diff --git a/ChocolArm64/AOptimizations.cs b/ChocolArm64/AOptimizations.cs index 2627c2368..e8c1f7c44 100644 --- a/ChocolArm64/AOptimizations.cs +++ b/ChocolArm64/AOptimizations.cs @@ -1,6 +1,12 @@ +using System.Runtime.Intrinsics.X86; + public static class AOptimizations { public static bool DisableMemoryChecks = false; public static bool GenerateCallStack = true; + + public static bool UseSse2IfAvailable = true; + + internal static bool UseSse2 = UseSse2IfAvailable && Sse2.IsSupported; } \ No newline at end of file diff --git a/ChocolArm64/ChocolArm64.csproj b/ChocolArm64/ChocolArm64.csproj index 47d66cf15..c878bdac1 100644 --- a/ChocolArm64/ChocolArm64.csproj +++ b/ChocolArm64/ChocolArm64.csproj @@ -1,7 +1,7 @@ - netcoreapp2.0 + netcoreapp2.1 @@ -12,4 +12,8 @@ true + + + + diff --git a/ChocolArm64/Instruction/AInstEmitSimdArithmetic.cs b/ChocolArm64/Instruction/AInstEmitSimdArithmetic.cs index 3a4b22106..0dfe0bd31 100644 --- a/ChocolArm64/Instruction/AInstEmitSimdArithmetic.cs +++ b/ChocolArm64/Instruction/AInstEmitSimdArithmetic.cs @@ -4,6 +4,7 @@ using ChocolArm64.Translation; using System; using System.Reflection; using System.Reflection.Emit; +using System.Runtime.Intrinsics.X86; using static ChocolArm64.Instruction.AInstEmitSimdHelper; @@ -41,7 +42,14 @@ namespace ChocolArm64.Instruction public static void Add_V(AILEmitterCtx Context) { - EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Add)); + if (AOptimizations.UseSse2) + { + EmitSse2Call(Context, nameof(Sse2.Add)); + } + else + { + EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Add)); + } } public static void Addhn_V(AILEmitterCtx Context) @@ -158,7 +166,7 @@ namespace ChocolArm64.Instruction Context.Emit(OpCodes.Conv_U1); - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.CountSetBits8)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.CountSetBits8)); Context.Emit(OpCodes.Conv_U8); @@ -303,12 +311,26 @@ namespace ChocolArm64.Instruction public static void Fadd_S(AILEmitterCtx Context) { - EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Add)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.AddScalar)); + } + else + { + EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Add)); + } } public static void Fadd_V(AILEmitterCtx Context) { - EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Add)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.Add)); + } + else + { + EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Add)); + } } public static void Faddp_V(AILEmitterCtx Context) @@ -345,12 +367,26 @@ namespace ChocolArm64.Instruction public static void Fdiv_S(AILEmitterCtx Context) { - EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Div)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.DivideScalar)); + } + else + { + EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Div)); + } } public static void Fdiv_V(AILEmitterCtx Context) { - EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Div)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.Divide)); + } + else + { + EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Div)); + } } public static void Fmadd_S(AILEmitterCtx Context) @@ -370,11 +406,11 @@ namespace ChocolArm64.Instruction { if (Op.Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.MaxF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.MaxF)); } else if (Op.Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Max)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Max)); } else { @@ -391,11 +427,11 @@ namespace ChocolArm64.Instruction { if (Op.Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.MaxF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.MaxF)); } else if (Op.Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Max)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Max)); } else { @@ -412,11 +448,11 @@ namespace ChocolArm64.Instruction { if (Op.Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.MinF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.MinF)); } else if (Op.Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Min)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Min)); } else { @@ -435,11 +471,11 @@ namespace ChocolArm64.Instruction { if (SizeF == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.MinF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.MinF)); } else if (SizeF == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Min)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Min)); } else { @@ -505,7 +541,14 @@ namespace ChocolArm64.Instruction public static void Fmul_S(AILEmitterCtx Context) { - EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Mul)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.MultiplyScalar)); + } + else + { + EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Mul)); + } } public static void Fmul_Se(AILEmitterCtx Context) @@ -515,7 +558,14 @@ namespace ChocolArm64.Instruction public static void Fmul_V(AILEmitterCtx Context) { - EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Mul)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.Multiply)); + } + else + { + EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Mul)); + } } public static void Fmul_Ve(AILEmitterCtx Context) @@ -716,11 +766,11 @@ namespace ChocolArm64.Instruction if (Op.Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.RoundF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.RoundF)); } else if (Op.Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Round)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Round)); } else { @@ -743,11 +793,11 @@ namespace ChocolArm64.Instruction if (SizeF == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.RoundF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.RoundF)); } else if (SizeF == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Round)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Round)); } else { @@ -819,11 +869,11 @@ namespace ChocolArm64.Instruction if (Op.Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.RoundF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.RoundF)); } else if (Op.Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Round)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Round)); } else { @@ -844,11 +894,11 @@ namespace ChocolArm64.Instruction if (Op.Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.RoundF)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.RoundF)); } else if (Op.Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.Round)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.Round)); } else { @@ -947,12 +997,26 @@ namespace ChocolArm64.Instruction public static void Fsub_S(AILEmitterCtx Context) { - EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Sub)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.SubtractScalar)); + } + else + { + EmitScalarBinaryOpF(Context, () => Context.Emit(OpCodes.Sub)); + } } public static void Fsub_V(AILEmitterCtx Context) { - EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Sub)); + if (AOptimizations.UseSse2) + { + EmitSse2CallF(Context, nameof(Sse2.Subtract)); + } + else + { + EmitVectorBinaryOpF(Context, () => Context.Emit(OpCodes.Sub)); + } } public static void Mla_V(AILEmitterCtx Context) @@ -1066,7 +1130,14 @@ namespace ChocolArm64.Instruction public static void Sub_V(AILEmitterCtx Context) { - EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Sub)); + if (AOptimizations.UseSse2) + { + EmitSse2Call(Context, nameof(Sse2.Subtract)); + } + else + { + EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Sub)); + } } public static void Subhn_V(AILEmitterCtx Context) diff --git a/ChocolArm64/Instruction/AInstEmitSimdCmp.cs b/ChocolArm64/Instruction/AInstEmitSimdCmp.cs index f155d7e86..583ad7020 100644 --- a/ChocolArm64/Instruction/AInstEmitSimdCmp.cs +++ b/ChocolArm64/Instruction/AInstEmitSimdCmp.cs @@ -3,6 +3,7 @@ using ChocolArm64.State; using ChocolArm64.Translation; using System; using System.Reflection.Emit; +using System.Runtime.Intrinsics.X86; using static ChocolArm64.Instruction.AInstEmitAluHelper; using static ChocolArm64.Instruction.AInstEmitSimdHelper; @@ -13,17 +14,38 @@ namespace ChocolArm64.Instruction { public static void Cmeq_V(AILEmitterCtx Context) { - EmitVectorCmp(Context, OpCodes.Beq_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2Call(Context, nameof(Sse2.CompareEqual)); + } + else + { + EmitVectorCmp(Context, OpCodes.Beq_S); + } } public static void Cmge_V(AILEmitterCtx Context) { - EmitVectorCmp(Context, OpCodes.Bge_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2Call(Context, nameof(Sse2.CompareGreaterThanOrEqual)); + } + else + { + EmitVectorCmp(Context, OpCodes.Bge_S); + } } public static void Cmgt_V(AILEmitterCtx Context) { - EmitVectorCmp(Context, OpCodes.Bgt_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2Call(Context, nameof(Sse2.CompareGreaterThan)); + } + else + { + EmitVectorCmp(Context, OpCodes.Bgt_S); + } } public static void Cmhi_V(AILEmitterCtx Context) @@ -112,32 +134,74 @@ namespace ChocolArm64.Instruction public static void Fcmeq_S(AILEmitterCtx Context) { - EmitScalarFcmp(Context, OpCodes.Beq_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2CallF(Context, nameof(Sse2.CompareEqualScalar)); + } + else + { + EmitScalarFcmp(Context, OpCodes.Beq_S); + } } public static void Fcmeq_V(AILEmitterCtx Context) { - EmitVectorFcmp(Context, OpCodes.Beq_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2CallF(Context, nameof(Sse2.CompareEqual)); + } + else + { + EmitVectorFcmp(Context, OpCodes.Beq_S); + } } public static void Fcmge_S(AILEmitterCtx Context) { - EmitScalarFcmp(Context, OpCodes.Bge_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2CallF(Context, nameof(Sse2.CompareGreaterThanOrEqualScalar)); + } + else + { + EmitScalarFcmp(Context, OpCodes.Bge_S); + } } public static void Fcmge_V(AILEmitterCtx Context) { - EmitVectorFcmp(Context, OpCodes.Bge_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2CallF(Context, nameof(Sse2.CompareGreaterThanOrEqual)); + } + else + { + EmitVectorFcmp(Context, OpCodes.Bge_S); + } } public static void Fcmgt_S(AILEmitterCtx Context) { - EmitScalarFcmp(Context, OpCodes.Bgt_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2CallF(Context, nameof(Sse2.CompareGreaterThanScalar)); + } + else + { + EmitScalarFcmp(Context, OpCodes.Bgt_S); + } } public static void Fcmgt_V(AILEmitterCtx Context) { - EmitVectorFcmp(Context, OpCodes.Bgt_S); + if (AOptimizations.UseSse2 && Context.CurrOp is AOpCodeSimdReg) + { + EmitSse2CallF(Context, nameof(Sse2.CompareGreaterThan)); + } + else + { + EmitVectorFcmp(Context, OpCodes.Bgt_S); + } } public static void Fcmle_S(AILEmitterCtx Context) diff --git a/ChocolArm64/Instruction/AInstEmitSimdCvt.cs b/ChocolArm64/Instruction/AInstEmitSimdCvt.cs index b9944e565..444638e60 100644 --- a/ChocolArm64/Instruction/AInstEmitSimdCvt.cs +++ b/ChocolArm64/Instruction/AInstEmitSimdCvt.cs @@ -382,15 +382,15 @@ namespace ChocolArm64.Instruction if (SizeF == 0) { - ASoftFallback.EmitCall(Context, Signed - ? nameof(ASoftFallback.SatF32ToS32) - : nameof(ASoftFallback.SatF32ToU32)); + AVectorHelper.EmitCall(Context, Signed + ? nameof(AVectorHelper.SatF32ToS32) + : nameof(AVectorHelper.SatF32ToU32)); } else /* if (SizeF == 1) */ { - ASoftFallback.EmitCall(Context, Signed - ? nameof(ASoftFallback.SatF64ToS64) - : nameof(ASoftFallback.SatF64ToU64)); + AVectorHelper.EmitCall(Context, Signed + ? nameof(AVectorHelper.SatF64ToS64) + : nameof(AVectorHelper.SatF64ToU64)); } if (SizeF == 0) @@ -420,22 +420,22 @@ namespace ChocolArm64.Instruction { if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF32ToS32)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF32ToS32)); } else /* if (Size == 1) */ { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF64ToS32)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF64ToS32)); } } else { if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF32ToS64)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF32ToS64)); } else /* if (Size == 1) */ { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF64ToS64)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF64ToS64)); } } } @@ -453,22 +453,22 @@ namespace ChocolArm64.Instruction { if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF32ToU32)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF32ToU32)); } else /* if (Size == 1) */ { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF64ToU32)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF64ToU32)); } } else { if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF32ToU64)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF32ToU64)); } else /* if (Size == 1) */ { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.SatF64ToU64)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.SatF64ToU64)); } } } diff --git a/ChocolArm64/Instruction/AInstEmitSimdHelper.cs b/ChocolArm64/Instruction/AInstEmitSimdHelper.cs index 9ef9d02f7..3caf2a3ed 100644 --- a/ChocolArm64/Instruction/AInstEmitSimdHelper.cs +++ b/ChocolArm64/Instruction/AInstEmitSimdHelper.cs @@ -3,6 +3,8 @@ using ChocolArm64.State; using ChocolArm64.Translation; using System; using System.Reflection; +using System.Runtime.Intrinsics; +using System.Runtime.Intrinsics.X86; namespace