr = opCode.Read(51, 3);
if (shr != 0)
{
int shift = (shr == 2) ? 15 : 7;
final = new ShaderIrOp(ShaderIrInst.Lsr, final, new ShaderIrOperImm(shift));
}
block.AddNode(new ShaderIrCmnt("Stubbed. Instruction is reduced to a * b + c"));
block.AddNode(opCode.PredNode(new ShaderIrAsg(opCode.Gpr0(), final)));
}
private static void EmitBinaryHalfOp(ShaderIrBlock block, long opCode, ShaderIrInst inst)
{
bool absB = opCode.Read(30);
bool negB = opCode.Read(31);
bool sat = opCode.Read(32);
bool absA = opCode.Read(44);
ShaderIrOperGpr[] vecA = opCode.GprHalfVec8();
ShaderIrOperGpr[] vecB = opCode.GprHalfVec20();
HalfOutputType outputType = (HalfOutputType)opCode.Read(49, 3);
int elems = outputType == HalfOutputType.PackedFp16 ? 2 : 1;
int first = outputType == HalfOutputType.MergeH1 ? 1 : 0;
for (int index = first; index < elems; index++)
{
ShaderIrNode operA = GetAluFabs(vecA[index], absA);
ShaderIrNode operB = GetAluFabsFneg(vecB[index], absB, negB);
ShaderIrNode op = new ShaderIrOp(inst, operA, operB);
ShaderIrOperGpr dst = GetHalfDst(opCode, outputType, index);
block.AddNode(opCode.PredNode(new ShaderIrAsg(dst, GetAluFsat(op, sat))));
}
}
private static void EmitTex(ShaderIrBlock Block, long OpCode, ShaderIrInst Inst)
{
//TODO: Support other formats.
ShaderIrNode OperA = GetOperGpr8(OpCode);
ShaderIrNode OperB = GetOperGpr20(OpCode);
ShaderIrNode OperC = GetOperImm13_36(OpCode);
for (int Ch = 0; Ch < 4; Ch++)
{
//Assign it to a temp because the destination registers
//may be used as texture coord input aswell.
ShaderIrOperGpr Dst = new ShaderIrOperGpr(0x100 + Ch);
ShaderIrMetaTex Meta = new ShaderIrMetaTex(Ch);
ShaderIrOp Op = new ShaderIrOp(Inst, OperA, OperB, OperC, Meta);
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Op), OpCode));
}
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOperGpr Src = new ShaderIrOperGpr(0x100 + Ch);
ShaderIrOperGpr Dst = (Ch >> 1) != 0
? GetOperGpr28(OpCode)
: GetOperGpr0(OpCode);
Dst.Index += Ch & 1;
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Src), OpCode));
}
}
private static void EmitTex(ShaderIrBlock Block, long OpCode, bool GprHandle)
{
//TODO: Support other formats.
ShaderIrOperGpr[] Coords = new ShaderIrOperGpr[2];
for (int Index = 0; Index < Coords.Length; Index++)
{
Coords[Index] = GetOperGpr8(OpCode);
Coords[Index].Index += Index;
if (Coords[Index].Index > ShaderIrOperGpr.ZRIndex)
{
Coords[Index].Index = ShaderIrOperGpr.ZRIndex;
}
}
int ChMask = (int)(OpCode >> 31) & 0xf;
ShaderIrNode OperC = GprHandle
? (ShaderIrNode)GetOperGpr20(OpCode)
: (ShaderIrNode)GetOperImm13_36(OpCode);
ShaderIrInst Inst = GprHandle ? ShaderIrInst.Texb : ShaderIrInst.Texs;
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOperGpr Dst = new ShaderIrOperGpr(TempRegStart + Ch);
ShaderIrMetaTex Meta = new ShaderIrMetaTex(Ch);
ShaderIrOp Op = new ShaderIrOp(Inst, Coords[0], Coords[1], OperC, Meta);
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Op), OpCode));
}
int RegInc = 0;
for (int Ch = 0; Ch < 4; Ch++)
{
if (!IsChannelUsed(ChMask, Ch))
{
continue;
}
ShaderIrOperGpr Src = new ShaderIrOperGpr(TempRegStart + Ch);
ShaderIrOperGpr Dst = GetOperGpr0(OpCode);
Dst.Index += RegInc++;
if (Dst.Index >= ShaderIrOperGpr.ZRIndex)
{
continue;
}
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Src), OpCode));
}
}
private static void EmitTex(ShaderIrBlock Block, long OpCode, bool GprHandle)
{
//TODO: Support other formats.
