Arm64 Support (#1108)

* Adds ARM64 toolchain file and logic to determine target architecture * Enables non-sse version of rgbz transformation * Adds arm64 cross compile support to Dockerfile * Add ARM64 builds to pipeline * Enabling Functional Testing on ARM64 * Updating Display for ARM64 and glxinfo * fixing up build artifact names * Switch to publishing GNU binaries * Moving DepthEngine properly * Moving DepthEngine properly * fixing case * Don't run multi device tests on ARM64 Co-authored-by: Matt Schulte <schultetwin1@gmail.com> Co-authored-by: Matt Schulte <maschu@microsoft.com>
2020-03-06 09:06:21 -08:00 · 2020-03-06 09:06:21 -08:00 · 6dea37e560
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@ -135,7 +135,8 @@ if ("${CMAKE_SYSTEM_NAME}" STREQUAL "Linux")
    set(CMAKE_BUILD_RPATH "\$ORIGIN")
 endif()

-set(TARGET_ARCH ${CMAKE_SYSTEM_PROCESSOR})
+include(DetermineTargetArch)
+determine_target_arch(TARGET_ARCH)

 # CMake doesn't set the target processor correctly for MSVC
 if ("${CMAKE_C_COMPILER_ID}" STREQUAL "MSVC")
@ -149,16 +150,14 @@ if ("${CMAKE_C_COMPILER_ID}" STREQUAL "MSVC")

    # Check what architecture we are building for. This assumes all 64-bit architectures are amd64, which will break
    # if we decide to support arm.
-    if ("${CMAKE_SIZEOF_VOID_P}" STREQUAL "8")
-        set(TARGET_ARCH "amd64")
+    if ("${TARGET_ARCH}" STREQUAL "x86_64")
        configure_file(k4a.props.in ${CMAKE_CURRENT_SOURCE_DIR}/src/csharp/k4a.x64.props)
        configure_file(StubGenerator.xml.in ${CMAKE_CURRENT_SOURCE_DIR}/src/csharp/StubGenerator.x64.xml)
-    elseif("${CMAKE_SIZEOF_VOID_P}" STREQUAL "4")
-        set(TARGET_ARCH "x86")
+    elseif("${TARGET_ARCH}" STREQUAL "i686")
        configure_file(k4a.props.in ${CMAKE_CURRENT_SOURCE_DIR}/src/csharp/k4a.x86.props)
        configure_file(StubGenerator.xml.in ${CMAKE_CURRENT_SOURCE_DIR}/src/csharp/StubGenerator.x86.xml)
    else()
-        message(FATAL_ERROR "Unknown architecture with size of void* = ${CMAKE_SIZEOF_VOID_P}")
+        message(FATAL_ERROR "Unknown architecture for MSVC: ${TARGET_ARCH}")
    endif()
 endif()

--- a/azure-pipelines.yml
+++ b/azure-pipelines.yml
@ -369,9 +369,6 @@ jobs:

 - job: LinuxK4ABuildTest
  displayName: Linux
-  pool:
-    vmImage: 'ubuntu-16.04'
-  container: mcr.microsoft.com/akbuilder-linux:v3
  strategy:
    maxParallel: 100
    matrix:
@ -379,18 +376,40 @@ jobs:
        CMakeLinuxTargetTriple: 'x86_64-linux-clang'
        CMakeConfiguration: 'debug'
        UsesOpenCV: 'TRUE'
+        container: 'mcr.microsoft.com/akbuilder-linux:v5-amd64'
      x64-gnu_debug_ninja:
        CMakeLinuxTargetTriple: 'x86_64-linux-gnu'
        CMakeConfiguration: 'debug'
        UsesOpenCV: 'TRUE'
+        container: 'mcr.microsoft.com/akbuilder-linux:v5-amd64'
      x64-clang_rel_ninja:
        CMakeLinuxTargetTriple: 'x86_64-linux-clang'
        CMakeConfiguration: 'relwithdebinfo'
        UsesOpenCV: 'TRUE'
+        container: 'mcr.microsoft.com/akbuilder-linux:v5-amd64'
      x64-gnu_rel_ninja:
        CMakeLinuxTargetTriple: 'x86_64-linux-gnu'
        CMakeConfiguration: 'relwithdebinfo'
        UsesOpenCV: 'TRUE'
+        container: 'mcr.microsoft.com/akbuilder-linux:v5-amd64'
+      # arm64-clang_debug_ninja:
+      #   CMakeLinuxTargetTriple: 'arm64-linux-clang'
+      #   CMakeConfiguration: 'debug'
+      #   UsesOpenCV: 'FALSE'
+      arm64-gnu_debug_ninja:
+        CMakeLinuxTargetTriple: 'arm64-linux-gnu'
+        CMakeConfiguration: 'debug'
+        UsesOpenCV: 'FALSE'
+        container: 'mcr.microsoft.com/akbuilder-linux:v5-arm64'
+      # arm64-clang_rel_ninja:
+      #   CMakeLinuxTargetTriple: 'arm64-linux-clang'
+      #   CMakeConfiguration: 'relwithdebinfo'
+      #   UsesOpenCV: 'FALSE'
+      arm64-gnu_rel_ninja:
+        CMakeLinuxTargetTriple: 'arm64-linux-gnu'
+        CMakeConfiguration: 'relwithdebinfo'
+        UsesOpenCV: 'FALSE'
+        container: 'mcr.microsoft.com/akbuilder-linux:v5-arm64'

