R/Ryujinx.HLE/HOS/Kernel/Threading/KThread.cs

1437 lines
43 KiB
C#
Raw Normal View History

using Ryujinx.Common.Logging;
using Ryujinx.Cpu;
using Ryujinx.HLE.HOS.Kernel.Common;
using Ryujinx.HLE.HOS.Kernel.Process;
using Ryujinx.HLE.HOS.Kernel.SupervisorCall;
using System;
using System.Collections.Generic;
using System.Numerics;
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
using System.Threading;
namespace Ryujinx.HLE.HOS.Kernel.Threading
2018-02-04 23:08:20 +00:00
{
class KThread : KSynchronizationObject, IKFutureSchedulerObject
2018-02-04 23:08:20 +00:00
{
private const int TlsUserDisableCountOffset = 0x100;
private const int TlsUserInterruptFlagOffset = 0x102;
public const int MaxWaitSyncObjects = 64;
private ManualResetEvent _schedulerWaitEvent;
public ManualResetEvent SchedulerWaitEvent => _schedulerWaitEvent;
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
public Thread HostThread { get; private set; }
public IExecutionContext Context { get; private set; }
2018-02-04 23:08:20 +00:00
public KThreadContext ThreadContext { get; private set; }
public int DynamicPriority { get; set; }
public ulong AffinityMask { get; set; }
public ulong ThreadUid { get; private set; }
private long _totalTimeRunning;
public long TotalTimeRunning => _totalTimeRunning;
public KSynchronizationObject SignaledObj { get; set; }
public ulong CondVarAddress { get; set; }
private ulong _entrypoint;
private ThreadStart _customThreadStart;
private bool _forcedUnschedulable;
public bool IsSchedulable => _customThreadStart == null && !_forcedUnschedulable;
public ulong MutexAddress { get; set; }
public int KernelWaitersCount { get; private set; }
public KProcess Owner { get; private set; }
private ulong _tlsAddress;
public ulong TlsAddress => _tlsAddress;
public KSynchronizationObject[] WaitSyncObjects { get; }
public int[] WaitSyncHandles { get; }
public long LastScheduledTime { get; set; }
public LinkedListNode<KThread>[] SiblingsPerCore { get; private set; }
public LinkedList<KThread> Withholder { get; set; }
public LinkedListNode<KThread> WithholderNode { get; set; }
public LinkedListNode<KThread> ProcessListNode { get; set; }
private LinkedList<KThread> _mutexWaiters;
private LinkedListNode<KThread> _mutexWaiterNode;
private LinkedList<KThread> _pinnedWaiters;
public KThread MutexOwner { get; private set; }
2018-06-26 04:09:32 +00:00
public int ThreadHandleForUserMutex { get; set; }
private ThreadSchedState _forcePauseFlags;
private ThreadSchedState _forcePausePermissionFlags;
public KernelResult ObjSyncResult { get; set; }
public int BasePriority { get; set; }
public int PreferredCore { get; set; }
public int CurrentCore { get; set; }
public int ActiveCore { get; set; }
public bool IsPinned { get; private set; }
private ulong _originalAffinityMask;
private int _originalPreferredCore;
private int _originalBasePriority;
private int _coreMigrationDisableCount;
public ThreadSchedState SchedFlags { get; private set; }
private int _shallBeTerminated;
public bool ShallBeTerminated
{
get => _shallBeTerminated != 0;
set => _shallBeTerminated = value ? 1 : 0;
}
public bool TerminationRequested => ShallBeTerminated || SchedFlags == ThreadSchedState.TerminationPending;
public bool SyncCancelled { get; set; }
public bool WaitingSync { get; set; }
private int _hasExited;
private bool _hasBeenInitialized;
private bool _hasBeenReleased;
public bool WaitingInArbitration { get; set; }
private object _activityOperationLock;
public KThread(KernelContext context) : base(context)
2018-02-04 23:08:20 +00:00
{
WaitSyncObjects = new KSynchronizationObject[MaxWaitSyncObjects];
WaitSyncHandles = new int[MaxWaitSyncObjects];
SiblingsPerCore = new LinkedListNode<KThread>[KScheduler.CpuCoresCount];
_mutexWaiters = new LinkedList<KThread>();
_pinnedWaiters = new LinkedList<KThread>();
_activityOperationLock = new object();
}
public KernelResult Initialize(
ulong entrypoint,
ulong argsPtr,
ulong stackTop,
int priority,
int cpuCore,
KProcess owner,
ThreadType type,
ThreadStart customThreadStart = null)
{
if ((uint)type > 3)
{
throw new ArgumentException($"Invalid thread type \"{type}\".");
}
ThreadContext = new KThreadContext();
PreferredCore = cpuCore;
AffinityMask |= 1UL << cpuCore;
SchedFlags = type == ThreadType.Dummy
? ThreadSchedState.Running
: ThreadSchedState.None;
ActiveCore = cpuCore;
