[executorch] Populate cadence cpu ops

.lintrunner.toml

@@ -74,6 +74,7 @@ exclude_patterns = [
     # NB: Objective-C is not supported
     'examples/apple/**',
     'examples/demo-apps/apple_ios/**',
+    'examples/demo-apps/react-native/rnllama/ios/**',
     # File contains @generated
     'extension/llm/custom_ops/spinquant/fast_hadamard_transform_special.h',
     'extension/llm/custom_ops/spinquant/test/fast_hadamard_transform_special_unstrided_cpu.h',

backends/cadence/CMakeLists.txt

@@ -23,7 +23,6 @@ include(${EXECUTORCH_ROOT}/build/Utils.cmake)
 
 # Let files say "include <executorch/path/to/header.h>".
 set(_common_include_directories ${EXECUTORCH_ROOT}/..)
-set(TARGET_DIR reference)
 
 if(EXECUTORCH_CADENCE_CPU_RUNNER)
   include(${EXECUTORCH_ROOT}/build/Codegen.cmake)
@@ -61,6 +60,9 @@ if(EXECUTORCH_CADENCE_CPU_RUNNER)
                                       ${_common_include_directories}
   )
 
+  set(TARGET_DIR reference)
+  add_subdirectory(${CMAKE_CURRENT_SOURCE_DIR}/${TARGET_DIR}/kernels)
+
   target_link_libraries(
   cadence_runner
   executorch

backends/cadence/aot/functions.yaml

@@ -142,6 +142,41 @@
     - arg_meta: null
       kernel_name: torch::executor::where_out
 
+- op: transpose_copy.int_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::transpose_copy_int_out
+
+- op: eq.Scalar_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::eq_scalar_out
+
+- op: logical_not.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::logical_not_out
+
+- op: any.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::any_out
+
+- op: native_group_norm.out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::native_group_norm_out
+
+- op: sum.IntList_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::sum_dim_out
+
+- op: select_copy.int_out
+  kernels:
+    - arg_meta: null
+      kernel_name: torch::executor::select_copy_int_out
+
 # custom ops
 - func: cadence::quantize_per_tensor.out(Tensor input, float scale, int zero_point, int quant_min, int quant_max, ScalarType dtype, *, Tensor(a!) out) -> Tensor(a!)
   variants: function
@@ -183,3 +218,18 @@
   kernels:
     - arg_meta: null
       kernel_name: impl::reference::quantized_matmul_out
+
+- func: cadence::quantized_linear.per_tensor_out(Tensor src, Tensor weight, Tensor bias, SymInt src_zero_point, SymInt weight_zero_point, SymInt out_multiplier, SymInt out_shift, SymInt out_zero_point, Tensor? offset, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::reference::quantized_linear_per_tensor_out
+
+- func: cadence::im2row.out(Tensor input, int[2] kernel_size, int[2] dilation, int[2] padding, int[2] stride, Tensor in_zero_point, bool channel_last=False, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::reference::im2row_out
+
+- func: cadence::quantized_conv.per_tensor_out(Tensor input, Tensor weight, Tensor bias, int[] stride, SymInt[] padding, int[] dilation, int groups, int input_zero_point, int weight_zero_point, float bias_scale, float out_scale, int out_zero_point, int out_multiplier, int out_shift, bool channel_last=False, *, Tensor(a!) out) -> Tensor(a!)
+  kernels:
+    - arg_meta: null
+      kernel_name: impl::reference::quantized_conv_per_tensor_out

backends/cadence/reference/operators/CMakeLists.txt

@@ -55,6 +55,16 @@ set(_aten_ops__srcs
     "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_expand_copy.cpp"
     "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_gelu.cpp"
     "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_empty.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_transpose_copy.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_eq.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_logical_not.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_any.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_native_group_norm.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_sum.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/op_select_copy.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/util/dtype_util.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/util/normalization_ops_util.cpp"
+    "${EXECUTORCH_ROOT}/kernels/portable/cpu/util/select_copy_util.cpp"
 )
 add_library(aten_ops_cadence ${_aten_ops__srcs})
 target_link_libraries(aten_ops_cadence PUBLIC executorch)
@@ -72,12 +82,13 @@ target_include_directories(
 add_library(
   custom_ops
   "quantized_linear_out.cpp"
-  "quantized_conv_out.cpp"
   "quantized_relu_out.cpp"
   "quantized_layer_norm.cpp"
   "quantize_per_tensor.cpp"
   "dequantize_per_tensor.cpp"
   "quantized_matmul_out.cpp"
+  "im2row_out.cpp"
+  "convolution_out.cpp"
 )
 target_include_directories(
   custom_ops PUBLIC ${ROOT_DIR}/.. ${CMAKE_BINARY_DIR}