Ryujinx/ARMeilleure/CodeGen/X86/CodeGenerator.cs
FICTURE7 22b2cb39af
Reduce JIT GC allocations (#2515)
* Turn `MemoryOperand` into a struct

* Remove `IntrinsicOperation`

* Remove `PhiNode`

* Remove `Node`

* Turn `Operand` into a struct

* Turn `Operation` into a struct

* Clean up pool management methods

* Add `Arena` allocator

* Move `OperationHelper` to `Operation.Factory`

* Move `OperandHelper` to `Operand.Factory`

* Optimize `Operation` a bit

* Fix `Arena` initialization

* Rename `NativeList<T>` to `ArenaList<T>`

* Reduce `Operand` size from 88 to 56 bytes

* Reduce `Operation` size from 56 to 40 bytes

* Add optimistic interning of Register & Constant operands

* Optimize `RegisterUsage` pass a bit

* Optimize `RemoveUnusedNodes` pass a bit

Iterating in reverse-order allows killing dependency chains in a single
pass.

* Fix PPTC symbols

* Optimize `BasicBlock` a bit

Reduce allocations from `_successor` & `DominanceFrontiers`

* Fix `Operation` resize

* Make `Arena` expandable

Change the arena allocator to be expandable by allocating in pages, with
some of them being pooled. Currently 32 pages are pooled. An LRU removal
mechanism should probably be added to it.

Apparently MHR can allocate bitmaps large enough to exceed the 16MB
limit for the type.

* Move `Arena` & `ArenaList` to `Common`

* Remove `ThreadStaticPool` & co

* Add `PhiOperation`

* Reduce `Operand` size from 56 from 48 bytes

* Add linear-probing to `Operand` intern table

* Optimize `HybridAllocator` a bit

* Add `Allocators` class

* Tune `ArenaAllocator` sizes

* Add page removal mechanism to `ArenaAllocator`

Remove pages which have not been used for more than 5s after each reset.

I am on fence if this would be better using a Gen2 callback object like
the one in System.Buffers.ArrayPool<T>, to trim the pool. Because right
now if a large translation happens, the pages will be freed only after a
reset. This reset may not happen for a while because no new translation
is hit, but the arena base sizes are rather small.

* Fix `OOM` when allocating larger than page size in `ArenaAllocator`

Tweak resizing mechanism for Operand.Uses and Assignemnts.

* Optimize `Optimizer` a bit

* Optimize `Operand.Add<T>/Remove<T>` a bit

* Clean up `PreAllocator`

* Fix phi insertion order

Reduce codegen diffs.

* Fix code alignment

* Use new heuristics for degree of parallelism

* Suppress warnings

* Address gdkchan's feedback

Renamed `GetValue()` to `GetValueUnsafe()` to make it more clear that
`Operand.Value` should usually not be modified directly.

* Add fast path to `ArenaAllocator`

* Assembly for `ArenaAllocator.Allocate(ulong)`:

  .L0:
    mov rax, [rcx+0x18]
    lea r8, [rax+rdx]
    cmp r8, [rcx+0x10]
    ja short .L2
  .L1:
    mov rdx, [rcx+8]
    add rax, [rdx+8]
    mov [rcx+0x18], r8
    ret
  .L2:
    jmp ArenaAllocator.AllocateSlow(UInt64)

  A few variable/field had to be changed to ulong so that RyuJIT avoids
  emitting zero-extends.

* Implement a new heuristic to free pooled pages.

  If an arena is used often, it is more likely that its pages will be
  needed, so the pages are kept for longer (e.g: during PPTC rebuild or
  burst sof compilations). If is not used often, then it is more likely
  that its pages will not be needed (e.g: after PPTC rebuild or bursts
  of compilations).

