RyuKen/ARMeilleure/Instructions/InstEmitSimdMove.cs
gdkchan a731ab3a2a 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 21:56:22 +03:00

794 lines
No EOL
26 KiB
C#

using ARMeilleure.Decoders;
using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.Translation;
using System;
using static ARMeilleure.Instructions.InstEmitHelper;
using static ARMeilleure.Instructions.InstEmitSimdHelper;
using static ARMeilleure.IntermediateRepresentation.OperandHelper;
namespace ARMeilleure.Instructions
{
static partial class InstEmit
{
#region "Masks"
private static readonly long[] _masksE0_TrnUzpXtn = new long[]
{
14L << 56 | 12L << 48 | 10L << 40 | 08L << 32 | 06L << 24 | 04L << 16 | 02L << 8 | 00L << 0,
13L << 56 | 12L << 48 | 09L << 40 | 08L << 32 | 05L << 24 | 04L << 16 | 01L << 8 | 00L << 0,
11L << 56 | 10L << 48 | 09L << 40 | 08L << 32 | 03L << 24 | 02L << 16 | 01L << 8 | 00L << 0
};
private static readonly long[] _masksE1_TrnUzp = new long[]
{
15L << 56 | 13L << 48 | 11L << 40 | 09L << 32 | 07L << 24 | 05L << 16 | 03L << 8 | 01L << 0,
15L << 56 | 14L << 48 | 11L << 40 | 10L << 32 | 07L << 24 | 06L << 16 | 03L << 8 | 02L << 0,
15L << 56 | 14L << 48 | 13L << 40 | 12L << 32 | 07L << 24 | 06L << 16 | 05L << 8 | 04L << 0
};
private static readonly long[] _masksE0_Uzp = new long[]
{
13L << 56 | 09L << 48 | 05L << 40 | 01L << 32 | 12L << 24 | 08L << 16 | 04L << 8 | 00L << 0,
11L << 56 | 10L << 48 | 03L << 40 | 02L << 32 | 09L << 24 | 08L << 16 | 01L << 8 | 00L << 0
};
private static readonly long[] _masksE1_Uzp = new long[]
{
15L << 56 | 11L << 48 | 07L << 40 | 03L << 32 | 14L << 24 | 10L << 16 | 06L << 8 | 02L << 0,
15L << 56 | 14L << 48 | 07L << 40 | 06L << 32 | 13L << 24 | 12L << 16 | 05L << 8 | 04L << 0
};
#endregion
public static void Dup_Gp(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
Operand n = GetIntOrZR(context, op.Rn);
if (Optimizations.UseSse2)
{
switch (op.Size)
{
case 0: n = context.ZeroExtend8 (n.Type, n); n = context.Multiply(n, Const(n.Type, 0x01010101)); break;
case 1: n = context.ZeroExtend16(n.Type, n); n = context.Multiply(n, Const(n.Type, 0x00010001)); break;
case 2: n = context.ZeroExtend32(n.Type, n); break;
}
Operand res = context.VectorInsert(context.VectorZero(), n, 0);
if (op.Size < 3)
{
if (op.RegisterSize == RegisterSize.Simd64)
{
res = context.AddIntrinsic(Intrinsic.X86Shufps, res, res, Const(0xf0));
}
else
{
res = context.AddIntrinsic(Intrinsic.X86Shufps, res, res, Const(0));
}
}
else
{
res = context.AddIntrinsic(Intrinsic.X86Movlhps, res, res);
}
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
res = EmitVectorInsert(context, res, n, index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
}
public static void Dup_S(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
Operand ne = EmitVectorExtractZx(context, op.Rn, op.DstIndex, op.Size);
context.Copy(GetVec(op.Rd), EmitVectorInsert(context, context.VectorZero(), ne, 0, op.Size));
}
public static void Dup_V(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
if (Optimizations.UseSse2)
{
Operand res = GetVec(op.Rn);
if (op.Size == 0)
{
if (op.DstIndex != 0)
{
res = context.AddIntrinsic(Intrinsic.X86Psrldq, res, Const(op.DstIndex));
}
res = context.AddIntrinsic(Intrinsic.X86Punpcklbw, res, res);
res = context.AddIntrinsic(Intrinsic.X86Punpcklwd, res, res);
res = context.AddIntrinsic(Intrinsic.X86Shufps, res, res, Const(0));
}
else if (op.Size == 1)
{
if (op.DstIndex != 0)
{
res = context.AddIntrinsic(Intrinsic.X86Psrldq, res, Const(op.DstIndex * 2));
}
res = context.AddIntrinsic(Intrinsic.X86Punpcklwd, res, res);
res = context.AddIntrinsic(Intrinsic.X86Shufps, res, res, Const(0));
