Local spiller optimization:

Turn this:
movswl  %ax, %eax
movl    %eax, -36(%ebp)
xorl    %edi, -36(%ebp)
into
movswl  %ax, %eax
xorl    %edi, %eax
movl    %eax, -36(%ebp)
by unfolding the load / store xorl into an xorl and a store when we know the
value in the spill slot is available in a register. This doesn't change the
number of instructions but reduce the number of times memory is accessed.

Also unfold some load folding instructions and reuse the value when similar
situation presents itself.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@42947 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 4 insertions, 0 deletions

diff --git a/lib/CodeGen/VirtRegMap.h b/lib/CodeGen/VirtRegMap.h
index 8385f3d..f6d305e 100644
--- a/lib/CodeGen/VirtRegMap.h
+++ b/lib/CodeGen/VirtRegMap.h
@@ -181,6 +181,10 @@ namespace llvm {
     void virtFolded(unsigned VirtReg, MachineInstr *OldMI, unsigned OpNum,
                     MachineInstr *NewMI);
 
+    /// @brief Updates information about the specified virtual register's value
+    /// folded into the specified machine instruction.
+    void virtFolded(unsigned VirtReg, MachineInstr *MI, ModRef MRInfo);
+
     /// @brief returns the virtual registers' values folded in memory
     /// operands of this instruction
     std::pair<MI2VirtMapTy::const_iterator, MI2VirtMapTy::const_iterator>