Add Fcvtns_S, Fcvtns_V, Fcvtnu_S, Fcvtnu_V (AOpCodeSimd) FP & Umlal_V, Umlsl_V, Saddl_V, Ssubl_V, Usubl_V instructions; add 8 FP & 16 S/Umlal_V, S/Umlsl_V, S/Uaddl_V, S/Usubl_V Tests. (#390)

* Update AOpCodeTable.cs

* Update AInstEmitSimdCvt.cs

* Update Pseudocode.cs

* Update Instructions.cs

* Update CpuTestSimd.cs

* Update AOpCodeTable.cs

* Update AInstEmitSimdArithmetic.cs

* Update Instructions.cs

* Update CpuTestSimdReg.cs

* Update CpuTestSimd.cs

* Update AOpCodeTable.cs

* Update AInstEmitSimdArithmetic.cs

* Update Instructions.cs

* Update CpuTestSimdReg.cs

* Add QCFlagBit.

* Add QCFlagBit.
This commit is contained in:
LDj3SNuD 2018-09-01 16:52:51 +02:00 committed by gdkchan
parent 326777ca4a
commit 42e4e02a64
7 changed files with 2208 additions and 286 deletions

View file

@ -255,6 +255,10 @@ namespace ChocolArm64
SetA64("x00111100x110000000000xxxxxxxxxx", AInstEmit.Fcvtms_Gp, typeof(AOpCodeSimdCvt));
SetA64("x00111100x110001000000xxxxxxxxxx", AInstEmit.Fcvtmu_Gp, typeof(AOpCodeSimdCvt));
SetA64("0x0011100x100001011010xxxxxxxxxx", AInstEmit.Fcvtn_V, typeof(AOpCodeSimd));
SetA64("010111100x100001101010xxxxxxxxxx", AInstEmit.Fcvtns_S, typeof(AOpCodeSimd));
SetA64("0>0011100<100001101010xxxxxxxxxx", AInstEmit.Fcvtns_V, typeof(AOpCodeSimd));
SetA64("011111100x100001101010xxxxxxxxxx", AInstEmit.Fcvtnu_S, typeof(AOpCodeSimd));
SetA64("0>1011100<100001101010xxxxxxxxxx", AInstEmit.Fcvtnu_V, typeof(AOpCodeSimd));
SetA64("x00111100x101000000000xxxxxxxxxx", AInstEmit.Fcvtps_Gp, typeof(AOpCodeSimdCvt));
SetA64("x00111100x101001000000xxxxxxxxxx", AInstEmit.Fcvtpu_Gp, typeof(AOpCodeSimdCvt));
SetA64("x00111100x111000000000xxxxxxxxxx", AInstEmit.Fcvtzs_Gp, typeof(AOpCodeSimdCvt));
@ -365,6 +369,7 @@ namespace ChocolArm64
SetA64("0x001110<<1xxxxx011101xxxxxxxxxx", AInstEmit.Sabd_V, typeof(AOpCodeSimdReg));
SetA64("0x001110<<1xxxxx011100xxxxxxxxxx", AInstEmit.Sabdl_V, typeof(AOpCodeSimdReg));
SetA64("0x001110<<100000011010xxxxxxxxxx", AInstEmit.Sadalp_V, typeof(AOpCodeSimd));
SetA64("0x001110<<1xxxxx000000xxxxxxxxxx", AInstEmit.Saddl_V, typeof(AOpCodeSimdReg));
SetA64("0x001110<<100000001010xxxxxxxxxx", AInstEmit.Saddlp_V, typeof(AOpCodeSimd));
SetA64("0x001110<<1xxxxx000100xxxxxxxxxx", AInstEmit.Saddw_V, typeof(AOpCodeSimdReg));
SetA64("x0011110xx100010000000xxxxxxxxxx", AInstEmit.Scvtf_Gp, typeof(AOpCodeSimdCvt));
@ -419,6 +424,7 @@ namespace ChocolArm64
SetA64("0100111101xxxxxx000001xxxxxxxxxx", AInstEmit.Sshr_V, typeof(AOpCodeSimdShImm));
SetA64("0x00111100>>>xxx000101xxxxxxxxxx", AInstEmit.Ssra_V, typeof(AOpCodeSimdShImm));
SetA64("0100111101xxxxxx000101xxxxxxxxxx", AInstEmit.Ssra_V, typeof(AOpCodeSimdShImm));
SetA64("0x001110<<1xxxxx001000xxxxxxxxxx", AInstEmit.Ssubl_V, typeof(AOpCodeSimdReg));
SetA64("0x001110<<1xxxxx001100xxxxxxxxxx", AInstEmit.Ssubw_V, typeof(AOpCodeSimdReg));
SetA64("0x00110000000000xxxxxxxxxxxxxxxx", AInstEmit.St__Vms, typeof(AOpCodeSimdMemMs));
SetA64("0x001100100xxxxxxxxxxxxxxxxxxxxx", AInstEmit.St__Vms, typeof(AOpCodeSimdMemMs));
@ -457,6 +463,8 @@ namespace ChocolArm64
SetA64("0x101110<<1xxxxx101001xxxxxxxxxx", AInstEmit.Umaxp_V, typeof(AOpCodeSimdReg));
SetA64("0x101110<<1xxxxx011011xxxxxxxxxx", AInstEmit.Umin_V, typeof(AOpCodeSimdReg));
SetA64("0x101110<<1xxxxx101011xxxxxxxxxx", AInstEmit.Uminp_V, typeof(AOpCodeSimdReg));
SetA64("0x101110<<1xxxxx100000xxxxxxxxxx", AInstEmit.Umlal_V, typeof(AOpCodeSimdReg));
SetA64("0x101110<<1xxxxx101000xxxxxxxxxx", AInstEmit.Umlsl_V, typeof(AOpCodeSimdReg));
SetA64("0x001110000xxxxx001111xxxxxxxxxx", AInstEmit.Umov_S, typeof(AOpCodeSimdIns));
SetA64("0x101110<<1xxxxx110000xxxxxxxxxx", AInstEmit.Umull_V, typeof(AOpCodeSimdReg));
SetA64("01111110xx1xxxxx000011xxxxxxxxxx", AInstEmit.Uqadd_S, typeof(AOpCodeSimdReg));
@ -475,6 +483,7 @@ namespace ChocolArm64
SetA64("0>101110<<100000001110xxxxxxxxxx", AInstEmit.Usqadd_V, typeof(AOpCodeSimd));
SetA64("0x10111100>>>xxx000101xxxxxxxxxx", AInstEmit.Usra_V, typeof(AOpCodeSimdShImm));
SetA64("0110111101xxxxxx000101xxxxxxxxxx", AInstEmit.Usra_V, typeof(AOpCodeSimdShImm));
SetA64("0x101110<<1xxxxx001000xxxxxxxxxx", AInstEmit.Usubl_V, typeof(AOpCodeSimdReg));
SetA64("0x101110<<1xxxxx001100xxxxxxxxxx", AInstEmit.Usubw_V, typeof(AOpCodeSimdReg));
SetA64("0>001110<<0xxxxx000110xxxxxxxxxx", AInstEmit.Uzp1_V, typeof(AOpCodeSimdReg));
SetA64("0>001110<<0xxxxx010110xxxxxxxxxx", AInstEmit.Uzp2_V, typeof(AOpCodeSimdReg));