}
else if (op.Size == 2)
{
int mask = op.DstIndex * 0b01010101;
res = context.AddIntrinsic(Intrinsic.X86Shufps, res, res, Const(mask));
}
else if (op.DstIndex == 0 && op.RegisterSize != RegisterSize.Simd64)
{
res = context.AddIntrinsic(Intrinsic.X86Movlhps, res, res);
}
else if (op.DstIndex == 1)
{
res = context.AddIntrinsic(Intrinsic.X86Movhlps, res, res);
}
if (op.RegisterSize == RegisterSize.Simd64)
{
res = context.VectorZeroUpper64(res);
}
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand ne = EmitVectorExtractZx(context, op.Rn, op.DstIndex, op.Size);
Operand res = context.VectorZero();
int elems = op.GetBytesCount() >> op.Size;
for (int index = 0; index < elems; index++)
{
res = EmitVectorInsert(context, res, ne, index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
}
public static void Ext_V(ArmEmitterContext context)
{
OpCodeSimdExt op = (OpCodeSimdExt)context.CurrOp;
if (Optimizations.UseSse2)
{
Operand nShifted = GetVec(op.Rn);
if (op.RegisterSize == RegisterSize.Simd64)
{
nShifted = context.AddIntrinsic(Intrinsic.X86Movlhps, nShifted, context.VectorZero());
}
nShifted = context.AddIntrinsic(Intrinsic.X86Psrldq, nShifted, Const(op.Imm4));
Operand mShifted = GetVec(op.Rm);
mShifted = context.AddIntrinsic(Intrinsic.X86Pslldq, mShifted, Const(op.GetBytesCount() - op.Imm4));
if (op.RegisterSize == RegisterSize.Simd64)
{
mShifted = context.AddIntrinsic(Intrinsic.X86Movlhps, mShifted, context.VectorZero());
}
Operand res = context.AddIntrinsic(Intrinsic.X86Por, nShifted, mShifted);
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand res = context.VectorZero();
int bytes = op.GetBytesCount();
int position = op.Imm4 & (bytes - 1);
for (int index = 0; index < bytes; index++)
{
int reg = op.Imm4 + index < bytes ? op.Rn : op.Rm;
Operand e = EmitVectorExtractZx(context, reg, position, 0);
position = (position + 1) & (bytes - 1);
res = EmitVectorInsert(context, res, e, index, 0);
}
context.Copy(GetVec(op.Rd), res);
}
}
public static void Fcsel_S(ArmEmitterContext context)
{
OpCodeSimdFcond op = (OpCodeSimdFcond)context.CurrOp;
Operand lblTrue = Label();
Operand lblEnd = Label();
Operand isTrue = InstEmitFlowHelper.GetCondTrue(context, op.Cond);
context.BranchIfTrue(lblTrue, isTrue);
OperandType type = op.Size == 0 ? OperandType.FP32 : OperandType.FP64;
Operand me = context.VectorExtract(type, GetVec(op.Rm), 0);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), me, 0));
context.Branch(lblEnd);
context.MarkLabel(lblTrue);
Operand ne = context.VectorExtract(type, GetVec(op.Rn), 0);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), ne, 0));
context.MarkLabel(lblEnd);
}
public static void Fmov_Ftoi(ArmEmitterContext context)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand ne = EmitVectorExtractZx(context, op.Rn, 0, op.Size + 2);
SetIntOrZR(context, op.Rd, ne);
}
public static void Fmov_Ftoi1(ArmEmitterContext context)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand ne = EmitVectorExtractZx(context, op.Rn, 1, 3);
SetIntOrZR(context, op.Rd, ne);
}
public static void Fmov_Itof(ArmEmitterContext context)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand n = GetIntOrZR(context, op.Rn);
context.Copy(GetVec(op.Rd), EmitVectorInsert(context, context.VectorZero(), n, 0, op.Size + 2));
}
public static void Fmov_Itof1(ArmEmitterContext context)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
Operand n = GetIntOrZR(context, op.Rn);
context.Copy(GetVec(op.Rd), EmitVectorInsert(context, GetVec(op.Rd), n, 1, 3));
}
public static void Fmov_S(ArmEmitterContext context)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