      # 32-bit builds are currently broken
      # i386-unknown-linux-clang_debug_ninja:
@ -405,6 +424,9 @@ jobs:
      # i386-unknown-linux-gnu_relwithdebinfo_ninja:
      #   CMakeLinuxTargetTriple: 'i386-linux-gnu'
      #   CMakeConfiguration: 'relwithdebinfo'
+  pool:
+    vmImage: 'ubuntu-18.04'
+  container: $[ variables['container'] ]

  steps:
  - checkout: self
@ -470,7 +492,7 @@ jobs:
      ArtifactName: '$(CMakeLinuxTargetTriple)-$(CMakeConfiguration)'
      parallel: true
      parallelCount: 8
-    condition: and(succeeded(), contains(variables['CMakeLinuxTargetTriple'], 'clang'))
+    condition: and(succeeded(), contains(variables['CMakeLinuxTargetTriple'], 'gnu'))

 - job: DocumentationBuild
  displayName: Documentation Build
@ -657,7 +679,7 @@ jobs:
      inputs:
        sourceFolder: "$(System.ArtifactsDirectory)/depthengineplugin/windows/amd64/release/"
        contents: "depthengine*.dll"
-        targetFolder: "$(System.ArtifactsDirectory)/amd64-windows-msvc-RelWithDebInfo/bin"
+        targetFolder: "$(System.ArtifactsDirectory)/amd64-windows-msvc-relwithdebinfo/bin"
        flattenFolders: true

    - task: CopyFiles@2
@ -665,33 +687,33 @@ jobs:
      inputs:
        sourceFolder: "$(System.ArtifactsDirectory)/depthengineplugin/windows/amd64/release/"
        contents: "depthengine*.dll"
-        targetFolder: "$(System.ArtifactsDirectory)/amd64-windows-msvc-RelWithDebInfo/bin/Release/Microsoft.AzureKinect.FunctionalTests/netcoreapp2.1/"
+        targetFolder: "$(System.ArtifactsDirectory)/amd64-windows-msvc-relwithdebinfo/bin/Release/Microsoft.AzureKinect.FunctionalTests/netcoreapp2.1/"
        flattenFolders: true

-    - script: '.\amd64-windows-msvc-RelWithDebInfo\bin\AzureKinectFirmwareTool.exe -r'
+    - script: '.\amd64-windows-msvc-relwithdebinfo\bin\AzureKinectFirmwareTool.exe -r'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Reset K4A Device'

-    - script: '.\amd64-windows-msvc-RelWithDebInfo\bin\AzureKinectFirmwareTool.exe -u firmware/AzureKinectDK_Fw_$(firmware_version).bin'
+    - script: '.\amd64-windows-msvc-relwithdebinfo\bin\AzureKinectFirmwareTool.exe -u firmware/AzureKinectDK_Fw_$(firmware_version).bin'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Update Device'

-    - script: '.\amd64-windows-msvc-RelWithDebInfo\bin\enumerate_devices.exe'
+    - script: '.\amd64-windows-msvc-relwithdebinfo\bin\enumerate_devices.exe'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Check Device Health'

