Ryujinx/ARMeilleure/Instructions/InstEmitSimdHelper.cs

1576 lines
55 KiB
C#
Raw Normal View History

Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
using ARMeilleure.Decoders;
using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.State;
using ARMeilleure.Translation;
using System;
using System.Diagnostics;
using static ARMeilleure.Instructions.InstEmitHelper;
using static ARMeilleure.IntermediateRepresentation.OperandHelper;
namespace ARMeilleure.Instructions
{
using Func1I = Func<Operand, Operand>;
using Func2I = Func<Operand, Operand, Operand>;
using Func3I = Func<Operand, Operand, Operand, Operand>;
static class InstEmitSimdHelper
{
#region "Masks"
public static readonly long[] EvenMasks = new long[]
{
14L << 56 | 12L << 48 | 10L << 40 | 08L << 32 | 06L << 24 | 04L << 16 | 02L << 8 | 00L << 0, // B
13L << 56 | 12L << 48 | 09L << 40 | 08L << 32 | 05L << 24 | 04L << 16 | 01L << 8 | 00L << 0, // H
11L << 56 | 10L << 48 | 09L << 40 | 08L << 32 | 03L << 24 | 02L << 16 | 01L << 8 | 00L << 0 // S
};
public static readonly long[] OddMasks = new long[]
{
15L << 56 | 13L << 48 | 11L << 40 | 09L << 32 | 07L << 24 | 05L << 16 | 03L << 8 | 01L << 0, // B
15L << 56 | 14L << 48 | 11L << 40 | 10L << 32 | 07L << 24 | 06L << 16 | 03L << 8 | 02L << 0, // H
15L << 56 | 14L << 48 | 13L << 40 | 12L << 32 | 07L << 24 | 06L << 16 | 05L << 8 | 04L << 0 // S
};
public static readonly long ZeroMask = 128L << 56 | 128L << 48 | 128L << 40 | 128L << 32 | 128L << 24 | 128L << 16 | 128L << 8 | 128L << 0;
#endregion
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
#region "X86 SSE Intrinsics"
public static readonly Intrinsic[] X86PaddInstruction = new Intrinsic[]
{
Intrinsic.X86Paddb,
Intrinsic.X86Paddw,
Intrinsic.X86Paddd,
Intrinsic.X86Paddq
};
public static readonly Intrinsic[] X86PcmpeqInstruction = new Intrinsic[]
{
Intrinsic.X86Pcmpeqb,
Intrinsic.X86Pcmpeqw,
Intrinsic.X86Pcmpeqd,
Intrinsic.X86Pcmpeqq
};
public static readonly Intrinsic[] X86PcmpgtInstruction = new Intrinsic[]
{
Intrinsic.X86Pcmpgtb,
Intrinsic.X86Pcmpgtw,
Intrinsic.X86Pcmpgtd,
Intrinsic.X86Pcmpgtq
};
public static readonly Intrinsic[] X86PmaxsInstruction = new Intrinsic[]
{
Intrinsic.X86Pmaxsb,
Intrinsic.X86Pmaxsw,
Intrinsic.X86Pmaxsd
};
public static readonly Intrinsic[] X86PmaxuInstruction = new Intrinsic[]
{
Intrinsic.X86Pmaxub,
Intrinsic.X86Pmaxuw,
Intrinsic.X86Pmaxud
};
public static readonly Intrinsic[] X86PminsInstruction = new Intrinsic[]
{
Intrinsic.X86Pminsb,
Intrinsic.X86Pminsw,
Intrinsic.X86Pminsd
};
public static readonly Intrinsic[] X86PminuInstruction = new Intrinsic[]
{
Intrinsic.X86Pminub,
Intrinsic.X86Pminuw,
Intrinsic.X86Pminud
};
public static readonly Intrinsic[] X86PmovsxInstruction = new Intrinsic[]
{
Intrinsic.X86Pmovsxbw,
Intrinsic.X86Pmovsxwd,
Intrinsic.X86Pmovsxdq
};
public static readonly Intrinsic[] X86PmovzxInstruction = new Intrinsic[]
{
Intrinsic.X86Pmovzxbw,
Intrinsic.X86Pmovzxwd,
Intrinsic.X86Pmovzxdq
};
public static readonly Intrinsic[] X86PsllInstruction = new Intrinsic[]
{
0,
Intrinsic.X86Psllw,
Intrinsic.X86Pslld,
Intrinsic.X86Psllq
};
public static readonly Intrinsic[] X86PsraInstruction = new Intrinsic[]
{
0,
Intrinsic.X86Psraw,
Intrinsic.X86Psrad
};
public static readonly Intrinsic[] X86PsrlInstruction = new Intrinsic[]
{
0,
Intrinsic.X86Psrlw,
Intrinsic.X86Psrld,
Intrinsic.X86Psrlq
};
public static readonly Intrinsic[] X86PsubInstruction = new Intrinsic[]
{
Intrinsic.X86Psubb,
Intrinsic.X86Psubw,
Intrinsic.X86Psubd,
Intrinsic.X86Psubq
};
public static readonly Intrinsic[] X86PunpckhInstruction = new Intrinsic[]
{
Intrinsic.X86Punpckhbw,
Intrinsic.X86Punpckhwd,
Intrinsic.X86Punpckhdq,
Intrinsic.X86Punpckhqdq
};
public static readonly Intrinsic[] X86PunpcklInstruction = new Intrinsic[]
{
Intrinsic.X86Punpcklbw,
Intrinsic.X86Punpcklwd,
Intrinsic.X86Punpckldq,
Intrinsic.X86Punpcklqdq
};
#endregion
public static int GetImmShl(OpCodeSimdShImm op)
{
return op.Imm - (8 << op.Size);
}
public static int GetImmShr(OpCodeSimdShImm op)
{
return (8 << (op.Size + 1)) - op.Imm;
}
public static Operand X86GetScalar(ArmEmitterContext context, float value)
{
return X86GetScalar(context, BitConverter.SingleToInt32Bits(value));
}
public static Operand X86GetScalar(ArmEmitterContext context, double value)
{
return X86GetScalar(context, BitConverter.DoubleToInt64Bits(value));
}
public static Operand X86GetScalar(ArmEmitterContext context, int value)
{
return context.VectorCreateScalar(Const(value));
}
public static Operand X86GetScalar(ArmEmitterContext context, long value)
{
return context.VectorCreateScalar(Const(value));
}
public static Operand X86GetAllElements(ArmEmitterContext context, float value)
{
return X86GetAllElements(context, BitConverter.SingleToInt32Bits(value));
}
public static Operand X86GetAllElements(ArmEmitterContext context, double value)
{
return X86GetAllElements(context, BitConverter.DoubleToInt64Bits(value));
}
public static Operand X86GetAllElements(ArmEmitterContext context, int value)
{
Operand vector = context.VectorCreateScalar(Const(value));
vector = context.AddIntrinsic(Intrinsic.X86Shufps, vector, vector, Const(0));
return vector;
}
public static Operand X86GetAllElements(ArmEmitterContext context, long value)
{
Operand vector = context.VectorCreateScalar(Const(value));
vector = context.AddIntrinsic(Intrinsic.X86Movlhps, vector, vector);
return vector;
}
public static Operand X86GetElements(ArmEmitterContext context, long e1, long e0)
{
Operand vector0 = context.VectorCreateScalar(Const(e0));
Operand vector1 = context.VectorCreateScalar(Const(e1));
return context.AddIntrinsic(Intrinsic.X86Punpcklqdq, vector0, vector1);
}
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
public static int X86GetRoundControl(FPRoundingMode roundMode)
{
switch (roundMode)
{
case FPRoundingMode.ToNearest: return 8 | 0; // even
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
case FPRoundingMode.TowardsPlusInfinity: return 8 | 2;
case FPRoundingMode.TowardsMinusInfinity: return 8 | 1;
