rebase (undo apparently-accidental pin bump only in install_requirements), disable XNNPACK_ENABLE_AVX512VNNIGFNI

[ghstack-poisoned]

Author: swolchok

.ci/scripts/build_llama_android.sh

@@ -22,18 +22,12 @@ install_executorch_and_backend_lib() {
   ANDROID_NDK=/opt/ndk
   BUCK2=buck2
   ANDROID_ABI=arm64-v8a
-  cmake -DBUCK2="${BUCK2}" \
+  cmake --preset llm \
+    -DBUCK2="${BUCK2}" \
     -DCMAKE_TOOLCHAIN_FILE="${ANDROID_NDK}/build/cmake/android.toolchain.cmake" \
     -DANDROID_ABI="${ANDROID_ABI}" \
     -DCMAKE_INSTALL_PREFIX=cmake-android-out \
     -DCMAKE_BUILD_TYPE=Release \
-    -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
-    -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
-    -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
-    -DEXECUTORCH_BUILD_XNNPACK=ON \
-    -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \
     -DXNNPACK_ENABLE_ARM_BF16=OFF \
     -Bcmake-android-out .
 
@@ -51,11 +45,7 @@ build_llama_runner() {
     -DCMAKE_TOOLCHAIN_FILE="$ANDROID_NDK"/build/cmake/android.toolchain.cmake  \
     -DANDROID_ABI="${ANDROID_ABI}" \
     -DCMAKE_INSTALL_PREFIX=cmake-android-out \
-    -DCMAKE_BUILD_TYPE=Release -DPYTHON_EXECUTABLE=python \
-    -DEXECUTORCH_BUILD_XNNPACK=ON \
-    -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-    -DEXECUTORCH_BUILD_KERNELS_CUSTOM=ON \
+    -DCMAKE_BUILD_TYPE=Release \
     -Bcmake-android-out/examples/models/llama examples/models/llama
 
     cmake --build cmake-android-out/examples/models/llama -j4 --config Release

.ci/scripts/test_llama.sh

@@ -152,21 +152,11 @@ which "${PYTHON_EXECUTABLE}"
 cmake_install_executorch_libraries() {
     echo "Installing libexecutorch.a, libextension_module.so, libportable_ops_lib.a"
     rm -rf cmake-out
-    retry cmake \
+    retry cmake --preset llm \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \
-        -DEXECUTORCH_BUILD_EXTENSION_DATA_LOADER=ON \
-        -DEXECUTORCH_BUILD_EXTENSION_MODULE=ON \
-        -DEXECUTORCH_BUILD_EXTENSION_TENSOR=ON \
-        -DEXECUTORCH_BUILD_KERNELS_CUSTOM="$CUSTOM" \
-        -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-        -DEXECUTORCH_BUILD_KERNELS_QUANTIZED=ON \
-        -DEXECUTORCH_BUILD_XNNPACK="$XNNPACK" \
-        -DEXECUTORCH_BUILD_MPS="$MPS" \
-        -DEXECUTORCH_BUILD_COREML="$COREML" \
         -DEXECUTORCH_BUILD_QNN="$QNN" \
         -DQNN_SDK_ROOT="$QNN_SDK_ROOT" \
-        -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" \
         -Bcmake-out .
     cmake --build cmake-out -j9 --target install --config "$CMAKE_BUILD_TYPE"
 }
@@ -181,10 +171,6 @@ cmake_build_llama_runner() {
     retry cmake \
         -DCMAKE_INSTALL_PREFIX=cmake-out \
         -DCMAKE_BUILD_TYPE="$CMAKE_BUILD_TYPE" \
-        -DEXECUTORCH_BUILD_KERNELS_CUSTOM="$CUSTOM" \
-        -DEXECUTORCH_BUILD_KERNELS_OPTIMIZED=ON \
-        -DEXECUTORCH_BUILD_XNNPACK="$XNNPACK" \
-        -DPYTHON_EXECUTABLE="$PYTHON_EXECUTABLE" \
         -Bcmake-out/${dir} \
         ${dir}
     cmake --build cmake-out/${dir} -j9 --config "$CMAKE_BUILD_TYPE"

