Second installment of "BasicBlock operands to the back"
changes.

For InvokeInst now all arguments begin at op_begin().
The Callee, Cont and Fail are now faster to get by
access relative to op_end().

This patch introduces some temporary uglyness in CallSite.
Next I'll bring CallInst up to a similar scheme and then
the uglyness will magically vanish.

This patch also exposes all the reliance of the libraries
on InvokeInst's operand ordering. I am thinking of taking
care of that too.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@66920 91177308-0d34-0410-b5e6-96231b3b80d8
diff --git a/lib/Transforms/IPO/DeadArgumentElimination.cpp b/lib/Transforms/IPO/DeadArgumentElimination.cpp
index dd5fce6..8fd6708 100644
--- a/lib/Transforms/IPO/DeadArgumentElimination.cpp
+++ b/lib/Transforms/IPO/DeadArgumentElimination.cpp
@@ -356,14 +356,14 @@
         // argument, since if it was the function argument this would be an
         // indirect call and the we know can't be looking at a value of the
         // label type (for the invoke instruction).
-        unsigned ArgNo = CS.getArgumentNo(U.getOperandNo());
+        unsigned ArgNo = CS.getArgumentNo(U);
 
         if (ArgNo >= F->getFunctionType()->getNumParams())
           // The value is passed in through a vararg! Must be live.
           return Live;
 
-        assert(CS.getArgument(ArgNo) 
-               == CS.getInstruction()->getOperand(U.getOperandNo()) 
+        assert(CS.getArgument(ArgNo)
+               == CS.getInstruction()->getOperand(U.getOperandNo())
                && "Argument is not where we expected it");
 
         // Value passed to a normal call. It's only live when the corresponding