Ryujinx/ARMeilleure/Instructions/InstEmitFlowHelper.cs
LDj3SNuD 5e724cf24e
Add Profiled Persistent Translation Cache. (#769)
* Delete DelegateTypes.cs

* Delete DelegateCache.cs

* Add files via upload

* Update Horizon.cs

* Update Program.cs

* Update MainWindow.cs

* Update Aot.cs

* Update RelocEntry.cs

* Update Translator.cs

* Update MemoryManager.cs

* Update InstEmitMemoryHelper.cs

* Update Delegates.cs

* Nit.

* Nit.

* Nit.

* 10 fewer MSIL bytes for us

* Add comment. Nits.

* Update Translator.cs

* Update Aot.cs

* Nits.

* Opt..

* Opt..

* Opt..

* Opt..

* Allow to change compression level.

* Update MemoryManager.cs

* Update Translator.cs

* Manage corner cases during the save phase. Nits.

* Update Aot.cs

* Translator response tweak for Aot disabled. Nit.

* Nit.

* Nits.

* Create DelegateHelpers.cs

* Update Delegates.cs

* Nit.

* Nit.

* Nits.

* Fix due to #784.

* Fixes due to #757 & #841.

* Fix due to #846.

* Fix due to #847.

* Use MethodInfo for managed method calls.

Use IR methods instead of managed methods about Max/Min (S/U).
Follow-ups & Nits.

* Add missing exception messages.

Reintroduce slow path for Fmov_Vi.
Implement slow path for Fmov_Si.

* Switch to the new folder structure.

Nits.

* Impl. index-based relocation information. Impl. cache file version field.

* Nit.

* Address gdkchan comments.

Mainly:
- fixed cache file corruption issue on exit; - exposed a way to disable AOT on the GUI.

* Address AcK77 comment.

* Address Thealexbarney, jduncanator & emmauss comments.

Header magic, CpuId (FI) & Aot -> Ptc.

* Adaptation to the new application reloading system.

Improvements to the call system of managed methods.
Follow-ups.
Nits.

* Get the same boot times as on master when PTC is disabled.

* Profiled Aot.

* A32 support (#897).

* #975 support (1 of 2).

* #975 support (2 of 2).

* Rebase fix & nits.

* Some fixes and nits (still one bug left).

* One fix & nits.

* Tests fix (by gdk) & nits.

* Support translations not only in high quality and rejit.

Nits.

* Added possibility to skip translations and continue execution, using `ESC` key.

* Update SettingsWindow.cs

* Update GLRenderer.cs

* Update Ptc.cs

* Disabled Profiled PTC by default as requested in the past by gdk.

* Fix rejit bug. Increased number of parallel translations. Add stack unwinding stuffs support (1 of 2).

Nits.

* Add stack unwinding stuffs support (2 of 2). Tuned number of parallel translations.

* Restored the ability to assemble jumps with 8-bit offset when Profiled PTC is disabled or during profiling.

Modifications due to rebase.
Nits.

* Limited profiling of the functions to be translated to the addresses belonging to the range of static objects only.

* Nits.

* Nits.

* Update Delegates.cs

* Nit.

* Update InstEmitSimdArithmetic.cs

* Address riperiperi comments.

* Fixed the issue of unjustifiably longer boot times at the second boot than at the first boot, measured at the same time or reference point and with the same number of translated functions.

* Implemented a simple redundant load/save mechanism.

Halved the value of Decoder.MaxInstsPerFunction more appropriate for the current performance of the Translator.
Replaced by Logger.PrintError to Logger.PrintDebug in TexturePool.cs about the supposed invalid texture format to avoid the spawn of the log.
Nits.

* Nit.

Improved Logger.PrintError in TexturePool.cs to avoid log spawn.
Added missing code for FZ handling (in output) for fp max/min instructions (slow paths).

