Optimize address translation and write tracking on the MMU (#571)

* Implement faster address translation and write tracking on the MMU

* Rename MemoryAlloc to MemoryManagement, and other nits

* Support multi-level page tables

* Fix typo

* Reword comment a bit

* Support scalar vector loads/stores on the memory fast path, and minor fixes

* Add missing cast

* Alignment

* Fix VirtualFree function signature

* Change MemoryProtection enum to uint aswell for consistency

6 files changed, 88 insertions, 93 deletions

diff --git a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcHandler.cs b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcHandler.cs
index 071b3c20..cf881a79 100644
--- a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcHandler.cs
+++ b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcHandler.cs
@@ -1,5 +1,4 @@
 using ChocolArm64.Events;
-using ChocolArm64.Memory;
 using ChocolArm64.State;
 using Ryujinx.HLE.HOS.Kernel.Process;
 using System;
@@ -11,14 +10,12 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
         private Switch        _device;
         private KProcess      _process;
         private Horizon       _system;
-        private MemoryManager _memory;
 
         public SvcHandler(Switch device, KProcess process)
         {
             _device  = device;
             _process = process;
             _system  = device.System;
-            _memory  = process.CpuMemory;
         }
 
         public void SvcCall(object sender, InstExceptionEventArgs e)
diff --git a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcIpc.cs b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcIpc.cs
index 54939418..e19d9d26 100644
--- a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcIpc.cs
+++ b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcIpc.cs
@@ -93,7 +93,7 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
 
         private KernelResult SendSyncRequest(ulong messagePtr, ulong size, int handle)
         {
-            byte[] messageData = _memory.ReadBytes((long)messagePtr, (long)size);
+            byte[] messageData = _process.CpuMemory.ReadBytes((long)messagePtr, (long)size);
 
             KClientSession clientSession = _process.HandleTable.GetObject<KClientSession>(handle);
 
@@ -142,7 +142,7 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
             ipcMessage.Thread.ObjSyncResult = IpcHandler.IpcCall(
                 _device,
                 _process,
-                _memory,
+                _process.CpuMemory,
                 ipcMessage.Session,
                 ipcMessage.Message,
                 ipcMessage.MessagePtr);
diff --git a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcMemory.cs b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcMemory.cs
index 6f8180c5..f794d130 100644
--- a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcMemory.cs
+++ b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcMemory.cs
@@ -62,11 +62,6 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
                 attributeMask,
                 attributeValue);
 
-            if (result == KernelResult.Success)
-            {
-                _memory.StopObservingRegion((long)position, (long)size);
-            }
-
             return result;
         }
 
@@ -157,14 +152,14 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
         {
             KMemoryInfo blkInfo = _process.MemoryManager.QueryMemory(position);
 
-            _memory.WriteUInt64((long)infoPtr + 0x00, blkInfo.Address);
-            _memory.WriteUInt64((long)infoPtr + 0x08, blkInfo.Size);
-            _memory.WriteInt32 ((long)infoPtr + 0x10, (int)blkInfo.State & 0xff);
-            _memory.WriteInt32 ((long)infoPtr + 0x14, (int)blkInfo.Attribute);
-            _memory.WriteInt32 ((long)infoPtr + 0x18, (int)blkInfo.Permission);
-            _memory.WriteInt32 ((long)infoPtr + 0x1c, blkInfo.IpcRefCount);
-            _memory.WriteInt32 ((long)infoPtr + 0x20, blkInfo.DeviceRefCount);
-            _memory.WriteInt32 ((long)infoPtr + 0x24, 0);
+            _process.CpuMemory.WriteUInt64((long)infoPtr + 0x00, blkInfo.Address);
+            _process.CpuMemory.WriteUInt64((long)infoPtr + 0x08, blkInfo.Size);
+            _process.CpuMemory.WriteInt32 ((long)infoPtr + 0x10, (int)blkInfo.State & 0xff);
+            _process.CpuMemory.WriteInt32 ((long)infoPtr + 0x14, (int)blkInfo.Attribute);
+            _process.CpuMemory.WriteInt32 ((long)infoPtr + 0x18, (int)blkInfo.Permission);
+            _process.CpuMemory.WriteInt32 ((long)infoPtr + 0x1c, blkInfo.IpcRefCount);
+            _process.CpuMemory.WriteInt32 ((long)infoPtr + 0x20, blkInfo.DeviceRefCount);
+            _process.CpuMemory.WriteInt32 ((long)infoPtr + 0x24, 0);
 
