tilelang.autotuner.grouped_compile
==================================

.. py:module:: tilelang.autotuner.grouped_compile

.. autoapi-nested-parse::

   Grouped compilation helpers for autotuner.

   This module isolates backend-aware grouped compilation logic from AutoTuner.run
   so tuner.py can stay focused on orchestration.



Attributes
----------

.. autoapisummary::

   tilelang.autotuner.grouped_compile.CompileUnitResult


Functions
---------

.. autoapisummary::

   tilelang.autotuner.grouped_compile.compile_grouped_unit_tvm_ffi


Module Contents
---------------

.. py:data:: CompileUnitResult

.. py:function:: compile_grouped_unit_tvm_ffi(unit_items, compile_args, elaborate_func)

   Compile one grouped unit for CUDA+tvm_ffi backend.

   Flow:
   1. Elaborate each config into a PrimFunc.
   2. Lower each PrimFunc into host/device IR modules.
   3. Merge all device IR into one IRModule and compile device code once.
   4. Build host runtime module per config and import shared device module.
   5. Construct per-config JITKernel objects that share the grouped device module.