Generalize tail continues (#1298)

* Generalize tail continues

* Fix DecodeBasicBlock

`Next` and `Branch` would be null, which is not the state expected by
the branch instructions. They end up branching or falling into a block
which is never populated by the `Translator`. This causes an assert to
be fired when building the CFG.

* Clean up Decode overloads

* Do not synchronize when branching into exit block

If we're branching into an exit block, that exit block will tail
continue into another translation which already has a synchronization.

* Remove A32 predicate tail continue

If `block` is not an exit block then the `block.Next` must exist (as
per the last instruction of `block`).

* Throw if decoded 0 blocks

Address gdkchan's feedback

* Rebuild block list instead of setting to null

Address gdkchan's feedback

3 files changed, 87 insertions, 67 deletions

diff --git a/ARMeilleure/Decoders/Block.cs b/ARMeilleure/Decoders/Block.cs
index d38b5a8e..ef0517c0 100644
--- a/ARMeilleure/Decoders/Block.cs
+++ b/ARMeilleure/Decoders/Block.cs
@@ -12,6 +12,7 @@ namespace ARMeilleure.Decoders
         public Block Branch { get; set; }
 
         public bool TailCall { get; set; }
+        public bool Exit     { get; set; }
 
         public List<OpCode> OpCodes { get; private set; }
 
@@ -29,7 +30,7 @@ namespace ARMeilleure.Decoders
         {
             int splitIndex = BinarySearch(OpCodes, rightBlock.Address);
 
-            if ((ulong)OpCodes[splitIndex].Address < rightBlock.Address)
+            if (OpCodes[splitIndex].Address < rightBlock.Address)
             {
                 splitIndex++;
             }
diff --git a/ARMeilleure/Decoders/Decoder.cs b/ARMeilleure/Decoders/Decoder.cs
index 6117b807..de62d125 100644
--- a/ARMeilleure/Decoders/Decoder.cs
+++ b/ARMeilleure/Decoders/Decoder.cs
@@ -17,16 +17,7 @@ namespace ARMeilleure.Decoders
         // For lower code quality translation, we set a lower limit since we're blocking execution.
         private const int MaxInstsPerFunctionLowCq = 500;
 
-        public static Block[] DecodeBasicBlock(IMemoryManager memory, ulong address, ExecutionMode mode)
-        {
-            Block block = new Block(address);
-
-            FillBlock(memory, mode, block, ulong.MaxValue);
-
-            return new Block[] { block };
-        }
-
-        public static Block[] DecodeFunction(IMemoryManager memory, ulong address, ExecutionMode mode, bool highCq)
+        public static Block[] Decode(IMemoryManager memory, ulong address, ExecutionMode mode, bool highCq, bool singleBlock)
         {
             List<Block> blocks = new List<Block>();
 
@@ -42,13 +33,14 @@ namespace ARMeilleure.Decoders
             {
                 if (!visited.TryGetValue(blkAddress, out Block block))
                 {
-                    if (opsCount > instructionLimit || !memory.IsMapped(blkAddress))
+                    block = new Block(blkAddress);
+
+                    if ((singleBlock && visited.Count >= 1) || opsCount > instructionLimit || !memory.IsMapped(blkAddress))
                     {
-                        return null;
+                        block.Exit = true;
+                        block.EndAddress = blkAddress;
                     }
 
-                    block = new Block(blkAddress);
-
                     workQueue.Enqueue(block);
 
                     visited.Add(blkAddress, block);
@@ -71,6 +63,8 @@ namespace ARMeilleure.Decoders
                         throw new InvalidOperationException("Found duplicate block address on the list.");
                     }
 
+                    currBlock.Exit = false;
+
                     nBlock.Split(currBlock);
 
                     blocks.Insert(nBlkIndex + 1, currBlock);
@@ -78,47 +72,50 @@ namespace ARMeilleure.Decoders
                     continue;
                 }
 
-                // If we have a block after the current one, set the limit address.
-                ulong limitAddress = ulong.MaxValue;
-
-                if (nBlkIndex != blocks.Count)
+                if (!currBlock.Exit)
                 {
-                    Block nBlock = blocks[nBlkIndex];
-
-                    int nextIndex = nBlkIndex + 1;
+                    // If we have a block after the current one, set the limit address.
+                    ulong limitAddress = ulong.MaxValue;
 
-                    if (nBlock.Address < currBlock.Address && nextIndex < blocks.Count)
+                    if (nBlkIndex != blocks.Count)
                     {
-                        limitAddress = blocks[nextIndex].Address;
+                        Block nBlock = blocks[nBlkIndex];
+
+                        int nextIndex = nBlkIndex + 1;
+
+                        if (nBlock.Address < currBlock.Address && nextIndex < blocks.Count)
+                        {
+                            limitAddress = blocks[nextIndex].Address;
+                        }
+                        else if (nBlock.Address > currBlock.Address)
+                        {
+                            limitAddress = blocks[nBlkIndex].Address;
+                        }
                     }
-                    else if (nBlock.Address > currBlock.Address)
-                    {
-                        limitAddress = blocks[nBlkIndex].Address;
-                    }
-                }
 
-                FillBlock(memory, mode, currBlock, limitAddress);
+                    FillBlock(memory, mode, currBlock, limitAddress);
 
-                opsCount += currBlock.OpCodes.Count;
-
-                if (currBlock.OpCodes.Count != 0)
-                {
-                    // Set child blocks. "Branch" is the block the branch instruction
-                    // points to (when taken), "Next" is the block at the next address,
-                    // executed when the branch is not taken. For Unconditional Branches
-                    // (except BL/BLR that are sub calls) or end of executable, Next is null.
-                    OpCode lastOp = currBlock.GetLastOp();
+                    opsCount += currBlock.OpCodes.Count;
 
