TableGen: Streamline the semantics of NAME

[nhaehnle]

Summary

The new rules are straightforward. The main rules to keep in mind
are:

1. NAME is an implicit template argument of class and multiclass, and will be substituted by the name of the instantiating def/defm.

2. The name of a def/defm in a multiclass must contain a reference to NAME. If such a reference is not present, it is automatically prepended.

And for some additional subtleties, consider these:

3. defm with no name generates a unique name but has no special behavior otherwise.

4. def with no name generates an anonymous record, whose name is unique but undefined. In particular, the name won't contain a reference to NAME.

Keeping rules 1&2 in mind should allow a predictable behavior of
name resolution that is simple to follow.

The old "rules" were rather surprising: sometimes (but not always),
NAME would correspond to the name of the toplevel defm. They were
also plain bonkers in the edge cases, as the old version of the
name-resolution-consistency.td test case shows.

Having NAME correspond to the name of the toplevel defm introduces
"spooky action at a distance" and breaks composability:
refactoring the upper layers of a hierarchy of nested multiclass
instantiations can cause unexpected breakage by changing the value
of NAME at a lower level of the hierarchy. The new rules don't
suffer from this problem.

Some existing .td files have to be adjusted because they ended up
depending on the undocumented quirks of the old implementation.

Change-Id: I694095231565b30f563e6fd0417b41ee01a12589

[simon_tatham]

docs/TableGen/LangRef.rst

151

I haven't fully got my head round all of this, but this new description of NAME doesn't quite seem to match the text in index.rst:

Each def record has a special entry called "NAME". This is the name of the record ("ADD32rr" above). In the general case def names can be formed from various kinds of string processing expressions and NAME resolves to the final value obtained after resolving all of those expressions. The user may refer to NAME anywhere she desires to use the ultimate name of the def.

The general intention seems to be the same (if you refer to NAME in some expression in a class declaration, then each def that inherits that expression gets it evaluated in terms of that def's own name). But the old text says that NAME is also a property of the final def, whereas the new text seems to suggest that it's a more transient entity that never _really_ exists in the def but is just something classes can refer to while everything's being set up.

In my JSON backend (D46054) I manually filled in the NAME field in the JSON dict for every output def, because that text in index.rst said it should be there. But if that's not a reflection of reality any more, perhaps I should rename that field to !name to mark it as one of the extra pieces of JSON-level metadata?

[tra]

Perhaps unrelated documentation cleanup can be extracted into a separate patch.

docs/TableGen/LangRef.rst

245–249

[doc cleanup candidate] I don't think multiclass supports local variables.

foo.td:2:3: error: expected 'let', 'def', 'defm' or 'foreach' in multiclass body
  int Baz = Bar;
  ^

382–384

I'd add an example similar to def here as well.

include/llvm/TableGen/Record.h

1907

However fun puns are, I'd rather use a boring descriptive name. VarNameToTrack would make the purpose somewhat more obvious than Needle.

lib/TableGen/TGParser.cpp

249–250

Can this be folded into getNameInit() ?

286–300

This could use some comments explaining that this deals with expanding default template arguments.

296

Do I understand it correctly that this error will be produced if we can't get complete value for the default argument? If so, then we should probably change the message a bit. Saying "value not specified" for an implicit argument which may intentionally be left unspecified is somewhat confusing. Perhaps something along the lines of "default argument X has unspecified value" would work better?

lib/Target/AArch64/AArch64InstrFormats.td

8648

Given that it's an implicit argument of class/multiclass, it should be available here, so there must be another reason for the change.

IIUIC, this is needed because previously NAME would often be resolved to be the name of the top-level def/defm and this is what this code was relying on. Now that NAME is resolved to complete chain of templates from the top level to this multiclass, this no longer works. Is that so?

If this is indeed the case, I'd rename Name->BaseName to make it a tiny bit easier to understand.
I'd also add more details to the patch description as it would help explaining what specifically is wrong with using NAME this way.

[nhaehnle]

Address review comments.

[nhaehnle]

Thank you for the review!

docs/TableGen/LangRef.rst

151

Right. I don't have a very strong opinion on whether NAME should really occur in the final defs, but they currently don't occur consistently and nobody actually uses it, so I lean slightly against having it (which is reflected in the patch). Note that you can always get the name of a record in TableGen by using !cast<string>(...), which is a nice mirror construct of the more common !cast<ClassName>(some_string).

I'm updating the docs accordingly. I think your suggestion for the JSON dict makes sense as well.

245–249

Done as a separate patch.

382–384

Done.

include/llvm/TableGen/Record.h

1907

Done.

lib/TableGen/TGParser.cpp

249–250

Maybe I misunderstand you, but CurRec->getNameInit() returns the name of the class here, while what we want is "NAME" but with a scoping prefix because we treat it like a template arg. So if the class name is "Foo", the string we're looking for is "Foo:NAME".

I personally find the whole scoping thing is a bit awkward, but it works fine from what I can tell.

286–300

Done?

296

I'm not sure how I feel about this. The common case in which you'd run into this error is if you forgot a template argument, and for that case, the error message as it currently stands seems more likely to put you on the right track.

The deeper issue here is that we allow ? as an explicit template argument (and I've actually used that, although I could have worked around it), but we don't distinguish between a default template argument that was set to ? (which won't work) and a template argument that simply has no default.

It makes sense to clean that up at some point, but the same issue also exists for class template arguments, and it doesn't really fit into this change.

lib/Target/AArch64/AArch64InstrFormats.td

8648

You understand correctly. I'm making the change for the AArch64 .td file.

I'm sticking with Name in X86InstrAVX512.td, because that file already has precedent for using "Name" for the same purpose (in multiclass avx512_int_broadcastbw_reg).

[tra]

lib/TableGen/TGParser.cpp

249–250

My thinking was that all the information necessary to derive the name is contained in CurRec and so could be done in a CurRec's method (though maybe getNameInit() may not be the right one). Given that there's only one place where we do this, I'm OK to keep it as is.

296

Ah, we're actually dealing with two separate error conditions here. One would be "not enough template arguments" and the other is "we have to provide some default arguments, but can't get the values." OK. I've missed the former.

If there's no easy way to separate those the current error message should be OK to cover both cases.

lib/Target/AArch64/AArch64InstrFormats.td

8648

OK.

9418

defm "" pattern is something we should probably highlight in the docs (as examples for def/defm in a multiclass?) because it's somewhat unintuitive. We need to point that defm w/o a name will produce a unique name, while "" is effectively a synonym for NAME (IIUIC, due to rule 2).

[nhaehnle]

lib/Target/AArch64/AArch64InstrFormats.td

9418

I had already added a mention of this in the LangRef, but I've made it more explicit with an example.

docs/TableGen/LangRef.rst

.. productionlist::
    Object: `Class` | `Def` | `Defm` | `Defset` | `Let` | `MultiClass` |
            `Foreach`
 
 ``class``\es
 ------------
 
 .. productionlist::
    Class: "class" `TokIdentifier` [`TemplateArgList`] `ObjectBody`
+   TemplateArgList: "<" `Declaration` ("," `Declaration`)* ">"
 
 A ``class`` declaration creates a record which other records can inherit
 from. A class can be parametrized by a list of "template arguments", whose
 values can be used in the class body.
 
 A given class can only be defined once. A ``class`` declaration is
 considered to define the class if any of the following is true:
 
 .. break ObjectBody into its consituents so that they are present here?
 
 #. The :token:`TemplateArgList` is present.
 #. The :token:`Body` in the :token:`ObjectBody` is present and is not empty.
 #. The :token:`BaseClassList` in the :token:`ObjectBody` is present.
 
 You can declare an empty class by giving and empty :token:`TemplateArgList`
 and an empty :token:`ObjectBody`. This can serve as a restricted form of
 forward declaration: note that records deriving from the forward-declared
 class will inherit no fields from it since the record expansion is done
 when the record is parsed.
 
-.. productionlist::
-   TemplateArgList: "<" `Declaration` ("," `Declaration`)* ">"
+Every class has an implicit template argument called ``NAME``, which is set
+to the name of the instantiating ``def`` or ``defm``. The result is undefined
+if the class is instantiated by an anonymous record.
 
 Declarations
 ------------
 
 .. Omitting mention of arcane "field" prefix to discourage its use.
 
 The declaration syntax is pretty much what you would expect as a C++
 programmer.

 
 * value local to a ``multiclass``, such as the use of ``Bar`` in::
 
      multiclass Foo {
        int Bar = 5;
        int Baz = Bar;
      }
 
 * a template arg to a ``multiclass``, such as the use of ``Bar`` in::
 
      multiclass Foo<int Bar> {
        int Baz = Bar;
      }
 
+* the implicit template argument ``NAME`` in a ``class`` or ``multiclass``
+
 .. productionlist::
    SimpleValue: `TokInteger`
 
 This represents the numeric value of the integer.
 
 .. productionlist::
    SimpleValue: `TokString`+
 
.. productionlist::
    BodyItem: `Declaration` ";"
            :| "let" `TokIdentifier` [ "{" `RangeList` "}" ] "=" `Value` ";"
 
 The ``let`` form allows overriding the value of an inherited field.
 
 ``def``
 -------
 
-.. TODO::
-   There can be pastes in the names here, like ``#NAME#``. Look into that
-   and document it (it boils down to ParseIDValue with IDParseMode ==
-   ParseNameMode). ParseObjectName calls into the general ParseValue, with
-   the only different from "arbitrary expression parsing" being IDParseMode
-   == Mode.
-
 .. productionlist::
-   Def: "def" `TokIdentifier` `ObjectBody`
-
-Defines a record whose name is given by the :token:`TokIdentifier`. The
-fields of the record are inherited from the base classes and defined in the
-body.
+   Def: "def" [`Value`] `ObjectBody`
+
+Defines a record whose name is given by the optional :token:`Value`. The value
+is parsed in a special mode where global identifiers (records and variables
+defined by ``defset``) are not recognized, and all unrecognized identifiers
+are interpreted as strings.
+
+If no name is given, the record is anonymous. The final name of anonymous
+records is undefined, but globally unique.
 
 Special handling occurs if this ``def`` appears inside a ``multiclass`` or
 a ``foreach``.
 
+When a non-anonymous record is defined in a multiclass and the given name
+does not contain a reference to the implicit template argument ``NAME``, such
+a reference will automatically be prepended. That is, the following are
+equivalent inside a multiclass::
+
+    def Foo;
+    def NAME#Foo;
+
 ``defm``
 --------
 
 .. productionlist::
-   Defm: "defm" [`TokIdentifier`] ":" `BaseClassListNE` ";"
+   Defm: "defm" [`Value`] ":" `BaseClassListNE` ";"
+
+The :token:`BaseClassList` is a list of at least one ``multiclass`` and any
+number of ``class``'s. The ``multiclass``'s must occur before any ``class``'s.
+
+Instantiates all records defined in all given ``multiclass``'s and adds the
+given ``class``'s as superclasses.
+
+The name is parsed in the same special mode used by ``def``. If the name is
+missing, a globally unique string is used instead (but instantiated records
+are not considered to be anonymous, unless they were originally defined by an
+anonymous ``def``).
+
+When ``defm`` occurs in a multiclass and its name does not contain a reference
+to the implicit template argument ``NAME``, such a reference will automatically
+be prepended. That is, the following are equivalent inside a multiclass::
 
-Note that in the :token:`BaseClassList`, all of the ``multiclass``'s must
-precede any ``class``'s that appear.
+    defm Foo : SomeMultiClass<...>;
+    defm NAME#Foo : SomeMultiClass<...>;
 
 ``defset``
 ----------
 .. productionlist::
    Defset: "defset" `Type` `TokIdentifier` "=" "{" `Object`* "}"
 
 All records defined inside the braces via ``def`` and ``defm`` are collected
 in a globally accessible list of the given name (in addition to being added
 to the global collection of records as usual). Anonymous records created inside

docs/TableGen/index.rst

def ADD32rr  : I<0x01, MRMDestReg, (outs GR32:$dst),
                    [(set GR32:$dst, (add GR32:$src1, GR32:$src2))]>;
 
 This definition makes use of the custom class ``I`` (extended from the custom
 class ``X86Inst``), which is defined in the X86-specific TableGen file, to
 factor out the common features that instructions of its class share.  A key
 feature of TableGen is that it allows the end-user to define the abstractions
 they prefer to use when describing their information.
 
-Each ``def`` record has a special entry called "NAME".  This is the name of the
-record ("``ADD32rr``" above).  In the general case ``def`` names can be formed
-from various kinds of string processing expressions and ``NAME`` resolves to the
-final value obtained after resolving all of those expressions.  The user may
-refer to ``NAME`` anywhere she desires to use the ultimate name of the ``def``.
-``NAME`` should not be defined anywhere else in user code to avoid conflicts.
-
 Syntax
 ======
 
 TableGen has a syntax that is loosely based on C++ templates, with built-in
 types and specification. In addition, TableGen's syntax introduces some
 automation concepts like multiclass, foreach, let, etc.
 
 Basic concepts

include/llvm/TableGen/Record.h

class Record {
   RecordKeeper &TrackedRecords;
 
   DefInit *TheInit = nullptr;
 
   // Unique record ID.
   unsigned ID;
 
   bool IsAnonymous;
+  bool IsClass;
 
-  void init();
   void checkName();
 
 public:
   // Constructs a record.
   explicit Record(Init *N, ArrayRef<SMLoc> locs, RecordKeeper &records,
-                  bool Anonymous = false) :
-    Name(N), Locs(locs.begin(), locs.end()), TrackedRecords(records),
-    ID(LastID++), IsAnonymous(Anonymous) {
-    init();
+                  bool Anonymous = false, bool Class = false)
+    : Name(N), Locs(locs.begin(), locs.end()), TrackedRecords(records),
+      ID(LastID++), IsAnonymous(Anonymous), IsClass(Class) {
+    checkName();
   }
 
-  explicit Record(StringRef N, ArrayRef<SMLoc> locs, RecordKeeper &records)
-      : Record(StringInit::get(N), locs, records) {}
+  explicit Record(StringRef N, ArrayRef<SMLoc> locs, RecordKeeper &records,
+                  bool Class = false)
+      : Record(StringInit::get(N), locs, records, false, Class) {}
 
   // When copy-constructing a Record, we must still guarantee a globally unique
   // ID number.  Don't copy TheInit either since it's owned by the original
   // record. All other fields can be copied normally.
-  Record(const Record &O) :
-    Name(O.Name), Locs(O.Locs), TemplateArgs(O.TemplateArgs),
+  Record(const Record &O)
+    : Name(O.Name), Locs(O.Locs), TemplateArgs(O.TemplateArgs),
       Values(O.Values), SuperClasses(O.SuperClasses),
       TrackedRecords(O.TrackedRecords), ID(LastID++),
-    IsAnonymous(O.IsAnonymous) { }
+      IsAnonymous(O.IsAnonymous), IsClass(O.IsClass) { }
 
   static unsigned getNewUID() { return LastID++; }
 
   unsigned getID() const { return ID; }
 
   StringRef getName() const { return cast<StringInit>(Name)->getValue(); }
 
   Init *getNameInit() const {
public:
   void appendLoc(SMLoc Loc) { Locs.push_back(Loc); }
 
   // Make the type that this record should have based on its superclasses.
   RecordRecTy *getType();
 
   /// get the corresponding DefInit.
   DefInit *getDefInit();
 
+  bool isClass() const { return IsClass; }
+
   ArrayRef<Init *> getTemplateArgs() const {
     return TemplateArgs;
   }
 
   ArrayRef<RecordVal> getValues() const { return Values; }
 
   ArrayRef<std::pair<Record *, SMRange>>  getSuperClasses() const {
     return SuperClasses;
public:
   void addTemplateArg(Init *Name) {
     assert(!isTemplateArg(Name) && "Template arg already defined!");
     TemplateArgs.push_back(Name);
   }
 
   void addValue(const RecordVal &RV) {
     assert(getValue(RV.getNameInit()) == nullptr && "Value already added!");
     Values.push_back(RV);
-    if (Values.size() > 1)
-      // Keep NAME at the end of the list.  It makes record dumps a
-      // bit prettier and allows TableGen tests to be written more
-      // naturally.  Tests can use CHECK-NEXT to look for Record
-      // fields they expect to see after a def.  They can't do that if
-      // NAME is the first Record field.
-      std::swap(Values[Values.size() - 2], Values[Values.size() - 1]);
   }
 
   void removeValue(Init *Name) {
     for (unsigned i = 0, e = Values.size(); i != e; ++i)
       if (Values[i].getNameInit() == Name) {
         Values.erase(Values.begin()+i);
         return;
       }
public:
   explicit TrackUnresolvedResolver(Resolver *R = nullptr)
       : Resolver(R ? R->getCurrentRecord() : nullptr), R(R) {}
 
   bool foundUnresolved() const { return FoundUnresolved; }
 
   Init *resolve(Init *VarName) override;
 };
 
+/// Do not resolve anything, but keep track of whether a given variable was
+/// referenced.
+class HasReferenceResolver final : public Resolver {
+  Init *VarNameToTrack;
+  bool Found = false;
+
+public:
+  explicit HasReferenceResolver(Init *VarNameToTrack)
+      : Resolver(nullptr), VarNameToTrack(VarNameToTrack) {}
+
+  bool found() const { return Found; }
+
+  Init *resolve(Init *VarName) override;
+};
+
 } // end namespace llvm
 
 #endif // LLVM_TABLEGEN_RECORD_H

lib/TableGen/Record.cpp

DefInit *VarDefInit::instantiate() {
   if (!Def) {
     RecordKeeper &Records = Class->getRecords();
     auto NewRecOwner = make_unique<Record>(Records.getNewAnonymousName(),
                                            Class->getLoc(), Records,
                                            /*IsAnonymous=*/true);
     Record *NewRec = NewRecOwner.get();
 
     // Copy values from class to instance
-    for (const RecordVal &Val : Class->getValues()) {
-      if (Val.getName() != "NAME")
-        NewRec->addValue(Val);
-    }
+    for (const RecordVal &Val : Class->getValues())
 
     // Substitute and resolve template arguments
     ArrayRef<Init *> TArgs = Class->getTemplateArgs();
     MapResolver R(NewRec);
 
     for (unsigned i = 0, e = TArgs.size(); i != e; ++i) {
       if (i < args_size())
         R.set(TArgs[i], getArg(i));
void RecordVal::print(raw_ostream &OS, bool PrintSem) const {
   if (getValue())
     OS << " = " << *getValue();
 
   if (PrintSem) OS << ";\n";
 }
 
 unsigned Record::LastID = 0;
 
-void Record::init() {
-  checkName();
-
-  // Every record potentially has a def at the top.  This value is
-  // replaced with the top-level def name at instantiation time.
-  addValue(RecordVal(StringInit::get("NAME"), StringRecTy::get(), false));
-}
-
 void Record::checkName() {
   // Ensure the record name has string type.
   const TypedInit *TypedName = cast<const TypedInit>(Name);
   if (!isa<StringRecTy>(TypedName->getType()))
     PrintFatalError(getLoc(), Twine("Record name '") + Name->getAsString() +
                                   "' is not a string!");
 }
 
if (I && !FoundUnresolved) {
       FoundUnresolved |= Sub.FoundUnresolved;
     }
   }
 
   if (!I)
     FoundUnresolved = true;
   return I;
 }
+
+Init *HasReferenceResolver::resolve(Init *VarName)
+{
+  if (VarName == VarNameToTrack)
+    Found = true;
+  return nullptr;
+}

lib/TableGen/TGParser.h

struct IterRecord {
     Init *IterValue;
     IterRecord(VarInit *Var, Init *Val) : IterVar(Var), IterValue(Val) {}
   };
 
   // IterSet: The set of all iterator values at some point in the
   // iteration space.
   typedef std::vector<IterRecord> IterSet;
 
-  bool addDefOne(std::unique_ptr<Record> Rec, Init *DefmName,
-                 IterSet &IterVals);
-  bool addDefForeach(Record *Rec, Init *DefmName, IterSet &IterVals);
-  bool addDef(std::unique_ptr<Record> Rec, Init *DefmName);
+  bool addDefOne(std::unique_ptr<Record> Rec, IterSet &IterVals);
+  bool addDefForeach(Record *Rec, IterSet &IterVals);
+  bool addDef(std::unique_ptr<Record> Rec);
 
 private:  // Parser methods.
   bool ParseObjectList(MultiClass *MC = nullptr);
   bool ParseObject(MultiClass *MC);
   bool ParseClass();
   bool ParseMultiClass();
   bool ParseDefm(MultiClass *CurMultiClass);
   bool ParseDef(MultiClass *CurMultiClass);

lib/TableGen/TGParser.cpp

if (Init *V = RV.getValue()) {
                    "' in '" + R.getNameInitAsString() +
                    "' could not be fully resolved: " +
                    RV.getValue()->getAsString());
       }
     }
   }
 }
 
+/// Return the qualified version of the implicit 'NAME' template argument.
+static Init *QualifiedNameOfImplicitName(Record &Rec,
+                                         MultiClass *MC = nullptr) {
+  return QualifyName(Rec, MC, StringInit::get("NAME"), MC ? "::" : ":");
+}
+
+static Init *QualifiedNameOfImplicitName(MultiClass *MC) {
+  return QualifiedNameOfImplicitName(MC->Rec, MC);
+}
+
 bool TGParser::AddValue(Record *CurRec, SMLoc Loc, const RecordVal &RV) {
   if (!CurRec)
     CurRec = &CurMultiClass->Rec;
 
   if (RecordVal *ERV = CurRec->getValue(RV.getNameInit())) {
     // The value already exists in the class, treat this as a set.
     if (ERV->setValue(RV.getValue()))
       return Error(Loc, "New definition of '" + RV.getName() + "' of type '" +
if (i < SubClass.TemplateArgs.size()) {
                    ") of subclass '" + SC->getNameInitAsString() + "'!");
     }
 
     R.set(TArgs[i], CurRec->getValue(TArgs[i])->getValue());
 
