[PATCH] LiveRangeCalc: Don't start liveranges of PHI instruction at the block begin.

[Matthias Braun]

Hi qcolombet,

Letting them begin at the PHI instruction slightly simplifies the code
but more importantly avoids breaking the assumption that live ranges
starting at the block begin are also live at the end of the predecessor
blocks. The MachineVerifier checks that but was apparently never run in
the few instances where liveranges are calculated for machine-SSA
functions.

http://reviews.llvm.org/D7779

Files:
  lib/CodeGen/LiveRangeCalc.cpp

Index: lib/CodeGen/LiveRangeCalc.cpp
===================================================================
--- lib/CodeGen/LiveRangeCalc.cpp
+++ lib/CodeGen/LiveRangeCalc.cpp
@@ -43,11 +43,8 @@
 static void createDeadDef(SlotIndexes &Indexes, VNInfo::Allocator &Alloc,
                           LiveRange &LR, const MachineOperand &MO) {
     const MachineInstr *MI = MO.getParent();
-    SlotIndex DefIdx;
-    if (MI->isPHI())
-      DefIdx = Indexes.getMBBStartIdx(MI->getParent());
-    else
-      DefIdx = Indexes.getInstructionIndex(MI).getRegSlot(MO.isEarlyClobber());
+    SlotIndex DefIdx =
+        Indexes.getInstructionIndex(MI).getRegSlot(MO.isEarlyClobber());
 
     // Create the def in LR. This may find an existing def.
     LR.createDeadDef(DefIdx, Alloc);

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: D7779.20364.patch
Type: text/x-patch
Size: 775 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150220/c6f85a31/attachment.bin>