ObjSyncResult = KernelResult.ThreadNotStarted;
DynamicPriority = priority;
BasePriority = priority;
CurrentCore = cpuCore;
IsPinned = false;
_entrypoint = entrypoint;
_customThreadStart = customThreadStart;
if (type == ThreadType.User)
{
if (owner.AllocateThreadLocalStorage(out _tlsAddress) != KernelResult.Success)
{
return KernelResult.OutOfMemory;
}
MemoryHelper.FillWithZeros(owner.CpuMemory, _tlsAddress, KTlsPageInfo.TlsEntrySize);
}
bool is64Bits;
if (owner != null)
{
Owner = owner;
owner.IncrementReferenceCount();
owner.IncrementThreadCount();
is64Bits = owner.Flags.HasFlag(ProcessCreationFlags.Is64Bit);
}
else
{
is64Bits = true;
}
HostThread = new Thread(ThreadStart);
Context = owner?.CreateExecutionContext() ?? new ProcessExecutionContext();
Context.IsAarch32 = !is64Bits;
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
Context.SetX(0, argsPtr);
if (is64Bits)
{
Context.SetX(18, KSystemControl.GenerateRandom() | 1);
Context.SetX(31, stackTop);
}
else
{
Context.SetX(13, (uint)stackTop);
}
Context.TpidrroEl0 = (long)_tlsAddress;
ThreadUid = KernelContext.NewThreadUid();
HostThread.Name = customThreadStart != null ? $"HLE.OsThread.{ThreadUid}" : $"HLE.GuestThread.{ThreadUid}";
_hasBeenInitialized = true;
_forcePausePermissionFlags = ThreadSchedState.ForcePauseMask;
if (owner != null)
{
owner.AddThread(this);
if (owner.IsPaused)
{
KernelContext.CriticalSection.Enter();
if (TerminationRequested)
{
KernelContext.CriticalSection.Leave();
return KernelResult.Success;
}
_forcePauseFlags |= ThreadSchedState.ProcessPauseFlag;
CombineForcePauseFlags();
KernelContext.CriticalSection.Leave();
}
}
return KernelResult.Success;
}
public KernelResult Start()
{
if (!KernelContext.KernelInitialized)
{
KernelContext.CriticalSection.Enter();
if (!TerminationRequested)
{
_forcePauseFlags |= ThreadSchedState.KernelInitPauseFlag;
CombineForcePauseFlags();
}
KernelContext.CriticalSection.Leave();
}
KernelResult result = KernelResult.ThreadTerminating;
KernelContext.CriticalSection.Enter();
if (!ShallBeTerminated)
{
KThread currentThread = KernelStatic.GetCurrentThread();
while (SchedFlags != ThreadSchedState.TerminationPending && (currentThread == null || !currentThread.TerminationRequested))
{
if ((SchedFlags & ThreadSchedState.LowMask) != ThreadSchedState.None)
{
result = KernelResult.InvalidState;
break;
}
if (currentThread == null || currentThread._forcePauseFlags == ThreadSchedState.None)
{
if (Owner != null && _forcePauseFlags != ThreadSchedState.None)
{
CombineForcePauseFlags();
}
SetNewSchedFlags(ThreadSchedState.Running);
StartHostThread();
result = KernelResult.Success;
break;
}
else
{
currentThread.CombineForcePauseFlags();
KernelContext.CriticalSection.Leave();
KernelContext.CriticalSection.Enter();
if (currentThread.ShallBeTerminated)
{
break;
}
}
}
}
KernelContext.CriticalSection.Leave();
return result;
}
public ThreadSchedState PrepareForTermination()
{
KernelContext.CriticalSection.Enter();
if (Owner != null && Owner.PinnedThreads[KernelStatic.GetCurrentThread().CurrentCore] == this)
{
Owner.UnpinThread(this);
}
ThreadSchedState result;
if (Interlocked.CompareExchange(ref _shallBeTerminated, 1, 0) == 0)
{
if ((SchedFlags & ThreadSchedState.LowMask) == ThreadSchedState.None)
{
SchedFlags = ThreadSchedState.TerminationPending;
}
else
{
if (_forcePauseFlags != ThreadSchedState.None)
{
_forcePauseFlags &= ~ThreadSchedState.ThreadPauseFlag;
ThreadSchedState oldSchedFlags = SchedFlags;
SchedFlags &= ThreadSchedState.LowMask;
AdjustScheduling(oldSchedFlags);
}
if (BasePriority >= 0x10)
{
SetPriority(0xF);
}
if ((SchedFlags & ThreadSchedState.LowMask) == ThreadSchedState.Running)
{
// TODO: GIC distributor stuffs (sgir changes ect)
Context.RequestInterrupt();
}
SignaledObj = null;
ObjSyncResult = KernelResult.ThreadTerminating;
ReleaseAndResume();
}
}
result = SchedFlags;
KernelContext.CriticalSection.Leave();
return result & ThreadSchedState.LowMask;
}
public void Terminate()
{
ThreadSchedState state = PrepareForTermination();
if (state != ThreadSchedState.TerminationPending)
{
KernelContext.Synchronization.WaitFor(new KSynchronizationObject[] { this }, -1, out _);
}
}
public void HandlePostSyscall()
{
ThreadSchedState state;
do
{
if (TerminationRequested)
{
Exit();
// As the death of the thread is handled by the CPU emulator, we differ from the official kernel and return here.