case FPRoundingMode.TowardsZero: return 8 | 3;
}
throw new ArgumentException($"Invalid rounding mode \"{roundMode}\".");
}
public static void EmitScalarUnaryOpF(ArmEmitterContext context, Intrinsic inst32, Intrinsic inst64)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand n = GetVec(op.Rn);
Intrinsic inst = (op.Size & 1) != 0 ? inst64 : inst32;
Operand res = context.AddIntrinsic(inst, n);
if ((op.Size & 1) != 0)
{
res = context.VectorZeroUpper64(res);
}
else
{
res = context.VectorZeroUpper96(res);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitScalarBinaryOpF(ArmEmitterContext context, Intrinsic inst32, Intrinsic inst64)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
Intrinsic inst = (op.Size & 1) != 0 ? inst64 : inst32;
Operand res = context.AddIntrinsic(inst, n, m);
if ((op.Size & 1) != 0)
{
res = context.VectorZeroUpper64(res);
}
else
{
res = context.VectorZeroUpper96(res);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorUnaryOpF(ArmEmitterContext context, Intrinsic inst32, Intrinsic inst64)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand n = GetVec(op.Rn);
Intrinsic inst = (op.Size & 1) != 0 ? inst64 : inst32;
Operand res = context.AddIntrinsic(inst, n);
if (op.RegisterSize == RegisterSize.Simd64)
{
res = context.VectorZeroUpper64(res);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpF(ArmEmitterContext context, Intrinsic inst32, Intrinsic inst64)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
Intrinsic inst = (op.Size & 1) != 0 ? inst64 : inst32;
Operand res = context.AddIntrinsic(inst, n, m);
if (op.RegisterSize == RegisterSize.Simd64)
{
res = context.VectorZeroUpper64(res);
}
context.Copy(GetVec(op.Rd), res);
}
public static Operand EmitUnaryMathCall(ArmEmitterContext context, _F32_F32 f32, _F64_F64 f64, Operand n)
{
IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
return (op.Size & 1) == 0 ? context.Call(f32, n) : context.Call(f64, n);
}
public static Operand EmitRoundMathCall(ArmEmitterContext context, MidpointRounding roundMode, Operand n)
{
IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
Delegate dlg;
if ((op.Size & 1) == 0)
{
dlg = new _F32_F32_MidpointRounding(MathF.Round);
}
else /* if ((op.Size & 1) == 1) */
{
dlg = new _F64_F64_MidpointRounding(Math.Round);
}
return context.Call(dlg, n, Const((int)roundMode));
}
public static Operand EmitSoftFloatCall(
ArmEmitterContext context,
_F32_F32 f32,
_F64_F64 f64,
params Operand[] callArgs)
{
IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
Delegate dlg = (op.Size & 1) == 0 ? (Delegate)f32 : (Delegate)f64;
return context.Call(dlg, callArgs);
}
public static Operand EmitSoftFloatCall(
ArmEmitterContext context,
_F32_F32_F32 f32,
_F64_F64_F64 f64,
params Operand[] callArgs)
{
IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
Delegate dlg = (op.Size & 1) == 0 ? (Delegate)f32 : (Delegate)f64;
return context.Call(dlg, callArgs);
}
public static Operand EmitSoftFloatCall(
ArmEmitterContext context,
_F32_F32_F32_F32 f32,
_F64_F64_F64_F64 f64,
params Operand[] callArgs)
{
IOpCodeSimd op = (IOpCodeSimd)context.CurrOp;
Delegate dlg = (op.Size & 1) == 0 ? (Delegate)f32 : (Delegate)f64;
return context.Call(dlg, callArgs);
}
public static void EmitScalarBinaryOpByElemF(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdRegElemF op = (OpCodeSimdRegElemF)context.CurrOp;
OperandType type = (op.Size & 1) != 0 ? OperandType.FP64 : OperandType.FP32;
Operand n = context.VectorExtract(type, GetVec(op.Rn), 0);
Operand m = context.VectorExtract(type, GetVec(op.Rm), op.Index);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), emit(n, m), 0));
}
public static void EmitScalarTernaryOpByElemF(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdRegElemF op = (OpCodeSimdRegElemF)context.CurrOp;
OperandType type = (op.Size & 1) != 0 ? OperandType.FP64 : OperandType.FP32;
Operand d = context.VectorExtract(type, GetVec(op.Rd), 0);
Operand n = context.VectorExtract(type, GetVec(op.Rn), 0);
Operand m = context.VectorExtract(type, GetVec(op.Rm), op.Index);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), emit(d, n, m), 0));
}
public static void EmitScalarUnaryOpSx(ArmEmitterContext context, Func1I emit)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand n = EmitVectorExtractSx(context, op.Rn, 0, op.Size);
Operand d = EmitVectorInsert(context, context.VectorZero(), emit(n), 0, op.Size);
context.Copy(GetVec(op.Rd), d);
}
public static void EmitScalarBinaryOpSx(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand n = EmitVectorExtractSx(context, op.Rn, 0, op.Size);
Operand m = EmitVectorExtractSx(context, op.Rm, 0, op.Size);
Operand d = EmitVectorInsert(context, context.VectorZero(), emit(n, m), 0, op.Size);
context.Copy(GetVec(op.Rd), d);
}
public static void EmitScalarUnaryOpZx(ArmEmitterContext context, Func1I emit)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand n = EmitVectorExtractZx(context, op.Rn, 0, op.Size);
Operand d = EmitVectorInsert(context, context.VectorZero(), emit(n), 0, op.Size);
context.Copy(GetVec(op.Rd), d);
}
public static void EmitScalarBinaryOpZx(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand n = EmitVectorExtractZx(context, op.Rn, 0, op.Size);
Operand m = EmitVectorExtractZx(context, op.Rm, 0, op.Size);
Operand d = EmitVectorInsert(context, context.VectorZero(), emit(n, m), 0, op.Size);
context.Copy(GetVec(op.Rd), d);
}
public static void EmitScalarTernaryOpZx(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand d = EmitVectorExtractZx(context, op.Rd, 0, op.Size);
Operand n = EmitVectorExtractZx(context, op.Rn, 0, op.Size);
Operand m = EmitVectorExtractZx(context, op.Rm, 0, op.Size);
d = EmitVectorInsert(context, context.VectorZero(), emit(d, n, m), 0, op.Size);
context.Copy(GetVec(op.Rd), d);
}
public static void EmitScalarUnaryOpF(ArmEmitterContext context, Func1I emit)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
OperandType type = (op.Size & 1) != 0 ? OperandType.FP64 : OperandType.FP32;