View file

@ -1032,6 +1032,11 @@ namespace ChocolArm64.Instruction
EmitAddLongPairwise(Context, Signed: true, Accumulate: true);
}
public static void Saddl_V(AILEmitterCtx Context)
{
EmitVectorWidenRnRmBinaryOpSx(Context, () => Context.Emit(OpCodes.Add));
}
public static void Saddlp_V(AILEmitterCtx Context)
{
EmitAddLongPairwise(Context, Signed: true, Accumulate: false);
@ -1217,6 +1222,11 @@ namespace ChocolArm64.Instruction
});
}
public static void Ssubl_V(AILEmitterCtx Context)
{
EmitVectorWidenRnRmBinaryOpSx(Context, () => Context.Emit(OpCodes.Sub));
}
public static void Ssubw_V(AILEmitterCtx Context)
{
EmitVectorWidenRmBinaryOpSx(Context, () => Context.Emit(OpCodes.Sub));
@ -1391,6 +1401,24 @@ namespace ChocolArm64.Instruction
EmitVectorPairwiseOpZx(Context, () => Context.EmitCall(MthdInfo));
}
public static void Umlal_V(AILEmitterCtx Context)
{
EmitVectorWidenRnRmTernaryOpZx(Context, () =>
{
Context.Emit(OpCodes.Mul);
Context.Emit(OpCodes.Add);
});
}
public static void Umlsl_V(AILEmitterCtx Context)
{
EmitVectorWidenRnRmTernaryOpZx(Context, () =>
{
Context.Emit(OpCodes.Mul);
Context.Emit(OpCodes.Sub);
});
}
public static void Umull_V(AILEmitterCtx Context)
{
EmitVectorWidenRnRmBinaryOpZx(Context, () => Context.Emit(OpCodes.Mul));
@ -1450,6 +1478,11 @@ namespace ChocolArm64.Instruction
EmitVectorSaturatingBinaryOpZx(Context, SaturatingFlags.Accumulate);
}
public static void Usubl_V(AILEmitterCtx Context)
{
EmitVectorWidenRnRmBinaryOpZx(Context, () => Context.Emit(OpCodes.Sub));
}
public static void Usubw_V(AILEmitterCtx Context)
{
EmitVectorWidenRmBinaryOpZx(Context, () => Context.Emit(OpCodes.Sub));

