Add Sqdmulh_S, Sqdmulh_V, Sqrdmulh_S, Sqrdmulh_V instructions; add 6 Tests. Now all saturating methods are on ASoftFallback. (#334)

* Update Instructions.cs

* Update CpuTestSimd.cs

* Update CpuTestSimdReg.cs

* Update AOpCodeTable.cs

* Update AInstEmitSimdArithmetic.cs

* Update AInstEmitSimdHelper.cs

* Update ASoftFallback.cs

* Update CpuTestAlu.cs

* Update CpuTestAluImm.cs

* Update CpuTestAluRs.cs

* Update CpuTestAluRx.cs

* Update CpuTestBfm.cs

* Update CpuTestCcmpImm.cs

* Update CpuTestCcmpReg.cs

* Update CpuTestCsel.cs

* Update CpuTestMov.cs

* Update CpuTestMul.cs

* Update Ryujinx.Tests.csproj

* Update Ryujinx.csproj

* Update Luea.csproj

* Update Ryujinx.ShaderTools.csproj

* Address PR feedback (further tested).

* Address PR feedback.
This commit is contained in:
LDj3SNuD 2018-08-10 19:27:15 +02:00 committed by gdkchan
parent 267af1f0f7
commit 02a6fdcd13
21 changed files with 834 additions and 314 deletions

View file

@ -380,8 +380,16 @@ namespace ChocolArm64
SetA64("0>001110<<100000011110xxxxxxxxxx", AInstEmit.Sqabs_V, typeof(AOpCodeSimd));
SetA64("01011110xx1xxxxx000011xxxxxxxxxx", AInstEmit.Sqadd_S, typeof(AOpCodeSimdReg));
SetA64("0>001110<<1xxxxx000011xxxxxxxxxx", AInstEmit.Sqadd_V, typeof(AOpCodeSimdReg));
SetA64("01011110011xxxxx101101xxxxxxxxxx", AInstEmit.Sqdmulh_S, typeof(AOpCodeSimdReg));
SetA64("01011110101xxxxx101101xxxxxxxxxx", AInstEmit.Sqdmulh_S, typeof(AOpCodeSimdReg));
SetA64("0x001110011xxxxx101101xxxxxxxxxx", AInstEmit.Sqdmulh_V, typeof(AOpCodeSimdReg));
SetA64("0x001110101xxxxx101101xxxxxxxxxx", AInstEmit.Sqdmulh_V, typeof(AOpCodeSimdReg));
SetA64("01111110xx100000011110xxxxxxxxxx", AInstEmit.Sqneg_S, typeof(AOpCodeSimd));
SetA64("0>101110<<100000011110xxxxxxxxxx", AInstEmit.Sqneg_V, typeof(AOpCodeSimd));
SetA64("01111110011xxxxx101101xxxxxxxxxx", AInstEmit.Sqrdmulh_S, typeof(AOpCodeSimdReg));
SetA64("01111110101xxxxx101101xxxxxxxxxx", AInstEmit.Sqrdmulh_S, typeof(AOpCodeSimdReg));
SetA64("0x101110011xxxxx101101xxxxxxxxxx", AInstEmit.Sqrdmulh_V, typeof(AOpCodeSimdReg));
SetA64("0x101110101xxxxx101101xxxxxxxxxx", AInstEmit.Sqrdmulh_V, typeof(AOpCodeSimdReg));
SetA64("0x00111100>>>xxx100111xxxxxxxxxx", AInstEmit.Sqrshrn_V, typeof(AOpCodeSimdShImm));
SetA64("01011110xx1xxxxx001011xxxxxxxxxx", AInstEmit.Sqsub_S, typeof(AOpCodeSimdReg));
SetA64("0>001110<<1xxxxx001011xxxxxxxxxx", AInstEmit.Sqsub_V, typeof(AOpCodeSimdReg));