ChocolArm64.Instruction { @@ -32,6 +34,129 @@ namespace ChocolArm64.Instruction return (8 << (Op.Size + 1)) - Op.Imm; } + public static void EmitSse2Call(AILEmitterCtx Context, string Name) + { + AOpCodeSimd Op = (AOpCodeSimd)Context.CurrOp; + + int SizeF = Op.Size & 1; + + void Ldvec(int Reg) + { + Context.EmitLdvec(Reg); + + switch (Op.Size) + { + case 0: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorSingleToSByte)); break; + case 1: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorSingleToInt16)); break; + case 2: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorSingleToInt32)); break; + case 3: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorSingleToInt64)); break; + } + } + + Ldvec(Op.Rn); + + Type BaseType = null; + + Type[] Types; + + switch (Op.Size) + { + case 0: BaseType = typeof(Vector128); break; + case 1: BaseType = typeof(Vector128); break; + case 2: BaseType = typeof(Vector128); break; + case 3: BaseType = typeof(Vector128); break; + } + + if (Op is AOpCodeSimdReg BinOp) + { + Ldvec(BinOp.Rm); + + Types = new Type[] { BaseType, BaseType }; + } + else + { + Types = new Type[] { BaseType }; + } + + Context.EmitCall(typeof(Sse2).GetMethod(Name, Types)); + + switch (Op.Size) + { + case 0: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorSByteToSingle)); break; + case 1: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInt16ToSingle)); break; + case 2: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInt32ToSingle)); break; + case 3: AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInt64ToSingle)); break; + } + + Context.EmitStvec(Op.Rd); + + if (Op.RegisterSize == ARegisterSize.SIMD64) + { + EmitVectorZeroUpper(Context, Op.Rd); + } + } + + public static void EmitSse2CallF(AILEmitterCtx Context, string Name) + { + AOpCodeSimd Op = (AOpCodeSimd)Context.CurrOp; + + int SizeF = Op.Size & 1; + + void Ldvec(int Reg) + { + Context.EmitLdvec(Reg); + + if (SizeF == 1) + { + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorSingleToDouble)); + } + } + + Ldvec(Op.Rn); + + Type BaseType = SizeF == 0 + ? typeof(Vector128) + : typeof(Vector128); + + Type[] Types; + + if (Op is AOpCodeSimdReg BinOp) + { + Ldvec(BinOp.Rm); + + Types = new Type[] { BaseType, BaseType }; + } + else + { + Types = new Type[] { BaseType }; + } + + MethodInfo MthdInfo; + + if (SizeF == 0) + { + MthdInfo = typeof(Sse).GetMethod(Name, Types); + } + else /* if (SizeF == 1) */ + { + MthdInfo = typeof(Sse2).GetMethod(Name, Types); + } + + Context.EmitCall(MthdInfo); + + if (SizeF == 1) + { + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorDoubleToSingle)); + } + + Context.EmitStvec(Op.Rd); + + if (Op.RegisterSize == ARegisterSize.SIMD64) + { + EmitVectorZeroUpper(Context, Op.Rd); + } + } + public static void EmitUnaryMathCall(AILEmitterCtx Context, string Name) { IAOpCodeSimd Op = (IAOpCodeSimd)Context.CurrOp; @@ -596,9 +721,9 @@ namespace ChocolArm64.Instruction Context.EmitLdc_I4(Index); Context.EmitLdc_I4(Size); - ASoftFallback.EmitCall(Context, Signed - ? nameof(ASoftFallback.VectorExtractIntSx) - : nameof(ASoftFallback.VectorExtractIntZx)); + AVectorHelper.EmitCall(Context, Signed + ? nameof(AVectorHelper.VectorExtractIntSx) + : nameof(AVectorHelper.VectorExtractIntZx)); } public static void EmitVectorExtractF(AILEmitterCtx Context, int Reg, int Index, int Size) @@ -610,11 +735,11 @@ namespace ChocolArm64.Instruction if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorExtractSingle)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorExtractSingle)); } else if (Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorExtractDouble)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorExtractDouble)); } else { @@ -646,7 +771,7 @@ namespace ChocolArm64.Instruction Context.EmitLdc_I4(Index); Context.EmitLdc_I4(Size); - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertInt)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertInt)); Context.EmitStvec(Reg); } @@ -659,7 +784,7 @@ namespace ChocolArm64.Instruction Context.EmitLdc_I4(Index); Context.EmitLdc_I4(Size); - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertInt)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertInt)); Context.EmitStvectmp(); } @@ -673,7 +798,7 @@ namespace ChocolArm64.Instruction Context.EmitLdc_I4(Index); Context.EmitLdc_I4(Size); - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertInt)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertInt)); Context.EmitStvec(Reg); } @@ -687,11 +812,11 @@ namespace ChocolArm64.Instruction if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertSingle)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertSingle)); } else if (Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertDouble)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertDouble)); } else { @@ -710,11 +835,11 @@ namespace ChocolArm64.Instruction if (Size == 0) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertSingle)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertSingle)); } else if (Size == 1) { - ASoftFallback.EmitCall(Context, nameof(ASoftFallback.VectorInsertDouble)); + AVectorHelper.EmitCall(Context, nameof(AVectorHelper.VectorInsertDouble)); } else { diff --git a/ChocolArm64/Instruction/AInstEmitSimdLogical.cs b/ChocolArm64/Instruction/AInstEmitSimdLogical.cs index 25aa873bd..9c897bfeb 100644 --- a/ChocolArm64/Instruction/AInstEmitSimdLogical.cs +++ b/ChocolArm64/Instruction/AInstEmitSimdLogical.cs @@ -2,6 +2,7 @@ using ChocolArm64.Decoder; using ChocolArm64.State; using ChocolArm64.Translation; using System.Reflection.Emit; +using System.Runtime.Intrinsics.X86; using static ChocolArm64.Instruction.AInstEmitSimdHelper; @@ -11,7 +12,14 @@ namespace ChocolArm64.Instruction { public static void And_V(AILEmitterCtx Context) { - EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.And)); + if (AOptimizations.UseSse2) + { + EmitSse2Call(Context, nameof(Sse2.And)); + } + else + { + EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.And)); + } } public static void Bic_V(AILEmitterCtx Context) @@ -95,7 +103,14 @@ namespace ChocolArm64.Instruction public static void Eor_V(AILEmitterCtx Context) { - EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Xor)); + if (AOptimizations.UseSse2) + { + EmitSse2Call(Context, nameof(Sse2.Xor)); + } + else + { + EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Xor)); + } } public static void Not_V(AILEmitterCtx Context) @@ -114,7 +129,14 @@ namespace ChocolArm64.Instruction public static void Orr_V(AILEmitterCtx Context) { - EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Or)); + if (AOptimizations.UseSse2) + { + EmitSse2Call(Context, nameof(Sse2.Or)); + } + else + { + EmitVectorBinaryOpZx(Context, () => Context.Emit(OpCodes.Or)); + } } public static void Orr_Vi(AILEmitterCtx Context) diff --git a/ChocolArm64/Instruction/AInstEmitSimdMove.cs b/ChocolArm64/Instruction/AInstEmitSimdMove.cs index 20268d583..95fe59499 100644 --- a/ChocolArm64/Instruction/AInstEmitSimdMove.cs +++ b/ChocolArm64/Instruction/AInstEmitSimdMove.cs @@ -234,21 +234,21 @@ namespace ChocolArm64.Instruction switch (Op.Size) { - case 1: ASoftFallback.EmitCall(Context, - nameof(ASoftFallback.Tbl1_V64), - nameof(ASoftFallback.Tbl1_V128)); break; + case 1: AVectorHelper.EmitCall(Context, + nameof(AVectorHelper.Tbl1_V64), + nameof(AVectorHelper.Tbl1_V128)); break; - case 2: ASoftFallback.EmitCall(Context, - nameof(ASoftFallback.Tbl2_V64), - nameof(ASoftFallback.Tbl2_V128)); break; + case 2: AVectorHelper.EmitCall(Context, + nameof(AVectorHelper.Tbl2_V64), + nameof(AVectorHelper.Tbl2_V128)); break; - case 3: ASoftFallback.EmitCall(Context, - nameof(ASoftFallback.Tbl3_V64), - nameof(ASoftFallback.Tbl3_V128)); break; + case 3: AVectorHelper.EmitCall(Context, + nameof(AVectorHelper.Tbl3_V64), + nameof(AVectorHelper.Tbl3_V128)); break; - case 4: ASoftFallback.EmitCall(Context, - nameof(ASoftFallback.Tbl4_V64), - nameof(ASoftFallback.Tbl4_V128)); break; + case 4: AVectorHelper.EmitCall(Context, + nameof(AVectorHelper.Tbl4_V64), + nameof(AVectorHelper.Tbl4_V128)); break; default: throw new InvalidOperationException(); } diff --git a/ChocolArm64/Instruction/ASoftFallback.cs b/ChocolArm64/Instruction/ASoftFallback.cs index 497605a41..8ed55e204 100644 --- a/ChocolArm64/Instruction/ASoftFallback.cs +++ b/ChocolArm64/Instruction/ASoftFallback.cs @@ -1,20 +1,11 @@ -using ChocolArm64.State; using ChocolArm64.Translation; using System; using System.Numerics; -using System.Runtime.CompilerServices; namespace ChocolArm64.Instruction { static class ASoftFallback { - public static void EmitCall(AILEmitterCtx Context, string Name64, string Name128) - { - bool IsSimd64 = Context.CurrOp.RegisterSize == ARegisterSize.SIMD64; - - Context.EmitCall(typeof(ASoftFallback), IsSimd64 ? Name64 : Name128); - } - public static void EmitCall(AILEmitterCtx Context, string MthdName) { Context.EmitCall(typeof(ASoftFallback), MthdName); @@ -160,78 +151,6 @@ namespace ChocolArm64.Instruction throw new ArgumentException(nameof(Size)); } - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static int SatF32ToS32(float Value) - { - if (float.IsNaN(Value)) return 0; - - return Value > int.MaxValue ? int.MaxValue : - Value < int.MinValue ? int.MinValue : (int)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static long SatF32ToS64(float Value) - { - if (float.IsNaN(Value)) return 0; - - return Value > long.MaxValue ? long.MaxValue : - Value < long.MinValue ? long.MinValue : (long)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static uint SatF32ToU32(float Value) - { - if (float.IsNaN(Value)) return 0; - - return Value > uint.MaxValue ? uint.MaxValue : - Value < uint.MinValue ? uint.MinValue : (uint)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static ulong SatF32ToU64(float Value) - { - if (float.IsNaN(Value)) return 0; - - return Value > ulong.MaxValue ? ulong.MaxValue : - Value < ulong.MinValue ? ulong.MinValue : (ulong)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static int SatF64ToS32(double Value) - { - if (double.IsNaN(Value)) return 0; - - return Value > int.MaxValue ? int.MaxValue : - Value < int.MinValue ? int.MinValue : (int)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static long SatF64ToS64(double Value) - { - if (double.IsNaN(Value)) return 0; - - return Value > long.MaxValue ? long.MaxValue : - Value < long.MinValue ? long.MinValue : (long)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static uint SatF64ToU32(double Value) - { - if (double.IsNaN(Value)) return 0; - - return Value > uint.MaxValue ? uint.MaxValue : - Value < uint.MinValue ? uint.MinValue : (uint)Value; - } - - [MethodImpl(MethodImplOptions.AggressiveInlining)] - public static ulong SatF64ToU64(double Value) - { - if (double.IsNaN(Value)) return 0; - - return Value > ulong.MaxValue ? ulong.MaxValue : - Value < ulong.MinValue ? ulong.MinValue : (ulong)Value; - } - public static long SMulHi128(long LHS, long RHS) { return (long)(BigInteger.Multiply(LHS, RHS) >> 64); @@ -241,239 +160,5 @@ namespace ChocolArm64.Instruction { return (ulong)(BigInteger.Multiply(LHS, RHS) >> 64); } - - public static int CountSetBits8(byte Value) - { - return ((Value >> 0) & 1) + ((Value >> 1) & 1) + - ((Value >> 2) & 1) + ((Value >> 3) & 1) + - ((Value >> 4) & 1) + ((Value >> 5) & 1) + - ((Value >> 6) & 1) + (Value >> 7); - } - - public static float MaxF(float val1, float val2) - { - if (val1 == 0.0 && val2 == 0.0) - { - if (BitConverter.SingleToInt32Bits(val1) < 0 && BitConverter.SingleToInt32Bits(val2) < 0) - return -0.0f; - - return 0.0f; - } - - if (val1 > val2) - return val1; - - if (float.IsNaN(val1)) - return val1; - - return val2; - } - - public static double Max(double val1, double val2) - { - if (val1 == 0.0 && val2 == 0.0) - { - if (BitConverter.DoubleToInt64Bits(val1) < 0 && BitConverter.DoubleToInt64Bits(val2) < 0) - return -0.0; - - return 0.0; - } - - if (val1 > val2) - return val1; - - if (double.IsNaN(val1)) - return val1; - - return val2; - } - - public static float MinF(float val1, float val2) - { - if (val1 == 0.0 && val2 == 0.0) - { - if (BitConverter.SingleToInt32Bits(val1) < 0 || BitConverter.SingleToInt32Bits(val2) < 0) - return -0.0f; - - return 0.0f; - } - - if (val1 < val2) - return val1; - - if (float.IsNaN(val1)) - return val1; - - return val2; - } - - public static double Min(double val1, double val2) - { - if (val1 == 0.0 && val2 == 0.0) - { - if (BitConverter.DoubleToInt64Bits(val1) < 0 || BitConverter.DoubleToInt64Bits(val2) < 0) - return -0.0; - - return 0.0; - } - - if (val1 < val2) - return val1; - - if (double.IsNaN(val1)) - return val1; - - return val2; - } - - public static float RoundF(float Value, int Fpcr) - { - switch ((ARoundMode)((Fpcr >> 22) & 3)) - { - case ARoundMode.ToNearest: return MathF.Round (Value); - case ARoundMode.TowardsPlusInfinity: return MathF.Ceiling (Value); - case ARoundMode.TowardsMinusInfinity: return MathF.Floor (Value); - case ARoundMode.TowardsZero: return MathF.Truncate(Value); - } - - throw new InvalidOperationException(); - } - - public static double Round(double Value, int Fpcr) - { - switch ((ARoundMode)((Fpcr >> 22) & 3)) - { - case ARoundMode.ToNearest: return Math.Round (Value); - case ARoundMode.TowardsPlusInfinity: return Math.Ceiling (Value); - case ARoundMode.TowardsMinusInfinity: return Math.Floor (Value); - case ARoundMode.TowardsZero: return Math.Truncate(Value); - } - - throw new InvalidOperationException(); - } - - public static AVec Tbl1_V64(AVec Vector, AVec Tb0) - { - return Tbl(Vector, 8, Tb0); - } - - public static AVec Tbl1_V128(AVec Vector, AVec Tb0) - { - return Tbl(Vector, 16, Tb0); - } - - public static AVec Tbl2_V64(AVec Vector, AVec Tb0, AVec Tb1) - { - return Tbl(Vector, 8, Tb0, Tb1); - } - - public static AVec Tbl2_V128(AVec Vector, AVec Tb0, AVec Tb1) - { - return Tbl(Vector, 16, Tb0, Tb1); - } - - public static AVec Tbl3_V64(AVec Vector, AVec Tb0, AVec Tb1, AVec Tb2) - { - return Tbl(Vector, 8, Tb0, Tb1, Tb2); - } - - public static AVec Tbl3_V128(AVec Vector, AVec Tb0, AVec Tb1, AVec Tb2) - { - return Tbl(Vector, 16, Tb0, Tb1, Tb2); - } - - public static AVec Tbl4_V64(AVec Vector, AVec Tb0, AVec Tb1, AVec Tb2, AVec Tb3) - { - return Tbl(Vector, 8, Tb0, Tb1, Tb2, Tb3); - } - - public static AVec Tbl4_V128(AVec Vector, AVec Tb0, AVec Tb1, AVec Tb2, AVec Tb3) - { - return Tbl(Vector, 16, Tb0, Tb1, Tb2, Tb3); - } - - private static AVec Tbl(AVec Vector, int Bytes, params AVec[] Tb) - { - AVec Res = new AVec(); - - byte[] Table = new byte[Tb.Length * 16]; - - for (int Index = 0; Index < Tb.Length; Index++) - for (int Index2 = 0; Index2 < 16; Index2++) - { - Table[Index * 16 + Index2] = (byte)VectorExtractIntZx(Tb[Index], Index2, 0); - } - - for (int Index = 0; Index < Bytes; Index++) - { - byte TblIdx = (byte)VectorExtractIntZx(Vector, Index, 0); - - if (TblIdx < Table.Length) - { - Res = VectorInsertInt(Table[TblIdx], Res, Index, 0); - } - } - - return Res; - } - - public static ulong VectorExtractIntZx(AVec Vector, int Index, int Size) - { - switch (Size) - { - case 0: return Vector.ExtractByte (Index); - case 1: return Vector.ExtractUInt16(Index); - case 2: return Vector.ExtractUInt32(Index); - case 3: return Vector.ExtractUInt64(Index); - } - - throw new ArgumentOutOfRangeException(nameof(Size)); - } - - public static long VectorExtractIntSx(AVec Vector, int Index, int Size) - { - switch (Size) - { - case 0: return (sbyte)Vector.ExtractByte (Index); - case 1: return (short)Vector.ExtractUInt16(Index); - case 2: return (int)Vector.ExtractUInt32(Index); - case 3: return (long)Vector.ExtractUInt64(Index); - } - - throw new ArgumentOutOfRangeException(nameof(Size)); - } - - public static float VectorExtractSingle(AVec Vector, int Index) - { - return Vector.ExtractSingle(Index); - } - - public static double VectorExtractDouble(AVec Vector, int Index) - { - return Vector.ExtractDouble(Index); - } - - public static AVec VectorInsertSingle(float Value, AVec Vector, int Index) - { - return AVec.InsertSingle(Vector, Index, Value); - } - - public static AVec VectorInsertDouble(double Value, AVec Vector, int Index) - { - return AVec.InsertDouble(Vector, Index, Value); - } - - public static AVec VectorInsertInt(ulong Value, AVec Vector, int Index, int Size) - { - switch (Size) - { - case 0: return AVec.InsertByte (Vector, Index, (byte)Value); - case 1: return AVec.InsertUInt16(Vector, Index, (ushort)Value); - case 2: return AVec.InsertUInt32(Vector, Index, (uint)Value); - case 3: return AVec.InsertUInt64(Vector, Index, (ulong)Value); - } - - throw new ArgumentOutOfRangeException(nameof(Size)); - } } } diff --git a/ChocolArm64/Instruction/AVectorHelper.cs b/ChocolArm64/Instruction/AVectorHelper.cs new file mode 100644 index 000000000..1a2135923 --- /dev/null +++ b/ChocolArm64/Instruction/AVectorHelper.cs @@ -0,0 +1,626 @@ +using ChocolArm64.State; +using ChocolArm64.Translation; +using System; +using System.Runtime.CompilerServices; +using System.Runtime.Intrinsics; +using System.Runtime.Intrinsics.X86; + +namespace ChocolArm64.Instruction +{ + static class AVectorHelper + { + public static void EmitCall(AILEmitterCtx Context, string Name64, string Name128) + { + bool IsSimd64 = Context.CurrOp.RegisterSize == ARegisterSize.SIMD64; + + Context.EmitCall(typeof(AVectorHelper), IsSimd64 ? Name64 : Name128); + } + + public static void EmitCall(AILEmitterCtx Context, string MthdName) + { + Context.EmitCall(typeof(AVectorHelper), MthdName); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static int SatF32ToS32(float Value) + { + if (float.IsNaN(Value)) return 0; + + return Value > int.MaxValue ? int.MaxValue : + Value < int.MinValue ? int.MinValue : (int)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static long SatF32ToS64(float Value) + { + if (float.IsNaN(Value)) return 0; + + return Value > long.MaxValue ? long.MaxValue : + Value < long.MinValue ? long.MinValue : (long)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static uint SatF32ToU32(float Value) + { + if (float.IsNaN(Value)) return 0; + + return Value > uint.MaxValue ? uint.MaxValue : + Value < uint.MinValue ? uint.MinValue : (uint)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static ulong SatF32ToU64(float Value) + { + if (float.IsNaN(Value)) return 0; + + return Value > ulong.MaxValue ? ulong.MaxValue : + Value < ulong.MinValue ? ulong.MinValue : (ulong)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static int SatF64ToS32(double Value) + { + if (double.IsNaN(Value)) return 0; + + return Value > int.MaxValue ? int.MaxValue : + Value < int.MinValue ? int.MinValue : (int)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static long SatF64ToS64(double Value) + { + if (double.IsNaN(Value)) return 0; + + return Value > long.MaxValue ? long.MaxValue : + Value < long.MinValue ? long.MinValue : (long)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static uint SatF64ToU32(double Value) + { + if (double.IsNaN(Value)) return 0; + + return Value > uint.MaxValue ? uint.MaxValue : + Value < uint.MinValue ? uint.MinValue : (uint)Value; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static ulong SatF64ToU64(double Value) + { + if (double.IsNaN(Value)) return 0; + + return Value > ulong.MaxValue ? ulong.MaxValue : + Value < ulong.MinValue ? ulong.MinValue : (ulong)Value; + } + + public static int CountSetBits8(byte Value) + { + return ((Value >> 0) & 1) + ((Value >> 1) & 1) + + ((Value >> 2) & 1) + ((Value >> 3) & 1) + + ((Value >> 4) & 1) + ((Value >> 5) & 1) + + ((Value >> 6) & 1) + (Value >> 7); + } + + public static double Max(double LHS, double RHS) + { + if (LHS == 0.0 && RHS == 0.0) + { + if (BitConverter.DoubleToInt64Bits(LHS) < 0 && + BitConverter.DoubleToInt64Bits(RHS) < 0) + return -0.0; + + return 0.0; + } + + if (LHS > RHS) + return LHS; + + if (double.IsNaN(LHS)) + return LHS; + + return RHS; + } + + public static float MaxF(float LHS, float RHS) + { + if (LHS == 0.0 && RHS == 0.0) + { + if (BitConverter.SingleToInt32Bits(LHS) < 0 && + BitConverter.SingleToInt32Bits(RHS) < 0) + return -0.0f; + + return 0.0f; + } + + if (LHS > RHS) + return LHS; + + if (float.IsNaN(LHS)) + return LHS; + + return RHS; + } + + public static double Min(double LHS, double RHS) + { + if (LHS == 0.0 && RHS == 0.0) + { + if (BitConverter.DoubleToInt64Bits(LHS) < 0 || + BitConverter.DoubleToInt64Bits(RHS) < 0) + return -0.0; + + return 0.0; + } + + if (LHS < RHS) + return LHS; + + if (double.IsNaN(LHS)) + return LHS; + + return RHS; + } + + public static float MinF(float LHS, float RHS) + { + if (LHS == 0.0 && RHS == 0.0) + { + if (BitConverter.SingleToInt32Bits(LHS) < 0 || + BitConverter.SingleToInt32Bits(RHS) < 0) + return -0.0f; + + return 0.0f; + } + + if (LHS < RHS) + return LHS; + + if (float.IsNaN(LHS)) + return LHS; + + return RHS; + } + + public static double Round(double Value, int Fpcr) + { + switch ((ARoundMode)((Fpcr >> 22) & 3)) + { + case ARoundMode.ToNearest: return Math.Round (Value); + case ARoundMode.TowardsPlusInfinity: return Math.Ceiling (Value); + case ARoundMode.TowardsMinusInfinity: return Math.Floor (Value); + case ARoundMode.TowardsZero: return Math.Truncate(Value); + } + + throw new InvalidOperationException(); + } + + public static float RoundF(float Value, int Fpcr) + { + switch ((ARoundMode)((Fpcr >> 22) & 3)) + { + case ARoundMode.ToNearest: return MathF.Round (Value); + case ARoundMode.TowardsPlusInfinity: return MathF.Ceiling (Value); + case ARoundMode.TowardsMinusInfinity: return MathF.Floor (Value); + case ARoundMode.TowardsZero: return MathF.Truncate(Value); + } + + throw new InvalidOperationException(); + } + + public static Vector128 Tbl1_V64( + Vector128 Vector, + Vector128 Tb0) + { + return Tbl(Vector, 8, Tb0); + } + + public static Vector128 Tbl1_V128( + Vector128 Vector, + Vector128 Tb0) + { + return Tbl(Vector, 16, Tb0); + } + + public static Vector128 Tbl2_V64( + Vector128 Vector, + Vector128 Tb0, + Vector128 Tb1) + { + return Tbl(Vector, 8, Tb0, Tb1); + } + + public static Vector128 Tbl2_V128( + Vector128 Vector, + Vector128 Tb0, + Vector128 Tb1) + { + return Tbl(Vector, 16, Tb0, Tb1); + } + + public static Vector128 Tbl3_V64( + Vector128 Vector, + Vector128 Tb0, + Vector128 Tb1, + Vector128 Tb2) + { + return Tbl(Vector, 8, Tb0, Tb1, Tb2); + } + + public static Vector128 Tbl3_V128( + Vector128 Vector, + Vector128 Tb0, + Vector128 Tb1, + Vector128 Tb2) + { + return Tbl(Vector, 16, Tb0, Tb1, Tb2); + } + + public static Vector128 Tbl4_V64( + Vector128 Vector, + Vector128 Tb0, + Vector128 Tb1, + Vector128 Tb2, + Vector128 Tb3) + { + return Tbl(Vector, 8, Tb0, Tb1, Tb2, Tb3); + } + + public static Vector128 Tbl4_V128( + Vector128 Vector, + Vector128 Tb0, + Vector128 Tb1, + Vector128 Tb2, + Vector128 Tb3) + { + return Tbl(Vector, 16, Tb0, Tb1, Tb2, Tb3); + } + + private static Vector128 Tbl(Vector128 Vector, int Bytes, params Vector128[] Tb) + { + Vector128 Res = new Vector128(); + + byte[] Table = new byte[Tb.Length * 16]; + + for (byte Index = 0; Index < Tb.Length; Index++) + for (byte Index2 = 0; Index2 < 16; Index2++) + { + Table[Index * 16 + Index2] = (byte)VectorExtractIntZx(Tb[Index], Index2, 0); + } + + for (byte Index = 0; Index < Bytes; Index++) + { + byte TblIdx = (byte)VectorExtractIntZx(Vector, Index, 0); + + if (TblIdx < Table.Length) + { + Res = VectorInsertInt(Table[TblIdx], Res, Index, 0); + } + } + + return Res; + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static double VectorExtractDouble(Vector128 Vector, byte Index) + { + return BitConverter.Int64BitsToDouble(VectorExtractIntSx(Vector, Index, 3)); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static long VectorExtractIntSx(Vector128 Vector, byte Index, int Size) + { + if (Sse41.IsSupported) + { + switch (Size) + { + case 0: + return (sbyte)Sse41.Extract(Sse.StaticCast(Vector), Index); + + case 1: + return (short)Sse2.Extract(Sse.StaticCast(Vector), Index); + + case 2: + return Sse41.Extract(Sse.StaticCast(Vector), Index); + + case 3: + return Sse41.Extract(Sse.StaticCast(Vector), Index); + } + + throw new ArgumentOutOfRangeException(nameof(Size)); + } + else if (Sse2.IsSupported) + { + switch (Size) + { + case 0: + return (sbyte)VectorExtractIntZx(Vector, Index, Size); + + case 1: + return (short)VectorExtractIntZx(Vector, Index, Size); + + case 2: + return (int)VectorExtractIntZx(Vector, Index, Size); + + case 3: + return (long)VectorExtractIntZx(Vector, Index, Size); + } + + throw new ArgumentOutOfRangeException(nameof(Size)); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static ulong VectorExtractIntZx(Vector128 Vector, byte Index, int Size) + { + if (Sse41.IsSupported) + { + switch (Size) + { + case 0: + return Sse41.Extract(Sse.StaticCast(Vector), Index); + + case 1: + return Sse2.Extract(Sse.StaticCast(Vector), Index); + + case 2: + return Sse41.Extract(Sse.StaticCast(Vector), Index); + + case 3: + return Sse41.Extract(Sse.StaticCast(Vector), Index); + } + + throw new ArgumentOutOfRangeException(nameof(Size)); + } + else if (Sse2.IsSupported) + { + int ShortIdx = Size == 0 + ? Index >> 1 + : Index << (Size - 1); + + ushort Value = Sse2.Extract(Sse.StaticCast(Vector), (byte)ShortIdx); + + switch (Size) + { + case 0: + return (byte)(Value >> (Index & 1) * 8); + + case 1: + return Value; + + case 2: + case 3: + { + ushort Value1 = Sse2.Extract(Sse.StaticCast(Vector), (byte)(ShortIdx + 1)); + + if (Size == 2) + { + return (uint)(Value | (Value1 << 16)); + } + + ushort Value2 = Sse2.Extract(Sse.StaticCast(Vector), (byte)(ShortIdx + 2)); + ushort Value3 = Sse2.Extract(Sse.StaticCast(Vector), (byte)(ShortIdx + 3)); + + return ((ulong)Value << 0) | + ((ulong)Value1 << 16) | + ((ulong)Value2 << 32) | + ((ulong)Value3 << 48); + } + } + + throw new ArgumentOutOfRangeException(nameof(Size)); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static float VectorExtractSingle(Vector128 Vector, byte Index) + { + if (Sse41.IsSupported) + { + return Sse41.Extract(Vector, Index); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorInsertDouble(double Value, Vector128 Vector, byte Index) + { + return VectorInsertInt((ulong)BitConverter.DoubleToInt64Bits(Value), Vector, Index, 3); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorInsertInt(ulong Value, Vector128 Vector, byte Index, int Size) + { + if (Sse41.IsSupported) + { + switch (Size) + { + case 0: + return