    - script: 'python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_test_list.txt --bin bin/ --output=xml --gtest_filter=-*ONBOARDING*'
-      workingDirectory: '$(System.ArtifactsDirectory)/amd64-windows-msvc-RelWithDebInfo'
+      workingDirectory: '$(System.ArtifactsDirectory)/amd64-windows-msvc-relwithdebinfo'
      displayName: 'Run Functional Tests'
      timeoutInMinutes: 15

    - script: 'python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_custom_test_list.txt --bin bin/ --output=xml --gtest_filter=-*ONBOARDING*'
-      workingDirectory: '$(System.ArtifactsDirectory)/amd64-windows-msvc-RelWithDebInfo'
+      workingDirectory: '$(System.ArtifactsDirectory)/amd64-windows-msvc-relwithdebinfo'
      displayName: 'Run Custom Functional Tests'
      timeoutInMinutes: 15

    - script: 'python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_test_list.txt --bin bin/ --output=xml --gtest_filter=*ONBOARDING*'
-      workingDirectory: '$(System.ArtifactsDirectory)/amd64-windows-msvc-RelWithDebInfo'
+      workingDirectory: '$(System.ArtifactsDirectory)/amd64-windows-msvc-relwithdebinfo'
      displayName: 'Run Functional Tests - Onboarding'
      timeoutInMinutes: 15
      continueOnError: true
@ -714,10 +736,19 @@ jobs:
    dependsOn: LinuxK4ABuildTest
    variables:
      firmware_version: "$(linux_firmware_version)"
+    strategy:
+      maxParallel: 100
+      matrix:
+        x86_64:
+          CMakeLinuxTargetTriple: 'x86_64-linux-gnu'
+          EdenArch: 'x64'
+        arm64:
+          CMakeLinuxTargetTriple: 'arm64-linux-gnu'
+          EdenArch: 'arm64'
    pool:
      name: Analog-FwConnected
      demands:
-      - Eden
+      - EdenArch -equals $(EdenArch)
      - Agent.OS -equals Linux

    steps:
@ -730,7 +761,7 @@ jobs:
    - task: DownloadBuildArtifacts@0
      displayName: 'Download Build Artifacts'
      inputs:
-        artifactName: 'x86_64-linux-clang-RelWithDebInfo'
+        artifactName: '$(CMakeLinuxTargetTriple)-relwithdebinfo'
        parallelizationLimit: 8
        
    - task: NuGetToolInstaller@0
@ -755,19 +786,19 @@ jobs:
      displayName: "Temp List files"

    - task: CopyFiles@2
-      displayName: "Copy DepthEngine into amd64/debug artifacts folder"
+      displayName: "Copy DepthEngine into $(EdenArch)/debug artifacts folder"
      inputs:
-        sourceFolder: "$(System.ArtifactsDirectory)/NugetOutputDir/Microsoft.Azure.Kinect.Sensor.$(NuGetPackageVersion)/linux/lib/native/x64/release/"
+        sourceFolder: "$(System.ArtifactsDirectory)/NugetOutputDir/Microsoft.Azure.Kinect.Sensor.$(NuGetPackageVersion)/linux/lib/native/$(EdenArch)/release/"
        contents: "libdepthengine*"
-        targetFolder: "$(System.ArtifactsDirectory)/depthengineplugin/linux/x86_64/debug"
+        targetFolder: "$(System.ArtifactsDirectory)/depthengineplugin/linux/$(EdenArch)/debug"
        flattenFolders: true

    - task: CopyFiles@2
-      displayName: "Copy DepthEngine into amd64/release artifacts folder"
+      displayName: "Copy DepthEngine into $(EdenArch)/release artifacts folder"
      inputs:
-        sourceFolder: "$(System.ArtifactsDirectory)/NugetOutputDir/Microsoft.Azure.Kinect.Sensor.$(NuGetPackageVersion)/linux/lib/native/x64/release/"
+        sourceFolder: "$(System.ArtifactsDirectory)/NugetOutputDir/Microsoft.Azure.Kinect.Sensor.$(NuGetPackageVersion)/linux/lib/native/$(EdenArch)/release/"
        contents: "libdepthengine*"
-        targetFolder: "$(System.ArtifactsDirectory)/depthengineplugin/linux/x86_64/release"
+        targetFolder: "$(System.ArtifactsDirectory)/depthengineplugin/linux/$(EdenArch)/release"
        flattenFolders: true

