Implement Ssubw_V and Usubw_V instructions. (#287)

* Update AOpCodeTable.cs

* Update AInstEmitSimdHelper.cs

* Update AInstEmitSimdArithmetic.cs

* Update AInstEmitSimdMove.cs

* Update AInstEmitSimdCmp.cs

* Update Instructions.cs

* Update CpuTestSimd.cs

* Update CpuTestSimdReg.cs

ChocolArm64/AOpCodeTable.cs

@@ -387,6 +387,7 @@ 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<<1xxxxx001100xxxxxxxxxx", AInstEmit.Ssubw_V,       typeof(AOpCodeSimdReg));
             SetA64("0x00110000000000xxxxxxxxxxxxxxxx", AInstEmit.St__Vms,       typeof(AOpCodeSimdMemMs));
             SetA64("0x001100100xxxxxxxxxxxxxxxxxxxxx", AInstEmit.St__Vms,       typeof(AOpCodeSimdMemMs));
             SetA64("0x00110100x00000xxxxxxxxxxxxxxxx", AInstEmit.St__Vss,       typeof(AOpCodeSimdMemSs));
@@ -430,6 +431,7 @@ namespace ChocolArm64
             SetA64("0110111101xxxxxx000001xxxxxxxxxx", AInstEmit.Ushr_V,        typeof(AOpCodeSimdShImm));
             SetA64("0x10111100>>>xxx000101xxxxxxxxxx", AInstEmit.Usra_V,        typeof(AOpCodeSimdShImm));
             SetA64("0110111101xxxxxx000101xxxxxxxxxx", AInstEmit.Usra_V,        typeof(AOpCodeSimdShImm));
+            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));
             SetA64("0x001110<<100001001010xxxxxxxxxx", AInstEmit.Xtn_V,         typeof(AOpCodeSimd));

ChocolArm64/Instruction/AInstEmitSimdArithmetic.cs

@@ -1072,6 +1072,11 @@ namespace ChocolArm64.Instruction
             EmitVectorSaturatingNarrowOpSxZx(Context, () => { });
         }
 
+        public static void Ssubw_V(AILEmitterCtx Context)
+        {
+            EmitVectorWidenRmBinaryOpSx(Context, () => Context.Emit(OpCodes.Sub));
+        }
+
         public static void Sub_S(AILEmitterCtx Context)
         {
             EmitScalarBinaryOpZx(Context, () => Context.Emit(OpCodes.Sub));
@@ -1225,5 +1230,10 @@ namespace ChocolArm64.Instruction
         {
             EmitVectorSaturatingNarrowOpZxZx(Context, () => { });
         }
+
+        public static void Usubw_V(AILEmitterCtx Context)
+        {
+            EmitVectorWidenRmBinaryOpZx(Context, () => Context.Emit(OpCodes.Sub));
+        }
     }
 }

ChocolArm64/Instruction/AInstEmitSimdCmp.cs

@@ -364,7 +364,7 @@ namespace ChocolArm64.Instruction
             AOpCodeSimd Op = (AOpCodeSimd)Context.CurrOp;
 
             int Bytes = Op.GetBitsCount() >> 3;
-            int Elems = (!Scalar ? Bytes >> Op.Size : 1);
+            int Elems = !Scalar ? Bytes >> Op.Size : 1;
 
             ulong SzMask = ulong.MaxValue >> (64 - (8 << Op.Size));
 
@@ -408,7 +408,7 @@ namespace ChocolArm64.Instruction
             AOpCodeSimdReg Op = (AOpCodeSimdReg)Context.CurrOp;
 
             int Bytes = Op.GetBitsCount() >> 3;
-            int Elems = (!Scalar ? Bytes >> Op.Size : 1);
+            int Elems = !Scalar ? Bytes >> Op.Size : 1;
 
             ulong SzMask = ulong.MaxValue >> (64 - (8 << Op.Size));
 

ChocolArm64/Instruction/AInstEmitSimdHelper.cs

@@ -419,20 +419,25 @@ namespace ChocolArm64.Instruction
             int SizeF = Op.Size & 1;
 
             int Bytes = Op.GetBitsCount() >> 3;
+            int Elems = Bytes >> SizeF + 2;
 