OperandType type = op.Size == 0 ? OperandType.FP32 : OperandType.FP64;
Operand ne = context.VectorExtract(type, GetVec(op.Rn), 0);
context.Copy(GetVec(op.Rd), context.VectorInsert(context.VectorZero(), ne, 0));
}
public static void Fmov_Si(ArmEmitterContext context)
{
OpCodeSimdFmov op = (OpCodeSimdFmov)context.CurrOp;
if (op.Size == 0)
{
context.Copy(GetVec(op.Rd), X86GetScalar(context, (int)op.Immediate));
}
else
{
context.Copy(GetVec(op.Rd), X86GetScalar(context, op.Immediate));
}
}
public static void Fmov_Vi(ArmEmitterContext context)
{
OpCodeSimdImm op = (OpCodeSimdImm)context.CurrOp;
Operand e = Const(op.Immediate);
Operand res = context.VectorZero();
int elems = op.RegisterSize == RegisterSize.Simd128 ? 4 : 2;
for (int index = 0; index < (elems >> op.Size); index++)
{
res = EmitVectorInsert(context, res, e, index, op.Size + 2);
}
context.Copy(GetVec(op.Rd), res);
}
public static void Ins_Gp(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
Operand d = GetVec(op.Rd);
Operand n = GetIntOrZR(context, op.Rn);
context.Copy(d, EmitVectorInsert(context, d, n, op.DstIndex, op.Size));
}
public static void Ins_V(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
Operand d = GetVec(op.Rd);
Operand ne = EmitVectorExtractZx(context, op.Rn, op.SrcIndex, op.Size);
context.Copy(d, EmitVectorInsert(context, d, ne, op.DstIndex, op.Size));
}
public static void Movi_V(ArmEmitterContext context)
{
if (Optimizations.UseSse2)
{
EmitMoviMvni(context, not: false);
}
else
{
EmitVectorImmUnaryOp(context, (op1) => op1);
}
}
public static void Mvni_V(ArmEmitterContext context)
{
if (Optimizations.UseSse2)
{
EmitMoviMvni(context, not: true);
}
else
{
EmitVectorImmUnaryOp(context, (op1) => context.BitwiseNot(op1));
}
}
public static void Smov_S(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
Operand ne = EmitVectorExtractSx(context, op.Rn, op.DstIndex, op.Size);
if (op.RegisterSize == RegisterSize.Simd64)
{
ne = context.ZeroExtend32(OperandType.I64, ne);
}
SetIntOrZR(context, op.Rd, ne);
}
public static void Tbl_V(ArmEmitterContext context)
{
OpCodeSimdTbl op = (OpCodeSimdTbl)context.CurrOp;
if (Optimizations.UseSsse3)
{
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
Operand mask = X86GetAllElements(context, 0x0F0F0F0F0F0F0F0FL);
Operand mMask = context.AddIntrinsic(Intrinsic.X86Pcmpgtb, m, mask);
mMask = context.AddIntrinsic(Intrinsic.X86Por, mMask, m);
Operand res = context.AddIntrinsic(Intrinsic.X86Pshufb, n, mMask);
for (int index = 1; index < op.Size; index++)
{
Operand ni = GetVec((op.Rn + index) & 0x1f);
Operand indexMask = X86GetAllElements(context, 0x1010101010101010L * index);
Operand mMinusMask = context.AddIntrinsic(Intrinsic.X86Psubb, m, indexMask);
Operand mMask2 = context.AddIntrinsic(Intrinsic.X86Pcmpgtb, mMinusMask, mask);
mMask2 = context.AddIntrinsic(Intrinsic.X86Por, mMask2, mMinusMask);
Operand res2 = context.AddIntrinsic(Intrinsic.X86Pshufb, ni, mMask2);
res = context.AddIntrinsic(Intrinsic.X86Por, res, res2);
}
if (op.RegisterSize == RegisterSize.Simd64)
{
res = context.VectorZeroUpper64(res);
}
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand[] args = new Operand[1 + op.Size];
args[0] = GetVec(op.Rm);
for (int index = 0; index < op.Size; index++)
{
args[1 + index] = GetVec((op.Rn + index) & 0x1f);
}
Delegate dlg = null;
switch (op.Size)
{
case 1: dlg = op.RegisterSize == RegisterSize.Simd64
? (Delegate)new _V128_V128_V128(SoftFallback.Tbl1_V64)
: (Delegate)new _V128_V128_V128(SoftFallback.Tbl1_V128); break;
case 2: dlg = op.RegisterSize == RegisterSize.Simd64
? (Delegate)new _V128_V128_V128_V128(SoftFallback.Tbl2_V64)