View file

@ -106,6 +106,26 @@ namespace ChocolArm64.Instruction
}
}
public static void Fcvtns_S(AILEmitterCtx Context)
{
EmitFcvtn(Context, Signed: true, Scalar: true);
}
public static void Fcvtns_V(AILEmitterCtx Context)
{
EmitFcvtn(Context, Signed: true, Scalar: false);
}
public static void Fcvtnu_S(AILEmitterCtx Context)
{
EmitFcvtn(Context, Signed: false, Scalar: true);
}
public static void Fcvtnu_V(AILEmitterCtx Context)
{
EmitFcvtn(Context, Signed: false, Scalar: false);
}
public static void Fcvtps_Gp(AILEmitterCtx Context)
{
EmitFcvt_s_Gp(Context, () => EmitUnaryMathCall(Context, nameof(Math.Ceiling)));
@ -250,6 +270,54 @@ namespace ChocolArm64.Instruction
}
}
private static void EmitFcvtn(AILEmitterCtx Context, bool Signed, bool Scalar)
{
AOpCodeSimd Op = (AOpCodeSimd)Context.CurrOp;
int SizeF = Op.Size & 1;
int SizeI = SizeF + 2;
int Bytes = Op.GetBitsCount() >> 3;
int Elems = !Scalar ? Bytes >> SizeI : 1;
if (Scalar && (SizeF == 0))
{
EmitVectorZeroLowerTmp(Context);
}
for (int Index = 0; Index < Elems; Index++)
{
EmitVectorExtractF(Context, Op.Rn, Index, SizeF);
EmitRoundMathCall(Context, MidpointRounding.ToEven);
if (SizeF == 0)
{
AVectorHelper.EmitCall(Context, Signed
? nameof(AVectorHelper.SatF32ToS32)
: nameof(AVectorHelper.SatF32ToU32));
Context.Emit(OpCodes.Conv_U8);
}
else /* if (SizeF == 1) */
{
AVectorHelper.EmitCall(Context, Signed
? nameof(AVectorHelper.SatF64ToS64)
: nameof(AVectorHelper.SatF64ToU64));
}
EmitVectorInsertTmp(Context, Index, SizeI);
}
Context.EmitLdvectmp();
Context.EmitStvec(Op.Rd);
if ((Op.RegisterSize == ARegisterSize.SIMD64) || Scalar)
{
EmitVectorZeroUpper(Context, Op.Rd);
}
}
private static void EmitFcvt_s_Gp(AILEmitterCtx Context, Action Emit)
{
EmitFcvt___Gp(Context, Emit, true);
@ -569,4 +637,4 @@ namespace ChocolArm64.Instruction
}
}
}
}
}