* Address riperiperi's feedback

* Use `EqualityComparer<T>` in `IntrusiveList<T>`

Avoids a potential GC hole in `Equals(T, T)`.
2021-08-17 15:08:34 -03:00

1855 lines
No EOL
67 KiB
C#

using ARMeilleure.CodeGen.Optimizations;
using ARMeilleure.CodeGen.RegisterAllocators;
using ARMeilleure.CodeGen.Unwinding;
using ARMeilleure.Common;
using ARMeilleure.Diagnostics;
using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.Translation;
using ARMeilleure.Translation.PTC;
using System;
using System.Collections.Generic;
using System.Diagnostics;
using System.IO;
using System.Numerics;
using static ARMeilleure.IntermediateRepresentation.Operand.Factory;
namespace ARMeilleure.CodeGen.X86
{
static class CodeGenerator
{
private const int PageSize = 0x1000;
private const int StackGuardSize = 0x2000;
private static Action<CodeGenContext, Operation>[] _instTable;
static CodeGenerator()
{
_instTable = new Action<CodeGenContext, Operation>[EnumUtils.GetCount(typeof(Instruction))];
Add(Instruction.Add, GenerateAdd);
Add(Instruction.BitwiseAnd, GenerateBitwiseAnd);
Add(Instruction.BitwiseExclusiveOr, GenerateBitwiseExclusiveOr);
Add(Instruction.BitwiseNot, GenerateBitwiseNot);
Add(Instruction.BitwiseOr, GenerateBitwiseOr);
Add(Instruction.BranchIf, GenerateBranchIf);
Add(Instruction.ByteSwap, GenerateByteSwap);
Add(Instruction.Call, GenerateCall);
Add(Instruction.Clobber, GenerateClobber);
Add(Instruction.Compare, GenerateCompare);
Add(Instruction.CompareAndSwap, GenerateCompareAndSwap);
Add(Instruction.CompareAndSwap16, GenerateCompareAndSwap16);
Add(Instruction.CompareAndSwap8, GenerateCompareAndSwap8);
Add(Instruction.ConditionalSelect, GenerateConditionalSelect);
Add(Instruction.ConvertI64ToI32, GenerateConvertI64ToI32);
Add(Instruction.ConvertToFP, GenerateConvertToFP);
Add(Instruction.Copy, GenerateCopy);
Add(Instruction.CountLeadingZeros, GenerateCountLeadingZeros);
Add(Instruction.Divide, GenerateDivide);
Add(Instruction.DivideUI, GenerateDivideUI);
Add(Instruction.Fill, GenerateFill);
Add(Instruction.Load, GenerateLoad);
Add(Instruction.Load16, GenerateLoad16);
Add(Instruction.Load8, GenerateLoad8);
Add(Instruction.Multiply, GenerateMultiply);
Add(Instruction.Multiply64HighSI, GenerateMultiply64HighSI);
Add(Instruction.Multiply64HighUI, GenerateMultiply64HighUI);
Add(Instruction.Negate, GenerateNegate);
Add(Instruction.Return, GenerateReturn);
Add(Instruction.RotateRight, GenerateRotateRight);
Add(Instruction.ShiftLeft, GenerateShiftLeft);
Add(Instruction.ShiftRightSI, GenerateShiftRightSI);
Add(Instruction.ShiftRightUI, GenerateShiftRightUI);
Add(Instruction.SignExtend16, GenerateSignExtend16);
Add(Instruction.SignExtend32, GenerateSignExtend32);
Add(Instruction.SignExtend8, GenerateSignExtend8);
Add(Instruction.Spill, GenerateSpill);
Add(Instruction.SpillArg, GenerateSpillArg);
Add(Instruction.StackAlloc, GenerateStackAlloc);
Add(Instruction.Store, GenerateStore);
Add(Instruction.Store16, GenerateStore16);
Add(Instruction.Store8, GenerateStore8);
Add(Instruction.Subtract, GenerateSubtract);
Add(Instruction.Tailcall, GenerateTailcall);
Add(Instruction.VectorCreateScalar, GenerateVectorCreateScalar);
Add(Instruction.VectorExtract, GenerateVectorExtract);
Add(Instruction.VectorExtract16, GenerateVectorExtract16);
Add(Instruction.VectorExtract8, GenerateVectorExtract8);
Add(Instruction.VectorInsert, GenerateVectorInsert);
Add(Instruction.VectorInsert16, GenerateVectorInsert16);
Add(Instruction.VectorInsert8, GenerateVectorInsert8);
Add(Instruction.VectorOne, GenerateVectorOne);
Add(Instruction.VectorZero, GenerateVectorZero);
Add(Instruction.VectorZeroUpper64, GenerateVectorZeroUpper64);
Add(Instruction.VectorZeroUpper96, GenerateVectorZeroUpper96);
Add(Instruction.ZeroExtend16, GenerateZeroExtend16);
Add(Instruction.ZeroExtend32, GenerateZeroExtend32);
Add(Instruction.ZeroExtend8, GenerateZeroExtend8);
}
private static void Add(Instruction inst, Action<CodeGenContext, Operation> func)
{
_instTable[(int)inst] = func;
}
public static CompiledFunction Generate(CompilerContext cctx, PtcInfo ptcInfo = null)
{
ControlFlowGraph cfg = cctx.Cfg;
Logger.StartPass(PassName.Optimization);
if ((cctx.Options & CompilerOptions.SsaForm) != 0 &&
(cctx.Options & CompilerOptions.Optimize) != 0)
{
Optimizer.RunPass(cfg);
}
X86Optimizer.RunPass(cfg);
BlockPlacement.RunPass(cfg);
Logger.EndPass(PassName.Optimization, cfg);
Logger.StartPass(PassName.PreAllocation);
StackAllocator stackAlloc = new StackAllocator();
PreAllocator.RunPass(cctx, stackAlloc, out int maxCallArgs);
Logger.EndPass(PassName.PreAllocation, cfg);
Logger.StartPass(PassName.RegisterAllocation);
if ((cctx.Options & CompilerOptions.SsaForm) != 0)
{
Ssa.Deconstruct(cfg);
}
IRegisterAllocator regAlloc;
if ((cctx.Options & CompilerOptions.Lsra) != 0)
{
regAlloc = new LinearScanAllocator();
}
else
{
regAlloc = new HybridAllocator();
}
RegisterMasks regMasks = new RegisterMasks(
CallingConvention.GetIntAvailableRegisters(),
CallingConvention.GetVecAvailableRegisters(),
CallingConvention.GetIntCallerSavedRegisters(),
CallingConvention.GetVecCallerSavedRegisters(),
CallingConvention.GetIntCalleeSavedRegisters(),
CallingConvention.GetVecCalleeSavedRegisters());
AllocationResult allocResult = regAlloc.RunPass(cfg, stackAlloc, regMasks);
Logger.EndPass(PassName.RegisterAllocation, cfg);
Logger.StartPass(PassName.CodeGeneration);
using (MemoryStream stream = new MemoryStream())
{
CodeGenContext context = new CodeGenContext(stream, allocResult, maxCallArgs, cfg.Blocks.Count, ptcInfo);
UnwindInfo unwindInfo = WritePrologue(context);
ptcInfo?.WriteUnwindInfo(unwindInfo);
for (BasicBlock block = cfg.Blocks.First; block != null; block = block.ListNext)
{
context.EnterBlock(block);
for (Operation node = block.Operations.First; node != default; node = node.ListNext)
{
GenerateOperation(context, node);
}
if (block.SuccessorsCount == 0)
{
// The only blocks which can have 0 successors are exit blocks.
Operation last = block.Operations.Last;
Debug.Assert(last.Instruction == Instruction.Tailcall ||
last.Instruction == Instruction.Return);
}
else
{
