rjx-mirror/ChocolArm64/Instructions/InstEmitMemoryEx.cs
gdkchan 932224f051 ARM exclusive monitor and multicore fixes (#589)
* Implement ARM exclusive load/store with compare exchange insts, and enable multicore by default

* Fix comment typo

* Support Linux and OSX on MemoryAlloc and CompareExchange128, some cleanup

* Use intel syntax on assembly code

* Adjust identation

* Add CPUID check and fix exclusive reservation granule size

* Update schema multicore scheduling default value

* Make the cpu id check code lower case aswell
2019-02-19 10:52:06 +11:00

357 lines
No EOL
12 KiB
C#

using ChocolArm64.Decoders;
using ChocolArm64.Memory;
using ChocolArm64.State;
using ChocolArm64.Translation;
using System;
using System.Reflection.Emit;
using System.Threading;
using static ChocolArm64.Instructions.InstEmitMemoryHelper;
namespace ChocolArm64.Instructions
{
static partial class InstEmit
{
[Flags]
private enum AccessType
{
None = 0,
Ordered = 1,
Exclusive = 2,
OrderedEx = Ordered | Exclusive
}
public static void Clrex(ILEmitterCtx context)
{
context.EmitLdarg(TranslatedSub.StateArgIdx);
context.EmitPrivateCall(typeof(CpuThreadState), nameof(CpuThreadState.ClearExclusiveAddress));
}
public static void Dmb(ILEmitterCtx context) => EmitBarrier(context);
public static void Dsb(ILEmitterCtx context) => EmitBarrier(context);
public static void Ldar(ILEmitterCtx context) => EmitLdr(context, AccessType.Ordered);
public static void Ldaxr(ILEmitterCtx context) => EmitLdr(context, AccessType.OrderedEx);
public static void Ldxr(ILEmitterCtx context) => EmitLdr(context, AccessType.Exclusive);
public static void Ldxp(ILEmitterCtx context) => EmitLdp(context, AccessType.Exclusive);
public static void Ldaxp(ILEmitterCtx context) => EmitLdp(context, AccessType.OrderedEx);
private static void EmitLdr(ILEmitterCtx context, AccessType accType)
{
EmitLoad(context, accType, pair: false);
}
private static void EmitLdp(ILEmitterCtx context, AccessType accType)
{
EmitLoad(context, accType, pair: true);
}
private static void EmitLoad(ILEmitterCtx context, AccessType accType, bool pair)
{
OpCodeMemEx64 op = (OpCodeMemEx64)context.CurrOp;
bool ordered = (accType & AccessType.Ordered) != 0;
bool exclusive = (accType & AccessType.Exclusive) != 0;
if (ordered)
{
EmitBarrier(context);
}
context.EmitLdint(op.Rn);
context.EmitSttmp();
if (exclusive)
{
context.EmitLdarg(TranslatedSub.StateArgIdx);
context.EmitLdtmp();
context.EmitPrivateCall(typeof(CpuThreadState), nameof(CpuThreadState.SetExclusiveAddress));
}
void WriteExclusiveValue(string propName)
{
if (op.Size < 3)
{
context.Emit(OpCodes.Conv_U8);
}
context.EmitSttmp2();
context.EmitLdarg(TranslatedSub.StateArgIdx);
context.EmitLdtmp2();
context.EmitCallPrivatePropSet(typeof(CpuThreadState), propName);
context.EmitLdtmp2();
if (op.Size < 3)
{
context.Emit(OpCodes.Conv_U4);
}
}
if (pair)
{
//Exclusive loads should be atomic. For pairwise loads, we need to
//read all the data at once. For a 32-bits pairwise load, we do a
//simple 64-bits load, for a 128-bits load, we need to call a special
//method to read 128-bits atomically.
if (op.Size == 2)
{
context.EmitLdarg(TranslatedSub.MemoryArgIdx);
context.EmitLdtmp();
EmitReadZxCall(context, 3);
context.Emit(OpCodes.Dup);
//Mask low half.
context.Emit(OpCodes.Conv_U4);
if (exclusive)
{
WriteExclusiveValue(nameof(CpuThreadState.ExclusiveValueLow));
}
context.EmitStintzr(op.Rt);
//Shift high half.
context.EmitLsr(32);
context.Emit(OpCodes.Conv_U4);
if (exclusive)
{
WriteExclusiveValue(nameof(CpuThreadState.ExclusiveValueHigh));
}
context.EmitStintzr(op.Rt2);
}
else if (op.Size == 3)
{
context.EmitLdarg(TranslatedSub.MemoryArgIdx);
context.EmitLdtmp();
context.EmitPrivateCall(typeof(MemoryManager), nameof(MemoryManager.AtomicReadInt128));
context.Emit(OpCodes.Dup);
//Load low part of the vector.
context.EmitLdc_I4(0);
context.EmitLdc_I4(3);
VectorHelper.EmitCall(context, nameof(VectorHelper.VectorExtractIntZx));
if (exclusive)
{
WriteExclusiveValue(nameof(CpuThreadState.ExclusiveValueLow));
}
context.EmitStintzr(op.Rt);
//Load high part of the vector.
context.EmitLdc_I4(1);
context.EmitLdc_I4(3);
VectorHelper.EmitCall(context, nameof(VectorHelper.VectorExtractIntZx));
if (exclusive)
{
WriteExclusiveValue(nameof(CpuThreadState.ExclusiveValueHigh));
}
context.EmitStintzr(op.Rt2);
}
else
{
throw new InvalidOperationException($"Invalid store size of {1 << op.Size} bytes.");
}
}
else
{
//8, 16, 32 or 64-bits (non-pairwise) load.
context.EmitLdarg(TranslatedSub.MemoryArgIdx);