* Add configuration migration for PTC

Co-authored-by: Thog <me@thog.eu>
2020-06-16 20:28:02 +02:00

383 lines
14 KiB
C#

using ARMeilleure.Decoders;
using ARMeilleure.IntermediateRepresentation;
using ARMeilleure.State;
using ARMeilleure.Translation;
using ARMeilleure.Translation.PTC;
using System;
using static ARMeilleure.Instructions.InstEmitHelper;
using static ARMeilleure.IntermediateRepresentation.OperandHelper;
namespace ARMeilleure.Instructions
{
static class InstEmitFlowHelper
{
public const ulong CallFlag = 1;
public static void EmitCondBranch(ArmEmitterContext context, Operand target, Condition cond)
{
if (cond != Condition.Al)
{
context.BranchIfTrue(target, GetCondTrue(context, cond));
}
else
{
context.Branch(target);
}
}
public static Operand GetCondTrue(ArmEmitterContext context, Condition condition)
{
Operand cmpResult = context.TryGetComparisonResult(condition);
if (cmpResult != null)
{
return cmpResult;
}
Operand value = Const(1);
Operand Inverse(Operand val)
{
return context.BitwiseExclusiveOr(val, Const(1));
}
switch (condition)
{
case Condition.Eq:
value = GetFlag(PState.ZFlag);
break;
case Condition.Ne:
value = Inverse(GetFlag(PState.ZFlag));
break;
case Condition.GeUn:
value = GetFlag(PState.CFlag);
break;
case Condition.LtUn:
value = Inverse(GetFlag(PState.CFlag));
break;
case Condition.Mi:
value = GetFlag(PState.NFlag);
break;
case Condition.Pl:
value = Inverse(GetFlag(PState.NFlag));
break;
case Condition.Vs:
value = GetFlag(PState.VFlag);
break;
case Condition.Vc:
value = Inverse(GetFlag(PState.VFlag));
break;
case Condition.GtUn:
{
Operand c = GetFlag(PState.CFlag);
Operand z = GetFlag(PState.ZFlag);
value = context.BitwiseAnd(c, Inverse(z));
break;
}
case Condition.LeUn:
{
Operand c = GetFlag(PState.CFlag);
Operand z = GetFlag(PState.ZFlag);
value = context.BitwiseOr(Inverse(c), z);
break;
}
case Condition.Ge:
{
Operand n = GetFlag(PState.NFlag);
Operand v = GetFlag(PState.VFlag);
value = context.ICompareEqual(n, v);
break;
}
case Condition.Lt:
{
Operand n = GetFlag(PState.NFlag);
Operand v = GetFlag(PState.VFlag);
value = context.ICompareNotEqual(n, v);
break;
}
case Condition.Gt:
{
Operand n = GetFlag(PState.NFlag);
Operand z = GetFlag(PState.ZFlag);
Operand v = GetFlag(PState.VFlag);
value = context.BitwiseAnd(Inverse(z), context.ICompareEqual(n, v));
break;
}
case Condition.Le:
{
Operand n = GetFlag(PState.NFlag);
Operand z = GetFlag(PState.ZFlag);
Operand v = GetFlag(PState.VFlag);
value = context.BitwiseOr(z, context.ICompareNotEqual(n, v));
break;
}
}
return value;
}
public static void EmitCall(ArmEmitterContext context, ulong immediate)
{
EmitJumpTableBranch(context, Const(immediate));
}
private static void EmitNativeCall(ArmEmitterContext context, Operand nativeContextPtr, Operand funcAddr, bool isJump = false)
{
context.StoreToContext();
Operand returnAddress;
if (isJump)
{
context.Tailcall(funcAddr, nativeContextPtr);
}
else
{
returnAddress = context.Call(funcAddr, OperandType.I64, nativeContextPtr);
context.LoadFromContext();
EmitContinueOrReturnCheck(context, returnAddress);
}
}
private static void EmitNativeCall(ArmEmitterContext context, Operand funcAddr, bool isJump = false)
{
EmitNativeCall(context, context.LoadArgument(OperandType.I64, 0), funcAddr, isJump);
}
public static void EmitVirtualCall(ArmEmitterContext context, Operand target)
{
EmitVirtualCallOrJump(context, target, isJump: false);
}
public static void EmitVirtualJump(ArmEmitterContext context, Operand target, bool isReturn)
{
EmitVirtualCallOrJump(context, target, isJump: true, isReturn: isReturn);
}
private static void EmitVirtualCallOrJump(ArmEmitterContext context, Operand target, bool isJump, bool isReturn = false)
{
if (isReturn)
{
context.Return(target);
}
else
{
EmitJumpTableBranch(context, target, isJump);
}
}
private static void EmitContinueOrReturnCheck(ArmEmitterContext context, Operand returnAddress)
{
// Note: The return value of a translated function is always an Int64 with the
// address execution has returned to. We expect this address to be immediately after the
// current instruction, if it isn't we keep returning until we reach the dispatcher.
Operand nextAddr = Const(GetNextOpAddress(context.CurrOp));
// Try to continue within this block.
// If the return address isn't to our next instruction, we need to return so the JIT can figure out what to do.
Operand lblContinue = Label();
// We need to clear out the call flag for the return address before comparing it.
context.BranchIfTrue(lblContinue, context.ICompareEqual(context.BitwiseAnd(returnAddress, Const(~CallFlag)), nextAddr));
context.Return(returnAddress);
context.MarkLabel(lblContinue);
if (context.CurrBlock.Next == null)
{
// No code following this instruction, try and find the next block and jump to it.
EmitTailContinue(context, nextAddr);
}
}
private static ulong GetNextOpAddress(OpCode op)
{
return op.Address + (ulong)op.OpCodeSizeInBytes;
}
public static void EmitTailContinue(ArmEmitterContext context, Operand address, bool allowRejit = false)
{
bool useTailContinue = true; // Left option here as it may be useful if we need to return to managed rather than tail call in future. (eg. for debug)