BasicBlock succ = block.GetSuccessor(0);
if (succ != block.ListNext)
{
context.JumpTo(succ);
}
}
}
byte[] code = context.GetCode();
if (ptcInfo != null)
{
ptcInfo.Code = code;
}
Logger.EndPass(PassName.CodeGeneration);
return new CompiledFunction(code, unwindInfo);
}
}
private static void GenerateOperation(CodeGenContext context, Operation operation)
{
if (operation.Instruction == Instruction.Extended)
{
IntrinsicInfo info = IntrinsicTable.GetInfo(operation.Intrinsic);
switch (info.Type)
{
case IntrinsicType.Comis_:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
switch (operation.Intrinsic)
{
case Intrinsic.X86Comisdeq:
context.Assembler.Comisd(src1, src2);
context.Assembler.Setcc(dest, X86Condition.Equal);
break;
case Intrinsic.X86Comisdge:
context.Assembler.Comisd(src1, src2);
context.Assembler.Setcc(dest, X86Condition.AboveOrEqual);
break;
case Intrinsic.X86Comisdlt:
context.Assembler.Comisd(src1, src2);
context.Assembler.Setcc(dest, X86Condition.Below);
break;
case Intrinsic.X86Comisseq:
context.Assembler.Comiss(src1, src2);
context.Assembler.Setcc(dest, X86Condition.Equal);
break;
case Intrinsic.X86Comissge:
context.Assembler.Comiss(src1, src2);
context.Assembler.Setcc(dest, X86Condition.AboveOrEqual);
break;
case Intrinsic.X86Comisslt:
context.Assembler.Comiss(src1, src2);
context.Assembler.Setcc(dest, X86Condition.Below);
break;
}
context.Assembler.Movzx8(dest, dest, OperandType.I32);
break;
}
case IntrinsicType.Mxcsr:
{
Operand offset = operation.GetSource(0);
Operand bits = operation.GetSource(1);
Debug.Assert(offset.Kind == OperandKind.Constant && bits.Kind == OperandKind.Constant);
Debug.Assert(offset.Type == OperandType.I32 && bits.Type == OperandType.I32);
int offs = offset.AsInt32() + context.CallArgsRegionSize;
Operand rsp = Register(X86Register.Rsp);
Operand memOp = MemoryOp(OperandType.I32, rsp, default, Multiplier.x1, offs);
Debug.Assert(HardwareCapabilities.SupportsSse || HardwareCapabilities.SupportsVexEncoding);
context.Assembler.Stmxcsr(memOp);
if (operation.Intrinsic == Intrinsic.X86Mxcsrmb)
{
context.Assembler.Or(memOp, bits, OperandType.I32);
}
else /* if (intrinOp.Intrinsic == Intrinsic.X86Mxcsrub) */
{
Operand notBits = Const(~bits.AsInt32());
context.Assembler.And(memOp, notBits, OperandType.I32);
}
context.Assembler.Ldmxcsr(memOp);
break;
}
case IntrinsicType.PopCount:
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
EnsureSameType(dest, source);
Debug.Assert(dest.Type.IsInteger());
context.Assembler.Popcnt(dest, source, dest.Type);
break;
}
case IntrinsicType.Unary:
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
EnsureSameType(dest, source);
Debug.Assert(!dest.Type.IsInteger());
context.Assembler.WriteInstruction(info.Inst, dest, source);
break;
}
case IntrinsicType.UnaryToGpr:
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && !source.Type.IsInteger());
if (operation.Intrinsic == Intrinsic.X86Cvtsi2si)
{
if (dest.Type == OperandType.I32)
{
context.Assembler.Movd(dest, source); // int _mm_cvtsi128_si32(__m128i a)
}
else /* if (dest.Type == OperandType.I64) */
{
context.Assembler.Movq(dest, source); // __int64 _mm_cvtsi128_si64(__m128i a)
}
}
else
{
context.Assembler.WriteInstruction(info.Inst, dest, source, dest.Type);
}
break;
}
case IntrinsicType.Binary:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
EnsureSameType(dest, src1);
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(!dest.Type.IsInteger());
Debug.Assert(!src2.Type.IsInteger() || src2.Kind == OperandKind.Constant);
context.Assembler.WriteInstruction(info.Inst, dest, src1, src2);
break;
}
case IntrinsicType.BinaryGpr:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
EnsureSameType(dest, src1);
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(!dest.Type.IsInteger() && src2.Type.IsInteger());
context.Assembler.WriteInstruction(info.Inst, dest, src1, src2, src2.Type);
break;
}
case IntrinsicType.Crc32:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
EnsureSameReg(dest, src1);
Debug.Assert(dest.Type.IsInteger() && src1.Type.IsInteger() && src2.Type.IsInteger());
context.Assembler.WriteInstruction(info.Inst, dest, src2, dest.Type);
break;
}
case IntrinsicType.BinaryImm:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
EnsureSameType(dest, src1);
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(!dest.Type.IsInteger() && src2.Kind == OperandKind.Constant);
context.Assembler.WriteInstruction(info.Inst, dest, src1, src2.AsByte());
break;
}
case IntrinsicType.Ternary:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
EnsureSameType(dest, src1, src2, src3);
Debug.Assert(!dest.Type.IsInteger());
if (info.Inst == X86Instruction.Blendvpd && HardwareCapabilities.SupportsVexEncoding)
{
context.Assembler.WriteInstruction(X86Instruction.Vblendvpd, dest, src1, src2, src3);
}
else if (info.Inst == X86Instruction.Blendvps && HardwareCapabilities.SupportsVexEncoding)
{
context.Assembler.WriteInstruction(X86Instruction.Vblendvps, dest, src1, src2, src3);
}
else if (info.Inst == X86Instruction.Pblendvb && HardwareCapabilities.SupportsVexEncoding)
{
context.Assembler.WriteInstruction(X86Instruction.Vpblendvb, dest, src1, src2, src3);
}
else
{
EnsureSameReg(dest, src1);
Debug.Assert(src3.GetRegister().Index == 0);
context.Assembler.WriteInstruction(info.Inst, dest, src1, src2);
}
break;
}
case IntrinsicType.TernaryImm:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
EnsureSameType(dest, src1, src2);
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(!dest.Type.IsInteger() && src3.Kind == OperandKind.Constant);
context.Assembler.WriteInstruction(info.Inst, dest, src1, src2, src3.AsByte());
break;
}
case IntrinsicType.Fma:
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
Debug.Assert(HardwareCapabilities.SupportsVexEncoding);
Debug.Assert(dest.Kind == OperandKind.Register && src1.Kind == OperandKind.Register && src2.Kind == OperandKind.Register);
Debug.Assert(src3.Kind == OperandKind.Register || src3.Kind == OperandKind.Memory);
EnsureSameType(dest, src1, src2, src3);
Debug.Assert(dest.Type == OperandType.V128);
Debug.Assert(dest.Value == src1.Value);
