rjx-mirror/ARMeilleure/Instructions/InstEmitMemoryHelper.cs
riperiperi b4d8d893a4
Memory Read/Write Tracking using Region Handles (#1272)
* WIP Range Tracking

- Texture invalidation seems to have large problems
- Buffer/Pool invalidation may have problems
- Mirror memory tracking puts an additional `add` in compiled code, we likely just want to make HLE access slower if this is the final solution.
- Native project is in the messiest possible location.
- [HACK] JIT memory access always uses native "fast" path
- [HACK] Trying some things with texture invalidation and views.

It works :)

Still a few hacks, messy things, slow things

More work in progress stuff (also move to memory project)

Quite a bit faster now.
- Unmapping GPU VA and CPU VA will now correctly update write tracking regions, and invalidate textures for the former.
- The Virtual range list is now non-overlapping like the physical one.
- Fixed some bugs where regions could leak.
- Introduced a weird bug that I still need to track down (consistent invalid buffer in MK8 ribbon road)

Move some stuff.

I think we'll eventually just put the dll and so for this in a nuget package.

Fix rebase.

[WIP] MultiRegionHandle variable size ranges

- Avoid reprotecting regions that change often (needs some tweaking)
- There's still a bug in buffers, somehow.
- Might want different api for minimum granularity

Fix rebase issue

Commit everything needed for software only tracking.

Remove native components.

Remove more native stuff.

Cleanup

Use a separate window for the background context, update opentk. (fixes linux)

Some experimental changes

Should get things working up to scratch - still need to try some things with flush/modification and res scale.

Include address with the region action.

Initial work to make range tracking work

Still a ton of bugs

Fix some issues with the new stuff.

* Fix texture flush instability

There's still some weird behaviour, but it's much improved without this. (textures with cpu modified data were flushing over it)

* Find the destination texture for Buffer->Texture full copy

Greatly improves performance for nvdec videos (with range tracking)

* Further improve texture tracking

* Disable Memory Tracking for view parents

This is a temporary approach to better match behaviour on master (where invalidations would be soaked up by views, rather than trigger twice)

The assumption is that when views are created to a texture, they will cover all of its data anyways. Of course, this can easily be improved in future.

* Introduce some tracking tests.

WIP

* Complete base tests.

* Add more tests for multiregion, fix existing test.

* Cleanup Part 1

* Remove unnecessary code from memory tracking

* Fix some inconsistencies with 3D texture rule.

* Add dispose tests.

* Use a background thread for the background context.

Rather than setting and unsetting a context as current, doing the work on a dedicated thread with signals seems to be a bit faster.

Also nerf the multithreading test a bit.

* Copy to texture with matching alignment

This extends the copy to work for some videos with unusual size, such as tutorial videos in SMO. It will only occur if the destination texture already exists at XCount size.

* Track reads for buffer copies. Synchronize new buffers before copying overlaps.

* Remove old texture flushing mechanisms.

Range tracking all the way, baby.

* Wake the background thread when disposing.

Avoids a deadlock when games are closed.

* Address Feedback 1

* Separate TextureCopy instance for background thread

Also `BackgroundContextWorker.InBackground` for a more sensible idenfifier for if we're in a background thread.

* Add missing XML docs.

* Address Feedback

* Maybe I should start drinking coffee.

* Some more feedback.

* Remove flush warning, Refocus window after making background context
2020-10-16 17:18:35 -03:00