break;
}
KernelContext.CriticalSection.Enter();
if (TerminationRequested)
{
state = ThreadSchedState.TerminationPending;
}
else
{
if (_forcePauseFlags != ThreadSchedState.None)
{
CombineForcePauseFlags();
}
state = ThreadSchedState.Running;
}
KernelContext.CriticalSection.Leave();
} while (state == ThreadSchedState.TerminationPending);
}
public void Exit()
{
// TODO: Debug event.
if (Owner != null)
{
Owner.ResourceLimit?.Release(LimitableResource.Thread, 0, 1);
_hasBeenReleased = true;
}
KernelContext.CriticalSection.Enter();
_forcePauseFlags &= ~ThreadSchedState.ForcePauseMask;
_forcePausePermissionFlags = 0;
bool decRef = ExitImpl();
Context.StopRunning();
KernelContext.CriticalSection.Leave();
if (decRef)
{
DecrementReferenceCount();
}
}
private bool ExitImpl()
{
KernelContext.CriticalSection.Enter();
SetNewSchedFlags(ThreadSchedState.TerminationPending);
bool decRef = Interlocked.Exchange(ref _hasExited, 1) == 0;
Signal();
KernelContext.CriticalSection.Leave();
return decRef;
}
private int GetEffectiveRunningCore()
{
for (int coreNumber = 0; coreNumber < KScheduler.CpuCoresCount; coreNumber++)
{
if (KernelContext.Schedulers[coreNumber].CurrentThread == this)
{
return coreNumber;
}
}
return -1;
}
public KernelResult Sleep(long timeout)
{
KernelContext.CriticalSection.Enter();
if (ShallBeTerminated || SchedFlags == ThreadSchedState.TerminationPending)
{
KernelContext.CriticalSection.Leave();
return KernelResult.ThreadTerminating;
}
SetNewSchedFlags(ThreadSchedState.Paused);
if (timeout > 0)
{
KernelContext.TimeManager.ScheduleFutureInvocation(this, timeout);
}
KernelContext.CriticalSection.Leave();
if (timeout > 0)
{
KernelContext.TimeManager.UnscheduleFutureInvocation(this);
}
return 0;
}
public void SetPriority(int priority)
{
KernelContext.CriticalSection.Enter();
if (IsPinned)
{
_originalBasePriority = priority;
}
else
{
BasePriority = priority;
}
UpdatePriorityInheritance();
KernelContext.CriticalSection.Leave();
}
public void Suspend(ThreadSchedState type)
{
_forcePauseFlags |= type;
CombineForcePauseFlags();
}
public void Resume(ThreadSchedState type)
{
ThreadSchedState oldForcePauseFlags = _forcePauseFlags;
_forcePauseFlags &= ~type;
if ((oldForcePauseFlags & ~type) == ThreadSchedState.None)
{
ThreadSchedState oldSchedFlags = SchedFlags;
SchedFlags &= ThreadSchedState.LowMask;
AdjustScheduling(oldSchedFlags);
}
}
public KernelResult SetActivity(bool pause)
{
lock (_activityOperationLock)
{
KernelResult result = KernelResult.Success;
KernelContext.CriticalSection.Enter();
ThreadSchedState lowNibble = SchedFlags & ThreadSchedState.LowMask;
if (lowNibble != ThreadSchedState.Paused && lowNibble != ThreadSchedState.Running)
{
KernelContext.CriticalSection.Leave();
return KernelResult.InvalidState;
}
if (!ShallBeTerminated && SchedFlags != ThreadSchedState.TerminationPending)
{
if (pause)
{
// Pause, the force pause flag should be clear (thread is NOT paused).
if ((_forcePauseFlags & ThreadSchedState.ThreadPauseFlag) == 0)
{
Suspend(ThreadSchedState.ThreadPauseFlag);
}
else
{
result = KernelResult.InvalidState;
}
}
else
{
// Unpause, the force pause flag should be set (thread is paused).
