[llvm] [IR2Vec] Add support for Cmp predicates in vocabulary and embeddings (PR #156952)

[Mircea Trofin via llvm-commits]

================
@@ -162,16 +163,25 @@ using BBEmbeddingsMap = DenseMap<const BasicBlock *, Embedding>;
 /// embeddings.
 class Vocabulary {
   friend class llvm::IR2VecVocabAnalysis;
+  // Slot layout:
+  // [0 .. MaxOpcodes-1]                                   => Instruction
+  // opcodes [MaxOpcodes .. MaxOpcodes+MaxCanonicalTypeIDs-1]      =>
+  // Canonicalized types [MaxOpcodes+MaxCanonicalTypeIDs .. end of operands) =>
----------------
mtrofin wrote:

can you explain what canonicalized means here?

https://github.com/llvm/llvm-project/pull/156952