CodeGen: Refactor regallocator command line and target selection

This will allow targets more flexibility to replace the
register allocator core passes. In a future commit,
AMDGPU will run the core register assignment passes
twice, and will also want to disallow using the
standard -regalloc option.

llvm-svn: 356506

diff --git a/llvm/include/llvm/CodeGen/TargetPassConfig.h b/llvm/include/llvm/CodeGen/TargetPassConfig.h
index 1def50d..80a2684 100644 (file)
--- a/llvm/include/llvm/CodeGen/TargetPassConfig.h
+++ b/llvm/include/llvm/CodeGen/TargetPassConfig.h
@@ -360,11 +360,11 @@ protected:
 
   /// addFastRegAlloc - Add the minimum set of target-independent passes that
   /// are required for fast register allocation.
-  virtual void addFastRegAlloc(FunctionPass *RegAllocPass);
+  virtual void addFastRegAlloc();
 
   /// addOptimizedRegAlloc - Add passes related to register allocation.
   /// LLVMTargetMachine provides standard regalloc passes for most targets.
-  virtual void addOptimizedRegAlloc(FunctionPass *RegAllocPass);
+  virtual void addOptimizedRegAlloc();
 
   /// addPreRewrite - Add passes to the optimized register allocation pipeline
   /// after register allocation is complete, but before virtual registers are
@@ -374,6 +374,10 @@ protected:
   /// after RABasic or RAGreedy, they should take advantage of LiveRegMatrix.
   /// When these passes run, VirtRegMap contains legal physreg assignments for
   /// all virtual registers.
+  ///
+  /// Note if the target overloads addRegAssignAndRewriteOptimized, this may not
+  /// be honored. This is also not generally used for the the fast variant,
+  /// where the allocation and rewriting are done in one pass.
   virtual bool addPreRewrite() {
     return false;
   }
@@ -431,7 +435,12 @@ protected:
 
   /// addMachinePasses helper to create the target-selected or overriden
   /// regalloc pass.
-  FunctionPass *createRegAllocPass(bool Optimized);
+  virtual FunctionPass *createRegAllocPass(bool Optimized);
+
+  /// Add core register alloator passes which do the actual register assignment
+  /// and rewriting. \returns true if any passes were added.
+  virtual bool addRegAssignmentFast();
+  virtual bool addRegAssignmentOptimized();
 };
 
 } // end namespace llvm

diff --git a/llvm/lib/CodeGen/TargetPassConfig.cpp b/llvm/lib/CodeGen/TargetPassConfig.cpp
index 76275da..8ddde4b 100644 (file)
--- a/llvm/lib/CodeGen/TargetPassConfig.cpp
+++ b/llvm/lib/CodeGen/TargetPassConfig.cpp
@@ -897,13 +897,9 @@ void TargetPassConfig::addMachinePasses() {
   // Run register allocation and passes that are tightly coupled with it,
   // including phi elimination and scheduling.
   if (getOptimizeRegAlloc())
-    addOptimizedRegAlloc(createRegAllocPass(true));
-  else {
-    if (RegAlloc != &useDefaultRegisterAllocator &&
-        RegAlloc != &createFastRegisterAllocator)
-      report_fatal_error("Must use fast (default) register allocator for unoptimized regalloc.");
-    addFastRegAlloc(createRegAllocPass(false));
-  }
+    addOptimizedRegAlloc();
+  else
+    addFastRegAlloc();
 
   // Run post-ra passes.
   addPostRegAlloc();
@@ -1093,6 +1089,33 @@ FunctionPass *TargetPassConfig::createRegAllocPass(bool Optimized) {
   return createTargetRegisterAllocator(Optimized);
 }
 
+bool TargetPassConfig::addRegAssignmentFast() {
+  if (RegAlloc != &useDefaultRegisterAllocator &&
+      RegAlloc != &createFastRegisterAllocator)
+    report_fatal_error("Must use fast (default) register allocator for unoptimized regalloc.");
+
+  addPass(createRegAllocPass(false));
+  return true;
+}
+
+bool TargetPassConfig::addRegAssignmentOptimized() {
+  // Add the selected register allocation pass.
+  addPass(createRegAllocPass(true));
+
+  // Allow targets to change the register assignments before rewriting.
+  addPreRewrite();
+
+  // Finally rewrite virtual registers.
+  addPass(&VirtRegRewriterID);
+  // Perform stack slot coloring and post-ra machine LICM.
+  //
+  // FIXME: Re-enable coloring with register when it's capable of adding
+  // kill markers.
+  addPass(&StackSlotColoringID);
+
+  return true;
+}
+
 /// Return true if the default global register allocator is in use and
 /// has not be overriden on the command line with '-regalloc=...'
 bool TargetPassConfig::usingDefaultRegAlloc() const {
@@ -1101,18 +1124,17 @@ bool TargetPassConfig::usingDefaultRegAlloc() const {
 
 /// Add the minimum set of target-independent passes that are required for
 /// register allocation. No coalescing or scheduling.
-void TargetPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
+void TargetPassConfig::addFastRegAlloc() {
   addPass(&PHIEliminationID, false);
   addPass(&TwoAddressInstructionPassID, false);
 
