[NFC][IR2Vec] Minor refactoring of opcode access in vocabulary

[svkeerthy]

Refactored IR2Vec vocabulary handling to improve code organization and error handling. This would help in upcoming PRs related to the IR2Vec tool.

(Tracking issue - #141817)

[svkeerthy]

• [IR2Vec] Adding documentation for llvm-ir2vec tool #148719
• [IR2Vec] Add embeddings mode to llvm-ir2vec tool #147844
• [IR2Vec] Add llvm-ir2vec tool for generating triplet embeddings #147842
• [NFC][IR2Vec] Exposing helpers in IR2Vec Vocabulary #147841
• [NFC][IR2Vec] Minor refactoring of opcode access in vocabulary #147585 👈 (View in Graphite)
• [IR2Vec] Restructuring Vocabulary #145119
• [IR2Vec] Add out-of-place arithmetic operators to Embedding class #145118
• [NFC][IR2Vec] Increasing tolerance in approximatelyEquals() of Embedding  #145117
• [IR2Vec] Simplifying creation of Embedder #143999
• [IR2Vec] Scale embeddings once in vocab analysis instead of repetitive scaling #143986
• [MLGO][IR2Vec] Integrating IR2Vec with MLInliner #143479 : 1 other dependent PR (#144139 )
• [IR2Vec] Consider only reachable BBs and non-debug instructions #143476
• [IR2Vec] Minor vocab changes and exposing weights #143200
• [IR2Vec] Exposing Embedding as an data type wrapped around std::vector<double> #143197
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[llvmbot]

@llvm/pr-subscribers-llvm-analysis

@llvm/pr-subscribers-mlgo

Author: S. VenkataKeerthy (svkeerthy)

Changes

Refactored IR2Vec vocabulary handling to improve code organization and error handling. This would help in upcoming PRs related to the IR2Vec tool.

(Tracking issue - #141817)

---

Full diff: https://github.com/llvm/llvm-project/pull/147585.diff

2 Files Affected:

• (modified) llvm/include/llvm/Analysis/IR2Vec.h (+6-3)
• (modified) llvm/lib/Analysis/IR2Vec.cpp (+24-21)

diff --git a/llvm/include/llvm/Analysis/IR2Vec.h b/llvm/include/llvm/Analysis/IR2Vec.h
index f5a4e450cf160..176cdaf7b5378 100644
--- a/llvm/include/llvm/Analysis/IR2Vec.h
+++ b/llvm/include/llvm/Analysis/IR2Vec.h
@@ -163,15 +163,18 @@ class Vocabulary {
   static constexpr unsigned MaxOperandKinds =
       static_cast<unsigned>(OperandKind::MaxOperandKind);
 
+  /// Helper function to get vocabulary key for a given Opcode
+  static StringRef getVocabKeyForOpcode(unsigned Opcode);
+
+  /// Helper function to get vocabulary key for a given TypeID
+  static StringRef getVocabKeyForTypeID(Type::TypeID TypeID);
+
   /// Helper function to get vocabulary key for a given OperandKind
   static StringRef getVocabKeyForOperandKind(OperandKind Kind);
 
   /// Helper function to classify an operand into OperandKind
   static OperandKind getOperandKind(const Value *Op);
 
-  /// Helper function to get vocabulary key for a given TypeID
-  static StringRef getVocabKeyForTypeID(Type::TypeID TypeID);
-
 public:
   Vocabulary() = default;
   Vocabulary(VocabVector &&Vocab);
diff --git a/llvm/lib/Analysis/IR2Vec.cpp b/llvm/lib/Analysis/IR2Vec.cpp
index d3dc2e36fd23e..f97644b93a3d4 100644
--- a/llvm/lib/Analysis/IR2Vec.cpp
+++ b/llvm/lib/Analysis/IR2Vec.cpp
@@ -243,6 +243,17 @@ const ir2vec::Embedding &Vocabulary::operator[](const Value *Arg) const {
   return Vocab[MaxOpcodes + MaxTypeIDs + static_cast<unsigned>(ArgKind)];
 }
 