Operand n = context.VectorExtract(type, GetVec(op.Rn), 0);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), emit(n), 0));
}
public static void EmitScalarBinaryOpF(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
OperandType type = (op.Size & 1) != 0 ? OperandType.FP64 : OperandType.FP32;
Operand n = context.VectorExtract(type, GetVec(op.Rn), 0);
Operand m = context.VectorExtract(type, GetVec(op.Rm), 0);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), emit(n, m), 0));
}
public static void EmitScalarTernaryRaOpF(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
OperandType type = (op.Size & 1) != 0 ? OperandType.FP64 : OperandType.FP32;
Operand a = context.VectorExtract(type, GetVec(op.Ra), 0);
Operand n = context.VectorExtract(type, GetVec(op.Rn), 0);
Operand m = context.VectorExtract(type, GetVec(op.Rm), 0);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), emit(a, n, m), 0));
}
public static void EmitVectorUnaryOpF(ArmEmitterContext context, Func1I emit)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand res = context.VectorZero();
int sizeF = op.Size & 1;
OperandType type = sizeF != 0 ? OperandType.FP64 : OperandType.FP32;
int elems = op.GetBytesCount() >> sizeF + 2;
for (int index = 0; index < elems; index++)
{
Operand ne = context.VectorExtract(type, GetVec(op.Rn), index);
res = context.VectorInsert(res, emit(ne), index);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpF(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int sizeF = op.Size & 1;
OperandType type = sizeF != 0 ? OperandType.FP64 : OperandType.FP32;
int elems = op.GetBytesCount() >> sizeF + 2;
for (int index = 0; index < elems; index++)
{
Operand ne = context.VectorExtract(type, GetVec(op.Rn), index);
Operand me = context.VectorExtract(type, GetVec(op.Rm), index);
res = context.VectorInsert(res, emit(ne, me), index);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorTernaryOpF(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int sizeF = op.Size & 1;
OperandType type = sizeF != 0 ? OperandType.FP64 : OperandType.FP32;
int elems = op.GetBytesCount() >> sizeF + 2;
for (int index = 0; index < elems; index++)
{
Operand de = context.VectorExtract(type, GetVec(op.Rd), index);
Operand ne = context.VectorExtract(type, GetVec(op.Rn), index);
Operand me = context.VectorExtract(type, GetVec(op.Rm), index);
res = context.VectorInsert(res, emit(de, ne, me), index);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpByElemF(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdRegElemF op = (OpCodeSimdRegElemF)context.CurrOp;
Operand res = context.VectorZero();
int sizeF = op.Size & 1;
OperandType type = sizeF != 0 ? OperandType.FP64 : OperandType.FP32;
int elems = op.GetBytesCount() >> sizeF + 2;
for (int index = 0; index < elems; index++)
{
Operand ne = context.VectorExtract(type, GetVec(op.Rn), index);
Operand me = context.VectorExtract(type, GetVec(op.Rm), op.Index);
res = context.VectorInsert(res, emit(ne, me), index);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorTernaryOpByElemF(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdRegElemF op = (OpCodeSimdRegElemF)context.CurrOp;
Operand res = context.VectorZero();
int sizeF = op.Size & 1;
OperandType type = sizeF != 0 ? OperandType.FP64 : OperandType.FP32;
int elems = op.GetBytesCount() >> sizeF + 2;
for (int index = 0; index < elems; index++)
{
Operand de = context.VectorExtract(type, GetVec(op.Rd), index);
Operand ne = context.VectorExtract(type, GetVec(op.Rn), index);
Operand me = context.VectorExtract(type, GetVec(op.Rm), op.Index);
res = context.VectorInsert(res, emit(de, ne, me), index);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorUnaryOpSx(ArmEmitterContext context, Func1I emit)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractSx(context, op.Rn, index, op.Size);
res = EmitVectorInsert(context, res, emit(ne), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpSx(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractSx(context, op.Rn, index, op.Size);
Operand me = EmitVectorExtractSx(context, op.Rm, index, op.Size);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorTernaryOpSx(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand de = EmitVectorExtractSx(context, op.Rd, index, op.Size);
Operand ne = EmitVectorExtractSx(context, op.Rn, index, op.Size);
Operand me = EmitVectorExtractSx(context, op.Rm, index, op.Size);
res = EmitVectorInsert(context, res, emit(de, ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorUnaryOpZx(ArmEmitterContext context, Func1I emit)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractZx(context, op.Rn, index, op.Size);
res = EmitVectorInsert(context, res, emit(ne), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpZx(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractZx(context, op.Rn, index, op.Size);
Operand me = EmitVectorExtractZx(context, op.Rm, index, op.Size);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorTernaryOpZx(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand de = EmitVectorExtractZx(context, op.Rd, index, op.Size);
Operand ne = EmitVectorExtractZx(context, op.Rn, index, op.Size);
Operand me = EmitVectorExtractZx(context, op.Rm, index, op.Size);
res = EmitVectorInsert(context, res, emit(de, ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpByElemSx(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdRegElem op = (OpCodeSimdRegElem)context.CurrOp;
Operand res = context.VectorZero();
Operand me = EmitVectorExtractSx(context, op.Rm, op.Index, op.Size);
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractSx(context, op.Rn, index, op.Size);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorBinaryOpByElemZx(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdRegElem op = (OpCodeSimdRegElem)context.CurrOp;
Operand res = context.VectorZero();
Operand me = EmitVectorExtractZx(context, op.Rm, op.Index, op.Size);