             return KernelResult.Success;
         }
diff --git a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcSystem.cs b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcSystem.cs
index be136ff0..efc10512 100644
--- a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcSystem.cs
+++ b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcSystem.cs
@@ -201,7 +201,7 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
 
         private void OutputDebugString(ulong strPtr, ulong size)
         {
-            string str = MemoryHelper.ReadAsciiString(_memory, (long)strPtr, (long)size);
+            string str = MemoryHelper.ReadAsciiString(_process.CpuMemory, (long)strPtr, (long)size);
 
             Logger.PrintWarning(LogClass.KernelSvc, str);
         }
diff --git a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThread.cs b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThread.cs
index 64268ff2..fa0b3a6c 100644
--- a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThread.cs
+++ b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThread.cs
@@ -1,3 +1,4 @@
+using ChocolArm64.Memory;
 using Ryujinx.HLE.HOS.Kernel.Common;
 using Ryujinx.HLE.HOS.Kernel.Process;
 using Ryujinx.HLE.HOS.Kernel.Threading;
@@ -346,79 +347,81 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
                 return KernelResult.InvalidThread;
             }
 
-            _memory.WriteUInt64((long)address + 0x0,  thread.Context.ThreadState.X0);
-            _memory.WriteUInt64((long)address + 0x8,  thread.Context.ThreadState.X1);
-            _memory.WriteUInt64((long)address + 0x10, thread.Context.ThreadState.X2);
-            _memory.WriteUInt64((long)address + 0x18, thread.Context.ThreadState.X3);
-            _memory.WriteUInt64((long)address + 0x20, thread.Context.ThreadState.X4);
-            _memory.WriteUInt64((long)address + 0x28, thread.Context.ThreadState.X5);
-            _memory.WriteUInt64((long)address + 0x30, thread.Context.ThreadState.X6);
-            _memory.WriteUInt64((long)address + 0x38, thread.Context.ThreadState.X7);
-            _memory.WriteUInt64((long)address + 0x40, thread.Context.ThreadState.X8);
-            _memory.WriteUInt64((long)address + 0x48, thread.Context.ThreadState.X9);
-            _memory.WriteUInt64((long)address + 0x50, thread.Context.ThreadState.X10);
-            _memory.WriteUInt64((long)address + 0x58, thread.Context.ThreadState.X11);
-            _memory.WriteUInt64((long)address + 0x60, thread.Context.ThreadState.X12);
-            _memory.WriteUInt64((long)address + 0x68, thread.Context.ThreadState.X13);
-            _memory.WriteUInt64((long)address + 0x70, thread.Context.ThreadState.X14);
-            _memory.WriteUInt64((long)address + 0x78, thread.Context.ThreadState.X15);
-            _memory.WriteUInt64((long)address + 0x80, thread.Context.ThreadState.X16);
-            _memory.WriteUInt64((long)address + 0x88, thread.Context.ThreadState.X17);
-            _memory.WriteUInt64((long)address + 0x90, thread.Context.ThreadState.X18);
-            _memory.WriteUInt64((long)address + 0x98, thread.Context.ThreadState.X19);
-            _memory.WriteUInt64((long)address + 0xa0, thread.Context.ThreadState.X20);
-            _memory.WriteUInt64((long)address + 0xa8, thread.Context.ThreadState.X21);
-            _memory.WriteUInt64((long)address + 0xb0, thread.Context.ThreadState.X22);
-            _memory.WriteUInt64((long)address + 0xb8, thread.Context.ThreadState.X23);
-            _memory.WriteUInt64((long)address + 0xc0, thread.Context.ThreadState.X24);
-            _memory.WriteUInt64((long)address + 0xc8, thread.Context.ThreadState.X25);
