[ARM] Creating 'call_mangled' for Neon intrinsics definitions

[pratlucas]

Summary

As multiple versions of the same Neon intrinsic can be created through
the same TableGen definition with the same argument types, the existing
call operator is not always able to properly perform overload
resolutions.

As these different intrinsic versions are differentiated later on by the
NeonEmitter through name mangling, this patch introduces a new
call_mangled operator to the TableGen definitions, which allows a call
for an otherwise ambiguous intrinsic by matching its mangled name with
the mangled variation of the caller.

[dnsampaio]

clang/utils/TableGen/NeonEmitter.cpp

1881–1882

Can we move this above the loop just before? Perhaps, if false, can we just continue the outer loop?

[pratlucas]

Addressing review comment

[pratlucas]

Applying clang-format.

[dnsampaio]

LGTM, after a nit inline.

clang/include/clang/Basic/arm_neon_incl.td

66

As the example above, it would be nice to have an example here as well.

[pratlucas]

Adding example to call_mangled description comment.

clang/include/clang/Basic/arm_neon_incl.td

 // call - Invoke another intrinsic. The input types are type checked and
 //        disambiguated. If there is no intrinsic defined that takes
 //        the given types (or if there is a type ambiguity) an error is
 //        generated at tblgen time. The name of the intrinsic is the raw
 //        name as given to the Inst class (not mangled).
 // example: (call "vget_high", $p0) -> "vgetq_high_s16(__p0)"
 //            (assuming $p0 has type int16x8_t).
 def call;
+// call_mangled - Invoke another intrinsic matching the mangled name variation
+//                of the caller's base type. If there is no intrinsic defined
+//                that has the variation and takes the given types, an error
+//                is generated at tblgen time.
+// example: (call_mangled "vfma_lane", $p0, $p1) -> "vfma_lane(__p0, __p1)"
+//            (assuming non-LaneQ caller)
+//          (call_mangled "vfma_lane", $p0, $p1) -> "vfma_laneq(__p0, __p1)"
+//            (assuming LaneQ caller)
+def call_mangled;
 // cast - Perform a cast to a different type. This gets emitted as a static
 //        C-style cast. For a pure reinterpret cast (T x = *(T*)&y), use
 //        "bitcast".
 //
 //        The syntax is (cast MOD* VAL). The last argument is the value to
 //        cast, preceded by a sequence of type modifiers. The target type
 //        starts off as the type of VAL, and is modified by MOD in sequence.
 //        The available modifiers are:

clang/utils/TableGen/NeonEmitter.cpp

 // See also the documentation in include/clang/Basic/arm_neon.td.
 //
 //===----------------------------------------------------------------------===//
 
 #include "TableGenBackends.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/None.h"
-#include "llvm/ADT/SmallVector.h"
+#include "llvm/ADT/Optional.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/TableGen/Error.h"
 #include "llvm/TableGen/Record.h"
 #include "llvm/TableGen/SetTheory.h"
public:
     }
     std::pair<Type, std::string> emitDagArg(Init *Arg, std::string ArgName);
     std::pair<Type, std::string> emitDagSaveTemp(DagInit *DI);
     std::pair<Type, std::string> emitDagSplat(DagInit *DI);
     std::pair<Type, std::string> emitDagDup(DagInit *DI);
     std::pair<Type, std::string> emitDagDupTyped(DagInit *DI);
     std::pair<Type, std::string> emitDagShuffle(DagInit *DI);
     std::pair<Type, std::string> emitDagCast(DagInit *DI, bool IsBitCast);
-    std::pair<Type, std::string> emitDagCall(DagInit *DI);
+    std::pair<Type, std::string> emitDagCall(DagInit *DI,
+                                             bool MatchMangledName);
     std::pair<Type, std::string> emitDagNameReplace(DagInit *DI);
     std::pair<Type, std::string> emitDagLiteral(DagInit *DI);
     std::pair<Type, std::string> emitDagOp(DagInit *DI);
     std::pair<Type, std::string> emitDag(DagInit *DI);
   };
 };
 
 //===----------------------------------------------------------------------===//
class NeonEmitter {
   void genOverloadTypeCheckCode(raw_ostream &OS,
                                 SmallVectorImpl<Intrinsic *> &Defs);
   void genIntrinsicRangeCheckCode(raw_ostream &OS,
                                   SmallVectorImpl<Intrinsic *> &Defs);
 
 public:
   /// Called by Intrinsic - this attempts to get an intrinsic that takes
   /// the given types as arguments.
-  Intrinsic &getIntrinsic(StringRef Name, ArrayRef<Type> Types);
+  Intrinsic &getIntrinsic(StringRef Name, ArrayRef<Type> Types,
+                          Optional<std::string> MangledName);
 
   /// Called by Intrinsic - returns a globally-unique number.
   unsigned getUniqueNumber() { return UniqueNumber++; }
 
