[RFC][TableGen] Restructure TableGen Source

Lately, while trying to move as much code out of the GlobalISel Combiner Backend, I've been hitting more and more linker issues. It seems like the
library structure was due for an overhaul, so I created #80647

I'm proposing this refactor of the llvm-tblgen source into:
- a "Basic" library, which contains the bare minimum utilities to build `llvm-min-tablegen`
- a "Common" library which contains all of the helpers for TableGen backends. Such helpers can be shared by more than one backend, and even unit tested (e.g. CodeExpander is, maybe we can add more over time)

The backends remain in place and just make use of the new include paths.

Solves #80647

Author: Pierre-vh

llvm/unittests/TableGen/CMakeLists.txt

@@ -15,4 +15,4 @@ add_llvm_unittest(TableGenTests DISABLE_LLVM_LINK_LLVM_DYLIB
   ParserEntryPointTest.cpp
   )
 
-target_link_libraries(TableGenTests PRIVATE LLVMTableGenGlobalISel LLVMTableGen)
+target_link_libraries(TableGenTests PRIVATE LLVMTableGenCommon LLVMTableGen)

llvm/unittests/TableGen/CodeExpanderTest.cpp

@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "GlobalISel/CodeExpander.h"
-#include "GlobalISel/CodeExpansions.h"
+#include "Common/GlobalISel/CodeExpander.h"
+#include "Common/GlobalISel/CodeExpansions.h"
 
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/TableGen/Error.h"

llvm/utils/TableGen/AsmMatcherEmitter.cpp

@@ -95,12 +95,12 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenInstAlias.h"
-#include "CodeGenInstruction.h"
-#include "CodeGenRegisters.h"
-#include "CodeGenTarget.h"
-#include "SubtargetFeatureInfo.h"
-#include "Types.h"
+#include "Common/CodeGenInstAlias.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenRegisters.h"
+#include "Common/CodeGenTarget.h"
+#include "Common/SubtargetFeatureInfo.h"
+#include "Common/Types.h"
 #include "llvm/ADT/CachedHashString.h"
 #include "llvm/ADT/PointerUnion.h"
 #include "llvm/ADT/STLExtras.h"

llvm/utils/TableGen/AsmWriterEmitter.cpp

@@ -11,13 +11,13 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "AsmWriterInst.h"
-#include "CodeGenInstAlias.h"
-#include "CodeGenInstruction.h"
-#include "CodeGenRegisters.h"
-#include "CodeGenTarget.h"
-#include "SequenceToOffsetTable.h"
-#include "Types.h"
+#include "Basic/SequenceToOffsetTable.h"
+#include "Common/AsmWriterInst.h"
+#include "Common/CodeGenInstAlias.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenRegisters.h"
+#include "Common/CodeGenTarget.h"
+#include "Common/Types.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/STLExtras.h"

llvm/utils/TableGen/Basic/CMakeLists.txt

@@ -0,0 +1,21 @@
+# The basic TableGen library contains as little dependencies as possible.
+# In particular, it does not depend on vt_gen -> it does not use ValueTypes.
+#
+# This library is the only thing included in `llvm-min-tablegen`.
+
+set(LLVM_LINK_COMPONENTS
+  Support
+  TableGen
+  )
+
+add_llvm_library(LLVMTableGenBasic STATIC OBJECT EXCLUDE_FROM_ALL
+  CodeGenIntrinsics.cpp
+  SDNodeProperties.cpp
+)
+set_target_properties(LLVMTableGenBasic PROPERTIES FOLDER "Tablegenning")
+
+# Users may include its headers as "Basic/*.h"
+target_include_directories(LLVMTableGenBasic
+  INTERFACE
+  $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/..>
+  )

llvm/utils/TableGen/CodeGenIntrinsics.cpp renamed to llvm/utils/TableGen/Basic/CodeGenIntrinsics.cpp

@@ -1,26 +1,25 @@
-add_subdirectory(GlobalISel)
+# Basic utilities which is the strict minimum needed to build
+# llvm-min-tblgen.
+add_subdirectory(Basic)
+# Common utilities are all of the reusable components and helper
+# code needed by the backends.
+add_subdirectory(Common)
 
