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>
This commit is contained in:
LDj3SNuD 2020-06-16 20:28:02 +02:00 committed by GitHub
parent fa286d3535
commit 5e724cf24e
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
78 changed files with 3066 additions and 1207 deletions

View file

@ -13,22 +13,22 @@ using static ARMeilleure.IntermediateRepresentation.OperationHelper;
namespace ARMeilleure.Translation
{
using PTC;
public class Translator
{
private const ulong CallFlag = InstEmitFlowHelper.CallFlag;
private const bool AlwaysTranslateFunctions = true; // If false, only translates a single block for lowCq.
private readonly IMemoryManager _memory;
private readonly ConcurrentDictionary<ulong, TranslatedFunction> _funcs;
private readonly JumpTable _jumpTable;
private readonly PriorityQueue<RejitRequest> _backgroundQueue;
private readonly ConcurrentStack<RejitRequest> _backgroundStack;
private readonly AutoResetEvent _backgroundTranslatorEvent;
private readonly JumpTable _jumpTable;
private volatile int _threadCount;
public Translator(IJitMemoryAllocator allocator, IMemoryManager memory)
@ -37,32 +37,45 @@ namespace ARMeilleure.Translation
_funcs = new ConcurrentDictionary<ulong, TranslatedFunction>();
_jumpTable = new JumpTable(allocator);
_backgroundQueue = new PriorityQueue<RejitRequest>(2);
_backgroundStack = new ConcurrentStack<RejitRequest>();
_backgroundTranslatorEvent = new AutoResetEvent(false);
_jumpTable = new JumpTable(allocator);
JitCache.Initialize(allocator);
DirectCallStubs.InitializeStubs();
if (Ptc.State == PtcState.Enabled)
{
Ptc.LoadTranslations(_funcs, memory.PageTablePointer, _jumpTable);
}
}
private void TranslateQueuedSubs()
private void TranslateStackedSubs()
{
while (_threadCount != 0)
{
if (_backgroundQueue.TryDequeue(out RejitRequest request))
if (_backgroundStack.TryPop(out RejitRequest request))
{
TranslatedFunction func = Translate(request.Address, request.Mode, highCq: true);
TranslatedFunction func = Translate(_memory, _jumpTable, request.Address, request.Mode, highCq: true);
_funcs.AddOrUpdate(request.Address, func, (key, oldFunc) => func);
_jumpTable.RegisterFunction(request.Address, func);
if (PtcProfiler.Enabled)
{
PtcProfiler.UpdateEntry(request.Address, request.Mode, highCq: true);
}
}
else
{
_backgroundTranslatorEvent.WaitOne();
}
}
_backgroundTranslatorEvent.Set(); // Wake up any other background translator threads, to encourage them to exit.
}
@ -70,16 +83,27 @@ namespace ARMeilleure.Translation
{
if (Interlocked.Increment(ref _threadCount) == 1)
{
if (Ptc.State == PtcState.Enabled)
{
Ptc.MakeAndSaveTranslations(_funcs, _memory, _jumpTable);
}
PtcProfiler.Start();
Ptc.Disable();
// Simple heuristic, should be user configurable in future. (1 for 4 core/ht or less, 2 for 6 core+ht etc).
// All threads are normal priority except from the last, which just fills as much of the last core as the os lets it with a low priority.
// If we only have one rejit thread, it should be normal priority as highCq code is performance critical.
// TODO: Use physical cores rather than logical. This only really makes sense for processors with hyperthreading. Requires OS specific code.
int unboundedThreadCount = Math.Max(1, (Environment.ProcessorCount - 6) / 3);
int threadCount = Math.Min(4, unboundedThreadCount);
int threadCount = Math.Min(4, unboundedThreadCount);
for (int i = 0; i < threadCount; i++)
{
bool last = i != 0 && i == unboundedThreadCount - 1;
Thread backgroundTranslatorThread = new Thread(TranslateQueuedSubs)
Thread backgroundTranslatorThread = new Thread(TranslateStackedSubs)
{
Name = "CPU.BackgroundTranslatorThread." + i,
Priority = last ? ThreadPriority.Lowest : ThreadPriority.Normal
@ -130,13 +154,19 @@ namespace ARMeilleure.Translation
if (!_funcs.TryGetValue(address, out TranslatedFunction func))
{
func = Translate(address, mode, highCq: false);
func = Translate(_memory, _jumpTable, address, mode, highCq: false);
_funcs.TryAdd(address, func);
if (PtcProfiler.Enabled)
{
PtcProfiler.AddEntry(address, mode, highCq: false);
}
}
else if (isCallTarget && func.ShouldRejit())
if (isCallTarget && func.ShouldRejit())
{
_backgroundQueue.Enqueue(0, new RejitRequest(address, mode));
_backgroundStack.Push(new RejitRequest(address, mode));
_backgroundTranslatorEvent.Set();
}
@ -144,18 +174,16 @@ namespace ARMeilleure.Translation
return func;
}
private TranslatedFunction Translate(ulong address, ExecutionMode mode, bool highCq)
internal static TranslatedFunction Translate(IMemoryManager memory, JumpTable jumpTable, ulong address, ExecutionMode mode, bool highCq)
{
ArmEmitterContext context = new ArmEmitterContext(_memory, _jumpTable, (long)address, highCq, Aarch32Mode.User);
ArmEmitterContext context = new ArmEmitterContext(memory, jumpTable, (long)address, highCq, Aarch32Mode.User);
PrepareOperandPool(highCq);
PrepareOperationPool(highCq);
Logger.StartPass(PassName.Decoding);
Block[] blocks = AlwaysTranslateFunctions
? Decoder.DecodeFunction (_memory, address, mode, highCq)
: Decoder.DecodeBasicBlock(_memory, address, mode);
Block[] blocks = Decoder.DecodeFunction(memory, address, mode, highCq);
Logger.EndPass(PassName.Decoding);
@ -182,12 +210,26 @@ namespace ARMeilleure.Translation
CompilerOptions options = highCq ? CompilerOptions.HighCq : CompilerOptions.None;
GuestFunction func = Compiler.Compile<GuestFunction>(cfg, argTypes, OperandType.I64, options);
GuestFunction func;
if (Ptc.State == PtcState.Disabled)
{
func = Compiler.Compile<GuestFunction>(cfg, argTypes, OperandType.I64, options);
}
else
{
using (PtcInfo ptcInfo = new PtcInfo())
{
func = Compiler.Compile<GuestFunction>(cfg, argTypes, OperandType.I64, options, ptcInfo);
Ptc.WriteInfoCodeReloc((long)address, highCq, ptcInfo);
}
}
ResetOperandPool(highCq);
ResetOperationPool(highCq);
return new TranslatedFunction(func, rejit: !highCq);
return new TranslatedFunction(func, highCq);
}
private static ControlFlowGraph EmitAndGetCFG(ArmEmitterContext context, Block[] blocks)
@ -264,7 +306,7 @@ namespace ARMeilleure.Translation
context.BranchIfTrue(lblNonZero, count);
Operand running = context.Call(new _Bool(NativeInterface.CheckSynchronization));
Operand running = context.Call(typeof(NativeInterface).GetMethod(nameof(NativeInterface.CheckSynchronization)));
context.BranchIfTrue(lblExit, running);
@ -281,4 +323,4 @@ namespace ARMeilleure.Translation
context.MarkLabel(lblExit);
}
}
}
}