Setup dirsync for torchao experimental

Summary:
In order to build and test kernels internally for mac, android and ios we need
to add build support in xplat. Which means we need to setup mirroring between
fbcode and xplat. For now setting this up only for experimental folder, pending
larger decision on whether to sync the two folder entirely.

Reviewed By: mzlee

Differential Revision: D61480451

Author: kimishpatel

torchao/experimental/kernels/cpu/aarch64/benchmarks/build_and_run_benchmarks.sh

@@ -1,4 +1,6 @@
 #!/bin/bash
+# (c) Meta Platforms, Inc. and affiliates. Confidential and proprietary.
+
 # Call script with sh build_and_run_benchmarks.sh {BENCHAMRK}
 
 SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &> /dev/null && pwd)

torchao/experimental/kernels/cpu/aarch64/bitpacking/macro.h

@@ -1,3 +1,5 @@
+// (c) Meta Platforms, Inc. and affiliates. Confidential and proprietary.
+
 #pragma once
 
 #define TORCHAO_ALWAYS_INLINE __attribute__((always_inline))

torchao/experimental/kernels/cpu/aarch64/tests/build_and_run_tests.sh

@@ -1,4 +1,6 @@
 #!/bin/bash
+# (c) Meta Platforms, Inc. and affiliates. Confidential and proprietary.
+
 SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &> /dev/null && pwd)
 export TORCHAO_LIBRARIES=${SCRIPT_DIR}/../../../../../..
 export CMAKE_OUT=/tmp/cmake-out/torch_ao/tests

torchao/experimental/kernels/cpu/aarch64/tests/test_bitpacking.cpp

@@ -185,8 +185,7 @@ template <int nbit>
 void test_bitpacking_32_lowbit_values() {
   int unpacked_bytes = 32;
   int packed_bytes = unpacked_bytes * nbit / 8;
-  auto input_shifted =
-      torchao::get_random_lowbit_vector(unpacked_bytes, nbit);
+  auto input_shifted = torchao::get_random_lowbit_vector(unpacked_bytes, nbit);
   std::vector<int8_t> input(unpacked_bytes, 0);
   int8_t low = -(1 << (nbit - 1));
   int8_t high = (1 << (nbit - 1));
@@ -218,8 +217,7 @@ template <int nbit>
 void test_bitpacking_64_lowbit_values() {
   int unpacked_bytes = 64;
   int packed_bytes = unpacked_bytes * nbit / 8;
-  auto input_shifted =
-      torchao::get_random_lowbit_vector(unpacked_bytes, nbit);
+  auto input_shifted = torchao::get_random_lowbit_vector(unpacked_bytes, nbit);
   std::vector<int8_t> input(unpacked_bytes, 0);
   int8_t low = -(1 << (nbit - 1));
   int8_t high = (1 << (nbit - 1));
@@ -259,8 +257,7 @@ template <int nbit>
 void test_bitpacking_128_lowbit_values() {
   int unpacked_bytes = 128;
   int packed_bytes = unpacked_bytes * nbit / 8;
-  auto input_shifted =
-      torchao::get_random_lowbit_vector(unpacked_bytes, nbit);
+  auto input_shifted = torchao::get_random_lowbit_vector(unpacked_bytes, nbit);
   std::vector<int8_t> input(unpacked_bytes, 0);
   int8_t low = -(1 << (nbit - 1));
   int8_t high = (1 << (nbit - 1));

torchao/experimental/kernels/cpu/build_and_run_benchmarks.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+# (c) Meta Platforms, Inc. and affiliates. Confidential and proprietary.
+
+# Call script with sh build_and_run_benchmarks.sh {BENCHAMRK}
+
+SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &> /dev/null && pwd)
+export TORCHAO_LIBRARIES=${SCRIPT_DIR}/../../../..
+export CMAKE_OUT=/tmp/cmake-out/torch_ao/benchmarks
+cmake -DTORCHAO_LIBRARIES=${TORCHAO_LIBRARIES} \
+    -S ${TORCHAO_LIBRARIES}/torchao/experimental/kernels/cpu/aarch64/benchmarks \
+    -B ${CMAKE_OUT}
+
+cmake --build ${CMAKE_OUT}
+
+# Run
+case "$1" in
+    quantization) ${CMAKE_OUT}/benchmark_quantization; ;;
+    bitpacking) ${CMAKE_OUT}/benchmark_bitpacking; ;;
+    linear) ${CMAKE_OUT}/benchmark_linear; ;;
+    *) echo "Unknown benchmark: $1. Please specify quantization, bitpacking, or linear."; exit 1; ;;
+esac

torchao/experimental/kernels/cpu/build_and_run_tests.sh

@@ -0,0 +1,15 @@
+#!/bin/bash
+# (c) Meta Platforms, Inc. and affiliates. Confidential and proprietary.
+
+SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" &> /dev/null && pwd)
+export TORCHAO_LIBRARIES=${SCRIPT_DIR}/../../../..
+export CMAKE_OUT=/tmp/cmake-out/torch_ao/tests
+cmake -DTORCHAO_LIBRARIES=${TORCHAO_LIBRARIES} -S ${TORCHAO_LIBRARIES}/torchao/experimental/kernels/cpu/aarch64/tests -B ${CMAKE_OUT}
+
+cmake --build  ${CMAKE_OUT}
+
+# Run
+ ${CMAKE_OUT}/test_quantization
+ ${CMAKE_OUT}/test_bitpacking
+ ${CMAKE_OUT}/test_linear
+ ${CMAKE_OUT}/test_valpacking