    - task: CopyFiles@2
@ -788,17 +819,22 @@ jobs:
    - task: CopyFiles@2
      displayName: "Copy DepthEnginePlugin into Build Artifacts"
      inputs:
-        sourceFolder: "$(System.ArtifactsDirectory)/depthengineplugin/linux/x86_64/release/"
+        sourceFolder: "$(System.ArtifactsDirectory)/depthengineplugin/linux/$(EdenArch)/release/"
        contents: "libdepthengine.so*"
-        targetFolder: "$(System.ArtifactsDirectory)/x86_64-linux-clang-relwithdebinfo/bin/"
+        targetFolder: "$(System.ArtifactsDirectory)/$(CMakeLinuxTargetTriple)-relwithdebinfo/bin/"
        flattenFolders: true

-    - script: 'chmod +x ./x86_64-linux-clang-relwithdebinfo/bin/*'
+    - script: 'chmod +x ./$(CMakeLinuxTargetTriple)-relwithdebinfo/bin/*'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Add execution property to binary files'

      # Set the DISPLAY variable since DepthEngine needs to open a display window (even there is no visual display).
-    - script: 'echo "##vso[task.setvariable variable=DISPLAY]:0"'
+    - script: |
+        if [ $(EdenArch) == "arm64" ]; then
+          echo "##vso[task.setvariable variable=DISPLAY]:1"
+        else
+          echo "##vso[task.setvariable variable=DISPLAY]:0"
+        fi
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'set DISPLAY variable'

@ -807,34 +843,38 @@ jobs:
      displayName: 'Check openGL version'
      timeoutInMinutes: 5

-    - script: './x86_64-linux-clang-relwithdebinfo/bin/AzureKinectFirmwareTool -r'
+    - script: './$(CMakeLinuxTargetTriple)-relwithdebinfo/bin/AzureKinectFirmwareTool -r'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Reset K4A Device'
      env:
        K4A_LOG_LEVEL: 'I'

-    - script: './x86_64-linux-clang-relwithdebinfo/bin/AzureKinectFirmwareTool -u firmware/AzureKinectDK_Fw_$(firmware_version).bin'
+    - script: './$(CMakeLinuxTargetTriple)-relwithdebinfo/bin/AzureKinectFirmwareTool -u firmware/AzureKinectDK_Fw_$(firmware_version).bin'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Update Device'
      env:
        K4A_LOG_LEVEL: 'I'

-    - script: './x86_64-linux-clang-relwithdebinfo/bin/enumerate_devices'
+    - script: './$(CMakeLinuxTargetTriple)-relwithdebinfo/bin/enumerate_devices'
      workingDirectory: '$(System.ArtifactsDirectory)'
      displayName: 'Check Device Health'

    - script: 'python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_test_list.txt --bin bin/ --output=xml --gtest_filter=-*ONBOARDING*'
-      workingDirectory: '$(System.ArtifactsDirectory)/x86_64-linux-clang-relwithdebinfo'
+      workingDirectory: '$(System.ArtifactsDirectory)/$(CMakeLinuxTargetTriple)-relwithdebinfo'
      displayName: 'Run Functional Tests'
      timeoutInMinutes: 15

-    - script: 'python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_custom_test_list.txt --bin bin/ --output=xml --gtest_filter=-*ONBOARDING*'
-      workingDirectory: '$(System.ArtifactsDirectory)/x86_64-linux-clang-relwithdebinfo'
+    - script: |
+        # skip testing on ARM64 Jetson Nano doesn't like 2 USB devices running at once.
+        if [ $(EdenArch) == "x64" ]; then
+          python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_custom_test_list.txt --bin bin/ --output=xml --gtest_filter=-*ONBOARDING*
+        fi
+      workingDirectory: '$(System.ArtifactsDirectory)/$(CMakeLinuxTargetTriple)-relwithdebinfo'
      displayName: 'Run Custom Functional Tests'
      timeoutInMinutes: 15

