Inductor cpp wrapper: support None as output (#88560)

Map `None` to `at::Tensor()` in the cpp wrapper

Pull Request resolved: https://github.com/pytorch/pytorch/pull/88560
Approved by: https://github.com/jgong5, https://github.com/jansel, https://github.com/desertfire

test/inductor/test_torchinductor.py

@@ -4920,6 +4920,7 @@ class CommonTemplate:
         for name in [
             "test_as_strided",  # buffer reuse
             "test_cat",  # alias
+            "test_lowmem_dropout1",  # None as output
             "test_profiler_mark_wrapper_call",  # TODO: fallback to default wrapper for now
             "test_relu",  # multiple inputs
             "test_silu",  # single input, single output

torch/_inductor/codegen/wrapper.py

@@ -8,7 +8,7 @@ from typing import Any, Dict, List
 
 from .. import codecache, config, ir
 from ..codecache import cpp_compile_command, get_code_path
-from ..utils import dynamo_utils, has_triton, sympy_dot, sympy_product
+from ..utils import cache_on_self, dynamo_utils, has_triton, sympy_dot, sympy_product
 from ..virtualized import V
 from .common import CodeGen, DeferredLine, IndentedBuffer, Kernel
 from .triton import texpr
@@ -312,6 +312,10 @@ class WrapperCodeGen(CodeGen):
             self.write_get_cuda_stream
         )
 
+    @cache_on_self
+    def get_output_refs(self):
+        return [x.codegen_reference() for x in V.graph.graph_outputs]
+
     def write_prefix(self):
         self.prefix.splice(
             """
@@ -470,7 +474,7 @@ class WrapperCodeGen(CodeGen):
                 else:
                     self.wrapper_call.writeline(line)
 
-            output_refs = [x.codegen_reference() for x in V.graph.graph_outputs]
+            output_refs = self.get_output_refs()
             if config.triton.debug_sync_graph:
                 self.wrapper_call.writeline("torch.cuda.synchronize()")
             self.generate_return(output_refs)
@@ -562,6 +566,20 @@ class CppWrapperCodeGen(WrapperCodeGen):
         self._call_func_id = next(CppWrapperCodeGen.call_func_id)
         super().__init__()
 
+    @cache_on_self
+    def get_output_refs(self):
+        def has_cpp_codegen_func(x):
+            return hasattr(x, "cpp_wrapper_codegen_reference") and callable(
+                x.cpp_wrapper_codegen_reference
+            )
+
+        return [
+            x.cpp_wrapper_codegen_reference()
+            if has_cpp_codegen_func(x)
+            else x.codegen_reference()
+            for x in V.graph.graph_outputs
+        ]
+
     def write_prefix(self):
         self.prefix.splice(
             """
@@ -576,7 +594,7 @@ class CppWrapperCodeGen(WrapperCodeGen):
         )
         with self.wrapper_call.indent():
             inputs_len = len(V.graph.graph_inputs.keys())
-            output_refs = [x.codegen_reference() for x in V.graph.graph_outputs]
+            output_refs = self.get_output_refs()
             if output_refs:
                 if len(output_refs) == 1:
                     output_types = "at::Tensor"

torch/_inductor/graph.py

@@ -404,11 +404,6 @@ class GraphLowering(torch.fx.Interpreter):
             if value.get_dtype() != torch.float32:
                 self.disable_cpp_wrapper("inputs not FP32")
 
-    def check_output_for_cpp_buffer(self):
-        for item in self.graph_outputs:
-            if isinstance(item, ir.NoneAsConstantBuffer):
-                self.disable_cpp_wrapper("NoneAsConstantBuffer")
-
     def check_constant_for_cpp_buffer(self):
         if self.constants:
             self.disable_cpp_wrapper("Constants")
@@ -418,7 +413,6 @@ class GraphLowering(torch.fx.Interpreter):
         self.check_profiler_mark_wrapper_call()
         self.check_device_for_cpp_buffer()
         self.check_input_for_cpp_buffer()
-        self.check_output_for_cpp_buffer()
         self.check_constant_for_cpp_buffer()
 
     def init_wrapper_code(self):

torch/_inductor/ir.py

@@ -1998,6 +1998,9 @@ class NoneAsConstantBuffer(IRNode):
     def codegen_reference(self):
         return "None"
 
+    def cpp_wrapper_codegen_reference(self):
+        return "at::Tensor()"
+
 
 class ShapeAsConstantBuffer(IRNode):
     def __init__(self, shape):