ShaderIrOperGpr[] Coords = new ShaderIrOperGpr[2];
for (int Index = 0; Index < Coords.Length; Index++)
{
ShaderIrOperGpr CoordReg = OpCode.Gpr8();
CoordReg.Index += Index;
if (!CoordReg.IsValidRegister)
{
CoordReg.Index = ShaderIrOperGpr.ZRIndex;
}
Coords[Index] = ShaderIrOperGpr.MakeTemporary(Index);
Block.AddNode(new ShaderIrAsg(Coords[Index], CoordReg));
}
int ChMask = OpCode.Read(31, 0xf);
ShaderIrNode OperC = GprHandle
? (ShaderIrNode)OpCode.Gpr20()
: (ShaderIrNode)OpCode.Imm13_36();
ShaderIrInst Inst = GprHandle ? ShaderIrInst.Texb : ShaderIrInst.Texs;
int RegInc = 0;
for (int Ch = 0; Ch < 4; Ch++)
{
if (!IsChannelUsed(ChMask, Ch))
{
continue;
}
ShaderIrOperGpr Dst = OpCode.Gpr0();
Dst.Index += RegInc++;
if (!Dst.IsValidRegister || Dst.IsConst)
{
continue;
}
ShaderIrMetaTex Meta = new ShaderIrMetaTex(Ch);
ShaderIrOp Op = new ShaderIrOp(Inst, Coords[0], Coords[1], OperC, Meta);
Block.AddNode(OpCode.PredNode(new ShaderIrAsg(Dst, Op)));
}
}
private static ShaderIrOperGpr[] CoordsRegistersToTempRegisters(ShaderIrBlock block, params ShaderIrOperGpr[] registers)
{
ShaderIrOperGpr[] res = new ShaderIrOperGpr[registers.Length];
for (int index = 0; index < res.Length; index++)
{
res[index] = ShaderIrOperGpr.MakeTemporary(index);
block.AddNode(new ShaderIrAsg(res[index], registers[index]));
}
return(res);
}
private static void EmitTex(ShaderIrBlock Block, long OpCode, ShaderIrInst Inst)
{
//TODO: Support other formats.
ShaderIrNode OperA = GetOperGpr8(OpCode);
ShaderIrNode OperB = GetOperGpr20(OpCode);
ShaderIrNode OperC = GetOperImm13_36(OpCode);
int LutIndex;
LutIndex = GetOperGpr0(OpCode).Index != ShaderIrOperGpr.ZRIndex ? 1 : 0;
LutIndex |= GetOperGpr28(OpCode).Index != ShaderIrOperGpr.ZRIndex ? 2 : 0;
int ChMask = MaskLut[LutIndex, (OpCode >> 50) & 7];
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOperGpr Dst = new ShaderIrOperGpr(TempRegStart + Ch);
ShaderIrMetaTex Meta = new ShaderIrMetaTex(Ch);
ShaderIrOp Op = new ShaderIrOp(Inst, OperA, OperB, OperC, Meta);
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Op), OpCode));
}
int RegInc = 0;
for (int Ch = 0; Ch < 4; Ch++)
{
if (!IsChannelUsed(ChMask, Ch))
{
continue;
}
ShaderIrOperGpr Src = new ShaderIrOperGpr(TempRegStart + Ch);
ShaderIrOperGpr Dst = (RegInc >> 1) != 0
? GetOperGpr28(OpCode)
: GetOperGpr0(OpCode);
Dst.Index += RegInc++ & 1;
if (Dst.Index >= ShaderIrOperGpr.ZRIndex)
{
continue;
}
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Src), OpCode));
}
}
public static void Ld_A(ShaderIrBlock Block, long OpCode)
{
ShaderIrNode[] Opers = GetOperAbuf20(OpCode);
int Index = 0;
foreach (ShaderIrNode OperA in Opers)
{
ShaderIrOperGpr OperD = GetOperGpr0(OpCode);
OperD.Index += Index++;
Block.AddNode(GetPredNode(new ShaderIrAsg(OperD, OperA), OpCode));
}
}
public static void St_A(ShaderIrBlock block, long opCode, int position)
{
ShaderIrNode[] opers = opCode.Abuf20();
int index = 0;
foreach (ShaderIrNode operA in opers)
{
ShaderIrOperGpr operD = opCode.Gpr0();
operD.Index += index++;