-                    bool isCall = IsCall(lastOp);
-
-                    if (lastOp is IOpCodeBImm op && !isCall)
+                    if (currBlock.OpCodes.Count != 0)
                     {
-                        currBlock.Branch = GetBlock((ulong)op.Immediate);
-                    }
-
-                    if (!IsUnconditionalBranch(lastOp) || isCall)
-                    {
-                        currBlock.Next = GetBlock(currBlock.EndAddress);
+                        // Set child blocks. "Branch" is the block the branch instruction
+                        // points to (when taken), "Next" is the block at the next address,
+                        // executed when the branch is not taken. For Unconditional Branches
+                        // (except BL/BLR that are sub calls) or end of executable, Next is null.
+                        OpCode lastOp = currBlock.GetLastOp();
+
+                        bool isCall = IsCall(lastOp);
+
+                        if (lastOp is IOpCodeBImm op && !isCall)
+                        {
+                            currBlock.Branch = GetBlock((ulong)op.Immediate);
+                        }
+
+                        if (!IsUnconditionalBranch(lastOp) || isCall)
+                        {
+                            currBlock.Next = GetBlock(currBlock.EndAddress);
+                        }
                     }
                 }
 
@@ -135,7 +132,15 @@ namespace ARMeilleure.Decoders
                 }
             }
 
-            TailCallRemover.RunPass(address, blocks);
+            if (blocks.Count == 0)
+            {
+                throw new InvalidOperationException($"Decoded 0 blocks. Entry point = 0x{address:X}.");
+            }
+
+            if (!singleBlock)
+            {
+                return TailCallRemover.RunPass(address, blocks);
+            }
 
             return blocks.ToArray();
         }
diff --git a/ARMeilleure/Decoders/Optimizations/TailCallRemover.cs b/ARMeilleure/Decoders/Optimizations/TailCallRemover.cs
index 2d6439ba..e64f9a54 100644
--- a/ARMeilleure/Decoders/Optimizations/TailCallRemover.cs
+++ b/ARMeilleure/Decoders/Optimizations/TailCallRemover.cs
@@ -1,16 +1,15 @@
-using ARMeilleure.Decoders;
-using System;
+using System;
 using System.Collections.Generic;
 
 namespace ARMeilleure.Decoders.Optimizations
 {
     static class TailCallRemover
     {
-        public static void RunPass(ulong entryAddress, List<Block> blocks)
+        public static Block[] RunPass(ulong entryAddress, List<Block> blocks)
         {
             // Detect tail calls:
             // - Assume this function spans the space covered by contiguous code blocks surrounding the entry address.
-            // - Unconditional jump to an area outside this contiguous region will be treated as a tail call.
+            // - A jump to an area outside this contiguous region will be treated as an exit block.
             // - Include a small allowance for jumps outside the contiguous range.
 
             if (!Decoder.BinarySearch(blocks, entryAddress, out int entryBlockId))
@@ -19,57 +18,72 @@ namespace ARMeilleure.Decoders.Optimizations
             }
 
             const ulong allowance = 4;
+
             Block entryBlock = blocks[entryBlockId];
-            int startBlockIndex = entryBlockId;
+
             Block startBlock = entryBlock;
-            int endBlockIndex = entryBlockId;
-            Block endBlock = entryBlock;
+            Block endBlock   = entryBlock;
+
+            int startBlockIndex = entryBlockId;
+            int endBlockIndex   = entryBlockId;
 
             for (int i = entryBlockId + 1; i < blocks.Count; i++) // Search forwards.
             {
                 Block block = blocks[i];
+
                 if (endBlock.EndAddress < block.Address - allowance)
                 {
                     break; // End of contiguous function.
                 }
 
-                endBlock = block;
+                endBlock      = block;
                 endBlockIndex = i;
             }
 
             for (int i = entryBlockId - 1; i >= 0; i--) // Search backwards.
             {
                 Block block = blocks[i];
+
                 if (startBlock.Address > block.EndAddress + allowance)
                 {
                     break; // End of contiguous function.
                 }
 
-                startBlock = block;
+                startBlock      = block;
                 startBlockIndex = i;
             }
 
             if (startBlockIndex == 0 && endBlockIndex == blocks.Count - 1)
             {
-                return; // Nothing to do here.
+                return blocks.ToArray(); // Nothing to do here.
             }
-
-            // Replace all branches to blocks outside the range with null, and force a tail call.
-
+            
+            // Mark branches outside of contiguous region as exit blocks.
             for (int i = startBlockIndex; i <= endBlockIndex; i++)
             {
                 Block block = blocks[i];
+
                 if (block.Branch != null && (block.Branch.Address > endBlock.EndAddress || block.Branch.EndAddress < startBlock.Address))
                 {
-                    block.Branch = null;
-                    block.TailCall = true;
+                    block.Branch.Exit     = true;
+                    block.Branch.TailCall = true;
                 }
             }
 
-            // Finally, delete all blocks outside the contiguous range.
+           var newBlocks = new List<Block>(blocks.Count);
+
+            // Finally, rebuild decoded block list, ignoring blocks outside the contiguous range.
+            for (int i = 0; i < blocks.Count; i++)
+            {
+                Block block = blocks[i];
+
+                if (block.Exit || (i >= startBlockIndex && i <= endBlockIndex))
+                {
+                    newBlocks.Add(block);
+                }
+            }
 
-            blocks.RemoveRange(endBlockIndex + 1, (blocks.Count - endBlockIndex) - 1);
-            blocks.RemoveRange(0, startBlockIndex);
+            return newBlocks.ToArray();
         }
     }
 }