View file

@ -158,6 +158,42 @@ namespace ChocolArm64.Instruction
Context.MarkLabel(LblTrue);
}
private static void EmitDoublingMultiplyHighHalf(AILEmitterCtx Context, bool Round)
{
AOpCodeSimdReg Op = (AOpCodeSimdReg)Context.CurrOp;
int ESize = 8 << Op.Size;
Context.Emit(OpCodes.Mul);
if (!Round)
{
Context.EmitAsr(ESize - 1);
}
else
{
long RoundConst = 1L << (ESize - 1);
AILLabel LblTrue = new AILLabel();
Context.EmitLsl(1);
Context.EmitLdc_I8(RoundConst);
Context.Emit(OpCodes.Add);
Context.EmitAsr(ESize);
Context.Emit(OpCodes.Dup);
Context.EmitLdc_I8((long)int.MinValue);
Context.Emit(OpCodes.Bne_Un_S, LblTrue);
Context.Emit(OpCodes.Neg);
Context.MarkLabel(LblTrue);
}
}
private static void EmitHighNarrow(AILEmitterCtx Context, Action Emit, bool Round)
{
AOpCodeSimdReg Op = (AOpCodeSimdReg)Context.CurrOp;
@ -1040,6 +1076,16 @@ namespace ChocolArm64.Instruction
EmitVectorSaturatingBinaryOpSx(Context, SaturatingFlags.Add);
}
public static void Sqdmulh_S(AILEmitterCtx Context)
{
EmitSaturatingBinaryOp(Context, () => EmitDoublingMultiplyHighHalf(Context, Round: false), SaturatingFlags.ScalarSx);
}
public static void Sqdmulh_V(AILEmitterCtx Context)
{
EmitSaturatingBinaryOp(Context, () => EmitDoublingMultiplyHighHalf(Context, Round: false), SaturatingFlags.VectorSx);
}
public static void Sqneg_S(AILEmitterCtx Context)
{
EmitScalarSaturatingUnaryOpSx(Context, () => Context.Emit(OpCodes.Neg));
@ -1050,6 +1096,16 @@ namespace ChocolArm64.Instruction
EmitVectorSaturatingUnaryOpSx(Context, () => Context.Emit(OpCodes.Neg));
}
public static void Sqrdmulh_S(AILEmitterCtx Context)
{
EmitSaturatingBinaryOp(Context, () => EmitDoublingMultiplyHighHalf(Context, Round: true), SaturatingFlags.ScalarSx);
}
public static void Sqrdmulh_V(AILEmitterCtx Context)
{
EmitSaturatingBinaryOp(Context, () => EmitDoublingMultiplyHighHalf(Context, Round: true), SaturatingFlags.VectorSx);
}
public static void Sqsub_S(AILEmitterCtx Context)
{
EmitScalarSaturatingBinaryOpSx(Context, SaturatingFlags.Sub);

