Build optimized library with CMake

Author: GregoryComer

.gitmodules

@@ -59,6 +59,3 @@
 	path = third-party/lm-evaluation-harness
 	url = https://github.com/EleutherAI/lm-evaluation-harness
 	branch = v0.4.1
-[submodule "third-party/sleef"]
-	path = third-party/sleef
-	url = [email protected]:shibatch/sleef.git

kernels/optimized/CMakeLists.txt

@@ -39,10 +39,10 @@ include(${EXECUTORCH_ROOT}/build/Codegen.cmake)
 
 # Generate C++ bindings to register kernels into both PyTorch (for AOT) and
 # Executorch (for runtime). Here select all ops in optimized.yaml
-set(_yaml "${CMAKE_CURRENT_LIST_DIR}/optimized.yaml")
+set(_yaml "${CMAKE_CURRENT_LIST_DIR}/optimized-oss.yaml")
 gen_selected_ops("${_yaml}" "" "")
 
-generate_bindings_for_kernels(${CMAKE_CURRENT_SOURCE_DIR}/optimized.yaml "")
+generate_bindings_for_kernels(${CMAKE_CURRENT_SOURCE_DIR}/optimized-oss.yaml "")
 message("Generated files ${gen_command_sources}")
 
 list(TRANSFORM _optimized_kernels__srcs PREPEND "${EXECUTORCH_ROOT}/")

kernels/optimized/optimized-oss.yaml

@@ -0,0 +1,73 @@
+# Copyright (c) Meta Platforms, Inc. and affiliates.
+#
+# This yaml file contains operators that have optimized kernels available.
+
+- op: add.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_add_out
+
+- op: add.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_add_scalar_out
+
+- op: bmm.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_bmm_out
+
+- op: div.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_div_out
+
+- op: div.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_div_scalar_out
+
+- op: exp.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_exp_out
+
+- op: le.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_le_scalar_out
+
+- op: le.Tensor_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_le_tensor_out
+
+- op: mul.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_mul_out
+
+- op: mul.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_mul_scalar_out
+
+- op: native_layer_norm.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_native_layer_norm_out
+
+- op: neg.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_neg_out
+
+- op: sub.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_sub_out
+
+- op: sub.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::opt_sub_scalar_out