File diff suppressed because it is too large Load diff

File diff suppressed because it is too large Load diff

View file

@ -2685,6 +2685,154 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// fcvtns_advsimd.html#FCVTNS_asisdmisc_R
public static void Fcvtns_S(Bits sz, Bits Rn, Bits Rd)
{
const bool U = false;
const bool o2 = false;
const bool o1 = false;
/* Decode Scalar */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int esize = 32 << (int)UInt(sz);
int datasize = esize;
int elements = 1;
FPRounding rounding = FPDecodeRounding(Bits.Concat(o1, o2));
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand = V(datasize, n);
Bits element;
for (int e = 0; e <= elements - 1; e++)
{
element = Elem(operand, e, esize);
Elem(result, e, esize, FPToFixed(esize, element, 0, unsigned, FPCR, rounding));
}
V(d, result);
}
// fcvtns_advsimd.html#FCVTNS_asimdmisc_R
public static void Fcvtns_V(bool Q, Bits sz, Bits Rn, Bits Rd)
{
const bool U = false;
const bool o2 = false;
const bool o1 = false;
/* Decode Vector */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
/* if sz:Q == '10' then ReservedValue(); */
int esize = 32 << (int)UInt(sz);
int datasize = (Q ? 128 : 64);
int elements = datasize / esize;
FPRounding rounding = FPDecodeRounding(Bits.Concat(o1, o2));
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand = V(datasize, n);
Bits element;
for (int e = 0; e <= elements - 1; e++)
{
element = Elem(operand, e, esize);
Elem(result, e, esize, FPToFixed(esize, element, 0, unsigned, FPCR, rounding));
}
V(d, result);
}
// fcvtnu_advsimd.html#FCVTNU_asisdmisc_R
public static void Fcvtnu_S(Bits sz, Bits Rn, Bits Rd)
{
const bool U = true;
const bool o2 = false;
const bool o1 = false;
/* Decode Scalar */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int esize = 32 << (int)UInt(sz);
int datasize = esize;
int elements = 1;
FPRounding rounding = FPDecodeRounding(Bits.Concat(o1, o2));
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand = V(datasize, n);
Bits element;
for (int e = 0; e <= elements - 1; e++)
{
element = Elem(operand, e, esize);
Elem(result, e, esize, FPToFixed(esize, element, 0, unsigned, FPCR, rounding));
}
V(d, result);
}
// fcvtnu_advsimd.html#FCVTNU_asimdmisc_R
public static void Fcvtnu_V(bool Q, Bits sz, Bits Rn, Bits Rd)
{
const bool U = true;
const bool o2 = false;
const bool o1 = false;
/* Decode Vector */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
/* if sz:Q == '10' then ReservedValue(); */
int esize = 32 << (int)UInt(sz);
int datasize = (Q ? 128 : 64);
int elements = datasize / esize;
FPRounding rounding = FPDecodeRounding(Bits.Concat(o1, o2));
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand = V(datasize, n);
Bits element;
for (int e = 0; e <= elements - 1; e++)
{
element = Elem(operand, e, esize);
Elem(result, e, esize, FPToFixed(esize, element, 0, unsigned, FPCR, rounding));
}
V(d, result);
}
// neg_advsimd.html#NEG_asisdmisc_R
public static void Neg_S(Bits size, Bits Rn, Bits Rd)
{
@ -5122,6 +5270,57 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// saddl_advsimd.html
public static void Saddl_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = false;
const bool o1 = false;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
BigInteger element1;
BigInteger element2;
BigInteger sum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
if (sub_op)
{
sum = element1 - element2;
}
else
{
sum = element1 + element2;
}
Elem(result, e, 2 * esize, sum.SubBigInteger(2 * esize - 1, 0));
}
V(d, result);
}
// saddw_advsimd.html
public static void Saddw_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
@ -5333,6 +5532,116 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// smlal_advsimd_vec.html
public static void Smlal_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = false;
const bool o1 = false;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
Bits operand3 = V(2 * datasize, d);
BigInteger element1;
BigInteger element2;
Bits product;
Bits accum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
product = (element1 * element2).SubBigInteger(2 * esize - 1, 0);
if (sub_op)
{
accum = Elem(operand3, e, 2 * esize) - product;
}
else
{
accum = Elem(operand3, e, 2 * esize) + product;
}
Elem(result, e, 2 * esize, accum);
}
V(d, result);
}