View file

@ -804,7 +804,7 @@ namespace ChocolArm64.Instruction
ScalarZx = Scalar,
VectorSx = Signed,
VectorZx = 0,
VectorZx = 0
}
public static void EmitScalarSaturatingUnaryOpSx(AILEmitterCtx Context, Action Emit)
@ -837,7 +837,14 @@ namespace ChocolArm64.Instruction
Emit();
EmitUnarySignedSatQAbsOrNeg(Context, Op.Size);
if (Op.Size <= 2)
{
EmitSatQ(Context, Op.Size, true, true);
}
else /* if (Op.Size == 3) */
{
EmitUnarySignedSatQAbsOrNeg(Context);
}
EmitVectorInsertTmp(Context, Index, Op.Size);
}
@ -853,25 +860,25 @@ namespace ChocolArm64.Instruction
public static void EmitScalarSaturatingBinaryOpSx(AILEmitterCtx Context, SaturatingFlags Flags)
{
EmitSaturatingBinaryOp(Context, SaturatingFlags.ScalarSx | Flags);
EmitSaturatingBinaryOp(Context, () => { }, SaturatingFlags.ScalarSx | Flags);
}
public static void EmitScalarSaturatingBinaryOpZx(AILEmitterCtx Context, SaturatingFlags Flags)
{
EmitSaturatingBinaryOp(Context, SaturatingFlags.ScalarZx | Flags);
EmitSaturatingBinaryOp(Context, () => { }, SaturatingFlags.ScalarZx | Flags);
}
public static void EmitVectorSaturatingBinaryOpSx(AILEmitterCtx Context, SaturatingFlags Flags)
{
EmitSaturatingBinaryOp(Context, SaturatingFlags.VectorSx | Flags);
EmitSaturatingBinaryOp(Context, () => { }, SaturatingFlags.VectorSx | Flags);
}
public static void EmitVectorSaturatingBinaryOpZx(AILEmitterCtx Context, SaturatingFlags Flags)
{
EmitSaturatingBinaryOp(Context, SaturatingFlags.VectorZx | Flags);
EmitSaturatingBinaryOp(Context, () => { }, SaturatingFlags.VectorZx | Flags);
}
public static void EmitSaturatingBinaryOp(AILEmitterCtx Context, SaturatingFlags Flags)
public static void EmitSaturatingBinaryOp(AILEmitterCtx Context, Action Emit, SaturatingFlags Flags)
{
AOpCodeSimd Op = (AOpCodeSimd)Context.CurrOp;
@ -940,6 +947,20 @@ namespace ChocolArm64.Instruction
EmitVectorInsertTmp(Context, Index, Op.Size);
}
}
else
{
for (int Index = 0; Index < Elems; Index++)
{
EmitVectorExtract(Context, Op.Rn, Index, Op.Size, Signed);
EmitVectorExtract(Context, ((AOpCodeSimdReg)Op).Rm, Index, Op.Size, Signed);
Emit();
EmitSatQ(Context, Op.Size, true, Signed);
EmitVectorInsertTmp(Context, Index, Op.Size);
}
}
Context.EmitLdvectmp();
Context.EmitStvec(Op.Rd);
@ -1080,29 +1101,17 @@ namespace ChocolArm64.Instruction
}
}
// TSrc (8bit, 16bit, 32bit, 64bit) == TDst (8bit, 16bit, 32bit, 64bit); signed.
public static void EmitUnarySignedSatQAbsOrNeg(AILEmitterCtx Context, int Size)
// TSrc (64bit) == TDst (64bit); signed.
public static void EmitUnarySignedSatQAbsOrNeg(AILEmitterCtx Context)
{
int ESize = 8 << Size;
if (((AOpCodeSimd)Context.CurrOp).Size < 3)
{
throw new InvalidOperationException();
}
long TMaxValue = (1L << (ESize - 1)) - 1L;
long TMinValue = -(1L << (ESize - 1));
Context.EmitLdarg(ATranslatedSub.StateArgIdx);
AILLabel LblFalse = new AILLabel();
Context.Emit(OpCodes.Dup);
Context.Emit(OpCodes.Neg);
Context.EmitLdc_I8(TMinValue);
Context.Emit(OpCodes.Ceq);
Context.Emit(OpCodes.Brfalse_S, LblFalse);
Context.Emit(OpCodes.Pop);
EmitSetFpsrQCFlag(Context);
Context.EmitLdc_I8(TMaxValue);
Context.MarkLabel(LblFalse);
ASoftFallback.EmitCall(Context, nameof(ASoftFallback.UnarySignedSatQAbsOrNeg));
}
// TSrcs (64bit) == TDst (64bit); signed, unsigned.
@ -1150,22 +1159,6 @@ namespace ChocolArm64.Instruction
: nameof(ASoftFallback.BinaryUnsignedSatQAcc));
}
public static void EmitSetFpsrQCFlag(AILEmitterCtx Context)
{
const int QCFlagBit = 27;
Context.EmitLdarg(ATranslatedSub.StateArgIdx);
Context.EmitLdarg(ATranslatedSub.StateArgIdx);
Context.EmitCallPropGet(typeof(AThreadState), nameof(AThreadState.Fpsr));
Context.EmitLdc_I4(1 << QCFlagBit);
Context.Emit(OpCodes.Or);
Context.EmitCallPropSet(typeof(AThreadState), nameof(AThreadState.Fpsr));
}
public static void EmitScalarSet(AILEmitterCtx Context, int Reg, int Size)
{
EmitVectorZeroAll(Context, Reg);