-add_llvm_library(LLVMTableGenCommon STATIC OBJECT EXCLUDE_FROM_ALL
+set(LLVM_LINK_COMPONENTS Support)
+
+# llvm-min-tablegen only contains a subset of backends necessary to
+# build llvm/include. It must not depend on TableGenCommon, as
+# TableGenCommon depends on this already to generate things such as
+# ValueType definitions.
+add_tablegen(llvm-min-tblgen LLVM_HEADERS
+  TableGen.cpp
   Attributes.cpp
-  CodeGenIntrinsics.cpp
   DirectiveEmitter.cpp
   IntrinsicEmitter.cpp
   RISCVTargetDefEmitter.cpp
-  SDNodeProperties.cpp
   VTEmitter.cpp
-  PARTIAL_SOURCES_INTENDED
-
-  LINK_COMPONENTS
-  Support
-  TableGen
-  )
-set_target_properties(LLVMTableGenCommon PROPERTIES FOLDER "Tablegenning")
+  $<TARGET_OBJECTS:obj.LLVMTableGenBasic>
 
-set(LLVM_LINK_COMPONENTS Support)
-
-add_tablegen(llvm-min-tblgen LLVM_HEADERS
-  TableGen.cpp
-  $<TARGET_OBJECTS:obj.LLVMTableGenCommon>
   PARTIAL_SOURCES_INTENDED
   )
 set_target_properties(llvm-min-tblgen PROPERTIES FOLDER "Tablegenning")
@@ -35,63 +34,51 @@ add_tablegen(llvm-tblgen LLVM
   EXPORT LLVM
   AsmMatcherEmitter.cpp
   AsmWriterEmitter.cpp
-  AsmWriterInst.cpp
-  CTagsEmitter.cpp
+  Attributes.cpp
   CallingConvEmitter.cpp
   CodeEmitterGen.cpp
-  CodeGenDAGPatterns.cpp
-  CodeGenHwModes.cpp
-  CodeGenInstAlias.cpp
-  CodeGenInstruction.cpp
   CodeGenMapTable.cpp
-  CodeGenRegisters.cpp
-  CodeGenSchedule.cpp
-  CodeGenTarget.cpp
+  CompressInstEmitter.cpp
+  CTagsEmitter.cpp
   DAGISelEmitter.cpp
   DAGISelMatcherEmitter.cpp
   DAGISelMatcherGen.cpp
   DAGISelMatcherOpt.cpp
-  DAGISelMatcher.cpp
   DecoderEmitter.cpp
   DFAEmitter.cpp
   DFAPacketizerEmitter.cpp
+  DirectiveEmitter.cpp
   DisassemblerEmitter.cpp
   DXILEmitter.cpp
   ExegesisEmitter.cpp
   FastISelEmitter.cpp
   GlobalISelCombinerEmitter.cpp
   GlobalISelEmitter.cpp
-  GlobalISelMatchTable.cpp
-  GlobalISelMatchTableExecutorEmitter.cpp
-  InfoByHwMode.cpp
-  InstrInfoEmitter.cpp
   InstrDocsEmitter.cpp
-  OptEmitter.cpp
+  InstrInfoEmitter.cpp
+  IntrinsicEmitter.cpp
+  MacroFusionPredicatorEmitter.cpp
   OptParserEmitter.cpp
   OptRSTEmitter.cpp
-  PredicateExpander.cpp
   PseudoLoweringEmitter.cpp
-  CompressInstEmitter.cpp
-  MacroFusionPredicatorEmitter.cpp
   RegisterBankEmitter.cpp
   RegisterInfoEmitter.cpp
+  RISCVTargetDefEmitter.cpp
   SearchableTableEmitter.cpp
   SubtargetEmitter.cpp
-  SubtargetFeatureInfo.cpp
   TableGen.cpp
-  Types.cpp
-  VarLenCodeEmitterGen.cpp
-  X86DisassemblerTables.cpp
+  VTEmitter.cpp
+  WebAssemblyDisassemblerEmitter.cpp
   X86CompressEVEXTablesEmitter.cpp
+  X86DisassemblerTables.cpp
   X86FoldTablesEmitter.cpp
   X86MnemonicTables.cpp
   X86ModRMFilters.cpp
   X86RecognizableInstr.cpp
-  WebAssemblyDisassemblerEmitter.cpp
   $<TARGET_OBJECTS:obj.LLVMTableGenCommon>
 