// smlsl_advsimd_vec.html
public static void Smlsl_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = false;
const bool o1 = true;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
Bits operand3 = V(2 * datasize, d);
BigInteger element1;
BigInteger element2;
Bits product;
Bits accum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
product = (element1 * element2).SubBigInteger(2 * esize - 1, 0);
if (sub_op)
{
accum = Elem(operand3, e, 2 * esize) - product;
}
else
{
accum = Elem(operand3, e, 2 * esize) + product;
}
Elem(result, e, 2 * esize, accum);
}
V(d, result);
}
// sqadd_advsimd.html#SQADD_asisdsame_only
public static void Sqadd_S(Bits size, Bits Rm, Bits Rn, Bits Rd)
{
@ -5771,6 +6080,57 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// ssubl_advsimd.html
public static void Ssubl_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = false;
const bool o1 = true;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
BigInteger element1;
BigInteger element2;
BigInteger sum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
if (sub_op)
{
sum = element1 - element2;
}
else
{
sum = element1 + element2;
}
Elem(result, e, 2 * esize, sum.SubBigInteger(2 * esize - 1, 0));
}
V(d, result);
}
// ssubw_advsimd.html
public static void Ssubw_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
@ -6212,6 +6572,57 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// uaddl_advsimd.html
public static void Uaddl_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = true;
const bool o1 = false;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
BigInteger element1;
BigInteger element2;
BigInteger sum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
if (sub_op)
{
sum = element1 - element2;
}
else
{
sum = element1 + element2;
}
Elem(result, e, 2 * esize, sum.SubBigInteger(2 * esize - 1, 0));
}
V(d, result);
}
// uaddw_advsimd.html
public static void Uaddw_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
@ -6345,6 +6756,116 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// umlal_advsimd_vec.html
public static void Umlal_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = true;
const bool o1 = false;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
Bits operand3 = V(2 * datasize, d);
BigInteger element1;
BigInteger element2;
Bits product;
Bits accum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
product = (element1 * element2).SubBigInteger(2 * esize - 1, 0);
if (sub_op)
{
accum = Elem(operand3, e, 2 * esize) - product;
}
else
{
accum = Elem(operand3, e, 2 * esize) + product;
}
Elem(result, e, 2 * esize, accum);
}
V(d, result);
}
// umlsl_advsimd_vec.html
public static void Umlsl_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = true;
const bool o1 = true;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
Bits operand3 = V(2 * datasize, d);
BigInteger element1;
BigInteger element2;
Bits product;
Bits accum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
product = (element1 * element2).SubBigInteger(2 * esize - 1, 0);
if (sub_op)
{
accum = Elem(operand3, e, 2 * esize) - product;
}
else
{
accum = Elem(operand3, e, 2 * esize) + product;
}
Elem(result, e, 2 * esize, accum);
}
V(d, result);
}
// uqadd_advsimd.html#UQADD_asisdsame_only
public static void Uqadd_S(Bits size, Bits Rm, Bits Rn, Bits Rd)
{
@ -6579,6 +7100,57 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// usubl_advsimd.html
public static void Usubl_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = true;
const bool o1 = true;
/* Decode */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = 64;
int part = (int)UInt(Q);
int elements = datasize / esize;
bool sub_op = (o1 == true);
bool unsigned = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(2 * datasize);
Bits operand1 = Vpart(datasize, n, part);
Bits operand2 = Vpart(datasize, m, part);
BigInteger element1;
BigInteger element2;
BigInteger sum;
for (int e = 0; e <= elements - 1; e++)
{
element1 = Int(Elem(operand1, e, esize), unsigned);
element2 = Int(Elem(operand2, e, esize), unsigned);
if (sub_op)
{
sum = element1 - element2;
}
else
{
sum = element1 + element2;
}
Elem(result, e, 2 * esize, sum.SubBigInteger(2 * esize - 1, 0));
}
V(d, result);
}
// usubw_advsimd.html
public static void Usubw_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{