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractZx(context, op.Rn, index, op.Size);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorTernaryOpByElemZx(ArmEmitterContext context, Func3I emit)
{
OpCodeSimdRegElem op = (OpCodeSimdRegElem)context.CurrOp;
Operand res = context.VectorZero();
Operand me = EmitVectorExtractZx(context, op.Rm, op.Index, op.Size);
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand de = EmitVectorExtractZx(context, op.Rd, index, op.Size);
Operand ne = EmitVectorExtractZx(context, op.Rn, index, op.Size);
res = EmitVectorInsert(context, res, emit(de, ne, me), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorImmUnaryOp(ArmEmitterContext context, Func1I emit)
{
OpCodeSimdImm op = (OpCodeSimdImm)context.CurrOp;
Operand imm = Const(op.Immediate);
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
res = EmitVectorInsert(context, res, emit(imm), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorImmBinaryOp(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdImm op = (OpCodeSimdImm)context.CurrOp;
Operand imm = Const(op.Immediate);
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
Operand de = EmitVectorExtractZx(context, op.Rd, index, op.Size);
res = EmitVectorInsert(context, res, emit(de, imm), index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorWidenRmBinaryOpSx(ArmEmitterContext context, Func2I emit)
{
EmitVectorWidenRmBinaryOp(context, emit, signed: true);
}
public static void EmitVectorWidenRmBinaryOpZx(ArmEmitterContext context, Func2I emit)
{
EmitVectorWidenRmBinaryOp(context, emit, signed: false);
}
private static void EmitVectorWidenRmBinaryOp(ArmEmitterContext context, Func2I emit, bool signed)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = 8 >> op.Size;
int part = op.RegisterSize == RegisterSize.Simd128 ? elems : 0;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtract(context, op.Rn, index, op.Size + 1, signed);
Operand me = EmitVectorExtract(context, op.Rm, part + index, op.Size, signed);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size + 1);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorWidenRnRmBinaryOpSx(ArmEmitterContext context, Func2I emit)
{
EmitVectorWidenRnRmBinaryOp(context, emit, signed: true);
}
public static void EmitVectorWidenRnRmBinaryOpZx(ArmEmitterContext context, Func2I emit)
{
EmitVectorWidenRnRmBinaryOp(context, emit, signed: false);
}
private static void EmitVectorWidenRnRmBinaryOp(ArmEmitterContext context, Func2I emit, bool signed)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = 8 >> op.Size;
int part = op.RegisterSize == RegisterSize.Simd128 ? elems : 0;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtract(context, op.Rn, part + index, op.Size, signed);
Operand me = EmitVectorExtract(context, op.Rm, part + index, op.Size, signed);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size + 1);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorWidenRnRmTernaryOpSx(ArmEmitterContext context, Func3I emit)
{
EmitVectorWidenRnRmTernaryOp(context, emit, signed: true);
}
public static void EmitVectorWidenRnRmTernaryOpZx(ArmEmitterContext context, Func3I emit)
{
EmitVectorWidenRnRmTernaryOp(context, emit, signed: false);
}
private static void EmitVectorWidenRnRmTernaryOp(ArmEmitterContext context, Func3I emit, bool signed)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int elems = 8 >> op.Size;
int part = op.RegisterSize == RegisterSize.Simd128 ? elems : 0;
for (int index = 0; index < elems; index++)
{
Operand de = EmitVectorExtract(context, op.Rd, index, op.Size + 1, signed);
Operand ne = EmitVectorExtract(context, op.Rn, part + index, op.Size, signed);
Operand me = EmitVectorExtract(context, op.Rm, part + index, op.Size, signed);
res = EmitVectorInsert(context, res, emit(de, ne, me), index, op.Size + 1);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorWidenBinaryOpByElemSx(ArmEmitterContext context, Func2I emit)
{
EmitVectorWidenBinaryOpByElem(context, emit, signed: true);
}
public static void EmitVectorWidenBinaryOpByElemZx(ArmEmitterContext context, Func2I emit)
{
EmitVectorWidenBinaryOpByElem(context, emit, signed: false);
}
private static void EmitVectorWidenBinaryOpByElem(ArmEmitterContext context, Func2I emit, bool signed)
{
OpCodeSimdRegElem op = (OpCodeSimdRegElem)context.CurrOp;
Operand res = context.VectorZero();
Operand me = EmitVectorExtract(context, op.Rm, op.Index, op.Size, signed);
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
int elems = 8 >> op.Size;
int part = op.RegisterSize == RegisterSize.Simd128 ? elems : 0;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtract(context, op.Rn, part + index, op.Size, signed);
res = EmitVectorInsert(context, res, emit(ne, me), index, op.Size + 1);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorWidenTernaryOpByElemSx(ArmEmitterContext context, Func3I emit)
{
EmitVectorWidenTernaryOpByElem(context, emit, signed: true);
}
public static void EmitVectorWidenTernaryOpByElemZx(ArmEmitterContext context, Func3I emit)
{
EmitVectorWidenTernaryOpByElem(context, emit, signed: false);
}
private static void EmitVectorWidenTernaryOpByElem(ArmEmitterContext context, Func3I emit, bool signed)
{
OpCodeSimdRegElem op = (OpCodeSimdRegElem)context.CurrOp;
Operand res = context.VectorZero();
Operand me = EmitVectorExtract(context, op.Rm, op.Index, op.Size, signed);
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
int elems = 8 >> op.Size;
int part = op.RegisterSize == RegisterSize.Simd128 ? elems : 0;
for (int index = 0; index < elems; index++)
{
Operand de = EmitVectorExtract(context, op.Rd, index, op.Size + 1, signed);
Operand ne = EmitVectorExtract(context, op.Rn, part + index, op.Size, signed);
res = EmitVectorInsert(context, res, emit(de, ne, me), index, op.Size + 1);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitVectorPairwiseOpSx(ArmEmitterContext context, Func2I emit)
{
EmitVectorPairwiseOp(context, emit, signed: true);
}
public static void EmitVectorPairwiseOpZx(ArmEmitterContext context, Func2I emit)
{