context.Assembler.WriteInstruction(info.Inst, dest, src2, src3);
break;
}
}
}
else
{
Action<CodeGenContext, Operation> func = _instTable[(int)operation.Instruction];
if (func != null)
{
func(context, operation);
}
else
{
throw new ArgumentException($"Invalid instruction \"{operation.Instruction}\".");
}
}
}
private static void GenerateAdd(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
if (dest.Type.IsInteger())
{
// If Destination and Source 1 Operands are the same, perform a standard add as there are no benefits to using LEA.
if (dest.Kind == src1.Kind && dest.Value == src1.Value)
{
ValidateBinOp(dest, src1, src2);
context.Assembler.Add(dest, src2, dest.Type);
}
else
{
EnsureSameType(dest, src1, src2);
int offset;
Operand index;
if (src2.Kind == OperandKind.Constant)
{
offset = src2.AsInt32();
index = default;
}
else
{
offset = 0;
index = src2;
}
Operand memOp = MemoryOp(dest.Type, src1, index, Multiplier.x1, offset);
context.Assembler.Lea(dest, memOp, dest.Type);
}
}
else
{
ValidateBinOp(dest, src1, src2);
if (dest.Type == OperandType.FP32)
{
context.Assembler.Addss(dest, src1, src2);
}
else /* if (dest.Type == OperandType.FP64) */
{
context.Assembler.Addsd(dest, src1, src2);
}
}
}
private static void GenerateBitwiseAnd(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateBinOp(dest, src1, src2);
Debug.Assert(dest.Type.IsInteger());
// Note: GenerateCompareCommon makes the assumption that BitwiseAnd will emit only a single `and`
// instruction.
context.Assembler.And(dest, src2, dest.Type);
}
private static void GenerateBitwiseExclusiveOr(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateBinOp(dest, src1, src2);
if (dest.Type.IsInteger())
{
context.Assembler.Xor(dest, src2, dest.Type);
}
else
{
context.Assembler.Xorps(dest, src1, src2);
}
}
private static void GenerateBitwiseNot(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
ValidateUnOp(dest, source);
Debug.Assert(dest.Type.IsInteger());
context.Assembler.Not(dest);
}
private static void GenerateBitwiseOr(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateBinOp(dest, src1, src2);
Debug.Assert(dest.Type.IsInteger());
context.Assembler.Or(dest, src2, dest.Type);
}
private static void GenerateBranchIf(CodeGenContext context, Operation operation)
{
Operand comp = operation.GetSource(2);
Debug.Assert(comp.Kind == OperandKind.Constant);
var cond = ((Comparison)comp.AsInt32()).ToX86Condition();
GenerateCompareCommon(context, operation);
context.JumpTo(cond, context.CurrBlock.GetSuccessor(1));
}
private static void GenerateByteSwap(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
ValidateUnOp(dest, source);
Debug.Assert(dest.Type.IsInteger());
context.Assembler.Bswap(dest);
}
private static void GenerateCall(CodeGenContext context, Operation operation)
{
context.Assembler.Call(operation.GetSource(0));
}
private static void GenerateClobber(CodeGenContext context, Operation operation)
{
// This is only used to indicate that a register is clobbered to the
// register allocator, we don't need to produce any code.
}
private static void GenerateCompare(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand comp = operation.GetSource(2);
Debug.Assert(dest.Type == OperandType.I32);
Debug.Assert(comp.Kind == OperandKind.Constant);
var cond = ((Comparison)comp.AsInt32()).ToX86Condition();
GenerateCompareCommon(context, operation);
context.Assembler.Setcc(dest, cond);
context.Assembler.Movzx8(dest, dest, OperandType.I32);
}
private static void GenerateCompareCommon(CodeGenContext context, Operation operation)
{
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
EnsureSameType(src1, src2);
Debug.Assert(src1.Type.IsInteger());
if (src2.Kind == OperandKind.Constant && src2.Value == 0)
{
if (MatchOperation(operation.ListPrevious, Instruction.BitwiseAnd, src1.Type, src1.GetRegister()))
{
// Since the `test` and `and` instruction set the status flags in the same way, we can omit the
// `test r,r` instruction when it is immediately preceded by an `and r,*` instruction.
//
// For example:
//
// and eax, 0x3
// test eax, eax
// jz .L0
//
// =>
//
// and eax, 0x3
// jz .L0
}
else
{
context.Assembler.Test(src1, src1, src1.Type);
}
}
else
{
context.Assembler.Cmp(src1, src2, src1.Type);
}
}
private static void GenerateCompareAndSwap(CodeGenContext context, Operation operation)
{
Operand src1 = operation.GetSource(0);
if (operation.SourcesCount == 5) // CompareAndSwap128 has 5 sources, compared to CompareAndSwap64/32's 3.
{
Operand memOp = MemoryOp(OperandType.I64, src1);
context.Assembler.Cmpxchg16b(memOp);
}
else
{
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
EnsureSameType(src2, src3);
Operand memOp = MemoryOp(src3.Type, src1);
context.Assembler.Cmpxchg(memOp, src3);
}
}
private static void GenerateCompareAndSwap16(CodeGenContext context, Operation operation)
{
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
EnsureSameType(src2, src3);
Operand memOp = MemoryOp(src3.Type, src1);
context.Assembler.Cmpxchg16(memOp, src3);
}
private static void GenerateCompareAndSwap8(CodeGenContext context, Operation operation)
{
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
EnsureSameType(src2, src3);
Operand memOp = MemoryOp(src3.Type, src1);
context.Assembler.Cmpxchg8(memOp, src3);
}
private static void GenerateConditionalSelect(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
Operand src3 = operation.GetSource(2);
EnsureSameReg (dest, src3);
EnsureSameType(dest, src2, src3);
Debug.Assert(dest.Type.IsInteger());
Debug.Assert(src1.Type == OperandType.I32);
context.Assembler.Test (src1, src1, src1.Type);
context.Assembler.Cmovcc(dest, src2, dest.Type, X86Condition.NotEqual);
}
private static void GenerateConvertI64ToI32(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type == OperandType.I32 && source.Type == OperandType.I64);
context.Assembler.Mov(dest, source, OperandType.I32);
}
private static void GenerateConvertToFP(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type == OperandType.FP32 || dest.Type == OperandType.FP64);
if (dest.Type == OperandType.FP32)
{