if (useTailContinue)
{
if (context.HighCq)
{
// If we're doing a tail continue in HighCq, reserve a space in the jump table to avoid calling back to the translator.
// This will always try to get a HighCq version of our continue target as well.
EmitJumpTableBranch(context, address, true);
}
else
{
if (allowRejit)
{
address = context.BitwiseOr(address, Const(CallFlag));
}
Operand fallbackAddr = context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFunctionAddress)), address);
EmitNativeCall(context, fallbackAddr, true);
}
}
else
{
context.Return(address);
}
}
private static void EmitNativeCallWithGuestAddress(ArmEmitterContext context, Operand funcAddr, Operand guestAddress, bool isJump)
{
Operand nativeContextPtr = context.LoadArgument(OperandType.I64, 0);
context.Store(context.Add(nativeContextPtr, Const(NativeContext.GetCallAddressOffset())), guestAddress);
EmitNativeCall(context, nativeContextPtr, funcAddr, isJump);
}
private static void EmitBranchFallback(ArmEmitterContext context, Operand address, bool isJump)
{
address = context.BitwiseOr(address, Const(address.Type, (long)CallFlag)); // Set call flag.
Operand fallbackAddr = context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.GetFunctionAddress)), address);
EmitNativeCall(context, fallbackAddr, isJump);
}
public static void EmitDynamicTableCall(ArmEmitterContext context, Operand tableAddress, Operand address, bool isJump)
{
// Loop over elements of the dynamic table. Unrolled loop.
Operand endLabel = Label();
Operand fallbackLabel = Label();
Action<Operand> emitTableEntry = (Operand entrySkipLabel) =>
{
// Try to take this entry in the table if its guest address equals 0.
Operand gotResult = context.CompareAndSwap(tableAddress, Const(0L), address);
// Is the address ours? (either taken via CompareAndSwap (0), or what was already here)
context.BranchIfFalse(entrySkipLabel, context.BitwiseOr(context.ICompareEqual(gotResult, address), context.ICompareEqual(gotResult, Const(0L))));
// It's ours, so what function is it pointing to?
Operand targetFunctionPtr = context.Add(tableAddress, Const(8L));
Operand targetFunction = context.Load(OperandType.I64, targetFunctionPtr);
// Call the function.
// We pass in the entry address as the guest address, as the entry may need to be updated by the indirect call stub.
EmitNativeCallWithGuestAddress(context, targetFunction, tableAddress, isJump);
context.Branch(endLabel);
};
// Currently this uses a size of 1, as higher values inflate code size for no real benefit.
for (int i = 0; i < JumpTable.DynamicTableElems; i++)
{
if (i == JumpTable.DynamicTableElems - 1)
{
emitTableEntry(fallbackLabel); // If this is the last entry, avoid emitting the additional label and add.
}
else
{
Operand nextLabel = Label();
emitTableEntry(nextLabel);
context.MarkLabel(nextLabel);
tableAddress = context.Add(tableAddress, Const((long)JumpTable.JumpTableStride)); // Move to the next table entry.
}
}
context.MarkLabel(fallbackLabel);
EmitBranchFallback(context, address, isJump);
context.MarkLabel(endLabel);
}
public static void EmitJumpTableBranch(ArmEmitterContext context, Operand address, bool isJump = false)
{
if (address.Type == OperandType.I32)
{
address = context.ZeroExtend32(OperandType.I64, address);
}
// TODO: Constant folding. Indirect calls are slower in the best case and emit more code so we want to avoid them when possible.
bool isConst = address.Kind == OperandKind.Constant;
long constAddr = (long)address.Value;
if (!context.HighCq)
{
// Don't emit indirect calls or jumps if we're compiling in lowCq mode.
// This avoids wasting space on the jump and indirect tables.
// Just ask the translator for the function address.
EmitBranchFallback(context, address, isJump);
}
else if (!isConst)
{
// Virtual branch/call - store first used addresses on a small table for fast lookup.
int entry = context.JumpTable.ReserveDynamicEntry(isJump);
int jumpOffset = entry * JumpTable.JumpTableStride * JumpTable.DynamicTableElems;
Operand dynTablePtr;
if (Ptc.State == PtcState.Disabled)
{
dynTablePtr = Const(context.JumpTable.DynamicPointer.ToInt64() + jumpOffset);
}
else
{
dynTablePtr = Const(context.JumpTable.DynamicPointer.ToInt64(), true, Ptc.DynamicPointerIndex);
dynTablePtr = context.Add(dynTablePtr, Const((long)jumpOffset));
}
EmitDynamicTableCall(context, dynTablePtr, address, isJump);
}
else
{
int entry = context.JumpTable.ReserveTableEntry(context.BaseAddress & (~3L), constAddr, isJump);
int jumpOffset = entry * JumpTable.JumpTableStride + 8; // Offset directly to the host address.
Operand tableEntryPtr;
if (Ptc.State == PtcState.Disabled)
{
tableEntryPtr = Const(context.JumpTable.JumpPointer.ToInt64() + jumpOffset);
}
else
{
tableEntryPtr = Const(context.JumpTable.JumpPointer.ToInt64(), true, Ptc.JumpPointerIndex);
tableEntryPtr = context.Add(tableEntryPtr, Const((long)jumpOffset));
}
Operand funcAddr = context.Load(OperandType.I64, tableEntryPtr);
EmitNativeCallWithGuestAddress(context, funcAddr, address, isJump); // Call the function directly. If it's not present yet, this will call the direct call stub.
}
}
}
}