View file

@ -11,6 +11,107 @@ namespace ChocolArm64.Instruction
Context.EmitCall(typeof(ASoftFallback), MthdName);
}
#region "Saturating"
public static long SignedSrcSignedDstSatQ(long op, int Size, AThreadState State)
{
int ESize = 8 << Size;
long TMaxValue = (1L << (ESize - 1)) - 1L;
long TMinValue = -(1L << (ESize - 1));
if (op > TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else if (op < TMinValue)
{
SetFpsrQCFlag(State);
return TMinValue;
}
else
{
return op;
}
}
public static ulong SignedSrcUnsignedDstSatQ(long op, int Size, AThreadState State)
{
int ESize = 8 << Size;
ulong TMaxValue = (1UL << ESize) - 1UL;
ulong TMinValue = 0UL;
if (op > (long)TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else if (op < (long)TMinValue)
{
SetFpsrQCFlag(State);
return TMinValue;
}
else
{
return (ulong)op;
}
}
public static long UnsignedSrcSignedDstSatQ(ulong op, int Size, AThreadState State)
{
int ESize = 8 << Size;
long TMaxValue = (1L << (ESize - 1)) - 1L;
if (op > (ulong)TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else
{
return (long)op;
}
}
public static ulong UnsignedSrcUnsignedDstSatQ(ulong op, int Size, AThreadState State)
{
int ESize = 8 << Size;
ulong TMaxValue = (1UL << ESize) - 1UL;
if (op > TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else
{
return op;
}
}
public static long UnarySignedSatQAbsOrNeg(long op, AThreadState State)
{
if (op == long.MinValue)
{
SetFpsrQCFlag(State);
return long.MaxValue;
}
else
{
return op;
}
}
public static long BinarySignedSatQAdd(long op1, long op2, AThreadState State)
{
long Add = op1 + op2;
@ -185,99 +286,15 @@ namespace ChocolArm64.Instruction
}
}
public static long SignedSrcSignedDstSatQ(long op, int Size, AThreadState State)
{
int ESize = 8 << Size;
long TMaxValue = (1L << (ESize - 1)) - 1L;
long TMinValue = -(1L << (ESize - 1));
if (op > TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else if (op < TMinValue)
{
SetFpsrQCFlag(State);
return TMinValue;
}
else
{
return op;
}
}
public static ulong SignedSrcUnsignedDstSatQ(long op, int Size, AThreadState State)
{
int ESize = 8 << Size;
ulong TMaxValue = (1UL << ESize) - 1UL;
ulong TMinValue = 0UL;
if (op > (long)TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else if (op < (long)TMinValue)
{
SetFpsrQCFlag(State);
return TMinValue;
}
else
{
return (ulong)op;
}
}
public static long UnsignedSrcSignedDstSatQ(ulong op, int Size, AThreadState State)
{
int ESize = 8 << Size;
long TMaxValue = (1L << (ESize - 1)) - 1L;
if (op > (ulong)TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else
{
return (long)op;
}
}
public static ulong UnsignedSrcUnsignedDstSatQ(ulong op, int Size, AThreadState State)
{
int ESize = 8 << Size;
ulong TMaxValue = (1UL << ESize) - 1UL;
if (op > TMaxValue)
{
SetFpsrQCFlag(State);
return TMaxValue;
}
else
{
return op;
}
}
private static void SetFpsrQCFlag(AThreadState State)
{
const int QCFlagBit = 27;
State.Fpsr |= 1 << QCFlagBit;
}
#endregion
#region "Count"
public static ulong CountLeadingSigns(ulong Value, int Size)
{
Value ^= Value >> 1;
@ -325,7 +342,9 @@ namespace ChocolArm64.Instruction
return (Value >> 4) + (Value & 0x0f);
}
#endregion
#region "Crc32"
private const uint Crc32RevPoly = 0xedb88320;
private const uint Crc32cRevPoly = 0x82f63b78;
@ -384,7 +403,9 @@ namespace ChocolArm64.Instruction
return Crc;
}
#endregion
#region "Reverse"
public static uint ReverseBits8(uint Value)
{
Value = ((Value & 0xaa) >> 1) | ((Value & 0x55) << 1);
@ -453,7 +474,9 @@ namespace ChocolArm64.Instruction
throw new ArgumentException(nameof(Size));
}
#endregion
#region "MultiplyHigh"
public static long SMulHi128(long LHS, long RHS)
{
long Result = (long)UMulHi128((ulong)LHS, (ulong)RHS);
@ -479,5 +502,6 @@ namespace ChocolArm64.Instruction
return LHigh * RHigh + Z0 + (Z1 >> 32);
}
#endregion
}
}

View file

@ -1,9 +1,9 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
<TargetFramework>netcoreapp2.1</TargetFramework>
<RuntimeIdentifiers>win10-x64;osx-x64;linux-x64</RuntimeIdentifiers>
<OutputType>Exe</OutputType>
</PropertyGroup>
</Project>

View file

@ -1,13 +1,13 @@
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<TargetFramework>netcoreapp2.1</TargetFramework>
<RuntimeIdentifiers>win10-x64;osx-x64;linux-x64</RuntimeIdentifiers>
<OutputType>Exe</OutputType>
</PropertyGroup>
<ItemGroup>
<ProjectReference Include="..\Ryujinx.Graphics\Ryujinx.Graphics.csproj" />
</ItemGroup>
<PropertyGroup>
<OutputType>Exe</OutputType>
<TargetFramework>netcoreapp2.1</TargetFramework>
<RuntimeIdentifiers>win10-x64;osx-x64;linux-x64</RuntimeIdentifiers>
</PropertyGroup>
</Project>

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("Alu"), Ignore("Tested: first half of 2018.")]
[Category("Alu"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestAlu : CpuTest
{
#if Alu

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("AluImm"), Ignore("Tested: first half of 2018.")]
[Category("AluImm"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestAluImm : CpuTest
{
#if AluImm

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("AluRs"), Ignore("Tested: first half of 2018.")]
[Category("AluRs"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestAluRs : CpuTest
{
#if AluRs

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("AluRx"), Ignore("Tested: first half of 2018.")]
[Category("AluRx"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestAluRx : CpuTest
{
#if AluRx