EmitVectorPairwiseOp(context, emit, signed: false);
}
private static void EmitVectorPairwiseOp(ArmEmitterContext context, Func2I emit, bool signed)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int pairs = op.GetPairsCount() >> op.Size;
for (int index = 0; index < pairs; index++)
{
int pairIndex = index << 1;
Operand n0 = EmitVectorExtract(context, op.Rn, pairIndex, op.Size, signed);
Operand n1 = EmitVectorExtract(context, op.Rn, pairIndex + 1, op.Size, signed);
Operand m0 = EmitVectorExtract(context, op.Rm, pairIndex, op.Size, signed);
Operand m1 = EmitVectorExtract(context, op.Rm, pairIndex + 1, op.Size, signed);
res = EmitVectorInsert(context, res, emit(n0, n1), index, op.Size);
res = EmitVectorInsert(context, res, emit(m0, m1), pairs + index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitSsse3VectorPairwiseOp(ArmEmitterContext context, Intrinsic[] inst)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
if (op.RegisterSize == RegisterSize.Simd64)
{
Operand zeroEvenMask = X86GetElements(context, ZeroMask, EvenMasks[op.Size]);
Operand zeroOddMask = X86GetElements(context, ZeroMask, OddMasks [op.Size]);
Operand mN = context.AddIntrinsic(Intrinsic.X86Punpcklqdq, n, m); // m:n
Operand left = context.AddIntrinsic(Intrinsic.X86Pshufb, mN, zeroEvenMask); // 0:even from m:n
Operand right = context.AddIntrinsic(Intrinsic.X86Pshufb, mN, zeroOddMask); // 0:odd from m:n
context.Copy(GetVec(op.Rd), context.AddIntrinsic(inst[op.Size], left, right));
}
else if (op.Size < 3)
{
Operand oddEvenMask = X86GetElements(context, OddMasks[op.Size], EvenMasks[op.Size]);
Operand oddEvenN = context.AddIntrinsic(Intrinsic.X86Pshufb, n, oddEvenMask); // odd:even from n
Operand oddEvenM = context.AddIntrinsic(Intrinsic.X86Pshufb, m, oddEvenMask); // odd:even from m
Operand left = context.AddIntrinsic(Intrinsic.X86Punpcklqdq, oddEvenN, oddEvenM);
Operand right = context.AddIntrinsic(Intrinsic.X86Punpckhqdq, oddEvenN, oddEvenM);
context.Copy(GetVec(op.Rd), context.AddIntrinsic(inst[op.Size], left, right));
}
else
{
Operand left = context.AddIntrinsic(Intrinsic.X86Punpcklqdq, n, m);
Operand right = context.AddIntrinsic(Intrinsic.X86Punpckhqdq, n, m);
context.Copy(GetVec(op.Rd), context.AddIntrinsic(inst[3], left, right));
}
}
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
public static void EmitVectorAcrossVectorOpSx(ArmEmitterContext context, Func2I emit)
{
EmitVectorAcrossVectorOp(context, emit, signed: true, isLong: false);
}
public static void EmitVectorAcrossVectorOpZx(ArmEmitterContext context, Func2I emit)
{
EmitVectorAcrossVectorOp(context, emit, signed: false, isLong: false);
}
public static void EmitVectorLongAcrossVectorOpSx(ArmEmitterContext context, Func2I emit)
{
EmitVectorAcrossVectorOp(context, emit, signed: true, isLong: true);
}
public static void EmitVectorLongAcrossVectorOpZx(ArmEmitterContext context, Func2I emit)
{
EmitVectorAcrossVectorOp(context, emit, signed: false, isLong: true);
}
private static void EmitVectorAcrossVectorOp(
ArmEmitterContext context,
Func2I emit,
bool signed,
bool isLong)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
int elems = op.GetBytesCount() >> op.Size;
Operand res = EmitVectorExtract(context, op.Rn, 0, op.Size, signed);
for (int index = 1; index < elems; index++)
{
Operand n = EmitVectorExtract(context, op.Rn, index, op.Size, signed);
res = emit(res, n);
}
int size = isLong ? op.Size + 1 : op.Size;
Operand d = EmitVectorInsert(context, context.VectorZero(), res, 0, size);
context.Copy(GetVec(op.Rd), d);
}
public static void EmitVectorPairwiseOpF(ArmEmitterContext context, Func2I emit)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand res = context.VectorZero();
int sizeF = op.Size & 1;
OperandType type = sizeF != 0 ? OperandType.FP64 : OperandType.FP32;
int pairs = op.GetPairsCount() >> sizeF + 2;
for (int index = 0; index < pairs; index++)
{
int pairIndex = index << 1;
Operand n0 = context.VectorExtract(type, GetVec(op.Rn), pairIndex);
Operand n1 = context.VectorExtract(type, GetVec(op.Rn), pairIndex + 1);
Operand m0 = context.VectorExtract(type, GetVec(op.Rm), pairIndex);
Operand m1 = context.VectorExtract(type, GetVec(op.Rm), pairIndex + 1);
res = context.VectorInsert(res, emit(n0, n1), index);
res = context.VectorInsert(res, emit(m0, m1), pairs + index);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitSse2VectorPairwiseOpF(ArmEmitterContext context, Intrinsic inst32, Intrinsic inst64)
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
int sizeF = op.Size & 1;
if (sizeF == 0)
{
if (op.RegisterSize == RegisterSize.Simd64)
{
Operand unpck = context.AddIntrinsic(Intrinsic.X86Unpcklps, n, m);
Operand zero = context.VectorZero();
Operand part0 = context.AddIntrinsic(Intrinsic.X86Movlhps, unpck, zero);
Operand part1 = context.AddIntrinsic(Intrinsic.X86Movhlps, zero, unpck);
context.Copy(GetVec(op.Rd), context.AddIntrinsic(inst32, part0, part1));
}
else /* if (op.RegisterSize == RegisterSize.Simd128) */
{
const int sm0 = 2 << 6 | 0 << 4 | 2 << 2 | 0 << 0;
const int sm1 = 3 << 6 | 1 << 4 | 3 << 2 | 1 << 0;
Operand part0 = context.AddIntrinsic(Intrinsic.X86Shufps, n, m, Const(sm0));
Operand part1 = context.AddIntrinsic(Intrinsic.X86Shufps, n, m, Const(sm1));
context.Copy(GetVec(op.Rd), context.AddIntrinsic(inst32, part0, part1));
}
}
else /* if (sizeF == 1) */
{
Operand part0 = context.AddIntrinsic(Intrinsic.X86Unpcklpd, n, m);
Operand part1 = context.AddIntrinsic(Intrinsic.X86Unpckhpd, n, m);
context.Copy(GetVec(op.Rd), context.AddIntrinsic(inst64, part0, part1));
}
}
public enum CmpCondition
{
// Legacy Sse.
Equal = 0, // Ordered, non-signaling.
LessThan = 1, // Ordered, signaling.
LessThanOrEqual = 2, // Ordered, signaling.
UnorderedQ = 3, // Non-signaling.
NotLessThan = 5, // Unordered, signaling.
NotLessThanOrEqual = 6, // Unordered, signaling.
OrderedQ = 7, // Non-signaling.
// Vex.
GreaterThanOrEqual = 13, // Ordered, signaling.
GreaterThan = 14, // Ordered, signaling.
OrderedS = 23 // Signaling.