-            _memory.WriteUInt64((long)address + 0xd0, thread.Context.ThreadState.X26);
-            _memory.WriteUInt64((long)address + 0xd8, thread.Context.ThreadState.X27);
-            _memory.WriteUInt64((long)address + 0xe0, thread.Context.ThreadState.X28);
-            _memory.WriteUInt64((long)address + 0xe8, thread.Context.ThreadState.X29);
-            _memory.WriteUInt64((long)address + 0xf0, thread.Context.ThreadState.X30);
-            _memory.WriteUInt64((long)address + 0xf8, thread.Context.ThreadState.X31);
-
-            _memory.WriteInt64((long)address + 0x100, thread.LastPc);
-
-            _memory.WriteUInt64((long)address + 0x108, (ulong)thread.Context.ThreadState.Psr);
-
-            _memory.WriteVector128((long)address + 0x110, thread.Context.ThreadState.V0);
-            _memory.WriteVector128((long)address + 0x120, thread.Context.ThreadState.V1);
-            _memory.WriteVector128((long)address + 0x130, thread.Context.ThreadState.V2);
-            _memory.WriteVector128((long)address + 0x140, thread.Context.ThreadState.V3);
-            _memory.WriteVector128((long)address + 0x150, thread.Context.ThreadState.V4);
-            _memory.WriteVector128((long)address + 0x160, thread.Context.ThreadState.V5);
-            _memory.WriteVector128((long)address + 0x170, thread.Context.ThreadState.V6);
-            _memory.WriteVector128((long)address + 0x180, thread.Context.ThreadState.V7);
-            _memory.WriteVector128((long)address + 0x190, thread.Context.ThreadState.V8);
-            _memory.WriteVector128((long)address + 0x1a0, thread.Context.ThreadState.V9);
-            _memory.WriteVector128((long)address + 0x1b0, thread.Context.ThreadState.V10);
-            _memory.WriteVector128((long)address + 0x1c0, thread.Context.ThreadState.V11);
-            _memory.WriteVector128((long)address + 0x1d0, thread.Context.ThreadState.V12);
-            _memory.WriteVector128((long)address + 0x1e0, thread.Context.ThreadState.V13);
-            _memory.WriteVector128((long)address + 0x1f0, thread.Context.ThreadState.V14);
-            _memory.WriteVector128((long)address + 0x200, thread.Context.ThreadState.V15);
-            _memory.WriteVector128((long)address + 0x210, thread.Context.ThreadState.V16);
-            _memory.WriteVector128((long)address + 0x220, thread.Context.ThreadState.V17);
-            _memory.WriteVector128((long)address + 0x230, thread.Context.ThreadState.V18);
-            _memory.WriteVector128((long)address + 0x240, thread.Context.ThreadState.V19);
-            _memory.WriteVector128((long)address + 0x250, thread.Context.ThreadState.V20);
-            _memory.WriteVector128((long)address + 0x260, thread.Context.ThreadState.V21);
-            _memory.WriteVector128((long)address + 0x270, thread.Context.ThreadState.V22);
-            _memory.WriteVector128((long)address + 0x280, thread.Context.ThreadState.V23);
-            _memory.WriteVector128((long)address + 0x290, thread.Context.ThreadState.V24);
-            _memory.WriteVector128((long)address + 0x2a0, thread.Context.ThreadState.V25);
-            _memory.WriteVector128((long)address + 0x2b0, thread.Context.ThreadState.V26);
-            _memory.WriteVector128((long)address + 0x2c0, thread.Context.ThreadState.V27);
-            _memory.WriteVector128((long)address + 0x2d0, thread.Context.ThreadState.V28);
-            _memory.WriteVector128((long)address + 0x2e0, thread.Context.ThreadState.V29);
-            _memory.WriteVector128((long)address + 0x2f0, thread.Context.ThreadState.V30);
-            _memory.WriteVector128((long)address + 0x300, thread.Context.ThreadState.V31);
