Revert "Revert "[MIR] Target specific MIR formating and parsing""

There was an unguarded dereference of MF in a function that permitted
nullptr. Fixed

This reverts commit 71d64f7.

Author: dsandersllvm

llvm/include/llvm/CodeGen/MIRFormatter.h

@@ -0,0 +1,83 @@
+//===-- llvm/CodeGen/MIRFormatter.h -----------------------------*- C++ -*-===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file contains the declaration of the MIRFormatter class.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_CODEGEN_MIRFORMATTER_H
+#define LLVM_CODEGEN_MIRFORMATTER_H
+
+#include "llvm/CodeGen/MachineInstr.h"
+#include "llvm/CodeGen/PseudoSourceValue.h"
+
+namespace llvm {
+
+struct PerFunctionMIParsingState;
+struct SlotMapping;
+
+/// MIRFormater - Interface to format MIR operand based on target
+class MIRFormatter {
+public:
+  typedef function_ref<bool(StringRef::iterator Loc, const Twine &)>
+      ErrorCallbackType;
+
+  MIRFormatter() {}
+  virtual ~MIRFormatter() = default;
+
+  /// Implement target specific printing for machine operand immediate value, so
+  /// that we can have more meaningful mnemonic than a 64-bit integer. Passing
+  /// None to OpIdx means the index is unknown.
+  virtual void printImm(raw_ostream &OS, const MachineInstr &MI,
+                        Optional<unsigned> OpIdx, int64_t Imm) const {
+    OS << Imm;
+  }
+
+  /// Implement target specific parsing of immediate mnemonics. The mnemonic is
+  /// dot seperated strings.
+  virtual bool parseImmMnemonic(const unsigned OpCode, const unsigned OpIdx,
+                                StringRef Src, int64_t &Imm,
+                                ErrorCallbackType ErrorCallback) const {
+    llvm_unreachable("target did not implement parsing MIR immediate mnemonic");
+  }
+
+  /// Implement target specific printing of target custom pseudo source value.
+  /// Default implementation is not necessarily the correct MIR serialization
+  /// format.
+  virtual void
+  printCustomPseudoSourceValue(raw_ostream &OS, ModuleSlotTracker &MST,
+                               const PseudoSourceValue &PSV) const {
+    PSV.printCustom(OS);
+  }
+
+  /// Implement target specific parsing of target custom pseudo source value.
+  virtual bool parseCustomPseudoSourceValue(
+      StringRef Src, MachineFunction &MF, PerFunctionMIParsingState &PFS,
+      const PseudoSourceValue *&PSV, ErrorCallbackType ErrorCallback) const {
+    llvm_unreachable(
+        "target did not implement parsing MIR custom pseudo source value");
+  }
+
+  /// Helper functions to print IR value as MIR serialization format which will
+  /// be useful for target specific printer, e.g. for printing IR value in
+  /// custom pseudo source value.
+  static void printIRValue(raw_ostream &OS, const Value &V,
+                           ModuleSlotTracker &MST);
+
+  /// Helper functions to parse IR value from MIR serialization format which
+  /// will be useful for target specific parser, e.g. for parsing IR value for
+  /// custom pseudo source value.
+  static bool parseIRValue(StringRef Src, MachineFunction &MF,
+                           PerFunctionMIParsingState &PFS, const Value *&V,
+                           ErrorCallbackType ErrorCallback);
+};
+
+} // end namespace llvm
+
+#endif

llvm/include/llvm/CodeGen/MIRParser/MIParser.h

@@ -171,12 +171,16 @@ struct PerFunctionMIParsingState {
   DenseMap<unsigned, unsigned> ConstantPoolSlots;
   DenseMap<unsigned, unsigned> JumpTableSlots;
 
+  /// Maps from slot numbers to function's unnamed values.
+  DenseMap<unsigned, const Value *> Slots2Values;
+
   PerFunctionMIParsingState(MachineFunction &MF, SourceMgr &SM,
                             const SlotMapping &IRSlots,
                             PerTargetMIParsingState &Target);
 
   VRegInfo &getVRegInfo(unsigned Num);
   VRegInfo &getVRegInfoNamed(StringRef RegName);
+  const Value *getIRValue(unsigned Slot);
 };
 
 /// Parse the machine basic block definitions, and skip the machine

llvm/include/llvm/CodeGen/MachineMemOperand.h

@@ -26,6 +26,7 @@ namespace llvm {
 
 class FoldingSetNodeID;
 class MDNode;
+class MIRFormatter;
 class raw_ostream;
 class MachineFunction;
 class ModuleSlotTracker;
@@ -295,7 +296,8 @@ class MachineMemOperand {
   /// @{
   void print(raw_ostream &OS, ModuleSlotTracker &MST,
              SmallVectorImpl<StringRef> &SSNs, const LLVMContext &Context,
-             const MachineFrameInfo *MFI, const TargetInstrInfo *TII) const;
+             const MachineFrameInfo *MFI, const TargetInstrInfo *TII,
+             const MIRFormatter *MIRF) const;
   /// @}
 