}
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
[Flags]
public enum SaturatingFlags
{
Scalar = 1 << 0,
Signed = 1 << 1,
Add = 1 << 2,
Sub = 1 << 3,
Accumulate = 1 << 4,
ScalarSx = Scalar | Signed,
ScalarZx = Scalar,
VectorSx = Signed,
VectorZx = 0
}
public static void EmitScalarSaturatingUnaryOpSx(ArmEmitterContext context, Func1I emit)
{
EmitSaturatingUnaryOpSx(context, emit, SaturatingFlags.ScalarSx);
}
public static void EmitVectorSaturatingUnaryOpSx(ArmEmitterContext context, Func1I emit)
{
EmitSaturatingUnaryOpSx(context, emit, SaturatingFlags.VectorSx);
}
private static void EmitSaturatingUnaryOpSx(ArmEmitterContext context, Func1I emit, SaturatingFlags flags)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand res = context.VectorZero();
bool scalar = (flags & SaturatingFlags.Scalar) != 0;
int elems = !scalar ? op.GetBytesCount() >> op.Size : 1;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractSx(context, op.Rn, index, op.Size);
Operand de;
if (op.Size <= 2)
{
de = EmitSatQ(context, emit(ne), op.Size, signedSrc: true, signedDst: true);
}
else /* if (op.Size == 3) */
{
de = EmitUnarySignedSatQAbsOrNeg(context, emit(ne));
}
res = EmitVectorInsert(context, res, de, index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
public static void EmitScalarSaturatingBinaryOpSx(ArmEmitterContext context, SaturatingFlags flags)
{
EmitSaturatingBinaryOp(context, null, SaturatingFlags.ScalarSx | flags);
}
public static void EmitScalarSaturatingBinaryOpZx(ArmEmitterContext context, SaturatingFlags flags)
{
EmitSaturatingBinaryOp(context, null, SaturatingFlags.ScalarZx | flags);
}
public static void EmitVectorSaturatingBinaryOpSx(ArmEmitterContext context, SaturatingFlags flags)
{
EmitSaturatingBinaryOp(context, null, SaturatingFlags.VectorSx | flags);
}
public static void EmitVectorSaturatingBinaryOpZx(ArmEmitterContext context, SaturatingFlags flags)
{
EmitSaturatingBinaryOp(context, null, SaturatingFlags.VectorZx | flags);
}
public static void EmitSaturatingBinaryOp(ArmEmitterContext context, Func2I emit, SaturatingFlags flags)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand res = context.VectorZero();
bool scalar = (flags & SaturatingFlags.Scalar) != 0;
bool signed = (flags & SaturatingFlags.Signed) != 0;
bool add = (flags & SaturatingFlags.Add) != 0;
bool sub = (flags & SaturatingFlags.Sub) != 0;
bool accumulate = (flags & SaturatingFlags.Accumulate) != 0;
int elems = !scalar ? op.GetBytesCount() >> op.Size : 1;
if (add || sub)
{
OpCodeSimdReg opReg = (OpCodeSimdReg)op;
for (int index = 0; index < elems; index++)
{
Operand de;
Operand ne = EmitVectorExtract(context, opReg.Rn, index, op.Size, signed);
Operand me = EmitVectorExtract(context, opReg.Rm, index, op.Size, signed);
if (op.Size <= 2)
{
Operand temp = add ? context.Add(ne, me) : context.Subtract(ne, me);
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
de = EmitSatQ(context, temp, op.Size, signedSrc: true, signedDst: signed);
}
else if (add) /* if (op.Size == 3) */
{
de = EmitBinarySatQAdd(context, ne, me, signed);
}
else /* if (sub) */
{
de = EmitBinarySatQSub(context, ne, me, signed);
}
res = EmitVectorInsert(context, res, de, index, op.Size);
}
}
else if (accumulate)
{
for (int index = 0; index < elems; index++)
{
Operand de;
Operand ne = EmitVectorExtract(context, op.Rn, index, op.Size, !signed);
Operand me = EmitVectorExtract(context, op.Rd, index, op.Size, signed);
if (op.Size <= 2)
{
Operand temp = context.Add(ne, me);
de = EmitSatQ(context, temp, op.Size, signedSrc: true, signedDst: signed);
}
else /* if (op.Size == 3) */
{
de = EmitBinarySatQAccumulate(context, ne, me, signed);
}
res = EmitVectorInsert(context, res, de, index, op.Size);
}
}
else
{
OpCodeSimdReg opReg = (OpCodeSimdReg)op;
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtract(context, opReg.Rn, index, op.Size, signed);
Operand me = EmitVectorExtract(context, opReg.Rm, index, op.Size, signed);
Operand de = EmitSatQ(context, emit(ne, me), op.Size, true, signed);
res = EmitVectorInsert(context, res, de, index, op.Size);
}
}
context.Copy(GetVec(op.Rd), res);
}
[Flags]
public enum SaturatingNarrowFlags
{
Scalar = 1 << 0,
SignedSrc = 1 << 1,
SignedDst = 1 << 2,
ScalarSxSx = Scalar | SignedSrc | SignedDst,
ScalarSxZx = Scalar | SignedSrc,
ScalarZxZx = Scalar,
VectorSxSx = SignedSrc | SignedDst,
VectorSxZx = SignedSrc,
VectorZxZx = 0
}
public static void EmitSaturatingNarrowOp(ArmEmitterContext context, SaturatingNarrowFlags flags)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
bool scalar = (flags & SaturatingNarrowFlags.Scalar) != 0;
bool signedSrc = (flags & SaturatingNarrowFlags.SignedSrc) != 0;
bool signedDst = (flags & SaturatingNarrowFlags.SignedDst) != 0;
int elems = !scalar ? 8 >> op.Size : 1;
int part = !scalar && (op.RegisterSize == RegisterSize.Simd128) ? elems : 0;
Operand d = GetVec(op.Rd);
Operand res = part == 0 ? context.VectorZero() : context.Copy(d);
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtract(context, op.Rn, index, op.Size + 1, signedSrc);
Operand temp = EmitSatQ(context, ne, op.Size, signedSrc, signedDst);
res = EmitVectorInsert(context, res, temp, part + index, op.Size);
}
context.Copy(d, res);
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
}
// TSrc (16bit, 32bit, 64bit; signed, unsigned) > TDst (8bit, 16bit, 32bit; signed, unsigned).
public static Operand EmitSatQ(ArmEmitterContext context, Operand op, int sizeDst, bool signedSrc, bool signedDst)
{
if ((uint)sizeDst > 2u)
{
throw new ArgumentOutOfRangeException(nameof(sizeDst));
}
Delegate dlg;
if (signedSrc)
{
dlg = signedDst
? (Delegate)new _S64_S64_S32(SoftFallback.SignedSrcSignedDstSatQ)
: (Delegate)new _U64_S64_S32(SoftFallback.SignedSrcUnsignedDstSatQ);
}
else
{
dlg = signedDst
? (Delegate)new _S64_U64_S32(SoftFallback.UnsignedSrcSignedDstSatQ)
: (Delegate)new _U64_U64_S32(SoftFallback.UnsignedSrcUnsignedDstSatQ);
}
return context.Call(dlg, op, Const(sizeDst));
}
// TSrc (64bit) == TDst (64bit); signed.
public static Operand EmitUnarySignedSatQAbsOrNeg(ArmEmitterContext context, Operand op)
{
Debug.Assert(((OpCodeSimd)context.CurrOp).Size == 3, "Invalid element size.");
return context.Call(new _S64_S64(SoftFallback.UnarySignedSatQAbsOrNeg), op);
}
// TSrcs (64bit) == TDst (64bit); signed, unsigned.
public static Operand EmitBinarySatQAdd(ArmEmitterContext context, Operand op1, Operand op2, bool signed)
{
Debug.Assert(((OpCodeSimd)context.CurrOp).Size == 3, "Invalid element size.");
Delegate dlg = signed
? (Delegate)new _S64_S64_S64(SoftFallback.BinarySignedSatQAdd)
: (Delegate)new _U64_U64_U64(SoftFallback.BinaryUnsignedSatQAdd);
return context.Call(dlg, op1, op2);
}
// TSrcs (64bit) == TDst (64bit); signed, unsigned.