     CurRec->removeValue(TArgs[i]);
   }
 
+  Init *Name;
+  if (CurRec->isClass())
+    Name =
+        VarInit::get(QualifiedNameOfImplicitName(*CurRec), StringRecTy::get());
+  else
+    Name = CurRec->getNameInit();
+  R.set(QualifiedNameOfImplicitName(*SC), Name);
+
   CurRec->resolveReferences(R);
 
   // Since everything went well, we can now set the "superclass" list for the
   // current record.
   ArrayRef<std::pair<Record *, SMRange>> SCs = SC->getSuperClasses();
   for (const auto &SCPair : SCs) {
     if (CurRec->isSubClassOf(SCPair.first))
       return Error(SubClass.RefRange.Start,

 }
 
 /// AddSubMultiClass - Add SubMultiClass as a subclass to
 /// CurMC, resolving its template args as SubMultiClass's
 /// template arguments.
 bool TGParser::AddSubMultiClass(MultiClass *CurMC,
                                 SubMultiClassReference &SubMultiClass) {
   MultiClass *SMC = SubMultiClass.MC;
-  Record *CurRec = &CurMC->Rec;
-
-  // Add all of the values in the subclass into the current class.
-  for (const auto &SMCVal : SMC->Rec.getValues())
-    if (AddValue(CurRec, SubMultiClass.RefRange.Start, SMCVal))
-      return true;
-
-  unsigned newDefStart = CurMC->DefPrototypes.size();
-
-  // Add all of the defs in the subclass into the current multiclass.
-  for (const std::unique_ptr<Record> &R : SMC->DefPrototypes) {
-    // Clone the def and add it to the current multiclass
-    auto NewDef = make_unique<Record>(*R);
-
-    // Add all of the values in the superclass into the current def.
-    for (const auto &MCVal : CurRec->getValues())
-      if (AddValue(NewDef.get(), SubMultiClass.RefRange.Start, MCVal))
-        return true;
-
-    CurMC->DefPrototypes.push_back(std::move(NewDef));
-  }
 
   ArrayRef<Init *> SMCTArgs = SMC->Rec.getTemplateArgs();
-
-  // Ensure that an appropriate number of template arguments are
-  // specified.
   if (SMCTArgs.size() < SubMultiClass.TemplateArgs.size())
     return Error(SubMultiClass.RefRange.Start,
                  "More template args specified than expected");
 
-  // Loop over all of the template arguments, setting them to the specified
-  // value or leaving them as the default if necessary.
-  MapResolver CurRecResolver(CurRec);
-
+  // Prepare the mapping of template argument name to value, filling in default
+  // values if necessary.
+  SmallVector<std::pair<Init *, Init *>, 8> TemplateArgs;
   for (unsigned i = 0, e = SMCTArgs.size(); i != e; ++i) {
     if (i < SubMultiClass.TemplateArgs.size()) {
-      // If a value is specified for this template arg, set it in the
-      // superclass now.
-      if (SetValue(CurRec, SubMultiClass.RefRange.Start, SMCTArgs[i],
-                   None, SubMultiClass.TemplateArgs[i]))
-        return true;
-
-      // If a value is specified for this template arg, set it in the
-      // new defs now.
-      for (const auto &Def :
-             makeArrayRef(CurMC->DefPrototypes).slice(newDefStart)) {
-        if (SetValue(Def.get(), SubMultiClass.RefRange.Start, SMCTArgs[i],
-                     None, SubMultiClass.TemplateArgs[i]))
-          return true;
-      }
-    } else if (!CurRec->getValue(SMCTArgs[i])->getValue()->isComplete()) {
+      TemplateArgs.emplace_back(SMCTArgs[i], SubMultiClass.TemplateArgs[i]);
+    } else {
+      Init *Default = SMC->Rec.getValue(SMCTArgs[i])->getValue();
+      if (!Default->isComplete()) {
         return Error(SubMultiClass.RefRange.Start,
-                   "Value not specified for template argument #" +
-                   Twine(i) + " (" + SMCTArgs[i]->getAsUnquotedString() +
-                   ") of subclass '" + SMC->Rec.getNameInitAsString() + "'!");
+                     "value not specified for template argument #" + Twine(i) +
+                         " (" + SMCTArgs[i]->getAsUnquotedString() +
+                         ") of multiclass '" + SMC->Rec.getNameInitAsString() +
+                         "'");
+      }
+      TemplateArgs.emplace_back(SMCTArgs[i], Default);
     }
-
-    CurRecResolver.set(SMCTArgs[i], CurRec->getValue(SMCTArgs[i])->getValue());
-
-    CurRec->removeValue(SMCTArgs[i]);
   }
 
-  CurRec->resolveReferences(CurRecResolver);
+  TemplateArgs.emplace_back(
+      QualifiedNameOfImplicitName(SMC),
+      VarInit::get(QualifiedNameOfImplicitName(CurMC), StringRecTy::get()));
 
-  for (const auto &Def :
-       makeArrayRef(CurMC->DefPrototypes).slice(newDefStart)) {
-    MapResolver R(Def.get());
+  // Add all of the defs in the subclass into the current multiclass.
+  for (const std::unique_ptr<Record> &Rec : SMC->DefPrototypes) {
+    auto NewDef = make_unique<Record>(*Rec);
 
-    for (Init *SMCTArg : SMCTArgs) {
-      R.set(SMCTArg, Def->getValue(SMCTArg)->getValue());
-      Def->removeValue(SMCTArg);
-    }
+    MapResolver R(NewDef.get());
+    for (const auto &TArg : TemplateArgs)
+      R.set(TArg.first, TArg.second);
+    NewDef->resolveReferences(R);
 
-    Def->resolveReferences(R);
+    CurMC->DefPrototypes.push_back(std::move(NewDef));
   }
 
   return false;
 }
 
 /// Add a record that results from 'def' or 'defm', after template arguments
 /// and the external let stack have been resolved.
 ///
 /// Apply foreach loops, resolve internal variable references, and add to the
 /// current multi class or the global record keeper as appropriate.
-bool TGParser::addDef(std::unique_ptr<Record> Rec, Init *DefmName) {
+bool TGParser::addDef(std::unique_ptr<Record> Rec) {
   IterSet IterVals;
 
   if (Loops.empty())
-    return addDefOne(std::move(Rec), DefmName, IterVals);
+    return addDefOne(std::move(Rec), IterVals);
 
-  return addDefForeach(Rec.get(), DefmName, IterVals);
+  return addDefForeach(Rec.get(), IterVals);
 }
 
 /// Recursive helper function for addDef/addDefOne to resolve references to
 /// foreach variables.
-bool TGParser::addDefForeach(Record *Rec, Init *DefmName, IterSet &IterVals) {
+bool TGParser::addDefForeach(Record *Rec, IterSet &IterVals) {
   if (IterVals.size() != Loops.size()) {
     assert(IterVals.size() < Loops.size());
     ForeachLoop &CurLoop = Loops[IterVals.size()];
     ListInit *List = CurLoop.ListValue;
 
     // Process each value.
     for (unsigned i = 0; i < List->size(); ++i) {
       IterVals.push_back(IterRecord(CurLoop.IterVar, List->getElement(i)));
-      if (addDefForeach(Rec, DefmName, IterVals))
+      if (addDefForeach(Rec, IterVals))
         return true;
       IterVals.pop_back();
     }
     return false;
   }
 
   // This is the bottom of the recursion. We have all of the iterator values
   // for this point in the iteration space.  Instantiate a new record to
   // reflect this combination of values.
   auto IterRec = make_unique<Record>(*Rec);
-  return addDefOne(std::move(IterRec), DefmName, IterVals);
+  return addDefOne(std::move(IterRec), IterVals);
 }
 
 /// After resolving foreach loops, add the record as a prototype to the
 /// current multiclass, or resolve fully and add to the record keeper.
-bool TGParser::addDefOne(std::unique_ptr<Record> Rec, Init *DefmName,
-                         IterSet &IterVals) {
+bool TGParser::addDefOne(std::unique_ptr<Record> Rec, IterSet &IterVals) {
   MapResolver R(Rec.get());
 
   for (IterRecord &IR : IterVals)
     R.set(IR.IterVar->getNameInit(), IR.IterValue);
 
   Rec->resolveReferences(R);
 
   if (CurMultiClass) {
+    if (!Rec->isAnonymous()) {
       for (const auto &Proto : CurMultiClass->DefPrototypes) {
         if (Proto->getNameInit() == Rec->getNameInit()) {
-        if (!Rec->isAnonymous()) {
           PrintError(Rec->getLoc(),
                     Twine("def '") + Rec->getNameInitAsString() +
                         "' already defined in this multiclass!");
           PrintNote(Proto->getLoc(), "location of previous definition");
           return true;
         }
-        Rec->setName(Records.getNewAnonymousName());
-        break;
       }
     }
     CurMultiClass->DefPrototypes.emplace_back(std::move(Rec));
     return false;
   }
 
-  // Name construction is an incoherent mess. Unfortunately, existing .td
-  // files rely on pretty much all the quirks and implementation details of
-  // this.
-  if (DefmName) {
-    MapResolver R(Rec.get());
-    R.set(StringInit::get("NAME"), DefmName);
-    Rec->resolveReferences(R);
-  }
-
   if (Record *Prev = Records.getDef(Rec->getNameInitAsString())) {
     if (!Rec->isAnonymous()) {
       PrintError(Rec->getLoc(),
                  "def already exists: " + Rec->getNameInitAsString());
       PrintNote(Prev->getLoc(), "location of previous definition");
       return true;
     }
     Rec->setName(Records.getNewAnonymousName());
Init *TGParser::ParseObjectName(MultiClass *CurMultiClass) {
   default:
     break;
   }
 
   Record *CurRec = nullptr;
   if (CurMultiClass)
     CurRec = &CurMultiClass->Rec;
 
-  return ParseValue(CurRec, StringRecTy::get(), ParseNameMode);
+  Init *Name = ParseValue(CurRec, StringRecTy::get(), ParseNameMode);
+  if (!Name)
+    return nullptr;
+
+  if (CurMultiClass) {
+    Init *NameStr = QualifiedNameOfImplicitName(CurMultiClass);
+    HasReferenceResolver R(NameStr);
+    Name->resolveReferences(R);
+    if (!R.found())
+      Name = BinOpInit::getStrConcat(VarInit::get(NameStr, StringRecTy::get()),
+                                     Name);
+  }
+
+  return Name;
 }
 
 /// ParseClassID - Parse and resolve a reference to a class name.  This returns
 /// null on error.
 ///
 ///    ClassID ::= ID
 ///
 Record *TGParser::ParseClassID() {

 
 /// ParseIDValue - This is just like ParseIDValue above, but it assumes the ID
 /// has already been read.
 Init *TGParser::ParseIDValue(Record *CurRec, StringInit *Name, SMLoc NameLoc,
                              IDParseMode Mode) {
   if (CurRec) {
     if (const RecordVal *RV = CurRec->getValue(Name))
       return VarInit::get(Name, RV->getType());
-
-    Init *TemplateArgName = QualifyName(*CurRec, CurMultiClass, Name, ":");
-
-    if (CurMultiClass)
-      TemplateArgName = QualifyName(CurMultiClass->Rec, CurMultiClass, Name,
-                                    "::");
-
-    if (CurRec->isTemplateArg(TemplateArgName)) {
-      const RecordVal *RV = CurRec->getValue(TemplateArgName);
-      assert(RV && "Template arg doesn't exist??");
-      return VarInit::get(TemplateArgName, RV->getType());
-    }
   }
 
+  if ((CurRec && CurRec->isClass()) || CurMultiClass) {
+    Init *TemplateArgName;
     if (CurMultiClass) {
-    if (Name->getValue() == "NAME")
-      return VarInit::get(Name, StringRecTy::get());
-
-    Init *MCName = QualifyName(CurMultiClass->Rec, CurMultiClass, Name, "::");
-
-    if (CurMultiClass->Rec.isTemplateArg(MCName)) {
-      const RecordVal *RV = CurMultiClass->Rec.getValue(MCName);
+      TemplateArgName =
+          QualifyName(CurMultiClass->Rec, CurMultiClass, Name, "::");
+    } else
+      TemplateArgName = QualifyName(*CurRec, CurMultiClass, Name, ":");
+
+    Record *TemplateRec = CurMultiClass ? &CurMultiClass->Rec : CurRec;
+    if (TemplateRec->isTemplateArg(TemplateArgName)) {
+      const RecordVal *RV = TemplateRec->getValue(TemplateArgName);
       assert(RV && "Template arg doesn't exist??");
-      return VarInit::get(MCName, RV->getType());
+      return VarInit::get(TemplateArgName, RV->getType());
+    } else if (Name->getValue() == "NAME") {
+      return VarInit::get(TemplateArgName, StringRecTy::get());
     }
   }
 
   // If this is in a foreach loop, make sure it's not a loop iterator
   for (const auto &L : Loops) {
     VarInit *IterVar = dyn_cast<VarInit>(L.IterVar);
     if (IterVar && IterVar->getNameInit() == Name)
       return IterVar;
   }
 
   if (Mode == ParseNameMode)
     return Name;
 
   if (Init *I = Records.getGlobal(Name->getValue()))
     return I;
 
   // Allow self-references of concrete defs, but delay the lookup so that we
   // get the correct type.
-  if (CurRec && !CurMultiClass && CurRec->getNameInit() == Name)
+  if (CurRec && !CurRec->isClass() && !CurMultiClass &&
+      CurRec->getNameInit() == Name)
     return UnOpInit::get(UnOpInit::CAST, Name, CurRec->getType());
 
-  if (Mode == ParseValueMode) {
-    Error(NameLoc, "Variable not defined: '" + Name->getValue() + "'");
-    return nullptr;
-  }
 
-  return Name;
-}
-
 /// ParseOperation - Parse an operator.  This returns null on error.
 ///
 /// Operation ::= XOperator ['<' Type '>'] '(' Args ')'
 ///
 Init *TGParser::ParseOperation(Record *CurRec, RecTy *ItemType) {
   switch (Lex.getCode()) {
   default:
     TokError("unknown operation");
Init *TGParser::ParseDeclaration(Record *CurRec,
   RecTy *Type = ParseType();
   if (!Type) return nullptr;
 
   if (Lex.getCode() != tgtok::Id) {
     TokError("Expected identifier in declaration");
     return nullptr;
   }
 
+  std::string Str = Lex.getCurStrVal();
+  if (Str == "NAME") {
+    TokError("'" + Str + "' is a reserved variable name");
+    return nullptr;
+  }
+
   SMLoc IdLoc = Lex.getLoc();
-  Init *DeclName = StringInit::get(Lex.getCurStrVal());
+  Init *DeclName = StringInit::get(Str);
   Lex.Lex();
 
   if (ParsingTemplateArgs) {
     if (CurRec)
       DeclName = QualifyName(*CurRec, CurMultiClass, DeclName, ":");
     else
       assert(CurMultiClass);
     if (CurMultiClass)
if (isa<UnsetInit>(Name))
     CurRec = make_unique<Record>(Records.getNewAnonymousName(), DefLoc, Records,
                                  /*Anonymous=*/true);
   else
     CurRec = make_unique<Record>(Name, DefLoc, Records);
 
   if (ParseObjectBody(CurRec.get()))
     return true;
 
-  return addDef(std::move(CurRec), nullptr);
+  return addDef(std::move(CurRec));
 }
 
 /// ParseDefset - Parse a defset statement.
 ///
 ///   Defset ::= DEFSET Type Id '=' '{' ObjectList '}'
 ///
 bool TGParser::ParseDefset() {
   assert(Lex.getCode() == tgtok::Defset);
bool TGParser::ParseClass() {
   Lex.Lex();
 
   if (Lex.getCode() != tgtok::Id)
     return TokError("expected class name after 'class' keyword");
 
   Record *CurRec = Records.getClass(Lex.getCurStrVal());
   if (CurRec) {
     // If the body was previously defined, this is an error.
-    if (CurRec->getValues().size() > 1 ||  // Account for NAME.
+    if (!CurRec->getValues().empty() ||
         !CurRec->getSuperClasses().empty() ||
         !CurRec->getTemplateArgs().empty())
       return TokError("Class '" + CurRec->getNameInitAsString() +
                       "' already defined");
   } else {
     // If this is the first reference to this class, create and add it.
     auto NewRec =
-        llvm::make_unique<Record>(Lex.getCurStrVal(), Lex.getLoc(), Records);
+        llvm::make_unique<Record>(Lex.getCurStrVal(), Lex.getLoc(), Records,
+                                  /*Class=*/true);
     CurRec = NewRec.get();
     Records.addClass(std::move(NewRec));
   }
   Lex.Lex(); // eat the name.
 
   // If there are template args, parse them.
   if (Lex.getCode() == tgtok::less)
     if (ParseTemplateArgList(CurRec))
       return true;
 
-  // Finally, parse the object body.
   return ParseObjectBody(CurRec);
 }
 
 /// ParseLetList - Parse a non-empty list of assignment expressions into a list
 /// of LetRecords.
 ///
 ///   LetList ::= LetItem (',' LetItem)*
 ///   LetItem ::= ID OptionalRangeList '=' Value

 ///
 bool TGParser::ParseDefm(MultiClass *CurMultiClass) {
   assert(Lex.getCode() == tgtok::Defm && "Unexpected token!");
   Lex.Lex(); // eat the defm
 
   Init *DefmName = ParseObjectName(CurMultiClass);
   if (!DefmName)
     return true;
-  if (isa<UnsetInit>(DefmName))
+  if (isa<UnsetInit>(DefmName)) {
     DefmName = Records.getNewAnonymousName();
+    if (CurMultiClass)
+      DefmName = BinOpInit::getStrConcat(
+          VarInit::get(QualifiedNameOfImplicitName(CurMultiClass),
+                       StringRecTy::get()),
+          DefmName);
+  }
 
   if (Lex.getCode() != tgtok::colon)
     return TokError("expected ':' after defm identifier");
 
   // Keep track of the new generated record definitions.
   SmallVector<std::unique_ptr<Record>, 8> NewRecDefs;
 
   // This record also inherits from a regular class (non-multiclass)?
while (true) {
     ArrayRef<Init*> TemplateVals = Ref.TemplateArgs;
 
     // Verify that the correct number of template arguments were specified.
     ArrayRef<Init *> TArgs = MC->Rec.getTemplateArgs();
     if (TArgs.size() < TemplateVals.size())
       return Error(SubClassLoc,
                    "more template args specified than multiclass expects");
 
-    DenseMap<Init *, Init *> TemplateArgs;
+    SmallVector<std::pair<Init *, Init *>, 8> TemplateArgs;
     for (unsigned i = 0, e = TArgs.size(); i != e; ++i) {
       if (i < TemplateVals.size()) {
-        TemplateArgs.insert({TArgs[i], TemplateVals[i]});
+        TemplateArgs.emplace_back(TArgs[i], TemplateVals[i]);
       } else {
         Init *Default = MC->Rec.getValue(TArgs[i])->getValue();
         if (!Default->isComplete()) {
           return Error(SubClassLoc,
                        "value not specified for template argument #" +
                            Twine(i) + " (" + TArgs[i]->getAsUnquotedString() +
                            ") of multiclass '" + MC->Rec.getNameInitAsString() +
                            "'");
         }
-        TemplateArgs.insert({TArgs[i], Default});
+        TemplateArgs.emplace_back(TArgs[i], Default);
       }
     }
 
+    TemplateArgs.emplace_back(QualifiedNameOfImplicitName(MC), DefmName);
+
     // Loop over all the def's in the multiclass, instantiating each one.
     for (const std::unique_ptr<Record> &DefProto : MC->DefPrototypes) {
-      bool ResolveName = true;
       auto CurRec = make_unique<Record>(*DefProto);
       CurRec->appendLoc(SubClassLoc);
 
-      if (StringInit *NameString =
-              dyn_cast<StringInit>(CurRec->getNameInit())) {
-        // We have a fully expanded string so there are no operators to
-        // resolve.  We should concatenate the given prefix and name.
-        //
-        // TODO: This MUST happen before template argument resolution. This
-        //       does not make sense and should be changed, but at the time of
-        //       writing, there are existing .td files which rely on this
-        //       implementation detail. It's a bad idea and should be fixed.
-        //       See test/TableGen/name-resolution-consistency.td for some
-        //       examples.
-        CurRec->setName(BinOpInit::getStrConcat(DefmName, NameString));
-        ResolveName = false;
-      }
-
       MapResolver R(CurRec.get());
-
-      if (ResolveName) {
-        // If the proto's name wasn't resolved, we probably have a reference to
-        // NAME and need to replace it.
-        //
-        // TODO: Whether the name is resolved is basically determined by magic.
-        //       Unfortunately, existing .td files depend on it.
-        R.set(StringInit::get("NAME"), DefmName);
-      }
-
       for (const auto &TArg : TemplateArgs)
         R.set(TArg.first, TArg.second);
-
       CurRec->resolveReferences(R);
 
       NewRecDefs.emplace_back(std::move(CurRec));
     }
 
     if (Lex.getCode() != tgtok::comma) break;
     Lex.Lex(); // eat ','.
 
while (true) {
       SubClass = ParseSubClassReference(nullptr, false);
     }
   }
 
   for (auto &CurRec : NewRecDefs) {
     if (ApplyLetStack(CurRec.get()))
       return true;
 
-    addDef(std::move(CurRec), DefmName);
+    addDef(std::move(CurRec));
   }
 
   if (Lex.getCode() != tgtok::semi)
     return TokError("expected ';' at end of defm");
   Lex.Lex();
 
   return false;
 }

lib/Target/AArch64/AArch64InstrFormats.td

class BaseSIMDLdStPost<bit Q, bit L, bits<4> opcode, bits<2> size,
   let Inst{15-12} = opcode;
   let Inst{11-10} = size;
   let Inst{9-5} = Rn;
   let Inst{4-0} = Vt;
 }
 
