Add patterns for CALL32m and CALL64m. They aren't matched in most

cases due to an isel deficiency already noted in
lib/Target/X86/README.txt, but they can be matched in this fold-call.ll
testcase, for example.

This is interesting mainly because it exposes a tricky tblgen bug;
tblgen was incorrectly computing the starting index for variable_ops
in the case of a complex pattern.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@51706 91177308-0d34-0410-b5e6-96231b3b80d8

lib/Target/X86/X86Instr64bit.td

@@ -101,7 +101,7 @@ let isCall = 1 in
     def CALL64r       : I<0xFF, MRM2r, (outs), (ins GR64:$dst, variable_ops),
                           "call\t{*}$dst", [(X86call GR64:$dst)]>;
     def CALL64m       : I<0xFF, MRM2m, (outs), (ins i64mem:$dst, variable_ops),
-                          "call\t{*}$dst", []>;
+                          "call\t{*}$dst", [(X86call (loadi64 addr:$dst))]>;
   }
 
 

lib/Target/X86/X86InstrInfo.td

@@ -397,7 +397,7 @@ let isCall = 1 in
     def CALL32r     : I<0xFF, MRM2r, (outs), (ins GR32:$dst, variable_ops),
                         "call\t{*}$dst", [(X86call GR32:$dst)]>;
     def CALL32m     : I<0xFF, MRM2m, (outs), (ins i32mem:$dst, variable_ops),
-                        "call\t{*}$dst", []>;
+                        "call\t{*}$dst", [(X86call (loadi32 addr:$dst))]>;
   }
 
 // Tail call stuff.

test/CodeGen/X86/fold-call.ll

@@ -0,0 +1,10 @@
+; RUN: llvm-as < %s | llc -march=x86 | not grep mov
+; RUN: llvm-as < %s | llc -march=x86-64 | not grep mov
+
+declare void @bar()
+
+define void @foo(i32 %i0, i32 %i1, i32 %i2, i32 %i3, i32 %i4, i32 %i5, void()* %arg) nounwind {
+	call void @bar()
+	call void %arg()
+	ret void
+}

utils/TableGen/DAGISelEmitter.cpp

@@ -947,7 +947,7 @@ public:
       // instruction operands to do this.
       std::vector<std::string> AllOps;
       unsigned NumEAInputs = 0; // # of synthesized 'execute always' inputs.
-      unsigned NumDiscardedInputs = 0; // # of 'discard' inputs to skip.
+      unsigned InputIndex = 0;
       for (unsigned ChildNo = 0, InstOpNo = NumResults;
            InstOpNo != II.OperandList.size(); ++InstOpNo) {
         std::vector<std::string> Ops;
@@ -956,7 +956,7 @@ public:
         Record *OperandNode = II.OperandList[InstOpNo].Rec;
         if (OperandNode->getName() == "discard") {
           // This is a "discard" operand; emit nothing. Just note it.
-          ++NumDiscardedInputs;
+          ++InputIndex;
         } else if ((OperandNode->isSubClassOf("PredicateOperand") ||
                     OperandNode->isSubClassOf("OptionalDefOperand")) &&
                    !CGP.getDefaultOperand(OperandNode).DefaultOps.empty()) {
@@ -970,6 +970,7 @@ public:
             AllOps.insert(AllOps.end(), Ops.begin(), Ops.end());
             NumEAInputs += Ops.size();
           }
+          ++InputIndex;
         } else {
           // Otherwise this is a normal operand or a predicate operand without
           // 'execute always'; emit it.
@@ -977,6 +978,7 @@ public:
                                InFlagDecled, ResNodeDecled);
           AllOps.insert(AllOps.end(), Ops.begin(), Ops.end());
           ++ChildNo;
+          ++InputIndex;
         }
       }
 
@@ -1062,10 +1064,7 @@ public:
           // number of operands that are 'execute always'. This is the index
           // where we should start copying operands into the 'variable_ops'
           // portion of the output.
-          unsigned InputIndex = AllOps.size() +
-                                NumDiscardedInputs +
-                                NodeHasChain -
-                                NumEAInputs;
+          InputIndex += NodeHasChain - NumEAInputs;
         
           for (unsigned i = 0, e = AllOps.size(); i != e; ++i)
             emitCode("Ops" + utostr(OpsNo) + ".push_back(" + AllOps[i] + ");");