620 lines
22 KiB
C#

using ARMeilleure.Decoders;
using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.Translation;
using ARMeilleure.Translation.PTC;
using System;
using System.Reflection;
using static ARMeilleure.Instructions.InstEmitHelper;
using static ARMeilleure.IntermediateRepresentation.OperandHelper;
namespace ARMeilleure.Instructions
{
static class InstEmitMemoryHelper
{
private const int PageBits = 12;
private const int PageMask = (1 << PageBits) - 1;
private enum Extension
{
Zx,
Sx32,
Sx64
}
public static void EmitLoadZx(ArmEmitterContext context, Operand address, int rt, int size)
{
EmitLoad(context, address, Extension.Zx, rt, size);
}
public static void EmitLoadSx32(ArmEmitterContext context, Operand address, int rt, int size)
{
EmitLoad(context, address, Extension.Sx32, rt, size);
}
public static void EmitLoadSx64(ArmEmitterContext context, Operand address, int rt, int size)
{
EmitLoad(context, address, Extension.Sx64, rt, size);
}
private static void EmitLoad(ArmEmitterContext context, Operand address, Extension ext, int rt, int size)
{
bool isSimd = IsSimd(context);
if ((uint)size > (isSimd ? 4 : 3))
{
throw new ArgumentOutOfRangeException(nameof(size));
}
if (isSimd)
{
EmitReadVector(context, address, context.VectorZero(), rt, 0, size);
}
else
{
EmitReadInt(context, address, rt, size);
}
if (!isSimd && !(context.CurrOp is OpCode32 && rt == State.RegisterAlias.Aarch32Pc))
{
Operand value = GetInt(context, rt);
if (ext == Extension.Sx32 || ext == Extension.Sx64)
{
OperandType destType = ext == Extension.Sx64 ? OperandType.I64 : OperandType.I32;
switch (size)
{
case 0: value = context.SignExtend8 (destType, value); break;
case 1: value = context.SignExtend16(destType, value); break;
case 2: value = context.SignExtend32(destType, value); break;
}
}
SetInt(context, rt, value);
}
}
public static void EmitLoadSimd(
ArmEmitterContext context,
Operand address,
Operand vector,
int rt,
int elem,
int size)
{
EmitReadVector(context, address, vector, rt, elem, size);
}
public static void EmitStore(ArmEmitterContext context, Operand address, int rt, int size)
{
bool isSimd = IsSimd(context);
if ((uint)size > (isSimd ? 4 : 3))
{
throw new ArgumentOutOfRangeException(nameof(size));
}
if (isSimd)
{
EmitWriteVector(context, address, rt, 0, size);
}
else
{
EmitWriteInt(context, address, rt, size);
}
}
public static void EmitStoreSimd(
ArmEmitterContext context,
Operand address,
int rt,
int elem,
int size)
{
EmitWriteVector(context, address, rt, elem, size);
}
private static bool IsSimd(ArmEmitterContext context)
{
return context.CurrOp is IOpCodeSimd &&
!(context.CurrOp is OpCodeSimdMemMs ||
context.CurrOp is OpCodeSimdMemSs);
}
private static void EmitReadInt(ArmEmitterContext context, Operand address, int rt, int size)
{
Operand lblSlowPath = Label();
Operand lblEnd = Label();
Operand isUnalignedAddr = EmitAddressCheck(context, address, size);
context.BranchIfTrue(lblSlowPath, isUnalignedAddr);
Operand physAddr = EmitPtPointerLoad(context, address, lblSlowPath, write: false);
Operand value = null;
switch (size)
{
case 0: value = context.Load8 (physAddr); break;
case 1: value = context.Load16(physAddr); break;
case 2: value = context.Load (OperandType.I32, physAddr); break;
case 3: value = context.Load (OperandType.I64, physAddr); break;
}
SetInt(context, rt, value);
context.Branch(lblEnd);
context.MarkLabel(lblSlowPath, BasicBlockFrequency.Cold);
EmitReadIntFallback(context, address, rt, size);
context.MarkLabel(lblEnd);
}
public static Operand EmitReadIntAligned(ArmEmitterContext context, Operand address, int size)
{
if ((uint)size > 4)
{
throw new ArgumentOutOfRangeException(nameof(size));
}
Operand isUnalignedAddr = EmitAddressCheck(context, address, size);
Operand lblFastPath = Label();
context.BranchIfFalse(lblFastPath, isUnalignedAddr, BasicBlockFrequency.Cold);
// The call is not expected to return (it should throw).
context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.ThrowInvalidMemoryAccess)), address);
context.MarkLabel(lblFastPath);
Operand physAddr = EmitPtPointerLoad(context, address, null, write: false);