Sse.StaticCast(Sse41.Insert(Sse.StaticCast(Vector), (byte)Value, Index)); + + case 1: + return Sse.StaticCast(Sse2.Insert(Sse.StaticCast(Vector), (ushort)Value, Index)); + + case 2: + return Sse.StaticCast(Sse41.Insert(Sse.StaticCast(Vector), (uint)Value, Index)); + + case 3: + return Sse.StaticCast(Sse41.Insert(Sse.StaticCast(Vector), Value, Index)); + } + + throw new ArgumentOutOfRangeException(nameof(Size)); + } + else if (Sse2.IsSupported) + { + Vector128 ShortVector = Sse.StaticCast(Vector); + + int ShortIdx = Size == 0 + ? Index >> 1 + : Index << (Size - 1); + + switch (Size) + { + case 0: + { + ushort ShortVal = Sse2.Extract(Sse.StaticCast(Vector), (byte)ShortIdx); + + int Shift = (Index & 1) * 8; + + ShortVal &= (ushort)(0xff00 >> Shift); + + ShortVal |= (ushort)((byte)Value << Shift); + + return Sse.StaticCast(Sse2.Insert(ShortVector, ShortVal, (byte)ShortIdx)); + } + + case 1: + return Sse.StaticCast(Sse2.Insert(Sse.StaticCast(Vector), (ushort)Value, Index)); + + case 2: + case 3: + { + ShortVector = Sse2.Insert(ShortVector, (ushort)(Value >> 0), (byte)(ShortIdx + 0)); + ShortVector = Sse2.Insert(ShortVector, (ushort)(Value >> 16), (byte)(ShortIdx + 1)); + + if (Size == 3) + { + ShortVector = Sse2.Insert(ShortVector, (ushort)(Value >> 32), (byte)(ShortIdx + 2)); + ShortVector = Sse2.Insert(ShortVector, (ushort)(Value >> 48), (byte)(ShortIdx + 3)); + } + + return Sse.StaticCast(ShortVector); + } + } + + throw new ArgumentOutOfRangeException(nameof(Size)); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorInsertSingle(float Value, Vector128 Vector, byte Index) + { + if (Sse41.IsSupported) + { + return Sse41.Insert(Vector, Value, (byte)(Index << 4)); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorSingleToSByte(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorSingleToInt16(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorSingleToInt32(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorSingleToInt64(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorSingleToDouble(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorSByteToSingle(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorInt16ToSingle(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorInt32ToSingle(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorInt64ToSingle(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + + [MethodImpl(MethodImplOptions.AggressiveInlining)] + public static Vector128 VectorDoubleToSingle(Vector128 Vector) + { + if (Sse.IsSupported) + { + return Sse.StaticCast(Vector); + } + + throw new PlatformNotSupportedException(); + } + } +} \ No newline at end of file diff --git a/ChocolArm64/Memory/AMemory.cs b/ChocolArm64/Memory/AMemory.cs index 6625966f1..c74e11b2c 100644 --- a/ChocolArm64/Memory/AMemory.cs +++ b/ChocolArm64/Memory/AMemory.cs @@ -3,6 +3,8 @@ using ChocolArm64.State; using System; using System.Collections.Generic; using System.Runtime.InteropServices; +using System.Runtime.Intrinsics; +using System.Runtime.Intrinsics.X86; namespace ChocolArm64.Memory { @@ -189,33 +191,73 @@ namespace ChocolArm64.Memory return ReadUInt64Unchecked(Position); } - public AVec ReadVector8(long Position) + public Vector128 ReadVector8(long Position) { - return new AVec() { B0 = ReadByte(Position) }; - } - - public AVec ReadVector16(long Position) - { - return new AVec() { H0 = ReadUInt16(Position) }; - } - - public AVec ReadVector32(long Position) - { - return new AVec() { W0 = ReadUInt32(Position) }; - } - - public AVec ReadVector64(long Position) - { - return new AVec() { X0 = ReadUInt64(Position) }; - } - - public AVec ReadVector128(long Position) - { - return new AVec() + if (Sse2.IsSupported) { - X0 = ReadUInt64(Position + 0), - X1 = ReadUInt64(Position + 8) - }; + return Sse.StaticCast(Sse2.SetVector128(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ReadByte(Position))); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector16(long Position) + { + if (Sse2.IsSupported) + { + return Sse.StaticCast(Sse2.Insert(Sse2.SetZeroVector128(), ReadUInt16(Position), 0)); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector32(long Position) + { + EnsureAccessIsValid(Position + 0, AMemoryPerm.Read); + EnsureAccessIsValid(Position + 3, AMemoryPerm.Read); + + if (Sse.IsSupported) + { + return Sse.LoadScalarVector128((float*)(RamPtr + (uint)Position)); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector64(long Position) + { + EnsureAccessIsValid(Position + 0, AMemoryPerm.Read); + EnsureAccessIsValid(Position + 7, AMemoryPerm.Read); + + if (Sse2.IsSupported) + { + return Sse.StaticCast(Sse2.LoadScalarVector128((double*)(RamPtr + (uint)Position))); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector128(long Position) + { + EnsureAccessIsValid(Position + 0, AMemoryPerm.Read); + EnsureAccessIsValid(Position + 15, AMemoryPerm.Read); + + if (Sse.IsSupported) + { + return Sse.LoadVector128((float*)(RamPtr + (uint)Position)); + } + else + { + throw new PlatformNotSupportedException(); + } } public sbyte ReadSByteUnchecked(long Position) @@ -258,33 +300,64 @@ namespace ChocolArm64.Memory return *((ulong*)(RamPtr + (uint)Position)); } - public AVec ReadVector8Unchecked(long Position) + public Vector128 ReadVector8Unchecked(long Position) { - return new AVec() { B0 = ReadByteUnchecked(Position) }; - } - - public AVec ReadVector16Unchecked(long Position) - { - return new AVec() { H0 = ReadUInt16Unchecked(Position) }; - } - - public AVec ReadVector32Unchecked(long Position) - { - return new AVec() { W0 = ReadUInt32Unchecked(Position) }; - } - - public AVec ReadVector64Unchecked(long Position) - { - return new AVec() { X0 = ReadUInt64Unchecked(Position) }; - } - - public AVec ReadVector128Unchecked(long Position) - { - return new AVec() + if (Sse2.IsSupported) { - X0 = ReadUInt64Unchecked(Position + 0), - X1 = ReadUInt64Unchecked(Position + 8) - }; + return Sse.StaticCast(Sse2.SetVector128(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ReadByte(Position))); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector16Unchecked(long Position) + { + if (Sse2.IsSupported) + { + return Sse.StaticCast(Sse2.Insert(Sse2.SetZeroVector128(), ReadUInt16Unchecked(Position), 0)); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector32Unchecked(long Position) + { + if (Sse.IsSupported) + { + return Sse.LoadScalarVector128((float*)(RamPtr + (uint)Position)); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector64Unchecked(long Position) + { + if (Sse2.IsSupported) + { + return Sse.StaticCast(Sse2.LoadScalarVector128((double*)(RamPtr + (uint)Position))); + } + else + { + throw new PlatformNotSupportedException(); + } + } + + public Vector128 ReadVector128Unchecked(long Position) + { + if (Sse.IsSupported) + { + return Sse.LoadVector128((float*)(RamPtr + (uint)Position)); + } + else + { + throw new PlatformNotSupportedException(); + } } public void WriteSByte(long Position, sbyte Value) @@ -338,30 +411,77 @@ namespace ChocolArm64.Memory WriteUInt64Unchecked(Position, Value); } - public void WriteVector8(long Position, AVec Value) + public void WriteVector8(long Position, Vector128 Value) { - WriteByte(Position, Value.B0); + if (Sse41.IsSupported) + { + WriteByte(Position, Sse41.Extract(Sse.StaticCast(Value), 0)); + } + else if (Sse2.IsSupported) + { + WriteByte(Position, (byte)Sse2.Extract(Sse.StaticCast(Value), 0)); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector16(long Position, AVec Value) + public void WriteVector16(long Position, Vector128 Value) { - WriteUInt16(Position, Value.H0); + if (Sse2.IsSupported) + { + WriteUInt16(Position, Sse2.Extract(Sse.StaticCast(Value), 0)); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector32(long Position, AVec Value) + public void WriteVector32(long Position, Vector128 Value) { - WriteUInt32(Position, Value.W0); + EnsureAccessIsValid(Position + 0, AMemoryPerm.Write); + EnsureAccessIsValid(Position + 3, AMemoryPerm.Write); + + if (Sse.IsSupported) + { + Sse.StoreScalar((float*)(RamPtr + (uint)Position), Value); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector64(long Position, AVec Value) + public void WriteVector64(long Position, Vector128 Value) { - WriteUInt64(Position, Value.X0); + EnsureAccessIsValid(Position + 0, AMemoryPerm.Write); + EnsureAccessIsValid(Position + 7, AMemoryPerm.Write); + + if (Sse2.IsSupported) + { + Sse2.StoreScalar((double*)(RamPtr + (uint)Position), Sse.StaticCast(Value)); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector128(long Position, AVec Value) + public void WriteVector128(long Position, Vector128 Value) { - WriteUInt64(Position + 0, Value.X0); - WriteUInt64(Position + 8, Value.X1); + EnsureAccessIsValid(Position + 0, AMemoryPerm.Write); + EnsureAccessIsValid(Position + 15, AMemoryPerm.Write); + + if (Sse.IsSupported) + { + Sse.Store((float*)(RamPtr + (uint)Position), Value); + } + else + { + throw new PlatformNotSupportedException(); + } } public void WriteSByteUnchecked(long Position, sbyte Value) @@ -404,30 +524,68 @@ namespace ChocolArm64.Memory *((ulong*)(RamPtr + (uint)Position)) = Value; } - public void WriteVector8Unchecked(long Position, AVec Value) + public void WriteVector8Unchecked(long Position, Vector128 Value) { - WriteByteUnchecked(Position, Value.B0); + if (Sse41.IsSupported) + { + WriteByteUnchecked(Position, Sse41.Extract(Sse.StaticCast(Value), 0)); + } + else if (Sse2.IsSupported) + { + WriteByteUnchecked(Position, (byte)Sse2.Extract(Sse.StaticCast(Value), 0)); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector16Unchecked(long Position, AVec Value) + public void WriteVector16Unchecked(long Position, Vector128 Value) { - WriteUInt16Unchecked(Position, Value.H0); + if (Sse2.IsSupported) + { + WriteUInt16Unchecked(Position, Sse2.Extract(Sse.StaticCast(Value), 0)); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector32Unchecked(long Position, AVec Value) + public void WriteVector32Unchecked(long Position, Vector128 Value) { - WriteUInt32Unchecked(Position, Value.W0); + if (Sse.IsSupported) + { + Sse.StoreScalar((float*)(RamPtr + (uint)Position), Value); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector64Unchecked(long Position, AVec Value) + public void WriteVector64Unchecked(long Position, Vector128 Value) { - WriteUInt64Unchecked(Position, Value.X0); + if (Sse2.IsSupported) + { + Sse2.StoreScalar((double*)(RamPtr + (uint)Position), Sse.StaticCast(Value)); + } + else + { + throw new PlatformNotSupportedException(); + } } - public void WriteVector128Unchecked(long Position, AVec Value) + public void WriteVector128Unchecked(long Position, Vector128 Value) { - WriteUInt64Unchecked(Position + 0, Value.X0); - WriteUInt64Unchecked(Position + 8, Value.X1); + if (Sse.IsSupported) + { + Sse.Store((float*)(RamPtr + (uint)Position), Value); + } + else + { + throw new PlatformNotSupportedException(); + } } private void EnsureAccessIsValid(long Position, AMemoryPerm Perm) diff --git a/ChocolArm64/State/AThreadState.cs b/ChocolArm64/State/AThreadState.cs index ce1278866..93211813f 100644 --- a/ChocolArm64/State/AThreadState.cs +++ b/ChocolArm64/State/AThreadState.cs @@ -2,6 +2,7 @@ using ChocolArm64.Events; using System; using System.Collections.Generic; using System.Diagnostics; +using System.Runtime.Intrinsics; namespace ChocolArm64.State { @@ -18,10 +19,10 @@ namespace ChocolArm64.State X16, X17, X18, X19, X20, X21, X22, X23, X24, X25, X26, X27, X28, X29, X30, X31; - public AVec V0, V1, V2, V3, V4, V5, V6, V7, - V8, V9, V10, V11, V12, V13, V14, V15, - V16, V17, V18, V19, V20, V21, V22, V23, - V24, V25, V26, V27, V28, V29, V30, V31; + public Vector128 V0, V1, V2, V3, V4, V5, V6, V7, + V8, V9, V10, V11, V12, V13, V14, V15, + V16, V17, V18, V19, V20, V21, V22, V23, + V24, V25, V26, V27, V28, V29, V30, V31; public bool Overflow; public bool Carry; diff --git a/ChocolArm64/State/AVec.cs b/ChocolArm64/State/AVec.cs deleted file mode 100644 index f7eb2e221..000000000 --- a/ChocolArm64/State/AVec.cs +++ /dev/null @@ -1,243 +0,0 @@ -using System; -using System.Runtime.InteropServices; - -namespace ChocolArm64.State -{ - [StructLayout(LayoutKind.Explicit, Size = 16)] - public struct AVec - { - [FieldOffset(0x0)] public byte B0; - [FieldOffset(0x1)] public byte B1; - [FieldOffset(0x2)] public byte B2; - [FieldOffset(0x3)] public byte B3; - [FieldOffset(0x4)] public byte B4; - [FieldOffset(0x5)] public byte B5; - [FieldOffset(0x6)] public byte B6; - [FieldOffset(0x7)] public byte B7; - [FieldOffset(0x8)] public byte B8; - [FieldOffset(0x9)] public byte B9; - [FieldOffset(0xa)] public byte B10; - [FieldOffset(0xb)] public byte B11; - [FieldOffset(0xc)] public byte B12; - [FieldOffset(0xd)] public byte B13; - [FieldOffset(0xe)] public byte B14; - [FieldOffset(0xf)] public byte B15; - - [FieldOffset(0x0)] public ushort H0; - [FieldOffset(0x2)] public ushort H1; - [FieldOffset(0x4)] public ushort H2; - [FieldOffset(0x6)] public ushort H3; - [FieldOffset(0x8)] public ushort H4; - [FieldOffset(0xa)] public ushort H5; - [FieldOffset(0xc)] public ushort H6; - [FieldOffset(0xe)] public ushort H7; - - [FieldOffset(0x0)] public uint W0; - [FieldOffset(0x4)] public uint W1; - [FieldOffset(0x8)] public uint W2; - [FieldOffset(0xc)] public uint