    - script: 'python $(Build.SourcesDirectory)/scripts/RunTestList.py --list bin/functional_test_list.txt --bin bin/ --output=xml --gtest_filter=*ONBOARDING*'
-      workingDirectory: '$(System.ArtifactsDirectory)/x86_64-linux-clang-relwithdebinfo'
+      workingDirectory: '$(System.ArtifactsDirectory)/$(CMakeLinuxTargetTriple)-relwithdebinfo'
      displayName: 'Run Functional Tests - Onboarding'
      timeoutInMinutes: 15
      continueOnError: true
--- a/cmake/DetermineTargetArch.cmake
+++ b/cmake/DetermineTargetArch.cmake
@ -0,0 +1,53 @@
+function(determine_target_arch OUTPUT_VARIABLE)
+    if("${CMAKE_C_COMPILER_ID}" STREQUAL "MSVC")
+        if("${MSVC_C_ARCHITECTURE_ID}" STREQUAL "X86")
+            set(ARCH "i686")
+        elseif("${MSVC_C_ARCHITECTURE_ID}" STREQUAL "x64")
+            set(ARCH "x86_64")
+        elseif("${MSVC_C_ARCHITECTURE_ID}" STREQUAL "ARM")
+            set(ARCH "arm")
+        elseif("${MSVC_C_ARCHITECTURE_ID}" STREQUAL "ARM64")
+            set(ARCH "arm64")
+        else()
+            message(FATAL_ERROR "Unrecognized architecture ${MSVC_C_ARCHITECTURE_ID} from ${CMAKE_C_COMPILER}")
+        endif()
+    elseif("${CMAKE_C_COMPILER_ID}" STREQUAL "Clang" OR "${CMAKE_C_COMPILER_ID}" STREQUAL "GNU")
+        if("${CMAKE_C_COMPILER_ID}" STREQUAL "Clang")
+            set(CMD_ARGS "-target" "${CMAKE_C_COMPILER_TARGET}" "-dumpmachine")
+        else()
+            set(CMD_ARGS "-dumpmachine")
+        endif()
+        execute_process(
+            COMMAND 
+                ${CMAKE_C_COMPILER} ${CMD_ARGS}
+            RESULT_VARIABLE
+                RESULT
+            OUTPUT_VARIABLE
+                ARCH
+            ERROR_QUIET)
+        if(RESULT)
+            message(FATAL_ERROR "Failed to determine target architecture triplet: ${RESULT}")
+        endif()
+        string(REGEX MATCH "([^-]+).*" ARCH_MATCH ${ARCH})
+        if(NOT CMAKE_MATCH_1 OR NOT ARCH_MATCH)
+            message(FATAL_ERROR "Failed to match the target architecture triplet: ${ARCH}")
+        endif()
+        set(ARCH ${CMAKE_MATCH_1})
+
+        if("${ARCH}" STREQUAL "x86_64")
+            # Do nothing
+        elseif("${ARCH}" STREQUAL "aarch64")
+            set(ARCH "arm64")
+        elseif("${ARCH}" STREQUAL "i686")
+            # Do nothing
+        elseif("${ARCH}" STREQUAL "i386")