Debug.Assert(source.Type.IsInteger() || source.Type == OperandType.FP64);
if (source.Type.IsInteger())
{
context.Assembler.Xorps (dest, dest, dest);
context.Assembler.Cvtsi2ss(dest, dest, source, source.Type);
}
else /* if (source.Type == OperandType.FP64) */
{
context.Assembler.Cvtsd2ss(dest, dest, source);
GenerateZeroUpper96(context, dest, dest);
}
}
else /* if (dest.Type == OperandType.FP64) */
{
Debug.Assert(source.Type.IsInteger() || source.Type == OperandType.FP32);
if (source.Type.IsInteger())
{
context.Assembler.Xorps (dest, dest, dest);
context.Assembler.Cvtsi2sd(dest, dest, source, source.Type);
}
else /* if (source.Type == OperandType.FP32) */
{
context.Assembler.Cvtss2sd(dest, dest, source);
GenerateZeroUpper64(context, dest, dest);
}
}
}
private static void GenerateCopy(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
EnsureSameType(dest, source);
Debug.Assert(dest.Type.IsInteger() || source.Kind != OperandKind.Constant);
// Moves to the same register are useless.
if (dest.Kind == source.Kind && dest.Value == source.Value)
{
return;
}
if (dest.Kind == OperandKind.Register &&
source.Kind == OperandKind.Constant && source.Value == 0)
{
// Assemble "mov reg, 0" as "xor reg, reg" as the later is more efficient.
context.Assembler.Xor(dest, dest, OperandType.I32);
}
else if (dest.Type.IsInteger())
{
context.Assembler.Mov(dest, source, dest.Type);
}
else
{
context.Assembler.Movdqu(dest, source);
}
}
private static void GenerateCountLeadingZeros(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
EnsureSameType(dest, source);
Debug.Assert(dest.Type.IsInteger());
context.Assembler.Bsr(dest, source, dest.Type);
int operandSize = dest.Type == OperandType.I32 ? 32 : 64;
int operandMask = operandSize - 1;
// When the input operand is 0, the result is undefined, however the
// ZF flag is set. We are supposed to return the operand size on that
// case. So, add an additional jump to handle that case, by moving the
// operand size constant to the destination register.
context.JumpToNear(X86Condition.NotEqual);
context.Assembler.Mov(dest, Const(operandSize | operandMask), OperandType.I32);
context.JumpHere();
// BSR returns the zero based index of the last bit set on the operand,
// starting from the least significant bit. However we are supposed to
// return the number of 0 bits on the high end. So, we invert the result
// of the BSR using XOR to get the correct value.
context.Assembler.Xor(dest, Const(operandMask), OperandType.I32);
}
private static void GenerateDivide(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand dividend = operation.GetSource(0);
Operand divisor = operation.GetSource(1);
if (!dest.Type.IsInteger())
{
ValidateBinOp(dest, dividend, divisor);
}
if (dest.Type.IsInteger())
{
divisor = operation.GetSource(2);
EnsureSameType(dest, divisor);
if (divisor.Type == OperandType.I32)
{
context.Assembler.Cdq();
}
else
{
context.Assembler.Cqo();
}
context.Assembler.Idiv(divisor);
}
else if (dest.Type == OperandType.FP32)
{
context.Assembler.Divss(dest, dividend, divisor);
}
else /* if (dest.Type == OperandType.FP64) */
{
context.Assembler.Divsd(dest, dividend, divisor);
}
}
private static void GenerateDivideUI(CodeGenContext context, Operation operation)
{
Operand divisor = operation.GetSource(2);
Operand rdx = Register(X86Register.Rdx);
Debug.Assert(divisor.Type.IsInteger());
context.Assembler.Xor(rdx, rdx, OperandType.I32);
context.Assembler.Div(divisor);
}
private static void GenerateFill(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand offset = operation.GetSource(0);
Debug.Assert(offset.Kind == OperandKind.Constant);
int offs = offset.AsInt32() + context.CallArgsRegionSize;
Operand rsp = Register(X86Register.Rsp);
Operand memOp = MemoryOp(dest.Type, rsp, default, Multiplier.x1, offs);
GenerateLoad(context, memOp, dest);
}
private static void GenerateLoad(CodeGenContext context, Operation operation)
{
Operand value = operation.Destination;
Operand address = Memory(operation.GetSource(0), value.Type);
GenerateLoad(context, address, value);
}
private static void GenerateLoad16(CodeGenContext context, Operation operation)
{
Operand value = operation.Destination;
Operand address = Memory(operation.GetSource(0), value.Type);
Debug.Assert(value.Type.IsInteger());
context.Assembler.Movzx16(value, address, value.Type);
}
private static void GenerateLoad8(CodeGenContext context, Operation operation)
{
Operand value = operation.Destination;
Operand address = Memory(operation.GetSource(0), value.Type);
Debug.Assert(value.Type.IsInteger());
context.Assembler.Movzx8(value, address, value.Type);
}
private static void GenerateMultiply(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
if (src2.Kind != OperandKind.Constant)
{
EnsureSameReg(dest, src1);
}
EnsureSameType(dest, src1, src2);
if (dest.Type.IsInteger())
{
if (src2.Kind == OperandKind.Constant)
{
context.Assembler.Imul(dest, src1, src2, dest.Type);
}
else
{
context.Assembler.Imul(dest, src2, dest.Type);
}
}
else if (dest.Type == OperandType.FP32)
{
context.Assembler.Mulss(dest, src1, src2);
}
else /* if (dest.Type == OperandType.FP64) */
{
context.Assembler.Mulsd(dest, src1, src2);
}
}
private static void GenerateMultiply64HighSI(CodeGenContext context, Operation operation)
{
Operand source = operation.GetSource(1);
Debug.Assert(source.Type == OperandType.I64);
context.Assembler.Imul(source);
}
private static void GenerateMultiply64HighUI(CodeGenContext context, Operation operation)
{
Operand source = operation.GetSource(1);
Debug.Assert(source.Type == OperandType.I64);
context.Assembler.Mul(source);
}
private static void GenerateNegate(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
ValidateUnOp(dest, source);
Debug.Assert(dest.Type.IsInteger());
context.Assembler.Neg(dest);
}
private static void GenerateReturn(CodeGenContext context, Operation operation)
{
WriteEpilogue(context);
context.Assembler.Return();
}
private static void GenerateRotateRight(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateShift(dest, src1, src2);
context.Assembler.Ror(dest, src2, dest.Type);
}
