Adding better register tracking through locals and fixing branches.

LLVM does an amazing job of optimizing this. There are many opportunities
to make it better, too, by preventing spilling where not required or only
spilling/filling things when needed.

include/xenia/cpu/codegen/function_generator.h

@@ -44,16 +44,23 @@ public:
   void GenerateBasicBlocks();
   llvm::BasicBlock* GetBasicBlock(uint32_t address);
   llvm::BasicBlock* GetNextBasicBlock();
+  llvm::BasicBlock* GetReturnBasicBlock();
 
   llvm::Function* GetFunction(sdb::FunctionSymbol* fn);
 
+  int GenerateIndirectionBranch(uint32_t cia, llvm::Value* target,
+                                bool lk, bool likely_local);
+
   llvm::Value* LoadStateValue(uint32_t offset, llvm::Type* type,
                               const char* name = "");
   void StoreStateValue(uint32_t offset, llvm::Type* type, llvm::Value* value);
 
   llvm::Value* cia_value();
 
-  void FlushRegisters();
+  llvm::Value* SetupRegisterLocal(uint32_t offset, llvm::Type* type,
+                                  const char* name);
+  void FillRegisters();
+  void SpillRegisters();
 
   llvm::Value* xer_value();
   void update_xer_value(llvm::Value* value);
@@ -74,6 +81,7 @@ public:
   void WriteMemory(llvm::Value* addr, uint32_t size, llvm::Value* value);
 
 private:
+  void GenerateSharedBlocks();
   void GenerateBasicBlock(sdb::FunctionBlock* block, llvm::BasicBlock* bb);
 
   xe_memory_ref         memory_;
@@ -83,6 +91,9 @@ private:
   llvm::Module*         gen_module_;
   llvm::Function*       gen_fn_;
   sdb::FunctionBlock*   fn_block_;
+  llvm::BasicBlock*     return_block_;
+  llvm::BasicBlock*     internal_indirection_block_;
+  llvm::BasicBlock*     external_indirection_block_;
   llvm::BasicBlock*     bb_;
   llvm::IRBuilder<>*    builder_;
 
@@ -92,19 +103,15 @@ private:
   uint32_t      cia_;
 
   struct {
+    llvm::Value*  indirection_target;
+    llvm::Value*  indirection_cia;
+
     llvm::Value*  xer;
-    bool          xer_dirty;
     llvm::Value*  lr;
-    bool          lr_dirty;
     llvm::Value*  ctr;
-    bool          ctr_dirty;
-
     llvm::Value*  cr;
-    bool          cr_dirty;
-
     llvm::Value*  gpr[32];
-    uint32_t      gpr_dirty_bits;
-  } values_;
+  } locals_;
 };
 
 

include/xenia/cpu/ppc/state.h

@@ -40,6 +40,10 @@
 // }  // FPRF
 
 
+#define kXEPPCRegLR     0xFFFF0001
+#define kXEPPCRegCTR    0xFFFF0002
+
+
 typedef struct XECACHEALIGN xe_float4 {
   union {
     struct {