Prevent remat of partial register redefinitions.

An instruction that redefines only part of a larger register can never
be rematerialized since the virtual register value depends on the old
value in other parts of the register.

This was fixed for the inline spiller in r138794.  This patch fixes the
problem for all register allocators, and includes a small test case.

<rdar://problem/10032939>

llvm-svn: 138944

1 files changed, 9 insertions, 0 deletions

diff --git a/llvm/lib/CodeGen/TargetInstrInfoImpl.cpp b/llvm/lib/CodeGen/TargetInstrInfoImpl.cpp
index ea0cfe2b940..1cd42ca2ae0 100644
--- a/llvm/lib/CodeGen/TargetInstrInfoImpl.cpp
+++ b/llvm/lib/CodeGen/TargetInstrInfoImpl.cpp
@@ -362,6 +362,15 @@ isReallyTriviallyReMaterializableGeneric(const MachineInstr *MI,
   const TargetInstrInfo &TII = *TM.getInstrInfo();
   const TargetRegisterInfo &TRI = *TM.getRegisterInfo();
 
+  // A sub-register definition can only be rematerialized if the instruction
+  // doesn't read the other parts of the register.  Otherwise it is really a
+  // read-modify-write operation on the full virtual register which cannot be
+  // moved safely.
+  unsigned Reg = MI->getOperand(0).getReg();
+  if (TargetRegisterInfo::isVirtualRegister(Reg) &&
+      MI->getOperand(0).getSubReg() && MI->readsVirtualRegister(Reg))
+    return false;
+
   // A load from a fixed stack slot can be rematerialized. This may be
   // redundant with subsequent checks, but it's target-independent,
   // simple, and a common case.