block.AddNode(opCode.PredNode(new ShaderIrAsg(operA, operD)));
}
}
public static void St_A(ShaderIrBlock Block, long OpCode, int Position)
{
ShaderIrNode[] Opers = OpCode.Abuf20();
int Index = 0;
foreach (ShaderIrNode OperA in Opers)
{
ShaderIrOperGpr OperD = OpCode.Gpr0();
OperD.Index += Index++;
Block.AddNode(OpCode.PredNode(new ShaderIrAsg(OperA, OperD)));
}
}
public static ShaderIrOperAbuf[] GetOperAbuf20(long OpCode)
{
int Abuf = (int)(OpCode >> 20) & 0x3ff;
int Size = (int)(OpCode >> 47) & 3;
ShaderIrOperGpr Vertex = GetOperGpr39(OpCode);
ShaderIrOperAbuf[] Opers = new ShaderIrOperAbuf[Size + 1];
for (int Index = 0; Index <= Size; Index++)
{
Opers[Index] = new ShaderIrOperAbuf(Abuf + Index * 4, Vertex);
}
return(Opers);
}
private static ShaderIrOperAbuf[] Abuf20(this long OpCode)
{
int Abuf = OpCode.Read(20, 0x3ff);
int Size = OpCode.Read(47, 3);
ShaderIrOperGpr Vertex = OpCode.Gpr39();
ShaderIrOperAbuf[] Opers = new ShaderIrOperAbuf[Size + 1];
for (int Index = 0; Index <= Size; Index++)
{
Opers[Index] = new ShaderIrOperAbuf(Abuf + Index * 4, Vertex);
}
return(Opers);
}
private static ShaderIrOperAbuf[] Abuf20(this long opCode)
{
int abuf = opCode.Read(20, 0x3ff);
int size = opCode.Read(47, 3);
ShaderIrOperGpr vertex = opCode.Gpr39();
ShaderIrOperAbuf[] opers = new ShaderIrOperAbuf[size + 1];
for (int index = 0; index <= size; index++)
{
opers[index] = new ShaderIrOperAbuf(abuf + index * 4, vertex);
}
return(opers);
}
public static void Ld_C(ShaderIrBlock block, long opCode, int position)
{
int cbufPos = opCode.Read(22, 0x3fff);
int cbufIndex = opCode.Read(36, 0x1f);
int type = opCode.Read(48, 7);
if (type > 5)
{
throw new InvalidOperationException();
}
ShaderIrOperGpr temp = ShaderIrOperGpr.MakeTemporary();
block.AddNode(new ShaderIrAsg(temp, opCode.Gpr8()));
int count = type == 5 ? 2 : 1;
for (int index = 0; index < count; index++)
{
ShaderIrOperCbuf operA = new ShaderIrOperCbuf(cbufIndex, cbufPos, temp);
ShaderIrOperGpr operD = opCode.Gpr0();
operA.Pos += index;
operD.Index += index;
if (!operD.IsValidRegister)
{
break;
}
ShaderIrNode node = operA;
if (type < 4)
{
//This is a 8 or 16 bits type.
bool signed = (type & 1) != 0;
int size = 8 << (type >> 1);
node = ExtendTo32(node, signed, size);
}
block.AddNode(opCode.PredNode(new ShaderIrAsg(operD, node)));
}
}
public static void Ld_C(ShaderIrBlock Block, long OpCode, long Position)
{
int CbufPos = (int)(OpCode >> 22) & 0x3fff;
int CbufIndex = (int)(OpCode >> 36) & 0x1f;
int Type = (int)(OpCode >> 48) & 7;
if (Type > 5)
{
throw new InvalidOperationException();
}
ShaderIrOperGpr Temp = ShaderIrOperGpr.MakeTemporary();
Block.AddNode(new ShaderIrAsg(Temp, GetOperGpr8(OpCode)));
int Count = Type == 5 ? 2 : 1;
for (int Index = 0; Index < Count; Index++)
{
ShaderIrOperCbuf OperA = new ShaderIrOperCbuf(CbufIndex, CbufPos, Temp);
ShaderIrOperGpr OperD = GetOperGpr0(OpCode);
OperA.Pos += Index;
OperD.Index += Index;
if (!OperD.IsValidRegister)
{
break;
}
ShaderIrNode Node = OperA;
if (Type < 4)
{
//This is a 8 or 16 bits type.