-
-            _memory.WriteInt32((long)address + 0x310, thread.Context.ThreadState.Fpcr);
-            _memory.WriteInt32((long)address + 0x314, thread.Context.ThreadState.Fpsr);
-            _memory.WriteInt64((long)address + 0x318, thread.Context.ThreadState.Tpidr);
+            MemoryManager memory = currentProcess.CpuMemory;
+
+            memory.WriteUInt64((long)address + 0x0,  thread.Context.ThreadState.X0);
+            memory.WriteUInt64((long)address + 0x8,  thread.Context.ThreadState.X1);
+            memory.WriteUInt64((long)address + 0x10, thread.Context.ThreadState.X2);
+            memory.WriteUInt64((long)address + 0x18, thread.Context.ThreadState.X3);
+            memory.WriteUInt64((long)address + 0x20, thread.Context.ThreadState.X4);
+            memory.WriteUInt64((long)address + 0x28, thread.Context.ThreadState.X5);
+            memory.WriteUInt64((long)address + 0x30, thread.Context.ThreadState.X6);
+            memory.WriteUInt64((long)address + 0x38, thread.Context.ThreadState.X7);
+            memory.WriteUInt64((long)address + 0x40, thread.Context.ThreadState.X8);
+            memory.WriteUInt64((long)address + 0x48, thread.Context.ThreadState.X9);
+            memory.WriteUInt64((long)address + 0x50, thread.Context.ThreadState.X10);
+            memory.WriteUInt64((long)address + 0x58, thread.Context.ThreadState.X11);
+            memory.WriteUInt64((long)address + 0x60, thread.Context.ThreadState.X12);
+            memory.WriteUInt64((long)address + 0x68, thread.Context.ThreadState.X13);
+            memory.WriteUInt64((long)address + 0x70, thread.Context.ThreadState.X14);
+            memory.WriteUInt64((long)address + 0x78, thread.Context.ThreadState.X15);
+            memory.WriteUInt64((long)address + 0x80, thread.Context.ThreadState.X16);
+            memory.WriteUInt64((long)address + 0x88, thread.Context.ThreadState.X17);
+            memory.WriteUInt64((long)address + 0x90, thread.Context.ThreadState.X18);
+            memory.WriteUInt64((long)address + 0x98, thread.Context.ThreadState.X19);
+            memory.WriteUInt64((long)address + 0xa0, thread.Context.ThreadState.X20);
+            memory.WriteUInt64((long)address + 0xa8, thread.Context.ThreadState.X21);
+            memory.WriteUInt64((long)address + 0xb0, thread.Context.ThreadState.X22);
+            memory.WriteUInt64((long)address + 0xb8, thread.Context.ThreadState.X23);
+            memory.WriteUInt64((long)address + 0xc0, thread.Context.ThreadState.X24);
+            memory.WriteUInt64((long)address + 0xc8, thread.Context.ThreadState.X25);
+            memory.WriteUInt64((long)address + 0xd0, thread.Context.ThreadState.X26);
+            memory.WriteUInt64((long)address + 0xd8, thread.Context.ThreadState.X27);
+            memory.WriteUInt64((long)address + 0xe0, thread.Context.ThreadState.X28);
+            memory.WriteUInt64((long)address + 0xe8, thread.Context.ThreadState.X29);
+            memory.WriteUInt64((long)address + 0xf0, thread.Context.ThreadState.X30);
+            memory.WriteUInt64((long)address + 0xf8, thread.Context.ThreadState.X31);
+
+            memory.WriteInt64((long)address + 0x100, thread.LastPc);
+
+            memory.WriteUInt64((long)address + 0x108, (ulong)thread.Context.ThreadState.Psr);
+
+            memory.WriteVector128((long)address + 0x110, thread.Context.ThreadState.V0);
+            memory.WriteVector128((long)address + 0x120, thread.Context.ThreadState.V1);
+            memory.WriteVector128((long)address + 0x130, thread.Context.ThreadState.V2);
+            memory.WriteVector128((long)address + 0x140, thread.Context.ThreadState.V3);