-            for (int Index = 0; Index < (Bytes >> SizeF + 2); Index++)
+            bool Rd = (Opers & OperFlags.Rd) != 0;
+            bool Rn = (Opers & OperFlags.Rn) != 0;
+            bool Rm = (Opers & OperFlags.Rm) != 0;
+
+            for (int Index = 0; Index < Elems; Index++)
             {
-                if (Opers.HasFlag(OperFlags.Rd))
+                if (Rd)
                 {
                     EmitVectorExtractF(Context, Op.Rd, Index, SizeF);
                 }
 
-                if (Opers.HasFlag(OperFlags.Rn))
+                if (Rn)
                 {
                     EmitVectorExtractF(Context, Op.Rn, Index, SizeF);
                 }
 
-                if (Opers.HasFlag(OperFlags.Rm))
+                if (Rm)
                 {
                     EmitVectorExtractF(Context, ((AOpCodeSimdReg)Op).Rm, Index, SizeF);
                 }
@@ -469,8 +474,9 @@ namespace ChocolArm64.Instruction
             int SizeF = Op.Size & 1;
 
             int Bytes = Op.GetBitsCount() >> 3;
+            int Elems = Bytes >> SizeF + 2;
 
-            for (int Index = 0; Index < (Bytes >> SizeF + 2); Index++)
+            for (int Index = 0; Index < Elems; Index++)
             {
                 if (Ternary)
                 {
@@ -531,19 +537,23 @@ namespace ChocolArm64.Instruction
             int Bytes = Op.GetBitsCount() >> 3;
             int Elems = Bytes >> Op.Size;
 
+            bool Rd = (Opers & OperFlags.Rd) != 0;
+            bool Rn = (Opers & OperFlags.Rn) != 0;
+            bool Rm = (Opers & OperFlags.Rm) != 0;
+
             for (int Index = 0; Index < Elems; Index++)
             {
-                if (Opers.HasFlag(OperFlags.Rd))
+                if (Rd)
                 {
                     EmitVectorExtract(Context, Op.Rd, Index, Op.Size, Signed);
                 }
 
-                if (Opers.HasFlag(OperFlags.Rn))
+                if (Rn)
                 {
                     EmitVectorExtract(Context, Op.Rn, Index, Op.Size, Signed);
                 }
 
-                if (Opers.HasFlag(OperFlags.Rm))
+                if (Rm)
                 {
                     EmitVectorExtract(Context, ((AOpCodeSimdReg)Op).Rm, Index, Op.Size, Signed);
                 }
@@ -662,9 +672,6 @@ namespace ChocolArm64.Instruction
         {
             AOpCodeSimdReg Op = (AOpCodeSimdReg)Context.CurrOp;
 
-            Context.EmitLdvec(Op.Rd);
-            Context.EmitStvectmp();
-
             int Elems = 8 >> Op.Size;
 
             int Part = Op.RegisterSize == ARegisterSize.SIMD128 ? Elems : 0;
@@ -707,9 +714,6 @@ namespace ChocolArm64.Instruction
         {
             AOpCodeSimdReg Op = (AOpCodeSimdReg)Context.CurrOp;
 
-            Context.EmitLdvec(Op.Rd);
-            Context.EmitStvectmp();
-
             int Elems = 8 >> Op.Size;
 
             int Part = Op.RegisterSize == ARegisterSize.SIMD128 ? Elems : 0;
@@ -747,20 +751,24 @@ namespace ChocolArm64.Instruction
         {
             AOpCodeSimdReg Op = (AOpCodeSimdReg)Context.CurrOp;
 
-            int Bytes = Op.GetBitsCount() >> 3;
+            int Words = Op.GetBitsCount() >> 4;
+            int Pairs = Words >> Op.Size;
 
-            int Elems = Bytes >> Op.Size;
+            for (int Index = 0; Index < Pairs; Index++)
-            int Half  = Elems >> 1;
-
-            for (int Index = 0; Index < Elems; Index++)
             {
-                int Elem = (Index & (Half - 1)) << 1;
+                int Idx = Index << 1;
 