W3; - - [FieldOffset(0x0)] public float S0; - [FieldOffset(0x4)] public float S1; - [FieldOffset(0x8)] public float S2; - [FieldOffset(0xc)] public float S3; - - [FieldOffset(0x0)] public ulong X0; - [FieldOffset(0x8)] public ulong X1; - - [FieldOffset(0x0)] public double D0; - [FieldOffset(0x8)] public double D1; - - public byte ExtractByte(int Index) - { - switch (Index) - { - case 0: return B0; - case 1: return B1; - case 2: return B2; - case 3: return B3; - case 4: return B4; - case 5: return B5; - case 6: return B6; - case 7: return B7; - case 8: return B8; - case 9: return B9; - case 10: return B10; - case 11: return B11; - case 12: return B12; - case 13: return B13; - case 14: return B14; - case 15: return B15; - } - - throw new ArgumentOutOfRangeException(nameof(Index)); - } - - public ushort ExtractUInt16(int Index) - { - switch (Index) - { - case 0: return H0; - case 1: return H1; - case 2: return H2; - case 3: return H3; - case 4: return H4; - case 5: return H5; - case 6: return H6; - case 7: return H7; - } - - throw new ArgumentOutOfRangeException(nameof(Index)); - } - - public uint ExtractUInt32(int Index) - { - switch (Index) - { - case 0: return W0; - case 1: return W1; - case 2: return W2; - case 3: return W3; - } - - throw new ArgumentOutOfRangeException(nameof(Index)); - } - - public float ExtractSingle(int Index) - { - switch (Index) - { - case 0: return S0; - case 1: return S1; - case 2: return S2; - case 3: return S3; - } - - throw new ArgumentOutOfRangeException(nameof(Index)); - } - - public ulong ExtractUInt64(int Index) - { - switch (Index) - { - case 0: return X0; - case 1: return X1; - } - - throw new ArgumentOutOfRangeException(nameof(Index)); - } - - public double ExtractDouble(int Index) - { - switch (Index) - { - case 0: return D0; - case 1: return D1; - } - - throw new ArgumentOutOfRangeException(nameof(Index)); - } - - public static AVec InsertByte(AVec Vec, int Index, byte Value) - { - switch (Index) - { - case 0: Vec.B0 = Value; break; - case 1: Vec.B1 = Value; break; - case 2: Vec.B2 = Value; break; - case 3: Vec.B3 = Value; break; - case 4: Vec.B4 = Value; break; - case 5: Vec.B5 = Value; break; - case 6: Vec.B6 = Value; break; - case 7: Vec.B7 = Value; break; - case 8: Vec.B8 = Value; break; - case 9: Vec.B9 = Value; break; - case 10: Vec.B10 = Value; break; - case 11: Vec.B11 = Value; break; - case 12: Vec.B12 = Value; break; - case 13: Vec.B13 = Value; break; - case 14: Vec.B14 = Value; break; - case 15: Vec.B15 = Value; break; - - default: throw new ArgumentOutOfRangeException(nameof(Index)); - } - - return Vec; - } - - public static AVec InsertUInt16(AVec Vec, int Index, ushort Value) - { - switch (Index) - { - case 0: Vec.H0 = Value; break; - case 1: Vec.H1 = Value; break; - case 2: Vec.H2 = Value; break; - case 3: Vec.H3 = Value; break; - case 4: Vec.H4 = Value; break; - case 5: Vec.H5 = Value; break; - case 6: Vec.H6 = Value; break; - case 7: Vec.H7 = Value; break; - - default: throw new ArgumentOutOfRangeException(nameof(Index)); - } - - return Vec; - } - - public static AVec InsertUInt32(AVec Vec, int Index, uint Value) - { - switch (Index) - { - case 0: Vec.W0 = Value; break; - case 1: Vec.W1 = Value; break; - case 2: Vec.W2 = Value; break; - case 3: Vec.W3 = Value; break; - - default: throw new ArgumentOutOfRangeException(nameof(Index)); - } - - return Vec; - } - - public static AVec InsertSingle(AVec Vec, int Index, float Value) - { - switch (Index) - { - case 0: Vec.S0 = Value; break; - case 1: Vec.S1 = Value; break; - case 2: Vec.S2 = Value; break; - case 3: Vec.S3 = Value; break; - - default: throw new ArgumentOutOfRangeException(nameof(Index)); - } - - return Vec; - } - - public static AVec InsertUInt64(AVec Vec, int Index, ulong Value) - { - switch (Index) - { - case 0: Vec.X0 = Value; break; - case 1: Vec.X1 = Value; break; - - default: throw new ArgumentOutOfRangeException(nameof(Index)); - } - - return Vec; - } - - public static AVec InsertDouble(AVec Vec, int Index, double Value) - { - switch (Index) - { - case 0: Vec.D0 = Value; break; - case 1: Vec.D1 = Value; break; - - default: throw new ArgumentOutOfRangeException(nameof(Index)); - } - - return Vec; - } - } -} \ No newline at end of file diff --git a/ChocolArm64/Translation/AILEmitter.cs b/ChocolArm64/Translation/AILEmitter.cs index 55b1751f6..8c7805353 100644 --- a/ChocolArm64/Translation/AILEmitter.cs +++ b/ChocolArm64/Translation/AILEmitter.cs @@ -3,6 +3,7 @@ using ChocolArm64.State; using System; using System.Collections.Generic; using System.Reflection.Emit; +using System.Runtime.Intrinsics; namespace ChocolArm64.Translation { @@ -157,7 +158,7 @@ namespace ChocolArm64.Translation { case ARegisterType.Flag: return typeof(bool); case ARegisterType.Int: return typeof(ulong); - case ARegisterType.Vector: return typeof(AVec); + case ARegisterType.Vector: return typeof(Vector128); } throw new ArgumentException(nameof(RegType)); diff --git a/Ryujinx.Audio/Ryujinx.Audio.csproj b/Ryujinx.Audio/Ryujinx.Audio.csproj index 2e7c86fa0..faeedb908 100644 --- a/Ryujinx.Audio/Ryujinx.Audio.csproj +++ b/Ryujinx.Audio/Ryujinx.Audio.csproj @@ -1,7 +1,7 @@ - netcoreapp2.0 + netcoreapp2.1 diff --git a/Ryujinx.Core/Ryujinx.Core.csproj b/Ryujinx.Core/Ryujinx.Core.csproj index b9374af1d..27e43626d 100644 --- a/Ryujinx.Core/Ryujinx.Core.csproj +++ b/Ryujinx.Core/Ryujinx.Core.csproj @@ -1,7 +1,7 @@ - netcoreapp2.0 + netcoreapp2.1 diff --git a/Ryujinx.Graphics/Ryujinx.Graphics.csproj b/Ryujinx.Graphics/Ryujinx.Graphics.csproj index 01bf41f83..0212ca2ac 100644 --- a/Ryujinx.Graphics/Ryujinx.Graphics.csproj +++ b/Ryujinx.Graphics/Ryujinx.Graphics.csproj @@ -1,7 +1,7 @@ - netcoreapp2.0 + netcoreapp2.1 diff --git a/Ryujinx.Tests/Cpu/CpuTest.cs b/Ryujinx.Tests/Cpu/CpuTest.cs index 67e836f0f..3ff1e4a0d 100644 --- a/Ryujinx.Tests/Cpu/CpuTest.cs +++ b/Ryujinx.Tests/Cpu/CpuTest.cs @@ -4,7 +4,10 @@ using ChocolArm64.State; using NUnit.Framework; +using System; using System.Threading; +using System.Runtime.Intrinsics; +using System.Runtime.Intrinsics.X86; namespace Ryujinx.Tests.Cpu { @@ -54,7 +57,9 @@ namespace Ryujinx.Tests.Cpu } protected void SetThreadState(ulong X0 = 0, ulong X1 = 0, ulong X2 = 0, ulong X3 = 0, ulong X31 = 0, - AVec V0 = default(AVec), AVec V1 = default(AVec), AVec V2 = default(AVec), + Vector128 V0 = default(Vector128), + Vector128 V1 = default(Vector128), + Vector128 V2 = default(Vector128), bool Overflow = false, bool Carry = false, bool Zero = false, bool Negative = false, int Fpcr = 0x0, int Fpsr = 0x0) { @@ -93,7 +98,9 @@ namespace Ryujinx.Tests.Cpu protected AThreadState SingleOpcode(uint Opcode, ulong X0 = 0, ulong X1 = 0, ulong X2 = 0, ulong X3 = 0, ulong X31 = 0, - AVec V0 = default(AVec), AVec V1 = default(AVec), AVec V2 = default(AVec), + Vector128 V0 = default(Vector128), + Vector128 V1 = default(Vector128), + Vector128 V2 = default(Vector128), bool Overflow = false, bool Carry = false, bool Zero = false, bool Negative = false, int Fpcr = 0x0, int Fpsr = 0x0) { @@ -105,5 +112,42 @@ namespace Ryujinx.Tests.Cpu return GetThreadState(); } + + protected static double VectorExtractDouble(Vector128 Vector, byte Index) + { + long Value = Sse41.Extract(Sse.StaticCast(Vector), Index); + + return BitConverter.Int64BitsToDouble(Value); + } + + protected static Vector128 MakeVectorE0(double A) + { + return Sse.StaticCast(Sse2.SetVector128(0, BitConverter.DoubleToInt64Bits(A))); + } + + protected static Vector128 MakeVectorE0(ulong A) + { + return Sse.StaticCast(Sse2.SetVector128(0, A)); + } + + protected static Vector128 MakeVectorE0E1(ulong A, ulong B) + { + return Sse.StaticCast(Sse2.SetVector128(B, A)); + } + + protected static Vector128 MakeVectorE1(ulong B) + { + return Sse.StaticCast(Sse2.SetVector128(B, 0)); + } + + protected static ulong GetVectorE0(Vector128 Vector) + { + return Sse41.Extract(Sse.StaticCast(Vector), 0); + } + + protected static ulong GetVectorE1(Vector128 Vector) + { + return Sse41.Extract(Sse.StaticCast(Vector), 1); + } } } diff --git a/Ryujinx.Tests/Cpu/CpuTestAlu.cs b/Ryujinx.Tests/Cpu/CpuTestAlu.cs index 564fadec2..1efb2a015 100644 --- a/Ryujinx.Tests/Cpu/CpuTestAlu.cs +++ b/Ryujinx.Tests/Cpu/CpuTestAlu.cs @@ -1,14 +1,9 @@ //#define Alu -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("Alu"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestAlu : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestAluImm.cs b/Ryujinx.Tests/Cpu/CpuTestAluImm.cs index 5d1f0b6ba..73f7bff34 100644 --- a/Ryujinx.Tests/Cpu/CpuTestAluImm.cs +++ b/Ryujinx.Tests/Cpu/CpuTestAluImm.cs @@ -1,14 +1,9 @@ //#define AluImm -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("AluImm"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestAluImm : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestAluRs.cs b/Ryujinx.Tests/Cpu/CpuTestAluRs.cs index b81f7100c..2ecaf9973 100644 --- a/Ryujinx.Tests/Cpu/CpuTestAluRs.cs +++ b/Ryujinx.Tests/Cpu/CpuTestAluRs.cs @@ -1,14 +1,9 @@ //#define AluRs -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("AluRs"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestAluRs : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestAluRx.cs b/Ryujinx.Tests/Cpu/CpuTestAluRx.cs index 26169bca6..c60d86c14 100644 --- a/Ryujinx.Tests/Cpu/CpuTestAluRx.cs +++ b/Ryujinx.Tests/Cpu/CpuTestAluRx.cs @@ -1,14 +1,9 @@ //#define AluRx -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("AluRx"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestAluRx : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestBfm.cs b/Ryujinx.Tests/Cpu/CpuTestBfm.cs index 2952bca4c..f6fd02ce6 100644 --- a/Ryujinx.Tests/Cpu/CpuTestBfm.cs +++ b/Ryujinx.Tests/Cpu/CpuTestBfm.cs @@ -1,14 +1,9 @@ //#define Bfm -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("Bfm"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestBfm : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestCcmpImm.cs b/Ryujinx.Tests/Cpu/CpuTestCcmpImm.cs index 38d73878a..915acc3eb 100644 --- a/Ryujinx.Tests/Cpu/CpuTestCcmpImm.cs +++ b/Ryujinx.Tests/Cpu/CpuTestCcmpImm.cs @@ -1,14 +1,9 @@ //#define CcmpImm -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("CcmpImm"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestCcmpImm : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestCcmpReg.cs b/Ryujinx.Tests/Cpu/CpuTestCcmpReg.cs index eb1c3abf2..6bec214b3 100644 --- a/Ryujinx.Tests/Cpu/CpuTestCcmpReg.cs +++ b/Ryujinx.Tests/Cpu/CpuTestCcmpReg.cs @@ -1,14 +1,9 @@ //#define CcmpReg -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("CcmpReg"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestCcmpReg : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestCsel.cs b/Ryujinx.Tests/Cpu/CpuTestCsel.cs index 9dd61957f..93683adcb 100644 --- a/Ryujinx.Tests/Cpu/CpuTestCsel.cs +++ b/Ryujinx.Tests/Cpu/CpuTestCsel.cs @@ -1,14 +1,9 @@ //#define Csel -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("Csel"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestCsel : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestMisc.cs b/Ryujinx.Tests/Cpu/CpuTestMisc.cs index 647d46c00..6db653d3f 100644 --- a/Ryujinx.Tests/Cpu/CpuTestMisc.cs +++ b/Ryujinx.Tests/Cpu/CpuTestMisc.cs @@ -2,6 +2,8 @@ using ChocolArm64.State; using NUnit.Framework; +using System.Runtime.Intrinsics.X86; + namespace Ryujinx.Tests.Cpu { [Category("Misc"), Explicit] @@ -73,7 +75,9 @@ namespace Ryujinx.Tests.Cpu RET */ - SetThreadState(V0: new AVec { S0 = A }, V1: new AVec { S0 = B }); + SetThreadState( + V0: Sse.SetScalarVector128(A), + V1: Sse.SetScalarVector128(B)); Opcode(0x1E2E1002); Opcode(0x1E201840); Opcode(0x1E211841); @@ -84,7 +88,7 @@ namespace Ryujinx.Tests.Cpu Opcode(0xD65F03C0); ExecuteOpcodes(); - Assert.That(GetThreadState().V0.S0, Is.EqualTo(16f)); + Assert.That(Sse41.Extract(GetThreadState().V0, 0), Is.EqualTo(16f)); } [TestCase(-20d, -5d)] // 18 integer solutions. @@ -120,7 +124,9 @@ namespace Ryujinx.Tests.Cpu RET */ - SetThreadState(V0: new AVec { D0 = A }, V1: new AVec { D0 = B }); + SetThreadState( + V0: Sse.StaticCast(Sse2.SetScalarVector128(A)), + V1: Sse.StaticCast(Sse2.SetScalarVector128(B))); Opcode(0x1E6E1002); Opcode(0x1E601840); Opcode(0x1E611841); @@ -131,7 +137,7 @@ namespace Ryujinx.Tests.Cpu Opcode(0xD65F03C0); ExecuteOpcodes(); - Assert.That(GetThreadState().V0.D0, Is.EqualTo(16d)); + Assert.That(VectorExtractDouble(GetThreadState().V0, 0), Is.EqualTo(16d)); } [Test] diff --git a/Ryujinx.Tests/Cpu/CpuTestMov.cs b/Ryujinx.Tests/Cpu/CpuTestMov.cs index 9c7e3255a..bd753b213 100644 --- a/Ryujinx.Tests/Cpu/CpuTestMov.cs +++ b/Ryujinx.Tests/Cpu/CpuTestMov.cs @@ -1,14 +1,9 @@ //#define Mov -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("Mov"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestMov : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestMul.cs b/Ryujinx.Tests/Cpu/CpuTestMul.cs index 9bdc1fa65..23aab0370 100644 --- a/Ryujinx.Tests/Cpu/CpuTestMul.cs +++ b/Ryujinx.Tests/Cpu/CpuTestMul.cs @@ -1,14 +1,9 @@ //#define Mul -using ChocolArm64.State; - using NUnit.Framework; namespace Ryujinx.Tests.Cpu { - using Tester; - using Tester.Types; - [Category("Mul"), Ignore("Tested: first half of 2018.")] public sealed class CpuTestMul : CpuTest { diff --git a/Ryujinx.Tests/Cpu/CpuTestScalar.cs b/Ryujinx.Tests/Cpu/CpuTestScalar.cs index c7c3aa051..d1a0c3131 100644 --- a/Ryujinx.Tests/Cpu/CpuTestScalar.cs +++ b/Ryujinx.Tests/Cpu/CpuTestScalar.cs @@ -1,6 +1,9 @@ using ChocolArm64.State; + using NUnit.Framework; +using System.Runtime.Intrinsics.X86; + namespace Ryujinx.Tests.Cpu { public class CpuTestScalar : CpuTest @@ -25,8 +28,10 @@ namespace Ryujinx.Tests.Cpu public void Fmax_S(uint Opcode, ulong A, ulong B, ulong Result) { // FMAX S0, S1, S2 - AThreadState ThreadState = SingleOpcode(Opcode, V1: new AVec { X0 = A }, V2: new AVec { X0 = B }); - Assert.AreEqual(Result, ThreadState.V0.X0); + AThreadState ThreadState = SingleOpcode(Opcode, + V1: Sse.StaticCast(Sse2.SetVector128(0, A)), + V2: Sse.StaticCast(Sse2.SetVector128(0, B))); + Assert.AreEqual(Result, Sse41.Extract(Sse.StaticCast(ThreadState.V0), 0)); } [TestCase(0x1E225820u, 0x0000000000000000ul, 0x0000000080000000ul, 0x0000000080000000ul)] @@ -49,8 +54,10 @@ namespace Ryujinx.Tests.Cpu public void Fmin_S(uint Opcode, ulong A, ulong B, ulong Result) { // FMIN S0, S1, S2 - AThreadState ThreadState = SingleOpcode(Opcode, V1: new AVec { X0 = A }, V2: new AVec { X0 = B }); - Assert.AreEqual(Result, ThreadState.V0.X0); + AThreadState ThreadState = SingleOpcode(Opcode, + V1: Sse.StaticCast(Sse2.SetVector128(0, A)), + V2: Sse.StaticCast(Sse2.SetVector128(0, B))); + Assert.AreEqual(Result, Sse41.Extract(Sse.StaticCast(ThreadState.V0), 0)); } } } diff --git a/Ryujinx.Tests/Cpu/CpuTestSimd.cs b/Ryujinx.Tests/Cpu/CpuTestSimd.cs index 95fca86a3..e2adc9955 100644 --- a/Ryujinx.Tests/Cpu/CpuTestSimd.cs +++ b/Ryujinx.Tests/Cpu/CpuTestSimd.cs @@ -4,6 +4,8 @@ using ChocolArm64.State; using NUnit.Framework; +using System.Runtime.Intrinsics; + namespace Ryujinx.Tests.Cpu { using Tester; @@ -74,8 +76,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x5EE0B820; // ABS D0, D1 Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.V(1, new Bits(A)); @@ -83,8 +85,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -96,8 +98,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.V(1, new Bits(A)); @@ -105,8 +107,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -119,7 +121,7 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -128,8 +130,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -140,8 +142,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x5EF1B820; // ADDP D0, V1.2D Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -150,8 +152,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -163,9 +165,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = TestContext.CurrentContext.Random.NextULong(), - X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE0E1(TestContext.CurrentContext.Random.NextULong(), + TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(0, 0, new Bits(TestContext.CurrentContext.Random.NextULong())); @@ -174,8 +176,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -188,9 +190,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = TestContext.CurrentContext.Random.NextULong(), - X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V0 = MakeVectorE0E1(TestContext.CurrentContext.Random.NextULong(), + TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(0, 0, new Bits(TestContext.CurrentContext.Random.NextULong())); @@ -200,8 +202,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -213,8 +215,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.V(1, new Bits(A)); @@ -222,8 +224,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -236,7 +238,7 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -245,8 +247,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -258,8 +260,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.V(1, new Bits(A)); @@ -267,8 +269,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -281,7 +283,7 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -290,8 +292,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -301,8 +303,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x7EE0B820; // NEG D0, D1 Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.V(1, new Bits(A)); @@ -310,8 +312,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -323,8 +325,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.V(1, new Bits(A)); @@ -332,8 +334,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -346,7 +348,7 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -355,8 +357,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -368,9 +370,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = TestContext.CurrentContext.Random.NextULong(), - X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE0E1(TestContext.CurrentContext.Random.NextULong(), + TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(0, 0, new Bits(TestContext.CurrentContext.Random.NextULong())); @@ -379,8 +381,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27])); // FIXME: Temporary solution. } @@ -394,8 +396,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -404,8 +406,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27])); // FIXME: Temporary solution. } @@ -420,8 +422,8 @@ namespace Ryujinx.Tests.Cpu Bits Op = new Bits(Opcode); ulong _X0 = TestContext.CurrentContext.Random.NextULong(); - AVec V0 = new AVec { X0 = _X0 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V0 = MakeVectorE0(_X0); + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -430,8 +432,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(_X0)); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(_X0)); + Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); }); Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27])); // FIXME: Temporary solution. } @@ -444,9 +446,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = TestContext.CurrentContext.Random.NextULong(), - X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; + Vector128 V0 = MakeVectorE0E1(TestContext.CurrentContext.Random.NextULong(), + TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(0, 0, new Bits(TestContext.CurrentContext.Random.NextULong())); @@ -455,8 +457,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27])); // FIXME: Temporary solution. } @@ -470,8 +472,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -480,8 +482,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27])); // FIXME: Temporary solution. } @@ -496,8 +498,8 @@ namespace Ryujinx.Tests.Cpu Bits Op = new Bits(Opcode); ulong _X0 = TestContext.CurrentContext.Random.NextULong(); - AVec V0 = new AVec { X0 = _X0 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; + Vector128 V0 = MakeVectorE0(_X0); + Vector128 V1 = MakeVectorE0E1(A0, A1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1); AArch64.Vpart(1, 0, new Bits(A0)); @@ -506,8 +508,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(_X0)); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(_X0)); + Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); }); Assert.That(((ThreadState.Fpsr >> 27) & 1) != 0, Is.EqualTo(Shared.FPSR[27])); // FIXME: Temporary solution. } diff --git a/Ryujinx.Tests/Cpu/CpuTestSimdArithmetic.cs b/Ryujinx.Tests/Cpu/CpuTestSimdArithmetic.cs index f32fe398d..f4982c1b2 100644 --- a/Ryujinx.Tests/Cpu/CpuTestSimdArithmetic.cs +++ b/Ryujinx.Tests/Cpu/CpuTestSimdArithmetic.cs @@ -1,6 +1,10 @@ using ChocolArm64.State; + using NUnit.Framework; +using System.Runtime.Intrinsics; +using System.Runtime.Intrinsics.X86; + namespace Ryujinx.Tests.Cpu { public class CpuTestSimdArithmetic : CpuTest @@ -35,13 +39,13 @@ namespace Ryujinx.Tests.Cpu [TestCase(0x4EE28420u, 0x0102030405060708ul, 0xAAAAAAAAAAAAAAAAul, 0x0807060504030201ul, 0x2222222222222222ul, 0x0909090909090909ul, 0xCCCCCCCCCCCCCCCCul)] public void Add_V(uint Opcode, ulong A0, ulong A1, ulong B0, ulong B1, ulong Result0, ulong Result1) { - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -61,13 +65,13 @@ namespace Ryujinx.Tests.Cpu public void Fmax_V(uint A, uint B, uint C, uint D, uint Result0, uint Result1) { uint Opcode = 0x4E22F420; - AVec V1 = new AVec { X0 = A, X1 = B }; - AVec V2 = new AVec { X0 = C, X1 = D }; + Vector128 V1 = MakeVectorE0E1(A, B); + Vector128 V2 = MakeVectorE0E1(C, D); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -87,63 +91,68 @@ namespace Ryujinx.Tests.Cpu public void Fmin_V(uint A, uint B, uint C, uint D, uint Result0, uint Result1) { uint Opcode = 0x4EA2F420; - AVec V1 = new AVec { X0 = A, X1 = B }; - AVec V2 = new AVec { X0 = C, X1 = D }; + Vector128 V1 = MakeVectorE0E1(A, B); + Vector128 V2 = MakeVectorE0E1(C, D); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } [Test, Description("fmul s6, s1, v0.s[2]")] public void Fmul_Se([Random(10)] float A, [Random(10)] float B) { - AThreadState ThreadState = SingleOpcode(0x5F809826, V1: new AVec { S0 = A }, V0: new AVec { S2 = B }); + AThreadState ThreadState = SingleOpcode(0x5F809826, + V1: Sse.SetVector128(0, 0, 0, A), + V0: Sse.SetVector128(0, B, 0, 0)); - Assert.That(ThreadState.V6.S0, Is.EqualTo(A * B)); + Assert.That(Sse41.Extract(ThreadState.V6, (byte)0), Is.EqualTo(A * B)); } [Test, Description("frecpe v2.4s, v0.4s")] public void Frecpe_V([Random(100)] float A) { - AThreadState ThreadState = SingleOpcode(0x4EA1D802, V0: new AVec { S0 = A, S1 = A, S2 = A, S3 = A }); + AThreadState ThreadState = SingleOpcode(0x4EA1D802, V0: Sse.SetAllVector128(A)); - Assert.That(ThreadState.V2.S0, Is.EqualTo(1 / A)); - Assert.That(ThreadState.V2.S1, Is.EqualTo(1 / A)); - Assert.That(ThreadState.V2.S2, Is.EqualTo(1 / A)); - Assert.That(ThreadState.V2.S3, Is.EqualTo(1 / A)); + Assert.That(Sse41.Extract(ThreadState.V2, (byte)0), Is.EqualTo(1 / A)); + Assert.That(Sse41.Extract(ThreadState.V2, (byte)1), Is.EqualTo(1 / A)); + Assert.That(Sse41.Extract(ThreadState.V2, (byte)2), Is.EqualTo(1 / A)); + Assert.That(Sse41.Extract(ThreadState.V2, (byte)3), Is.EqualTo(1 / A)); } [Test, Description("frecpe d0, d1")] public void Frecpe_S([Random(100)] double A) { - AThreadState ThreadState = SingleOpcode(0x5EE1D820, V1: new AVec { D0 = A }); + AThreadState ThreadState = SingleOpcode(0x5EE1D820, V1: MakeVectorE0(A)); - Assert.That(ThreadState.V0.D0, Is.EqualTo(1 / A)); + Assert.That(VectorExtractDouble(ThreadState.V0, 0), Is.EqualTo(1 / A)); } [Test, Description("frecps v4.4s, v2.4s, v0.4s")] public void Frecps_V([Random(10)] float A, [Random(10)] float B) { - AThreadState ThreadState = SingleOpcode(0x4E20FC44, V2: new AVec { S0 = A, S1 = A, S2 = A, S3 = A }, - V0: new AVec { S0 = B, S1 = B, S2 = B, S3 = B }); + AThreadState ThreadState = SingleOpcode(0x4E20FC44, + V2: Sse.SetAllVector128(A), + V0: Sse.SetAllVector128(B)); - Assert.That(ThreadState.V4.S0, Is.EqualTo(2 - (A * B))); - Assert.That(ThreadState.V4.S1, Is.EqualTo(2 - (A * B))); - Assert.That(ThreadState.V4.S2, Is.EqualTo(2 - (A * B))); - Assert.That(ThreadState.V4.S3, Is.EqualTo(2 - (A * B))); + Assert.That(Sse41.Extract(ThreadState.V4, (byte)0), Is.EqualTo(2 - (A * B))); + Assert.That(Sse41.Extract(ThreadState.V4, (byte)1), Is.EqualTo(2 - (A * B))); + Assert.That(Sse41.Extract(ThreadState.V4, (byte)2), Is.EqualTo(2 - (A * B))); + Assert.That(Sse41.Extract(ThreadState.V4, (byte)3), Is.EqualTo(2 - (A * B))); } [Test, Description("frecps d0, d1, d2")] public void Frecps_S([Random(10)] double A, [Random(10)] double B) { - AThreadState ThreadState = SingleOpcode(0x5E62FC20, V1: new AVec { D0 = A }, V2: new AVec { D0 = B }); + AThreadState ThreadState = SingleOpcode(0x5E62FC20, + V1: MakeVectorE0(A), + V2: MakeVectorE0(B)); - Assert.That(ThreadState.V0.D0, Is.EqualTo(2 - (A * B))); + Assert.That(VectorExtractDouble(ThreadState.V0, 0), Is.EqualTo(2 - (A * B))); } - + [TestCase(0x3FE66666u, false, 0x40000000u)] [TestCase(0x3F99999Au, false, 0x3F800000u)] [TestCase(0x404CCCCDu, false, 0x40400000u)] @@ -189,17 +198,17 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x1E264020, V1: V1, Fpcr: FpcrTemp); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } [TestCase(0x6E618820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x6E618820u, 0x3FFCCCCCCCCCCCCDul, 0x3FFCCCCCCCCCCCCDul, false, 0x4000000000000000ul, 0x4000000000000000ul)] [TestCase(0x6E618820u, 0x3FF8000000000000ul, 0x3FF8000000000000ul, false, 0x4000000000000000ul, 0x4000000000000000ul)] [TestCase(0x6E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f80000040000000ul, 0x3f80000040000000ul)] - [TestCase(0x6E219820u, 0x3fc000003fc00000ul, 0x3fc000003fc00000ul, false, 0x4000000040000000ul, 0x4000000040000000ul)] - [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f80000040000000ul, 0x0000000000000000ul)] + [TestCase(0x6E219820u, 0x3fc000003fc00000ul, 0x3fc000003fc00000ul, false, 0x4000000040000000ul, 0x4000000040000000ul)] + [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f80000040000000ul, 0x0000000000000000ul)] [TestCase(0x2E219820u, 0x3fc000003fc00000ul, 0x3fc000003fc00000ul, false, 0x4000000040000000ul, 0x0000000000000000ul)] [TestCase(0x2E218820u, 0x0000000080000000ul, 0x0000000000000000ul, false, 0x0000000080000000ul, 0x0000000000000000ul)] [TestCase(0x2E218820u, 0x7F800000FF800000ul, 0x0000000000000000ul, false, 0x7F800000FF800000ul, 0x0000000000000000ul)] @@ -212,12 +221,12 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A, X1 = B }; + Vector128 V1 = MakeVectorE0E1(A, B); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, Fpcr: FpcrTemp); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -286,9 +295,9 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp |= 1 << 25; } - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x1E27C020, V1: V1, Fpcr: FpcrTemp); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } [TestCase(0x6EE19820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, 'N', false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] @@ -300,11 +309,11 @@ namespace Ryujinx.Tests.Cpu [TestCase(0x6EE19820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, 'Z', false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x6EE19820u, 0x3FFCCCCCCCCCCCCDul, 0x3FFCCCCCCCCCCCCDul, 'Z', false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x6EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'N', false, 0x3f80000040000000ul, 0x3f80000040000000ul)] - [TestCase(0x6EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x4000000040000000ul)] + [TestCase(0x6EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x4000000040000000ul)] [TestCase(0x6EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'M', false, 0x3f8000003f800000ul, 0x3f8000003f800000ul)] [TestCase(0x6EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'Z', false, 0x3f8000003f800000ul, 0x3f8000003f800000ul)] [TestCase(0x2EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'N', false, 0x3f80000040000000ul, 0x0000000000000000ul)] - [TestCase(0x2EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x0000000000000000ul)] + [TestCase(0x2EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x0000000000000000ul)] [TestCase(0x2EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'M', false, 0x3f8000003f800000ul, 0x0000000000000000ul)] [TestCase(0x2EA19820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'Z', false, 0x3f8000003f800000ul, 0x0000000000000000ul)] [TestCase(0x2EA19820u, 0x0000000080000000ul, 0x0000000000000000ul, 'N', false, 0x0000000080000000ul, 0x0000000000000000ul)] @@ -348,12 +357,12 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp |= 1 << 25; } - AVec V1 = new AVec { X0 = A, X1 = B }; + Vector128 V1 = MakeVectorE0E1(A, B); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, Fpcr: FpcrTemp); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -402,15 +411,15 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x1E254020, V1: V1, Fpcr: FpcrTemp); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } [TestCase(0x4E619820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x4E619820u, 0x3FFCCCCCCCCCCCCDul, 0x3FFCCCCCCCCCCCCDul, false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] - [TestCase(0x4E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f8000003f800000ul, 0x3f8000003f800000ul)] - [TestCase(0xE219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f8000003f800000ul, 0x0000000000000000ul)] + [TestCase(0x4E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f8000003f800000ul, 0x3f8000003f800000ul)] + [TestCase(0xE219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f8000003f800000ul, 0x0000000000000000ul)] [TestCase(0xE219820u, 0x0000000080000000ul, 0x0000000000000000ul, false, 0x0000000080000000ul, 0x0000000000000000ul)] [TestCase(0xE219820u, 0x7F800000FF800000ul, 0x0000000000000000ul, false, 0x7F800000FF800000ul, 0x0000000000000000ul)] [TestCase(0xE219820u, 0xFF8000017FC00002ul, 0x0000000000000000ul, false, 0xFFC000017FC00002ul, 0x0000000000000000ul, Ignore = "NaN test.")] @@ -422,12 +431,12 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A, X1 = B }; + Vector128 V1 = MakeVectorE0E1(A, B); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, Fpcr: FpcrTemp); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -476,17 +485,17 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x1E264020, V1: V1, Fpcr: FpcrTemp); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } [TestCase(0x4E618820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x4E618820u, 0x3FFCCCCCCCCCCCCDul, 0x3FFCCCCCCCCCCCCDul, false, 0x4000000000000000ul, 0x4000000000000000ul)] [TestCase(0x4E618820u, 0x3FF8000000000000ul, 0x3FF8000000000000ul, false, 0x4000000000000000ul, 0x4000000000000000ul)] [TestCase(0x4E218820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f80000040000000ul, 0x3f80000040000000ul)] - [TestCase(0x4E218820u, 0x3fc000003fc00000ul, 0x3fc000003fc00000ul, false, 0x4000000040000000ul, 0x4000000040000000ul)] - [TestCase(0xE218820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f80000040000000ul, 0x0000000000000000ul)] + [TestCase(0x4E218820u, 0x3fc000003fc00000ul, 0x3fc000003fc00000ul, false, 0x4000000040000000ul, 0x4000000040000000ul)] + [TestCase(0xE218820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x3f80000040000000ul, 0x0000000000000000ul)] [TestCase(0xE218820u, 0x3fc000003fc00000ul, 0x3fc000003fc00000ul, false, 0x4000000040000000ul, 0x0000000000000000ul)] [TestCase(0xE218820u, 0x0000000080000000ul, 0x0000000000000000ul, false, 0x0000000080000000ul, 0x0000000000000000ul)] [TestCase(0xE218820u, 0x7F800000FF800000ul, 0x0000000000000000ul, false, 0x7F800000FF800000ul, 0x0000000000000000ul)] @@ -499,12 +508,12 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A, X1 = B }; + Vector128 V1 = MakeVectorE0E1(A, B); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, Fpcr: FpcrTemp); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -553,15 +562,15 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x1E24C020, V1: V1, Fpcr: FpcrTemp); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } [TestCase(0x4EE18820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, false, 0x4000000000000000ul, 0x4000000000000000ul)] [TestCase(0x4EE18820u, 0x3FFCCCCCCCCCCCCDul, 0x3FFCCCCCCCCCCCCDul, false, 0x4000000000000000ul, 0x4000000000000000ul)] - [TestCase(0x4EA18820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x4000000040000000ul, 0x4000000040000000ul)] - [TestCase(0xEA18820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x4000000040000000ul, 0x0000000000000000ul)] + [TestCase(0x4EA18820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x4000000040000000ul, 0x4000000040000000ul)] + [TestCase(0xEA18820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, false, 0x4000000040000000ul, 0x0000000000000000ul)] [TestCase(0xEA18820u, 0x0000000080000000ul, 0x0000000000000000ul, false, 0x0000000080000000ul, 0x0000000000000000ul)] [TestCase(0xEA18820u, 0x7F800000FF800000ul, 0x0000000000000000ul, false, 0x7F800000FF800000ul, 0x0000000000000000ul)] [TestCase(0xEA18820u, 0xFF8000017FC00002ul, 0x0000000000000000ul, false, 0xFFC000017FC00002ul, 0x0000000000000000ul, Ignore = "NaN test.")] @@ -573,12 +582,12 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp = 0x2000000; } - AVec V1 = new AVec { X0 = A, X1 = B }; + Vector128 V1 = MakeVectorE0E1(A, B); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, Fpcr: FpcrTemp); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } @@ -647,9 +656,9 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp |= 1 << 25; } - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x1E274020, V1: V1, Fpcr: FpcrTemp); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } [TestCase(0x6E619820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, 'N', false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] @@ -661,11 +670,11 @@ namespace Ryujinx.Tests.Cpu [TestCase(0x6E619820u, 0x3FF3333333333333ul, 0x3FF3333333333333ul, 'Z', false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x6E619820u, 0x3FFCCCCCCCCCCCCDul, 0x3FFCCCCCCCCCCCCDul, 'Z', false, 0x3FF0000000000000ul, 0x3FF0000000000000ul)] [TestCase(0x6E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'N', false, 0x3f80000040000000ul, 0x3f80000040000000ul)] - [TestCase(0x6E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x4000000040000000ul)] + [TestCase(0x6E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x4000000040000000ul)] [TestCase(0x6E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'M', false, 0x3f8000003f800000ul, 0x3f8000003f800000ul)] [TestCase(0x6E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'Z', false, 0x3f8000003f800000ul, 0x3f8000003f800000ul)] [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'N', false, 0x3f80000040000000ul, 0x0000000000000000ul)] - [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x0000000000000000ul)] + [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'P', false, 0x4000000040000000ul, 0x0000000000000000ul)] [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'M', false, 0x3f8000003f800000ul, 0x0000000000000000ul)] [TestCase(0x2E219820u, 0x3f99999a3fe66666ul, 0x3f99999a3fe66666ul, 'Z', false, 0x3f8000003f800000ul, 0x0000000000000000ul)] [TestCase(0x2E219820u, 0x0000000080000000ul, 0x0000000000000000ul, 'N', false, 0x0000000080000000ul, 0x0000000000000000ul)] @@ -709,21 +718,21 @@ namespace Ryujinx.Tests.Cpu { FpcrTemp |= 1 << 25; } - AVec V1 = new AVec { X0 = A, X1 = B }; + Vector128 V1 = MakeVectorE0E1(A, B); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, Fpcr: FpcrTemp); Assert.Multiple(() => { - Assert.AreEqual(Result0, ThreadState.V0.X0); - Assert.AreEqual(Result1, ThreadState.V0.X1); + Assert.AreEqual(Result0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result1, GetVectorE1(ThreadState.V0)); }); } [TestCase(0x41200000u, 0x3EA18000u)] public void Frsqrte_S(uint A, uint Result) { - AVec V1 = new AVec { X0 = A }; + Vector128 V1 = MakeVectorE0(A); AThreadState ThreadState = SingleOpcode(0x7EA1D820, V1: V1); - Assert.AreEqual(Result, ThreadState.V0.X0); + Assert.AreEqual(Result, GetVectorE0(ThreadState.V0)); } } } diff --git a/Ryujinx.Tests/Cpu/CpuTestSimdMove.cs b/Ryujinx.Tests/Cpu/CpuTestSimdMove.cs index 0681b6139..e2766a166 100644 --- a/Ryujinx.Tests/Cpu/CpuTestSimdMove.cs +++ b/Ryujinx.Tests/Cpu/CpuTestSimdMove.cs @@ -1,6 +1,10 @@ using ChocolArm64.State; + using NUnit.Framework; +using System.Runtime.Intrinsics; +using System.Runtime.Intrinsics.X86; + namespace Ryujinx.Tests.Cpu { public class CpuTestSimdMove : CpuTest @@ -10,15 +14,17 @@ namespace Ryujinx.Tests.Cpu [Random(2)] uint B0, [Random(2)] uint B1, [Random(2)] uint B2, [Random(2)] uint B3) { uint Opcode = 0x4E822820; - AVec V1 = new AVec { W0 = A0, W1 = A1, W2 = A2, W3 = A3 }; - AVec V2 = new AVec { W0 = B0, W1 = B1, W2 = B2, W3 = B3 }; + Vector128 V1 = Sse.StaticCast(Sse2.SetVector128(A3, A2, A1, A0)); + Vector128 V2 = Sse.StaticCast(Sse2.SetVector128(B3, B2, B1, B0)); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); - Assert.That(ThreadState.V0.W0, Is.EqualTo(A0)); - Assert.That(ThreadState.V0.W1, Is.EqualTo(B0)); - Assert.That(ThreadState.V0.W2, Is.EqualTo(A2)); - Assert.That(ThreadState.V0.W3, Is.EqualTo(B2)); + Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)0); + + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)0), Is.EqualTo(A0)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)1), Is.EqualTo(B0)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)2), Is.EqualTo(A2)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)3), Is.EqualTo(B2)); } [Test, Description("trn1 v0.8b, v1.8b, v2.8b")] @@ -28,19 +34,19 @@ namespace Ryujinx.Tests.Cpu [Random(2)] byte B4, [Random(1)] byte B5, [Random(2)] byte B6, [Random(1)] byte B7) { uint Opcode = 0x0E022820; - AVec V1 = new AVec { B0 = A0, B1 = A1, B2 = A2, B3 = A3, B4 = A4, B5 = A5, B6 = A6, B7 = A7 }; - AVec V2 = new AVec { B0 = B0, B1 = B1, B2 = B2, B3 = B3, B4 = B4, B5 = B5, B6 = B6, B7 = B7 }; + Vector128 V1 = Sse.StaticCast(Sse2.SetVector128(0, 0, 0, 0, 0, 0, 0, 0, A7, A6, A5, A4, A3, A2, A1, A0)); + Vector128 V2 = Sse.StaticCast(Sse2.SetVector128(0, 0, 0, 0, 0, 0, 0, 0, B7, B6, B5, B4, B3, B2, B1, B0)); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); - Assert.That(ThreadState.V0.B0, Is.EqualTo(A0)); - Assert.That(ThreadState.V0.B1, Is.EqualTo(B0)); - Assert.That(ThreadState.V0.B2, Is.EqualTo(A2)); - Assert.That(ThreadState.V0.B3, Is.EqualTo(B2)); - Assert.That(ThreadState.V0.B4, Is.EqualTo(A4)); - Assert.That(ThreadState.V0.B5, Is.EqualTo(B4)); - Assert.That(ThreadState.V0.B6, Is.EqualTo(A6)); - Assert.That(ThreadState.V0.B7, Is.EqualTo(B6)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)0), Is.EqualTo(A0)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)1), Is.EqualTo(B0)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)2), Is.EqualTo(A2)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)3), Is.EqualTo(B2)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)4), Is.EqualTo(A4)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)5), Is.EqualTo(B4)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)6), Is.EqualTo(A6)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)7), Is.EqualTo(B6)); } [Test, Description("trn2 v0.4s, v1.4s, v2.4s")] @@ -48,15 +54,15 @@ namespace Ryujinx.Tests.Cpu [Random(2)] uint B0, [Random(2)] uint B1, [Random(2)] uint B2, [Random(2)] uint B3) { uint Opcode = 0x4E826820; - AVec V1 = new AVec { W0 = A0, W1 = A1, W2 = A2, W3 = A3 }; - AVec V2 = new AVec { W0 = B0, W1 = B1, W2 = B2, W3 = B3 }; + Vector128 V1 = Sse.StaticCast(Sse2.SetVector128(A3, A2, A1, A0)); + Vector128 V2 = Sse.StaticCast(Sse2.SetVector128(B3, B2, B1, B0)); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); - Assert.That(ThreadState.V0.W0, Is.EqualTo(A1)); - Assert.That(ThreadState.V0.W1, Is.EqualTo(B1)); - Assert.That(ThreadState.V0.W2, Is.EqualTo(A3)); - Assert.That(ThreadState.V0.W3, Is.EqualTo(B3)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)0), Is.EqualTo(A1)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)1), Is.EqualTo(B1)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)2), Is.EqualTo(A3)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)3), Is.EqualTo(B3)); } [Test, Description("trn2 v0.8b, v1.8b, v2.8b")] @@ -66,19 +72,19 @@ namespace Ryujinx.Tests.Cpu [Random(1)] byte B4, [Random(2)] byte B5, [Random(1)] byte B6, [Random(2)] byte B7) { uint Opcode = 0x0E026820; - AVec V1 = new AVec { B0 = A0, B1 = A1, B2 = A2, B3 = A3, B4 = A4, B5 = A5, B6 = A6, B7 = A7 }; - AVec V2 = new AVec { B0 = B0, B1 = B1, B2 = B2, B3 = B3, B4 = B4, B5 = B5, B6 = B6, B7 = B7 }; + Vector128 V1 = Sse.StaticCast(Sse2.SetVector128(0, 0, 0, 0, 0, 0, 0, 0, A7, A6, A5, A4, A3, A2, A1, A0)); + Vector128 V2 = Sse.StaticCast(Sse2.SetVector128(0, 0, 0, 0, 0, 0, 0, 0, B7, B6, B5, B4, B3, B2, B1, B0)); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); - Assert.That(ThreadState.V0.B0, Is.EqualTo(A1)); - Assert.That(ThreadState.V0.B1, Is.EqualTo(B1)); - Assert.That(ThreadState.V0.B2, Is.EqualTo(A3)); - Assert.That(ThreadState.V0.B3, Is.EqualTo(B3)); - Assert.That(ThreadState.V0.B4, Is.EqualTo(A5)); - Assert.That(ThreadState.V0.B5, Is.EqualTo(B5)); - Assert.That(ThreadState.V0.B6, Is.EqualTo(A7)); - Assert.That(ThreadState.V0.B7, Is.EqualTo(B7)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)0), Is.EqualTo(A1)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)1), Is.EqualTo(B1)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)2), Is.EqualTo(A3)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)3), Is.EqualTo(B3)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)4), Is.EqualTo(A5)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)5), Is.EqualTo(B5)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)6), Is.EqualTo(A7)); + Assert.That(Sse41.Extract(Sse.StaticCast(ThreadState.V0), (byte)7), Is.EqualTo(B7)); } [TestCase(0u, 0u, 0x2313221221112010ul, 0x0000000000000000ul)] @@ -92,11 +98,11 @@ namespace Ryujinx.Tests.Cpu { // ZIP1 V0., V1., V2. uint Opcode = 0x0E023820 | (Q << 30) | (size << 22); - AVec V1 = new AVec { X0 = 0x1716151413121110, X1 = 0x1F1E1D1C1B1A1918 }; - AVec V2 = new AVec { X0 = 0x2726252423222120, X1 = 0x2F2E2D2C2B2A2928 }; + Vector128 V1 = MakeVectorE0E1(0x1716151413121110, 0x1F1E1D1C1B1A1918); + Vector128 V2 = MakeVectorE0E1(0x2726252423222120, 0x2F2E2D2C2B2A2928); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); - Assert.AreEqual(Result_0, ThreadState.V0.X0); - Assert.AreEqual(Result_1, ThreadState.V0.X1); + Assert.AreEqual(Result_0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result_1, GetVectorE1(ThreadState.V0)); } [TestCase(0u, 0u, 0x2717261625152414ul, 0x0000000000000000ul)] @@ -110,11 +116,11 @@ namespace Ryujinx.Tests.Cpu { // ZIP2 V0., V1., V2. uint Opcode = 0x0E027820 | (Q << 30) | (size << 22); - AVec V1 = new AVec { X0 = 0x1716151413121110, X1 = 0x1F1E1D1C1B1A1918 }; - AVec V2 = new AVec { X0 = 0x2726252423222120, X1 = 0x2F2E2D2C2B2A2928 }; + Vector128 V1 = MakeVectorE0E1(0x1716151413121110, 0x1F1E1D1C1B1A1918); + Vector128 V2 = MakeVectorE0E1(0x2726252423222120, 0x2F2E2D2C2B2A2928); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); - Assert.AreEqual(Result_0, ThreadState.V0.X0); - Assert.AreEqual(Result_1, ThreadState.V0.X1); + Assert.AreEqual(Result_0, GetVectorE0(ThreadState.V0)); + Assert.AreEqual(Result_1, GetVectorE1(ThreadState.V0)); } } } diff --git a/Ryujinx.Tests/Cpu/CpuTestSimdReg.cs b/Ryujinx.Tests/Cpu/CpuTestSimdReg.cs index e9d53eb76..88aebfbe7 100644 --- a/Ryujinx.Tests/Cpu/CpuTestSimdReg.cs +++ b/Ryujinx.Tests/Cpu/CpuTestSimdReg.cs @@ -4,6 +4,8 @@ using ChocolArm64.State; using NUnit.Framework; +using System.Runtime.Intrinsics; + namespace Ryujinx.Tests.Cpu { using Tester; @@ -65,9 +67,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x5EE28420; // ADD D0, D1, D2 Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -76,8 +78,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -90,9 +92,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -101,8 +103,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -117,8 +119,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -129,8 +131,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -145,9 +147,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -158,8 +160,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -175,9 +177,9 @@ namespace Ryujinx.Tests.Cpu Bits Op = new Bits(Opcode); ulong _X0 = TestContext.CurrentContext.Random.NextULong(); - AVec V0 = new AVec { X0 = _X0 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0(_X0); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -188,8 +190,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(_X0)); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(_X0)); + Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); }); } @@ -202,9 +204,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -213,8 +215,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -229,8 +231,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -241,8 +243,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -253,9 +255,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x0E221C20; // AND V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -264,8 +266,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -278,8 +280,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x4E221C20; // AND V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -290,8 +292,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -302,9 +304,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x0E621C20; // BIC V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -313,8 +315,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -327,8 +329,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x4E621C20; // BIC V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -339,8 +341,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -352,9 +354,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x2EE21C20; // BIF V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = _Z, X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE0E1(_Z, TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(0, 0, new Bits(_Z)); @@ -364,8 +366,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -380,9 +382,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x6EE21C20; // BIF V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = _Z0, X1 = _Z1 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0E1(_Z0, _Z1); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(0, 0, new Bits(_Z0)); @@ -395,8 +397,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -408,9 +410,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x2EA21C20; // BIT V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = _Z, X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE0E1(_Z, TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(0, 0, new Bits(_Z)); @@ -420,8 +422,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -436,9 +438,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x6EA21C20; // BIT V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = _Z0, X1 = _Z1 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0E1(_Z0, _Z1); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(0, 0, new Bits(_Z0)); @@ -451,8 +453,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -464,9 +466,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x2E621C20; // BSL V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = _Z, X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE0E1(_Z, TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(0, 0, new Bits(_Z)); @@ -476,8 +478,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -492,9 +494,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x6E621C20; // BSL V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X0 = _Z0, X1 = _Z1 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0E1(_Z0, _Z1); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(0, 0, new Bits(_Z0)); @@ -507,8 +509,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -519,9 +521,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x0EE21C20; // ORN V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -530,8 +532,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -544,8 +546,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x4EE21C20; // ORN V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -556,8 +558,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -568,9 +570,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x0EA21C20; // ORR V0.8B, V1.8B, V2.8B Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -579,8 +581,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -593,8 +595,8 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x4EA21C20; // ORR V0.16B, V1.16B, V2.16B Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -605,8 +607,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -621,9 +623,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -634,8 +636,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -651,9 +653,9 @@ namespace Ryujinx.Tests.Cpu Bits Op = new Bits(Opcode); ulong _X0 = TestContext.CurrentContext.Random.NextULong(); - AVec V0 = new AVec { X0 = _X0 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0(_X0); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -664,8 +666,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(_X0)); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(_X0)); + Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); }); } @@ -680,9 +682,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -693,8 +695,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -710,9 +712,9 @@ namespace Ryujinx.Tests.Cpu Bits Op = new Bits(Opcode); ulong _X0 = TestContext.CurrentContext.Random.NextULong(); - AVec V0 = new AVec { X0 = _X0 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0(_X0); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -723,8 +725,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(_X0)); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(_X0)); + Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); }); } @@ -735,9 +737,9 @@ namespace Ryujinx.Tests.Cpu uint Opcode = 0x7EE28420; // SUB D0, D1, D2 Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -746,8 +748,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -760,9 +762,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A }; - AVec V2 = new AVec { X0 = B }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0(A); + Vector128 V2 = MakeVectorE0(B); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.V(1, new Bits(A)); @@ -771,8 +773,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -787,8 +789,8 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -799,8 +801,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.Vpart(64, 0, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + 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())); }); } @@ -815,9 +817,9 @@ namespace Ryujinx.Tests.Cpu Opcode |= ((size & 3) << 22); Bits Op = new Bits(Opcode); - AVec V0 = new AVec { X1 = TestContext.CurrentContext.Random.NextULong() }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE1(TestContext.CurrentContext.Random.NextULong()); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -828,8 +830,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(AArch64.V(64, 0).ToUInt64())); - Assert.That(ThreadState.V0.X1, Is.Zero); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(AArch64.V(64, 0).ToUInt64())); + Assert.That(GetVectorE1(ThreadState.V0), Is.Zero); }); } @@ -845,9 +847,9 @@ namespace Ryujinx.Tests.Cpu Bits Op = new Bits(Opcode); ulong _X0 = TestContext.CurrentContext.Random.NextULong(); - AVec V0 = new AVec { X0 = _X0 }; - AVec V1 = new AVec { X0 = A0, X1 = A1 }; - AVec V2 = new AVec { X0 = B0, X1 = B1 }; + Vector128 V0 = MakeVectorE0(_X0); + Vector128 V1 = MakeVectorE0E1(A0, A1); + Vector128 V2 = MakeVectorE0E1(B0, B1); AThreadState ThreadState = SingleOpcode(Opcode, V0: V0, V1: V1, V2: V2); AArch64.Vpart(1, 0, new Bits(A0)); @@ -858,8 +860,8 @@ namespace Ryujinx.Tests.Cpu Assert.Multiple(() => { - Assert.That(ThreadState.V0.X0, Is.EqualTo(_X0)); - Assert.That(ThreadState.V0.X1, Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); + Assert.That(GetVectorE0(ThreadState.V0), Is.EqualTo(_X0)); + Assert.That(GetVectorE1(ThreadState.V0), Is.EqualTo(AArch64.Vpart(64, 0, 1).ToUInt64())); }); } #endif diff --git a/Ryujinx.Tests/Ryujinx.Tests.csproj b/Ryujinx.Tests/Ryujinx.Tests.csproj index 77d86ec4f..b8a1ac541 100644 --- a/Ryujinx.Tests/Ryujinx.Tests.csproj +++ b/Ryujinx.Tests/Ryujinx.Tests.csproj @@ -1,6 +1,6 @@ - netcoreapp2.0 + netcoreapp2.1 win10-x64 Exe false @@ -12,6 +12,7 @@ + diff --git a/Ryujinx/Ryujinx.csproj b/Ryujinx/Ryujinx.csproj index f2d9cafe9..dc6203d52 100644 --- a/Ryujinx/Ryujinx.csproj +++ b/Ryujinx/Ryujinx.csproj @@ -1,7 +1,7 @@ Exe - netcoreapp2.0 + netcoreapp2.1 true win10-x64;osx-x64