Teach main builtin TableGen to use direct enums, strings, and info

This moves the main builtins and several targets to use nice generated
string tables and info structures rather than X-macros. Even without
obvious prefixes to factor out, the resulting tables are significantly
smaller and much cheaper to compile with out all the X-macro overhead.

This leaves the X-macros in place for atomic builtins which have a wide
range of uses that don't seem reasonable to fold into TableGen.

As future work, these should move to their own file (whether as X-macros
or just generated patterns) so the AST headers don't have to include all
the data for other builtins.

Author: chandlerc

clang/include/clang/AST/Expr.h

@@ -6678,7 +6678,6 @@ class PseudoObjectExpr final
 class AtomicExpr : public Expr {
 public:
   enum AtomicOp {
-#define BUILTIN(ID, TYPE, ATTRS)
 #define ATOMIC_BUILTIN(ID, TYPE, ATTRS) AO ## ID,
 #include "clang/Basic/Builtins.inc"
     // Avoid trailing comma
@@ -6742,7 +6741,6 @@ class AtomicExpr : public Expr {
   AtomicOp getOp() const { return Op; }
   StringRef getOpAsString() const {
     switch (Op) {
-#define BUILTIN(ID, TYPE, ATTRS)
 #define ATOMIC_BUILTIN(ID, TYPE, ATTRS)                                        \
   case AO##ID:                                                                 \
     return #ID;

clang/include/clang/Basic/Builtins.h

@@ -64,9 +64,10 @@ struct HeaderDesc {
 
 namespace Builtin {
 enum ID {
-  NotBuiltin  = 0,      // This is not a builtin function.
-#define BUILTIN(ID, TYPE, ATTRS) BI##ID,
+  NotBuiltin = 0, // This is not a builtin function.
+#define GET_BUILTIN_ENUMERATORS
 #include "clang/Basic/Builtins.inc"
+#undef GET_BUILTIN_ENUMERATORS
   FirstTSBuiltin
 };
 
@@ -75,14 +76,14 @@ struct Info {
   // Rather than store pointers to the string literals describing these four
   // aspects of builtins, we store offsets into a common string table.
   struct StrOffsets {
-    llvm::StringTable::Offset Name;
-    llvm::StringTable::Offset Type;
-    llvm::StringTable::Offset Attributes;
-    llvm::StringTable::Offset Features;
+    llvm::StringTable::Offset Name = 0;
+    llvm::StringTable::Offset Type = 0;
+    llvm::StringTable::Offset Attributes = 0;
+    llvm::StringTable::Offset Features = 0;
   } Offsets;
 
-  HeaderDesc Header;
-  LanguageID Langs;
+  HeaderDesc Header = HeaderDesc::NO_HEADER;
+  LanguageID Langs = ALL_LANGUAGES;
 };
 
 /// A constexpr function to construct an infos array from X-macros.

clang/include/clang/Basic/IdentifierTable.h

@@ -101,8 +101,9 @@ enum class InterestingIdentifier {
   NUM_OBJC_KEYWORDS_AND_NOTABLE_IDENTIFIERS,
 
   NotBuiltin,
-#define BUILTIN(ID, TYPE, ATTRS) BI##ID,
+#define GET_BUILTIN_ENUMERATORS
 #include "clang/Basic/Builtins.inc"
+#undef GET_BUILTIN_ENUMERATORS
   FirstTSBuiltin,
 
   NotInterestingIdentifier = 65534

clang/include/clang/Basic/TargetBuiltins.h

@@ -93,8 +93,9 @@ namespace clang {
   namespace BPF {
   enum {
     LastTIBuiltin = clang::Builtin::FirstTSBuiltin - 1,
-  #define BUILTIN(ID, TYPE, ATTRS) BI##ID,
-  #include "clang/Basic/BuiltinsBPF.inc"
+#define GET_BUILTIN_ENUMERATORS
+#include "clang/Basic/BuiltinsBPF.inc"
+#undef GET_BUILTIN_ENUMERATORS
     LastTSBuiltin
   };
   }
@@ -133,8 +134,9 @@ namespace clang {
   namespace X86 {
   enum {
     LastTIBuiltin = clang::Builtin::FirstTSBuiltin - 1,
-#define BUILTIN(ID, TYPE, ATTRS) BI##ID,
+#define GET_BUILTIN_ENUMERATORS
 #include "clang/Basic/BuiltinsX86.inc"
+#undef GET_BUILTIN_ENUMERATORS
     FirstX86_64Builtin,
     LastX86CommonBuiltin = FirstX86_64Builtin - 1,
 #define BUILTIN(ID, TYPE, ATTRS) BI##ID,
@@ -172,8 +174,9 @@ namespace clang {
     LastTIBuiltin = clang::Builtin::FirstTSBuiltin - 1,
     FirstRVVBuiltin = clang::Builtin::FirstTSBuiltin,
     LastRVVBuiltin = RISCVVector::FirstTSBuiltin - 1,
-#define BUILTIN(ID, TYPE, ATTRS) BI##ID,
+#define GET_BUILTIN_ENUMERATORS
 #include "clang/Basic/BuiltinsRISCV.inc"
+#undef GET_BUILTIN_ENUMERATORS
     LastTSBuiltin
   };
   } // namespace RISCV