-                EmitVectorExtract(Context, Index < Half ? Op.Rn : Op.Rm, Elem + 0, Op.Size, Signed);
+                EmitVectorExtract(Context, Op.Rn, Idx,     Op.Size, Signed);
-                EmitVectorExtract(Context, Index < Half ? Op.Rn : Op.Rm, Elem + 1, Op.Size, Signed);
+                EmitVectorExtract(Context, Op.Rn, Idx + 1, Op.Size, Signed);
 
                 Emit();
 
+                EmitVectorExtract(Context, Op.Rm, Idx,     Op.Size, Signed);
+                EmitVectorExtract(Context, Op.Rm, Idx + 1, Op.Size, Signed);
+
+                Emit();
+
+                EmitVectorInsertTmp(Context, Pairs + Index, Op.Size);
                 EmitVectorInsertTmp(Context, Index,         Op.Size);
             }
 
@@ -818,7 +826,7 @@ namespace ChocolArm64.Instruction
 
             int Part = !Scalar && (Op.RegisterSize == ARegisterSize.SIMD128) ? Elems : 0;
 
-            long TMaxValue = SignedDst ? (1 << (ESize - 1)) - 1 : (1L << ESize) - 1L;
+            long TMaxValue = SignedDst ? (1 << (ESize - 1)) - 1 : (long)(~0UL >> (64 - ESize));
             long TMinValue = SignedDst ? -((1 << (ESize - 1))) : 0;
 
             Context.EmitLdc_I8(0L);
@@ -871,7 +879,7 @@ namespace ChocolArm64.Instruction
 
                 if (Scalar)
                 {
-                    EmitVectorZeroLower(Context, Op.Rd);
+                    EmitVectorZeroLowerTmp(Context);
                 }
 
                 EmitVectorInsertTmp(Context, Part + Index, Op.Size);
@@ -963,6 +971,11 @@ namespace ChocolArm64.Instruction
             EmitVectorInsert(Context, Rd, 0, 3, 0);
         }
 
+        public static void EmitVectorZeroLowerTmp(AILEmitterCtx Context)
+        {
+            EmitVectorInsertTmp(Context, 0, 3, 0);
+        }
+
         public static void EmitVectorZeroUpper(AILEmitterCtx Context, int Rd)
         {
             EmitVectorInsert(Context, Rd, 1, 3, 0);
@@ -1008,6 +1021,20 @@ namespace ChocolArm64.Instruction
             Context.EmitStvec(Reg);
         }
 
+        public static void EmitVectorInsertTmp(AILEmitterCtx Context, int Index, int Size, long Value)
+        {
+            ThrowIfInvalid(Index, Size);
+
+            Context.EmitLdc_I8(Value);
+            Context.EmitLdvectmp();
+            Context.EmitLdc_I4(Index);
+            Context.EmitLdc_I4(Size);
+
+            AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertInt));
+
+            Context.EmitStvectmp();
+        }
+
         public static void EmitVectorInsertF(AILEmitterCtx Context, int Reg, int Index, int Size)
         {
             ThrowIfInvalidF(Index, Size);

ChocolArm64/Instruction/AInstEmitSimdMove.cs

@@ -295,13 +295,22 @@ namespace ChocolArm64.Instruction
 
             int Part = Op.RegisterSize == ARegisterSize.SIMD128 ? Elems : 0;
 
+            if (Part != 0)
+            {
+                Context.EmitLdvec(Op.Rd);
+                Context.EmitStvectmp();
+            }
+
             for (int Index = 0; Index < Elems; Index++)
             {
                 EmitVectorExtractZx(Context, Op.Rn, Index, Op.Size + 1);
 
-                EmitVectorInsert(Context, Op.Rd, Part + Index, Op.Size);
+                EmitVectorInsertTmp(Context, Part + Index, Op.Size);
             }
 
+            Context.EmitLdvectmp();
+            Context.EmitStvec(Op.Rd);
+
             if (Part == 0)
             {
                 EmitVectorZeroUpper(Context, Op.Rd);
@@ -342,7 +351,7 @@ namespace ChocolArm64.Instruction
                 EmitVectorExtractZx(Context, Op.Rm, Idx + Part, Op.Size);
 