if ((_forcePauseFlags & ThreadSchedState.ThreadPauseFlag) != 0)
{
Resume(ThreadSchedState.ThreadPauseFlag);
}
else
{
result = KernelResult.InvalidState;
}
}
}
KernelContext.CriticalSection.Leave();
if (result == KernelResult.Success && pause)
{
bool isThreadRunning = true;
while (isThreadRunning)
{
KernelContext.CriticalSection.Enter();
if (TerminationRequested)
{
KernelContext.CriticalSection.Leave();
break;
}
isThreadRunning = false;
if (IsPinned)
{
KThread currentThread = KernelStatic.GetCurrentThread();
if (currentThread.TerminationRequested)
{
KernelContext.CriticalSection.Leave();
result = KernelResult.ThreadTerminating;
break;
}
_pinnedWaiters.AddLast(currentThread);
currentThread.Reschedule(ThreadSchedState.Paused);
}
else
{
isThreadRunning = GetEffectiveRunningCore() >= 0;
}
KernelContext.CriticalSection.Leave();
}
}
return result;
}
}
public KernelResult GetThreadContext3(out ThreadContext context)
{
context = default;
lock (_activityOperationLock)
{
KernelContext.CriticalSection.Enter();
if ((_forcePauseFlags & ThreadSchedState.ThreadPauseFlag) == 0)
{
KernelContext.CriticalSection.Leave();
return KernelResult.InvalidState;
}
if (!TerminationRequested)
{
context = GetCurrentContext();
}
KernelContext.CriticalSection.Leave();
}
return KernelResult.Success;
}
private static uint GetPsr(IExecutionContext context)
{
return context.Pstate & 0xFF0FFE20;
}
private ThreadContext GetCurrentContext()
{
const int MaxRegistersAArch32 = 15;
const int MaxFpuRegistersAArch32 = 16;
ThreadContext context = new ThreadContext();
if (Owner.Flags.HasFlag(ProcessCreationFlags.Is64Bit))
{
for (int i = 0; i < context.Registers.Length; i++)
{
context.Registers[i] = Context.GetX(i);
}
for (int i = 0; i < context.FpuRegisters.Length; i++)
{
context.FpuRegisters[i] = Context.GetV(i);
}
context.Fp = Context.GetX(29);
context.Lr = Context.GetX(30);
context.Sp = Context.GetX(31);
context.Pc = Context.Pc;
context.Pstate = GetPsr(Context);
context.Tpidr = (ulong)Context.TpidrroEl0;
}
else
{
for (int i = 0; i < MaxRegistersAArch32; i++)
{
context.Registers[i] = (uint)Context.GetX(i);
}
for (int i = 0; i < MaxFpuRegistersAArch32; i++)
{
context.FpuRegisters[i] = Context.GetV(i);
}
context.Pc = (uint)Context.Pc;
context.Pstate = GetPsr(Context);
context.Tpidr = (uint)Context.TpidrroEl0;
}
context.Fpcr = (uint)Context.Fpcr;
context.Fpsr = (uint)Context.Fpsr;
return context;
}
public void CancelSynchronization()
{
KernelContext.CriticalSection.Enter();
if ((SchedFlags & ThreadSchedState.LowMask) != ThreadSchedState.Paused || !WaitingSync)
{
SyncCancelled = true;
}
else if (Withholder != null)
{
Withholder.Remove(WithholderNode);
SetNewSchedFlags(ThreadSchedState.Running);
Withholder = null;
SyncCancelled = true;
}
else
{
SignaledObj = null;
ObjSyncResult = KernelResult.Cancelled;
SetNewSchedFlags(ThreadSchedState.Running);
SyncCancelled = false;
}
KernelContext.CriticalSection.Leave();
}
public KernelResult SetCoreAndAffinityMask(int newCore, ulong newAffinityMask)
{
lock (_activityOperationLock)
{
KernelContext.CriticalSection.Enter();
bool isCoreMigrationDisabled = _coreMigrationDisableCount != 0;
// The value -3 is "do not change the preferred core".