private static void GenerateShiftLeft(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateShift(dest, src1, src2);
context.Assembler.Shl(dest, src2, dest.Type);
}
private static void GenerateShiftRightSI(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateShift(dest, src1, src2);
context.Assembler.Sar(dest, src2, dest.Type);
}
private static void GenerateShiftRightUI(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateShift(dest, src1, src2);
context.Assembler.Shr(dest, src2, dest.Type);
}
private static void GenerateSignExtend16(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && source.Type.IsInteger());
context.Assembler.Movsx16(dest, source, dest.Type);
}
private static void GenerateSignExtend32(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && source.Type.IsInteger());
context.Assembler.Movsx32(dest, source, dest.Type);
}
private static void GenerateSignExtend8(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && source.Type.IsInteger());
context.Assembler.Movsx8(dest, source, dest.Type);
}
private static void GenerateSpill(CodeGenContext context, Operation operation)
{
GenerateSpill(context, operation, context.CallArgsRegionSize);
}
private static void GenerateSpillArg(CodeGenContext context, Operation operation)
{
GenerateSpill(context, operation, 0);
}
private static void GenerateSpill(CodeGenContext context, Operation operation, int baseOffset)
{
Operand offset = operation.GetSource(0);
Operand source = operation.GetSource(1);
Debug.Assert(offset.Kind == OperandKind.Constant);
int offs = offset.AsInt32() + baseOffset;
Operand rsp = Register(X86Register.Rsp);
Operand memOp = MemoryOp(source.Type, rsp, default, Multiplier.x1, offs);
GenerateStore(context, memOp, source);
}
private static void GenerateStackAlloc(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand offset = operation.GetSource(0);
Debug.Assert(offset.Kind == OperandKind.Constant);
int offs = offset.AsInt32() + context.CallArgsRegionSize;
Operand rsp = Register(X86Register.Rsp);
Operand memOp = MemoryOp(OperandType.I64, rsp, default, Multiplier.x1, offs);
context.Assembler.Lea(dest, memOp, OperandType.I64);
}
private static void GenerateStore(CodeGenContext context, Operation operation)
{
Operand value = operation.GetSource(1);
Operand address = Memory(operation.GetSource(0), value.Type);
GenerateStore(context, address, value);
}
private static void GenerateStore16(CodeGenContext context, Operation operation)
{
Operand value = operation.GetSource(1);
Operand address = Memory(operation.GetSource(0), value.Type);
Debug.Assert(value.Type.IsInteger());
context.Assembler.Mov16(address, value);
}
private static void GenerateStore8(CodeGenContext context, Operation operation)
{
Operand value = operation.GetSource(1);
Operand address = Memory(operation.GetSource(0), value.Type);
Debug.Assert(value.Type.IsInteger());
context.Assembler.Mov8(address, value);
}
private static void GenerateSubtract(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0);
Operand src2 = operation.GetSource(1);
ValidateBinOp(dest, src1, src2);
if (dest.Type.IsInteger())
{
context.Assembler.Sub(dest, src2, dest.Type);
}
else if (dest.Type == OperandType.FP32)
{
context.Assembler.Subss(dest, src1, src2);
}
else /* if (dest.Type == OperandType.FP64) */
{
context.Assembler.Subsd(dest, src1, src2);
}
}
private static void GenerateTailcall(CodeGenContext context, Operation operation)
{
WriteEpilogue(context);
context.Assembler.Jmp(operation.GetSource(0));
}
private static void GenerateVectorCreateScalar(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(!dest.Type.IsInteger() && source.Type.IsInteger());
if (source.Type == OperandType.I32)
{
context.Assembler.Movd(dest, source); // (__m128i _mm_cvtsi32_si128(int a))
}
else /* if (source.Type == OperandType.I64) */
{
context.Assembler.Movq(dest, source); // (__m128i _mm_cvtsi64_si128(__int64 a))
}
}
private static void GenerateVectorExtract(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination; //Value
Operand src1 = operation.GetSource(0); //Vector
Operand src2 = operation.GetSource(1); //Index
Debug.Assert(src1.Type == OperandType.V128);
Debug.Assert(src2.Kind == OperandKind.Constant);
byte index = src2.AsByte();
Debug.Assert(index < OperandType.V128.GetSizeInBytes() / dest.Type.GetSizeInBytes());
if (dest.Type == OperandType.I32)
{
if (index == 0)
{
context.Assembler.Movd(dest, src1);
}
else if (HardwareCapabilities.SupportsSse41)
{
context.Assembler.Pextrd(dest, src1, index);
}
else
{
int mask0 = 0b11_10_01_00;
int mask1 = 0b11_10_01_00;
mask0 = BitUtils.RotateRight(mask0, index * 2, 8);
mask1 = BitUtils.RotateRight(mask1, 8 - index * 2, 8);
context.Assembler.Pshufd(src1, src1, (byte)mask0);
context.Assembler.Movd (dest, src1);
context.Assembler.Pshufd(src1, src1, (byte)mask1);
}
}
else if (dest.Type == OperandType.I64)
{
if (index == 0)
{
context.Assembler.Movq(dest, src1);
}
else if (HardwareCapabilities.SupportsSse41)
{
context.Assembler.Pextrq(dest, src1, index);
}
else
{
const byte mask = 0b01_00_11_10;
context.Assembler.Pshufd(src1, src1, mask);
context.Assembler.Movq (dest, src1);
context.Assembler.Pshufd(src1, src1, mask);
}
}
else
{
// Floating-point types.
if ((index >= 2 && dest.Type == OperandType.FP32) ||
(index == 1 && dest.Type == OperandType.FP64))
{
context.Assembler.Movhlps(dest, dest, src1);
context.Assembler.Movq (dest, dest);
}
else
{
context.Assembler.Movq(dest, src1);
}
if (dest.Type == OperandType.FP32)
{
context.Assembler.Pshufd(dest, dest, (byte)(0xfc | (index & 1)));
}
}
}
private static void GenerateVectorExtract16(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination; //Value
Operand src1 = operation.GetSource(0); //Vector
Operand src2 = operation.GetSource(1); //Index
Debug.Assert(src1.Type == OperandType.V128);
Debug.Assert(src2.Kind == OperandKind.Constant);
byte index = src2.AsByte();
Debug.Assert(index < 8);
context.Assembler.Pextrw(dest, src1, index);
}
private static void GenerateVectorExtract8(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination; //Value
Operand src1 = operation.GetSource(0); //Vector