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("Bfm"), Ignore("Tested: first half of 2018.")]
[Category("Bfm"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestBfm : CpuTest
{
#if Bfm

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("CcmpImm"), Ignore("Tested: first half of 2018.")]
[Category("CcmpImm"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestCcmpImm : CpuTest
{
#if CcmpImm

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("CcmpReg"), Ignore("Tested: first half of 2018.")]
[Category("CcmpReg"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestCcmpReg : CpuTest
{
#if CcmpReg

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("Csel"), Ignore("Tested: first half of 2018.")]
[Category("Csel"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestCsel : CpuTest
{
#if Csel

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("Mov"), Ignore("Tested: first half of 2018.")]
[Category("Mov"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestMov : CpuTest
{
#if Mov

View file

@ -9,7 +9,7 @@ namespace Ryujinx.Tests.Cpu
using Tester;
using Tester.Types;
[Category("Mul"), Ignore("Tested: first half of 2018.")]
[Category("Mul"), Ignore("Tested: second half of 2018.")]
public sealed class CpuTestMul : CpuTest
{
#if Mul

View file

@ -39,6 +39,21 @@ namespace Ryujinx.Tests.Cpu
0x8000000000000000ul, 0xFFFFFFFFFFFFFFFFul };
}
private static ulong[] _1H1S_()
{
return new ulong[] { 0x0000000000000000ul, 0x0000000000007FFFul,
0x0000000000008000ul, 0x000000000000FFFFul,
0x000000007FFFFFFFul, 0x0000000080000000ul,
0x00000000FFFFFFFFul };
}
private static ulong[] _4H2S_()
{
return new ulong[] { 0x0000000000000000ul, 0x7FFF7FFF7FFF7FFFul,
0x8000800080008000ul, 0x7FFFFFFF7FFFFFFFul,
0x8000000080000000ul, 0xFFFFFFFFFFFFFFFFul };
}
private static ulong[] _4H2S1D_()
{
return new ulong[] { 0x0000000000000000ul, 0x7FFF7FFF7FFF7FFFul,
@ -1837,6 +1852,216 @@ namespace Ryujinx.Tests.Cpu
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQDMULH <V><d>, <V><n>, <V><m>")]
public void Sqdmulh_S_H_S([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,
[Values(2u, 0u)] uint Rm,
[ValueSource("_1H1S_")] [Random(RndCnt)] ulong Z,
[ValueSource("_1H1S_")] [Random(RndCnt)] ulong A,
[ValueSource("_1H1S_")] [Random(RndCnt)] ulong B,
[Values(0b01u, 0b10u)] uint size) // <H, S>
{
uint Opcode = 0x5E20B400; // SQDMULH B0, B0, B0 (RESERVED)
Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
Opcode |= ((size & 3) << 22);
Bits Op = new Bits(Opcode);
int Fpsr = (int)TestContext.CurrentContext.Random.NextUInt();
Vector128<float> V0 = MakeVectorE0E1(Z, Z);
Vector128<float> V1 = MakeVectorE0(A);
Vector128<float> V2 = MakeVectorE0(B);
AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2, Fpsr: Fpsr);
AArch64.Vpart(0, 0, new Bits(Z)); AArch64.Vpart(0, 1, new Bits(Z));
AArch64.V(1, new Bits(A));
AArch64.V(2, new Bits(B));
Shared.FPSR = new Bits((uint)Fpsr);
SimdFp.Sqdmulh_S(Op[23, 22], Op[20, 16], Op[9, 5], Op[4, 0]);
Assert.Multiple(() =>
{
Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64()));
Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64()));
});
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQDMULH <Vd>.<T>, <Vn>.<T>, <Vm>.<T>")]
public void Sqdmulh_V_4H_2S([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,
[Values(2u, 0u)] uint Rm,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong Z,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong A,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong B,
[Values(0b01u, 0b10u)] uint size) // <4H, 2S>
{
uint Opcode = 0x0E20B400; // SQDMULH V0.8B, V0.8B, V0.8B (RESERVED)
Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
Opcode |= ((size & 3) << 22);
Bits Op = new Bits(Opcode);
int Fpsr = (int)TestContext.CurrentContext.Random.NextUInt();
Vector128<float> V0 = MakeVectorE0E1(Z, Z);
Vector128<float> V1 = MakeVectorE0(A);
Vector128<float> V2 = MakeVectorE0(B);
AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2, Fpsr: Fpsr);
AArch64.Vpart(0, 0, new Bits(Z)); AArch64.Vpart(0, 1, new Bits(Z));
AArch64.V(1, new Bits(A));
AArch64.V(2, new Bits(B));
Shared.FPSR = new Bits((uint)Fpsr);
SimdFp.Sqdmulh_V(Op[30], Op[23, 22], Op[20, 16], Op[9, 5], Op[4, 0]);
Assert.Multiple(() =>
{
Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64()));
Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64()));
});
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQDMULH <Vd>.<T>, <Vn>.<T>, <Vm>.<T>")]
public void Sqdmulh_V_8H_4S([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,
[Values(2u, 0u)] uint Rm,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong Z,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong A,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong B,
[Values(0b01u, 0b10u)] uint size) // <8H, 4S>
{
uint Opcode = 0x4E20B400; // SQDMULH V0.16B, V0.16B, V0.16B (RESERVED)
Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
Opcode |= ((size & 3) << 22);
Bits Op = new Bits(Opcode);
int Fpsr = (int)TestContext.CurrentContext.Random.NextUInt();
Vector128<float> V0 = MakeVectorE0E1(Z, Z);
Vector128<float> V1 = MakeVectorE0E1(A, A);
Vector128<float> V2 = MakeVectorE0E1(B, B);
AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2, Fpsr: Fpsr);
AArch64.Vpart(0, 0, new Bits(Z)); AArch64.Vpart(0, 1, new Bits(Z));
AArch64.Vpart(1, 0, new Bits(A)); AArch64.Vpart(1, 1, new Bits(A));
AArch64.Vpart(2, 0, new Bits(B)); AArch64.Vpart(2, 1, new Bits(B));
Shared.FPSR = new Bits((uint)Fpsr);
SimdFp.Sqdmulh_V(Op[30], Op[23, 22], Op[20, 16], Op[9, 5], Op[4, 0]);
Assert.Multiple(() =>
{
Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64()));
Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64()));
});
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQRDMULH <V><d>, <V><n>, <V><m>")]
public void Sqrdmulh_S_H_S([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,
[Values(2u, 0u)] uint Rm,
[ValueSource("_1H1S_")] [Random(RndCnt)] ulong Z,
[ValueSource("_1H1S_")] [Random(RndCnt)] ulong A,
[ValueSource("_1H1S_")] [Random(RndCnt)] ulong B,
[Values(0b01u, 0b10u)] uint size) // <H, S>
{
uint Opcode = 0x7E20B400; // SQRDMULH B0, B0, B0 (RESERVED)
Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
Opcode |= ((size & 3) << 22);
Bits Op = new Bits(Opcode);
int Fpsr = (int)TestContext.CurrentContext.Random.NextUInt();
Vector128<float> V0 = MakeVectorE0E1(Z, Z);
Vector128<float> V1 = MakeVectorE0(A);
Vector128<float> V2 = MakeVectorE0(B);
AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2, Fpsr: Fpsr);
AArch64.Vpart(0, 0, new Bits(Z)); AArch64.Vpart(0, 1, new Bits(Z));
AArch64.V(1, new Bits(A));
AArch64.V(2, new Bits(B));
Shared.FPSR = new Bits((uint)Fpsr);
SimdFp.Sqrdmulh_S(Op[23, 22], Op[20, 16], Op[9, 5], Op[4, 0]);
Assert.Multiple(() =>
{
Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64()));
Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64()));
});
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQRDMULH <Vd>.<T>, <Vn>.<T>, <Vm>.<T>")]
public void Sqrdmulh_V_4H_2S([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,
[Values(2u, 0u)] uint Rm,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong Z,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong A,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong B,
[Values(0b01u, 0b10u)] uint size) // <4H, 2S>
{
uint Opcode = 0x2E20B400; // SQRDMULH V0.8B, V0.8B, V0.8B (RESERVED)
Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
Opcode |= ((size & 3) << 22);
Bits Op = new Bits(Opcode);
int Fpsr = (int)TestContext.CurrentContext.Random.NextUInt();
Vector128<float> V0 = MakeVectorE0E1(Z, Z);
Vector128<float> V1 = MakeVectorE0(A);
Vector128<float> V2 = MakeVectorE0(B);
AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2, Fpsr: Fpsr);
AArch64.Vpart(0, 0, new Bits(Z)); AArch64.Vpart(0, 1, new Bits(Z));
AArch64.V(1, new Bits(A));
AArch64.V(2, new Bits(B));
Shared.FPSR = new Bits((uint)Fpsr);
SimdFp.Sqrdmulh_V(Op[30], Op[23, 22], Op[20, 16], Op[9, 5], Op[4, 0]);
Assert.Multiple(() =>
{
Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64()));
Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64()));
});
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQRDMULH <Vd>.<T>, <Vn>.<T>, <Vm>.<T>")]
public void Sqrdmulh_V_8H_4S([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,
[Values(2u, 0u)] uint Rm,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong Z,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong A,
[ValueSource("_4H2S_")] [Random(RndCnt)] ulong B,
[Values(0b01u, 0b10u)] uint size) // <8H, 4S>
{
uint Opcode = 0x6E20B400; // SQRDMULH V0.16B, V0.16B, V0.16B (RESERVED)
Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
Opcode |= ((size & 3) << 22);
Bits Op = new Bits(Opcode);
int Fpsr = (int)TestContext.CurrentContext.Random.NextUInt();
Vector128<float> V0 = MakeVectorE0E1(Z, Z);
Vector128<float> V1 = MakeVectorE0E1(A, A);
Vector128<float> V2 = MakeVectorE0E1(B, B);
AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2, Fpsr: Fpsr);
AArch64.Vpart(0, 0, new Bits(Z)); AArch64.Vpart(0, 1, new Bits(Z));
AArch64.Vpart(1, 0, new Bits(A)); AArch64.Vpart(1, 1, new Bits(A));
AArch64.Vpart(2, 0, new Bits(B)); AArch64.Vpart(2, 1, new Bits(B));
Shared.FPSR = new Bits((uint)Fpsr);
SimdFp.Sqrdmulh_V(Op[30], Op[23, 22], Op[20, 16], Op[9, 5], Op[4, 0]);
Assert.Multiple(() =>
{
Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64()));
Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64()));
});
Assert.That(ThreadState.Fpsr, Is.EqualTo((int)Shared.FPSR.ToUInt32()));
}
[Test, Pairwise, Description("SQSUB <V><d>, <V><n>, <V><m>")]
public void Sqsub_S_B_H_S_D([Values(0u)] uint Rd,
[Values(1u, 0u)] uint Rn,

