From c075fb8c37856365fb76d986ad3aefa2400b3240 Mon Sep 17 00:00:00 2001
From: Hank <49036880+hankluo6@users.noreply.github.com>
Date: Wed, 20 Aug 2025 06:03:26 -0700
Subject: [MLIR] Fix duplicated attribute nodes in MLIR bytecode
 deserialization (#151267)

Fixes #150163

MLIR bytecode does not preserve alias definitions, so each attribute
encountered during deserialization is treated as a new one. This can
generate duplicate `DISubprogram` nodes during deserialization.

The patch adds a `StringMap` cache that records attributes and fetches
them when encountered again.
---
 mlir/lib/Bytecode/Reader/BytecodeReader.cpp | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

(limited to 'mlir/lib/Bytecode/Reader/BytecodeReader.cpp')

diff --git a/mlir/lib/Bytecode/Reader/BytecodeReader.cpp b/mlir/lib/Bytecode/Reader/BytecodeReader.cpp
index 44458d0..0f97443 100644
--- a/mlir/lib/Bytecode/Reader/BytecodeReader.cpp
+++ b/mlir/lib/Bytecode/Reader/BytecodeReader.cpp
@@ -895,6 +895,10 @@ private:
   SmallVector<AttrEntry> attributes;
   SmallVector<TypeEntry> types;
 
+  /// The map of cached attributes, used to avoid re-parsing the same
+  /// attribute multiple times.
+  llvm::StringMap<Attribute> attributesCache;
+
   /// A location used for error emission.
   Location fileLoc;
 
@@ -1235,7 +1239,7 @@ LogicalResult AttrTypeReader::parseAsmEntry(T &result, EncodingReader &reader,
         ::parseType(asmStr, context, &numRead, /*isKnownNullTerminated=*/true);
   else
     result = ::parseAttribute(asmStr, context, Type(), &numRead,
-                              /*isKnownNullTerminated=*/true);
+                              /*isKnownNullTerminated=*/true, &attributesCache);
   if (!result)
     return failure();
 
-- 
cgit v1.1