                 EmitVectorInsertTmp(Context, Idx + 1, Op.Size);
-                EmitVectorInsertTmp(Context, Idx    , Op.Size);
+                EmitVectorInsertTmp(Context, Idx,     Op.Size);
             }
 
             Context.EmitLdvectmp();
@@ -398,7 +407,7 @@ namespace ChocolArm64.Instruction
                 EmitVectorExtractZx(Context, Op.Rm, Base + Index, Op.Size);
 
                 EmitVectorInsertTmp(Context, Idx + 1, Op.Size);
-                EmitVectorInsertTmp(Context, Idx    , Op.Size);
+                EmitVectorInsertTmp(Context, Idx,     Op.Size);
             }
 
             Context.EmitLdvectmp();

Ryujinx.Tests/Cpu/CpuTestSimd.cs

@@ -1377,6 +1377,60 @@ namespace Ryujinx.Tests.Cpu
             });
             Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27]));
         }
+
+        [Test, Description("XTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
+        public void Xtn_V_8H8B_4S4H_2D2S([Values(0u)] uint Rd,
+                                         [Values(1u, 0u)] uint Rn,
+                                         [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                         [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                         [Values(0b00u, 0b01u, 0b10u)] uint size) // <8H8B, 4S4H, 2D2S>
+        {
+            uint Opcode = 0x0E212800; // XTN V0.8B, V0.8H
+            Opcode |= ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1);
+
+            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));
+            SimdFp.Xtn_V(Op[30], Op[23, 22], 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()));
+            });
+        }
+
+        [Test, Description("XTN{2} <Vd>.<Tb>, <Vn>.<Ta>")]
+        public void Xtn_V_8H16B_4S8H_2D4S([Values(0u)] uint Rd,
+                                          [Values(1u, 0u)] uint Rn,
+                                          [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                          [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                          [Values(0b00u, 0b01u, 0b10u)] uint size) // <8H16B, 4S8H, 2D4S>
+        {
+            uint Opcode = 0x4E212800; // XTN2 V0.16B, V0.8H
+            Opcode |= ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1);
+
+            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));
+            SimdFp.Xtn_V(Op[30], Op[23, 22], 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()));
+            });
+        }
 #endif
     }
 }

Ryujinx.Tests/Cpu/CpuTestSimdReg.cs

@@ -1659,6 +1659,130 @@ namespace Ryujinx.Tests.Cpu
             });
         }
 
+        [Test, Pairwise, Description("SADDW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Saddw_V_8B8H8H_4H4S4S_2S2D2D([Values(0u)] uint Rd,
+                                                 [Values(1u, 0u)] uint Rn,
+                                                 [Values(2u, 0u)] uint Rm,
+                                                 [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                 [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                 [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                 [Values(0b00u, 0b01u, 0b10u)] uint size) // <8B8H8H, 4H4S4S, 2S2D2D>
+        {
+            uint Opcode = 0x0E201000; // SADDW V0.8H, V0.8H, V0.8B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE0(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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));
+            SimdFp.Saddw_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()));
+            });
+        }
+
+        [Test, Pairwise, Description("SADDW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Saddw_V_16B8H8H_8H4S4S_4S2D2D([Values(0u)] uint Rd,
+                                                  [Values(1u, 0u)] uint Rn,
+                                                  [Values(2u, 0u)] uint Rm,
+                                                  [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                  [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                  [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                  [Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H8H, 8H4S4S, 4S2D2D>
+        {
+            uint Opcode = 0x4E201000; // SADDW2 V0.8H, V0.8H, V0.16B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE1(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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, 1, new Bits(B));
+            SimdFp.Saddw_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()));
+            });
+        }
+
+        [Test, Pairwise, Description("SSUBW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Ssubw_V_8B8H8H_4H4S4S_2S2D2D([Values(0u)] uint Rd,
+                                                 [Values(1u, 0u)] uint Rn,
+                                                 [Values(2u, 0u)] uint Rm,
+                                                 [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                 [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                 [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                 [Values(0b00u, 0b01u, 0b10u)] uint size) // <8B8H8H, 4H4S4S, 2S2D2D>
+        {
+            uint Opcode = 0x0E203000; // SSUBW V0.8H, V0.8H, V0.8B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE0(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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));
+            SimdFp.Ssubw_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()));
+            });
+        }
+
+        [Test, Pairwise, Description("SSUBW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Ssubw_V_16B8H8H_8H4S4S_4S2D2D([Values(0u)] uint Rd,
+                                                  [Values(1u, 0u)] uint Rn,
+                                                  [Values(2u, 0u)] uint Rm,
+                                                  [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                  [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                  [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                  [Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H8H, 8H4S4S, 4S2D2D>
+        {
+            uint Opcode = 0x4E203000; // SSUBW2 V0.8H, V0.8H, V0.16B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE1(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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, 1, new Bits(B));
+            SimdFp.Ssubw_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()));
+            });
+        }
+
         [Test, Pairwise, Description("SUB <V><d>, <V><n>, <V><m>")]
         public void Sub_S_D([Values(0u)] uint Rd,
                             [Values(1u, 0u)] uint Rn,
@@ -2184,6 +2308,130 @@ namespace Ryujinx.Tests.Cpu
             });
         }
 