bool Signed = (Type & 1) != 0;
int Size = 8 << (Type >> 1);
Node = ExtendTo32(Node, Signed, Size);
}
Block.AddNode(GetPredNode(new ShaderIrAsg(OperD, Node), OpCode));
}
}
public static void Out_R(ShaderIrBlock Block, long OpCode, long Position)
{
//TODO: Those registers have to be used for something
ShaderIrOperGpr Gpr0 = GetOperGpr0(OpCode);
ShaderIrOperGpr Gpr8 = GetOperGpr8(OpCode);
ShaderIrOperGpr Gpr20 = GetOperGpr20(OpCode);
int Type = (int)((OpCode >> 39) & 3);
if ((Type & 1) != 0)
{
Block.AddNode(GetPredNode(new ShaderIrOp(ShaderIrInst.Emit), OpCode));
}
if ((Type & 2) != 0)
{
Block.AddNode(GetPredNode(new ShaderIrOp(ShaderIrInst.Cut), OpCode));
}
}
public static void Texs(ShaderIrBlock Block, long OpCode)
{
//TODO: Support other formats.
ShaderIrNode OperA = GetOperGpr8(OpCode);
ShaderIrNode OperB = GetOperGpr20(OpCode);
ShaderIrNode OperC = GetOperGpr28(OpCode);
ShaderIrNode OperD = GetOperImm13_36(OpCode);
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOp Op = new ShaderIrOp(ShaderIrInst.Texr + Ch, OperA, OperB, OperD);
ShaderIrOperGpr Dst = GetOperGpr0(OpCode);
Dst.Index += Ch;
Block.AddNode(new ShaderIrAsg(Dst, Op));
}
}
private static ShaderIrOperGpr[] GetGprHalfVec2(int gpr, int mask)
{
if (mask == 1)
{
//This value is used for FP32, the whole 32-bits register
//is used as each element on the vector.
return(new ShaderIrOperGpr[]
{
new ShaderIrOperGpr(gpr),
new ShaderIrOperGpr(gpr)
});
}
ShaderIrOperGpr low = new ShaderIrOperGpr(gpr, 0);
ShaderIrOperGpr high = new ShaderIrOperGpr(gpr, 1);
return(new ShaderIrOperGpr[]
{
(mask & 1) != 0 ? high : low,
(mask & 2) != 0 ? high : low
});
}
private static void EmitSel(ShaderIrBlock Block, long OpCode, ShaderOper Oper)
{
ShaderIrOperGpr Dst = GetOperGpr0(OpCode);
ShaderIrNode Pred = GetOperPred39N(OpCode);
ShaderIrNode ResultA = GetOperGpr8(OpCode);
ShaderIrNode ResultB;
switch (Oper)
{
case ShaderOper.CR: ResultB = GetOperCbuf34(OpCode); break;
case ShaderOper.Imm: ResultB = GetOperImm19_20(OpCode); break;
case ShaderOper.RR: ResultB = GetOperGpr20(OpCode); break;
default: throw new ArgumentException(nameof(Oper));
}
Block.AddNode(GetPredNode(new ShaderIrCond(Pred, new ShaderIrAsg(Dst, ResultA), false), OpCode));
Block.AddNode(GetPredNode(new ShaderIrCond(Pred, new ShaderIrAsg(Dst, ResultB), true), OpCode));
}
private static void EmitSel(ShaderIrBlock block, long opCode, ShaderOper oper)
{
ShaderIrOperGpr dst = opCode.Gpr0();
ShaderIrNode pred = opCode.Pred39N();
ShaderIrNode resultA = opCode.Gpr8();
ShaderIrNode resultB;
switch (oper)
{
case ShaderOper.Cr: resultB = opCode.Cbuf34(); break;
case ShaderOper.Imm: resultB = opCode.Imm19_20(); break;
case ShaderOper.Rr: resultB = opCode.Gpr20(); break;
default: throw new ArgumentException(nameof(oper));
}
block.AddNode(opCode.PredNode(new ShaderIrCond(pred, new ShaderIrAsg(dst, resultA), false)));