   NeonEmitter(RecordKeeper &R) : Records(R), UniqueNumber(0) {
     Record *SI = R.getClass("SInst");
     Record *II = R.getClass("IInst");
     Record *WI = R.getClass("WInst");
std::pair<Type, std::string> Intrinsic::DagEmitter::emitDag(DagInit *DI) {
   if (Op == "dup_typed")
     return emitDagDupTyped(DI);
   if (Op == "splat")
     return emitDagSplat(DI);
   if (Op == "save_temp")
     return emitDagSaveTemp(DI);
   if (Op == "op")
     return emitDagOp(DI);
-  if (Op == "call")
-    return emitDagCall(DI);
+  if (Op == "call" || Op == "call_mangled")
+    return emitDagCall(DI, Op == "call_mangled");
   if (Op == "name_replace")
     return emitDagNameReplace(DI);
   if (Op == "literal")
     return emitDagLiteral(DI);
   assert_with_loc(false, "Unknown operation!");
   return std::make_pair(Type::getVoid(), "");
 }
 
std::pair<Type, std::string> R1 =
         emitDagArg(DI->getArg(1), std::string(DI->getArgNameStr(1)));
     std::pair<Type, std::string> R2 =
         emitDagArg(DI->getArg(2), std::string(DI->getArgNameStr(2)));
     assert_with_loc(R1.first == R2.first, "Argument type mismatch!");
     return std::make_pair(R1.first, R1.second + " " + Op + " " + R2.second);
   }
 }
 
-std::pair<Type, std::string> Intrinsic::DagEmitter::emitDagCall(DagInit *DI) {
+std::pair<Type, std::string>
+Intrinsic::DagEmitter::emitDagCall(DagInit *DI, bool MatchMangledName) {
   std::vector<Type> Types;
   std::vector<std::string> Values;
   for (unsigned I = 0; I < DI->getNumArgs() - 1; ++I) {
     std::pair<Type, std::string> R =
         emitDagArg(DI->getArg(I + 1), std::string(DI->getArgNameStr(I + 1)));
     Types.push_back(R.first);
     Values.push_back(R.second);
   }
 
   // Look up the called intrinsic.
   std::string N;
   if (StringInit *SI = dyn_cast<StringInit>(DI->getArg(0)))
     N = SI->getAsUnquotedString();
   else
     N = emitDagArg(DI->getArg(0), "").second;
-  Intrinsic &Callee = Intr.Emitter.getIntrinsic(N, Types);
+  Optional<std::string> MangledName;
+  if (MatchMangledName) {
+    if (Intr.getRecord()->getValueAsBit("isLaneQ"))
+      N += "q";
+    MangledName = Intr.mangleName(N, ClassS);
+  }
+  Intrinsic &Callee = Intr.Emitter.getIntrinsic(N, Types, MangledName);
 
   // Make sure the callee is known as an early def.
   Callee.setNeededEarly();
   Intr.Dependencies.insert(&Callee);
 
   // Now create the call itself.
   std::string S = "";
   if (!Callee.isBigEndianSafe())
void Intrinsic::indexBody() {
 
   CurrentRecord = nullptr;
 }
 
 //===----------------------------------------------------------------------===//
 // NeonEmitter implementation
 //===----------------------------------------------------------------------===//
 
-Intrinsic &NeonEmitter::getIntrinsic(StringRef Name, ArrayRef<Type> Types) {
+Intrinsic &NeonEmitter::getIntrinsic(StringRef Name, ArrayRef<Type> Types,
+                                     Optional<std::string> MangledName) {
   // First, look up the name in the intrinsic map.
   assert_with_loc(IntrinsicMap.find(Name.str()) != IntrinsicMap.end(),
                   ("Intrinsic '" + Name + "' not found!").str());
   auto &V = IntrinsicMap.find(Name.str())->second;
   std::vector<Intrinsic *> GoodVec;
 
   // Create a string to print if we end up failing.
   std::string ErrMsg = "looking up intrinsic '" + Name.str() + "(";
for (auto &I : V) {
     ErrMsg += "(";
     for (unsigned A = 0; A < I.getNumParams(); ++A) {
       if (A != 0)
         ErrMsg += ", ";
       ErrMsg += I.getParamType(A).str();
     }
     ErrMsg += ")\n";
 
+    if (MangledName && MangledName != I.getMangledName(true))
+      continue;
+
     if (I.getNumParams() != Types.size())
       continue;
 
-    bool Good = true;
-    for (unsigned Arg = 0; Arg < Types.size(); ++Arg) {
-      if (I.getParamType(Arg) != Types[Arg]) {
-        Good = false;
-        break;
-      }
-    }
-    if (Good)
+    unsigned ArgNum = 0;
+    bool MatchingArgumentTypes =
+        std::all_of(Types.begin(), Types.end(), [&](const auto &Type) {
+          return Type == I.getParamType(ArgNum++);
+        });
+
+    if (MatchingArgumentTypes)
       GoodVec.push_back(&I);
   }
 
   assert_with_loc(!GoodVec.empty(),
                   "No compatible intrinsic found - " + ErrMsg);
   assert_with_loc(GoodVec.size() == 1, "Multiple overloads found - " + ErrMsg);
 
   return *GoodVec.front();