[llvm] r197691 - Add support for positionally-encoded operands to FixedLenDecoderEmitter

Hal Finkel hfinkel at anl.gov
Thu Dec 19 08:12:53 PST 2013


Author: hfinkel
Date: Thu Dec 19 10:12:53 2013
New Revision: 197691

URL: http://llvm.org/viewvc/llvm-project?rev=197691&view=rev
Log:
Add support for positionally-encoded operands to FixedLenDecoderEmitter

Unfortunately, the PowerPC instruction definitions make heavy use of the
positional operand encoding heuristic to map operands onto bitfield variables
in the instruction definitions. Changing this to use name-based mapping is not
trivial, however, because additional infrastructure needs to be designed to
handle mapping of complex operands (with multiple suboperands) onto multiple
bitfield variables.

In the mean time, this adds support for positionally encoded operands to
FixedLenDecoderEmitter, so that we can generate a disassembler for the PowerPC
backend. To prevent an accidental reliance on this feature, and to prevent an
undesirable interaction with existing disassemblers, a backend must opt-in to
this support by setting the new decodePositionallyEncodedOperands
instruction-set bit to true.

When enabled, this iterates the variables that contribute to the instruction
encoding, just as the encoder does, and emulates the procedure the encoder uses
to map "numbered" operands to variables. The bit range for each variable is
also determined as the encoder determines them. This map is then consulted
during the decoder-generator's loop over operands to decode, allowing the
decoder to understand both position-based and name-based operand-to-variable
mappings.

As noted in the comment on the decodePositionallyEncodedOperands definition,
this support should be removed once it is no longer needed. There should be no
change to existing disassemblers.

Modified:
    llvm/trunk/include/llvm/Target/Target.td
    llvm/trunk/utils/TableGen/FixedLenDecoderEmitter.cpp

Modified: llvm/trunk/include/llvm/Target/Target.td
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/include/llvm/Target/Target.td?rev=197691&r1=197690&r2=197691&view=diff
==============================================================================
--- llvm/trunk/include/llvm/Target/Target.td (original)
+++ llvm/trunk/include/llvm/Target/Target.td Thu Dec 19 10:12:53 2013
@@ -685,6 +685,18 @@ class InstrInfo {
   //
   // This option is a temporary migration help. It will go away.
   bit guessInstructionProperties = 1;
+
+  // TableGen's instruction encoder generator has support for matching operands
+  // to bit-field variables both by name and by position. While matching by
+  // name is preferred, this is currently not possible for complex operands,
+  // and some targets still reply on the positional encoding rules. When
+  // generating a decoder for such targets, the positional encoding rules must
+  // be used by the decoder generator as well.
+  //
+  // This option is temporary; it will go away once the TableGen decoder
+  // generator has better support for complex operands and targets have
+  // migrated away from using positionally encoded operands.
+  bit decodePositionallyEncodedOperands = 0;
 }
 
 // Standard Pseudo Instructions.

Modified: llvm/trunk/utils/TableGen/FixedLenDecoderEmitter.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/utils/TableGen/FixedLenDecoderEmitter.cpp?rev=197691&r1=197690&r2=197691&view=diff
==============================================================================
--- llvm/trunk/utils/TableGen/FixedLenDecoderEmitter.cpp (original)
+++ llvm/trunk/utils/TableGen/FixedLenDecoderEmitter.cpp Thu Dec 19 10:12:53 2013
@@ -1698,7 +1698,8 @@ void FilterChooser::emitTableEntries(Dec
   }
 }
 