 // The immediate form of AdvSIMD post-indexed addressing is encoded with
 // register post-index addressing from the zero register.
-multiclass SIMDLdStAliases<string asm, string layout, string Count,
+multiclass SIMDLdStAliases<string BaseName, string asm, string layout, string Count,
                            int Offset, int Size> {
   // E.g. "ld1 { v0.8b, v1.8b }, [x1], #16"
   //      "ld1\t$Vt, [$Rn], #16"
   // may get mapped to
   //      (LD1Twov8b_POST VecListTwo8b:$Vt, GPR64sp:$Rn, XZR)
   def : InstAlias<asm # "\t$Vt, [$Rn], #" # Offset,
-                  (!cast<Instruction>(NAME # Count # "v" # layout # "_POST")
+                  (!cast<Instruction>(BaseName # Count # "v" # layout # "_POST")
                       GPR64sp:$Rn,
                       !cast<RegisterOperand>("VecList" # Count # layout):$Vt,
                       XZR), 1>;
 
   // E.g. "ld1.8b { v0, v1 }, [x1], #16"
   //      "ld1.8b\t$Vt, [$Rn], #16"
   // may get mapped to
   //      (LD1Twov8b_POST VecListTwo64:$Vt, GPR64sp:$Rn, XZR)
   def : InstAlias<asm # "." # layout # "\t$Vt, [$Rn], #" # Offset,
-                  (!cast<Instruction>(NAME # Count # "v" # layout # "_POST")
+                  (!cast<Instruction>(BaseName # Count # "v" # layout # "_POST")
                       GPR64sp:$Rn,
                       !cast<RegisterOperand>("VecList" # Count # Size):$Vt,
                       XZR), 0>;
 
   // E.g. "ld1.8b { v0, v1 }, [x1]"
   //      "ld1\t$Vt, [$Rn]"
   // may get mapped to
   //      (LD1Twov8b VecListTwo64:$Vt, GPR64sp:$Rn)
   def : InstAlias<asm # "." # layout # "\t$Vt, [$Rn]",
-                  (!cast<Instruction>(NAME # Count # "v" # layout)
+                  (!cast<Instruction>(BaseName # Count # "v" # layout)
                       !cast<RegisterOperand>("VecList" # Count # Size):$Vt,
                       GPR64sp:$Rn), 0>;
 
   // E.g. "ld1.8b { v0, v1 }, [x1], x2"
   //      "ld1\t$Vt, [$Rn], $Xm"
   // may get mapped to
   //      (LD1Twov8b_POST VecListTwo64:$Vt, GPR64sp:$Rn, GPR64pi8:$Xm)
   def : InstAlias<asm # "." # layout # "\t$Vt, [$Rn], $Xm",
-                  (!cast<Instruction>(NAME # Count # "v" # layout # "_POST")
+                  (!cast<Instruction>(BaseName # Count # "v" # layout # "_POST")
                       GPR64sp:$Rn,
                       !cast<RegisterOperand>("VecList" # Count # Size):$Vt,
                       !cast<RegisterOperand>("GPR64pi" # Offset):$Xm), 0>;
 }
 
-multiclass BaseSIMDLdN<string Count, string asm, string veclist, int Offset128,
-                       int Offset64, bits<4> opcode> {
+multiclass BaseSIMDLdN<string BaseName, string Count, string asm, string veclist,
+                       int Offset128, int Offset64, bits<4> opcode> {
   let hasSideEffects = 0, mayLoad = 1, mayStore = 0 in {
     def v16b: BaseSIMDLdSt<1, 1, opcode, 0b00, asm,
                            (outs !cast<RegisterOperand>(veclist # "16b"):$Vt),
                            (ins GPR64sp:$Rn), []>;
     def v8h : BaseSIMDLdSt<1, 1, opcode, 0b01, asm,
                            (outs !cast<RegisterOperand>(veclist # "8h"):$Vt),
                            (ins GPR64sp:$Rn), []>;
     def v4s : BaseSIMDLdSt<1, 1, opcode, 0b10, asm,
def v4h_POST : BaseSIMDLdStPost<0, 1, opcode, 0b01, asm,
                             !cast<RegisterOperand>("GPR64pi" # Offset64):$Xm)>;
     def v2s_POST : BaseSIMDLdStPost<0, 1, opcode, 0b10, asm,
                        (outs GPR64sp:$wback,
                              !cast<RegisterOperand>(veclist # "2s"):$Vt),
                        (ins GPR64sp:$Rn,
                             !cast<RegisterOperand>("GPR64pi" # Offset64):$Xm)>;
   }
 
-  defm : SIMDLdStAliases<asm, "16b", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "8h", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "4s", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "2d", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "8b", Count, Offset64, 64>;
-  defm : SIMDLdStAliases<asm, "4h", Count, Offset64, 64>;
-  defm : SIMDLdStAliases<asm, "2s", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "16b", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "8h", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "4s", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "2d", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "8b", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "4h", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "2s", Count, Offset64, 64>;
 }
 
 // Only ld1/st1 has a v1d version.
-multiclass BaseSIMDStN<string Count, string asm, string veclist, int Offset128,
-                       int Offset64, bits<4> opcode> {
+multiclass BaseSIMDStN<string BaseName, string Count, string asm, string veclist,
+                       int Offset128, int Offset64, bits<4> opcode> {
   let hasSideEffects = 0, mayStore = 1, mayLoad = 0 in {
     def v16b : BaseSIMDLdSt<1, 0, opcode, 0b00, asm, (outs),
                             (ins !cast<RegisterOperand>(veclist # "16b"):$Vt,
                                  GPR64sp:$Rn), []>;
     def v8h : BaseSIMDLdSt<1, 0, opcode, 0b01, asm, (outs),
                            (ins !cast<RegisterOperand>(veclist # "8h"):$Vt,
                                 GPR64sp:$Rn), []>;
     def v4s : BaseSIMDLdSt<1, 0, opcode, 0b10, asm, (outs),
def v4h_POST : BaseSIMDLdStPost<0, 0, opcode, 0b01, asm,
                             !cast<RegisterOperand>("GPR64pi" # Offset64):$Xm)>;
     def v2s_POST : BaseSIMDLdStPost<0, 0, opcode, 0b10, asm,
                        (outs GPR64sp:$wback),
                        (ins !cast<RegisterOperand>(veclist # "2s"):$Vt,
                             GPR64sp:$Rn,
                             !cast<RegisterOperand>("GPR64pi" # Offset64):$Xm)>;
   }
 
-  defm : SIMDLdStAliases<asm, "16b", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "8h", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "4s", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "2d", Count, Offset128, 128>;
-  defm : SIMDLdStAliases<asm, "8b", Count, Offset64, 64>;
-  defm : SIMDLdStAliases<asm, "4h", Count, Offset64, 64>;
-  defm : SIMDLdStAliases<asm, "2s", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "16b", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "8h", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "4s", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "2d", Count, Offset128, 128>;
+  defm : SIMDLdStAliases<BaseName, asm, "8b", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "4h", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "2s", Count, Offset64, 64>;
 }
 
-multiclass BaseSIMDLd1<string Count, string asm, string veclist,
+multiclass BaseSIMDLd1<string BaseName, string Count, string asm, string veclist,
                        int Offset128, int Offset64, bits<4> opcode>
-  : BaseSIMDLdN<Count, asm, veclist, Offset128, Offset64, opcode> {
+  : BaseSIMDLdN<BaseName, Count, asm, veclist, Offset128, Offset64, opcode> {
 
   // LD1 instructions have extra "1d" variants.
   let hasSideEffects = 0, mayLoad = 1, mayStore = 0 in {
     def v1d : BaseSIMDLdSt<0, 1, opcode, 0b11, asm,
                            (outs !cast<RegisterOperand>(veclist # "1d"):$Vt),
                            (ins GPR64sp:$Rn), []>;
 
     def v1d_POST : BaseSIMDLdStPost<0, 1, opcode, 0b11, asm,
                        (outs GPR64sp:$wback,
                              !cast<RegisterOperand>(veclist # "1d"):$Vt),
                        (ins GPR64sp:$Rn,
                             !cast<RegisterOperand>("GPR64pi" # Offset64):$Xm)>;
   }
 
-  defm : SIMDLdStAliases<asm, "1d", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "1d", Count, Offset64, 64>;
 }
 
-multiclass BaseSIMDSt1<string Count, string asm, string veclist,
+multiclass BaseSIMDSt1<string BaseName, string Count, string asm, string veclist,
                        int Offset128, int Offset64, bits<4> opcode>
-  : BaseSIMDStN<Count, asm, veclist, Offset128, Offset64, opcode> {
+  : BaseSIMDStN<BaseName, Count, asm, veclist, Offset128, Offset64, opcode> {
 
   // ST1 instructions have extra "1d" variants.
   let hasSideEffects = 0, mayLoad = 0, mayStore = 1 in {
     def v1d : BaseSIMDLdSt<0, 0, opcode, 0b11, asm, (outs),
                            (ins !cast<RegisterOperand>(veclist # "1d"):$Vt,
                                 GPR64sp:$Rn), []>;
 
     def v1d_POST : BaseSIMDLdStPost<0, 0, opcode, 0b11, asm,
                        (outs GPR64sp:$wback),
                        (ins !cast<RegisterOperand>(veclist # "1d"):$Vt,
                             GPR64sp:$Rn,
                             !cast<RegisterOperand>("GPR64pi" # Offset64):$Xm)>;
   }
 
-  defm : SIMDLdStAliases<asm, "1d", Count, Offset64, 64>;
+  defm : SIMDLdStAliases<BaseName, asm, "1d", Count, Offset64, 64>;
 }
 
 multiclass SIMDLd1Multiple<string asm> {
-  defm One   : BaseSIMDLd1<"One", asm, "VecListOne", 16, 8,  0b0111>;
-  defm Two   : BaseSIMDLd1<"Two", asm, "VecListTwo", 32, 16, 0b1010>;
-  defm Three : BaseSIMDLd1<"Three", asm, "VecListThree", 48, 24, 0b0110>;
-  defm Four  : BaseSIMDLd1<"Four", asm, "VecListFour", 64, 32, 0b0010>;
+  defm One   : BaseSIMDLd1<NAME, "One", asm, "VecListOne", 16, 8,  0b0111>;
+  defm Two   : BaseSIMDLd1<NAME, "Two", asm, "VecListTwo", 32, 16, 0b1010>;
+  defm Three : BaseSIMDLd1<NAME, "Three", asm, "VecListThree", 48, 24, 0b0110>;
+  defm Four  : BaseSIMDLd1<NAME, "Four", asm, "VecListFour", 64, 32, 0b0010>;
 }
 
 multiclass SIMDSt1Multiple<string asm> {
-  defm One   : BaseSIMDSt1<"One", asm, "VecListOne", 16, 8,  0b0111>;
-  defm Two   : BaseSIMDSt1<"Two", asm, "VecListTwo", 32, 16, 0b1010>;
-  defm Three : BaseSIMDSt1<"Three", asm, "VecListThree", 48, 24, 0b0110>;
-  defm Four  : BaseSIMDSt1<"Four", asm, "VecListFour", 64, 32, 0b0010>;
+  defm One   : BaseSIMDSt1<NAME, "One", asm, "VecListOne", 16, 8,  0b0111>;
+  defm Two   : BaseSIMDSt1<NAME, "Two", asm, "VecListTwo", 32, 16, 0b1010>;
+  defm Three : BaseSIMDSt1<NAME, "Three", asm, "VecListThree", 48, 24, 0b0110>;
+  defm Four  : BaseSIMDSt1<NAME, "Four", asm, "VecListFour", 64, 32, 0b0010>;
 }
 
 multiclass SIMDLd2Multiple<string asm> {
-  defm Two : BaseSIMDLdN<"Two", asm, "VecListTwo", 32, 16, 0b1000>;
+  defm Two : BaseSIMDLdN<NAME, "Two", asm, "VecListTwo", 32, 16, 0b1000>;
 }
 
 multiclass SIMDSt2Multiple<string asm> {
-  defm Two : BaseSIMDStN<"Two", asm, "VecListTwo", 32, 16, 0b1000>;
+  defm Two : BaseSIMDStN<NAME, "Two", asm, "VecListTwo", 32, 16, 0b1000>;
 }
 
 multiclass SIMDLd3Multiple<string asm> {
-  defm Three : BaseSIMDLdN<"Three", asm, "VecListThree", 48, 24, 0b0100>;
+  defm Three : BaseSIMDLdN<NAME, "Three", asm, "VecListThree", 48, 24, 0b0100>;
 }
 
 multiclass SIMDSt3Multiple<string asm> {
-  defm Three : BaseSIMDStN<"Three", asm, "VecListThree", 48, 24, 0b0100>;
+  defm Three : BaseSIMDStN<NAME, "Three", asm, "VecListThree", 48, 24, 0b0100>;
 }
 
 multiclass SIMDLd4Multiple<string asm> {
-  defm Four : BaseSIMDLdN<"Four", asm, "VecListFour", 64, 32, 0b0000>;
+  defm Four : BaseSIMDLdN<NAME, "Four", asm, "VecListFour", 64, 32, 0b0000>;
 }
 
 multiclass SIMDSt4Multiple<string asm> {
-  defm Four : BaseSIMDStN<"Four", asm, "VecListFour", 64, 32, 0b0000>;
+  defm Four : BaseSIMDStN<NAME, "Four", asm, "VecListFour", 64, 32, 0b0000>;
 }
 
 //---
 // AdvSIMD Load/store single-element
 //---
 
 class BaseSIMDLdStSingle<bit L, bit R, bits<3> opcode,
                          string asm, string operands, string cst,
class BaseSIMDLdRPost<bit Q, bit R, bits<3> opcode, bit S, bits<2> size,
   bits<5> Xm;
   let Inst{30} = Q;
   let Inst{23} = 1;
   let Inst{20-16} = Xm;
   let Inst{12} = S;
   let Inst{11-10} = size;
 }
 
-multiclass SIMDLdrAliases<string asm, string layout, string Count,
+multiclass SIMDLdrAliases<string BaseName, string asm, string layout, string Count,
                           int Offset, int Size> {
   // E.g. "ld1r { v0.8b }, [x1], #1"
   //      "ld1r.8b\t$Vt, [$Rn], #1"
   // may get mapped to
   //      (LD1Rv8b_POST VecListOne8b:$Vt, GPR64sp:$Rn, XZR)
   def : InstAlias<asm # "\t$Vt, [$Rn], #" # Offset,
-                  (!cast<Instruction>(NAME # "v" # layout # "_POST")
+                  (!cast<Instruction>(BaseName # "v" # layout # "_POST")
                       GPR64sp:$Rn,
                       !cast<RegisterOperand>("VecList" # Count # layout):$Vt,
                       XZR), 1>;
 
   // E.g. "ld1r.8b { v0 }, [x1], #1"
   //      "ld1r.8b\t$Vt, [$Rn], #1"
   // may get mapped to
   //      (LD1Rv8b_POST VecListOne64:$Vt, GPR64sp:$Rn, XZR)
   def : InstAlias<asm # "." # layout # "\t$Vt, [$Rn], #" # Offset,
-                  (!cast<Instruction>(NAME # "v" # layout # "_POST")
+                  (!cast<Instruction>(BaseName # "v" # layout # "_POST")
                       GPR64sp:$Rn,
                       !cast<RegisterOperand>("VecList" # Count # Size):$Vt,
                       XZR), 0>;
 
   // E.g. "ld1r.8b { v0 }, [x1]"
   //      "ld1r.8b\t$Vt, [$Rn]"
   // may get mapped to
   //      (LD1Rv8b VecListOne64:$Vt, GPR64sp:$Rn)
   def : InstAlias<asm # "." # layout # "\t$Vt, [$Rn]",
-                  (!cast<Instruction>(NAME # "v" # layout)
+                  (!cast<Instruction>(BaseName # "v" # layout)
                       !cast<RegisterOperand>("VecList" # Count # Size):$Vt,
                       GPR64sp:$Rn), 0>;
 
   // E.g. "ld1r.8b { v0 }, [x1], x2"
   //      "ld1r.8b\t$Vt, [$Rn], $Xm"
   // may get mapped to
   //      (LD1Rv8b_POST VecListOne64:$Vt, GPR64sp:$Rn, GPR64pi1:$Xm)
   def : InstAlias<asm # "." # layout # "\t$Vt, [$Rn], $Xm",
-                  (!cast<Instruction>(NAME # "v" # layout # "_POST")
+                  (!cast<Instruction>(BaseName # "v" # layout # "_POST")
                       GPR64sp:$Rn,
                       !cast<RegisterOperand>("VecList" # Count # Size):$Vt,
                       !cast<RegisterOperand>("GPR64pi" # Offset):$Xm), 0>;
 }
 
 multiclass SIMDLdR<bit R, bits<3> opcode, bit S, string asm, string Count,
   int Offset1, int Offset2, int Offset4, int Offset8> {
   def v8b : BaseSIMDLdR<0, R, opcode, S, 0b00, asm,
def v4s_POST : BaseSIMDLdRPost<1, R, opcode, S, 0b10, asm,
                                  !cast<DAGOperand>("GPR64pi" # Offset4)>;
   def v1d_POST : BaseSIMDLdRPost<0, R, opcode, S, 0b11, asm,
                                  !cast<DAGOperand>("VecList" # Count # "1d"),
                                  !cast<DAGOperand>("GPR64pi" # Offset8)>;
   def v2d_POST : BaseSIMDLdRPost<1, R, opcode, S, 0b11, asm,
                                  !cast<DAGOperand>("VecList" # Count # "2d"),
                                  !cast<DAGOperand>("GPR64pi" # Offset8)>;
 
-  defm : SIMDLdrAliases<asm, "8b",  Count, Offset1,  64>;
-  defm : SIMDLdrAliases<asm, "16b", Count, Offset1, 128>;
-  defm : SIMDLdrAliases<asm, "4h",  Count, Offset2,  64>;
-  defm : SIMDLdrAliases<asm, "8h",  Count, Offset2, 128>;
-  defm : SIMDLdrAliases<asm, "2s",  Count, Offset4,  64>;
-  defm : SIMDLdrAliases<asm, "4s",  Count, Offset4, 128>;
-  defm : SIMDLdrAliases<asm, "1d",  Count, Offset8,  64>;
-  defm : SIMDLdrAliases<asm, "2d",  Count, Offset8, 128>;
+  defm : SIMDLdrAliases<NAME, asm, "8b",  Count, Offset1,  64>;
+  defm : SIMDLdrAliases<NAME, asm, "16b", Count, Offset1, 128>;
+  defm : SIMDLdrAliases<NAME, asm, "4h",  Count, Offset2,  64>;
+  defm : SIMDLdrAliases<NAME, asm, "8h",  Count, Offset2, 128>;
+  defm : SIMDLdrAliases<NAME, asm, "2s",  Count, Offset4,  64>;
+  defm : SIMDLdrAliases<NAME, asm, "4s",  Count, Offset4, 128>;
+  defm : SIMDLdrAliases<NAME, asm, "1d",  Count, Offset8,  64>;
+  defm : SIMDLdrAliases<NAME, asm, "2d",  Count, Offset8, 128>;
 }
 
 class SIMDLdStSingleB<bit L, bit R, bits<3> opcode, string asm,
                       dag oops, dag iops, list<dag> pattern>
   : BaseSIMDLdStSingle<L, R, opcode, asm, "\t$Vt$idx, [$Rn]", "", oops, iops,
                        pattern> {
   // idx encoded in Q:S:size fields.
   bits<4> idx;
def : InstAlias<asm # "." # layout # "\t$Vt$idx, [$Rn], $Xm",
                       (!cast<Instruction>(NAME # Type # "_POST")
                          GPR64sp:$Rn,
                          !cast<RegisterOperand>("VecList" # Count # "128"):$Vt,
                          idxtype:$idx,
                          !cast<RegisterOperand>("GPR64pi" # Offset):$Xm), 0>;
 }
 
 multiclass SIMDLdSt1SingleAliases<string asm> {
-  defm : SIMDLdStSingleAliases<asm, "b", "i8",  "One", 1, VectorIndexB>;
-  defm : SIMDLdStSingleAliases<asm, "h", "i16", "One", 2, VectorIndexH>;
-  defm : SIMDLdStSingleAliases<asm, "s", "i32", "One", 4, VectorIndexS>;
-  defm : SIMDLdStSingleAliases<asm, "d", "i64", "One", 8, VectorIndexD>;
+  defm "" : SIMDLdStSingleAliases<asm, "b", "i8",  "One", 1, VectorIndexB>;
+  defm "" : SIMDLdStSingleAliases<asm, "h", "i16", "One", 2, VectorIndexH>;
+  defm "" : SIMDLdStSingleAliases<asm, "s", "i32", "One", 4, VectorIndexS>;
+  defm "" : SIMDLdStSingleAliases<asm, "d", "i64", "One", 8, VectorIndexD>;
 }
 
 multiclass SIMDLdSt2SingleAliases<string asm> {
-  defm : SIMDLdStSingleAliases<asm, "b", "i8",  "Two", 2,  VectorIndexB>;
-  defm : SIMDLdStSingleAliases<asm, "h", "i16", "Two", 4,  VectorIndexH>;
-  defm : SIMDLdStSingleAliases<asm, "s", "i32", "Two", 8,  VectorIndexS>;
-  defm : SIMDLdStSingleAliases<asm, "d", "i64", "Two", 16, VectorIndexD>;
+  defm "" : SIMDLdStSingleAliases<asm, "b", "i8",  "Two", 2,  VectorIndexB>;
+  defm "" : SIMDLdStSingleAliases<asm, "h", "i16", "Two", 4,  VectorIndexH>;
+  defm "" : SIMDLdStSingleAliases<asm, "s", "i32", "Two", 8,  VectorIndexS>;
+  defm "" : SIMDLdStSingleAliases<asm, "d", "i64", "Two", 16, VectorIndexD>;
 }
 
 multiclass SIMDLdSt3SingleAliases<string asm> {
-  defm : SIMDLdStSingleAliases<asm, "b", "i8",  "Three", 3,  VectorIndexB>;
-  defm : SIMDLdStSingleAliases<asm, "h", "i16", "Three", 6,  VectorIndexH>;
-  defm : SIMDLdStSingleAliases<asm, "s", "i32", "Three", 12, VectorIndexS>;
-  defm : SIMDLdStSingleAliases<asm, "d", "i64", "Three", 24, VectorIndexD>;
+  defm "" : SIMDLdStSingleAliases<asm, "b", "i8",  "Three", 3,  VectorIndexB>;
+  defm "" : SIMDLdStSingleAliases<asm, "h", "i16", "Three", 6,  VectorIndexH>;
+  defm "" : SIMDLdStSingleAliases<asm, "s", "i32", "Three", 12, VectorIndexS>;
+  defm "" : SIMDLdStSingleAliases<asm, "d", "i64", "Three", 24, VectorIndexD>;
 }
 
 multiclass SIMDLdSt4SingleAliases<string asm> {
-  defm : SIMDLdStSingleAliases<asm, "b", "i8",  "Four", 4,  VectorIndexB>;
-  defm : SIMDLdStSingleAliases<asm, "h", "i16", "Four", 8,  VectorIndexH>;
-  defm : SIMDLdStSingleAliases<asm, "s", "i32", "Four", 16, VectorIndexS>;
-  defm : SIMDLdStSingleAliases<asm, "d", "i64", "Four", 32, VectorIndexD>;
+  defm "" : SIMDLdStSingleAliases<asm, "b", "i8",  "Four", 4,  VectorIndexB>;
+  defm "" : SIMDLdStSingleAliases<asm, "h", "i16", "Four", 8,  VectorIndexH>;
+  defm "" : SIMDLdStSingleAliases<asm, "s", "i32", "Four", 16, VectorIndexS>;
+  defm "" : SIMDLdStSingleAliases<asm, "d", "i64", "Four", 32, VectorIndexD>;
 }
 } // end of 'let Predicates = [HasNEON]'
 