View file

@ -5,21 +5,19 @@
// https://alastairreid.github.io/asl-lexical-syntax/
// | ------------------------|----------------------------------- |
// | ASL | C# |
// | ------------------------|----------------------------------- |
// | bit, bits(1); boolean | bool |
// | bits | Bits |
// | integer | BigInteger, int |
// | real | decimal |
// | ------------------------|----------------------------------- |
// | '0'; FALSE | false |
// | '1'; TRUE | true |
// | '010' | "010" |
// | bitsX IN {bitsY, bitsZ} | (bitsX == bitsY || bitsX == bitsZ) |
// | DIV | / |
// | MOD | % |
// | ------------------------|----------------------------------- |
// | ------------------------|-------------------------------- |
// | ASL | C# |
// | ------------------------|-------------------------------- |
// | bit, bits(1); boolean | bool |
// | bits | Bits |
// | integer | BigInteger, int |
// | real | decimal; double, float |
// | ------------------------|-------------------------------- |
// | '0'; FALSE | false |
// | '1'; TRUE | true |
// | '010' | "010" |
// | DIV, MOD | /, % |
// | ------------------------|-------------------------------- |
using System;
using System.Numerics;
@ -107,6 +105,7 @@ namespace Ryujinx.Tests.Cpu.Tester
/* SP_EL1 = bits(64) UNKNOWN; */
SP_EL1.SetAll(false);
FPCR.SetAll(false); // TODO: Add named fields.
FPSR.SetAll(false); // TODO: Add named fields.
}
@ -458,6 +457,7 @@ namespace Ryujinx.Tests.Cpu.Tester
#endregion
#region "instrs/vector/reduce/reduceop/"
// shared_pseudocode.html#impl-aarch64.Reduce.3
public static Bits Reduce(ReduceOp op, Bits input, int esize)
{
int N = input.Count;
@ -528,6 +528,7 @@ namespace Ryujinx.Tests.Cpu.Tester
SP_EL0 = new Bits(64, false);
SP_EL1 = new Bits(64, false);
FPCR = new Bits(32, false); // TODO: Add named fields.
FPSR = new Bits(32, false); // TODO: Add named fields.
PSTATE.N = false;
@ -817,6 +818,36 @@ namespace Ryujinx.Tests.Cpu.Tester
return (decimal)value;
}
/* */
public static float Real_32(BigInteger value)
{
if (value == BigInteger.Pow((BigInteger)2.0f, 1000))
{
return float.PositiveInfinity;
}
if (value == -BigInteger.Pow((BigInteger)2.0f, 1000))
{
return float.NegativeInfinity;
}
return (float)value;
}
/* */
public static double Real_64(BigInteger value)
{
if (value == BigInteger.Pow((BigInteger)2.0, 10000))
{
return double.PositiveInfinity;
}
if (value == -BigInteger.Pow((BigInteger)2.0, 10000))
{
return double.NegativeInfinity;
}
return (double)value;
}
// shared_pseudocode.html#impl-shared.ROR.2
public static Bits ROR(Bits x, int shift)
{
@ -881,6 +912,36 @@ namespace Ryujinx.Tests.Cpu.Tester
return (BigInteger)Decimal.Floor(x);
}
/* */
public static BigInteger RoundDown_32(float x)
{
if (float.IsPositiveInfinity(x))
{
return BigInteger.Pow((BigInteger)2.0f, 1000);
}
if (float.IsNegativeInfinity(x))
{
return -BigInteger.Pow((BigInteger)2.0f, 1000);
}
return (BigInteger)MathF.Floor(x);
}
/* */
public static BigInteger RoundDown_64(double x)
{
if (double.IsPositiveInfinity(x))
{
return BigInteger.Pow((BigInteger)2.0, 10000);
}
if (double.IsNegativeInfinity(x))
{
return -BigInteger.Pow((BigInteger)2.0, 10000);
}
return (BigInteger)Math.Floor(x);
}
// shared_pseudocode.html#impl-shared.RoundTowardsZero.1
public static BigInteger RoundTowardsZero(decimal x)
{
@ -1091,6 +1152,398 @@ namespace Ryujinx.Tests.Cpu.Tester
}
#endregion
#region "functions/float/fpdecoderounding/"
/* shared_pseudocode.html#impl-shared.FPDecodeRounding.1 */
public static FPRounding FPDecodeRounding(Bits rmode)
{
switch (rmode)
{
default:
case Bits bits when bits == "00":
return FPRounding.FPRounding_TIEEVEN; // N
case Bits bits when bits == "01":
return FPRounding.FPRounding_POSINF; // P
case Bits bits when bits == "10":
return FPRounding.FPRounding_NEGINF; // M
case Bits bits when bits == "11":
return FPRounding.FPRounding_ZERO; // Z
}
}
#endregion
#region "functions/float/fpexc/"
// shared_pseudocode.html#FPExc
public enum FPExc {FPExc_InvalidOp, FPExc_DivideByZero, FPExc_Overflow,
FPExc_Underflow, FPExc_Inexact, FPExc_InputDenorm};
#endregion
#region "functions/float/fpprocessexception/"
// shared_pseudocode.html#impl-shared.FPProcessException.2
public static void FPProcessException(FPExc exception, Bits _fpcr)
{
Bits fpcr = new Bits(_fpcr); // Clone.
int cumul;
// Determine the cumulative exception bit number
switch (exception)
{
default:
case FPExc.FPExc_InvalidOp: cumul = 0; break;
case FPExc.FPExc_DivideByZero: cumul = 1; break;