block.AddNode(opCode.PredNode(new ShaderIrCond(pred, new ShaderIrAsg(dst, resultB), true)));
}
public static void Mov_R(ShaderIrBlock Block, long OpCode, int Position)
{
ShaderIrOperGpr Gpr = OpCode.Gpr20();
Block.AddNode(OpCode.PredNode(new ShaderIrAsg(OpCode.Gpr0(), Gpr)));
}
public static void Mov_R(ShaderIrBlock Block, long OpCode, long Position)
{
ShaderIrOperGpr Gpr = GetOperGpr20(OpCode);
Block.AddNode(GetPredNode(new ShaderIrAsg(GetOperGpr0(OpCode), Gpr), OpCode));
}
public static void Mov_R(ShaderIrBlock block, long opCode, int position)
{
ShaderIrOperGpr gpr = opCode.Gpr20();
block.AddNode(opCode.PredNode(new ShaderIrAsg(opCode.Gpr0(), gpr)));
}
private static void EmitTexs(ShaderIrBlock block,
long opCode,
ShaderIrInst inst,
GalTextureTarget textureTarget,
TextureInstructionSuffix textureInstructionSuffix)
{
if (inst == ShaderIrInst.Txlf && textureTarget == GalTextureTarget.CubeArray)
{
throw new InvalidOperationException("TLDS instructions cannot use CUBE modifier!");
}
bool isArray = ImageUtils.IsArray(textureTarget);
ShaderIrOperGpr[] coords = new ShaderIrOperGpr[ImageUtils.GetCoordsCountTextureTarget(textureTarget)];
ShaderIrOperGpr operA = opCode.Gpr8();
ShaderIrOperGpr operB = opCode.Gpr20();
ShaderIrOperGpr suffixExtra = opCode.Gpr20();
suffixExtra.Index += 1;
int coordStartIndex = 0;
if (isArray)
{
coordStartIndex++;
coords[coords.Length - 1] = opCode.Gpr8();
}
switch (coords.Length - coordStartIndex)
{
case 1:
coords[0] = opCode.Gpr8();
break;
case 2:
coords[0] = opCode.Gpr8();
coords[0].Index += coordStartIndex;
break;
case 3:
coords[0] = opCode.Gpr8();
coords[0].Index += coordStartIndex;
coords[1] = opCode.Gpr8();
coords[1].Index += 1 + coordStartIndex;
break;
default:
throw new NotSupportedException($"{coords.Length - coordStartIndex} coords textures aren't supported in TEXS");
}
int operBIndex = 0;
ShaderIrOperGpr levelOfDetail = null;
ShaderIrOperGpr offset = null;
ShaderIrOperGpr depthCompare = null;
// OperB is always the last value
// Not applicable to 1d textures
if (coords.Length - coordStartIndex != 1)
{
coords[coords.Length - coordStartIndex - 1] = operB;
operBIndex++;
}
// Encoding of TEXS/TLDS is a bit special and change for 2d textures
// NOTE: OperA seems to hold at best two args.
// On 2D textures, if no suffix need an additional values, Y is stored in OperB, otherwise coords are in OperA and the additional values is in OperB.
if (textureInstructionSuffix != TextureInstructionSuffix.None && textureInstructionSuffix != TextureInstructionSuffix.Lz && textureTarget == GalTextureTarget.TwoD)
{
coords[coords.Length - coordStartIndex - 1] = opCode.Gpr8();
coords[coords.Length - coordStartIndex - 1].Index += coords.Length - coordStartIndex - 1;
operBIndex--;
}
// TODO: Find what MZ does and what changes about the encoding (Maybe Multisample?)
