Fix operand latency computation in cases where the definition operand is

implicit. e.g.
%D6<def>, %D7<def> = VLD1q16 %R2<kill>, 0, ..., %Q3<imp-def>                                                                                                                                                                              
%Q1<def> = VMULv8i16 %Q1<kill>, %Q3<kill>, ...

The real definition indices are 0,1.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@116080 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 11 insertions, 0 deletions

diff --git a/lib/CodeGen/ScheduleDAGInstrs.cpp b/lib/CodeGen/ScheduleDAGInstrs.cpp
index bf68c0e..c0fa9dd 100644
--- a/lib/CodeGen/ScheduleDAGInstrs.cpp
+++ b/lib/CodeGen/ScheduleDAGInstrs.cpp
@@ -527,6 +527,17 @@ void ScheduleDAGInstrs::ComputeOperandLatency(SUnit *Def, SUnit *Use,
   MachineInstr *DefMI = Def->getInstr();
   int DefIdx = DefMI->findRegisterDefOperandIdx(Reg);
   if (DefIdx != -1) {
+    const MachineOperand &MO = DefMI->getOperand(DefIdx);
+    if (MO.isReg() && MO.isImplicit() &&
+        DefIdx >= DefMI->getDesc().getNumOperands()) {
+      // This is an implicit def, getOperandLatency() won't return the correct
+      // latency. e.g.
+      //   %D6<def>, %D7<def> = VLD1q16 %R2<kill>, 0, ..., %Q3<imp-def>
+      //   %Q1<def> = VMULv8i16 %Q1<kill>, %Q3<kill>, ...
+      // What we want is to compute latency between def of %D6/%D7 and use of
+      // %Q3 instead.
+      DefIdx = DefMI->findRegisterDefOperandIdx(Reg, false, true, TRI);
+    }
     MachineInstr *UseMI = Use->getInstr();
     // For all uses of the register, calculate the maxmimum latency
     int Latency = -1;