-static bool populateInstruction(const CodeGenInstruction &CGI, unsigned Opc,
+static bool populateInstruction(CodeGenTarget &Target,
+                       const CodeGenInstruction &CGI, unsigned Opc,
                        std::map<unsigned, std::vector<OperandInfo> > &Operands){
   const Record &Def = *CGI.TheDef;
   // If all the bit positions are not specified; do not decode this instruction.
@@ -1747,14 +1748,165 @@ static bool populateInstruction(const Co
   for (unsigned i = 0; i < CGI.Operands.size(); ++i) {
     int tiedTo = CGI.Operands[i].getTiedRegister();
     if (tiedTo != -1) {
-      TiedNames[InOutOperands[i].second] = InOutOperands[tiedTo].second;
-      TiedNames[InOutOperands[tiedTo].second] = InOutOperands[i].second;
+      std::pair<unsigned, unsigned> SO =
+        CGI.Operands.getSubOperandNumber(tiedTo);
+      TiedNames[InOutOperands[i].second] = InOutOperands[SO.first].second;
+      TiedNames[InOutOperands[SO.first].second] = InOutOperands[i].second;
+    }
+  }
+
+  std::map<std::string, std::vector<OperandInfo> > NumberedInsnOperands;
+  std::set<std::string> NumberedInsnOperandsNoTie;
+  if (Target.getInstructionSet()->
+        getValueAsBit("decodePositionallyEncodedOperands")) {
+    const std::vector<RecordVal> &Vals = Def.getValues();
+    unsigned NumberedOp = 0;
+
+    for (unsigned i = 0, e = Vals.size(); i != e; ++i) {
+      // Ignore fixed fields in the record, we're looking for values like:
+      //    bits<5> RST = { ?, ?, ?, ?, ? };
+      if (Vals[i].getPrefix() || Vals[i].getValue()->isComplete())
+        continue;
+
+      // Determine if Vals[i] actually contributes to the Inst encoding.
+      unsigned bi = 0;
+      for (; bi < Bits.getNumBits(); ++bi) {
+        VarInit *Var = 0;
+        VarBitInit *BI = dyn_cast<VarBitInit>(Bits.getBit(bi));
+        if (BI)
+          Var = dyn_cast<VarInit>(BI->getBitVar());
+        else
+          Var = dyn_cast<VarInit>(Bits.getBit(bi));
+
+        if (Var && Var->getName() == Vals[i].getName())
+          break;
+      }
+
+      if (bi == Bits.getNumBits())
+        continue;
+
+      // Skip variables that correspond to explicitly-named operands.
+      unsigned OpIdx;
+      if (CGI.Operands.hasOperandNamed(Vals[i].getName(), OpIdx))
+        continue;
+
+      // Get the bit range for this operand:
+      unsigned bitStart = bi++, bitWidth = 1;
+      for (; bi < Bits.getNumBits(); ++bi) {
+        VarInit *Var = 0;
+        VarBitInit *BI = dyn_cast<VarBitInit>(Bits.getBit(bi));
+        if (BI)
+          Var = dyn_cast<VarInit>(BI->getBitVar());
+        else
+          Var = dyn_cast<VarInit>(Bits.getBit(bi));
+
+        if (!Var)
+          break;
+
+        if (Var->getName() != Vals[i].getName())
+          break;
+
+        ++bitWidth;
+      }
+
+      unsigned NumberOps = CGI.Operands.size();
+      while (NumberedOp < NumberOps &&
+             CGI.Operands.isFlatOperandNotEmitted(NumberedOp))
+        ++NumberedOp;
+
+      OpIdx = NumberedOp++;
+
+      // OpIdx now holds the ordered operand number of Vals[i].
+      std::pair<unsigned, unsigned> SO =
+        CGI.Operands.getSubOperandNumber(OpIdx);
+      const std::string &Name = CGI.Operands[SO.first].Name;
+
+      DEBUG(dbgs() << "Numbered operand mapping for " << Def.getName() << ": " <<
+                      Name << "(" << SO.first << ", " << SO.second << ") => " <<
+                      Vals[i].getName() << "\n");
+
+      std::string Decoder = "";
+      Record *TypeRecord = CGI.Operands[SO.first].Rec;
+
+      RecordVal *DecoderString = TypeRecord->getValue("DecoderMethod");
+      StringInit *String = DecoderString ?
+        dyn_cast<StringInit>(DecoderString->getValue()) : 0;
+      if (String && String->getValue() != "")
+        Decoder = String->getValue();
+
+      if (Decoder == "" &&
+          CGI.Operands[SO.first].MIOperandInfo &&
+          CGI.Operands[SO.first].MIOperandInfo->getNumArgs()) {
+        Init *Arg = CGI.Operands[SO.first].MIOperandInfo->
+                      getArg(SO.second);
+        if (TypedInit *TI = cast<TypedInit>(Arg)) {
+          RecordRecTy *Type = cast<RecordRecTy>(TI->getType());
+          TypeRecord = Type->getRecord();
+        }
+      }
+
+      bool isReg = false;
+      if (TypeRecord->isSubClassOf("RegisterOperand"))
+        TypeRecord = TypeRecord->getValueAsDef("RegClass");
+      if (TypeRecord->isSubClassOf("RegisterClass")) {
+        Decoder = "Decode" + TypeRecord->getName() + "RegisterClass";
+        isReg = true;
+      } else if (TypeRecord->isSubClassOf("PointerLikeRegClass")) {
+        Decoder = "DecodePointerLikeRegClass" +
+                  utostr(TypeRecord->getValueAsInt("RegClassKind"));
+        isReg = true;
+      }
+
+      DecoderString = TypeRecord->getValue("DecoderMethod");
+      String = DecoderString ?
+        dyn_cast<StringInit>(DecoderString->getValue()) : 0;
+      if (!isReg && String && String->getValue() != "")
+        Decoder = String->getValue();
+
+      OperandInfo OpInfo(Decoder);
+      OpInfo.addField(bitStart, bitWidth, 0);
+
+      NumberedInsnOperands[Name].push_back(OpInfo);
+
+      // FIXME: For complex operands with custom decoders we can't handle tied
+      // sub-operands automatically. Skip those here and assume that this is
+      // fixed up elsewhere.
+      if (CGI.Operands[SO.first].MIOperandInfo &&
+          CGI.Operands[SO.first].MIOperandInfo->getNumArgs() > 1 &&
+          String && String->getValue() != "")
+        NumberedInsnOperandsNoTie.insert(Name);
     }
   }
 