+            # Do nothing
+        else()
+            message(FATAL_ERROR "Unrecognized architecture ${ARCH} from ${CMAKE_C_COMPILER}")
+        endif()
+    else()
+        message(FATAL_ERROR "Unrecognized Compiler ${CMAKE_C_COMPILER_ID}")
+    endif()
+    message(STATUS "Target architecture - ${ARCH}")
+    set(${OUTPUT_VARIABLE} ${ARCH} PARENT_SCOPE)
+endfunction()
--- a/cmake/toolchains/arm64-linux-clang.cmake
+++ b/cmake/toolchains/arm64-linux-clang.cmake
@ -0,0 +1,22 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+# the name of the target OS and arch
+SET(CMAKE_SYSTEM_NAME Linux)
+SET(CMAKE_SYSTEM_PROCESSOR arm64)
+SET(triple aarch64-linux-gnu)
+
+# which compilers to use
+SET(CMAKE_C_COMPILER "clang-6.0")
+SET(CMAKE_C_COMPILER_TARGET ${triple})
+SET(CMAKE_CXX_COMPILER "clang++-6.0")
+SET(CMAKE_CXX_COMPILER_TARGET ${triple})
+
+# Tell pkgconfig to use aarch64
+SET(ENV{PKG_CONFIG_PATH} "/usr/lib/aarch64-linux-gnu/pkgconfig")
+
+# Tell CMake to use qemu to emulate
+# Note: This should be automatically done by Ubuntu using binfmt_misc, but that
+# seems to be broken on WSL (https://github.com/microsoft/WSL/issues/2620) so
+# explicitly setting the emulator for now.
+SET(CMAKE_CROSSCOMPILING_EMULATOR qemu-aarch64-static)
--- a/cmake/toolchains/arm64-linux-gnu.cmake
+++ b/cmake/toolchains/arm64-linux-gnu.cmake
@ -0,0 +1,19 @@
+# Copyright (c) Microsoft Corporation. All rights reserved.
+# Licensed under the MIT License.
+
+# the name of the target OS and arch
+SET(CMAKE_SYSTEM_NAME Linux)
+SET(CMAKE_SYSTEM_PROCESSOR arm64)
+
+# which compilers to use
+SET(CMAKE_C_COMPILER aarch64-linux-gnu-gcc)
+SET(CMAKE_CXX_COMPILER aarch64-linux-gnu-g++)
+
+# Tell pkgconfig to use arm64
+SET(ENV{PKG_CONFIG_PATH} "/usr/lib/aarch64-linux-gnu/pkgconfig")
+
+# Tell CMake to use qemu to emulate
+# Note: This should be automatically done by Ubuntu using binfmt_misc, but that
+# seems to be broken on WSL (https://github.com/microsoft/WSL/issues/2620) so
+# explicitly setting the emulator for now.
+SET(CMAKE_CROSSCOMPILING_EMULATOR qemu-aarch64-static)
--- a/cmake/toolchains/x86_64-linux-clang.cmake
+++ b/cmake/toolchains/x86_64-linux-clang.cmake
@ -4,12 +4,13 @@
 # the name of the target OS and arch
 SET(CMAKE_SYSTEM_NAME Linux)
 SET(CMAKE_SYSTEM_PROCESSOR x86_64)
+SET(triple x86_64-linux-gnu)

