The coalescer does not coalesce a virtual register to a physical register if any of the physical register's sub-register live intervals overlaps with the virtual register. This is overly conservative. It prevents a extract_subreg from being coalesced away:

v1024 = EDI  // not killed
      =
      = EDI

One possible solution is for the coalescer to examine the sub-register live intervals in the same manner as the physical register. Another possibility is to examine defs and uses (when needed) of sub-registers. Both solutions are too expensive. For now, look for "short virtual intervals" and scan instructions to look for conflict instead.

This is a small win on x86-64. e.g. It shaves 403.gcc by ~80 instructions.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@61847 91177308-0d34-0410-b5e6-96231b3b80d8

include/llvm/CodeGen/LiveIntervalAnalysis.h

@@ -24,6 +24,7 @@
 #include "llvm/CodeGen/LiveInterval.h"
 #include "llvm/ADT/BitVector.h"
 #include "llvm/ADT/DenseMap.h"
+#include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/Support/Allocator.h"
 #include <cmath>
@@ -256,6 +257,12 @@ namespace llvm {
     bool conflictsWithPhysRegDef(const LiveInterval &li, VirtRegMap &vrm,
                                  unsigned reg);
 
+    /// conflictsWithPhysRegRef - Similar to conflictsWithPhysRegRef except
+    /// it can check use as well.
+    bool conflictsWithPhysRegRef(LiveInterval &li, unsigned Reg,
+                                 bool CheckUse,
+                                 SmallPtrSet<MachineInstr*,32> &JoinedCopies);
+
     /// findLiveInMBBs - Given a live range, if the value of the range
     /// is live in any MBB returns true as well as the list of basic blocks
     /// in which the value is live.