   friend bool operator==(const MachineMemOperand &LHS,

llvm/include/llvm/CodeGen/MachineOperand.h

@@ -278,6 +278,9 @@ class MachineOperand {
   /// More complex way of printing a MachineOperand.
   /// \param TypeToPrint specifies the generic type to be printed on uses and
   /// defs. It can be determined using MachineInstr::getTypeToPrint.
+  /// \param OpIdx - specifies the index of the operand in machine instruction.
+  /// This will be used by target dependent MIR formatter. Could be None if the
+  /// index is unknown, e.g. called by dump().
   /// \param PrintDef - whether we want to print `def` on an operand which
   /// isDef. Sometimes, if the operand is printed before '=', we don't print
   /// `def`.
@@ -294,8 +297,9 @@ class MachineOperand {
   /// information from it's parent.
   /// \param IntrinsicInfo - same as \p TRI.
   void print(raw_ostream &os, ModuleSlotTracker &MST, LLT TypeToPrint,
-             bool PrintDef, bool IsStandalone, bool ShouldPrintRegisterTies,
-             unsigned TiedOperandIdx, const TargetRegisterInfo *TRI,
+             Optional<unsigned> OpIdx, bool PrintDef, bool IsStandalone,
+             bool ShouldPrintRegisterTies, unsigned TiedOperandIdx,
+             const TargetRegisterInfo *TRI,
              const TargetIntrinsicInfo *IntrinsicInfo) const;
 
   /// Same as print(os, TRI, IntrinsicInfo), but allows to specify the low-level

llvm/include/llvm/CodeGen/PseudoSourceValue.h

@@ -22,6 +22,7 @@ namespace llvm {
 
 class MachineFrameInfo;
 class MachineMemOperand;
+class MIRFormatter;
 class raw_ostream;
 class TargetInstrInfo;
 
@@ -52,6 +53,7 @@ class PseudoSourceValue {
                                        const PseudoSourceValue* PSV);
 
   friend class MachineMemOperand; // For printCustom().
+  friend class MIRFormatter;      // For printCustom().
 
   /// Implement printing for PseudoSourceValue. This is called from
   /// Value::print or Value's operator<<.

llvm/include/llvm/CodeGen/TargetInstrInfo.h

@@ -18,6 +18,7 @@
 #include "llvm/ADT/DenseMapInfo.h"
 #include "llvm/ADT/None.h"
 #include "llvm/CodeGen/LiveRegUnits.h"
+#include "llvm/CodeGen/MIRFormatter.h"
 #include "llvm/CodeGen/MachineBasicBlock.h"
 #include "llvm/CodeGen/MachineCombinerPattern.h"
 #include "llvm/CodeGen/MachineFunction.h"
@@ -1807,6 +1808,7 @@ class TargetInstrInfo : public MCInstrInfo {
                                                          Register Reg) const;
 
 private:
+  mutable std::unique_ptr<MIRFormatter> Formatter;
   unsigned CallFrameSetupOpcode, CallFrameDestroyOpcode;
   unsigned CatchRetOpcode;
   unsigned ReturnOpcode;

llvm/include/llvm/Target/TargetMachine.h

@@ -33,6 +33,7 @@ class MCInstrInfo;
 class MCRegisterInfo;
 class MCSubtargetInfo;
 class MCSymbol;
+class MIRFormatter;
 class raw_pwrite_stream;
 class PassManagerBuilder;
 struct PerFunctionMIParsingState;
@@ -94,6 +95,7 @@ class TargetMachine {
   std::unique_ptr<const MCRegisterInfo> MRI;
   std::unique_ptr<const MCInstrInfo> MII;
   std::unique_ptr<const MCSubtargetInfo> STI;
+  std::unique_ptr<const MIRFormatter> MIRF;
 
   unsigned RequireStructuredCFG : 1;
   unsigned O0WantsFastISel : 1;
@@ -197,6 +199,10 @@ class TargetMachine {
     return nullptr;
   }
 
+  /// Return MIR formatter to format/parse MIR operands.  Target can override
+  /// this virtual function and return target specific MIR formatter.
+  virtual const MIRFormatter *getMIRFormatter() const { return MIRF.get(); }
+
   bool requiresStructuredCFG() const { return RequireStructuredCFG; }
   void setRequiresStructuredCFG(bool Value) { RequireStructuredCFG = Value; }
 

llvm/lib/CodeGen/MIRParser/MILexer.cpp

@@ -242,6 +242,7 @@ static MIToken::TokenKind getIdentifierKind(StringRef Identifier) {
       .Case("jump-table", MIToken::kw_jump_table)
       .Case("constant-pool", MIToken::kw_constant_pool)
       .Case("call-entry", MIToken::kw_call_entry)
+      .Case("custom", MIToken::kw_custom)
       .Case("liveout", MIToken::kw_liveout)
       .Case("address-taken", MIToken::kw_address_taken)
       .Case("landing-pad", MIToken::kw_landing_pad)

llvm/lib/CodeGen/MIRParser/MILexer.h

@@ -110,6 +110,7 @@ struct MIToken {
     kw_jump_table,
     kw_constant_pool,
     kw_call_entry,
+    kw_custom,
     kw_liveout,
     kw_address_taken,
     kw_landing_pad,