return size switch
{
0 => context.Load8(physAddr),
1 => context.Load16(physAddr),
2 => context.Load(OperandType.I32, physAddr),
3 => context.Load(OperandType.I64, physAddr),
_ => context.Load(OperandType.V128, physAddr)
};
}
private static void EmitReadVector(
ArmEmitterContext context,
Operand address,
Operand vector,
int rt,
int elem,
int size)
{
Operand lblSlowPath = Label();
Operand lblEnd = Label();
Operand isUnalignedAddr = EmitAddressCheck(context, address, size);
context.BranchIfTrue(lblSlowPath, isUnalignedAddr);
Operand physAddr = EmitPtPointerLoad(context, address, lblSlowPath, write: false);
Operand value = null;
switch (size)
{
case 0: value = context.VectorInsert8 (vector, context.Load8(physAddr), elem); break;
case 1: value = context.VectorInsert16(vector, context.Load16(physAddr), elem); break;
case 2: value = context.VectorInsert (vector, context.Load(OperandType.I32, physAddr), elem); break;
case 3: value = context.VectorInsert (vector, context.Load(OperandType.I64, physAddr), elem); break;
case 4: value = context.Load (OperandType.V128, physAddr); break;
}
context.Copy(GetVec(rt), value);
context.Branch(lblEnd);
context.MarkLabel(lblSlowPath, BasicBlockFrequency.Cold);
EmitReadVectorFallback(context, address, vector, rt, elem, size);
context.MarkLabel(lblEnd);
}
private static Operand VectorCreate(ArmEmitterContext context, Operand value)
{
return context.VectorInsert(context.VectorZero(), value, 0);
}
private static void EmitWriteInt(ArmEmitterContext context, Operand address, int rt, int size)
{
Operand lblSlowPath = Label();
Operand lblEnd = Label();
Operand isUnalignedAddr = EmitAddressCheck(context, address, size);
context.BranchIfTrue(lblSlowPath, isUnalignedAddr);
Operand physAddr = EmitPtPointerLoad(context, address, lblSlowPath, write: true);
Operand value = GetInt(context, rt);
if (size < 3 && value.Type == OperandType.I64)
{
value = context.ConvertI64ToI32(value);
}
switch (size)
{
case 0: context.Store8 (physAddr, value); break;
case 1: context.Store16(physAddr, value); break;
case 2: context.Store (physAddr, value); break;
case 3: context.Store (physAddr, value); break;
}
context.Branch(lblEnd);
context.MarkLabel(lblSlowPath, BasicBlockFrequency.Cold);
EmitWriteIntFallback(context, address, rt, size);
context.MarkLabel(lblEnd);
}
public static void EmitWriteIntAligned(ArmEmitterContext context, Operand address, Operand value, int size)
{
if ((uint)size > 4)
{
throw new ArgumentOutOfRangeException(nameof(size));
}
Operand isUnalignedAddr = EmitAddressCheck(context, address, size);
Operand lblFastPath = Label();
context.BranchIfFalse(lblFastPath, isUnalignedAddr, BasicBlockFrequency.Cold);
// The call is not expected to return (it should throw).
context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.ThrowInvalidMemoryAccess)), address);
context.MarkLabel(lblFastPath);
Operand physAddr = EmitPtPointerLoad(context, address, null, write: true);
if (size < 3 && value.Type == OperandType.I64)
{
value = context.ConvertI64ToI32(value);
}
if (size == 0)
{
context.Store8(physAddr, value);
}
else if (size == 1)
{
context.Store16(physAddr, value);
}
else
{
context.Store(physAddr, value);
}
}
private static void EmitWriteVector(
ArmEmitterContext context,
Operand address,
int rt,
int elem,
int size)
{
Operand lblSlowPath = Label();
Operand lblEnd = Label();
Operand isUnalignedAddr = EmitAddressCheck(context, address, size);
context.BranchIfTrue(lblSlowPath, isUnalignedAddr);
Operand physAddr = EmitPtPointerLoad(context, address, lblSlowPath, write: true);
Operand value = GetVec(rt);
switch (size)
{
case 0: context.Store8 (physAddr, context.VectorExtract8(value, elem)); break;
case 1: context.Store16(physAddr, context.VectorExtract16(value, elem)); break;
case 2: context.Store (physAddr, context.VectorExtract(OperandType.I32, value, elem)); break;
case 3: context.Store (physAddr, context.VectorExtract(OperandType.I64, value, elem)); break;
case 4: context.Store (physAddr, value); break;
}
context.Branch(lblEnd);