.github/workflows/build-presets.yml

@@ -20,7 +20,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        preset: [macos-arm64, pybind]
+        preset: [macos, ios, ios-simulator, pybind, llm]
     with:
       job-name: build
       ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
@@ -32,14 +32,14 @@ jobs:
         set -eux
         ${CONDA_RUN} ./install_requirements.sh > /dev/null
         ${CONDA_RUN} cmake --preset ${{ matrix.preset }}
-        ${CONDA_RUN} cmake --build cmake-out --parallel
+        ${CONDA_RUN} cmake --build cmake-out -j$(( $(sysctl -n hw.ncpu) - 1 ))
 
   linux:
     uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
     strategy:
       fail-fast: false
       matrix:
-        preset: [pybind]
+        preset: [linux, pybind, llm]
         runner: [linux.2xlarge, linux.arm64.2xlarge]
         docker-image: [executorch-ubuntu-22.04-clang12, executorch-ubuntu-22.04-gcc11-aarch64]
         # Excluding specific runner + docker image combinations that don't make sense:
@@ -65,4 +65,4 @@ jobs:
 
         ./install_requirements.sh > /dev/null
         cmake --preset ${{ matrix.preset }}
-        cmake --build cmake-out --parallel
+        cmake --build cmake-out -j$(( $(nproc) - 1 ))

.github/workflows/trunk.yml

@@ -692,3 +692,29 @@ jobs:
       build-mode: Release
       build-tool: cmake
       docker-image: executorch-ubuntu-22.04-clang12
+
+  unittest-nxp-neutron:
+    uses: pytorch/test-infra/.github/workflows/linux_job_v2.yml@main
+    permissions:
+      id-token: write
+      contents: read
+    with:
+      runner: linux.2xlarge
+      docker-image: executorch-ubuntu-22.04-clang12
+      submodules: 'true'
+      ref: ${{ github.event_name == 'pull_request' && github.event.pull_request.head.sha || github.sha }}
+      timeout: 90
+      script: |
+        set -eux
+        
+        # The generic Linux job chooses to use base env, not the one setup by the image
+        CONDA_ENV=$(conda env list --json | jq -r ".envs | .[-1]")
+        conda activate "${CONDA_ENV}"
+        
+        # Build and install Executorch
+        PYTHON_EXECUTABLE=python \
+        CMAKE_ARGS="-DEXECUTORCH_BUILD_NXP_NEUTRON=ON" \
+        .ci/scripts/setup-linux.sh --build-tool "cmake"
+        
+        # Run pytest
+        PYTHON_EXECUTABLE=python bash backends/nxp/run_unittests.sh

.gitignore

@@ -41,5 +41,18 @@ xcuserdata/
 *.xcworkspace/
 *.xcframework/
 
+# misc
+/.vscode/
+*.so
+*.dylib
+/cmake_wrapper.sh
+/data/
+/devtools/bundled_program/serialize/
+/exir/_serialize/program.fbs
+/exir/_serialize/scalar_type.fbs
+/include/
+/share/
+/version.py
+
 # Android
 *.aar

.lintrunner.toml

@@ -384,6 +384,22 @@ code = "TORCH_AO_IMPORT"
 include_patterns = ["**/*.py"]
 exclude_patterns = [
     "third-party/**",
+    # TODO: remove exceptions as we migrate
+    # backends
+    "backends/arm/quantizer/**",
+    "backends/arm/test/ops/**",
+    "backends/vulkan/quantizer/**",
+    "backends/vulkan/test/**",
+    "backends/qualcomm/quantizer/**",
+    "examples/qualcomm/**",
+    "backends/xnnpack/quantizer/**",
+    "backends/xnnpack/test/**",
+    "exir/tests/test_passes.py",
+    "extension/llm/export/builder.py",
+    "extension/llm/export/quantizer_lib.py",
+    "exir/tests/test_memory_planning.py",
+    "backends/transforms/duplicate_dynamic_quant_chain.py",
+    "exir/backend/test/demos/test_xnnpack_qnnpack.py",
 ]
 
 command = [
@@ -392,7 +408,7 @@ command = [
   "lintrunner_adapters",
   "run",
   "grep_linter",
-  "--pattern=\\bfrom torch\\.ao\\.quantization\\.(?:quantize_pt2e)(?:\\.[A-Za-z0-9_]+)*\\b",
+  "--pattern=\\bfrom torch\\.ao\\.quantization\\.(?:quantizer|observer|quantize_pt2e|pt2e)(?:\\.[A-Za-z0-9_]+)*\\b",
   "--linter-name=TorchAOImport",
   "--error-name=Prohibited torch.ao.quantization import",
   """--error-description=\

CMakeLists.txt

@@ -362,6 +362,9 @@ add_library(executorch_core ${_executorch_core__srcs})
 add_library(executorch_no_prim_ops ALIAS executorch_core)
 
 target_link_libraries(executorch_core PRIVATE program_schema)
+if(ANDROID)
+  target_link_libraries(executorch_core PUBLIC log)
+endif()
 if(EXECUTORCH_USE_DL)
   # Check if dl exists for this toolchain and only then link it.
   find_library(DL_LIBRARY_EXISTS NAMES dl)
@@ -653,10 +656,6 @@ if(EXECUTORCH_BUILD_PYBIND)
   install(TARGETS portable_lib
           LIBRARY DESTINATION executorch/extension/pybindings
   )
-
-  if(EXECUTORCH_BUILD_EXTENSION_TRAINING)
-
-  endif()
 endif()
 
 if(EXECUTORCH_BUILD_EXTENSION_TRAINING)