case FPExc.FPExc_Overflow: cumul = 2; break;
case FPExc.FPExc_Underflow: cumul = 3; break;
case FPExc.FPExc_Inexact: cumul = 4; break;
case FPExc.FPExc_InputDenorm: cumul = 7; break;
}
int enable = cumul + 8;
if (fpcr[enable])
{
// Trapping of the exception enabled.
// It is IMPLEMENTATION DEFINED whether the enable bit may be set at all, and
// if so then how exceptions may be accumulated before calling FPTrapException()
/* IMPLEMENTATION_DEFINED "floating-point trap handling"; */
throw new NotImplementedException();
}/*
else if (UsingAArch32())
{
// Set the cumulative exception bit
FPSCR<cumul> = '1';
}*/
else
{
// Set the cumulative exception bit
FPSR[cumul] = true;
}
}
#endregion
#region "functions/float/fprounding/"
// shared_pseudocode.html#FPRounding
public enum FPRounding {FPRounding_TIEEVEN, FPRounding_POSINF,
FPRounding_NEGINF, FPRounding_ZERO,
FPRounding_TIEAWAY, FPRounding_ODD};
#endregion
#region "functions/float/fptofixed/"
/* shared_pseudocode.html#impl-shared.FPToFixed.5 */
public static Bits FPToFixed(int M, Bits op, int fbits, bool unsigned, Bits _fpcr, FPRounding rounding)
{
int N = op.Count;
/* assert N IN {16,32,64}; */
/* assert M IN {16,32,64}; */
/* assert fbits >= 0; */
/* assert rounding != FPRounding_ODD; */
Bits fpcr = new Bits(_fpcr); // Clone.
if (N == 16)
{
throw new NotImplementedException();
}
else if (N == 32)
{
// Unpack using fpcr to determine if subnormals are flushed-to-zero
(FPType type, bool sign, float value) = FPUnpack_32(op, fpcr);
// If NaN, set cumulative flag or take exception
if (type == FPType.FPType_SNaN || type == FPType.FPType_QNaN)
{
FPProcessException(FPExc.FPExc_InvalidOp, fpcr);
}
// Scale by fractional bits and produce integer rounded towards minus-infinity
value = value * MathF.Pow(2.0f, fbits);
BigInteger int_result = RoundDown_32(value);
float error = value - Real_32(int_result);
if (float.IsNaN(error))
{
error = 0.0f;
}
// Determine whether supplied rounding mode requires an increment
bool round_up;
switch (rounding)
{
default:
case FPRounding.FPRounding_TIEEVEN:
round_up = (error > 0.5f || (error == 0.5f && int_result.SubBigInteger(0)));
break;
case FPRounding.FPRounding_POSINF:
round_up = (error != 0.0f);
break;
case FPRounding.FPRounding_NEGINF:
round_up = false;
break;
case FPRounding.FPRounding_ZERO:
round_up = (error != 0.0f && int_result < (BigInteger)0);
break;
case FPRounding.FPRounding_TIEAWAY:
round_up = (error > 0.5f || (error == 0.5f && int_result >= (BigInteger)0));
break;
}
if (round_up)
{
int_result = int_result + 1;
}
// Generate saturated result and exceptions
(Bits result, bool overflow) = SatQ(int_result, M, unsigned);
if (overflow)
{
FPProcessException(FPExc.FPExc_InvalidOp, fpcr);
}
else if (error != 0.0f)
{
FPProcessException(FPExc.FPExc_Inexact, fpcr);
}
return result;
}
else /* if (N == 64) */
{
// Unpack using fpcr to determine if subnormals are flushed-to-zero
(FPType type, bool sign, double value) = FPUnpack_64(op, fpcr);
// If NaN, set cumulative flag or take exception
if (type == FPType.FPType_SNaN || type == FPType.FPType_QNaN)
{
FPProcessException(FPExc.FPExc_InvalidOp, fpcr);
}
// Scale by fractional bits and produce integer rounded towards minus-infinity
value = value * Math.Pow(2.0, fbits);
BigInteger int_result = RoundDown_64(value);
double error = value - Real_64(int_result);
if (double.IsNaN(error))
{
error = 0.0;
}
// Determine whether supplied rounding mode requires an increment
bool round_up;
switch (rounding)
{
default:
case FPRounding.FPRounding_TIEEVEN:
round_up = (error > 0.5 || (error == 0.5 && int_result.SubBigInteger(0)));
break;
case FPRounding.FPRounding_POSINF:
round_up = (error != 0.0);
break;
case FPRounding.FPRounding_NEGINF:
round_up = false;
break;
case FPRounding.FPRounding_ZERO:
round_up = (error != 0.0 && int_result < (BigInteger)0);
break;
case FPRounding.FPRounding_TIEAWAY:
round_up = (error > 0.5 || (error == 0.5 && int_result >= (BigInteger)0));
break;
}
if (round_up)
{
int_result = int_result + 1;
}
// Generate saturated result and exceptions
(Bits result, bool overflow) = SatQ(int_result, M, unsigned);
if (overflow)
{
FPProcessException(FPExc.FPExc_InvalidOp, fpcr);
}
else if (error != 0.0)
{
FPProcessException(FPExc.FPExc_Inexact, fpcr);
}
return result;
}
}
#endregion
#region "functions/float/fptype/"
// shared_pseudocode.html#FPType
public enum FPType {FPType_Nonzero, FPType_Zero, FPType_Infinity,