clang/lib/AST/StmtPrinter.cpp

@@ -1928,7 +1928,6 @@ void StmtPrinter::VisitPseudoObjectExpr(PseudoObjectExpr *Node) {
 void StmtPrinter::VisitAtomicExpr(AtomicExpr *Node) {
   const char *Name = nullptr;
   switch (Node->getOp()) {
-#define BUILTIN(ID, TYPE, ATTRS)
 #define ATOMIC_BUILTIN(ID, TYPE, ATTRS) \
   case AtomicExpr::AO ## ID: \
     Name = #ID "("; \

clang/lib/Basic/Builtins.cpp

@@ -29,23 +29,19 @@ const char *HeaderDesc::getName() const {
   llvm_unreachable("Unknown HeaderDesc::HeaderID enum");
 }
 
-static constexpr llvm::StringTable BuiltinStrings =
-    CLANG_BUILTIN_STR_TABLE_START
-        // We inject a non-builtin string into the table.
-        CLANG_BUILTIN_STR_TABLE("not a builtin function", "", "")
-#define BUILTIN CLANG_BUILTIN_STR_TABLE
+static constexpr unsigned NumBuiltins = Builtin::FirstTSBuiltin;
+
+#define GET_BUILTIN_STR_TABLE
 #include "clang/Basic/Builtins.inc"
-    ;
-static_assert(BuiltinStrings.size() < 100'000);
-
-static constexpr auto BuiltinInfos =
-    Builtin::MakeInfos<Builtin::FirstTSBuiltin>(
-        {CLANG_BUILTIN_ENTRY("not a builtin function", "", "")
-#define BUILTIN CLANG_BUILTIN_ENTRY
-#define LANGBUILTIN CLANG_LANGBUILTIN_ENTRY
-#define LIBBUILTIN CLANG_LIBBUILTIN_ENTRY
+#undef GET_BUILTIN_STR_TABLE
+
+static constexpr Builtin::Info BuiltinInfos[] = {
+    Builtin::Info{}, // No-builtin info entry.
+#define GET_BUILTIN_INFOS
 #include "clang/Basic/Builtins.inc"
-        });
+#undef GET_BUILTIN_INFOS
+};
+static_assert(std::size(BuiltinInfos) == NumBuiltins);
 
 std::pair<const Builtin::InfosShard &, const Builtin::Info &>
 Builtin::Context::getShardAndInfo(unsigned ID) const {

clang/lib/Basic/Targets/BPF.cpp

@@ -22,16 +22,16 @@ using namespace clang::targets;
 static constexpr int NumBuiltins =
     clang::BPF::LastTSBuiltin - Builtin::FirstTSBuiltin;
 
-static constexpr llvm::StringTable BuiltinStrings =
-    CLANG_BUILTIN_STR_TABLE_START
-#define BUILTIN CLANG_BUILTIN_STR_TABLE
+#define GET_BUILTIN_STR_TABLE
 #include "clang/Basic/BuiltinsBPF.inc"
-    ;
+#undef GET_BUILTIN_STR_TABLE
 
-static constexpr auto BuiltinInfos = Builtin::MakeInfos<NumBuiltins>({
-#define BUILTIN CLANG_BUILTIN_ENTRY
+static constexpr Builtin::Info BuiltinInfos[] = {
+#define GET_BUILTIN_INFOS
 #include "clang/Basic/BuiltinsBPF.inc"
-});
+#undef GET_BUILTIN_INFOS
+};
+static_assert(std::size(BuiltinInfos) == NumBuiltins);
 
 void BPFTargetInfo::getTargetDefines(const LangOptions &Opts,
                                      MacroBuilder &Builder) const {

clang/lib/Basic/Targets/RISCV.cpp

@@ -277,18 +277,16 @@ static constexpr std::array<Builtin::Info, NumRVVSiFiveBuiltins> BuiltinInfos =
 };
 } // namespace RVVSiFive
 