CMakePresets.json

@@ -7,13 +7,13 @@
       "binaryDir": "${sourceDir}/cmake-out"
     },
     {
-      "name": "macos-arm64",
-      "displayName": "Build everything buildable on macOS arm64",
+      "name": "macos",
+      "displayName": "Build everything buildable on macOS",
       "inherits": ["common"],
       "generator": "Xcode",
       "cacheVariables": {
         "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/third-party/ios-cmake/ios.toolchain.cmake",
-        "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/macos-arm64.cmake",
+        "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/macos.cmake",
         "PLATFORM": "MAC_ARM64",
         "DEPLOYMENT_TARGET": "10.15"
       },
@@ -23,6 +23,54 @@
         "rhs": "Darwin"
       }
     },
+    {
+      "name": "ios",
+      "displayName": "Build everything buildable on iOS",
+      "inherits": ["common"],
+      "generator": "Xcode",
+      "cacheVariables": {
+        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/third-party/ios-cmake/ios.toolchain.cmake",
+        "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/ios.cmake",
+        "PLATFORM": "OS64",
+        "DEPLOYMENT_TARGET": "17.0"
+      },
+      "condition": {
+        "lhs": "${hostSystemName}",
+        "type": "equals",
+        "rhs": "Darwin"
+      }
+    },
+    {
+      "name": "ios-simulator",
+      "displayName": "Build everything buildable on iOS simulator",
+      "inherits": ["common"],
+      "generator": "Xcode",
+      "cacheVariables": {
+        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/third-party/ios-cmake/ios.toolchain.cmake",
+        "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/ios.cmake",
+        "PLATFORM": "SIMULATORARM64",
+        "DEPLOYMENT_TARGET": "17.0"
+      },
+      "condition": {
+        "lhs": "${hostSystemName}",
+        "type": "equals",
+        "rhs": "Darwin"
+      }
+    },
+    {
+      "name": "linux",
+      "displayName": "Build everything buildable on Linux",
+      "inherits": ["common"],
+      "cacheVariables": {
+        "CMAKE_SYSTEM_NAME": "Linux",
+        "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/linux.cmake"
+      },
+      "condition": {
+        "lhs": "${hostSystemName}",
+        "type": "equals",
+        "rhs": "Linux"
+      }
+    },
     {
       "name": "pybind",
       "displayName": "Build pybindings exported in the wheel",
@@ -36,6 +84,26 @@
         "string": "${hostSystemName}",
         "list": ["Darwin", "Linux", "Windows"]
       }
+    },
+    {
+        "name": "llm",
+        "displayName": "Build LLM libraries",
+        "inherits": [
+            "common"
+        ],
+        "cacheVariables": {
+            "EXECUTORCH_BUILD_PRESET_FILE": "${sourceDir}/tools/cmake/preset/llm.cmake",
+            "CMAKE_OSX_DEPLOYMENT_TARGET": "10.15"
+        },
+        "condition": {
+            "type": "inList",
+            "string": "${hostSystemName}",
+            "list": [
+                "Darwin",
+                "Linux",
+                "Windows"
+            ]
+        }
     }
   ]
 }

backends/apple/coreml/CMakeLists.txt

@@ -25,8 +25,6 @@ endif()
 
 option(COREML_BUILD_EXECUTOR_RUNNER "Build CoreML executor runner." OFF)
 
-set(CMAKE_OSX_DEPLOYMENT_TARGET 10.15)
-
 # inmemoryfs sources
 set(INMEMORYFS_SOURCES
     runtime/inmemoryfs/inmemory_filesystem.cpp
@@ -240,7 +238,6 @@ if(EXECUTORCH_BUILD_COREML AND EXECUTORCH_BUILD_PYBIND)
 
   pybind11_add_module(executorchcoreml SHARED runtime/inmemoryfs/inmemory_filesystem_py.cpp)
 
-  target_compile_options(executorchcoreml PRIVATE -mmacosx-version-min=${CMAKE_OSX_DEPLOYMENT_TARGET})
   if(CMAKE_BUILD_TYPE STREQUAL "Debug" OR CMAKE_BUILD_TYPE STREQUAL "RelWithDebInfo")
     target_compile_options(executorchcoreml PRIVATE -g)
   endif()