public static Operand EmitBinarySatQSub(ArmEmitterContext context, Operand op1, Operand op2, bool signed)
{
Debug.Assert(((OpCodeSimd)context.CurrOp).Size == 3, "Invalid element size.");
Delegate dlg = signed
? (Delegate)new _S64_S64_S64(SoftFallback.BinarySignedSatQSub)
: (Delegate)new _U64_U64_U64(SoftFallback.BinaryUnsignedSatQSub);
return context.Call(dlg, op1, op2);
}
// TSrcs (64bit) == TDst (64bit); signed, unsigned.
public static Operand EmitBinarySatQAccumulate(ArmEmitterContext context, Operand op1, Operand op2, bool signed)
{
Debug.Assert(((OpCodeSimd)context.CurrOp).Size == 3, "Invalid element size.");
Delegate dlg = signed
? (Delegate)new _S64_U64_S64(SoftFallback.BinarySignedSatQAcc)
: (Delegate)new _U64_S64_U64(SoftFallback.BinaryUnsignedSatQAcc);
return context.Call(dlg, op1, op2);
}
public static Operand EmitFloatAbs(ArmEmitterContext context, Operand value, bool single, bool vector)
{
Operand mask;
if (single)
{
mask = vector ? X86GetAllElements(context, -0f) : X86GetScalar(context, -0f);
}
else
{
mask = vector ? X86GetAllElements(context, -0d) : X86GetScalar(context, -0d);
}
return context.AddIntrinsic(single ? Intrinsic.X86Andnps : Intrinsic.X86Andnpd, mask, value);
}
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
public static Operand EmitVectorExtractSx(ArmEmitterContext context, int reg, int index, int size)
{
return EmitVectorExtract(context, reg, index, size, true);
}
public static Operand EmitVectorExtractZx(ArmEmitterContext context, int reg, int index, int size)
{
return EmitVectorExtract(context, reg, index, size, false);
}
public static Operand EmitVectorExtract(ArmEmitterContext context, int reg, int index, int size, bool signed)
{
ThrowIfInvalid(index, size);
Operand res = null;
switch (size)
{
case 0:
res = context.VectorExtract8(GetVec(reg), index);
break;
case 1:
res = context.VectorExtract16(GetVec(reg), index);
break;
case 2:
res = context.VectorExtract(OperandType.I32, GetVec(reg), index);
break;
case 3:
res = context.VectorExtract(OperandType.I64, GetVec(reg), index);
break;
}
if (signed)
{
switch (size)
{
case 0: res = context.SignExtend8 (OperandType.I64, res); break;
case 1: res = context.SignExtend16(OperandType.I64, res); break;
case 2: res = context.SignExtend32(OperandType.I64, res); break;
}
}
else
{
switch (size)
{
case 0: res = context.ZeroExtend8 (OperandType.I64, res); break;
case 1: res = context.ZeroExtend16(OperandType.I64, res); break;
case 2: res = context.ZeroExtend32(OperandType.I64, res); break;
}
}
return res;
}
public static Operand EmitVectorInsert(ArmEmitterContext context, Operand vector, Operand value, int index, int size)
{
ThrowIfInvalid(index, size);
Add most of the A32 instruction set to ARMeilleure (#897) * Implement TEQ and MOV (Imm16) * Initial work on A32 instructions + SVC. No tests yet, hangs in rtld. * Implement CLZ, fix BFI and BFC Now stops on SIMD initialization. * Exclusive access instructions, fix to mul, system instructions. Now gets to a break after SignalProcessWideKey64. * Better impl of UBFX, add UDIV and SDIV Now boots way further - now stuck on VMOV instruction. * Many more instructions, start on SIMD and testing framework. * Fix build issues * svc: Rework 32 bit codepath Fixing once and for all argument ordering issues. * Fix 32 bits stacktrace * hle debug: Add 32 bits dynamic section parsing * Fix highCq mode, add many tests, fix some instruction bugs Still suffers from critical malloc failure :weary: * Fix incorrect opcode decoders and a few more instructions. * Add a few instructions and fix others. re-disable highCq for now. Disabled the svc memory clear since i'm not sure about it. * Fix build * Fix typo in ordered/exclusive stores. * Implement some more instructions, fix others. Uxtab16/Sxtab16 are untested. * Begin impl of pairwise, some other instructions. * Add a few more instructions, a quick hack to fix svcs for now. * Add tests and fix issues with VTRN, VZIP, VUZP * Add a few more instructions, fix Vmul_1 encoding. * Fix way too many instruction bugs, add tests for some of the more important ones. * Fix HighCq, enable FastFP paths for some floating point instructions (not entirely sure why these were disabled, so important to note this commit exists) Branching has been removed in A32 shifts until I figure out if it's worth it * Cleanup Part 1 There should be no functional change between these next few commits. Should is the key word. (except for removing break handler) * Implement 32 bits syscalls Co-authored-by: riperiperi <rhy3756547@hotmail.com> Implement all 32 bits counterparts of the 64 bits syscalls we currently have. * Refactor part 2: Move index/subindex logic to Operand May have inadvertently fixed one (1) bug * Add FlushProcessDataCache32 * Address jd's comments * Remove 16 bit encodings from OpCodeTable Still need to catch some edge cases (operands that use the "F" flag) and make Q encodings with non-even indexes undefined. * Correct Fpscr handling for FP vector slow paths WIP * Add StandardFPSCRValue behaviour for all Arithmetic instructions * Add StandardFPSCRValue behaviour to compare instructions. * Force passing of fpcr to FPProcessException and FPUnpack. Reduces potential for code error significantly * OpCode cleanup * Remove urgency from DMB comment in MRRC DMB is currently a no-op via the instruction, so it should likely still be a no-op here. * Test Cleanup * Fix FPDefaultNaN on Ryzen CPUs * Improve some tests, fix some shift instructions, add slow path for Vadd * Fix Typo * More test cleanup * Flip order of Fx and index, to indicate that the operand's is the "base" * Remove Simd32 register type, use Int32 and Int64 for scalars like A64 does. * Reintroduce alignment to DecoderHelper (removed by accident) * One more realign as reading diffs is hard * Use I32 registers in A32 (part 2) Swap default integer register type based on current execution mode. * FPSCR flags as Registers (part 1) Still need to change NativeContext and ExecutionContext to allow getting/setting with the flag values. * Use I32 registers in A32 (part 1) * FPSCR flags as registers (part 2) Only CMP flags are on the registers right now. It could be useful to use more of the space in non-fast-float when implementing A32 flags accurately in the fast path. * Address Feedback * Correct FP->Int behaviour (should saturate) * Make branches made by writing to PC eligible for Rejit Greatly improves performance in most games. * Remove unused branching for Vtbl * RejitRequest as a class rather than a tuple Makes a lot more sense than storing tuples on a dictionary. * Add VMOVN, VSHR (imm), VSHRN (imm) and related tests * Re-order InstEmitSystem32 Alphabetical sorting. * Address Feedback Feedback from Ac_K, remove and sort usings. * Address Feedback 2 * Address Feedback from LDj3SNuD Opcode table reordered to have alphabetical sorting within groups, Vmaxnm and Vminnm have split names to be less ambiguous, SoftFloat nits, Test nits and Test simplification with ValueSource. * Add Debug Asserts to A32 helpers Mainly to prevent the shift ones from being used on I64 operands, as they expect I32 input for most operations (eg. carry flag setting), and expect I32 input for shift and boolean amounts. Most other helper functions don't take Operands, throw on out of range values, and take specific types of OpCode, so didn't need any asserts. * Use ConstF rather than creating an operand. (useful for pooling in future) * Move exclusive load to helper, reference call flag rather than literal 1. * Address LDj feedback (minus table flatten) one final look before it's all gone. the world is so beautiful. * Flatten OpCodeTable oh no * Address more table ordering * Call Flag as int on A32 Co-authored-by: Natalie C. <cyuubiapps@gmail.com> Co-authored-by: Thog <thog@protonmail.com>