Operand src2 = operation.GetSource(1); //Index
Debug.Assert(src1.Type == OperandType.V128);
Debug.Assert(src2.Kind == OperandKind.Constant);
byte index = src2.AsByte();
Debug.Assert(index < 16);
if (HardwareCapabilities.SupportsSse41)
{
context.Assembler.Pextrb(dest, src1, index);
}
else
{
context.Assembler.Pextrw(dest, src1, (byte)(index >> 1));
if ((index & 1) != 0)
{
context.Assembler.Shr(dest, Const(8), OperandType.I32);
}
else
{
context.Assembler.Movzx8(dest, dest, OperandType.I32);
}
}
}
private static void GenerateVectorInsert(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0); //Vector
Operand src2 = operation.GetSource(1); //Value
Operand src3 = operation.GetSource(2); //Index
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(src1.Type == OperandType.V128);
Debug.Assert(src3.Kind == OperandKind.Constant);
byte index = src3.AsByte();
void InsertIntSse2(int words)
{
if (dest.GetRegister() != src1.GetRegister())
{
context.Assembler.Movdqu(dest, src1);
}
for (int word = 0; word < words; word++)
{
// Insert lower 16-bits.
context.Assembler.Pinsrw(dest, dest, src2, (byte)(index * words + word));
// Move next word down.
context.Assembler.Ror(src2, Const(16), src2.Type);
}
}
if (src2.Type == OperandType.I32)
{
Debug.Assert(index < 4);
if (HardwareCapabilities.SupportsSse41)
{
context.Assembler.Pinsrd(dest, src1, src2, index);
}
else
{
InsertIntSse2(2);
}
}
else if (src2.Type == OperandType.I64)
{
Debug.Assert(index < 2);
if (HardwareCapabilities.SupportsSse41)
{
context.Assembler.Pinsrq(dest, src1, src2, index);
}
else
{
InsertIntSse2(4);
}
}
else if (src2.Type == OperandType.FP32)
{
Debug.Assert(index < 4);
if (index != 0)
{
if (HardwareCapabilities.SupportsSse41)
{
context.Assembler.Insertps(dest, src1, src2, (byte)(index << 4));
}
else
{
if (src1.GetRegister() == src2.GetRegister())
{
int mask = 0b11_10_01_00;
mask &= ~(0b11 << index * 2);
context.Assembler.Pshufd(dest, src1, (byte)mask);
}
else
{
int mask0 = 0b11_10_01_00;
int mask1 = 0b11_10_01_00;
mask0 = BitUtils.RotateRight(mask0, index * 2, 8);
mask1 = BitUtils.RotateRight(mask1, 8 - index * 2, 8);
context.Assembler.Pshufd(src1, src1, (byte)mask0); // Lane to be inserted in position 0.
context.Assembler.Movss (dest, src1, src2); // dest[127:0] = src1[127:32] | src2[31:0]
context.Assembler.Pshufd(dest, dest, (byte)mask1); // Inserted lane in original position.
if (dest.GetRegister() != src1.GetRegister())
{
context.Assembler.Pshufd(src1, src1, (byte)mask1); // Restore src1.
}
}
}
}
else
{
context.Assembler.Movss(dest, src1, src2);
}
}
else /* if (src2.Type == OperandType.FP64) */
{
Debug.Assert(index < 2);
if (index != 0)
{
context.Assembler.Movlhps(dest, src1, src2);
}
else
{
context.Assembler.Movsd(dest, src1, src2);
}
}
}
private static void GenerateVectorInsert16(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0); //Vector
Operand src2 = operation.GetSource(1); //Value
Operand src3 = operation.GetSource(2); //Index
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(src1.Type == OperandType.V128);
Debug.Assert(src3.Kind == OperandKind.Constant);
byte index = src3.AsByte();
context.Assembler.Pinsrw(dest, src1, src2, index);
}
private static void GenerateVectorInsert8(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand src1 = operation.GetSource(0); //Vector
Operand src2 = operation.GetSource(1); //Value
Operand src3 = operation.GetSource(2); //Index
// It's not possible to emulate this instruction without
// SSE 4.1 support without the use of a temporary register,
// so we instead handle that case on the pre-allocator when
// SSE 4.1 is not supported on the CPU.
Debug.Assert(HardwareCapabilities.SupportsSse41);
if (!HardwareCapabilities.SupportsVexEncoding)
{
EnsureSameReg(dest, src1);
}
Debug.Assert(src1.Type == OperandType.V128);
Debug.Assert(src3.Kind == OperandKind.Constant);
byte index = src3.AsByte();
context.Assembler.Pinsrb(dest, src1, src2, index);
}
private static void GenerateVectorOne(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Debug.Assert(!dest.Type.IsInteger());
context.Assembler.Pcmpeqw(dest, dest, dest);
}
private static void GenerateVectorZero(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Debug.Assert(!dest.Type.IsInteger());
context.Assembler.Xorps(dest, dest, dest);
}
private static void GenerateVectorZeroUpper64(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type == OperandType.V128 && source.Type == OperandType.V128);
GenerateZeroUpper64(context, dest, source);
}
private static void GenerateVectorZeroUpper96(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type == OperandType.V128 && source.Type == OperandType.V128);
GenerateZeroUpper96(context, dest, source);
}
private static void GenerateZeroExtend16(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && source.Type.IsInteger());
context.Assembler.Movzx16(dest, source, OperandType.I32);
}
private static void GenerateZeroExtend32(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && source.Type.IsInteger());
context.Assembler.Mov(dest, source, OperandType.I32);
}
private static void GenerateZeroExtend8(CodeGenContext context, Operation operation)
{
Operand dest = operation.Destination;
Operand source = operation.GetSource(0);
Debug.Assert(dest.Type.IsInteger() && source.Type.IsInteger());
context.Assembler.Movzx8(dest, source, OperandType.I32);
}
private static void GenerateLoad(CodeGenContext context, Operand address, Operand value)
{
switch (value.Type)
{
case OperandType.I32: context.Assembler.Mov (value, address, OperandType.I32); break;
case OperandType.I64: context.Assembler.Mov (value, address, OperandType.I64); break;
case OperandType.FP32: context.Assembler.Movd (value, address); break;
case OperandType.FP64: context.Assembler.Movq (value, address); break;
case OperandType.V128: context.Assembler.Movdqu(value, address); break;
default: Debug.Assert(false); break;
}
}
private static void GenerateStore(CodeGenContext context, Operand address, Operand value)
{
switch (value.Type)
{
case OperandType.I32: context.Assembler.Mov (address, value, OperandType.I32); break;