 # which compilers to use
 SET(CMAKE_C_COMPILER "clang-6.0")
-SET(CMAKE_C_FLAGS -m64)
+SET(CMAKE_C_COMPILER_TARGET ${triple})
 SET(CMAKE_CXX_COMPILER "clang++-6.0")
-SET(CMAKE_CXX_FLAGS -m64)
+SET(CMAKE_CXX_COMPILER_TARGET ${triple})

 # save flags to cache
 SET(CMAKE_C_FLAGS   ${CMAKE_C_FLAGS}   CACHE STRING "C Flags"   FORCE)
--- a/scripts/Dockerfile
+++ b/scripts/Dockerfile
@ -1,34 +0,0 @@
-# Use the official ubuntu:18.04 image as the parent image
-FROM ubuntu:18.04
-
-# Tell Ubuntu non-interactive install
-ARG DEBIAN_FRONTEND=noninteractive
-
-# Set the working directory to /app
-WORKDIR /app
-
-# Run apt-get steps
-RUN apt-get update && apt-get install -y \
-    pkg-config \
-    ninja-build \
-    doxygen \
-    clang \
-    gcc-multilib \
-    g++-multilib \
-    python3 \
-    git-lfs \
-    nasm \
-    cmake \
-    libgl1-mesa-dev \
-    libsoundio-dev \
-    libvulkan-dev \
-    libx11-dev \
-    libxcursor-dev \
-    libxinerama-dev \
-    libxrandr-dev \
-    libusb-1.0-0-dev \
-    libssl-dev \
-    libudev-dev \
-    mesa-common-dev \
-    uuid-dev \
-    libopencv-dev
--- a/scripts/docker/Dockerfile
+++ b/scripts/docker/Dockerfile
@ -0,0 +1,20 @@
+ARG UBUNTU_VERSION=18.04
+# Use the official ubuntu:18.04 image as the parent image
+FROM ubuntu:${UBUNTU_VERSION}
+
+# Set the working directory to /app
+WORKDIR /app
+
+ARG ARCH=amd64
+
+ADD setup-ubuntu.sh /app
+ADD sources.list /app
+
+# Tell Ubuntu non-interactive install
+ARG DEBIAN_FRONTEND=noninteractive
+
+
+RUN ./setup-ubuntu.sh ${ARCH}
+
+RUN apt-get install -y ca-certificates
+RUN update-ca-certificates
--- a/scripts/docker/setup-ubuntu.sh
+++ b/scripts/docker/setup-ubuntu.sh
@ -0,0 +1,75 @@
+#!/bin/bash
+
+# Usage:
+# ./setup-ubuntu.sh [arm64 | amd64]
+
+# Warning! This will override your sources.list file!!
+
+arch=amd64
+
+# Copy off old sources.list file
+cp /etc/apt/sources.list /etc/apt/sources.list.old
+echo "Backed up /etc/apt/sources.list to /etc/apt/sources.list.old"
+
+# Copy over the new file
+cp sources.list /etc/apt/sources.list
+echo "Overwrote /etc/apt/sources.list with sources.list"
+
+apt-get update
+
+apt-get install wget -y
+
+# Add Public microsoft repo keys to the image
+wget -q https://packages.microsoft.com/config/ubuntu/18.04/packages-microsoft-prod.deb
+dpkg -i packages-microsoft-prod.deb
+
+if [ "$1" = "arm64" ]; then
+    arch="arm64"
+fi
+
+echo "Setting up for building $arch binaries"
+
+dpkg --add-architecture amd64
+dpkg --add-architecture arm64
+
+apt-get update
+
+packages=(\
+    gcc-aarch64-linux-gnu \
+    g++-aarch64-linux-gnu \
+    file \
+    dpkg-dev \
+    qemu \
+    binfmt-support \
+    qemu-user-static \
+    pkg-config \
+    ninja-build \
+    doxygen \
+    clang \
+    python3 \
+    gcc \
+    g++ \
+    git \
+    git-lfs \
+    nasm \
+    cmake \
+    powershell \
+    libgl1-mesa-dev:$arch \
+    libsoundio-dev:$arch \
+    libjpeg-dev:$arch \
+    libvulkan-dev:$arch \
+    libx11-dev:$arch \
+    libxcursor-dev:$arch \
+    libxinerama-dev:$arch \
+    libxrandr-dev:$arch \
+    libusb-1.0-0-dev:$arch \
+    libssl-dev:$arch \
+    libudev-dev:$arch \
+    mesa-common-dev:$arch \
+    uuid-dev:$arch )
+
+if [ "$arch" = "amd64" ]; then
+    packages+=(libopencv-dev:$arch)
+fi
+
+apt-get install -y --no-install-recommends ${packages[@]}
--- a/scripts/docker/sources.list
+++ b/scripts/docker/sources.list
@ -0,0 +1,26 @@
+
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic main restricted
+deb [arch=arm64] http://ports.ubuntu.com/ bionic main restricted
+
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic-updates main restricted
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-updates main restricted
+
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic universe
+deb [arch=arm64] http://ports.ubuntu.com/ bionic universe
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic-updates universe
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-updates universe
+
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic multiverse
+deb [arch=arm64] http://ports.ubuntu.com/ bionic multiverse
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic-updates multiverse
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-updates multiverse
+
+deb [arch=amd64] http://archive.ubuntu.com/ubuntu/ bionic-backports main restricted universe multiverse
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-backports main restricted universe multiverse
+
+deb [arch=amd64] http://security.ubuntu.com/ubuntu/ bionic-security main restricted
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-security main restricted
+deb [arch=amd64] http://security.ubuntu.com/ubuntu/ bionic-security universe
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-security universe
+deb [arch=amd64] http://security.ubuntu.com/ubuntu/ bionic-security multiverse
+deb [arch=arm64] http://ports.ubuntu.com/ bionic-security multiverse
--- a/src/transformation/rgbz.c
+++ b/src/transformation/rgbz.c
@ -8,9 +8,12 @@
 #include <limits.h>
 #include <math.h>