2020-02-23 22:20:40 +01:00
if (size < 3 && value.Type == OperandType.I64)
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
{
value = context.ConvertI64ToI32(value);
}
switch (size)
{
case 0: vector = context.VectorInsert8 (vector, value, index); break;
case 1: vector = context.VectorInsert16(vector, value, index); break;
case 2: vector = context.VectorInsert (vector, value, index); break;
case 3: vector = context.VectorInsert (vector, value, index); break;
}
return vector;
}
Add most of the A32 instruction set to ARMeilleure (#897) * Implement TEQ and MOV (Imm16) * Initial work on A32 instructions + SVC. No tests yet, hangs in rtld. * Implement CLZ, fix BFI and BFC Now stops on SIMD initialization. * Exclusive access instructions, fix to mul, system instructions. Now gets to a break after SignalProcessWideKey64. * Better impl of UBFX, add UDIV and SDIV Now boots way further - now stuck on VMOV instruction. * Many more instructions, start on SIMD and testing framework. * Fix build issues * svc: Rework 32 bit codepath Fixing once and for all argument ordering issues. * Fix 32 bits stacktrace * hle debug: Add 32 bits dynamic section parsing * Fix highCq mode, add many tests, fix some instruction bugs Still suffers from critical malloc failure :weary: * Fix incorrect opcode decoders and a few more instructions. * Add a few instructions and fix others. re-disable highCq for now. Disabled the svc memory clear since i'm not sure about it. * Fix build * Fix typo in ordered/exclusive stores. * Implement some more instructions, fix others. Uxtab16/Sxtab16 are untested. * Begin impl of pairwise, some other instructions. * Add a few more instructions, a quick hack to fix svcs for now. * Add tests and fix issues with VTRN, VZIP, VUZP * Add a few more instructions, fix Vmul_1 encoding. * Fix way too many instruction bugs, add tests for some of the more important ones. * Fix HighCq, enable FastFP paths for some floating point instructions (not entirely sure why these were disabled, so important to note this commit exists) Branching has been removed in A32 shifts until I figure out if it's worth it * Cleanup Part 1 There should be no functional change between these next few commits. Should is the key word. (except for removing break handler) * Implement 32 bits syscalls Co-authored-by: riperiperi <rhy3756547@hotmail.com> Implement all 32 bits counterparts of the 64 bits syscalls we currently have. * Refactor part 2: Move index/subindex logic to Operand May have inadvertently fixed one (1) bug * Add FlushProcessDataCache32 * Address jd's comments * Remove 16 bit encodings from OpCodeTable Still need to catch some edge cases (operands that use the "F" flag) and make Q encodings with non-even indexes undefined. * Correct Fpscr handling for FP vector slow paths WIP * Add StandardFPSCRValue behaviour for all Arithmetic instructions * Add StandardFPSCRValue behaviour to compare instructions. * Force passing of fpcr to FPProcessException and FPUnpack. Reduces potential for code error significantly * OpCode cleanup * Remove urgency from DMB comment in MRRC DMB is currently a no-op via the instruction, so it should likely still be a no-op here. * Test Cleanup * Fix FPDefaultNaN on Ryzen CPUs * Improve some tests, fix some shift instructions, add slow path for Vadd * Fix Typo * More test cleanup * Flip order of Fx and index, to indicate that the operand's is the "base" * Remove Simd32 register type, use Int32 and Int64 for scalars like A64 does. * Reintroduce alignment to DecoderHelper (removed by accident) * One more realign as reading diffs is hard * Use I32 registers in A32 (part 2) Swap default integer register type based on current execution mode. * FPSCR flags as Registers (part 1) Still need to change NativeContext and ExecutionContext to allow getting/setting with the flag values. * Use I32 registers in A32 (part 1) * FPSCR flags as registers (part 2) Only CMP flags are on the registers right now. It could be useful to use more of the space in non-fast-float when implementing A32 flags accurately in the fast path. * Address Feedback * Correct FP->Int behaviour (should saturate) * Make branches made by writing to PC eligible for Rejit Greatly improves performance in most games. * Remove unused branching for Vtbl * RejitRequest as a class rather than a tuple Makes a lot more sense than storing tuples on a dictionary. * Add VMOVN, VSHR (imm), VSHRN (imm) and related tests * Re-order InstEmitSystem32 Alphabetical sorting. * Address Feedback Feedback from Ac_K, remove and sort usings. * Address Feedback 2 * Address Feedback from LDj3SNuD Opcode table reordered to have alphabetical sorting within groups, Vmaxnm and Vminnm have split names to be less ambiguous, SoftFloat nits, Test nits and Test simplification with ValueSource. * Add Debug Asserts to A32 helpers Mainly to prevent the shift ones from being used on I64 operands, as they expect I32 input for most operations (eg. carry flag setting), and expect I32 input for shift and boolean amounts. Most other helper functions don't take Operands, throw on out of range values, and take specific types of OpCode, so didn't need any asserts. * Use ConstF rather than creating an operand. (useful for pooling in future) * Move exclusive load to helper, reference call flag rather than literal 1. * Address LDj feedback (minus table flatten) one final look before it's all gone. the world is so beautiful. * Flatten OpCodeTable oh no * Address more table ordering * Call Flag as int on A32 Co-authored-by: Natalie C. <cyuubiapps@gmail.com> Co-authored-by: Thog <thog@protonmail.com>
2020-02-23 22:20:40 +01:00
public static void ThrowIfInvalid(int index, int size)
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 20:56:22 +02:00
{
if ((uint)size > 3u)
{
throw new ArgumentOutOfRangeException(nameof(size));
}
if ((uint)index >= 16u >> size)
{
throw new ArgumentOutOfRangeException(nameof(index));
}
}
}
}