 //----------------------------------------------------------------------------
 // AdvSIMD v8.1 Rounding Double Multiply Add/Subtract
 //----------------------------------------------------------------------------
 
 let Predicates = [HasNEON, HasRDM] in {

lib/Target/AMDGPU/BUFInstructions.td

string ret =
     !if(!eq(addrKind, BUFAddrKind.Offset), "offset",
     !if(!eq(addrKind, BUFAddrKind.OffEn),  "offen",
     !if(!eq(addrKind, BUFAddrKind.IdxEn),  "idxen",
     !if(!eq(addrKind, BUFAddrKind.BothEn), "bothen",
     !if(!eq(addrKind, BUFAddrKind.Addr64), "addr64",
     "")))));
 }
 
-class MUBUFAddr64Table <bit is_addr64, string suffix = ""> {
+class MUBUFAddr64Table <bit is_addr64, string Name> {
   bit IsAddr64 = is_addr64;
-  string OpName = NAME # suffix;
+  string OpName = Name;
 }
 
-class MUBUFLdsTable <bit is_lds, string suffix> {
+class MUBUFLdsTable <bit is_lds, string Name> {
   bit IsLds = is_lds;
-  string OpName = NAME # suffix;
+  string OpName = Name;
 }
 
-class MTBUFAddr64Table <bit is_addr64, string suffix = ""> {
+class MTBUFAddr64Table <bit is_addr64, string Name> {
   bit IsAddr64 = is_addr64;
-  string OpName = NAME # suffix;
+  string OpName = Name;
 }
 
 //===----------------------------------------------------------------------===//
 // MTBUF classes
 //===----------------------------------------------------------------------===//
 
 class MTBUF_Pseudo <string opName, dag outs, dag ins,
                     string asmOps, list<dag> pattern=[]> :

 multiclass MTBUF_Pseudo_Loads<string opName, RegisterClass vdataClass,
                               ValueType load_vt = i32,
                               SDPatternOperator ld = null_frag> {
 
   def _OFFSET : MTBUF_Load_Pseudo <opName, BUFAddrKind.Offset, vdataClass,
     [(set load_vt:$vdata,
      (ld (MUBUFOffset v4i32:$srsrc, i32:$soffset, i16:$offset, i8:$dfmt,
                       i8:$nfmt, i1:$glc, i1:$slc, i1:$tfe)))]>,
-    MTBUFAddr64Table<0>;
+    MTBUFAddr64Table<0, NAME>;
 
   def _ADDR64 : MTBUF_Load_Pseudo <opName, BUFAddrKind.Addr64, vdataClass,
     [(set load_vt:$vdata,
      (ld (MUBUFAddr64 v4i32:$srsrc, i64:$vaddr, i32:$soffset, i16:$offset,
                       i8:$dfmt, i8:$nfmt, i1:$glc, i1:$slc, i1:$tfe)))]>,
-    MTBUFAddr64Table<1>;
+    MTBUFAddr64Table<1, NAME>;
 
   def _OFFEN  : MTBUF_Load_Pseudo <opName, BUFAddrKind.OffEn, vdataClass>;
   def _IDXEN  : MTBUF_Load_Pseudo <opName, BUFAddrKind.IdxEn, vdataClass>;
   def _BOTHEN : MTBUF_Load_Pseudo <opName, BUFAddrKind.BothEn, vdataClass>;
 
   let DisableWQM = 1 in {
     def _OFFSET_exact : MTBUF_Load_Pseudo <opName, BUFAddrKind.Offset, vdataClass>;
     def _OFFEN_exact  : MTBUF_Load_Pseudo <opName, BUFAddrKind.OffEn, vdataClass>;

 multiclass MTBUF_Pseudo_Stores<string opName, RegisterClass vdataClass,
                                ValueType store_vt = i32,
                                SDPatternOperator st = null_frag> {
 
   def _OFFSET : MTBUF_Store_Pseudo <opName, BUFAddrKind.Offset, vdataClass,
     [(st store_vt:$vdata, (MUBUFOffset v4i32:$srsrc, i32:$soffset,
                                        i16:$offset, i8:$dfmt, i8:$nfmt, i1:$glc,
                                        i1:$slc, i1:$tfe))]>,
-    MTBUFAddr64Table<0>;
+    MTBUFAddr64Table<0, NAME>;
 
   def _ADDR64 : MTBUF_Store_Pseudo <opName, BUFAddrKind.Addr64, vdataClass,
     [(st store_vt:$vdata, (MUBUFAddr64 v4i32:$srsrc, i64:$vaddr, i32:$soffset,
                                        i16:$offset, i8:$dfmt, i8:$nfmt, i1:$glc,
                                        i1:$slc, i1:$tfe))]>,
-    MTBUFAddr64Table<1>;
+    MTBUFAddr64Table<1, NAME>;
 
   def _OFFEN  : MTBUF_Store_Pseudo <opName, BUFAddrKind.OffEn, vdataClass>;
   def _IDXEN  : MTBUF_Store_Pseudo <opName, BUFAddrKind.IdxEn, vdataClass>;
   def _BOTHEN : MTBUF_Store_Pseudo <opName, BUFAddrKind.BothEn, vdataClass>;
 
   let DisableWQM = 1 in {
     def _OFFSET_exact : MTBUF_Store_Pseudo <opName, BUFAddrKind.Offset, vdataClass>;
     def _OFFEN_exact  : MTBUF_Store_Pseudo <opName, BUFAddrKind.OffEn, vdataClass>;
multiclass MUBUF_Pseudo_Loads<string opName, RegisterClass vdataClass,
                               bit isLds = 0> {
 
   def _OFFSET : MUBUF_Load_Pseudo <opName, BUFAddrKind.Offset, vdataClass,
     TiedDest, isLds,
     !if(isLds,
         [],
         [(set load_vt:$vdata,
          (ld (MUBUFOffset v4i32:$srsrc, i32:$soffset, i16:$offset, i1:$glc, i1:$slc, i1:$tfe)))])>,
-    MUBUFAddr64Table<0, !if(isLds, "_LDS", "")>;
+    MUBUFAddr64Table<0, NAME # !if(isLds, "_LDS", "")>;
 
   def _ADDR64 : MUBUF_Load_Pseudo <opName, BUFAddrKind.Addr64, vdataClass,
     TiedDest, isLds,
     !if(isLds,
         [],
         [(set load_vt:$vdata,
          (ld (MUBUFAddr64 v4i32:$srsrc, i64:$vaddr, i32:$soffset, i16:$offset, i1:$glc, i1:$slc, i1:$tfe)))])>,
-    MUBUFAddr64Table<1, !if(isLds, "_LDS", "")>;
+    MUBUFAddr64Table<1, NAME # !if(isLds, "_LDS", "")>;
 
   def _OFFEN  : MUBUF_Load_Pseudo <opName, BUFAddrKind.OffEn, vdataClass, TiedDest, isLds>;
   def _IDXEN  : MUBUF_Load_Pseudo <opName, BUFAddrKind.IdxEn, vdataClass, TiedDest, isLds>;
   def _BOTHEN : MUBUF_Load_Pseudo <opName, BUFAddrKind.BothEn, vdataClass, TiedDest, isLds>;
 
   let DisableWQM = 1 in {
     def _OFFSET_exact : MUBUF_Load_Pseudo <opName, BUFAddrKind.Offset, vdataClass, TiedDest, isLds>;
     def _OFFEN_exact  : MUBUF_Load_Pseudo <opName, BUFAddrKind.OffEn, vdataClass, TiedDest, isLds>;

 
 multiclass MUBUF_Pseudo_Stores<string opName, RegisterClass vdataClass,
                                ValueType store_vt = i32,
                                SDPatternOperator st = null_frag> {
 
   def _OFFSET : MUBUF_Store_Pseudo <opName, BUFAddrKind.Offset, vdataClass,
     [(st store_vt:$vdata, (MUBUFOffset v4i32:$srsrc, i32:$soffset,
                                        i16:$offset, i1:$glc, i1:$slc, i1:$tfe))]>,
-    MUBUFAddr64Table<0>;
+    MUBUFAddr64Table<0, NAME>;
 
   def _ADDR64 : MUBUF_Store_Pseudo <opName, BUFAddrKind.Addr64, vdataClass,
     [(st store_vt:$vdata, (MUBUFAddr64 v4i32:$srsrc, i64:$vaddr, i32:$soffset,
                                        i16:$offset, i1:$glc, i1:$slc, i1:$tfe))]>,
-    MUBUFAddr64Table<1>;
+    MUBUFAddr64Table<1, NAME>;
 
   def _OFFEN  : MUBUF_Store_Pseudo <opName, BUFAddrKind.OffEn, vdataClass>;
   def _IDXEN  : MUBUF_Store_Pseudo <opName, BUFAddrKind.IdxEn, vdataClass>;
   def _BOTHEN : MUBUF_Store_Pseudo <opName, BUFAddrKind.BothEn, vdataClass>;
 
   let DisableWQM = 1 in {
     def _OFFSET_exact : MUBUF_Store_Pseudo <opName, BUFAddrKind.Offset, vdataClass>;
     def _OFFEN_exact  : MUBUF_Store_Pseudo <opName, BUFAddrKind.OffEn, vdataClass>;

 }
 
 multiclass MUBUF_Pseudo_Atomics <string opName,
                                  RegisterClass vdataClass,
                                  ValueType vdataType,
                                  SDPatternOperator atomic> {
 
   def _OFFSET : MUBUF_AtomicNoRet_Pseudo <opName, BUFAddrKind.Offset, vdataClass>,
-                MUBUFAddr64Table <0>;
+                MUBUFAddr64Table <0, NAME>;
   def _ADDR64 : MUBUF_AtomicNoRet_Pseudo <opName, BUFAddrKind.Addr64, vdataClass>,
-                MUBUFAddr64Table <1>;
+                MUBUFAddr64Table <1, NAME>;
   def _OFFEN  : MUBUF_AtomicNoRet_Pseudo <opName, BUFAddrKind.OffEn,  vdataClass>;
   def _IDXEN  : MUBUF_AtomicNoRet_Pseudo <opName, BUFAddrKind.IdxEn,  vdataClass>;
   def _BOTHEN : MUBUF_AtomicNoRet_Pseudo <opName, BUFAddrKind.BothEn, vdataClass>;
 
   def _OFFSET_RTN : MUBUF_AtomicRet_Pseudo <opName, BUFAddrKind.Offset, vdataClass,
     [(set vdataType:$vdata,
      (atomic (MUBUFOffsetAtomic v4i32:$srsrc, i32:$soffset, i16:$offset, i1:$slc),
              vdataType:$vdata_in))]>,
-    MUBUFAddr64Table <0, "_RTN">;
+    MUBUFAddr64Table <0, NAME # "_RTN">;
 
   def _ADDR64_RTN : MUBUF_AtomicRet_Pseudo <opName, BUFAddrKind.Addr64, vdataClass,
     [(set vdataType:$vdata,
      (atomic (MUBUFAddr64Atomic v4i32:$srsrc, i64:$vaddr, i32:$soffset, i16:$offset, i1:$slc),
              vdataType:$vdata_in))]>,
-    MUBUFAddr64Table <1, "_RTN">;
+    MUBUFAddr64Table <1, NAME # "_RTN">;
 
   def _OFFEN_RTN  : MUBUF_AtomicRet_Pseudo <opName, BUFAddrKind.OffEn,  vdataClass>;
   def _IDXEN_RTN  : MUBUF_AtomicRet_Pseudo <opName, BUFAddrKind.IdxEn,  vdataClass>;
   def _BOTHEN_RTN : MUBUF_AtomicRet_Pseudo <opName, BUFAddrKind.BothEn, vdataClass>;
 }
 
 
 //===----------------------------------------------------------------------===//
multiclass MUBUF_Real_AllAddr_si<bits<7> op> {
   def _OFFEN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_OFFEN")>;
   def _IDXEN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_IDXEN")>;
   def _BOTHEN_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_BOTHEN")>;
 }
 
 multiclass MUBUF_Real_AllAddr_Lds_si<bits<7> op> {
 
   def _OFFSET_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_OFFSET")>,
-                   MUBUFLdsTable<0, "_OFFSET_si">;
+                   MUBUFLdsTable<0, NAME # "_OFFSET_si">;
   def _ADDR64_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_ADDR64")>,
-                   MUBUFLdsTable<0, "_ADDR64_si">;
+                   MUBUFLdsTable<0, NAME # "_ADDR64_si">;
   def _OFFEN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_OFFEN")>,
-                   MUBUFLdsTable<0, "_OFFEN_si">;
+                   MUBUFLdsTable<0, NAME # "_OFFEN_si">;
   def _IDXEN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_IDXEN")>,
-                   MUBUFLdsTable<0, "_IDXEN_si">;
+                   MUBUFLdsTable<0, NAME # "_IDXEN_si">;
   def _BOTHEN_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_BOTHEN")>,
-                   MUBUFLdsTable<0, "_BOTHEN_si">;
+                   MUBUFLdsTable<0, NAME # "_BOTHEN_si">;
 
   def _LDS_OFFSET_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_OFFSET")>,
-                       MUBUFLdsTable<1, "_OFFSET_si">;
+                       MUBUFLdsTable<1, NAME # "_OFFSET_si">;
   def _LDS_ADDR64_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_ADDR64")>,
-                       MUBUFLdsTable<1, "_ADDR64_si">;
+                       MUBUFLdsTable<1, NAME # "_ADDR64_si">;
   def _LDS_OFFEN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_OFFEN")>,
-                       MUBUFLdsTable<1, "_OFFEN_si">;
+                       MUBUFLdsTable<1, NAME # "_OFFEN_si">;
   def _LDS_IDXEN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_IDXEN")>,
-                       MUBUFLdsTable<1, "_IDXEN_si">;
+                       MUBUFLdsTable<1, NAME # "_IDXEN_si">;
   def _LDS_BOTHEN_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_BOTHEN")>,
-                       MUBUFLdsTable<1, "_BOTHEN_si">;
+                       MUBUFLdsTable<1, NAME # "_BOTHEN_si">;
 }
 
 multiclass MUBUF_Real_Atomic_si<bits<7> op> : MUBUF_Real_AllAddr_si<op> {
   def _OFFSET_RTN_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_OFFSET_RTN")>;
   def _ADDR64_RTN_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_ADDR64_RTN")>;
   def _OFFEN_RTN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_OFFEN_RTN")>;
   def _IDXEN_RTN_si  : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_IDXEN_RTN")>;
   def _BOTHEN_RTN_si : MUBUF_Real_si <op, !cast<MUBUF_Pseudo>(NAME#"_BOTHEN_RTN")>;
multiclass MUBUF_Real_AllAddr_vi<bits<7> op> {
   def _OFFEN_vi  : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_OFFEN")>;
   def _IDXEN_vi  : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_IDXEN")>;
   def _BOTHEN_vi : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_BOTHEN")>;
 }
 
 multiclass MUBUF_Real_AllAddr_Lds_vi<bits<7> op> {
 
   def _OFFSET_vi : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_OFFSET")>,
-                   MUBUFLdsTable<0, "_OFFSET_vi">;
+                   MUBUFLdsTable<0, NAME # "_OFFSET_vi">;
   def _OFFEN_vi  : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_OFFEN")>,
-                   MUBUFLdsTable<0, "_OFFEN_vi">;
+                   MUBUFLdsTable<0, NAME # "_OFFEN_vi">;
   def _IDXEN_vi  : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_IDXEN")>,
-                   MUBUFLdsTable<0, "_IDXEN_vi">;
+                   MUBUFLdsTable<0, NAME # "_IDXEN_vi">;
   def _BOTHEN_vi : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_BOTHEN")>,
-                   MUBUFLdsTable<0, "_BOTHEN_vi">;
+                   MUBUFLdsTable<0, NAME # "_BOTHEN_vi">;
 
   def _LDS_OFFSET_vi : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_OFFSET")>,
-                       MUBUFLdsTable<1, "_OFFSET_vi">;
+                       MUBUFLdsTable<1, NAME # "_OFFSET_vi">;
   def _LDS_OFFEN_vi  : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_OFFEN")>,
-                       MUBUFLdsTable<1, "_OFFEN_vi">;
+                       MUBUFLdsTable<1, NAME # "_OFFEN_vi">;
   def _LDS_IDXEN_vi  : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_IDXEN")>,
-                       MUBUFLdsTable<1, "_IDXEN_vi">;
+                       MUBUFLdsTable<1, NAME # "_IDXEN_vi">;
   def _LDS_BOTHEN_vi : MUBUF_Real_vi <op, !cast<MUBUF_Pseudo>(NAME#"_LDS_BOTHEN")>,
-                       MUBUFLdsTable<1, "_BOTHEN_vi">;
+                       MUBUFLdsTable<1, NAME # "_BOTHEN_vi">;
 }
 
 class MUBUF_Real_gfx80 <bits<7> op, MUBUF_Pseudo ps> :
   MUBUF_Real<op, ps>,
   Enc64,
   SIMCInstr<ps.PseudoInstr, SIEncodingFamily.GFX80> {
   let AssemblerPredicate=HasUnpackedD16VMem;
   let DecoderNamespace="GFX80_UNPACKED";

lib/Target/AMDGPU/MIMGInstructions.td

def _V8 # suffix : MIMG_Sampler_Helper <op, asm, dst_rc, VReg_256, wqm, d16_bit>,
                        MIMG_Mask<asm#"_V8"#suffix, channels>;
     def _V16 # suffix : MIMG_Sampler_Helper <op, asm, dst_rc, VReg_512, wqm, d16_bit>,
                         MIMG_Mask<asm#"_V16"#suffix, channels>;
 }
 
 multiclass MIMG_Sampler_Src_Helper <bits<7> op, string asm,
                                     RegisterClass dst_rc,
                                     int channels, bit wqm> {
-  defm : MIMG_Sampler_Src_Helper_Helper <op, asm, dst_rc, channels, wqm, 0, "">;
+  defm "" : MIMG_Sampler_Src_Helper_Helper <op, asm, dst_rc, channels, wqm, 0, "">;
 
   let d16 = 1 in {
     let SubtargetPredicate = HasPackedD16VMem in {
-      defm : MIMG_Sampler_Src_Helper_Helper <op, asm, dst_rc, channels, wqm, 1, "_D16">;
+      defm "" : MIMG_Sampler_Src_Helper_Helper <op, asm, dst_rc, channels, wqm, 1, "_D16">;
     } // End HasPackedD16VMem.
 
     let SubtargetPredicate = HasUnpackedD16VMem, DecoderNamespace = "GFX80_UNPACKED" in {
-      defm : MIMG_Sampler_Src_Helper_Helper <op, asm, dst_rc, channels, wqm, 1, "_D16_gfx80">;
+      defm "" : MIMG_Sampler_Src_Helper_Helper <op, asm, dst_rc, channels, wqm, 1, "_D16_gfx80">;
     } // End HasUnpackedD16VMem.
   } // End d16 = 1.
 }
 
 multiclass MIMG_Sampler <bits<7> op, string asm, bit wqm=0> {
   defm _V1 : MIMG_Sampler_Src_Helper<op, asm, VGPR_32, 1, wqm>;
   defm _V2 : MIMG_Sampler_Src_Helper<op, asm, VReg_64, 2, wqm>;
   defm _V3 : MIMG_Sampler_Src_Helper<op, asm, VReg_96, 3, wqm>;
multiclass MIMG_Gather_Src_Helper <bits<7> op, string asm,
   def prefix # _V2 # suffix : MIMG_Gather_Helper <op, asm, dst_rc, VReg_64, wqm, d16_bit>;
   def prefix # _V3 # suffix : MIMG_Gather_Helper <op, asm, dst_rc, VReg_96, wqm, d16_bit>;
   def prefix # _V4 # suffix : MIMG_Gather_Helper <op, asm, dst_rc, VReg_128, wqm, d16_bit>;
   def prefix # _V8 # suffix : MIMG_Gather_Helper <op, asm, dst_rc, VReg_256, wqm, d16_bit>;
   def prefix # _V16 # suffix : MIMG_Gather_Helper <op, asm, dst_rc, VReg_512, wqm, d16_bit>;
 }
 
 multiclass MIMG_Gather <bits<7> op, string asm, bit wqm=0> {
-  defm : MIMG_Gather_Src_Helper<op, asm, VReg_128, wqm, 0, "_V4", "">;
+  defm "" : MIMG_Gather_Src_Helper<op, asm, VReg_128, wqm, 0, "_V4", "">;
 
   let d16 = 1 in {
     let AssemblerPredicate = HasPackedD16VMem in {
-      defm : MIMG_Gather_Src_Helper<op, asm, VReg_64, wqm, 1, "_V2", "_D16">;
+      defm "" : MIMG_Gather_Src_Helper<op, asm, VReg_64, wqm, 1, "_V2", "_D16">;
     } // End HasPackedD16VMem.
 