+#if defined(__amd64__) || defined(_M_AMD64) || defined(__i386__) || defined(_M_X86)
+#define K4A_USING_SSE
 #include <emmintrin.h> // SSE2
 #include <tmmintrin.h> // SSE3
 #include <smmintrin.h> // SSE4.1
+#endif

 typedef struct _k4a_transformation_input_image_t
 {
@ -1055,42 +1058,45 @@ k4a_buffer_result_t transformation_color_image_to_depth_camera_internal(
    return K4A_BUFFER_RESULT_SUCCEEDED;
 }

+#if !defined(K4A_USING_SSE)
 // This is the same function as transformation_depth_to_xyz without the SSE
 // instructions. This code is kept here for readability.
-// static void transformation_depth_to_xyz(k4a_transformation_xy_tables_t *xy_tables,
-//                                         const void *depth_image_data,
-//                                         void *xyz_image_data)
-// {
-//     const uint16_t *depth_image_data_uint16 = (const uint16_t *)depth_image_data;
-//     int16_t *xyz_data_int16 = (int16_t *)xyz_image_data;
-//     int16_t x, y, z;
+static void transformation_depth_to_xyz(k4a_transformation_xy_tables_t *xy_tables,
+                                        const void *depth_image_data,
+                                        void *xyz_image_data)
+{
+    const uint16_t *depth_image_data_uint16 = (const uint16_t *)depth_image_data;
+    int16_t *xyz_data_int16 = (int16_t *)xyz_image_data;
+    int16_t x, y, z;

-//      for (int i = 0; i < xy_tables->width * xy_tables->height; i++)
-//     {
-//         float x_tab = xy_tables->x_table[i];
+    for (int i = 0; i < xy_tables->width * xy_tables->height; i++)
+    {
+        float x_tab = xy_tables->x_table[i];

-//          if (!isnan(x_tab))
-//         {
-//             z = (int16_t)depth_image_data_uint16[i];
-//             x = (int16_t)(floorf(x_tab * (float)z + 0.5f));
-//             y = (int16_t)(floorf(xy_tables->y_table[i] * (float)z + 0.5f));
-//         }
-//         else
-//         {
-//             x = 0;
-//             y = 0;
-//             z = 0;
-//         }
+        if (!isnan(x_tab))
+        {
+            z = (int16_t)depth_image_data_uint16[i];
+            x = (int16_t)(floorf(x_tab * (float)z + 0.5f));
+            y = (int16_t)(floorf(xy_tables->y_table[i] * (float)z + 0.5f));
+        }
+        else
+        {
+            x = 0;
+            y = 0;
+            z = 0;
+        }

-//          xyz_data_int16[3 * i + 0] = x;
-//         xyz_data_int16[3 * i + 1] = y;
-//         xyz_data_int16[3 * i + 2] = z;
-//     }
-// }
+        xyz_data_int16[3 * i + 0] = x;
+        xyz_data_int16[3 * i + 1] = y;
+        xyz_data_int16[3 * i + 2] = z;
+    }
+}

-static void transformation_depth_to_xyz_sse(k4a_transformation_xy_tables_t *xy_tables,
-                                            const void *depth_image_data,
-                                            void *xyz_image_data)
+#else
+
+static void transformation_depth_to_xyz(k4a_transformation_xy_tables_t *xy_tables,
+                                        const void *depth_image_data,
+                                        void *xyz_image_data)
 {
    const __m128i *depth_image_data_m128i = (const __m128i *)depth_image_data;
 #if defined(__clang__) || defined(__GNUC__)
@ -1159,6 +1165,7 @@ static void transformation_depth_to_xyz_sse(k4a_transformation_xy_tables_t *xy_t
        *xyz_data_m128i++ = _mm_blend_epi16(_mm_blend_epi16(x, y, 0x49), z, 0x92);
    }
 }
+#endif

 k4a_buffer_result_t
 transformation_depth_image_to_point_cloud_internal(k4a_transformation_xy_tables_t *xy_tables,
@ -1207,7 +1214,7 @@ transformation_depth_image_to_point_cloud_internal(k4a_transformation_xy_tables_
        return K4A_BUFFER_RESULT_FAILED;
    }

-    transformation_depth_to_xyz_sse(xy_tables, (const void *)depth_image_data, (void *)xyz_image_data);
+    transformation_depth_to_xyz(xy_tables, (const void *)depth_image_data, (void *)xyz_image_data);

    return K4A_BUFFER_RESULT_SUCCEEDED;
 }