-  if (RegAllocPass)
-    addPass(RegAllocPass);
+  addRegAssignmentFast();
 }
 
 /// Add standard target-independent passes that are tightly coupled with
 /// optimized register allocation, including coalescing, machine instruction
 /// scheduling, and register allocation itself.
-void TargetPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
+void TargetPassConfig::addOptimizedRegAlloc() {
   addPass(&DetectDeadLanesID, false);
 
   addPass(&ProcessImplicitDefsID, false);
@@ -1144,22 +1166,7 @@ void TargetPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
   // PreRA instruction scheduling.
   addPass(&MachineSchedulerID);
 
-  if (RegAllocPass) {
-    // Add the selected register allocation pass.
-    addPass(RegAllocPass);
-
-    // Allow targets to change the register assignments before rewriting.
-    addPreRewrite();
-
-    // Finally rewrite virtual registers.
-    addPass(&VirtRegRewriterID);
-
-    // Perform stack slot coloring and post-ra machine LICM.
-    //
-    // FIXME: Re-enable coloring with register when it's capable of adding
-    // kill markers.
-    addPass(&StackSlotColoringID);
-
+  if (addRegAssignmentOptimized()) {
     // Copy propagate to forward register uses and try to eliminate COPYs that
     // were not coalesced.
     addPass(&MachineCopyPropagationID);

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 4d57193..e592814 100644 (file)
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -579,8 +579,8 @@ public:
   bool addLegalizeMachineIR() override;
   bool addRegBankSelect() override;
   bool addGlobalInstructionSelect() override;
-  void addFastRegAlloc(FunctionPass *RegAllocPass) override;
-  void addOptimizedRegAlloc(FunctionPass *RegAllocPass) override;
+  void addFastRegAlloc() override;
+  void addOptimizedRegAlloc() override;
   void addPreRegAlloc() override;
   void addPostRegAlloc() override;
   void addPreSched2() override;
@@ -865,7 +865,7 @@ void GCNPassConfig::addPreRegAlloc() {
   addPass(createSIWholeQuadModePass());
 }
 
-void GCNPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
+void GCNPassConfig::addFastRegAlloc() {
   // FIXME: We have to disable the verifier here because of PHIElimination +
   // TwoAddressInstructions disabling it.
 
@@ -878,10 +878,10 @@ void GCNPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
   // machine-level CFG, but before register allocation.
   insertPass(&SILowerControlFlowID, &SIFixWWMLivenessID, false);
 
-  TargetPassConfig::addFastRegAlloc(RegAllocPass);
+  TargetPassConfig::addFastRegAlloc();
 }
 
-void GCNPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
+void GCNPassConfig::addOptimizedRegAlloc() {
   insertPass(&MachineSchedulerID, &SIOptimizeExecMaskingPreRAID);
 
   insertPass(&SIOptimizeExecMaskingPreRAID, &SIFormMemoryClausesID);
@@ -895,7 +895,7 @@ void GCNPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
   // machine-level CFG, but before register allocation.
   insertPass(&SILowerControlFlowID, &SIFixWWMLivenessID, false);
 
-  TargetPassConfig::addOptimizedRegAlloc(RegAllocPass);
+  TargetPassConfig::addOptimizedRegAlloc();
 }
 
 void GCNPassConfig::addPostRegAlloc() {

diff --git a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp
index eb8e40a..a5c6f34 100644 (file)
--- a/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp
+++ b/llvm/lib/Target/NVPTX/NVPTXTargetMachine.cpp
@@ -166,8 +166,16 @@ public:
   void addMachineSSAOptimization() override;
 
   FunctionPass *createTargetRegisterAllocator(bool) override;
-  void addFastRegAlloc(FunctionPass *RegAllocPass) override;
-  void addOptimizedRegAlloc(FunctionPass *RegAllocPass) override;
+  void addFastRegAlloc() override;
+  void addOptimizedRegAlloc() override;
+
+  bool addRegAssignmentFast() override {
+    llvm_unreachable("should not be used");
+  }
+
+  bool addRegAssignmentOptimized() override {
+    llvm_unreachable("should not be used");
+  }
 
 private:
   // If the opt level is aggressive, add GVN; otherwise, add EarlyCSE. This
@@ -322,15 +330,12 @@ FunctionPass *NVPTXPassConfig::createTargetRegisterAllocator(bool) {
   return nullptr; // No reg alloc
 }
 
-void NVPTXPassConfig::addFastRegAlloc(FunctionPass *RegAllocPass) {
-  assert(!RegAllocPass && "NVPTX uses no regalloc!");
+void NVPTXPassConfig::addFastRegAlloc() {
   addPass(&PHIEliminationID);
   addPass(&TwoAddressInstructionPassID);
 }
 
-void NVPTXPassConfig::addOptimizedRegAlloc(FunctionPass *RegAllocPass) {
-  assert(!RegAllocPass && "NVPTX uses no regalloc!");
-
+void NVPTXPassConfig::addOptimizedRegAlloc() {
   addPass(&ProcessImplicitDefsID);
   addPass(&LiveVariablesID);
   addPass(&MachineLoopInfoID);

diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
index 9b10c8d..414f055 100644 (file)
--- a/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyTargetMachine.cpp
@@ -191,6 +191,12 @@ public:
   void addPostRegAlloc() override;
   bool addGCPasses() override { return false; }
   void addPreEmitPass() override;
+
+  // No reg alloc
+  bool addRegAssignmentFast() override { return false; }
+
+  // No reg alloc
+  bool addRegAssignmentOptimized() override { return false; }
 };
 } // end anonymous namespace