[NFC][IR2Vec] Exposing helpers in IR2Vec Vocabulary

[svkeerthy]

Minor refactoring IR2Vec vocabulary. This would help in upcoming PRs related to the IR2Vec tool.

(Tracking issue - #141817)

[svkeerthy]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• [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 👈 (View in Graphite)
• [NFC][IR2Vec] Minor refactoring of opcode access in vocabulary #147585
• [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

Minor refactoring IR2Vec vocabulary. This would help in upcoming PRs related to the IR2Vec tool.

(Tracking issue - #141817)

---

Full diff: https://github.com/llvm/llvm-project/pull/147841.diff

2 Files Affected:

• (modified) llvm/include/llvm/Analysis/IR2Vec.h (+8-8)
• (modified) llvm/lib/Analysis/IR2Vec.cpp (+2-1)

diff --git a/llvm/include/llvm/Analysis/IR2Vec.h b/llvm/include/llvm/Analysis/IR2Vec.h
index 176cdaf7b5378..4209cf8d476b9 100644
--- a/llvm/include/llvm/Analysis/IR2Vec.h
+++ b/llvm/include/llvm/Analysis/IR2Vec.h
@@ -163,6 +163,14 @@ class Vocabulary {
   static constexpr unsigned MaxOperandKinds =
       static_cast<unsigned>(OperandKind::MaxOperandKind);
 
+public:
+  Vocabulary() = default;
+  Vocabulary(VocabVector &&Vocab);
+
+  bool isValid() const;
+  unsigned getDimension() const;
+  size_t size() const;
+
   /// Helper function to get vocabulary key for a given Opcode
   static StringRef getVocabKeyForOpcode(unsigned Opcode);
 
@@ -175,14 +183,6 @@ class Vocabulary {
   /// Helper function to classify an operand into OperandKind
   static OperandKind getOperandKind(const Value *Op);
 
-public:
-  Vocabulary() = default;
-  Vocabulary(VocabVector &&Vocab);
-
-  bool isValid() const;
-  unsigned getDimension() const;
-  size_t size() const;
-
   /// Accessors to get the embedding for a given entity.
   const ir2vec::Embedding &operator[](unsigned Opcode) const;
   const ir2vec::Embedding &operator[](Type::TypeID TypeId) const;
diff --git a/llvm/lib/Analysis/IR2Vec.cpp b/llvm/lib/Analysis/IR2Vec.cpp
index f97644b93a3d4..d1eb709012a57 100644
--- a/llvm/lib/Analysis/IR2Vec.cpp
+++ b/llvm/lib/Analysis/IR2Vec.cpp
@@ -518,7 +518,8 @@ IR2VecVocabAnalysis::run(Module &M, ModuleAnalysisManager &AM) {
   // Otherwise, try to read from the vocabulary file.
   if (VocabFile.empty()) {
     // FIXME: Use default vocabulary
-    Ctx->emitError("IR2Vec vocabulary file path not specified");
+    Ctx->emitError("IR2Vec vocabulary file path not specified; You may need to "
+                   "set it using --ir2vec-vocab-path");
     return Vocabulary(); // Return invalid result
   }
   if (auto Err = readVocabulary()) {

[boomanaiden154]

What's the point of moving these declarations? Keep them in the same order but also make the getVocabKeyFor* functions part of the public interface?

[boomanaiden154]

This doesn't really have anything to do with the patch (based on the description) and shouldn't really be here, but separating it out is probably too much of a hassle for what it is. Leaving it should be fine.