if (newCore == -3)
{
newCore = isCoreMigrationDisabled ? _originalPreferredCore : PreferredCore;
if ((newAffinityMask & (1UL << newCore)) == 0)
{
KernelContext.CriticalSection.Leave();
return KernelResult.InvalidCombination;
}
}
if (isCoreMigrationDisabled)
{
_originalPreferredCore = newCore;
_originalAffinityMask = newAffinityMask;
}
else
{
ulong oldAffinityMask = AffinityMask;
PreferredCore = newCore;
AffinityMask = newAffinityMask;
if (oldAffinityMask != newAffinityMask)
{
int oldCore = ActiveCore;
if (oldCore >= 0 && ((AffinityMask >> oldCore) & 1) == 0)
{
if (PreferredCore < 0)
{
ActiveCore = sizeof(ulong) * 8 - 1 - BitOperations.LeadingZeroCount(AffinityMask);
}
else
{
ActiveCore = PreferredCore;
}
}
AdjustSchedulingForNewAffinity(oldAffinityMask, oldCore);
}
}
KernelContext.CriticalSection.Leave();
bool targetThreadPinned = true;
while (targetThreadPinned)
{
KernelContext.CriticalSection.Enter();
if (TerminationRequested)
{
KernelContext.CriticalSection.Leave();
break;
}
targetThreadPinned = false;
int coreNumber = GetEffectiveRunningCore();
bool isPinnedThreadCurrentlyRunning = coreNumber >= 0;
if (isPinnedThreadCurrentlyRunning && ((1UL << coreNumber) & AffinityMask) == 0)
{
if (IsPinned)
{
KThread currentThread = KernelStatic.GetCurrentThread();
if (currentThread.TerminationRequested)
{
KernelContext.CriticalSection.Leave();
return KernelResult.ThreadTerminating;
}
_pinnedWaiters.AddLast(currentThread);
currentThread.Reschedule(ThreadSchedState.Paused);
}
else
{
targetThreadPinned = true;
}
}
KernelContext.CriticalSection.Leave();
}
return KernelResult.Success;
}
}
private void CombineForcePauseFlags()
{
ThreadSchedState oldFlags = SchedFlags;
ThreadSchedState lowNibble = SchedFlags & ThreadSchedState.LowMask;
SchedFlags = lowNibble | (_forcePauseFlags & _forcePausePermissionFlags);
AdjustScheduling(oldFlags);
}
private void SetNewSchedFlags(ThreadSchedState newFlags)
{
KernelContext.CriticalSection.Enter();
ThreadSchedState oldFlags = SchedFlags;
SchedFlags = (oldFlags & ThreadSchedState.HighMask) | newFlags;
if ((oldFlags & ThreadSchedState.LowMask) != newFlags)
{
AdjustScheduling(oldFlags);
}
KernelContext.CriticalSection.Leave();
}
public void ReleaseAndResume()
{
KernelContext.CriticalSection.Enter();
if ((SchedFlags & ThreadSchedState.LowMask) == ThreadSchedState.Paused)
{
if (Withholder != null)
{
Withholder.Remove(WithholderNode);
SetNewSchedFlags(ThreadSchedState.Running);
Withholder = null;
}
else
{
SetNewSchedFlags(ThreadSchedState.Running);
}
}
KernelContext.CriticalSection.Leave();
}
public void Reschedule(ThreadSchedState newFlags)
{
KernelContext.CriticalSection.Enter();
ThreadSchedState oldFlags = SchedFlags;
SchedFlags = (oldFlags & ThreadSchedState.HighMask) |
(newFlags & ThreadSchedState.LowMask);
AdjustScheduling(oldFlags);
KernelContext.CriticalSection.Leave();
}
public void AddMutexWaiter(KThread requester)
{
AddToMutexWaitersList(requester);
requester.MutexOwner = this;
UpdatePriorityInheritance();
}
public void RemoveMutexWaiter(KThread thread)
{
if (thread._mutexWaiterNode?.List != null)
{
_mutexWaiters.Remove(thread._mutexWaiterNode);
}
thread.MutexOwner = null;
UpdatePriorityInheritance();
}
public KThread RelinquishMutex(ulong mutexAddress, out int count)
{
count = 0;
if (_mutexWaiters.First == null)
{
return null;
}
KThread newMutexOwner = null;
LinkedListNode<KThread> currentNode = _mutexWaiters.First;
do
{
// Skip all threads that are not waiting for this mutex.
while (currentNode != null && currentNode.Value.MutexAddress != mutexAddress)
{
currentNode = currentNode.Next;
}
if (currentNode == null)
{
break;
}
LinkedListNode<KThread> nextNode = currentNode.Next;
_mutexWaiters.Remove(currentNode);
currentNode.Value.MutexOwner = newMutexOwner;
if (newMutexOwner != null)
{
// New owner was already selected, re-insert on new owner list.
newMutexOwner.AddToMutexWaitersList(currentNode.Value);
}
else
{
// New owner not selected yet, use current thread.
newMutexOwner = currentNode.Value;
}
count++;
currentNode = nextNode;
}
while (currentNode != null);
if (newMutexOwner != null)
{
UpdatePriorityInheritance();
newMutexOwner.UpdatePriorityInheritance();
}
return newMutexOwner;
}
private void UpdatePriorityInheritance()
{
// If any of the threads waiting for the mutex has
// higher priority than the current thread, then
// the current thread inherits that priority.
int highestPriority = BasePriority;
if (_mutexWaiters.First != null)
{
int waitingDynamicPriority = _mutexWaiters.First.Value.DynamicPriority;
if (waitingDynamicPriority < highestPriority)
{
highestPriority = waitingDynamicPriority;
}
}
if (highestPriority != DynamicPriority)
{
int oldPriority = DynamicPriority;
DynamicPriority = highestPriority;
AdjustSchedulingForNewPriority(oldPriority);
if (MutexOwner != null)
{
// Remove and re-insert to ensure proper sorting based on new priority.