case OperandType.I64: context.Assembler.Mov (address, value, OperandType.I64); break;
case OperandType.FP32: context.Assembler.Movd (address, value); break;
case OperandType.FP64: context.Assembler.Movq (address, value); break;
case OperandType.V128: context.Assembler.Movdqu(address, value); break;
default: Debug.Assert(false); break;
}
}
private static void GenerateZeroUpper64(CodeGenContext context, Operand dest, Operand source)
{
context.Assembler.Movq(dest, source);
}
private static void GenerateZeroUpper96(CodeGenContext context, Operand dest, Operand source)
{
context.Assembler.Movq(dest, source);
context.Assembler.Pshufd(dest, dest, 0xfc);
}
private static bool MatchOperation(Operation node, Instruction inst, OperandType destType, Register destReg)
{
if (node == default || node.DestinationsCount == 0)
{
return false;
}
if (node.Instruction != inst)
{
return false;
}
Operand dest = node.Destination;
return dest.Kind == OperandKind.Register &&
dest.Type == destType &&
dest.GetRegister() == destReg;
}
[Conditional("DEBUG")]
private static void ValidateUnOp(Operand dest, Operand source)
{
EnsureSameReg (dest, source);
EnsureSameType(dest, source);
}
[Conditional("DEBUG")]
private static void ValidateBinOp(Operand dest, Operand src1, Operand src2)
{
EnsureSameReg (dest, src1);
EnsureSameType(dest, src1, src2);
}
[Conditional("DEBUG")]
private static void ValidateShift(Operand dest, Operand src1, Operand src2)
{
EnsureSameReg (dest, src1);
EnsureSameType(dest, src1);
Debug.Assert(dest.Type.IsInteger() && src2.Type == OperandType.I32);
}
private static void EnsureSameReg(Operand op1, Operand op2)
{
if (!op1.Type.IsInteger() && HardwareCapabilities.SupportsVexEncoding)
{
return;
}
Debug.Assert(op1.Kind == OperandKind.Register || op1.Kind == OperandKind.Memory);
Debug.Assert(op1.Kind == op2.Kind);
Debug.Assert(op1.Value == op2.Value);
}
private static void EnsureSameType(Operand op1, Operand op2)
{
Debug.Assert(op1.Type == op2.Type);
}
private static void EnsureSameType(Operand op1, Operand op2, Operand op3)
{
Debug.Assert(op1.Type == op2.Type);
Debug.Assert(op1.Type == op3.Type);
}
private static void EnsureSameType(Operand op1, Operand op2, Operand op3, Operand op4)
{
Debug.Assert(op1.Type == op2.Type);
Debug.Assert(op1.Type == op3.Type);
Debug.Assert(op1.Type == op4.Type);
}
private static UnwindInfo WritePrologue(CodeGenContext context)
{
List<UnwindPushEntry> pushEntries = new List<UnwindPushEntry>();
Operand rsp = Register(X86Register.Rsp);
int mask = CallingConvention.GetIntCalleeSavedRegisters() & context.AllocResult.IntUsedRegisters;
while (mask != 0)
{
int bit = BitOperations.TrailingZeroCount(mask);
context.Assembler.Push(Register((X86Register)bit));
pushEntries.Add(new UnwindPushEntry(UnwindPseudoOp.PushReg, context.StreamOffset, regIndex: bit));
mask &= ~(1 << bit);
}
int reservedStackSize = context.CallArgsRegionSize + context.AllocResult.SpillRegionSize;
reservedStackSize += context.XmmSaveRegionSize;
if (reservedStackSize >= StackGuardSize)
{
GenerateInlineStackProbe(context, reservedStackSize);
}
if (reservedStackSize != 0)
{
context.Assembler.Sub(rsp, Const(reservedStackSize), OperandType.I64);
pushEntries.Add(new UnwindPushEntry(UnwindPseudoOp.AllocStack, context.StreamOffset, stackOffsetOrAllocSize: reservedStackSize));
}
int offset = reservedStackSize;
mask = CallingConvention.GetVecCalleeSavedRegisters() & context.AllocResult.VecUsedRegisters;
while (mask != 0)
{
int bit = BitOperations.TrailingZeroCount(mask);
offset -= 16;
Operand memOp = MemoryOp(OperandType.V128, rsp, default, Multiplier.x1, offset);
context.Assembler.Movdqu(memOp, Xmm((X86Register)bit));
pushEntries.Add(new UnwindPushEntry(UnwindPseudoOp.SaveXmm128, context.StreamOffset, bit, offset));
mask &= ~(1 << bit);
}
return new UnwindInfo(pushEntries.ToArray(), context.StreamOffset);
}
private static void WriteEpilogue(CodeGenContext context)
{
Operand rsp = Register(X86Register.Rsp);
int reservedStackSize = context.CallArgsRegionSize + context.AllocResult.SpillRegionSize;
reservedStackSize += context.XmmSaveRegionSize;
int offset = reservedStackSize;
int mask = CallingConvention.GetVecCalleeSavedRegisters() & context.AllocResult.VecUsedRegisters;
while (mask != 0)
{
int bit = BitOperations.TrailingZeroCount(mask);
offset -= 16;
Operand memOp = MemoryOp(OperandType.V128, rsp, default, Multiplier.x1, offset);
context.Assembler.Movdqu(Xmm((X86Register)bit), memOp);
mask &= ~(1 << bit);
}
if (reservedStackSize != 0)
{
context.Assembler.Add(rsp, Const(reservedStackSize), OperandType.I64);
}
mask = CallingConvention.GetIntCalleeSavedRegisters() & context.AllocResult.IntUsedRegisters;
while (mask != 0)
{
int bit = BitUtils.HighestBitSet(mask);
context.Assembler.Pop(Register((X86Register)bit));
mask &= ~(1 << bit);
}
}
private static void GenerateInlineStackProbe(CodeGenContext context, int size)
{
// Windows does lazy stack allocation, and there are just 2
// guard pages on the end of the stack. So, if the allocation
// size we make is greater than this guard size, we must ensure
// that the OS will map all pages that we'll use. We do that by
// doing a dummy read on those pages, forcing a page fault and
// the OS to map them. If they are already mapped, nothing happens.
const int pageMask = PageSize - 1;
size = (size + pageMask) & ~pageMask;
Operand rsp = Register(X86Register.Rsp);
Operand temp = Register(CallingConvention.GetIntReturnRegister());
for (int offset = PageSize; offset < size; offset += PageSize)
{
Operand memOp = MemoryOp(OperandType.I32, rsp, default, Multiplier.x1, -offset);
context.Assembler.Mov(temp, memOp, OperandType.I32);
}
}
private static Operand Memory(Operand operand, OperandType type)
{
if (operand.Kind == OperandKind.Memory)
{
return operand;
}
return MemoryOp(type, operand);
}
private static Operand Register(X86Register register, OperandType type = OperandType.I64)
{
return Operand.Factory.Register((int)register, RegisterType.Integer, type);
}
private static Operand Xmm(X86Register register)
{
return Operand.Factory.Register((int)register, RegisterType.Vector, OperandType.V128);
}
}
}