+        [Test, Pairwise, Description("UADDW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Uaddw_V_8B8H8H_4H4S4S_2S2D2D([Values(0u)] uint Rd,
+                                                 [Values(1u, 0u)] uint Rn,
+                                                 [Values(2u, 0u)] uint Rm,
+                                                 [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                 [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                 [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                 [Values(0b00u, 0b01u, 0b10u)] uint size) // <8B8H8H, 4H4S4S, 2S2D2D>
+        {
+            uint Opcode = 0x2E201000; // UADDW V0.8H, V0.8H, V0.8B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE0(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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));
+            SimdFp.Uaddw_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()));
+            });
+        }
+
+        [Test, Pairwise, Description("UADDW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Uaddw_V_16B8H8H_8H4S4S_4S2D2D([Values(0u)] uint Rd,
+                                                  [Values(1u, 0u)] uint Rn,
+                                                  [Values(2u, 0u)] uint Rm,
+                                                  [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                  [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                  [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                  [Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H8H, 8H4S4S, 4S2D2D>
+        {
+            uint Opcode = 0x6E201000; // UADDW2 V0.8H, V0.8H, V0.16B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE1(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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, 1, new Bits(B));
+            SimdFp.Uaddw_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()));
+            });
+        }
+
+        [Test, Pairwise, Description("USUBW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Usubw_V_8B8H8H_4H4S4S_2S2D2D([Values(0u)] uint Rd,
+                                                 [Values(1u, 0u)] uint Rn,
+                                                 [Values(2u, 0u)] uint Rm,
+                                                 [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                 [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                 [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                 [Values(0b00u, 0b01u, 0b10u)] uint size) // <8B8H8H, 4H4S4S, 2S2D2D>
+        {
+            uint Opcode = 0x2E203000; // USUBW V0.8H, V0.8H, V0.8B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE0(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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));
+            SimdFp.Usubw_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()));
+            });
+        }
+
+        [Test, Pairwise, Description("USUBW{2} <Vd>.<Ta>, <Vn>.<Ta>, <Vm>.<Tb>")]
+        public void Usubw_V_16B8H8H_8H4S4S_4S2D2D([Values(0u)] uint Rd,
+                                                  [Values(1u, 0u)] uint Rn,
+                                                  [Values(2u, 0u)] uint Rm,
+                                                  [ValueSource("_8B4H2S1D_")] [Random(RndCnt)] ulong Z,
+                                                  [ValueSource("_4H2S1D_")] [Random(RndCnt)] ulong A,
+                                                  [ValueSource("_8B4H2S_")] [Random(RndCnt)] ulong B,
+                                                  [Values(0b00u, 0b01u, 0b10u)] uint size) // <16B8H8H, 8H4S4S, 4S2D2D>
+        {
+            uint Opcode = 0x6E203000; // USUBW2 V0.8H, V0.8H, V0.16B
+            Opcode |= ((Rm & 31) << 16) | ((Rn & 31) << 5) | ((Rd & 31) << 0);
+            Opcode |= ((size & 3) << 22);
+            Bits Op = new Bits(Opcode);
+
+            Vector128<float> V0 = MakeVectorE0E1(Z, Z);
+            Vector128<float> V1 = MakeVectorE0E1(A, A);
+            Vector128<float> V2 = MakeVectorE1(B);
+            AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2);
+
+            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, 1, new Bits(B));
+            SimdFp.Usubw_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()));
+            });
+        }
+
         [Test, Pairwise, Description("UZP1 <Vd>.<T>, <Vn>.<T>, <Vm>.<T>")]
         public void Uzp1_V_8B_4H_2S([Values(0u)] uint Rd,
                                     [Values(1u, 0u)] uint Rn,