   // For each operand, see if we can figure out where it is encoded.
   for (std::vector<std::pair<Init*, std::string> >::const_iterator
        NI = InOutOperands.begin(), NE = InOutOperands.end(); NI != NE; ++NI) {
+    if (!NumberedInsnOperands[NI->second].empty()) {
+      InsnOperands.insert(InsnOperands.end(),
+                          NumberedInsnOperands[NI->second].begin(),
+                          NumberedInsnOperands[NI->second].end());
+      continue;
+    } else if (!NumberedInsnOperands[TiedNames[NI->second]].empty()) {
+      if (!NumberedInsnOperandsNoTie.count(TiedNames[NI->second])) {
+        // Figure out to which (sub)operand we're tied.
+        unsigned i = CGI.Operands.getOperandNamed(TiedNames[NI->second]);
+        int tiedTo = CGI.Operands[i].getTiedRegister();
+        if (tiedTo == -1) {
+          i = CGI.Operands.getOperandNamed(NI->second);
+          tiedTo = CGI.Operands[i].getTiedRegister();
+        }
+
+        if (tiedTo != -1) {
+          std::pair<unsigned, unsigned> SO =
+            CGI.Operands.getSubOperandNumber(tiedTo);
+
+          InsnOperands.push_back(NumberedInsnOperands[TiedNames[NI->second]]
+                                   [SO.second]);
+        }
+      }
+      continue;
+    }
+
     std::string Decoder = "";
 
     // At this point, we can locate the field, but we need to know how to
@@ -2037,7 +2189,7 @@ void FixedLenDecoderEmitter::run(raw_ost
     std::string DecoderNamespace = Def->getValueAsString("DecoderNamespace");
 
     if (Size) {
-      if (populateInstruction(*Inst, i, Operands)) {
+      if (populateInstruction(Target, *Inst, i, Operands)) {
         OpcMap[std::make_pair(DecoderNamespace, Size)].push_back(i);
       }
     }





More information about the llvm-commits mailing list