FPType_QNaN, FPType_SNaN};
#endregion
#region "functions/float/fpunpack/"
/* shared_pseudocode.html#impl-shared.FPUnpack.2 */
/* shared_pseudocode.html#impl-shared.FPUnpackBase.2 */
/*public static (FPType, bool, real) FPUnpack_16(Bits fpval, Bits _fpcr)
{
int N = fpval.Count;
// assert N == 16;
Bits fpcr = new Bits(_fpcr); // Clone.
fpcr[26] = false;
return FPUnpackBase_16(fpval, fpcr);
}*/
public static (FPType, bool, float) FPUnpack_32(Bits fpval, Bits _fpcr)
{
int N = fpval.Count;
/* assert N == 32; */
Bits fpcr = new Bits(_fpcr); // Clone.
FPType type;
float value;
bool sign = fpval[31];
Bits exp32 = fpval[30, 23];
Bits frac32 = fpval[22, 0];
if (IsZero(exp32))
{
// Produce zero if value is zero or flush-to-zero is selected.
if (IsZero(frac32) || fpcr[24])
{
type = FPType.FPType_Zero;
value = 0.0f;
// Denormalized input flushed to zero
if (!IsZero(frac32))
{
FPProcessException(FPExc.FPExc_InputDenorm, fpcr);
}
}
else
{
type = FPType.FPType_Nonzero;
value = MathF.Pow(2.0f, -126) * (Real_32(UInt(frac32)) * MathF.Pow(2.0f, -23));
}
}
else if (IsOnes(exp32))
{
if (IsZero(frac32))
{
type = FPType.FPType_Infinity;
/* value = 2.0^1000000; */
value = MathF.Pow(2.0f, 1000);
}
else
{
type = frac32[22] ? FPType.FPType_QNaN : FPType.FPType_SNaN;
value = 0.0f;
}
}
else
{
type = FPType.FPType_Nonzero;
value = MathF.Pow(2.0f, (int)UInt(exp32) - 127) * (1.0f + Real_32(UInt(frac32)) * MathF.Pow(2.0f, -23));
}
if (sign)
{
value = -value;
}
return (type, sign, value);
}
public static (FPType, bool, double) FPUnpack_64(Bits fpval, Bits _fpcr)
{
int N = fpval.Count;
/* assert N == 64; */
Bits fpcr = new Bits(_fpcr); // Clone.
FPType type;
double value;
bool sign = fpval[63];
Bits exp64 = fpval[62, 52];
Bits frac64 = fpval[51, 0];
if (IsZero(exp64))
{
// Produce zero if value is zero or flush-to-zero is selected.
if (IsZero(frac64) || fpcr[24])
{
type = FPType.FPType_Zero;
value = 0.0;
// Denormalized input flushed to zero
if (!IsZero(frac64))
{
FPProcessException(FPExc.FPExc_InputDenorm, fpcr);
}
}
else
{
type = FPType.FPType_Nonzero;
value = Math.Pow(2.0, -1022) * (Real_64(UInt(frac64)) * Math.Pow(2.0, -52));
}
}
else if (IsOnes(exp64))
{
if (IsZero(frac64))
{
type = FPType.FPType_Infinity;
/* value = 2.0^1000000; */
value = Math.Pow(2.0, 10000);
}
else
{
type = frac64[51] ? FPType.FPType_QNaN : FPType.FPType_SNaN;
value = 0.0;
}
}
else
{
type = FPType.FPType_Nonzero;
value = Math.Pow(2.0, (int)UInt(exp64) - 1023) * (1.0 + Real_64(UInt(frac64)) * Math.Pow(2.0, -52));
}
if (sign)
{
value = -value;
}
return (type, sign, value);
}
/* shared_pseudocode.html#impl-shared.FPUnpackCV.2 */
/* shared_pseudocode.html#impl-shared.FPUnpackBase.2 */
/*public static (FPType, bool, real) FPUnpackCV_16(Bits fpval, Bits _fpcr)
{
int N = fpval.Count;
// assert N == 16;
Bits fpcr = new Bits(_fpcr); // Clone.
fpcr[19] = false;
return FPUnpackBase_16(fpval, fpcr);
}*/
public static (FPType, bool, float) FPUnpackCV_32(Bits fpval, Bits _fpcr)
{
return FPUnpack_32(fpval, _fpcr);
}
public static (FPType, bool, double) FPUnpackCV_64(Bits fpval, Bits _fpcr)
{
return FPUnpack_64(fpval, _fpcr);
}
#endregion
#region "functions/integer/"
/* shared_pseudocode.html#impl-shared.AddWithCarry.3 */
public static (Bits, Bits) AddWithCarry(int N, Bits x, Bits y, bool carry_in)
@ -1117,7 +1570,12 @@ namespace Ryujinx.Tests.Cpu.Tester
public static Bits SP_EL0;
public static Bits SP_EL1;
public static Bits FPCR; // TODO: Add named fields.
// [ 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 22 | 21 20 | 19 | 18 17 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 ]
// [ 0 | 0 | 0 | 0 | 0 | AHP | DN | FZ | RMode | Stride | FZ16 | Len | IDE | 0 | 0 | IXE | UFE | OFE | DZE | IOE | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 ]
public static Bits FPSR; // TODO: Add named fields.
// [ 31 | 30 | 29 | 28 | 27 | 26 | 25 | 24 | 23 | 22 | 21 | 20 | 19 | 18 | 17 | 16 | 15 | 14 | 13 | 12 | 11 | 10 | 9 | 8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0 ]
// [ N | Z | C | V | QC | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | IDC | 0 | 0 | IXC | UFC | OFC | DZC | IOC ]
#endregion
#region "functions/system/"
@ -1178,6 +1636,8 @@ namespace Ryujinx.Tests.Cpu.Tester
/* shared_pseudocode.html#impl-shared.HaveEL.1 */
public static bool HaveEL(Bits el)
{
// TODO: Implement ASL: "IN" as C#: "Bits.In()".
/* if el IN {EL1,EL0} then */
if (el == EL1 || el == EL0)
{
return true; // EL1 and EL0 must exist