if ((textureInstructionSuffix & TextureInstructionSuffix.Ll) != 0)
{
levelOfDetail = opCode.Gpr20();
levelOfDetail.Index += operBIndex;
operBIndex++;
}
if ((textureInstructionSuffix & TextureInstructionSuffix.AOffI) != 0)
{
offset = opCode.Gpr20();
offset.Index += operBIndex;
operBIndex++;
}
if ((textureInstructionSuffix & TextureInstructionSuffix.Dc) != 0)
{
depthCompare = opCode.Gpr20();
depthCompare.Index += operBIndex;
operBIndex++;
}
int lutIndex;
lutIndex = !opCode.Gpr0().IsConst ? 1 : 0;
lutIndex |= !opCode.Gpr28().IsConst ? 2 : 0;
if (lutIndex == 0)
{
//Both destination registers are RZ, do nothing.
return;
}
bool fp16 = !opCode.Read(59);
int dstIncrement = 0;
ShaderIrOperGpr GetDst()
{
ShaderIrOperGpr dst;
if (fp16)
{
//FP16 mode, two components are packed on the two
//halfs of a 32-bits register, as two half-float values.
int halfPart = dstIncrement & 1;
switch (lutIndex)
{
case 1: dst = opCode.GprHalf0(halfPart); break;
case 2: dst = opCode.GprHalf28(halfPart); break;
case 3: dst = (dstIncrement >> 1) != 0
? opCode.GprHalf28(halfPart)
: opCode.GprHalf0(halfPart); break;
default: throw new InvalidOperationException();
}
}
else
{
//32-bits mode, each component uses one register.
//Two components uses two consecutive registers.
switch (lutIndex)
{
case 1: dst = opCode.Gpr0(); break;
case 2: dst = opCode.Gpr28(); break;
case 3: dst = (dstIncrement >> 1) != 0
? opCode.Gpr28()
: opCode.Gpr0(); break;
default: throw new InvalidOperationException();
}
dst.Index += dstIncrement & 1;
}
dstIncrement++;
return(dst);
}
int chMask = _maskLut[lutIndex, opCode.Read(50, 7)];
if (chMask == 0)
{
//All channels are disabled, do nothing.
return;
}
ShaderIrNode operC = opCode.Imm13_36();
coords = CoordsRegistersToTempRegisters(block, coords);
for (int ch = 0; ch < 4; ch++)
{
if (!IsChannelUsed(chMask, ch))
{
continue;
}
ShaderIrMetaTex meta = new ShaderIrMetaTex(ch, textureTarget, textureInstructionSuffix, coords)
{
LevelOfDetail = levelOfDetail,
Offset = offset,
DepthCompare = depthCompare
};
ShaderIrOp op = new ShaderIrOp(inst, operA, operB, operC, meta);
ShaderIrOperGpr dst = GetDst();
if (dst.IsValidRegister && !dst.IsConst)
{
block.AddNode(opCode.PredNode(new ShaderIrAsg(dst, op)));
}
}
}
private static void EmitTexs(ShaderIrBlock Block, long OpCode, ShaderIrInst Inst)
{
//TODO: Support other formats.
ShaderIrNode OperA = GetOperGpr8(OpCode);
ShaderIrNode OperB = GetOperGpr20(OpCode);
ShaderIrNode OperC = GetOperImm13_36(OpCode);
int LutIndex;
LutIndex = GetOperGpr0(OpCode).Index != ShaderIrOperGpr.ZRIndex ? 1 : 0;
LutIndex |= GetOperGpr28(OpCode).Index != ShaderIrOperGpr.ZRIndex ? 2 : 0;
if (LutIndex == 0)
{
//Both registers are RZ, color is not written anywhere.
//So, the intruction is basically a no-op.