     let AssemblerPredicate = HasUnpackedD16VMem, DecoderNamespace = "GFX80_UNPACKED" in {
-      defm : MIMG_Gather_Src_Helper<op, asm, VReg_128, wqm, 1, "_V4", "_D16_gfx80">;
+      defm "" : MIMG_Gather_Src_Helper<op, asm, VReg_128, wqm, 1, "_V4", "_D16_gfx80">;
     } // End HasUnpackedD16VMem.
   } // End d16 = 1.
 }
 
 multiclass MIMG_Gather_WQM <bits<7> op, string asm> : MIMG_Gather<op, asm, 1>;
 
 //===----------------------------------------------------------------------===//
 // MIMG Instructions

lib/Target/X86/X86InstrAVX512.td

def VEXTRACTPSZmr : AVX512AIi8<0x17, MRMDestMem, (outs),
                           addr:$dst)]>,
       EVEX, VEX_WIG, EVEX_CD8<32, CD8VT1>, Sched<[WriteVecExtractSt]>;
 
 //===---------------------------------------------------------------------===//
 // AVX-512 BROADCAST
 //---
 // broadcast with a scalar argument.
 multiclass avx512_broadcast_scalar<bits<8> opc, string OpcodeStr,
+                            string Name,
                             X86VectorVTInfo DestInfo, X86VectorVTInfo SrcInfo> {
   def : Pat<(DestInfo.VT (X86VBroadcast SrcInfo.FRC:$src)),
-            (!cast<Instruction>(NAME#DestInfo.ZSuffix#r)
+            (!cast<Instruction>(Name#DestInfo.ZSuffix#r)
              (COPY_TO_REGCLASS SrcInfo.FRC:$src, SrcInfo.RC))>;
   def : Pat<(DestInfo.VT (vselect DestInfo.KRCWM:$mask,
                                   (X86VBroadcast SrcInfo.FRC:$src),
                                   DestInfo.RC:$src0)),
-            (!cast<Instruction>(NAME#DestInfo.ZSuffix#rk)
+            (!cast<Instruction>(Name#DestInfo.ZSuffix#rk)
              DestInfo.RC:$src0, DestInfo.KRCWM:$mask,
              (COPY_TO_REGCLASS SrcInfo.FRC:$src, SrcInfo.RC))>;
   def : Pat<(DestInfo.VT (vselect DestInfo.KRCWM:$mask,
                                   (X86VBroadcast SrcInfo.FRC:$src),
                                   DestInfo.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#DestInfo.ZSuffix#rkz)
+            (!cast<Instruction>(Name#DestInfo.ZSuffix#rkz)
              DestInfo.KRCWM:$mask, (COPY_TO_REGCLASS SrcInfo.FRC:$src, SrcInfo.RC))>;
 }
 
 // Split version to allow mask and broadcast node to be different types. This
 // helps support the 32x2 broadcasts.
 multiclass avx512_broadcast_rm_split<bits<8> opc, string OpcodeStr,
+                                     string Name,
                                      SchedWrite SchedRR, SchedWrite SchedRM,
                                      X86VectorVTInfo MaskInfo,
                                      X86VectorVTInfo DestInfo,
                                      X86VectorVTInfo SrcInfo,
                                      SDPatternOperator UnmaskedOp = X86VBroadcast> {
   let ExeDomain = DestInfo.ExeDomain, hasSideEffects = 0 in {
   defm r : AVX512_maskable_split<opc, MRMSrcReg, MaskInfo,
                    (outs MaskInfo.RC:$dst),
defm m : AVX512_maskable_split<opc, MRMSrcMem, MaskInfo,
                    Sched<[SchedRM]>;
   }
 
   def : Pat<(MaskInfo.VT
              (bitconvert
               (DestInfo.VT (UnmaskedOp
                             (SrcInfo.VT (scalar_to_vector
                                          (SrcInfo.ScalarLdFrag addr:$src))))))),
-            (!cast<Instruction>(NAME#MaskInfo.ZSuffix#m) addr:$src)>;
+            (!cast<Instruction>(Name#MaskInfo.ZSuffix#m) addr:$src)>;
   def : Pat<(MaskInfo.VT (vselect MaskInfo.KRCWM:$mask,
                           (bitconvert
                            (DestInfo.VT
                             (X86VBroadcast
                              (SrcInfo.VT (scalar_to_vector
                                           (SrcInfo.ScalarLdFrag addr:$src)))))),
                           MaskInfo.RC:$src0)),
-            (!cast<Instruction>(NAME#DestInfo.ZSuffix#mk)
+            (!cast<Instruction>(Name#DestInfo.ZSuffix#mk)
              MaskInfo.RC:$src0, MaskInfo.KRCWM:$mask, addr:$src)>;
   def : Pat<(MaskInfo.VT (vselect MaskInfo.KRCWM:$mask,
                           (bitconvert
                            (DestInfo.VT
                             (X86VBroadcast
                              (SrcInfo.VT (scalar_to_vector
                                           (SrcInfo.ScalarLdFrag addr:$src)))))),
                           MaskInfo.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#MaskInfo.ZSuffix#mkz)
+            (!cast<Instruction>(Name#MaskInfo.ZSuffix#mkz)
              MaskInfo.KRCWM:$mask, addr:$src)>;
 }
 
 // Helper class to force mask and broadcast result to same type.
-multiclass avx512_broadcast_rm<bits<8> opc, string OpcodeStr,
+multiclass avx512_broadcast_rm<bits<8> opc, string OpcodeStr, string Name,
                                SchedWrite SchedRR, SchedWrite SchedRM,
                                X86VectorVTInfo DestInfo,
                                X86VectorVTInfo SrcInfo> :
-  avx512_broadcast_rm_split<opc, OpcodeStr, SchedRR, SchedRM,
+  avx512_broadcast_rm_split<opc, OpcodeStr, Name, SchedRR, SchedRM,
                             DestInfo, DestInfo, SrcInfo>;
 
 multiclass avx512_fp_broadcast_sd<bits<8> opc, string OpcodeStr,
                                                        AVX512VLVectorVTInfo _> {
   let Predicates = [HasAVX512] in {
-    defm Z  : avx512_broadcast_rm<opc, OpcodeStr, WriteFShuffle256, 
+    defm Z  : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteFShuffle256,
                                   WriteFShuffle256Ld, _.info512, _.info128>,
-              avx512_broadcast_scalar<opc, OpcodeStr, _.info512, _.info128>,
-                                      EVEX_V512;
+              avx512_broadcast_scalar<opc, OpcodeStr, NAME, _.info512,
+                                      _.info128>,
   }
 
   let Predicates = [HasVLX] in {
-    defm Z256  : avx512_broadcast_rm<opc, OpcodeStr, WriteFShuffle256,
+    defm Z256  : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteFShuffle256,
                                      WriteFShuffle256Ld, _.info256, _.info128>,
-                 avx512_broadcast_scalar<opc, OpcodeStr, _.info256, _.info128>,
-                                         EVEX_V256;
+                 avx512_broadcast_scalar<opc, OpcodeStr, NAME, _.info256,
+                                         _.info128>,
   }
 }
 
 multiclass avx512_fp_broadcast_ss<bits<8> opc, string OpcodeStr,
                                                        AVX512VLVectorVTInfo _> {
   let Predicates = [HasAVX512] in {
-    defm Z  : avx512_broadcast_rm<opc, OpcodeStr, WriteFShuffle256,
+    defm Z  : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteFShuffle256,
                                   WriteFShuffle256Ld, _.info512, _.info128>,
-              avx512_broadcast_scalar<opc, OpcodeStr, _.info512, _.info128>,
-                               EVEX_V512;
+              avx512_broadcast_scalar<opc, OpcodeStr, NAME, _.info512,
+                                      _.info128>,
   }
 
   let Predicates = [HasVLX] in {
-    defm Z256  : avx512_broadcast_rm<opc, OpcodeStr, WriteFShuffle256,
+    defm Z256  : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteFShuffle256,
                                      WriteFShuffle256Ld, _.info256, _.info128>,
-                 avx512_broadcast_scalar<opc, OpcodeStr, _.info256, _.info128>,
-                             EVEX_V256;
-    defm Z128  : avx512_broadcast_rm<opc, OpcodeStr, WriteFShuffle256,
+                 avx512_broadcast_scalar<opc, OpcodeStr, NAME, _.info256,
+                                         _.info128>,
+    defm Z128  : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteFShuffle256,
                                      WriteFShuffle256Ld, _.info128, _.info128>,
-                 avx512_broadcast_scalar<opc, OpcodeStr, _.info128, _.info128>,
-                             EVEX_V128;
+                 avx512_broadcast_scalar<opc, OpcodeStr, NAME, _.info128,
+                                         _.info128>,
   }
 }
 defm VBROADCASTSS  : avx512_fp_broadcast_ss<0x18, "vbroadcastss",
                                        avx512vl_f32_info>;
 defm VBROADCASTSD  : avx512_fp_broadcast_sd<0x19, "vbroadcastsd",
                                        avx512vl_f64_info>, VEX_W;
 
 multiclass avx512_int_broadcast_reg<bits<8> opc, SchedWrite SchedRR,
defm VPBROADCASTWr : avx512_int_broadcastbw_reg_vl<0x7B, "VPBROADCASTWr",
                        HasBWI>;
 defm VPBROADCASTDr : avx512_int_broadcast_reg_vl<0x7C, avx512vl_i32_info,
                                                  X86VBroadcast, GR32, HasAVX512>;
 defm VPBROADCASTQr : avx512_int_broadcast_reg_vl<0x7C, avx512vl_i64_info,
                                                  X86VBroadcast, GR64, HasAVX512>, VEX_W;
 
 // Provide aliases for broadcast from the same register class that
 // automatically does the extract.
-multiclass avx512_int_broadcast_rm_lowering<X86VectorVTInfo DestInfo,
+multiclass avx512_int_broadcast_rm_lowering<string Name,
+                                            X86VectorVTInfo DestInfo,
                                             X86VectorVTInfo SrcInfo> {
   def : Pat<(DestInfo.VT (X86VBroadcast (SrcInfo.VT SrcInfo.RC:$src))),
-            (!cast<Instruction>(NAME#DestInfo.ZSuffix#"r")
+            (!cast<Instruction>(Name#DestInfo.ZSuffix#"r")
                 (EXTRACT_SUBREG (SrcInfo.VT SrcInfo.RC:$src), sub_xmm))>;
 }
 
 multiclass avx512_int_broadcast_rm_vl<bits<8> opc, string OpcodeStr,
                                         AVX512VLVectorVTInfo _, Predicate prd> {
   let Predicates = [prd] in {
-    defm Z :   avx512_broadcast_rm<opc, OpcodeStr, WriteShuffle256,
+    defm Z :   avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteShuffle256,
                                    WriteShuffle256Ld, _.info512, _.info128>,
-               avx512_int_broadcast_rm_lowering<_.info512, _.info256>,
+               avx512_int_broadcast_rm_lowering<NAME, _.info512, _.info256>,
                                   EVEX_V512;
     // Defined separately to avoid redefinition.
-    defm Z_Alt : avx512_int_broadcast_rm_lowering<_.info512, _.info512>;
+    defm Z_Alt : avx512_int_broadcast_rm_lowering<NAME, _.info512, _.info512>;
   }
   let Predicates = [prd, HasVLX] in {
-    defm Z256 : avx512_broadcast_rm<opc, OpcodeStr, WriteShuffle256,
+    defm Z256 : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteShuffle256,
                                     WriteShuffle256Ld, _.info256, _.info128>,
-                avx512_int_broadcast_rm_lowering<_.info256, _.info256>,
+                avx512_int_broadcast_rm_lowering<NAME, _.info256, _.info256>,
                                  EVEX_V256;
-    defm Z128 : avx512_broadcast_rm<opc, OpcodeStr, WriteShuffle,
+    defm Z128 : avx512_broadcast_rm<opc, OpcodeStr, NAME, WriteShuffle,
                                     WriteShuffleXLd, _.info128, _.info128>,
                                  EVEX_V128;
   }
 }
 
 defm VPBROADCASTB  : avx512_int_broadcast_rm_vl<0x78, "vpbroadcastb",
                                            avx512vl_i8_info, HasBWI>;
 defm VPBROADCASTW  : avx512_int_broadcast_rm_vl<0x79, "vpbroadcastw",
def : Pat<(vselect VK8WM:$mask,
                    (bc_v8i64 (v16i32 (X86SubVBroadcast (bc_v4i32 (loadv2i64 addr:$src))))),
                    VR512:$src0),
           (VBROADCASTI64X2rmk VR512:$src0, VK8WM:$mask, addr:$src)>;
 }
 
 multiclass avx512_common_broadcast_32x2<bits<8> opc, string OpcodeStr,
                          AVX512VLVectorVTInfo _Dst, AVX512VLVectorVTInfo _Src> {
   let Predicates = [HasDQI] in
-    defm Z :    avx512_broadcast_rm_split<opc, OpcodeStr, WriteShuffle256,
+    defm Z :    avx512_broadcast_rm_split<opc, OpcodeStr, NAME, WriteShuffle256,
                                           WriteShuffle256Ld, _Dst.info512,
                                           _Src.info512, _Src.info128, null_frag>,
                                           EVEX_V512;
   let Predicates = [HasDQI, HasVLX] in
-    defm Z256 : avx512_broadcast_rm_split<opc, OpcodeStr, WriteShuffle256,
+    defm Z256 : avx512_broadcast_rm_split<opc, OpcodeStr, NAME, WriteShuffle256,
                                           WriteShuffle256Ld, _Dst.info256,
                                           _Src.info256, _Src.info128, null_frag>,
                                           EVEX_V256;
 }
 
 multiclass avx512_common_broadcast_i32x2<bits<8> opc, string OpcodeStr,
                          AVX512VLVectorVTInfo _Dst, AVX512VLVectorVTInfo _Src> :
   avx512_common_broadcast_32x2<opc, OpcodeStr, _Dst, _Src> {
 
   let Predicates = [HasDQI, HasVLX] in
-    defm Z128 : avx512_broadcast_rm_split<opc, OpcodeStr, WriteShuffle,
+    defm Z128 : avx512_broadcast_rm_split<opc, OpcodeStr, NAME, WriteShuffle,
                                           WriteShuffleXLd, _Dst.info128,
                                           _Src.info128, _Src.info128, null_frag>,
                                           EVEX_V128;
 }
 
 defm VBROADCASTI32X2  : avx512_common_broadcast_i32x2<0x59, "vbroadcasti32x2",
                                           avx512vl_i32_info, avx512vl_i64_info>;
 defm VBROADCASTF32X2  : avx512_common_broadcast_32x2<0x19, "vbroadcastf32x2",

 // operand.
 def CommutePCMPCC : SDNodeXForm<imm, [{
   uint8_t Imm = N->getZExtValue() & 0x7;
   Imm = X86::getSwappedVPCMPImm(Imm);
   return getI8Imm(Imm, SDLoc(N));
 }]>;
 
 multiclass avx512_icmp_cc<bits<8> opc, string Suffix, SDNode OpNode,
-                          X86FoldableSchedWrite sched, X86VectorVTInfo _> {
+                          X86FoldableSchedWrite sched, X86VectorVTInfo _,
+                          string Name> {
   let isCommutable = 1 in
   def rri : AVX512AIi8<opc, MRMSrcReg,
              (outs _.KRC:$dst), (ins _.RC:$src1, _.RC:$src2, AVX512ICC:$cc),
              !strconcat("vpcmp${cc}", Suffix,
                         "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
              [(set _.KRC:$dst, (OpNode (_.VT _.RC:$src1), (_.VT _.RC:$src2),
                                        imm:$cc))]>,
              EVEX_4V, Sched<[sched]>;
def rmik_alt : AVX512AIi8<opc, MRMSrcMem,
                !strconcat("vpcmp", Suffix,
                           "\t{$cc, $src2, $src1, $dst {${mask}}|",
                           "$dst {${mask}}, $src1, $src2, $cc}"), []>,
                EVEX_4V, EVEX_K, Sched<[sched.Folded, ReadAfterLd]>;
   }
 
   def : Pat<(OpNode (bitconvert (_.LdFrag addr:$src2)),
                     (_.VT _.RC:$src1), imm:$cc),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmi") _.RC:$src1, addr:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmi") _.RC:$src1, addr:$src2,
                                                       (CommutePCMPCC imm:$cc))>;
 
   def : Pat<(and _.KRCWM:$mask, (OpNode (bitconvert (_.LdFrag addr:$src2)),
                                         (_.VT _.RC:$src1), imm:$cc)),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmik") _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmik") _.KRCWM:$mask,
                                                        _.RC:$src1, addr:$src2,
                                                        (CommutePCMPCC imm:$cc))>;
 }
 
 multiclass avx512_icmp_cc_rmb<bits<8> opc, string Suffix, SDNode OpNode,
-                              X86FoldableSchedWrite sched, X86VectorVTInfo _> :
-           avx512_icmp_cc<opc, Suffix, OpNode, sched, _> {
+                              X86FoldableSchedWrite sched, X86VectorVTInfo _,
+                              string Name> :
+           avx512_icmp_cc<opc, Suffix, OpNode, sched, _, Name> {
   def rmib : AVX512AIi8<opc, MRMSrcMem,
              (outs _.KRC:$dst), (ins _.RC:$src1, _.ScalarMemOp:$src2,
                                      AVX512ICC:$cc),
              !strconcat("vpcmp${cc}", Suffix,
                         "\t{${src2}", _.BroadcastStr, ", $src1, $dst|",
                         "$dst, $src1, ${src2}", _.BroadcastStr, "}"),
              [(set _.KRC:$dst, (OpNode (_.VT _.RC:$src1),
                                (X86VBroadcast (_.ScalarLdFrag addr:$src2)),
def rmibk_alt : AVX512AIi8<opc, MRMSrcMem,
                !strconcat("vpcmp", Suffix,
                   "\t{$cc, ${src2}", _.BroadcastStr, ", $src1, $dst {${mask}}|",
                   "$dst {${mask}}, $src1, ${src2}", _.BroadcastStr, ", $cc}"), []>,
                EVEX_4V, EVEX_K, EVEX_B, Sched<[sched.Folded, ReadAfterLd]>;
   }
 
   def : Pat<(OpNode (X86VBroadcast (_.ScalarLdFrag addr:$src2)),
                     (_.VT _.RC:$src1), imm:$cc),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmib") _.RC:$src1, addr:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmib") _.RC:$src1, addr:$src2,
                                                        (CommutePCMPCC imm:$cc))>;
 
   def : Pat<(and _.KRCWM:$mask, (OpNode (X86VBroadcast
                                          (_.ScalarLdFrag addr:$src2)),
                                         (_.VT _.RC:$src1), imm:$cc)),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmibk") _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmibk") _.KRCWM:$mask,
                                                        _.RC:$src1, addr:$src2,
                                                        (CommutePCMPCC imm:$cc))>;
 }
 
 multiclass avx512_icmp_cc_vl<bits<8> opc, string Suffix, SDNode OpNode,
                              X86SchedWriteWidths sched,
                              AVX512VLVectorVTInfo VTInfo, Predicate prd> {
   let Predicates = [prd] in
-  defm Z : avx512_icmp_cc<opc, Suffix, OpNode, sched.ZMM, VTInfo.info512>,
-                          EVEX_V512;
+  defm Z : avx512_icmp_cc<opc, Suffix, OpNode, sched.ZMM, VTInfo.info512, NAME>,
 
   let Predicates = [prd, HasVLX] in {
-    defm Z256 : avx512_icmp_cc<opc, Suffix, OpNode, sched.YMM, VTInfo.info256>,
-                               EVEX_V256;
-    defm Z128 : avx512_icmp_cc<opc, Suffix, OpNode, sched.XMM, VTInfo.info128>,
-                               EVEX_V128;
+    defm Z256 : avx512_icmp_cc<opc, Suffix, OpNode, sched.YMM, VTInfo.info256,
+                               NAME>,
+    defm Z128 : avx512_icmp_cc<opc, Suffix, OpNode, sched.XMM, VTInfo.info128,
+                               NAME>,
   }
 }
 
 multiclass avx512_icmp_cc_rmb_vl<bits<8> opc, string Suffix, SDNode OpNode,
                                  X86SchedWriteWidths sched,
                                  AVX512VLVectorVTInfo VTInfo, Predicate prd> {
   let Predicates = [prd] in
   defm Z : avx512_icmp_cc_rmb<opc, Suffix, OpNode, sched.ZMM,
-                              VTInfo.info512>, EVEX_V512;
+                              VTInfo.info512, NAME>, EVEX_V512;
 
   let Predicates = [prd, HasVLX] in {
     defm Z256 : avx512_icmp_cc_rmb<opc, Suffix, OpNode, sched.YMM,
-                                    VTInfo.info256>, EVEX_V256;
+                                    VTInfo.info256, NAME>, EVEX_V256;
     defm Z128 : avx512_icmp_cc_rmb<opc, Suffix, OpNode, sched.XMM,
-                                   VTInfo.info128>, EVEX_V128;
+                                   VTInfo.info128, NAME>, EVEX_V128;
   }
 }
 
 // FIXME: Is there a better scheduler class for VPCMP/VPCMPU?
 defm VPCMPB : avx512_icmp_cc_vl<0x3F, "b", X86cmpm, SchedWriteVecALU,
                                 avx512vl_i8_info, HasBWI>, EVEX_CD8<8, CD8VF>;
 defm VPCMPUB : avx512_icmp_cc_vl<0x3E, "ub", X86cmpmu, SchedWriteVecALU,
                                  avx512vl_i8_info, HasBWI>, EVEX_CD8<8, CD8VF>;

 
 defm VPCMPQ : avx512_icmp_cc_rmb_vl<0x1F, "q", X86cmpm, SchedWriteVecALU,
                                     avx512vl_i64_info, HasAVX512>,
                                     VEX_W, EVEX_CD8<64, CD8VF>;
 defm VPCMPUQ : avx512_icmp_cc_rmb_vl<0x1E, "uq", X86cmpmu, SchedWriteVecALU,
                                      avx512vl_i64_info, HasAVX512>,
                                      VEX_W, EVEX_CD8<64, CD8VF>;
 