View file

@ -5075,6 +5075,210 @@ namespace Ryujinx.Tests.Cpu.Tester
V(d, result);
}
// sqdmulh_advsimd_vec.html#SQDMULH_asisdsame_only
public static void Sqdmulh_S(Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = false;
/* Decode Scalar */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' || size == '00' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = esize;
int elements = 1;
bool rounding = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand1 = V(datasize, n);
Bits operand2 = V(datasize, m);
BigInteger round_const = (rounding ? (BigInteger)1 << (esize - 1) : 0);
BigInteger element1;
BigInteger element2;
BigInteger product;
bool sat;
for (int e = 0; e <= elements - 1; e++)
{
element1 = SInt(Elem(operand1, e, esize));
element2 = SInt(Elem(operand2, e, esize));
product = (2 * element1 * element2) + round_const;
(Bits _result, bool _sat) = SignedSatQ(product >> esize, esize);
Elem(result, e, esize, _result);
sat = _sat;
if (sat)
{
/* FPSR.QC = '1'; */
FPSR[27] = true; // TODO: Add named fields.
}
}
V(d, result);
}
// sqdmulh_advsimd_vec.html#SQDMULH_asimdsame_only
public static void Sqdmulh_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = false;
/* Decode Vector */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' || size == '00' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = (Q ? 128 : 64);
int elements = datasize / esize;
bool rounding = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand1 = V(datasize, n);
Bits operand2 = V(datasize, m);
BigInteger round_const = (rounding ? (BigInteger)1 << (esize - 1) : 0);
BigInteger element1;
BigInteger element2;
BigInteger product;
bool sat;
for (int e = 0; e <= elements - 1; e++)
{
element1 = SInt(Elem(operand1, e, esize));
element2 = SInt(Elem(operand2, e, esize));
product = (2 * element1 * element2) + round_const;
(Bits _result, bool _sat) = SignedSatQ(product >> esize, esize);
Elem(result, e, esize, _result);
sat = _sat;
if (sat)
{
/* FPSR.QC = '1'; */
FPSR[27] = true; // TODO: Add named fields.
}
}
V(d, result);
}
// sqrdmulh_advsimd_vec.html#SQRDMULH_asisdsame_only
public static void Sqrdmulh_S(Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = true;
/* Decode Scalar */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' || size == '00' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = esize;
int elements = 1;
bool rounding = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand1 = V(datasize, n);
Bits operand2 = V(datasize, m);
BigInteger round_const = (rounding ? (BigInteger)1 << (esize - 1) : 0);
BigInteger element1;
BigInteger element2;
BigInteger product;
bool sat;
for (int e = 0; e <= elements - 1; e++)
{
element1 = SInt(Elem(operand1, e, esize));
element2 = SInt(Elem(operand2, e, esize));
product = (2 * element1 * element2) + round_const;
(Bits _result, bool _sat) = SignedSatQ(product >> esize, esize);
Elem(result, e, esize, _result);
sat = _sat;
if (sat)
{
/* FPSR.QC = '1'; */
FPSR[27] = true; // TODO: Add named fields.
}
}
V(d, result);
}
// sqrdmulh_advsimd_vec.html#SQRDMULH_asimdsame_only
public static void Sqrdmulh_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
{
const bool U = true;
/* Decode Vector */
int d = (int)UInt(Rd);
int n = (int)UInt(Rn);
int m = (int)UInt(Rm);
/* if size == '11' || size == '00' then ReservedValue(); */
int esize = 8 << (int)UInt(size);
int datasize = (Q ? 128 : 64);
int elements = datasize / esize;
bool rounding = (U == true);
/* Operation */
/* CheckFPAdvSIMDEnabled64(); */
Bits result = new Bits(datasize);
Bits operand1 = V(datasize, n);
Bits operand2 = V(datasize, m);
BigInteger round_const = (rounding ? (BigInteger)1 << (esize - 1) : 0);
BigInteger element1;
BigInteger element2;
BigInteger product;
bool sat;
for (int e = 0; e <= elements - 1; e++)
{
element1 = SInt(Elem(operand1, e, esize));
element2 = SInt(Elem(operand2, e, esize));
product = (2 * element1 * element2) + round_const;
(Bits _result, bool _sat) = SignedSatQ(product >> esize, esize);
Elem(result, e, esize, _result);
sat = _sat;
if (sat)
{
/* FPSR.QC = '1'; */
FPSR[27] = true; // TODO: Add named fields.
}
}
V(d, result);
}
// sqsub_advsimd.html#SQSUB_asisdsame_only
public static void Sqsub_S(Bits size, Bits Rm, Bits Rn, Bits Rd)
{