Ryujinx.Tests/Cpu/Tester/Instructions.cs

@@ -3315,6 +3315,37 @@ namespace Ryujinx.Tests.Cpu.Tester
 
             Vpart(d, part, result);
         }
+
+        // xtn_advsimd.html
+        public static void Xtn_V(bool Q, Bits size, Bits Rn, Bits Rd)
+        {
+            /* Decode Vector */
+            int d = (int)UInt(Rd);
+            int n = (int)UInt(Rn);
+
+            /* if size == '11' then ReservedValue(); */
+
+            int esize = 8 << (int)UInt(size);
+            int datasize = 64;
+            int part = (int)UInt(Q);
+            int elements = datasize / esize;
+
+            /* Operation */
+            /* CheckFPAdvSIMDEnabled64(); */
+
+            Bits result = new Bits(datasize);
+            Bits operand = V(2 * datasize, n);
+            Bits element;
+
+            for (int e = 0; e <= elements - 1; e++)
+            {
+                element = Elem(operand, e, 2 * esize);
+
+                Elem(result, e, esize, element[esize - 1, 0]);
+            }
+
+            Vpart(d, part, result);
+        }
 #endregion
 
 #region "SimdReg"
@@ -4395,8 +4426,8 @@ namespace Ryujinx.Tests.Cpu.Tester
             int part = (int)UInt(Q);
             int elements = datasize / esize;
 
-            bool unsigned = (U == true);
             bool accumulate = (op == false);
+            bool unsigned = (U == true);
 
             /* Operation */
             /* CheckFPAdvSIMDEnabled64(); */
@@ -4484,8 +4515,8 @@ namespace Ryujinx.Tests.Cpu.Tester
             int part = (int)UInt(Q);
             int elements = datasize / esize;
 
-            bool unsigned = (U == true);
             bool accumulate = (op == false);
+            bool unsigned = (U == true);
 
             /* Operation */
             /* CheckFPAdvSIMDEnabled64(); */
@@ -4511,6 +4542,108 @@ namespace Ryujinx.Tests.Cpu.Tester
             V(d, result);
         }
 
+        // saddw_advsimd.html
+        public static void Saddw_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 = V(2 * datasize, n);
+            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, 2 * 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)
+        {
+            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 = V(2 * datasize, n);
+            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, 2 * 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);
+        }
+
         // sub_advsimd.html#SUB_asisdsame_only
         public static void Sub_S(Bits size, Bits Rm, Bits Rn, Bits Rd)
         {
@@ -4785,8 +4918,8 @@ namespace Ryujinx.Tests.Cpu.Tester
             int part = (int)UInt(Q);
             int elements = datasize / esize;
 
-            bool unsigned = (U == true);
             bool accumulate = (op == false);
+            bool unsigned = (U == true);
 
             /* Operation */
             /* CheckFPAdvSIMDEnabled64(); */
@@ -4874,8 +5007,8 @@ namespace Ryujinx.Tests.Cpu.Tester
             int part = (int)UInt(Q);
             int elements = datasize / esize;
 
-            bool unsigned = (U == true);
             bool accumulate = (op == false);
+            bool unsigned = (U == true);
 
             /* Operation */
             /* CheckFPAdvSIMDEnabled64(); */
@@ -4901,6 +5034,108 @@ namespace Ryujinx.Tests.Cpu.Tester
             V(d, result);
         }
 
+        // uaddw_advsimd.html
+        public static void Uaddw_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 = V(2 * datasize, n);
+            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, 2 * 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)
+        {
+            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 = V(2 * datasize, n);
+            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, 2 * 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);
+        }
+
         // uzp1_advsimd.html
         public static void Uzp1_V(bool Q, Bits size, Bits Rm, Bits Rn, Bits Rd)
         {