-multiclass avx512_vcmp_common<X86FoldableSchedWrite sched, X86VectorVTInfo _> {
+multiclass avx512_vcmp_common<X86FoldableSchedWrite sched, X86VectorVTInfo _,
+                              string Name> {
   defm  rri  : AVX512_maskable_cmp<0xC2, MRMSrcReg, _,
                    (outs _.KRC:$dst), (ins _.RC:$src1, _.RC:$src2,AVXCC:$cc),
                    "vcmp${cc}"#_.Suffix,
                    "$src2, $src1", "$src1, $src2",
                    (X86cmpm (_.VT _.RC:$src1),
                          (_.VT _.RC:$src2),
                            imm:$cc), 1>,
                    Sched<[sched]>;
let mayLoad = 1 in {
                          "$src1, ${src2}"##_.BroadcastStr##", $cc">,
                          EVEX_B, Sched<[sched.Folded, ReadAfterLd]>;
     }
   }
 
   // Patterns for selecting with loads in other operand.
   def : Pat<(X86cmpm (_.LdFrag addr:$src2), (_.VT _.RC:$src1),
                      CommutableCMPCC:$cc),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmi") _.RC:$src1, addr:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmi") _.RC:$src1, addr:$src2,
                                                       imm:$cc)>;
 
   def : Pat<(and _.KRCWM:$mask, (X86cmpm (_.LdFrag addr:$src2),
                                          (_.VT _.RC:$src1),
                                          CommutableCMPCC:$cc)),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmik") _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmik") _.KRCWM:$mask,
                                                        _.RC:$src1, addr:$src2,
                                                        imm:$cc)>;
 
   def : Pat<(X86cmpm (X86VBroadcast (_.ScalarLdFrag addr:$src2)),
                      (_.VT _.RC:$src1), CommutableCMPCC:$cc),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmbi") _.RC:$src1, addr:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmbi") _.RC:$src1, addr:$src2,
                                                        imm:$cc)>;
 
   def : Pat<(and _.KRCWM:$mask, (X86cmpm (X86VBroadcast
                                           (_.ScalarLdFrag addr:$src2)),
                                          (_.VT _.RC:$src1),
                                          CommutableCMPCC:$cc)),
-            (!cast<Instruction>(NAME#_.ZSuffix#"rmbik") _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#"rmbik") _.KRCWM:$mask,
                                                         _.RC:$src1, addr:$src2,
                                                         imm:$cc)>;
 }
 
 multiclass avx512_vcmp_sae<X86FoldableSchedWrite sched, X86VectorVTInfo _> {
   // comparison code form (VCMP[EQ/LT/LE/...]
   defm  rrib  : AVX512_maskable_cmp<0xC2, MRMSrcReg, _,
                      (outs _.KRC:$dst),(ins _.RC:$src1, _.RC:$src2, AVXCC:$cc),
defm  rrib_alt  : AVX512_maskable_cmp_alt<0xC2, MRMSrcReg, _,
                          "$cc, {sae}, $src2, $src1",
                          "$src1, $src2, {sae}, $cc">,
                          EVEX_B, Sched<[sched]>;
    }
 }
 
 multiclass avx512_vcmp<X86SchedWriteWidths sched, AVX512VLVectorVTInfo _> {
   let Predicates = [HasAVX512] in {
-    defm Z    : avx512_vcmp_common<sched.ZMM, _.info512>,
+    defm Z    : avx512_vcmp_common<sched.ZMM, _.info512, NAME>,
                 avx512_vcmp_sae<sched.ZMM, _.info512>, EVEX_V512;
 
   }
   let Predicates = [HasAVX512,HasVLX] in {
-   defm Z128 : avx512_vcmp_common<sched.XMM, _.info128>, EVEX_V128;
-   defm Z256 : avx512_vcmp_common<sched.YMM, _.info256>, EVEX_V256;
+   defm Z128 : avx512_vcmp_common<sched.XMM, _.info128, NAME>, EVEX_V128;
+   defm Z256 : avx512_vcmp_common<sched.YMM, _.info256, NAME>, EVEX_V256;
   }
 }
 
 defm VCMPPD : avx512_vcmp<SchedWriteFCmp, avx512vl_f64_info>,
                           AVX512PDIi8Base, EVEX_4V, EVEX_CD8<64, CD8VF>, VEX_W;
 defm VCMPPS : avx512_vcmp<SchedWriteFCmp, avx512vl_f32_info>,
                           AVX512PSIi8Base, EVEX_4V, EVEX_CD8<32, CD8VF>;
 

 defm : operation_subvector_mask_lowering<VK16, v16i1, VK64, v64i1>;
 
 defm : operation_subvector_mask_lowering<VK32, v32i1, VK64, v64i1>;
 
 //===----------------------------------------------------------------------===//
 // AVX-512 - Aligned and unaligned load and store
 //
 
-multiclass avx512_load<bits<8> opc, string OpcodeStr,
+multiclass avx512_load<bits<8> opc, string OpcodeStr, string Name,
                        X86VectorVTInfo _, PatFrag ld_frag, PatFrag mload,
                        X86SchedWriteMoveLS Sched, bit NoRMPattern = 0,
                        SDPatternOperator SelectOprr = vselect> {
   let hasSideEffects = 0 in {
   let isMoveReg = 1 in
   def rr : AVX512PI<opc, MRMSrcReg, (outs _.RC:$dst), (ins _.RC:$src),
                     !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"), [],
                     _.ExeDomain>, EVEX, Sched<[Sched.RR]>;
def rmkz : AVX512PI<opc, MRMSrcMem, (outs _.RC:$dst),
                   (ins _.KRCWM:$mask, _.MemOp:$src),
                   OpcodeStr #"\t{$src, ${dst} {${mask}} {z}|"#
                                 "${dst} {${mask}} {z}, $src}",
                   [(set _.RC:$dst, (_.VT (vselect _.KRCWM:$mask,
                     (_.VT (bitconvert (ld_frag addr:$src))), _.ImmAllZerosV)))],
                   _.ExeDomain>, EVEX, EVEX_KZ, Sched<[Sched.RM]>;
   }
   def : Pat<(_.VT (mload addr:$ptr, _.KRCWM:$mask, undef)),
-            (!cast<Instruction>(NAME#_.ZSuffix##rmkz) _.KRCWM:$mask, addr:$ptr)>;
+            (!cast<Instruction>(Name#_.ZSuffix##rmkz) _.KRCWM:$mask, addr:$ptr)>;
 
   def : Pat<(_.VT (mload addr:$ptr, _.KRCWM:$mask, _.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#_.ZSuffix##rmkz) _.KRCWM:$mask, addr:$ptr)>;
+            (!cast<Instruction>(Name#_.ZSuffix##rmkz) _.KRCWM:$mask, addr:$ptr)>;
 
   def : Pat<(_.VT (mload addr:$ptr, _.KRCWM:$mask, (_.VT _.RC:$src0))),
-            (!cast<Instruction>(NAME#_.ZSuffix##rmk) _.RC:$src0,
+            (!cast<Instruction>(Name#_.ZSuffix##rmk) _.RC:$src0,
              _.KRCWM:$mask, addr:$ptr)>;
 }
 
 multiclass avx512_alignedload_vl<bits<8> opc, string OpcodeStr,
                                  AVX512VLVectorVTInfo _, Predicate prd,
                                  X86SchedWriteMoveLSWidths Sched,
                                  bit NoRMPattern = 0> {
   let Predicates = [prd] in
-  defm Z : avx512_load<opc, OpcodeStr, _.info512,
+  defm Z : avx512_load<opc, OpcodeStr, NAME, _.info512,
                        _.info512.AlignedLdFrag, masked_load_aligned512,
                        Sched.ZMM, NoRMPattern>, EVEX_V512;
 
   let Predicates = [prd, HasVLX] in {
-  defm Z256 : avx512_load<opc, OpcodeStr, _.info256,
+  defm Z256 : avx512_load<opc, OpcodeStr, NAME, _.info256,
                           _.info256.AlignedLdFrag, masked_load_aligned256,
                           Sched.YMM, NoRMPattern>, EVEX_V256;
-  defm Z128 : avx512_load<opc, OpcodeStr, _.info128,
+  defm Z128 : avx512_load<opc, OpcodeStr, NAME, _.info128,
                           _.info128.AlignedLdFrag, masked_load_aligned128,
                           Sched.XMM, NoRMPattern>, EVEX_V128;
   }
 }
 
 multiclass avx512_load_vl<bits<8> opc, string OpcodeStr,
                           AVX512VLVectorVTInfo _, Predicate prd,
                           X86SchedWriteMoveLSWidths Sched,
                           bit NoRMPattern = 0,
                           SDPatternOperator SelectOprr = vselect> {
   let Predicates = [prd] in
-  defm Z : avx512_load<opc, OpcodeStr, _.info512, _.info512.LdFrag,
+  defm Z : avx512_load<opc, OpcodeStr, NAME, _.info512, _.info512.LdFrag,
                        masked_load_unaligned, Sched.ZMM, NoRMPattern,
                        SelectOprr>, EVEX_V512;
 
   let Predicates = [prd, HasVLX] in {
-  defm Z256 : avx512_load<opc, OpcodeStr, _.info256, _.info256.LdFrag,
+  defm Z256 : avx512_load<opc, OpcodeStr, NAME, _.info256, _.info256.LdFrag,
                          masked_load_unaligned, Sched.YMM, NoRMPattern,
                          SelectOprr>, EVEX_V256;
-  defm Z128 : avx512_load<opc, OpcodeStr, _.info128, _.info128.LdFrag,
+  defm Z128 : avx512_load<opc, OpcodeStr, NAME, _.info128, _.info128.LdFrag,
                          masked_load_unaligned, Sched.XMM, NoRMPattern,
                          SelectOprr>, EVEX_V128;
   }
 }
 
-multiclass avx512_store<bits<8> opc, string OpcodeStr,
+multiclass avx512_store<bits<8> opc, string OpcodeStr, string BaseName,
                         X86VectorVTInfo _, PatFrag st_frag, PatFrag mstore,
                         string Name, X86SchedWriteMoveLS Sched,
                         bit NoMRPattern = 0> {
   let hasSideEffects = 0 in {
   let isMoveReg = 1 in
   def rr_REV  : AVX512PI<opc, MRMDestReg, (outs _.RC:$dst), (ins _.RC:$src),
                          OpcodeStr # ".s\t{$src, $dst|$dst, $src}",
                          [], _.ExeDomain>, EVEX, FoldGenData<Name#rr>,
def mr : AVX512PI<opc, MRMDestMem, (outs), (ins _.MemOp:$dst, _.RC:$src),
                         [(st_frag (_.VT _.RC:$src), addr:$dst)]),
                     _.ExeDomain>, EVEX, Sched<[Sched.MR]>;
   def mrk : AVX512PI<opc, MRMDestMem, (outs),
                      (ins _.MemOp:$dst, _.KRCWM:$mask, _.RC:$src),
               OpcodeStr # "\t{$src, ${dst} {${mask}}|${dst} {${mask}}, $src}",
                [], _.ExeDomain>, EVEX, EVEX_K, Sched<[Sched.MR]>;
 
   def: Pat<(mstore addr:$ptr, _.KRCWM:$mask, (_.VT _.RC:$src)),
-           (!cast<Instruction>(NAME#_.ZSuffix##mrk) addr:$ptr,
+           (!cast<Instruction>(BaseName#_.ZSuffix##mrk) addr:$ptr,
                                                         _.KRCWM:$mask, _.RC:$src)>;
 }
 
 multiclass avx512_store_vl< bits<8> opc, string OpcodeStr,
                             AVX512VLVectorVTInfo _, Predicate prd,
                             string Name, X86SchedWriteMoveLSWidths Sched,
                             bit NoMRPattern = 0> {
   let Predicates = [prd] in
-  defm Z : avx512_store<opc, OpcodeStr, _.info512, store,
+  defm Z : avx512_store<opc, OpcodeStr, NAME, _.info512, store,
                         masked_store_unaligned, Name#Z, Sched.ZMM,
                         NoMRPattern>, EVEX_V512;
   let Predicates = [prd, HasVLX] in {
-    defm Z256 : avx512_store<opc, OpcodeStr, _.info256, store,
+    defm Z256 : avx512_store<opc, OpcodeStr, NAME, _.info256, store,
                              masked_store_unaligned, Name#Z256, Sched.YMM,
                              NoMRPattern>, EVEX_V256;
-    defm Z128 : avx512_store<opc, OpcodeStr, _.info128, store,
+    defm Z128 : avx512_store<opc, OpcodeStr, NAME, _.info128, store,
                              masked_store_unaligned, Name#Z128, Sched.XMM,
                              NoMRPattern>, EVEX_V128;
   }
 }
 
 multiclass avx512_alignedstore_vl<bits<8> opc, string OpcodeStr,
                                   AVX512VLVectorVTInfo _, Predicate prd,
                                   string Name, X86SchedWriteMoveLSWidths Sched,
                                   bit NoMRPattern = 0> {
   let Predicates = [prd] in
-  defm Z : avx512_store<opc, OpcodeStr, _.info512, alignedstore,
+  defm Z : avx512_store<opc, OpcodeStr, NAME, _.info512, alignedstore,
                         masked_store_aligned512, Name#Z, Sched.ZMM,
                         NoMRPattern>, EVEX_V512;
 
   let Predicates = [prd, HasVLX] in {
-    defm Z256 : avx512_store<opc, OpcodeStr, _.info256, alignedstore,
+    defm Z256 : avx512_store<opc, OpcodeStr, NAME, _.info256, alignedstore,
                              masked_store_aligned256, Name#Z256, Sched.YMM,
                              NoMRPattern>, EVEX_V256;
-    defm Z128 : avx512_store<opc, OpcodeStr, _.info128, alignedstore,
+    defm Z128 : avx512_store<opc, OpcodeStr, NAME, _.info128, alignedstore,
                              masked_store_aligned128, Name#Z128, Sched.XMM,
                              NoMRPattern>, EVEX_V128;
   }
 }
 
 defm VMOVAPS : avx512_alignedload_vl<0x28, "vmovaps", avx512vl_f32_info,
                                      HasAVX512, SchedWriteFMoveLS>,
                avx512_alignedstore_vl<0x29, "vmovaps", avx512vl_f32_info,
defm VSCALEF : avx512_fp_scalef_all<0x2C, 0x2D, "vscalef", X86scalef, X86scalefs,
                                     SchedWriteFAdd>, T8PD;
 
 //===----------------------------------------------------------------------===//
 // AVX-512  VPTESTM instructions
 //===----------------------------------------------------------------------===//
 
 multiclass avx512_vptest<bits<8> opc, string OpcodeStr, PatFrag OpNode,
                          X86FoldableSchedWrite sched, X86VectorVTInfo _,
-                         string Suffix> {
+                         string Name> {
   let ExeDomain = _.ExeDomain in {
   let isCommutable = 1 in
   defm rr : AVX512_maskable_cmp<opc, MRMSrcReg, _, (outs _.KRC:$dst),
                    (ins _.RC:$src1, _.RC:$src2), OpcodeStr,
                       "$src2, $src1", "$src1, $src2",
                    (OpNode (bitconvert (_.i64VT (and _.RC:$src1, _.RC:$src2))),
                            _.ImmAllZerosV)>,
                    EVEX_4V, Sched<[sched]>;
   defm rm : AVX512_maskable_cmp<opc, MRMSrcMem, _, (outs _.KRC:$dst),
                    (ins _.RC:$src1, _.MemOp:$src2), OpcodeStr,
                        "$src2, $src1", "$src1, $src2",
                    (OpNode (bitconvert
                             (_.i64VT (and _.RC:$src1,
                                           (bitconvert (_.LdFrag addr:$src2))))),
                            _.ImmAllZerosV)>,
                    EVEX_4V, EVEX_CD8<_.EltSize, CD8VF>,
                    Sched<[sched.Folded, ReadAfterLd]>;
   }
 
   // Patterns for compare with 0 that just use the same source twice.
   def : Pat<(_.KVT (OpNode _.RC:$src, _.ImmAllZerosV)),
-            (_.KVT (!cast<Instruction>(NAME # Suffix # _.ZSuffix # "rr")
+            (_.KVT (!cast<Instruction>(Name # _.ZSuffix # "rr")
                                       _.RC:$src, _.RC:$src))>;
 
   def : Pat<(_.KVT (and _.KRC:$mask, (OpNode _.RC:$src, _.ImmAllZerosV))),
-            (_.KVT (!cast<Instruction>(NAME # Suffix # _.ZSuffix # "rrk")
+            (_.KVT (!cast<Instruction>(Name # _.ZSuffix # "rrk")
                                       _.KRC:$mask, _.RC:$src, _.RC:$src))>;
 }
 
 multiclass avx512_vptest_mb<bits<8> opc, string OpcodeStr, PatFrag OpNode,
                             X86FoldableSchedWrite sched, X86VectorVTInfo _> {
   let ExeDomain = _.ExeDomain in
   defm rmb : AVX512_maskable_cmp<opc, MRMSrcMem, _, (outs _.KRC:$dst),
                     (ins _.RC:$src1, _.ScalarMemOp:$src2), OpcodeStr,
                     "${src2}"##_.BroadcastStr##", $src1",
                     "$src1, ${src2}"##_.BroadcastStr,
                     (OpNode (and _.RC:$src1,
                                        (X86VBroadcast
                                         (_.ScalarLdFrag addr:$src2))),
                             _.ImmAllZerosV)>,
                     EVEX_B, EVEX_4V, EVEX_CD8<_.EltSize, CD8VF>,
                     Sched<[sched.Folded, ReadAfterLd]>;
 }
 
 // Use 512bit version to implement 128/256 bit in case NoVLX.
 multiclass avx512_vptest_lowering<PatFrag OpNode, X86VectorVTInfo ExtendInfo,
-                                  X86VectorVTInfo _, string Suffix> {
+                                  X86VectorVTInfo _, string Name> {
   def : Pat<(_.KVT (OpNode (bitconvert (_.i64VT (and _.RC:$src1, _.RC:$src2))),
                            _.ImmAllZerosV)),
             (_.KVT (COPY_TO_REGCLASS
-                     (!cast<Instruction>(NAME # Suffix # "Zrr")
+                     (!cast<Instruction>(Name # "Zrr")
                        (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                                       _.RC:$src1, _.SubRegIdx),
                        (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                                       _.RC:$src2, _.SubRegIdx)),
                    _.KRC))>;
 
   def : Pat<(_.KVT (and _.KRC:$mask,
                         (OpNode (bitconvert (_.i64VT (and _.RC:$src1, _.RC:$src2))),
                                 _.ImmAllZerosV))),
             (COPY_TO_REGCLASS
-             (!cast<Instruction>(NAME # Suffix # "Zrrk")
+             (!cast<Instruction>(Name # "Zrrk")
               (COPY_TO_REGCLASS _.KRC:$mask, ExtendInfo.KRC),
               (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                              _.RC:$src1, _.SubRegIdx),
               (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                              _.RC:$src2, _.SubRegIdx)),
              _.KRC)>;
 
   def : Pat<(_.KVT (OpNode _.RC:$src, _.ImmAllZerosV)),
             (_.KVT (COPY_TO_REGCLASS
-                     (!cast<Instruction>(NAME # Suffix # "Zrr")
+                     (!cast<Instruction>(Name # "Zrr")
                        (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                                       _.RC:$src, _.SubRegIdx),
                        (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                                       _.RC:$src, _.SubRegIdx)),
                    _.KRC))>;
 
   def : Pat<(_.KVT (and _.KRC:$mask, (OpNode _.RC:$src, _.ImmAllZerosV))),
             (COPY_TO_REGCLASS
-             (!cast<Instruction>(NAME # Suffix # "Zrrk")
+             (!cast<Instruction>(Name # "Zrrk")
               (COPY_TO_REGCLASS _.KRC:$mask, ExtendInfo.KRC),
               (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                              _.RC:$src, _.SubRegIdx),
               (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                              _.RC:$src, _.SubRegIdx)),
              _.KRC)>;
 }
 
 multiclass avx512_vptest_dq_sizes<bits<8> opc, string OpcodeStr, PatFrag OpNode,
-                                  X86SchedWriteWidths sched, AVX512VLVectorVTInfo _,
-                                  string Suffix> {
+                                  X86SchedWriteWidths sched, AVX512VLVectorVTInfo _> {
   let Predicates  = [HasAVX512] in
-  defm Z : avx512_vptest<opc, OpcodeStr, OpNode, sched.ZMM, _.info512, Suffix>,
+  defm Z : avx512_vptest<opc, OpcodeStr, OpNode, sched.ZMM, _.info512, NAME>,
            avx512_vptest_mb<opc, OpcodeStr, OpNode, sched.ZMM, _.info512>, EVEX_V512;
 
   let Predicates = [HasAVX512, HasVLX] in {
-  defm Z256 : avx512_vptest<opc, OpcodeStr, OpNode, sched.YMM, _.info256, Suffix>,
+  defm Z256 : avx512_vptest<opc, OpcodeStr, OpNode, sched.YMM, _.info256, NAME>,
               avx512_vptest_mb<opc, OpcodeStr, OpNode, sched.YMM, _.info256>, EVEX_V256;
-  defm Z128 : avx512_vptest<opc, OpcodeStr, OpNode, sched.XMM, _.info128, Suffix>,
+  defm Z128 : avx512_vptest<opc, OpcodeStr, OpNode, sched.XMM, _.info128, NAME>,
               avx512_vptest_mb<opc, OpcodeStr, OpNode, sched.XMM, _.info128>, EVEX_V128;
   }
   let Predicates = [HasAVX512, NoVLX] in {
-  defm Z256_Alt : avx512_vptest_lowering< OpNode, _.info512, _.info256, Suffix>;
-  defm Z128_Alt : avx512_vptest_lowering< OpNode, _.info512, _.info128, Suffix>;
+  defm Z256_Alt : avx512_vptest_lowering< OpNode, _.info512, _.info256, NAME>;
+  defm Z128_Alt : avx512_vptest_lowering< OpNode, _.info512, _.info128, NAME>;
   }
 }
 