   DEPENDS
   intrinsics_gen # via llvm-min-tablegen
   )
-target_link_libraries(llvm-tblgen PRIVATE LLVMTableGenGlobalISel)
+target_link_libraries(llvm-tblgen PRIVATE LLVMTableGenCommon)
 set_target_properties(llvm-tblgen PROPERTIES FOLDER "Tablegenning")

llvm/utils/TableGen/CodeGenIntrinsics.h renamed to llvm/utils/TableGen/Basic/CodeGenIntrinsics.h

@@ -11,7 +11,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenTarget.h"
+#include "Common/CodeGenTarget.h"
 #include "llvm/TableGen/Error.h"
 #include "llvm/TableGen/Record.h"
 #include "llvm/TableGen/TableGenBackend.h"

llvm/utils/TableGen/SDNodeProperties.cpp renamed to llvm/utils/TableGen/Basic/SDNodeProperties.cpp

@@ -22,11 +22,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenHwModes.h"
-#include "CodeGenInstruction.h"
-#include "CodeGenTarget.h"
-#include "InfoByHwMode.h"
-#include "VarLenCodeEmitterGen.h"
+#include "Common/CodeGenHwModes.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenTarget.h"
+#include "Common/InfoByHwMode.h"
+#include "Common/VarLenCodeEmitterGen.h"
 #include "llvm/ADT/APInt.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/StringExtras.h"

llvm/utils/TableGen/SDNodeProperties.h renamed to llvm/utils/TableGen/Basic/SDNodeProperties.h

@@ -75,8 +75,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenInstruction.h"
-#include "CodeGenTarget.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenTarget.h"
 #include "llvm/TableGen/Error.h"
 #include "llvm/TableGen/Record.h"
 using namespace llvm;

llvm/utils/TableGen/SequenceToOffsetTable.h renamed to llvm/utils/TableGen/Basic/SequenceToOffsetTable.h

@@ -0,0 +1,48 @@
+# The common library is similar to the basic library except it can
+# depend on vt_gen.
+#
+# This library contains the bulk of the supporting code for all
+# TableGen backends. It's split off as a separate library to
+# allow unit-testing those components.
+
+set(LLVM_LINK_COMPONENTS
+  Support
+  TableGen
+  )
+
+add_llvm_library(LLVMTableGenCommon STATIC OBJECT EXCLUDE_FROM_ALL
+  GlobalISel/CodeExpander.cpp
+  GlobalISel/CXXPredicates.cpp
+  GlobalISel/GlobalISelMatchTable.cpp
+  GlobalISel/GlobalISelMatchTableExecutorEmitter.cpp
+  GlobalISel/MatchDataInfo.cpp
+  GlobalISel/Patterns.cpp
+
+  AsmWriterInst.cpp
+  CodeGenDAGPatterns.cpp
+  CodeGenHwModes.cpp
+  CodeGenInstAlias.cpp
+  CodeGenInstruction.cpp
+  CodeGenRegisters.cpp
+  CodeGenSchedule.cpp
+  CodeGenTarget.cpp
+  DAGISelMatcher.cpp
+  InfoByHwMode.cpp
+  OptEmitter.cpp
+  PredicateExpander.cpp
+  SubtargetFeatureInfo.cpp
+  Types.cpp
+  VarLenCodeEmitterGen.cpp
+  $<TARGET_OBJECTS:obj.LLVMTableGenBasic>
+
+  DEPENDS
+  vt_gen
+  )
+set_target_properties(LLVMTableGenCommon PROPERTIES FOLDER "Tablegenning")
+target_link_libraries(LLVMTableGenCommon PUBLIC LLVMTableGenBasic)
+
+# Users may include its headers as "Common/*.h"
+target_include_directories(LLVMTableGenCommon
+  PUBLIC
+  $<BUILD_INTERFACE:${CMAKE_CURRENT_SOURCE_DIR}/..>
+  )

llvm/utils/TableGen/CMakeLists.txt

@@ -14,9 +14,9 @@
 #ifndef LLVM_UTILS_TABLEGEN_CODEGENDAGPATTERNS_H
 #define LLVM_UTILS_TABLEGEN_CODEGENDAGPATTERNS_H
 
-#include "CodeGenIntrinsics.h"
+#include "Basic/CodeGenIntrinsics.h"
+#include "Basic/SDNodeProperties.h"
 #include "CodeGenTarget.h"
-#include "SDNodeProperties.h"
 #include "llvm/ADT/IntrusiveRefCntPtr.h"
 #include "llvm/ADT/MapVector.h"
 #include "llvm/ADT/PointerUnion.h"