return;
}
int ChMask = MaskLut[LutIndex, (OpCode >> 50) & 7];
for (int Ch = 0; Ch < 4; Ch++)
{
ShaderIrOperGpr Dst = new ShaderIrOperGpr(TempRegStart + Ch);
ShaderIrMetaTex Meta = new ShaderIrMetaTex(Ch);
ShaderIrOp Op = new ShaderIrOp(Inst, OperA, OperB, OperC, Meta);
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Op), OpCode));
}
int RegInc = 0;
ShaderIrOperGpr GetDst()
{
ShaderIrOperGpr Dst;
switch (LutIndex)
{
case 1: Dst = GetOperGpr0(OpCode); break;
case 2: Dst = GetOperGpr28(OpCode); break;
case 3: Dst = (RegInc >> 1) != 0
? GetOperGpr28(OpCode)
: GetOperGpr0(OpCode); break;
default: throw new InvalidOperationException();
}
Dst.Index += RegInc++ & 1;
return(Dst);
}
for (int Ch = 0; Ch < 4; Ch++)
{
if (!IsChannelUsed(ChMask, Ch))
{
continue;
}
ShaderIrOperGpr Src = new ShaderIrOperGpr(TempRegStart + Ch);
ShaderIrOperGpr Dst = GetDst();
if (Dst.Index != ShaderIrOperGpr.ZRIndex)
{
Block.AddNode(GetPredNode(new ShaderIrAsg(Dst, Src), OpCode));
}
}
}
private static void EmitTld4(ShaderIrBlock block, long opCode, GalTextureTarget textureType, TextureInstructionSuffix textureInstructionSuffix, int chMask, int component, bool scalar)
{
ShaderIrOperGpr operA = opCode.Gpr8();
ShaderIrOperGpr operB = opCode.Gpr20();
ShaderIrOperImm operC = opCode.Imm13_36();
ShaderIrOperGpr[] coords = new ShaderIrOperGpr[ImageUtils.GetCoordsCountTextureTarget(textureType)];
ShaderIrOperGpr offset = null;
ShaderIrOperGpr depthCompare = null;
bool isArray = ImageUtils.IsArray(textureType);
int operBIndex = 0;
if (scalar)
{
int coordStartIndex = 0;
if (isArray)
{
coordStartIndex++;
coords[coords.Length - 1] = operB;
}
switch (coords.Length - coordStartIndex)
{
case 1:
coords[0] = opCode.Gpr8();
break;
case 2:
coords[0] = opCode.Gpr8();
coords[0].Index += coordStartIndex;
break;
case 3:
coords[0] = opCode.Gpr8();
coords[0].Index += coordStartIndex;
coords[1] = opCode.Gpr8();
coords[1].Index += 1 + coordStartIndex;
break;
default:
throw new NotSupportedException($"{coords.Length - coordStartIndex} coords textures aren't supported in TLD4S");
}
if (coords.Length - coordStartIndex != 1)
{
coords[coords.Length - coordStartIndex - 1] = operB;
operBIndex++;
}
if (textureInstructionSuffix != TextureInstructionSuffix.None && textureType == GalTextureTarget.TwoD)
{
coords[coords.Length - coordStartIndex - 1] = opCode.Gpr8();
coords[coords.Length - coordStartIndex - 1].Index += coords.Length - coordStartIndex - 1;
operBIndex--;
}
}
else
{
int indexExtraCoord = 0;
if (isArray)
{
indexExtraCoord++;
coords[coords.Length - 1] = opCode.Gpr8();
}
for (int index = 0; index < coords.Length - indexExtraCoord; index++)
{
coords[index] = opCode.Gpr8();
coords[index].Index += index;
coords[index].Index += indexExtraCoord;
if (coords[index].Index > ShaderIrOperGpr.ZrIndex)
{
coords[index].Index = ShaderIrOperGpr.ZrIndex;
}
}
}
if ((textureInstructionSuffix & TextureInstructionSuffix.AOffI) != 0)
{
offset = opCode.Gpr20();
offset.Index += operBIndex;
operBIndex++;
}
if ((textureInstructionSuffix & TextureInstructionSuffix.Dc) != 0)
{
depthCompare = opCode.Gpr20();
depthCompare.Index += operBIndex;
operBIndex++;
}
coords = CoordsRegistersToTempRegisters(block, coords);
int regInc = 0;
for (int ch = 0; ch < 4; ch++)
{
if (!IsChannelUsed(chMask, ch))
{
continue;
}
ShaderIrOperGpr dst = opCode.Gpr0();
dst.Index += regInc++;
if (!dst.IsValidRegister || dst.IsConst)
{
continue;
}
ShaderIrMetaTex meta = new ShaderIrMetaTex(ch, textureType, textureInstructionSuffix, coords)
{
Component = component,
Offset = offset,
DepthCompare = depthCompare
};
ShaderIrOp op = new ShaderIrOp(ShaderIrInst.Tld4, operA, operB, operC, meta);
block.AddNode(opCode.PredNode(new ShaderIrAsg(dst, op)));
}
}