 multiclass avx512_vptest_dq<bits<8> opc, string OpcodeStr, PatFrag OpNode,
                             X86SchedWriteWidths sched> {
   defm D : avx512_vptest_dq_sizes<opc, OpcodeStr#"d", OpNode, sched,
-                                 avx512vl_i32_info, "D">;
+                                 avx512vl_i32_info>;
   defm Q : avx512_vptest_dq_sizes<opc, OpcodeStr#"q", OpNode, sched,
-                                 avx512vl_i64_info, "Q">, VEX_W;
+                                 avx512vl_i64_info>, VEX_W;
 }
 
 multiclass avx512_vptest_wb<bits<8> opc, string OpcodeStr,
                             PatFrag OpNode, X86SchedWriteWidths sched> {
   let Predicates = [HasBWI] in {
   defm WZ:    avx512_vptest<opc, OpcodeStr#"w", OpNode, sched.ZMM,
-                            v32i16_info, "W">, EVEX_V512, VEX_W;
+                            v32i16_info, NAME#"W">, EVEX_V512, VEX_W;
   defm BZ:    avx512_vptest<opc, OpcodeStr#"b", OpNode, sched.ZMM,
-                            v64i8_info, "B">, EVEX_V512;
+                            v64i8_info, NAME#"B">, EVEX_V512;
   }
   let Predicates = [HasVLX, HasBWI] in {
 
   defm WZ256: avx512_vptest<opc, OpcodeStr#"w", OpNode, sched.YMM,
-                            v16i16x_info, "W">, EVEX_V256, VEX_W;
+                            v16i16x_info, NAME#"W">, EVEX_V256, VEX_W;
   defm WZ128: avx512_vptest<opc, OpcodeStr#"w", OpNode, sched.XMM,
-                            v8i16x_info, "W">, EVEX_V128, VEX_W;
+                            v8i16x_info, NAME#"W">, EVEX_V128, VEX_W;
   defm BZ256: avx512_vptest<opc, OpcodeStr#"b", OpNode, sched.YMM,
-                            v32i8x_info, "B">, EVEX_V256;
+                            v32i8x_info, NAME#"B">, EVEX_V256;
   defm BZ128: avx512_vptest<opc, OpcodeStr#"b", OpNode, sched.XMM,
-                            v16i8x_info, "B">, EVEX_V128;
+                            v16i8x_info, NAME#"B">, EVEX_V128;
   }
 
   let Predicates = [HasAVX512, NoVLX] in {
-  defm BZ256_Alt : avx512_vptest_lowering<OpNode, v64i8_info, v32i8x_info, "B">;
-  defm BZ128_Alt : avx512_vptest_lowering<OpNode, v64i8_info, v16i8x_info, "B">;
-  defm WZ256_Alt : avx512_vptest_lowering<OpNode, v32i16_info, v16i16x_info, "W">;
-  defm WZ128_Alt : avx512_vptest_lowering<OpNode, v32i16_info, v8i16x_info, "W">;
+  defm BZ256_Alt : avx512_vptest_lowering<OpNode, v64i8_info, v32i8x_info, NAME#"B">;
+  defm BZ128_Alt : avx512_vptest_lowering<OpNode, v64i8_info, v16i8x_info, NAME#"B">;
+  defm WZ256_Alt : avx512_vptest_lowering<OpNode, v32i16_info, v16i16x_info, NAME#"W">;
+  defm WZ128_Alt : avx512_vptest_lowering<OpNode, v32i16_info, v8i16x_info, NAME#"W">;
   }
 }
 
 // These patterns are used to match vptestm/vptestnm. We don't treat pcmpeqm
 // as commutable here because we already canonicalized all zeros vectors to the
 // RHS during lowering.
 def X86pcmpeqm : PatFrag<(ops node:$src1, node:$src2),
                          (X86cmpm node:$src1, node:$src2, (i8 0))>;
defm PSZ : avx512_sqrt_packed_round<opc, !strconcat(OpcodeStr, "ps"),
                                       sched.PS.ZMM, v16f32_info>,
                                       EVEX_V512, PS, EVEX_CD8<32, CD8VF>;
   defm PDZ : avx512_sqrt_packed_round<opc, !strconcat(OpcodeStr, "pd"),
                                       sched.PD.ZMM, v8f64_info>,
                                       EVEX_V512, VEX_W, PD, EVEX_CD8<64, CD8VF>;
 }
 
 multiclass avx512_sqrt_scalar<bits<8> opc, string OpcodeStr, X86FoldableSchedWrite sched,
-                              X86VectorVTInfo _, string SUFF, Intrinsic Intr> {
+                              X86VectorVTInfo _, string Name, Intrinsic Intr> {
   let ExeDomain = _.ExeDomain in {
     defm r_Int : AVX512_maskable_scalar<opc, MRMSrcReg, _, (outs _.RC:$dst),
                          (ins _.RC:$src1, _.RC:$src2), OpcodeStr,
                          "$src2, $src1", "$src1, $src2",
                          (X86fsqrtRnds (_.VT _.RC:$src1),
                                     (_.VT _.RC:$src2),
                                     (i32 FROUND_CURRENT))>,
                          Sched<[sched]>;
let isCodeGenOnly = 1, hasSideEffects = 0, Predicates=[HasAVX512] in {
                   (ins _.FRC:$src1, _.ScalarMemOp:$src2),
                   OpcodeStr#"\t{$src2, $src1, $dst|$dst, $src1, $src2}", []>,
                   Sched<[sched.Folded, ReadAfterLd]>;
     }
   }
 
   let Predicates = [HasAVX512] in {
     def : Pat<(_.EltVT (fsqrt _.FRC:$src)),
-              (!cast<Instruction>(NAME#SUFF#Zr)
+              (!cast<Instruction>(Name#Zr)
                   (_.EltVT (IMPLICIT_DEF)), _.FRC:$src)>;
 
      def : Pat<(Intr VR128X:$src),
-               (!cast<Instruction>(NAME#SUFF#Zr_Int) VR128X:$src,
+               (!cast<Instruction>(Name#Zr_Int) VR128X:$src,
                                  VR128X:$src)>;
   }
 
   let Predicates = [HasAVX512, OptForSize] in {
     def : Pat<(_.EltVT (fsqrt (load addr:$src))),
-              (!cast<Instruction>(NAME#SUFF#Zm)
+              (!cast<Instruction>(Name#Zm)
                   (_.EltVT (IMPLICIT_DEF)), addr:$src)>;
 
     def : Pat<(Intr _.ScalarIntMemCPat:$src2),
-              (!cast<Instruction>(NAME#SUFF#Zm_Int)
+              (!cast<Instruction>(Name#Zm_Int)
                     (_.VT (IMPLICIT_DEF)), addr:$src2)>;
   }
 }
 
 multiclass avx512_sqrt_scalar_all<bits<8> opc, string OpcodeStr,
                                   X86SchedWriteSizes sched> {
-  defm SSZ : avx512_sqrt_scalar<opc, OpcodeStr#"ss", sched.PS.Scl, f32x_info, "SS",
+  defm SSZ : avx512_sqrt_scalar<opc, OpcodeStr#"ss", sched.PS.Scl, f32x_info, NAME#"SS",
                         int_x86_sse_sqrt_ss>,
                         EVEX_CD8<32, CD8VT1>, EVEX_4V, XS, NotMemoryFoldable;
-  defm SDZ : avx512_sqrt_scalar<opc, OpcodeStr#"sd", sched.PD.Scl, f64x_info, "SD",
+  defm SDZ : avx512_sqrt_scalar<opc, OpcodeStr#"sd", sched.PD.Scl, f64x_info, NAME#"SD",
                         int_x86_sse2_sqrt_sd>,
                         EVEX_CD8<64, CD8VT1>, EVEX_4V, XD, VEX_W,
                         NotMemoryFoldable;
 }
 
 defm VSQRT : avx512_sqrt_packed_all<0x51, "vsqrt", SchedWriteFSqrtSizes>,
              avx512_sqrt_packed_all_round<0x51, "vsqrt", SchedWriteFSqrtSizes>;
 
def mrk : AVX512XS8I<opc, MRMDestMem, (outs),
                (ins x86memop:$dst, SrcInfo.KRCWM:$mask, SrcInfo.RC:$src),
                OpcodeStr # "\t{$src, $dst {${mask}}|$dst {${mask}}, $src}", []>,
                EVEX, EVEX_K, Sched<[sched.Folded]>;
   }//mayStore = 1, mayLoad = 1, hasSideEffects = 0
 }
 
 multiclass avx512_trunc_mr_lowering<X86VectorVTInfo SrcInfo,
                                     X86VectorVTInfo DestInfo,
-                                    PatFrag truncFrag, PatFrag mtruncFrag > {
+                                    PatFrag truncFrag, PatFrag mtruncFrag,
+                                    string Name> {
 
   def : Pat<(truncFrag (SrcInfo.VT SrcInfo.RC:$src), addr:$dst),
-            (!cast<Instruction>(NAME#SrcInfo.ZSuffix##mr)
+            (!cast<Instruction>(Name#SrcInfo.ZSuffix##mr)
                                     addr:$dst, SrcInfo.RC:$src)>;
 
   def : Pat<(mtruncFrag addr:$dst, SrcInfo.KRCWM:$mask,
                                                (SrcInfo.VT SrcInfo.RC:$src)),
-            (!cast<Instruction>(NAME#SrcInfo.ZSuffix##mrk)
+            (!cast<Instruction>(Name#SrcInfo.ZSuffix##mrk)
                             addr:$dst, SrcInfo.KRCWM:$mask, SrcInfo.RC:$src)>;
 }
 
 multiclass avx512_trunc<bits<8> opc, string OpcodeStr, SDNode OpNode128,
                         SDNode OpNode256, SDNode OpNode512, X86FoldableSchedWrite sched,
                         AVX512VLVectorVTInfo VTSrcInfo,
                         X86VectorVTInfo DestInfoZ128,
                         X86VectorVTInfo DestInfoZ256, X86VectorVTInfo DestInfoZ,
                         X86MemOperand x86memopZ128, X86MemOperand x86memopZ256,
                         X86MemOperand x86memopZ, PatFrag truncFrag,
                         PatFrag mtruncFrag, Predicate prd = HasAVX512>{
 
   let Predicates = [HasVLX, prd] in {
     defm Z128:  avx512_trunc_common<opc, OpcodeStr, OpNode128, sched,
                              VTSrcInfo.info128, DestInfoZ128, x86memopZ128>,
                 avx512_trunc_mr_lowering<VTSrcInfo.info128, DestInfoZ128,
-                             truncFrag, mtruncFrag>, EVEX_V128;
+                             truncFrag, mtruncFrag, NAME>, EVEX_V128;
 
     defm Z256:  avx512_trunc_common<opc, OpcodeStr, OpNode256, sched,
                              VTSrcInfo.info256, DestInfoZ256, x86memopZ256>,
                 avx512_trunc_mr_lowering<VTSrcInfo.info256, DestInfoZ256,
-                             truncFrag, mtruncFrag>, EVEX_V256;
+                             truncFrag, mtruncFrag, NAME>, EVEX_V256;
   }
   let Predicates = [prd] in
     defm Z:     avx512_trunc_common<opc, OpcodeStr, OpNode512, sched,
                              VTSrcInfo.info512, DestInfoZ, x86memopZ>,
                 avx512_trunc_mr_lowering<VTSrcInfo.info512, DestInfoZ,
-                             truncFrag, mtruncFrag>, EVEX_V512;
+                             truncFrag, mtruncFrag, NAME>, EVEX_V512;
 }
 
 multiclass avx512_trunc_qb<bits<8> opc, string OpcodeStr, SDNode OpNode,
                            X86FoldableSchedWrite sched, PatFrag StoreNode,
                            PatFrag MaskedStoreNode, SDNode InVecNode = OpNode> {
   defm NAME: avx512_trunc<opc, OpcodeStr, InVecNode, InVecNode, InVecNode, sched,
                           avx512vl_i64_info, v16i8x_info, v16i8x_info,
                           v16i8x_info, i16mem, i32mem, i64mem, StoreNode,
multiclass convert_vector_to_mask_common<bits<8> opc, X86VectorVTInfo _, string OpcodeStr > {
     def rr : AVX512XS8I<opc, MRMSrcReg, (outs _.KRC:$dst), (ins _.RC:$src),
                         !strconcat(OpcodeStr, "\t{$src, $dst|$dst, $src}"),
                         [(set _.KRC:$dst, (X86pcmpgtm _.ImmAllZerosV, (_.VT _.RC:$src)))]>,
                         EVEX, Sched<[WriteMove]>;
 }
 
 // Use 512bit version to implement 128/256 bit in case NoVLX.
 multiclass convert_vector_to_mask_lowering<X86VectorVTInfo ExtendInfo,
-                                                            X86VectorVTInfo _> {
+                                           X86VectorVTInfo _,
+                                           string Name> {
 
   def : Pat<(_.KVT (X86pcmpgtm _.ImmAllZerosV, (_.VT _.RC:$src))),
             (_.KVT (COPY_TO_REGCLASS
-                     (!cast<Instruction>(NAME#"Zrr")
+                     (!cast<Instruction>(Name#"Zrr")
                        (INSERT_SUBREG (ExtendInfo.VT (IMPLICIT_DEF)),
                                       _.RC:$src, _.SubRegIdx)),
                    _.KRC))>;
 }
 
 multiclass avx512_convert_vector_to_mask<bits<8> opc, string OpcodeStr,
                                    AVX512VLVectorVTInfo VTInfo, Predicate prd> {
   let Predicates = [prd] in
     defm Z : convert_vector_to_mask_common <opc, VTInfo.info512, OpcodeStr>,
                                             EVEX_V512;
 
   let Predicates = [prd, HasVLX] in {
     defm Z256 : convert_vector_to_mask_common<opc, VTInfo.info256, OpcodeStr>,
                                               EVEX_V256;
     defm Z128 : convert_vector_to_mask_common<opc, VTInfo.info128, OpcodeStr>,
                                                EVEX_V128;
   }
   let Predicates = [prd, NoVLX] in {
-    defm Z256_Alt : convert_vector_to_mask_lowering<VTInfo.info512, VTInfo.info256>;
-    defm Z128_Alt : convert_vector_to_mask_lowering<VTInfo.info512, VTInfo.info128>;
+    defm Z256_Alt : convert_vector_to_mask_lowering<VTInfo.info512, VTInfo.info256, NAME>;
+    defm Z128_Alt : convert_vector_to_mask_lowering<VTInfo.info512, VTInfo.info128, NAME>;
   }
 }
 
 defm VPMOVB2M : avx512_convert_vector_to_mask<0x29, "vpmovb2m",
                                               avx512vl_i8_info, HasBWI>;
 defm VPMOVW2M : avx512_convert_vector_to_mask<0x29, "vpmovw2m",
                                               avx512vl_i16_info, HasBWI>, VEX_W;
 defm VPMOVD2M : avx512_convert_vector_to_mask<0x39, "vpmovd2m",
multiclass compress_by_vec_width_common<bits<8> opc, X86VectorVTInfo _,
   def mrk : AVX5128I<opc, MRMDestMem, (outs),
               (ins _.MemOp:$dst, _.KRCWM:$mask, _.RC:$src),
               OpcodeStr # "\t{$src, $dst {${mask}}|$dst {${mask}}, $src}",
               []>,
               EVEX_K, EVEX_CD8<_.EltSize, CD8VT1>,
               Sched<[sched.Folded]>;
 }
 
-multiclass compress_by_vec_width_lowering<X86VectorVTInfo _ > {
+multiclass compress_by_vec_width_lowering<X86VectorVTInfo _, string Name> {
   def : Pat<(X86mCompressingStore addr:$dst, _.KRCWM:$mask,
                                                (_.VT _.RC:$src)),
-            (!cast<Instruction>(NAME#_.ZSuffix##mrk)
+            (!cast<Instruction>(Name#_.ZSuffix##mrk)
                             addr:$dst, _.KRCWM:$mask, _.RC:$src)>;
 }
 
 multiclass compress_by_elt_width<bits<8> opc, string OpcodeStr,
                                  X86FoldableSchedWrite sched,
                                  AVX512VLVectorVTInfo VTInfo,
                                  Predicate Pred = HasAVX512> {
   let Predicates = [Pred] in
   defm Z : compress_by_vec_width_common<opc, VTInfo.info512, OpcodeStr, sched>,
-           compress_by_vec_width_lowering<VTInfo.info512>, EVEX_V512;
+           compress_by_vec_width_lowering<VTInfo.info512, NAME>, EVEX_V512;
 
   let Predicates = [Pred, HasVLX] in {
     defm Z256 : compress_by_vec_width_common<opc, VTInfo.info256, OpcodeStr, sched>,
-                compress_by_vec_width_lowering<VTInfo.info256>, EVEX_V256;
+                compress_by_vec_width_lowering<VTInfo.info256, NAME>, EVEX_V256;
     defm Z128 : compress_by_vec_width_common<opc, VTInfo.info128, OpcodeStr, sched>,
-                compress_by_vec_width_lowering<VTInfo.info128>, EVEX_V128;
+                compress_by_vec_width_lowering<VTInfo.info128, NAME>, EVEX_V128;
   }
 }
 
 // FIXME: Is there a better scheduler class for VPCOMPRESS?
 defm VPCOMPRESSD : compress_by_elt_width <0x8B, "vpcompressd", WriteVarShuffle256,
                                           avx512vl_i32_info>, EVEX;
 defm VPCOMPRESSQ : compress_by_elt_width <0x8B, "vpcompressq", WriteVarShuffle256,
                                           avx512vl_i64_info>, EVEX, VEX_W;
multiclass expand_by_vec_width<bits<8> opc, X86VectorVTInfo _,
   defm rm : AVX512_maskable<opc, MRMSrcMem, _, (outs _.RC:$dst),
               (ins _.MemOp:$src1), OpcodeStr, "$src1", "$src1",
               (_.VT (X86expand (_.VT (bitconvert
                                       (_.LdFrag addr:$src1)))))>,
             AVX5128IBase, EVEX_CD8<_.EltSize, CD8VT1>,
             Sched<[sched.Folded, ReadAfterLd]>;
 }
 
-multiclass expand_by_vec_width_lowering<X86VectorVTInfo _ > {
+multiclass expand_by_vec_width_lowering<X86VectorVTInfo _, string Name> {
 
   def : Pat<(_.VT (X86mExpandingLoad addr:$src, _.KRCWM:$mask, undef)),
-            (!cast<Instruction>(NAME#_.ZSuffix##rmkz)
+            (!cast<Instruction>(Name#_.ZSuffix##rmkz)
                                         _.KRCWM:$mask, addr:$src)>;
 
   def : Pat<(_.VT (X86mExpandingLoad addr:$src, _.KRCWM:$mask,
                                                (_.VT _.RC:$src0))),
-            (!cast<Instruction>(NAME#_.ZSuffix##rmk)
+            (!cast<Instruction>(Name#_.ZSuffix##rmk)
                             _.RC:$src0, _.KRCWM:$mask, addr:$src)>;
 }
 
 multiclass expand_by_elt_width<bits<8> opc, string OpcodeStr,
                                X86FoldableSchedWrite sched,
                                AVX512VLVectorVTInfo VTInfo,
                                Predicate Pred = HasAVX512> {
   let Predicates = [Pred] in
   defm Z : expand_by_vec_width<opc, VTInfo.info512, OpcodeStr, sched>,
-           expand_by_vec_width_lowering<VTInfo.info512>, EVEX_V512;
+           expand_by_vec_width_lowering<VTInfo.info512, NAME>, EVEX_V512;
 
   let Predicates = [Pred, HasVLX] in {
     defm Z256 : expand_by_vec_width<opc, VTInfo.info256, OpcodeStr, sched>,
-                expand_by_vec_width_lowering<VTInfo.info256>, EVEX_V256;
+                expand_by_vec_width_lowering<VTInfo.info256, NAME>, EVEX_V256;
     defm Z128 : expand_by_vec_width<opc, VTInfo.info128, OpcodeStr, sched>,
-                expand_by_vec_width_lowering<VTInfo.info128>, EVEX_V128;
+                expand_by_vec_width_lowering<VTInfo.info128, NAME>, EVEX_V128;
   }
 }
 
 // FIXME: Is there a better scheduler class for VPEXPAND?
 defm VPEXPANDD : expand_by_elt_width <0x89, "vpexpandd", WriteVarShuffle256,
                                       avx512vl_i32_info>, EVEX;
 defm VPEXPANDQ : expand_by_elt_width <0x89, "vpexpandq", WriteVarShuffle256,
                                       avx512vl_i64_info>, EVEX, VEX_W;
def VPTERNLOG312_imm8 : SDNodeXForm<imm, [{
   if (Imm & 0x08) NewImm |= 0x20;
   if (Imm & 0x10) NewImm |= 0x04;
   if (Imm & 0x20) NewImm |= 0x40;
   if (Imm & 0x40) NewImm |= 0x08;
   return getI8Imm(NewImm, SDLoc(N));
 }]>;
 
 multiclass avx512_ternlog<bits<8> opc, string OpcodeStr, SDNode OpNode,
-                          X86FoldableSchedWrite sched, X86VectorVTInfo _>{
+                          X86FoldableSchedWrite sched, X86VectorVTInfo _,
+                          string Name>{
   let Constraints = "$src1 = $dst", ExeDomain = _.ExeDomain in {
   defm rri : AVX512_maskable_3src<opc, MRMSrcReg, _, (outs _.RC:$dst),
                       (ins _.RC:$src2, _.RC:$src3, u8imm:$src4),
                       OpcodeStr, "$src4, $src3, $src2", "$src2, $src3, $src4",
                       (OpNode (_.VT _.RC:$src1),
                               (_.VT _.RC:$src2),
                               (_.VT _.RC:$src3),
                               (i8 imm:$src4)), 1, 1>,
defm rmbi : AVX512_maskable_3src<opc, MRMSrcMem, _, (outs _.RC:$dst),
                     AVX512AIi8Base, EVEX_4V, EVEX_CD8<_.EltSize, CD8VF>,
                     Sched<[sched.Folded, ReadAfterLd]>;
   }// Constraints = "$src1 = $dst"
 
