From 437f3910383369ababcb30ca2a2b831f0bf9e664 Mon Sep 17 00:00:00 2001
From: Nishant Sachdeva <32475507+nishant-sachdeva@users.noreply.github.com>
Date: Wed, 21 Jan 2026 13:11:45 +0530
Subject: [PATCH] [IR2Vec] llvm-ir2vec python bindings scaffolding (#176571)

This patch adds the build infrastructure for Python bindings to
llvm-ir2vec.
Addresses https://github.com/llvm/llvm-project/issues/141839

Changes:
- Add `LLVM_IR2VEC_ENABLE_PYTHON_BINDINGS` CMake option (default OFF)
- Create minimal python module using nanobind
- Add lit test configuration that skips when the bindings have not been
requested

Build requires nanobind and is enabled via `cmake
-DLLVM_IR2VEC_ENABLE_PYTHON_BINDINGS=ON ... `

Future patches will add actual IR2Vec API functionality.
---
 llvm/test/lit.site.cfg.py.in                       |  1 +
 .../tools/llvm-ir2vec/bindings/ir2vec-bindings.py  |  7 +++++++
 llvm/test/tools/llvm-ir2vec/bindings/lit.local.cfg |  9 +++++++++
 llvm/tools/llvm-ir2vec/Bindings/CMakeLists.txt     | 14 ++++++++++++++
 llvm/tools/llvm-ir2vec/Bindings/PyIR2Vec.cpp       | 13 +++++++++++++
 llvm/tools/llvm-ir2vec/CMakeLists.txt              | 13 ++++++++++++-
 6 files changed, 56 insertions(+), 1 deletion(-)
 create mode 100644 llvm/test/tools/llvm-ir2vec/bindings/ir2vec-bindings.py
 create mode 100644 llvm/test/tools/llvm-ir2vec/bindings/lit.local.cfg
 create mode 100644 llvm/tools/llvm-ir2vec/Bindings/CMakeLists.txt
 create mode 100644 llvm/tools/llvm-ir2vec/Bindings/PyIR2Vec.cpp

diff --git a/llvm/test/lit.site.cfg.py.in b/llvm/test/lit.site.cfg.py.in
index 193f351ef52e..6fe48f21cc68 100644
--- a/llvm/test/lit.site.cfg.py.in
+++ b/llvm/test/lit.site.cfg.py.in
@@ -7,6 +7,7 @@ config.target_triple = "@LLVM_TARGET_TRIPLE@"
 config.llvm_src_root = path(r"@LLVM_SOURCE_DIR@")
 config.llvm_obj_root = path(r"@LLVM_BINARY_DIR@")
 config.llvm_tools_dir = lit_config.substitute(path(r"@LLVM_TOOLS_DIR@"))
+config.llvm_ir2vec_test_python_bindings = @LLVM_IR2VEC_TEST_PYTHON_BINDINGS@
 config.llvm_lib_dir = lit_config.substitute(path(r"@LLVM_LIBS_DIR@"))
 config.llvm_shlib_dir = lit_config.substitute(path(r"@SHLIBDIR@"))
 config.llvm_shlib_ext = "@SHLIBEXT@"
diff --git a/llvm/test/tools/llvm-ir2vec/bindings/ir2vec-bindings.py b/llvm/test/tools/llvm-ir2vec/bindings/ir2vec-bindings.py
new file mode 100644
index 000000000000..459f353a478c
--- /dev/null
+++ b/llvm/test/tools/llvm-ir2vec/bindings/ir2vec-bindings.py
@@ -0,0 +1,7 @@
+# RUN: env PYTHONPATH=%llvm_lib_dir %python %s
+
+import ir2vec
+
+print("SUCCESS: Module imported")
+
+# CHECK: SUCCESS: Module imported
diff --git a/llvm/test/tools/llvm-ir2vec/bindings/lit.local.cfg b/llvm/test/tools/llvm-ir2vec/bindings/lit.local.cfg
new file mode 100644
index 000000000000..af02c9e43492
--- /dev/null
+++ b/llvm/test/tools/llvm-ir2vec/bindings/lit.local.cfg
@@ -0,0 +1,9 @@
+import sys
+
+# Only run if user explicitly enabled bindings
+if not config.llvm_ir2vec_test_python_bindings:
+    config.unsupported = True
+else:
+    config.substitutions.append(('%python', sys.executable))
+    config.substitutions.append(('%llvm_lib_dir', config.llvm_lib_dir))
+    config.suffixes = ['.py']
diff --git a/llvm/tools/llvm-ir2vec/Bindings/CMakeLists.txt b/llvm/tools/llvm-ir2vec/Bindings/CMakeLists.txt
new file mode 100644
index 000000000000..5df3720a2477
--- /dev/null
+++ b/llvm/tools/llvm-ir2vec/Bindings/CMakeLists.txt
@@ -0,0 +1,14 @@
+find_package(Python COMPONENTS Interpreter Development.Module REQUIRED)
+
+execute_process(
+    COMMAND "${Python_EXECUTABLE}" -m nanobind --cmake_dir
+    OUTPUT_STRIP_TRAILING_WHITESPACE OUTPUT_VARIABLE nanobind_ROOT
+)
+find_package(nanobind CONFIG REQUIRED)
+
+set_target_properties(LLVMEmbUtils PROPERTIES POSITION_INDEPENDENT_CODE ON)
+
+nanobind_add_module(ir2vec MODULE PyIR2Vec.cpp)
+target_link_libraries(ir2vec PRIVATE LLVMEmbUtils)
+
+message(STATUS "Python bindings for llvm-ir2vec will be built with nanobind")
diff --git a/llvm/tools/llvm-ir2vec/Bindings/PyIR2Vec.cpp b/llvm/tools/llvm-ir2vec/Bindings/PyIR2Vec.cpp
new file mode 100644
index 000000000000..b3a46d429b6d
--- /dev/null
+++ b/llvm/tools/llvm-ir2vec/Bindings/PyIR2Vec.cpp
@@ -0,0 +1,13 @@
+//===- PyIR2Vec.cpp - Python Bindings for IR2Vec ------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include <nanobind/nanobind.h>
+
+namespace nb = nanobind;
+
+NB_MODULE(ir2vec, m) { m.doc() = "Python bindings for IR2Vec"; }
diff --git a/llvm/tools/llvm-ir2vec/CMakeLists.txt b/llvm/tools/llvm-ir2vec/CMakeLists.txt
index 018636aa536e..694dc77e274b 100644
--- a/llvm/tools/llvm-ir2vec/CMakeLists.txt
+++ b/llvm/tools/llvm-ir2vec/CMakeLists.txt
@@ -19,7 +19,7 @@ set(LLVM_LINK_COMPONENTS
   TargetParser
   )
 
-# Add the utils subdirectory for the library
+# Add the lib subdirectory
 add_subdirectory(lib)
 
 # Main executable
@@ -31,3 +31,14 @@ add_llvm_tool(llvm-ir2vec
 )
 
 target_link_libraries(llvm-ir2vec PRIVATE LLVMEmbUtils)
+
+option(LLVM_IR2VEC_ENABLE_PYTHON_BINDINGS 
+  "Build Python bindings for llvm-ir2vec" 
+  OFF)
+
+set(LLVM_IR2VEC_TEST_PYTHON_BINDINGS False CACHE INTERNAL "")
+
+if(LLVM_IR2VEC_ENABLE_PYTHON_BINDINGS)
+  add_subdirectory(Bindings)
+  set(LLVM_IR2VEC_TEST_PYTHON_BINDINGS True CACHE INTERNAL "" FORCE)
+endif()