: (Delegate)new _V128_V128_V128_V128(SoftFallback.Tbl2_V128); break;
case 3: dlg = op.RegisterSize == RegisterSize.Simd64
? (Delegate)new _V128_V128_V128_V128_V128(SoftFallback.Tbl3_V64)
: (Delegate)new _V128_V128_V128_V128_V128(SoftFallback.Tbl3_V128); break;
case 4: dlg = op.RegisterSize == RegisterSize.Simd64
? (Delegate)new _V128_V128_V128_V128_V128_V128(SoftFallback.Tbl4_V64)
: (Delegate)new _V128_V128_V128_V128_V128_V128(SoftFallback.Tbl4_V128); break;
}
context.Copy(GetVec(op.Rd), context.Call(dlg, args));
}
}
public static void Trn1_V(ArmEmitterContext context)
{
EmitVectorTranspose(context, part: 0);
}
public static void Trn2_V(ArmEmitterContext context)
{
EmitVectorTranspose(context, part: 1);
}
public static void Umov_S(ArmEmitterContext context)
{
OpCodeSimdIns op = (OpCodeSimdIns)context.CurrOp;
Operand ne = EmitVectorExtractZx(context, op.Rn, op.DstIndex, op.Size);
SetIntOrZR(context, op.Rd, ne);
}
public static void Uzp1_V(ArmEmitterContext context)
{
EmitVectorUnzip(context, part: 0);
}
public static void Uzp2_V(ArmEmitterContext context)
{
EmitVectorUnzip(context, part: 1);
}
public static void Xtn_V(ArmEmitterContext context)
{
OpCodeSimd op = (OpCodeSimd)context.CurrOp;
if (Optimizations.UseSsse3)
{
Operand d = GetVec(op.Rd);
Operand res = context.AddIntrinsic(Intrinsic.X86Movlhps, d, context.VectorZero());
Operand n = GetVec(op.Rn);
Operand mask = X86GetAllElements(context, _masksE0_TrnUzpXtn[op.Size]);
Operand res2 = context.AddIntrinsic(Intrinsic.X86Pshufb, n, mask);
Intrinsic movInst = op.RegisterSize == RegisterSize.Simd128
? Intrinsic.X86Movlhps
: Intrinsic.X86Movhlps;
res = context.AddIntrinsic(movInst, res, res2);
context.Copy(GetVec(op.Rd), res);
}
else
{
int elems = 8 >> op.Size;
int part = op.RegisterSize == RegisterSize.Simd128 ? elems : 0;
Operand res = part == 0 ? context.VectorZero() : context.Copy(GetVec(op.Rd));
for (int index = 0; index < elems; index++)
{
Operand ne = EmitVectorExtractZx(context, op.Rn, index, op.Size + 1);
res = EmitVectorInsert(context, res, ne, part + index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
}
public static void Zip1_V(ArmEmitterContext context)
{
EmitVectorZip(context, part: 0);
}
public static void Zip2_V(ArmEmitterContext context)
{
EmitVectorZip(context, part: 1);
}
private static void EmitMoviMvni(ArmEmitterContext context, bool not)
{
OpCodeSimdImm op = (OpCodeSimdImm)context.CurrOp;
long imm = op.Immediate;
switch (op.Size)
{
case 0: imm *= 0x01010101; break;
case 1: imm *= 0x00010001; break;
}
if (not)
{
imm = ~imm;
}
Operand mask;
if (op.Size < 3)
{
mask = X86GetAllElements(context, (int)imm);
}
else
{
mask = X86GetAllElements(context, imm);
}
if (op.RegisterSize == RegisterSize.Simd64)
{
mask = context.VectorZeroUpper64(mask);
}
context.Copy(GetVec(op.Rd), mask);
}
private static void EmitVectorTranspose(ArmEmitterContext context, int part)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
if (Optimizations.UseSsse3)
{
Operand mask = null;
if (op.Size < 3)
{
long maskE0 = _masksE0_TrnUzpXtn[op.Size];
long maskE1 = _masksE1_TrnUzp [op.Size];
mask = X86GetScalar(context, maskE0);
mask = EmitVectorInsert(context, mask, Const(maskE1), 1, 3);
}
Operand n = GetVec(op.Rn);
if (op.Size < 3)
{
n = context.AddIntrinsic(Intrinsic.X86Pshufb, n, mask);
}
Operand m = GetVec(op.Rm);
if (op.Size < 3)
{
m = context.AddIntrinsic(Intrinsic.X86Pshufb, m, mask);
}
Intrinsic punpckInst = part == 0
? X86PunpcklInstruction[op.Size]
: X86PunpckhInstruction[op.Size];
Operand res = context.AddIntrinsic(punpckInst, n, m);
if (op.RegisterSize == RegisterSize.Simd64)
{
res = context.VectorZeroUpper64(res);
}