-static constexpr llvm::StringTable BuiltinStrings =
-    CLANG_BUILTIN_STR_TABLE_START
-#define BUILTIN CLANG_BUILTIN_STR_TABLE
-#define TARGET_BUILTIN CLANG_TARGET_BUILTIN_STR_TABLE
+#define GET_BUILTIN_STR_TABLE
 #include "clang/Basic/BuiltinsRISCV.inc"
-    ;
+#undef GET_BUILTIN_STR_TABLE
 
-static constexpr auto BuiltinInfos = Builtin::MakeInfos<NumRISCVBuiltins>({
-#define BUILTIN CLANG_BUILTIN_ENTRY
-#define TARGET_BUILTIN CLANG_TARGET_BUILTIN_ENTRY
+static constexpr Builtin::Info BuiltinInfos[] = {
+#define GET_BUILTIN_INFOS
 #include "clang/Basic/BuiltinsRISCV.inc"
-});
+#undef GET_BUILTIN_INFOS
+};
+static_assert(std::size(BuiltinInfos) == NumRISCVBuiltins);
 
 llvm::SmallVector<Builtin::InfosShard>
 RISCVTargetInfo::getTargetBuiltins() const {

clang/lib/Basic/Targets/X86.cpp

@@ -31,13 +31,18 @@ static constexpr int NumX86_64Builtins =
 static constexpr int NumBuiltins = X86::LastTSBuiltin - Builtin::FirstTSBuiltin;
 static_assert(NumBuiltins == (NumX86Builtins + NumX86_64Builtins));
 
-static constexpr llvm::StringTable BuiltinX86Strings =
-    CLANG_BUILTIN_STR_TABLE_START
-#define BUILTIN CLANG_BUILTIN_STR_TABLE
-#define TARGET_BUILTIN CLANG_TARGET_BUILTIN_STR_TABLE
-#define TARGET_HEADER_BUILTIN CLANG_TARGET_HEADER_BUILTIN_STR_TABLE
+namespace X86 {
+#define GET_BUILTIN_STR_TABLE
 #include "clang/Basic/BuiltinsX86.inc"
-    ;
+#undef GET_BUILTIN_STR_TABLE
+
+static constexpr Builtin::Info BuiltinInfos[] = {
+#define GET_BUILTIN_INFOS
+#include "clang/Basic/BuiltinsX86.inc"
+#undef GET_BUILTIN_INFOS
+};
+static_assert(std::size(BuiltinInfos) == NumX86Builtins);
+} // namespace X86
 
 static constexpr llvm::StringTable BuiltinX86_64Strings =
     CLANG_BUILTIN_STR_TABLE_START
@@ -47,13 +52,6 @@ static constexpr llvm::StringTable BuiltinX86_64Strings =
 #include "clang/Basic/BuiltinsX86_64.def"
     ;
 
-static constexpr auto BuiltinX86Infos = Builtin::MakeInfos<NumX86Builtins>({
-#define BUILTIN CLANG_BUILTIN_ENTRY
-#define TARGET_BUILTIN CLANG_TARGET_BUILTIN_ENTRY
-#define TARGET_HEADER_BUILTIN CLANG_TARGET_HEADER_BUILTIN_ENTRY
-#include "clang/Basic/BuiltinsX86.inc"
-});
-
 static constexpr auto BuiltinX86_64Infos =
     Builtin::MakeInfos<NumX86_64Builtins>({
 #define BUILTIN CLANG_BUILTIN_ENTRY
@@ -1879,13 +1877,13 @@ ArrayRef<TargetInfo::AddlRegName> X86TargetInfo::getGCCAddlRegNames() const {
 
 llvm::SmallVector<Builtin::InfosShard>
 X86_32TargetInfo::getTargetBuiltins() const {
-  return {{&BuiltinX86Strings, BuiltinX86Infos}};
+  return {{&X86::BuiltinStrings, X86::BuiltinInfos}};
 }
 
 llvm::SmallVector<Builtin::InfosShard>
 X86_64TargetInfo::getTargetBuiltins() const {
   return {
-      {&BuiltinX86Strings, BuiltinX86Infos},
+      {&X86::BuiltinStrings, X86::BuiltinInfos},
       {&BuiltinX86_64Strings, BuiltinX86_64Infos},
   };
 }

clang/lib/Sema/SemaChecking.cpp

@@ -2446,7 +2446,6 @@ Sema::CheckBuiltinFunctionCall(FunctionDecl *FDecl, unsigned BuiltinID,
       CheckNonNullArgument(*this, TheCall->getArg(0), TheCall->getExprLoc());
     break;
   }
-#define BUILTIN(ID, TYPE, ATTRS)
 #define ATOMIC_BUILTIN(ID, TYPE, ATTRS)                                        \
   case Builtin::BI##ID:                                                        \
     return AtomicOpsOverloaded(TheCallResult, AtomicExpr::AO##ID);