View file

@ -1,20 +1,25 @@
<Project Sdk="Microsoft.NET.Sdk">
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<TargetFramework>netcoreapp2.1</TargetFramework>
<RuntimeIdentifiers>win10-x64;osx-x64;linux-x64</RuntimeIdentifiers>
<OutputType>Exe</OutputType>
<IsPackable>false</IsPackable>
</PropertyGroup>
<PropertyGroup>
<GenerateAssemblyInfo>false</GenerateAssemblyInfo>
</PropertyGroup>
<ItemGroup>
<PackageReference Include="Microsoft.NET.Test.Sdk" Version="15.7.0" />
<PackageReference Include="Microsoft.NET.Test.Sdk" Version="15.8.0" />
<PackageReference Include="NUnit" Version="3.10.1" />
<PackageReference Include="NUnit3TestAdapter" Version="3.10.0" />
<PackageReference Include="System.Runtime.Intrinsics.Experimental" Version="4.5.0-rc1" />
</ItemGroup>
<ItemGroup>
<ProjectReference Include="..\ChocolArm64\ChocolArm64.csproj" />
</ItemGroup>
</Project>

View file

@ -1,23 +1,28 @@
<Project Sdk="Microsoft.NET.Sdk">
<Project Sdk="Microsoft.NET.Sdk">
<PropertyGroup>
<OutputType>Exe</OutputType>
<TargetFramework>netcoreapp2.1</TargetFramework>
<AllowUnsafeBlocks>true</AllowUnsafeBlocks>
<RuntimeIdentifiers>win10-x64;osx-x64;linux-x64</RuntimeIdentifiers>
<OutputType>Exe</OutputType>
<AllowUnsafeBlocks>true</AllowUnsafeBlocks>
</PropertyGroup>
<ItemGroup>
<PackageReference Include="OpenTK.NetStandard" Version="1.0.4" />
<PackageReference Include="System.Runtime.CompilerServices.Unsafe" Version="4.4.0" />
</ItemGroup>
<ItemGroup>
<ProjectReference Include="..\ChocolArm64\ChocolArm64.csproj" />
<ProjectReference Include="..\Ryujinx.Audio\Ryujinx.Audio.csproj" />
<ProjectReference Include="..\Ryujinx.HLE\Ryujinx.HLE.csproj" />
<ProjectReference Include="..\Ryujinx.Graphics\Ryujinx.Graphics.csproj" />
<ProjectReference Include="..\Ryujinx.HLE\Ryujinx.HLE.csproj" />
</ItemGroup>
<ItemGroup>
<None Update="Ryujinx.conf">
<CopyToOutputDirectory>PreserveNewest</CopyToOutputDirectory>
</None>
</ItemGroup>
</Project>