Let each target determines whether a machine instruction is dead. If true, that allows late codeine passes to delete it.

This is considered a workaround. The problem is some targets are not modeling side effects correctly. PPC is apparently one of those. This patch allows ppc llvm-gcc to bootstrap on Darwin. Once we find out which instruction definitions are wrong, we can remove the PPCInstrInfo workaround.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@76703 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 6 insertions, 0 deletions

diff --git a/include/llvm/Target/TargetInstrInfo.h b/include/llvm/Target/TargetInstrInfo.h
index 0ab57ca..538d139 100644
--- a/include/llvm/Target/TargetInstrInfo.h
+++ b/include/llvm/Target/TargetInstrInfo.h
@@ -470,6 +470,10 @@ public:
     return true;
   }
 
+  /// isDeadInstruction - Return true if the instruction is considered dead.
+  /// This allows some late codegen passes to delete them.
+  virtual bool isDeadInstruction(const MachineInstr *MI) const = 0;
+
   /// GetInstSize - Returns the size of the specified Instruction.
   /// 
   virtual unsigned GetInstSizeInBytes(const MachineInstr *MI) const {
@@ -501,6 +505,8 @@ public:
                              MachineBasicBlock::iterator MI,
                              unsigned DestReg, unsigned SubReg,
                              const MachineInstr *Orig) const;
+  virtual bool isDeadInstruction(const MachineInstr *MI) const;
+
   virtual unsigned GetFunctionSizeInBytes(const MachineFunction &MF) const;
 };