   // Additional patterns for matching passthru operand in other positions.
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src3, _.RC:$src2, _.RC:$src1, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rrik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rrik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, _.RC:$src3, (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src2, _.RC:$src1, _.RC:$src3, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rrik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rrik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, _.RC:$src3, (VPTERNLOG213_imm8 imm:$src4))>;
 
   // Additional patterns for matching loads in other positions.
   def : Pat<(_.VT (OpNode (bitconvert (_.LdFrag addr:$src3)),
                           _.RC:$src2, _.RC:$src1, (i8 imm:$src4))),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmi) _.RC:$src1, _.RC:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#rmi) _.RC:$src1, _.RC:$src2,
                                    addr:$src3, (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (OpNode _.RC:$src1,
                           (bitconvert (_.LdFrag addr:$src3)),
                           _.RC:$src2, (i8 imm:$src4))),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmi) _.RC:$src1, _.RC:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#rmi) _.RC:$src1, _.RC:$src2,
                                    addr:$src3, (VPTERNLOG132_imm8 imm:$src4))>;
 
   // Additional patterns for matching zero masking with loads in other
   // positions.
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode (bitconvert (_.LdFrag addr:$src3)),
                     _.RC:$src2, _.RC:$src1, (i8 imm:$src4)),
                    _.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmikz) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmikz) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src1, (bitconvert (_.LdFrag addr:$src3)),
                     _.RC:$src2, (i8 imm:$src4)),
                    _.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmikz) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmikz) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG132_imm8 imm:$src4))>;
 
   // Additional patterns for matching masked loads with different
   // operand orders.
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src1, (bitconvert (_.LdFrag addr:$src3)),
                     _.RC:$src2, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG132_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode (bitconvert (_.LdFrag addr:$src3)),
                     _.RC:$src2, _.RC:$src1, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src2, _.RC:$src1,
                     (bitconvert (_.LdFrag addr:$src3)), (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG213_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src2, (bitconvert (_.LdFrag addr:$src3)),
                     _.RC:$src1, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG231_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode (bitconvert (_.LdFrag addr:$src3)),
                     _.RC:$src1, _.RC:$src2, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG312_imm8 imm:$src4))>;
 
   // Additional patterns for matching broadcasts in other positions.
   def : Pat<(_.VT (OpNode (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                           _.RC:$src2, _.RC:$src1, (i8 imm:$src4))),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbi) _.RC:$src1, _.RC:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbi) _.RC:$src1, _.RC:$src2,
                                    addr:$src3, (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (OpNode _.RC:$src1,
                           (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                           _.RC:$src2, (i8 imm:$src4))),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbi) _.RC:$src1, _.RC:$src2,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbi) _.RC:$src1, _.RC:$src2,
                                    addr:$src3, (VPTERNLOG132_imm8 imm:$src4))>;
 
   // Additional patterns for matching zero masking with broadcasts in other
   // positions.
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     _.RC:$src2, _.RC:$src1, (i8 imm:$src4)),
                    _.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbikz) _.RC:$src1,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbikz) _.RC:$src1,
              _.KRCWM:$mask, _.RC:$src2, addr:$src3,
              (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src1,
                     (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     _.RC:$src2, (i8 imm:$src4)),
                    _.ImmAllZerosV)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbikz) _.RC:$src1,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbikz) _.RC:$src1,
              _.KRCWM:$mask, _.RC:$src2, addr:$src3,
              (VPTERNLOG132_imm8 imm:$src4))>;
 
   // Additional patterns for matching masked broadcasts with different
   // operand orders.
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src1,
                     (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     _.RC:$src2, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG132_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     _.RC:$src2, _.RC:$src1, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG321_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src2, _.RC:$src1,
                     (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     (i8 imm:$src4)), _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG213_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode _.RC:$src2,
                     (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     _.RC:$src1, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG231_imm8 imm:$src4))>;
   def : Pat<(_.VT (vselect _.KRCWM:$mask,
                    (OpNode (X86VBroadcast (_.ScalarLdFrag addr:$src3)),
                     _.RC:$src1, _.RC:$src2, (i8 imm:$src4)),
                    _.RC:$src1)),
-            (!cast<Instruction>(NAME#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
+            (!cast<Instruction>(Name#_.ZSuffix#rmbik) _.RC:$src1, _.KRCWM:$mask,
              _.RC:$src2, addr:$src3, (VPTERNLOG312_imm8 imm:$src4))>;
 }
 
 multiclass avx512_common_ternlog<string OpcodeStr, X86SchedWriteWidths sched,
                                  AVX512VLVectorVTInfo _> {
   let Predicates = [HasAVX512] in
     defm Z    : avx512_ternlog<0x25, OpcodeStr, X86vpternlog, sched.ZMM,
-                               _.info512>, EVEX_V512;
+                               _.info512, NAME>, EVEX_V512;
   let Predicates = [HasAVX512, HasVLX] in {
     defm Z128 : avx512_ternlog<0x25, OpcodeStr, X86vpternlog, sched.XMM,
-                               _.info128>, EVEX_V128;
+                               _.info128, NAME>, EVEX_V128;
     defm Z256 : avx512_ternlog<0x25, OpcodeStr, X86vpternlog, sched.YMM,
-                               _.info256>, EVEX_V256;
+                               _.info256, NAME>, EVEX_V256;
   }
 }
 
 defm VPTERNLOGD : avx512_common_ternlog<"vpternlogd", SchedWriteVecALU,
                                         avx512vl_i32_info>;
 defm VPTERNLOGQ : avx512_common_ternlog<"vpternlogq", SchedWriteVecALU,
                                         avx512vl_i64_info>, VEX_W;
 

lib/Target/X86/X86InstrSSE.td

 /// For the non-AVX defs, we need $src1 to be tied to $dst because
 /// the HW instructions are 2 operand / destructive.
 multiclass sse_fp_unop_s<bits<8> opc, string OpcodeStr, RegisterClass RC,
                           ValueType vt, ValueType ScalarVT,
                           X86MemOperand x86memop,
                           Operand intmemop, ComplexPattern int_cpat,
                           Intrinsic Intr, SDNode OpNode, Domain d,
                           X86FoldableSchedWrite sched,
-                          Predicate target, string Suffix> {
+                          Predicate target> {
   let hasSideEffects = 0 in {
   def r : I<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1),
               !strconcat(OpcodeStr, "\t{$src1, $dst|$dst, $src1}"),
             [(set RC:$dst, (OpNode RC:$src1))], d>, Sched<[sched]>,
             Requires<[target]>;
   let mayLoad = 1 in
   def m : I<opc, MRMSrcMem, (outs RC:$dst), (ins x86memop:$src1),
             !strconcat(OpcodeStr, "\t{$src1, $dst|$dst, $src1}"),
def m_Int : I<opc, MRMSrcMem, (outs VR128:$dst), (ins VR128:$src1, intmemop:$src2),
                 Sched<[sched.Folded, ReadAfterLd]>;
   }
   }
 
   let Predicates = [target] in {
   // These are unary operations, but they are modeled as having 2 source operands
   // because the high elements of the destination are unchanged in SSE.
   def : Pat<(Intr VR128:$src),
-            (!cast<Instruction>(NAME#Suffix##r_Int) VR128:$src, VR128:$src)>;
+            (!cast<Instruction>(NAME#r_Int) VR128:$src, VR128:$src)>;
   }
   // We don't want to fold scalar loads into these instructions unless
   // optimizing for size. This is because the folded instruction will have a
   // partial register update, while the unfolded sequence will not, e.g.
   // movss mem, %xmm0
   // rcpss %xmm0, %xmm0
   // which has a clobber before the rcp, vs.
   // rcpss mem, %xmm0
   let Predicates = [target, OptForSize] in {
     def : Pat<(Intr int_cpat:$src2),
-               (!cast<Instruction>(NAME#Suffix##m_Int)
+               (!cast<Instruction>(NAME#m_Int)
                       (vt (IMPLICIT_DEF)), addr:$src2)>;
   }
 }
 
 multiclass avx_fp_unop_s<bits<8> opc, string OpcodeStr, RegisterClass RC,
                           ValueType vt, ValueType ScalarVT,
                           X86MemOperand x86memop,
                           Operand intmemop, ComplexPattern int_cpat,
                           Intrinsic Intr, SDNode OpNode, Domain d,
-                          X86FoldableSchedWrite sched, Predicate target,
-                          string Suffix> {
+                          X86FoldableSchedWrite sched, Predicate target> {
   let hasSideEffects = 0 in {
   def r : I<opc, MRMSrcReg, (outs RC:$dst), (ins RC:$src1, RC:$src2),
             !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
             [], d>, Sched<[sched]>;
   let mayLoad = 1 in
   def m : I<opc, MRMSrcMem, (outs RC:$dst), (ins RC:$src1, x86memop:$src2),
              !strconcat(OpcodeStr, "\t{$src2, $src1, $dst|$dst, $src1, $src2}"),
             [], d>, Sched<[sched.Folded, ReadAfterLd]>;
multiclass avx_fp_unop_s<bits<8> opc, string OpcodeStr, RegisterClass RC,
   // partial register update, while the unfolded sequence will not, e.g.
   // vmovss mem, %xmm0
   // vrcpss %xmm0, %xmm0, %xmm0
   // which has a clobber before the rcp, vs.
   // vrcpss mem, %xmm0, %xmm0
   // TODO: In theory, we could fold the load, and avoid the stall caused by
   // the partial register store, either in BreakFalseDeps or with smarter RA.
   let Predicates = [target] in {
-   def : Pat<(OpNode RC:$src),  (!cast<Instruction>("V"#NAME#Suffix##r)
+   def : Pat<(OpNode RC:$src),  (!cast<Instruction>(NAME#r)
                                 (ScalarVT (IMPLICIT_DEF)), RC:$src)>;
    def : Pat<(Intr VR128:$src),
-             (!cast<Instruction>("V"#NAME#Suffix##r_Int) VR128:$src,
+             (!cast<Instruction>(NAME#r_Int) VR128:$src,
                                  VR128:$src)>;
   }
   let Predicates = [target, OptForSize] in {
     def : Pat<(Intr int_cpat:$src2),
-              (!cast<Instruction>("V"#NAME#Suffix##m_Int)
+              (!cast<Instruction>(NAME#m_Int)
                     (vt (IMPLICIT_DEF)), addr:$src2)>;
     def : Pat<(ScalarVT (OpNode (load addr:$src))),
-              (!cast<Instruction>("V"#NAME#Suffix##m) (ScalarVT (IMPLICIT_DEF)),
+              (!cast<Instruction>(NAME#m) (ScalarVT (IMPLICIT_DEF)),
             addr:$src)>;
   }
 }
 
 /// sse1_fp_unop_p - SSE1 unops in packed form.
 multiclass sse1_fp_unop_p<bits<8> opc, string OpcodeStr, SDNode OpNode,
                           X86SchedWriteWidths sched, list<Predicate> prds> {
 let Predicates = prds in {
def PDm : PDI<opc, MRMSrcMem, (outs VR128:$dst), (ins f128mem:$src),
                 Sched<[sched.XMM.Folded]>;
 }
 
 multiclass sse1_fp_unop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
                           X86SchedWriteWidths sched, Predicate AVXTarget> {
   defm SS        :  sse_fp_unop_s<opc, OpcodeStr##ss, FR32, v4f32, f32, f32mem,
                       ssmem, sse_load_f32,
                       !cast<Intrinsic>("int_x86_sse_"##OpcodeStr##_ss), OpNode,
-                      SSEPackedSingle, sched.Scl, UseSSE1, "SS">, XS;
+                      SSEPackedSingle, sched.Scl, UseSSE1>, XS;
   defm V#NAME#SS  : avx_fp_unop_s<opc, "v"#OpcodeStr##ss, FR32, v4f32, f32,
                       f32mem, ssmem, sse_load_f32,
                       !cast<Intrinsic>("int_x86_sse_"##OpcodeStr##_ss), OpNode,
-                      SSEPackedSingle, sched.Scl, AVXTarget, "SS">, XS, VEX_4V,
+                      SSEPackedSingle, sched.Scl, AVXTarget>, XS, VEX_4V,
                       VEX_LIG, VEX_WIG, NotMemoryFoldable;
 }
 
 multiclass sse2_fp_unop_s<bits<8> opc, string OpcodeStr, SDNode OpNode,
                           X86SchedWriteWidths sched, Predicate AVXTarget> {
   defm SD         : sse_fp_unop_s<opc, OpcodeStr##sd, FR64, v2f64, f64, f64mem,
                          sdmem, sse_load_f64,
                          !cast<Intrinsic>("int_x86_sse2_"##OpcodeStr##_sd),
-                         OpNode, SSEPackedDouble, sched.Scl, UseSSE2, "SD">, XD;
+                         OpNode, SSEPackedDouble, sched.Scl, UseSSE2>, XD;
   defm V#NAME#SD  : avx_fp_unop_s<opc, "v"#OpcodeStr##sd, FR64, v2f64, f64,
                          f64mem, sdmem, sse_load_f64,
                          !cast<Intrinsic>("int_x86_sse2_"##OpcodeStr##_sd),
-                         OpNode, SSEPackedDouble, sched.Scl, AVXTarget, "SD">,
+                         OpNode, SSEPackedDouble, sched.Scl, AVXTarget>,
                          XD, VEX_4V, VEX_LIG, VEX_WIG, NotMemoryFoldable;
 }
 
 // Square root.
 defm SQRT  : sse1_fp_unop_s<0x51, "sqrt", fsqrt, SchedWriteFSqrt, UseAVX>,
              sse1_fp_unop_p<0x51, "sqrt", fsqrt, SchedWriteFSqrt, [HasAVX, NoVLX]>,
              sse2_fp_unop_s<0x51, "sqrt", fsqrt, SchedWriteFSqrt64, UseAVX>,
              sse2_fp_unop_p<0x51, "sqrt", fsqrt, SchedWriteFSqrt64>;

test/TableGen/Dag.td

 def foo2 : foo<somedef2>;
 
 def VAL3 : bar<foo1, somedef1>;
 
 // CHECK:     def VAL3 {  // bar
 // CHECK-NEXT:  dag Dag1 = (somedef1 1);
 // CHECK-NEXT:  dag Dag2 = (somedef1 2);
 // CHECK-NEXT:  dag Dag3 = (somedef1 2);
-// CHECK-NEXT:  NAME = ?
 // CHECK-NEXT: }
 
 
 def VAL4 : bar<foo2, somedef2>;
 // CHECK:      def VAL4 {
 // CHECK-NEXT:  dag Dag1 = (somedef1 1);
 // CHECK-NEXT:  dag Dag2 = (somedef2 2);
 // CHECK-NEXT:  dag Dag3 = (somedef2 2);
-// CHECK-NEXT:  NAME = ?
 // CHECK-NEXT: }
 
 def VAL5 : bar<foo2, somedef2> {
   // Named operands.
   let Dag1 = (somedef1 1:$name1);
 
   // Name, no node.
   let Dag2 = (somedef2 $name2, $name3);
 }
 
 // CHECK:      def VAL5 {
 // CHECK-NEXT:  dag Dag1 = (somedef1 1:$name1);
 // CHECK-NEXT:  dag Dag2 = (somedef2 ?:$name2, ?:$name3);

test/TableGen/MultiClassDefName.td

 // CHECK-NOT: MC2::name
 
 defm : MC2<"bar">;
 
 multiclass MC3<string s> {
   def ZFizz#s : C<s>;
 }
 
-defm : MC3<"Buzz">;
+defm "" : MC3<"Buzz">;
 
 // CHECK: def ZFizzBuzz
 // CHECK: string name = "Buzz";
 // CHECK-NOT: MC3::s
 
 multiclass MC4<string s> {
   def NAME#s : C<s>;
 }

test/TableGen/lisp.td

 // RUN: llvm-tblgen %s
 
 // CHECK:      def One {
 // CHECK-NEXT:   list<string> names = ["Jeffrey Sinclair"];
 // CHECK-NEXT:   string element = "Jeffrey Sinclair";
 // CHECK-NEXT:   list<string> rest = [];
 // CHECK-NEXT:   int null = 1;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def Three {
 // CHECK-NEXT:   list<string> names = ["Tom", "Dick", "Harry"];
 // CHECK-NEXT:   string element = "Tom";
 // CHECK-NEXT:   list<string> rest = ["Dick", "Harry"];
 // CHECK-NEXT:   int null = 0;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 
 class List<list<string> n> {
   list<string> names = n;
 }
 
 class CAR<string e> {
   string element = e;

test/TableGen/name-resolution-consistency.td

 // RUN: llvm-tblgen %s | FileCheck %s
 // XFAIL: vg_leak
 
-// This test demonstrates a number of inconsistencies in how NAME is resolved
-// and record names are constructed.
-//
-// The TODO lines describe a suggested consistent behavior that would result
-// from:
-//  (1) Treating NAME as an implicit multiclass template argument and
-//  (2) always storing the name of (non-anonymous) prototype records in
-//      multiclasses with at least one explicit reference to NAME.
-//
-// Unfortunately, several backends (including X86) rely quite heavily on the
-// current inconsistent behavior and would have to be fixed.
-
 // CHECK: def B0a {
 // CHECK:   string e = "B0";
 // CHECK: }
 
 // CHECK: def B0ba {
-// TODO: expect "B0b" here
-// CHECK:   string a = "B0";
+// CHECK:   string a = "B0b";
 // CHECK:   string b = "B0";
 // CHECK: }
 
+// CHECK: def B0bys {
+// CHECK:   string f = "B0b";
+// CHECK:   string g = "B0";
+// CHECK: }
+
 // CHECK: def B0cza {
-// TODO: expect "B0cz" here
-// CHECK:   string a = "B0";
+// CHECK:   string a = "B0cz";
 // CHECK:   string b = "B0";
 // CHECK: }
 
-// TODO: expect this to be named 'xB0b'
-// CHECK: def B0xb {
-// TODO: expect "B0b" here
-// CHECK:   string c = "b";
-// CHECK:   string d = "b";
+// CHECK: def B0czyt {
+// CHECK:   string f = "B0cz";
+// CHECK:   string g = "B0";
 // CHECK: }
 
-// TODO: expect this to be named B0bys
-// CHECK: def B0ys {
-// TODO: expect "B0b" here
-// CHECK:   string f = "b";
-// CHECK:   string g = "b";
+// CHECK: def C0 {
+// CHECK:   string a = "C0";
+// CHECK:   string b = "C0";
+// CHECK:   string c = "a";
 // CHECK: }
 
-// CHECK: def xB0cz {
-// CHECK:   string c = "B0cz";
-// CHECK:   string d = "B0cz";
+// CHECK: def D0a {
+// CHECK:   string a = "D0a";
+// CHECK:   string b = "D0a";
+// CHECK:   string c = "D0";
 // CHECK: }
 
-// TODO: expect this to be named B0czyt
-// CHECK: def yt {
-// CHECK:   string f = "B0cz";
-// CHECK:   string g = "B0cz";
+// CHECK: def D0b {
+// CHECK:   string a = "D0b";
+// CHECK:   string b = "D0b";
+// CHECK:   string c = "a";
+// CHECK: }
+
+// CHECK: def xB0b {
+// CHECK:   string c = "B0b";
+// CHECK:   string d = "B0";
+// CHECK: }
+
+// CHECK: def xB0cz {
+// CHECK:   string c = "B0cz";
+// CHECK:   string d = "B0";
 // CHECK: }
 
 multiclass A<string p, string q> {
   def a {
     string a = NAME;
     string b = p;
   }
 
multiclass B<string name, string t> {
   }
 
   defm b : A<NAME, "s">;
 
   defm NAME # c # name : A<NAME, t>;
 }
 
 defm B0 : B<"z", "t">;
+
+class Cbase {
+  string a = NAME;
+}
+
+class C<string arg> : Cbase {
+  string b = NAME;
+  string c = arg;
+}
+
+def C0 : C<"a">;
+
+multiclass D<string arg> {
+  def a : C<NAME>;
+  def b : C<arg>;
+}
+
+defm D0 : D<"a">;

test/TableGen/subst.td

 def JaneSmith : AName<"Jane LAST", Ms>;
 def JohnSmithJones : AName<"FIRST LAST-Jones", Mr>;
 def JimmyJohnson : AName<"Jimmy Johnson", Mr>;
 
 // CHECK:      ------------- Classes -----------------
 // CHECK-NEXT: class AName<string AName:name = ?, Honorific AName:honorific = ?> {
 // CHECK-NEXT:   string name = !subst("FIRST", "John", !subst("LAST", "Smith", AName:name));
 // CHECK-NEXT:   Honorific honorific = !subst(TVAR, Mr, AName:honorific);
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: class Honorific<string Honorific:t = ?> {
 // CHECK-NEXT:   string honorific = Honorific:t;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: class Name<string Name:n = ?, Honorific Name:t = ?> {
 // CHECK-NEXT:   string name = Name:n;
 // CHECK-NEXT:   Honorific honorific = Name:t;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: ------------- Defs -----------------
 // CHECK-NEXT: def JaneSmith {
 // CHECK-NEXT:   string name = "Jane Smith";
 // CHECK-NEXT:   Honorific honorific = Ms;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def JimmyJohnson {
 // CHECK-NEXT:   string name = "Jimmy Johnson";
 // CHECK-NEXT:   Honorific honorific = Mr;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def JohnSmith {
 // CHECK-NEXT:   string name = "John Smith";
 // CHECK-NEXT:   Honorific honorific = Mr;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def JohnSmithJones {
 // CHECK-NEXT:   string name = "John Smith-Jones";
 // CHECK-NEXT:   Honorific honorific = Mr;
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def Mr 
 // CHECK-NEXT:   string honorific = "Mr.";
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def Mrs {
 // CHECK-NEXT:   string honorific = "Mrs.";
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def Ms {
 // CHECK-NEXT:   string honorific = "Ms.";
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }
 // CHECK-NEXT: def TVAR {
 // CHECK-NEXT:   string honorific = "Bogus";
-// CHECK-NEXT:   string NAME = ?;
 // CHECK-NEXT: }