context.MarkLabel(lblSlowPath, BasicBlockFrequency.Cold);
EmitWriteVectorFallback(context, address, rt, elem, size);
context.MarkLabel(lblEnd);
}
public static Operand EmitAddressCheck(ArmEmitterContext context, Operand address, int size)
{
ulong addressCheckMask = ~((1UL << context.Memory.AddressSpaceBits) - 1);
addressCheckMask |= (1u << size) - 1;
return context.BitwiseAnd(address, Const(address.Type, (long)addressCheckMask));
}
public static Operand EmitPtPointerLoad(ArmEmitterContext context, Operand address, Operand lblSlowPath, bool write)
{
int ptLevelBits = context.Memory.AddressSpaceBits - 12; // 12 = Number of page bits.
int ptLevelSize = 1 << ptLevelBits;
int ptLevelMask = ptLevelSize - 1;
Operand pte = Ptc.State == PtcState.Disabled
? Const(context.Memory.PageTablePointer.ToInt64())
: Const(context.Memory.PageTablePointer.ToInt64(), true, Ptc.PageTablePointerIndex);
int bit = PageBits;
// Load page table entry from the page table.
// This was designed to support multi-level page tables of any size, however right
// now we only use flat page tables (so there's only one level).
// The page table entry contains the host address where the page is located.
// Additionally, the higher 16-bits of the host address may contain extra information
// used for write tracking, so this must be handled here aswell.
do
{
Operand addrPart = context.ShiftRightUI(address, Const(bit));
bit += ptLevelBits;
if (bit < context.Memory.AddressSpaceBits)
{
addrPart = context.BitwiseAnd(addrPart, Const(addrPart.Type, ptLevelMask));
}
Operand pteOffset = context.ShiftLeft(addrPart, Const(3));
if (pteOffset.Type == OperandType.I32)
{
pteOffset = context.ZeroExtend32(OperandType.I64, pteOffset);
}
Operand pteAddress = context.Add(pte, pteOffset);
pte = context.Load(OperandType.I64, pteAddress);
}
while (bit < context.Memory.AddressSpaceBits);
if (lblSlowPath != null)
{
ulong protection = (write ? 3UL : 1UL) << 48;
context.BranchIfTrue(lblSlowPath, context.BitwiseAnd(pte, Const(protection)));
}
else
{
// When no label is provided to jump to a slow path if the address is invalid,
// we do the validation ourselves, and throw if needed.
Operand lblNotWatched = Label();
// Is the page currently being tracked for read/write? If so we need to call MarkRegionAsModified.
context.BranchIf(lblNotWatched, pte, Const(0L), Comparison.GreaterOrEqual, BasicBlockFrequency.Cold);
// Mark the region as modified. Size here doesn't matter as address is assumed to be size aligned here.
context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.SignalMemoryTracking)), address, Const(1UL), Const(write ? 1 : 0));
context.MarkLabel(lblNotWatched);
Operand lblNonNull = Label();
// Skip exception if the PTE address is non-null (not zero).
context.BranchIfTrue(lblNonNull, pte, BasicBlockFrequency.Cold);
// The call is not expected to return (it should throw).
context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.ThrowInvalidMemoryAccess)), address);
context.MarkLabel(lblNonNull);
}
pte = context.BitwiseAnd(pte, Const(0xffffffffffffUL)); // Ignore any software protection bits. (they are still used by c# memory access)
Operand pageOffset = context.BitwiseAnd(address, Const(address.Type, PageMask));
if (pageOffset.Type == OperandType.I32)
{
pageOffset = context.ZeroExtend32(OperandType.I64, pageOffset);
}
return context.Add(pte, pageOffset);
}
private static void EmitReadIntFallback(ArmEmitterContext context, Operand address, int rt, int size)
{
MethodInfo info = null;
switch (size)
{
case 0: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadByte)); break;
case 1: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt16)); break;
case 2: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt32)); break;
case 3: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt64)); break;
}
SetInt(context, rt, context.Call(info, address));
}
private static void EmitReadVectorFallback(
ArmEmitterContext context,
Operand address,
Operand vector,
int rt,
int elem,
int size)
{
MethodInfo info = null;
switch (size)
{
case 0: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadByte)); break;