MutexOwner._mutexWaiters.Remove(_mutexWaiterNode);
MutexOwner.AddToMutexWaitersList(this);
MutexOwner.UpdatePriorityInheritance();
}
}
}
private void AddToMutexWaitersList(KThread thread)
{
LinkedListNode<KThread> nextPrio = _mutexWaiters.First;
int currentPriority = thread.DynamicPriority;
while (nextPrio != null && nextPrio.Value.DynamicPriority <= currentPriority)
{
nextPrio = nextPrio.Next;
}
if (nextPrio != null)
{
thread._mutexWaiterNode = _mutexWaiters.AddBefore(nextPrio, thread);
}
else
{
thread._mutexWaiterNode = _mutexWaiters.AddLast(thread);
}
}
private void AdjustScheduling(ThreadSchedState oldFlags)
{
if (oldFlags == SchedFlags)
{
return;
}
if (!IsSchedulable)
{
if (!_forcedUnschedulable)
{
// Ensure our thread is running and we have an event.
StartHostThread();
// If the thread is not schedulable, we want to just run or pause
// it directly as we don't care about priority or the core it is
// running on in this case.
if (SchedFlags == ThreadSchedState.Running)
{
_schedulerWaitEvent.Set();
}
else
{
_schedulerWaitEvent.Reset();
}
}
return;
}
if (oldFlags == ThreadSchedState.Running)
{
// Was running, now it's stopped.
if (ActiveCore >= 0)
{
KernelContext.PriorityQueue.Unschedule(DynamicPriority, ActiveCore, this);
}
for (int core = 0; core < KScheduler.CpuCoresCount; core++)
{
if (core != ActiveCore && ((AffinityMask >> core) & 1) != 0)
{
KernelContext.PriorityQueue.Unsuggest(DynamicPriority, core, this);
}
}
}
else if (SchedFlags == ThreadSchedState.Running)
{
// Was stopped, now it's running.
if (ActiveCore >= 0)
{
KernelContext.PriorityQueue.Schedule(DynamicPriority, ActiveCore, this);
}
for (int core = 0; core < KScheduler.CpuCoresCount; core++)
{
if (core != ActiveCore && ((AffinityMask >> core) & 1) != 0)
{
KernelContext.PriorityQueue.Suggest(DynamicPriority, core, this);
}
}
}
KernelContext.ThreadReselectionRequested = true;
}
private void AdjustSchedulingForNewPriority(int oldPriority)
{
if (SchedFlags != ThreadSchedState.Running || !IsSchedulable)
{
return;
}
// Remove thread from the old priority queues.
if (ActiveCore >= 0)
{
KernelContext.PriorityQueue.Unschedule(oldPriority, ActiveCore, this);
}
for (int core = 0; core < KScheduler.CpuCoresCount; core++)
{
if (core != ActiveCore && ((AffinityMask >> core) & 1) != 0)
{
KernelContext.PriorityQueue.Unsuggest(oldPriority, core, this);
}
}
// Add thread to the new priority queues.
KThread currentThread = KernelStatic.GetCurrentThread();
if (ActiveCore >= 0)
{
if (currentThread == this)
{
KernelContext.PriorityQueue.SchedulePrepend(DynamicPriority, ActiveCore, this);
}
else
{
KernelContext.PriorityQueue.Schedule(DynamicPriority, ActiveCore, this);
}
}
for (int core = 0; core < KScheduler.CpuCoresCount; core++)
{
if (core != ActiveCore && ((AffinityMask >> core) & 1) != 0)
{
KernelContext.PriorityQueue.Suggest(DynamicPriority, core, this);
}
}
KernelContext.ThreadReselectionRequested = true;
}
private void AdjustSchedulingForNewAffinity(ulong oldAffinityMask, int oldCore)
{
if (SchedFlags != ThreadSchedState.Running || DynamicPriority >= KScheduler.PrioritiesCount || !IsSchedulable)
{
return;
}
// Remove thread from the old priority queues.
for (int core = 0; core < KScheduler.CpuCoresCount; core++)
{
if (((oldAffinityMask >> core) & 1) != 0)
{
if (core == oldCore)
{
KernelContext.PriorityQueue.Unschedule(DynamicPriority, core, this);
}
else
{
KernelContext.PriorityQueue.Unsuggest(DynamicPriority, core, this);
}
}
}
// Add thread to the new priority queues.