+StringRef Vocabulary::getVocabKeyForOpcode(unsigned Opcode) {
+  assert(Opcode >= 1 && Opcode <= MaxOpcodes && "Invalid opcode");
+#define HANDLE_INST(NUM, OPCODE, CLASS)                                        \
+  if (Opcode == NUM) {                                                         \
+    return #OPCODE;                                                            \
+  }
+#include "llvm/IR/Instruction.def"
+#undef HANDLE_INST
+  return "UnknownOpcode";
+}
+
 StringRef Vocabulary::getVocabKeyForTypeID(Type::TypeID TypeID) {
   switch (TypeID) {
   case Type::VoidTyID:
@@ -280,6 +291,7 @@ StringRef Vocabulary::getVocabKeyForTypeID(Type::TypeID TypeID) {
   default:
     return "UnknownTy";
   }
+  return "UnknownTy";
 }
 
 // Operand kinds supported by IR2Vec - string mappings
@@ -297,9 +309,9 @@ StringRef Vocabulary::getVocabKeyForOperandKind(Vocabulary::OperandKind Kind) {
     OPERAND_KINDS
 #undef OPERAND_KIND
   case Vocabulary::OperandKind::MaxOperandKind:
-    llvm_unreachable("Invalid OperandKind");
+    return "UnknownOperand";
   }
-  llvm_unreachable("Unknown OperandKind");
+  return "UnknownOperand";
 }
 
 #undef OPERAND_KINDS
@@ -332,14 +344,8 @@ StringRef Vocabulary::getStringKey(unsigned Pos) {
   assert(Pos < MaxOpcodes + MaxTypeIDs + MaxOperandKinds &&
          "Position out of bounds in vocabulary");
   // Opcode
-  if (Pos < MaxOpcodes) {
-#define HANDLE_INST(NUM, OPCODE, CLASS)                                        \
-  if (Pos == NUM - 1) {                                                        \
-    return #OPCODE;                                                            \
-  }
-#include "llvm/IR/Instruction.def"
-#undef HANDLE_INST
-  }
+  if (Pos < MaxOpcodes)
+    return getVocabKeyForOpcode(Pos + 1);
   // Type
   if (Pos < MaxOpcodes + MaxTypeIDs)
     return getVocabKeyForTypeID(static_cast<Type::TypeID>(Pos - MaxOpcodes));
@@ -447,21 +453,18 @@ void IR2VecVocabAnalysis::generateNumMappedVocab() {
   // Handle Opcodes
   std::vector<Embedding> NumericOpcodeEmbeddings(Vocabulary::MaxOpcodes,
                                                  Embedding(Dim, 0));
-#define HANDLE_INST(NUM, OPCODE, CLASS)                                        \
-  {                                                                            \
-    auto It = OpcVocab.find(#OPCODE);                                          \
-    if (It != OpcVocab.end())                                                  \
-      NumericOpcodeEmbeddings[NUM - 1] = It->second;                           \
-    else                                                                       \
-      handleMissingEntity(#OPCODE);                                            \
+  for (unsigned Opcode : seq(0u, Vocabulary::MaxOpcodes)) {
+    StringRef VocabKey = Vocabulary::getVocabKeyForOpcode(Opcode + 1);
+    auto It = OpcVocab.find(VocabKey.str());
+    if (It != OpcVocab.end())
+      NumericOpcodeEmbeddings[Opcode] = It->second;
+    else
+      handleMissingEntity(VocabKey.str());
   }
-#include "llvm/IR/Instruction.def"
-#undef HANDLE_INST
   Vocab.insert(Vocab.end(), NumericOpcodeEmbeddings.begin(),
                NumericOpcodeEmbeddings.end());
 
-  // Handle Types using direct iteration through TypeID enum
-  // We iterate through all possible TypeID values and map them to embeddings
+  // Handle Types
   std::vector<Embedding> NumericTypeEmbeddings(Vocabulary::MaxTypeIDs,
                                                Embedding(Dim, 0));
   for (unsigned TypeID : seq(0u, Vocabulary::MaxTypeIDs)) {

[svkeerthy]

Merge activity

• Jul 14, 11:28 PM UTC: A user started a stack merge that includes this pull request via Graphite.
• Jul 14, 11:30 PM UTC: Graphite rebased this pull request as part of a merge.
• Jul 14, 11:33 PM UTC: Graphite rebased this pull request as part of a merge.
• Jul 14, 11:35 PM UTC: @svkeerthy merged this pull request with Graphite.