case 1: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt16)); break;
case 2: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt32)); break;
case 3: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadUInt64)); break;
case 4: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.ReadVector128)); break;
}
Operand value = context.Call(info, address);
switch (size)
{
case 0: value = context.VectorInsert8 (vector, value, elem); break;
case 1: value = context.VectorInsert16(vector, value, elem); break;
case 2: value = context.VectorInsert (vector, value, elem); break;
case 3: value = context.VectorInsert (vector, value, elem); break;
}
context.Copy(GetVec(rt), value);
}
private static void EmitWriteIntFallback(ArmEmitterContext context, Operand address, int rt, int size)
{
MethodInfo info = null;
switch (size)
{
case 0: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteByte)); break;
case 1: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt16)); break;
case 2: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt32)); break;
case 3: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt64)); break;
}
Operand value = GetInt(context, rt);
if (size < 3 && value.Type == OperandType.I64)
{
value = context.ConvertI64ToI32(value);
}
context.Call(info, address, value);
}
private static void EmitWriteVectorFallback(
ArmEmitterContext context,
Operand address,
int rt,
int elem,
int size)
{
MethodInfo info = null;
switch (size)
{
case 0: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteByte)); break;
case 1: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt16)); break;
case 2: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt32)); break;
case 3: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteUInt64)); break;
case 4: info = typeof(NativeInterface).GetMethod(nameof(NativeInterface.WriteVector128)); break;
}
Operand value = null;
if (size < 4)
{
switch (size)
{
case 0: value = context.VectorExtract8 (GetVec(rt), elem); break;
case 1: value = context.VectorExtract16(GetVec(rt), elem); break;
case 2: value = context.VectorExtract (OperandType.I32, GetVec(rt), elem); break;
case 3: value = context.VectorExtract (OperandType.I64, GetVec(rt), elem); break;
}
}
else
{
value = GetVec(rt);
}
context.Call(info, address, value);
}
private static Operand GetInt(ArmEmitterContext context, int rt)
{
return context.CurrOp is OpCode32 ? GetIntA32(context, rt) : GetIntOrZR(context, rt);
}
private static void SetInt(ArmEmitterContext context, int rt, Operand value)
{
if (context.CurrOp is OpCode32)
{
SetIntA32(context, rt, value);
}
else
{
SetIntOrZR(context, rt, value);
}
}
// ARM32 helpers.
public static Operand GetMemM(ArmEmitterContext context, bool setCarry = true)
{
switch (context.CurrOp)
{
case OpCode32MemRsImm op: return GetMShiftedByImmediate(context, op, setCarry);
case OpCode32MemReg op: return GetIntA32(context, op.Rm);
case OpCode32Mem op: return Const(op.Immediate);
case OpCode32SimdMemImm op: return Const(op.Immediate);
default: throw InvalidOpCodeType(context.CurrOp);
}
}
private static Exception InvalidOpCodeType(OpCode opCode)
{
return new InvalidOperationException($"Invalid OpCode type \"{opCode?.GetType().Name ?? "null"}\".");
}
public static Operand GetMShiftedByImmediate(ArmEmitterContext context, OpCode32MemRsImm op, bool setCarry)
{
Operand m = GetIntA32(context, op.Rm);
int shift = op.Immediate;
if (shift == 0)
{
switch (op.ShiftType)
{
case ShiftType.Lsr: shift = 32; break;
case ShiftType.Asr: shift = 32; break;
case ShiftType.Ror: shift = 1; break;
}
}
if (shift != 0)
{
setCarry &= false;
switch (op.ShiftType)
{
case ShiftType.Lsl: m = InstEmitAluHelper.GetLslC(context, m, setCarry, shift); break;
case ShiftType.Lsr: m = InstEmitAluHelper.GetLsrC(context, m, setCarry, shift); break;
case ShiftType.Asr: m = InstEmitAluHelper.GetAsrC(context, m, setCarry, shift); break;
case ShiftType.Ror:
if (op.Immediate != 0)
{
m = InstEmitAluHelper.GetRorC(context, m, setCarry, shift);
}
else
{
m = InstEmitAluHelper.GetRrxC(context, m, setCarry);
}
break;
}
}
return m;
}
}
}