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand res = context.VectorZero();
int pairs = op.GetPairsCount() >> op.Size;
for (int index = 0; index < pairs; index++)
{
int pairIndex = index << 1;
Operand ne = EmitVectorExtractZx(context, op.Rn, pairIndex + part, op.Size);
Operand me = EmitVectorExtractZx(context, op.Rm, pairIndex + part, op.Size);
res = EmitVectorInsert(context, res, ne, pairIndex, op.Size);
res = EmitVectorInsert(context, res, me, pairIndex + 1, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
}
private static void EmitVectorUnzip(ArmEmitterContext context, int part)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
if (Optimizations.UseSsse3)
{
if (op.RegisterSize == RegisterSize.Simd128)
{
Operand mask = null;
if (op.Size < 3)
{
long maskE0 = _masksE0_TrnUzpXtn[op.Size];
long maskE1 = _masksE1_TrnUzp [op.Size];
mask = X86GetScalar(context, maskE0);
mask = EmitVectorInsert(context, mask, Const(maskE1), 1, 3);
}
Operand n = GetVec(op.Rn);
if (op.Size < 3)
{
n = context.AddIntrinsic(Intrinsic.X86Pshufb, n, mask);
}
Operand m = GetVec(op.Rm);
if (op.Size < 3)
{
m = context.AddIntrinsic(Intrinsic.X86Pshufb, m, mask);
}
Intrinsic punpckInst = part == 0
? Intrinsic.X86Punpcklqdq
: Intrinsic.X86Punpckhqdq;
Operand res = context.AddIntrinsic(punpckInst, n, m);
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
Intrinsic punpcklInst = X86PunpcklInstruction[op.Size];
Operand res = context.AddIntrinsic(punpcklInst, n, m);
if (op.Size < 2)
{
long maskE0 = _masksE0_Uzp[op.Size];
long maskE1 = _masksE1_Uzp[op.Size];
Operand mask = X86GetScalar(context, maskE0);
mask = EmitVectorInsert(context, mask, Const(maskE1), 1, 3);
res = context.AddIntrinsic(Intrinsic.X86Pshufb, res, mask);
}
Intrinsic punpckInst = part == 0
? Intrinsic.X86Punpcklqdq
: Intrinsic.X86Punpckhqdq;
res = context.AddIntrinsic(punpckInst, res, context.VectorZero());
context.Copy(GetVec(op.Rd), res);
}
}
else
{
Operand res = context.VectorZero();
int pairs = op.GetPairsCount() >> op.Size;
for (int index = 0; index < pairs; index++)
{
int idx = index << 1;
Operand ne = EmitVectorExtractZx(context, op.Rn, idx + part, op.Size);
Operand me = EmitVectorExtractZx(context, op.Rm, idx + part, op.Size);
res = EmitVectorInsert(context, res, ne, index, op.Size);
res = EmitVectorInsert(context, res, me, pairs + index, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
}
private static void EmitVectorZip(ArmEmitterContext context, int part)
{
OpCodeSimdReg op = (OpCodeSimdReg)context.CurrOp;
if (Optimizations.UseSse2)
{
Operand n = GetVec(op.Rn);
Operand m = GetVec(op.Rm);
if (op.RegisterSize == RegisterSize.Simd128)
{
Intrinsic punpckInst = part == 0
? X86PunpcklInstruction[op.Size]
: X86PunpckhInstruction[op.Size];
Operand res = context.AddIntrinsic(punpckInst, n, m);
context.Copy(GetVec(op.Rd), res);
}
else
{
Operand res = context.AddIntrinsic(X86PunpcklInstruction[op.Size], n, m);
Intrinsic punpckInst = part == 0
? Intrinsic.X86Punpcklqdq
: Intrinsic.X86Punpckhqdq;
res = context.AddIntrinsic(punpckInst, res, context.VectorZero());
context.Copy(GetVec(op.Rd), res);
}
}
else
{
Operand res = context.VectorZero();
int pairs = op.GetPairsCount() >> op.Size;
int baseIndex = part != 0 ? pairs : 0;
for (int index = 0; index < pairs; index++)
{
int pairIndex = index << 1;
Operand ne = EmitVectorExtractZx(context, op.Rn, baseIndex + index, op.Size);
Operand me = EmitVectorExtractZx(context, op.Rm, baseIndex + index, op.Size);
res = EmitVectorInsert(context, res, ne, pairIndex, op.Size);
res = EmitVectorInsert(context, res, me, pairIndex + 1, op.Size);
}
context.Copy(GetVec(op.Rd), res);
}
}
}
}