JIT Sparse Function Table (#250)

More up to date build of the JIT Sparse PR for continued development.
JIT Sparse Function Table was originally developed by riperiperi for the
original Ryujinx project, and decreased the amount of layers in the
Function Table structure, to decrease lookup times at the cost of
slightly higher RAM usage.
This PR rebalances the JIT Sparse Function Table to be a bit more RAM
intensive, but faster in workloads where the JIT Function Table is a
bottleneck. Faster RAM will see a bigger impact and slower RAM (DDR3 and
potentially slow DDR4) will see a slight performance decrease.
This PR also implements a base for a PPTC profile system that could
allow for PPTC with ExeFS mods enabled in the future.
This PR also potentially fixes a strange issue where Avalonia would time
out in some rare instances, e.g. when running ExeFS mods with TotK and a
strange controller configuration.

---------

Co-authored-by: Evan Husted <gr33m11@gmail.com>

src/Ryujinx.Cpu/LightningJit/Arm32/Target/Arm64/InstEmitFlow.cs

@@ -140,6 +140,10 @@ namespace Ryujinx.Cpu.LightningJit.Arm32.Target.Arm64
             bool isTail = false)
         {
             int tempRegister;
+            int tempGuestAddress = -1;
+
+            bool inlineLookup = guestAddress.Kind != OperandKind.Constant && 
+                                funcTable is { Sparse: true };
 
             if (guestAddress.Kind == OperandKind.Constant)
             {
@@ -153,9 +157,16 @@ namespace Ryujinx.Cpu.LightningJit.Arm32.Target.Arm64
             else
             {
                 asm.StrRiUn(guestAddress, Register(regAlloc.FixedContextRegister), NativeContextOffsets.DispatchAddressOffset);
+
+                if (inlineLookup && guestAddress.Value == 0)
+                {
+                    // X0 will be overwritten. Move the address to a temp register.
+                    tempGuestAddress = regAlloc.AllocateTempGprRegister();
+                    asm.Mov(Register(tempGuestAddress), guestAddress);
+                }
             }
 
-            tempRegister = regAlloc.FixedContextRegister == 1 ? 2 : 1;
+            tempRegister = NextFreeRegister(1, tempGuestAddress);
 
             if (!isTail)
             {
@@ -176,6 +187,40 @@ namespace Ryujinx.Cpu.LightningJit.Arm32.Target.Arm64
                 asm.Mov(rn, funcPtrLoc & ~0xfffUL);
                 asm.LdrRiUn(rn, rn, (int)(funcPtrLoc & 0xfffUL));
             }
+            else if (inlineLookup)
+            {
+                // Inline table lookup. Only enabled when the sparse function table is enabled with 2 levels.
+
+                Operand indexReg = Register(NextFreeRegister(tempRegister + 1, tempGuestAddress));
+
+                if (tempGuestAddress != -1)
+                {
+                    guestAddress = Register(tempGuestAddress);
+                }
+
+                ulong tableBase = (ulong)funcTable.Base;
+
+                // Index into the table.
+                asm.Mov(rn, tableBase);
+
+                for (int i = 0; i < funcTable.Levels.Length; i++)
+                {
+                    var level = funcTable.Levels[i];
+                    asm.Ubfx(indexReg, guestAddress, level.Index, level.Length);
+                    asm.Lsl(indexReg, indexReg, Const(3));
+
+                    // Index into the page.
+                    asm.Add(rn, rn, indexReg);
+
+                    // Load the page address.
+                    asm.LdrRiUn(rn, rn, 0);
+                }
+
+                if (tempGuestAddress != -1)
+                {
+                    regAlloc.FreeTempGprRegister(tempGuestAddress);
+                }
+            }
             else
             {
                 asm.Mov(rn, (ulong)funcPtr);
@@ -252,5 +297,20 @@ namespace Ryujinx.Cpu.LightningJit.Arm32.Target.Arm64
         {
             return new Operand(register, RegisterType.Integer, type);
         }
+
+        private static Operand Const(long value, OperandType type = OperandType.I64)
+        {
+            return new Operand(type, (ulong)value);
+        }
+
+        private static int NextFreeRegister(int start, int avoid)
+        {
+            if (start == avoid)
+            {
+                start++;
+            }
+
+            return start;
+        }
     }
 }