+            memory.WriteVector128((long)address + 0x150, thread.Context.ThreadState.V4);
+            memory.WriteVector128((long)address + 0x160, thread.Context.ThreadState.V5);
+            memory.WriteVector128((long)address + 0x170, thread.Context.ThreadState.V6);
+            memory.WriteVector128((long)address + 0x180, thread.Context.ThreadState.V7);
+            memory.WriteVector128((long)address + 0x190, thread.Context.ThreadState.V8);
+            memory.WriteVector128((long)address + 0x1a0, thread.Context.ThreadState.V9);
+            memory.WriteVector128((long)address + 0x1b0, thread.Context.ThreadState.V10);
+            memory.WriteVector128((long)address + 0x1c0, thread.Context.ThreadState.V11);
+            memory.WriteVector128((long)address + 0x1d0, thread.Context.ThreadState.V12);
+            memory.WriteVector128((long)address + 0x1e0, thread.Context.ThreadState.V13);
+            memory.WriteVector128((long)address + 0x1f0, thread.Context.ThreadState.V14);
+            memory.WriteVector128((long)address + 0x200, thread.Context.ThreadState.V15);
+            memory.WriteVector128((long)address + 0x210, thread.Context.ThreadState.V16);
+            memory.WriteVector128((long)address + 0x220, thread.Context.ThreadState.V17);
+            memory.WriteVector128((long)address + 0x230, thread.Context.ThreadState.V18);
+            memory.WriteVector128((long)address + 0x240, thread.Context.ThreadState.V19);
+            memory.WriteVector128((long)address + 0x250, thread.Context.ThreadState.V20);
+            memory.WriteVector128((long)address + 0x260, thread.Context.ThreadState.V21);
+            memory.WriteVector128((long)address + 0x270, thread.Context.ThreadState.V22);
+            memory.WriteVector128((long)address + 0x280, thread.Context.ThreadState.V23);
+            memory.WriteVector128((long)address + 0x290, thread.Context.ThreadState.V24);
+            memory.WriteVector128((long)address + 0x2a0, thread.Context.ThreadState.V25);
+            memory.WriteVector128((long)address + 0x2b0, thread.Context.ThreadState.V26);
+            memory.WriteVector128((long)address + 0x2c0, thread.Context.ThreadState.V27);
+            memory.WriteVector128((long)address + 0x2d0, thread.Context.ThreadState.V28);
+            memory.WriteVector128((long)address + 0x2e0, thread.Context.ThreadState.V29);
+            memory.WriteVector128((long)address + 0x2f0, thread.Context.ThreadState.V30);
+            memory.WriteVector128((long)address + 0x300, thread.Context.ThreadState.V31);
+
+            memory.WriteInt32((long)address + 0x310, thread.Context.ThreadState.Fpcr);
+            memory.WriteInt32((long)address + 0x314, thread.Context.ThreadState.Fpsr);
+            memory.WriteInt64((long)address + 0x318, thread.Context.ThreadState.Tpidr);
 
             return KernelResult.Success;
         }
diff --git a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThreadSync.cs b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThreadSync.cs
index ecda9e2d..6e5b4782 100644
--- a/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThreadSync.cs
+++ b/Ryujinx.HLE/HOS/Kernel/SupervisorCall/SvcThreadSync.cs
@@ -25,7 +25,7 @@ namespace Ryujinx.HLE.HOS.Kernel.SupervisorCall
 
             for (int index = 0; index < handlesCount; index++)
             {
-                int handle = _memory.ReadInt32((long)handlesPtr + index * 4);
+                int handle = _process.CpuMemory.ReadInt32((long)handlesPtr + index * 4);
 
                 KSynchronizationObject syncObj = _process.HandleTable.GetObject<KSynchronizationObject>(handle);