for (int core = 0; core < KScheduler.CpuCoresCount; core++)
{
if (((AffinityMask >> core) & 1) != 0)
{
if (core == ActiveCore)
{
KernelContext.PriorityQueue.Schedule(DynamicPriority, core, this);
}
else
{
KernelContext.PriorityQueue.Suggest(DynamicPriority, core, this);
}
}
}
KernelContext.ThreadReselectionRequested = true;
}
public void SetEntryArguments(long argsPtr, int threadHandle)
{
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
Context.SetX(0, (ulong)argsPtr);
Context.SetX(1, (ulong)threadHandle);
}
public void TimeUp()
{
ReleaseAndResume();
}
public string GetGuestStackTrace()
{
return Owner.Debugger.GetGuestStackTrace(this);
}
public string GetGuestRegisterPrintout()
{
return Owner.Debugger.GetCpuRegisterPrintout(this);
}
public void PrintGuestStackTrace()
{
Logger.Info?.Print(LogClass.Cpu, $"Guest stack trace:\n{GetGuestStackTrace()}\n");
}
public void PrintGuestRegisterPrintout()
{
Logger.Info?.Print(LogClass.Cpu, $"Guest CPU registers:\n{GetGuestRegisterPrintout()}\n");
}
public void AddCpuTime(long ticks)
{
Interlocked.Add(ref _totalTimeRunning, ticks);
}
public void StartHostThread()
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
{
if (_schedulerWaitEvent == null)
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
{
var schedulerWaitEvent = new ManualResetEvent(false);
if (Interlocked.Exchange(ref _schedulerWaitEvent, schedulerWaitEvent) == null)
{
HostThread.Start();
}
else
{
schedulerWaitEvent.Dispose();
}
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
}
}
private void ThreadStart()
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
{
_schedulerWaitEvent.WaitOne();
KernelStatic.SetKernelContext(KernelContext, this);
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
if (_customThreadStart != null)
{
_customThreadStart();
}
else
{
Owner.Context.Execute(Context, _entrypoint);
}
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
Context.Dispose();
_schedulerWaitEvent.Dispose();
}
public void MakeUnschedulable()
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
{
_forcedUnschedulable = true;
Add a new JIT compiler for CPU code (#693) * Start of the ARMeilleure project * Refactoring around the old IRAdapter, now renamed to PreAllocator * Optimize the LowestBitSet method * Add CLZ support and fix CLS implementation * Add missing Equals and GetHashCode overrides on some structs, misc small tweaks * Implement the ByteSwap IR instruction, and some refactoring on the assembler * Implement the DivideUI IR instruction and fix 64-bits IDIV * Correct constant operand type on CSINC * Move division instructions implementation to InstEmitDiv * Fix destination type for the ConditionalSelect IR instruction * Implement UMULH and SMULH, with new IR instructions * Fix some issues with shift instructions * Fix constant types for BFM instructions * Fix up new tests using the new V128 struct * Update tests * Move DIV tests to a separate file * Add support for calls, and some instructions that depends on them * Start adding support for SIMD & FP types, along with some of the related ARM instructions * Fix some typos and the divide instruction with FP operands * Fix wrong method call on Clz_V * Implement ARM FP & SIMD move instructions, Saddlv_V, and misc. fixes * Implement SIMD logical instructions and more misc. fixes * Fix PSRAD x86 instruction encoding, TRN, UABD and UABDL implementations * Implement float conversion instruction, merge in LDj3SNuD fixes, and some other misc. fixes * Implement SIMD shift instruction and fix Dup_V * Add SCVTF and UCVTF (vector, fixed-point) variants to the opcode table * Fix check with tolerance on tester * Implement FP & SIMD comparison instructions, and some fixes * Update FCVT (Scalar) encoding on the table to support the Half-float variants * Support passing V128 structs, some cleanup on the register allocator, merge LDj3SNuD fixes * Use old memory access methods, made a start on SIMD memory insts support, some fixes * Fix float constant passed to functions, save and restore non-volatile XMM registers, other fixes * Fix arguments count with struct return values, other fixes * More instructions * Misc. fixes and integrate LDj3SNuD fixes * Update tests * Add a faster linear scan allocator, unwinding support on windows, and other changes * Update Ryujinx.HLE * Update Ryujinx.Graphics * Fix V128 return pointer passing, RCX is clobbered * Update Ryujinx.Tests * Update ITimeZoneService * Stop using GetFunctionPointer as that can't be called from native code, misc. fixes and tweaks * Use generic GetFunctionPointerForDelegate method and other tweaks * Some refactoring on the code generator, assert on invalid operations and use a separate enum for intrinsics * Remove some unused code on the assembler * Fix REX.W prefix regression on float conversion instructions, add some sort of profiler * Add hardware capability detection * Fix regression on Sha1h and revert Fcm** changes * Add SSE2-only paths on vector extract and insert, some refactoring on the pre-allocator * Fix silly mistake introduced on last commit on CpuId * Generate inline stack probes when the stack allocation is too large * Initial support for the System-V ABI * Support multiple destination operands * Fix SSE2 VectorInsert8 path, and other fixes * Change placement of XMM callee save and restore code to match other compilers * Rename Dest to Destination and Inst to Instruction * Fix a regression related to calls and the V128 type * Add an extra space on comments to match code style * Some refactoring * Fix vector insert FP32 SSE2 path * Port over the ARM32 instructions * Avoid memory protection races on JIT Cache * Another fix on VectorInsert FP32 (thanks to LDj3SNuD * Float operands don't need to use the same register when VEX is supported * Add a new register allocator, higher quality code for hot code (tier up), and other tweaks * Some nits, small improvements on the pre allocator * CpuThreadState is gone * Allow changing CPU emulators with a config entry * Add runtime identifiers on the ARMeilleure project * Allow switching between CPUs through a config entry (pt. 2) * Change win10-x64 to win-x64 on projects * Update the Ryujinx project to use ARMeilleure * Ensure that the selected register is valid on the hybrid allocator * Allow exiting on returns to 0 (should fix test regression) * Remove register assignments for most used variables on the hybrid allocator * Do not use fixed registers as spill temp * Add missing namespace and remove unneeded using * Address PR feedback * Fix types, etc * Enable AssumeStrictAbiCompliance by default * Ensure that Spill and Fill don't load or store any more than necessary
2019-08-08 18:56:22 +00:00
}
public override bool IsSignaled()
{
return _hasExited != 0;
}
protected override void Destroy()
{
if (_hasBeenInitialized)
{
FreeResources();
bool released = Owner != null || _hasBeenReleased;
if (Owner != null)
{
Owner.ResourceLimit?.Release(LimitableResource.Thread, 1, released ? 0 : 1);
Owner.DecrementReferenceCount();
}
else
{
KernelContext.ResourceLimit.Release(LimitableResource.Thread, 1, released ? 0 : 1);
}
}
}
private void FreeResources()
{
Owner?.RemoveThread(this);
if (_tlsAddress != 0 && Owner.FreeThreadLocalStorage(_tlsAddress) != KernelResult.Success)
{
throw new InvalidOperationException("Unexpected failure freeing thread local storage.");
}
KernelContext.CriticalSection.Enter();
// Wake up all threads that may be waiting for a mutex being held by this thread.
foreach (KThread thread in _mutexWaiters)
{
thread.MutexOwner = null;
thread._originalPreferredCore = 0;
thread.ObjSyncResult = KernelResult.InvalidState;
thread.ReleaseAndResume();
}
KernelContext.CriticalSection.Leave();
Owner?.DecrementThreadCountAndTerminateIfZero();
2018-02-04 23:08:20 +00:00
}
public void Pin()
{
IsPinned = true;
_coreMigrationDisableCount++;
int activeCore = ActiveCore;
_originalPreferredCore = PreferredCore;
_originalAffinityMask = AffinityMask;
ActiveCore = CurrentCore;
PreferredCore = CurrentCore;
AffinityMask = 1UL << CurrentCore;
if (activeCore != CurrentCore || _originalAffinityMask != AffinityMask)
{
AdjustSchedulingForNewAffinity(_originalAffinityMask, activeCore);
}
_originalBasePriority = BasePriority;
BasePriority = Math.Min(_originalBasePriority, BitOperations.TrailingZeroCount(Owner.Capabilities.AllowedThreadPriosMask) - 1);
UpdatePriorityInheritance();
// Disallows thread pausing
_forcePausePermissionFlags &= ~ThreadSchedState.ThreadPauseFlag;
CombineForcePauseFlags();
// TODO: Assign reduced SVC permissions
}
public void Unpin()
{
IsPinned = false;
_coreMigrationDisableCount--;
ulong affinityMask = AffinityMask;
int activeCore = ActiveCore;
PreferredCore = _originalPreferredCore;
AffinityMask = _originalAffinityMask;
if (AffinityMask != affinityMask)
{
if ((AffinityMask & 1UL << ActiveCore) != 0)
{
if (PreferredCore >= 0)
{
ActiveCore = PreferredCore;
}
else
{
ActiveCore = sizeof(ulong) * 8 - 1 - BitOperations.LeadingZeroCount((ulong)AffinityMask);
}
AdjustSchedulingForNewAffinity(affinityMask, activeCore);
}
}
BasePriority = _originalBasePriority;
UpdatePriorityInheritance();
if (!TerminationRequested)
{
// Allows thread pausing
_forcePausePermissionFlags |= ThreadSchedState.ThreadPauseFlag;
CombineForcePauseFlags();
// TODO: Restore SVC permissions
}
// Wake up waiters
foreach (KThread waiter in _pinnedWaiters)
{
waiter.ReleaseAndResume();
}
_pinnedWaiters.Clear();
}
public void SynchronizePreemptionState()
{
KernelContext.CriticalSection.Enter();
if (Owner != null && Owner.PinnedThreads[CurrentCore] == this)
{
ClearUserInterruptFlag();
Owner.UnpinThread(this);
}
KernelContext.CriticalSection.Leave();
}
public ushort GetUserDisableCount()
{
return Owner.CpuMemory.Read<ushort>(_tlsAddress + TlsUserDisableCountOffset);
}
public void SetUserInterruptFlag()
{
Owner.CpuMemory.Write<ushort>(_tlsAddress + TlsUserInterruptFlagOffset, 1);
}
public void ClearUserInterruptFlag()
{
Owner.CpuMemory.Write<ushort>(_tlsAddress + TlsUserInterruptFlagOffset, 0);
}
2018-02-04 23:08:20 +00:00
}
}