llvm/utils/TableGen/CallingConvEmitter.cpp

@@ -19,7 +19,7 @@
 #include "CodeGenHwModes.h"
 #include "CodeGenInstruction.h"
 #include "InfoByHwMode.h"
-#include "SDNodeProperties.h"
+#include "Basic/SDNodeProperties.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallVector.h"

llvm/utils/TableGen/CodeEmitterGen.cpp

@@ -7,8 +7,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "GlobalISelMatchTable.h"
-#include "CodeGenInstruction.h"
-#include "CodeGenRegisters.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenRegisters.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/LEB128.h"

llvm/utils/TableGen/CodeGenMapTable.cpp

@@ -16,7 +16,7 @@
 #ifndef LLVM_UTILS_TABLEGEN_GLOBALISELMATCHTABLE_H
 #define LLVM_UTILS_TABLEGEN_GLOBALISELMATCHTABLE_H
 
-#include "CodeGenDAGPatterns.h"
+#include "Common/CodeGenDAGPatterns.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallPtrSet.h"

llvm/utils/TableGen/AsmWriterInst.cpp renamed to llvm/utils/TableGen/Common/AsmWriterInst.cpp

@@ -15,7 +15,7 @@
 #ifndef LLVM_UTILS_TABLEGEN_GLOBALISELMATCHTABLEEXECUTOREMITTER_H
 #define LLVM_UTILS_TABLEGEN_GLOBALISELMATCHTABLEEXECUTOREMITTER_H
 
-#include "SubtargetFeatureInfo.h"
+#include "Common/SubtargetFeatureInfo.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/Twine.h"

llvm/utils/TableGen/AsmWriterInst.h renamed to llvm/utils/TableGen/Common/AsmWriterInst.h

@@ -7,11 +7,11 @@
 //===----------------------------------------------------------------------===//
 
 #include "Patterns.h"
-#include "../CodeGenInstruction.h"
-#include "../CodeGenIntrinsics.h"
+#include "Basic/CodeGenIntrinsics.h"
 #include "CXXPredicates.h"
 #include "CodeExpander.h"
 #include "CodeExpansions.h"
+#include "Common/CodeGenInstruction.h"
 #include "llvm/ADT/StringSet.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"

llvm/utils/TableGen/Common/CMakeLists.txt

@@ -64,9 +64,9 @@
 
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenInstruction.h"
-#include "CodeGenRegisters.h"
-#include "CodeGenTarget.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenRegisters.h"
+#include "Common/CodeGenTarget.h"
 #include "llvm/ADT/IndexedMap.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringMap.h"

llvm/utils/TableGen/CodeGenDAGPatterns.cpp renamed to llvm/utils/TableGen/Common/CodeGenDAGPatterns.cpp

@@ -10,10 +10,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenDAGPatterns.h"
-#include "CodeGenInstruction.h"
-#include "CodeGenTarget.h"
-#include "DAGISelMatcher.h"
+#include "Common/CodeGenDAGPatterns.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenTarget.h"
+#include "Common/DAGISelMatcher.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/TableGen/Record.h"
 #include "llvm/TableGen/TableGenBackend.h"

llvm/utils/TableGen/CodeGenDAGPatterns.h renamed to llvm/utils/TableGen/Common/CodeGenDAGPatterns.h

@@ -10,12 +10,12 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "CodeGenDAGPatterns.h"
-#include "CodeGenInstruction.h"
-#include "CodeGenRegisters.h"
-#include "CodeGenTarget.h"
-#include "DAGISelMatcher.h"
-#include "SDNodeProperties.h"
+#include "Basic/SDNodeProperties.h"
+#include "Common/CodeGenDAGPatterns.h"
+#include "Common/CodeGenInstruction.h"
+#include "Common/CodeGenRegisters.h"
+#include "Common/CodeGenTarget.h"
+#include "Common/DAGISelMatcher.h"
 #include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/MapVector.h"
 #include "llvm/ADT/StringMap.h"