context.EmitLdtmp();
EmitReadZxCall(context, op.Size);
if (exclusive)
{
WriteExclusiveValue(nameof(CpuThreadState.ExclusiveValueLow));
}
context.EmitStintzr(op.Rt);
}
}
public static void Pfrm(ILEmitterCtx context)
{
//Memory Prefetch, execute as no-op.
}
public static void Stlr(ILEmitterCtx context) => EmitStr(context, AccessType.Ordered);
public static void Stlxr(ILEmitterCtx context) => EmitStr(context, AccessType.OrderedEx);
public static void Stxr(ILEmitterCtx context) => EmitStr(context, AccessType.Exclusive);
public static void Stxp(ILEmitterCtx context) => EmitStp(context, AccessType.Exclusive);
public static void Stlxp(ILEmitterCtx context) => EmitStp(context, AccessType.OrderedEx);
private static void EmitStr(ILEmitterCtx context, AccessType accType)
{
EmitStore(context, accType, pair: false);
}
private static void EmitStp(ILEmitterCtx context, AccessType accType)
{
EmitStore(context, accType, pair: true);
}
private static void EmitStore(ILEmitterCtx context, AccessType accType, bool pair)
{
OpCodeMemEx64 op = (OpCodeMemEx64)context.CurrOp;
bool ordered = (accType & AccessType.Ordered) != 0;
bool exclusive = (accType & AccessType.Exclusive) != 0;
if (ordered)
{
EmitBarrier(context);
}
if (exclusive)
{
ILLabel lblEx = new ILLabel();
ILLabel lblEnd = new ILLabel();
context.EmitLdarg(TranslatedSub.StateArgIdx);
context.EmitLdint(op.Rn);
context.EmitPrivateCall(typeof(CpuThreadState), nameof(CpuThreadState.CheckExclusiveAddress));
context.Emit(OpCodes.Brtrue_S, lblEx);
//Address check failed, set error right away and do not store anything.
context.EmitLdc_I4(1);
context.EmitStintzr(op.Rs);
context.Emit(OpCodes.Br, lblEnd);
//Address check passsed.
context.MarkLabel(lblEx);
context.EmitLdarg(TranslatedSub.MemoryArgIdx);
context.EmitLdint(op.Rn);
context.EmitLdarg(TranslatedSub.StateArgIdx);
context.EmitCallPrivatePropGet(typeof(CpuThreadState), nameof(CpuThreadState.ExclusiveValueLow));
void EmitCast()
{
//The input should be always int64.
switch (op.Size)
{
case 0: context.Emit(OpCodes.Conv_U1); break;
case 1: context.Emit(OpCodes.Conv_U2); break;
case 2: context.Emit(OpCodes.Conv_U4); break;
}
}
EmitCast();
if (pair)
{
context.EmitLdarg(TranslatedSub.StateArgIdx);
context.EmitCallPrivatePropGet(typeof(CpuThreadState), nameof(CpuThreadState.ExclusiveValueHigh));
EmitCast();
context.EmitLdintzr(op.Rt);
EmitCast();
context.EmitLdintzr(op.Rt2);
EmitCast();
switch (op.Size)
{
case 2: context.EmitPrivateCall(typeof(MemoryManager), nameof(MemoryManager.AtomicCompareExchange2xInt32)); break;
case 3: context.EmitPrivateCall(typeof(MemoryManager), nameof(MemoryManager.AtomicCompareExchangeInt128)); break;
default: throw new InvalidOperationException($"Invalid store size of {1 << op.Size} bytes.");
}
}
else
{
context.EmitLdintzr(op.Rt);
EmitCast();
switch (op.Size)
{
case 0: context.EmitCall(typeof(MemoryManager), nameof(MemoryManager.AtomicCompareExchangeByte)); break;
case 1: context.EmitCall(typeof(MemoryManager), nameof(MemoryManager.AtomicCompareExchangeInt16)); break;
case 2: context.EmitCall(typeof(MemoryManager), nameof(MemoryManager.AtomicCompareExchangeInt32)); break;
case 3: context.EmitCall(typeof(MemoryManager), nameof(MemoryManager.AtomicCompareExchangeInt64)); break;
default: throw new InvalidOperationException($"Invalid store size of {1 << op.Size} bytes.");
}
}
//The value returned is a bool, true if the values compared
//were equal and the new value was written, false otherwise.
//We need to invert this result, as on ARM 1 indicates failure,
//and 0 success on those instructions.
context.EmitLdc_I4(1);
context.Emit(OpCodes.Xor);
context.Emit(OpCodes.Dup);
context.Emit(OpCodes.Conv_U8);
context.EmitStintzr(op.Rs);
//Only clear the exclusive monitor if the store was successful (Rs = false).
context.Emit(OpCodes.Brtrue_S, lblEnd);
Clrex(context);
context.MarkLabel(lblEnd);
}
else
{
void EmitWrite(int rt, long offset)
{
context.EmitLdarg(TranslatedSub.MemoryArgIdx);
context.EmitLdint(op.Rn);
if (offset != 0)
{
context.EmitLdc_I8(offset);
context.Emit(OpCodes.Add);
}
context.EmitLdintzr(rt);
EmitWriteCall(context, op.Size);
}
EmitWrite(op.Rt, 0);
if (pair)
{
EmitWrite(op.Rt2, 1 << op.Size);
}
}
}
private static void EmitBarrier(ILEmitterCtx context)
{
//Note: This barrier is most likely not necessary, and probably
//doesn't make any difference since we need to do a ton of stuff
//(software MMU emulation) to read or write anything anyway.
context.EmitCall(typeof(Thread), nameof(Thread.MemoryBarrier));
}
}
}