From 2d841bcb9f567f7acb9e400fd1d7d9712df343e6 Mon Sep 17 00:00:00 2001
From: Michael Lazos <mlazos@fb.com>
Date: Tue, 15 Aug 2023 05:32:39 +0000
Subject: [PATCH] Remove type promotion workaround (#107202)

Removes old type promotion workaround

Pull Request resolved: https://github.com/pytorch/pytorch/pull/107202
Approved by: https://github.com/xuzhao9, https://github.com/eellison
---
 test/inductor/test_foreach.py |  2 +-
 torch/_inductor/lowering.py   | 18 +-----------------
 2 files changed, 2 insertions(+), 18 deletions(-)

diff --git a/test/inductor/test_foreach.py b/test/inductor/test_foreach.py
index 8c8bdc40c23e53..17301989bc0387 100644
--- a/test/inductor/test_foreach.py
+++ b/test/inductor/test_foreach.py
@@ -223,7 +223,7 @@ def fn(a0, a1, b0, b1):
         actual = fn_opt(*inputs)
         expected = fn(*inputs)
         self.assertEqual(actual, expected)
-        self.assertEqual(torch._inductor.metrics.generated_kernel_count, 2)
+        self.assertEqual(torch._inductor.metrics.generated_kernel_count, 1)
 
     @requires_cuda()
     @bin_ops
diff --git a/torch/_inductor/lowering.py b/torch/_inductor/lowering.py
index 9218f5280e2dc0..932b3576dc3c3e 100644
--- a/torch/_inductor/lowering.py
+++ b/torch/_inductor/lowering.py
@@ -423,28 +423,12 @@ def is_dynamic(*args):
                 for t in args
             )
 
-        def has_type_promotion(*args):
-            if len(args) < 2:
-                return False
-            else:
-                dtype = None
-                for t in args:
-                    if isinstance(t, TensorBox):
-                        if dtype is None:
-                            dtype = t.data.get_dtype()  # type: ignore[attr-defined]
-                        elif dtype != t.data.get_dtype():
-                            return True
-                return False
-
         # group by device, whether any of the inputs are dynamic, and whether their types match
         # (proxy for type promotion)
-        # Note: we'll fallback on type promotion until
-        # https://github.com/openai/triton/commit/9820899b3845e461d9031dba66062efade65d420
-        # is in the pytorch triton version
         def group_args(arg_pairs):
             out = defaultdict(list)
             for i, args in enumerate(arg_pairs):
-                use_foreach = not (is_dynamic(*args) or has_type_promotion(*args))
+                use_foreach = not is_dynamic(*args)
                 device = None
                 for t in args:
                     if isinstance(t, TensorBox):