Merge branch 'main' into johnnynunez/main

Author: ssheorey

.github/workflows/macos.yml

@@ -73,21 +73,6 @@ jobs:
             ln -s $(brew --prefix gcc)/bin/gfortran-* /usr/local/bin/gfortran
           fi
 
-          # Install libomp 11.1.0 from old brew bottle for x64 catalina (10.15)
-          # / arm64 big sur (11.0). Directly installing the Ruby formula will
-          # install for the current OS.
-          # https://github.com/microsoft/LightGBM/issues/4229
-          if [[ ${{ runner.arch}} == "X64"  ]]; then
-            # x64 catalina (10.15) bottle
-            export LIBOMP_BOTTLE_HASH=45a5aa653bd45bd5ff5858580b1a4670c4b5a51ea29d68d45a53f72f56010e05
-          else # ARM64
-            # arm64 big_sur (11.0) bottle
-            export LIBOMP_BOTTLE_HASH=f87f7841eb8b72650fa771af39642361aec371ea1a1f94f081ecc0e8168a0e75
-          fi
-          curl -L -H "Authorization: Bearer QQ==" -o libomp-11.1.0.bottle.tar.gz \
-          https://ghcr.io/v2/homebrew/core/libomp/blobs/sha256:$LIBOMP_BOTTLE_HASH
-          brew install -f libomp-11.1.0.bottle.tar.gz
-
           ccache -M 2G  # See .github/workflows/readme.md for ccache strategy.
       - name: Config and build
         run: |
@@ -204,17 +189,6 @@ jobs:
             ln -s $(which gfortran-13) /usr/local/bin/gfortran
           fi
 
-          # Install libomp 11.1.0. See comment above.
-          if [[ ${{ runner.arch}} == "X64"  ]]; then
-            # x64 catalina (10.15) bottle
-            export LIBOMP_BOTTLE_HASH=45a5aa653bd45bd5ff5858580b1a4670c4b5a51ea29d68d45a53f72f56010e05
-          else # ARM64
-            # arm64 big_sur (11.0) bottle
-            export LIBOMP_BOTTLE_HASH=f87f7841eb8b72650fa771af39642361aec371ea1a1f94f081ecc0e8168a0e75
-          fi
-          curl -L -H "Authorization: Bearer QQ==" -o libomp-11.1.0.bottle.tar.gz \
-          https://ghcr.io/v2/homebrew/core/libomp/blobs/sha256:$LIBOMP_BOTTLE_HASH
-          brew install -f libomp-11.1.0.bottle.tar.gz
           brew install ccache
           ccache -M 2G  # See .github/workflows/readme.md for ccache strategy.
 

.github/workflows/style.yml

@@ -1,5 +1,4 @@
 name: Style Check
-permissions: {}
 
 on:
   workflow_dispatch:
@@ -16,18 +15,49 @@ concurrency:
 jobs:
   style-check:
     permissions:
-      contents: read
+      contents: write
     runs-on: ubuntu-latest
     steps:
       - name: Checkout source code
         uses: actions/checkout@v4
+        with:
+          repository: ${{ github.event.pull_request.head.repo.full_name || github.repository }}
+          ref: ${{ github.event.pull_request.head.sha || github.ref }}
+          token: ${{ secrets.GITHUB_TOKEN }}
       - name: Set up Python version
         uses: actions/setup-python@v5
         with:
           python-version: '3.12'
       - name: Install dependencies
         run: |
           pip install -U -r python/requirements_style.txt
-      - name: Run style check
+      - name: Apply style formatting
+        if: github.event_name == 'pull_request'
+        run: |
+          python util/check_style.py --apply
+      - name: Check style formatting
+        if: github.event_name != 'pull_request'
         run: |
           python util/check_style.py
+      - name: Commit and push changes
+        if: github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name == github.repository
+        run: |
+          git config --global user.name 'github-actions[bot]'
+          git config --global user.email 'github-actions[bot]@users.noreply.github.com'
+          git add -u
+          if ! git diff --staged --quiet; then
+            git commit -m "Apply automatic code formatting"
+            git push || exit 1
+          else
+            echo "No formatting changes needed"
+          fi
+      - name: Check for formatting changes from fork
+        if: github.event_name == 'pull_request' && github.event.pull_request.head.repo.full_name != github.repository
+        run: |
+          git add -u
+          if ! git diff --staged --quiet; then
+            echo "::error::This PR requires formatting changes but is from a fork. Please run 'python util/check_style.py --apply' locally and push the changes."
+            exit 1
+          else
+            echo "No formatting changes needed"
+          fi

3rdparty/find_dependencies.cmake

@@ -161,6 +161,7 @@ endfunction()
 # CMake arguments for configuring ExternalProjects. Use the second _hidden
 # version by default.
 set(ExternalProject_CMAKE_ARGS
+    -DCMAKE_POLICY_VERSION_MINIMUM=3.5      # for VTK 9.1
     -DCMAKE_C_COMPILER=${CMAKE_C_COMPILER}
     -DCMAKE_CXX_COMPILER=${CMAKE_CXX_COMPILER}
     -DCMAKE_CUDA_COMPILER=${CMAKE_CUDA_COMPILER}

3rdparty/uvatlas/uvatlas.cmake

@@ -4,9 +4,9 @@ include(ExternalProject)
 ExternalProject_Add(
     ext_directxheaders
     PREFIX uvatlas
-    GIT_REPOSITORY https://github.com/microsoft/DirectX-Headers.git
-    GIT_TAG v1.606.3
-    GIT_SHALLOW TRUE
+    URL https://github.com/microsoft/DirectX-Headers/archive/v1.606.3/DirectX-Headers-v1.606.3.tar.gz
+    URL_HASH
+    SHA256=bf0183981e505336e918609374907c934b99eb61c0826d75a5649f41568abc4b
     DOWNLOAD_DIR "${OPEN3D_THIRD_PARTY_DOWNLOAD_DIR}/uvatlas"
     UPDATE_COMMAND ""
     CMAKE_ARGS
@@ -22,9 +22,9 @@ ExternalProject_Add(
 ExternalProject_Add(
     ext_directxmath
     PREFIX uvatlas
-    GIT_REPOSITORY https://github.com/microsoft/DirectXMath.git
-    GIT_TAG may2022
-    GIT_SHALLOW TRUE
+    URL https://github.com/microsoft/DirectXMath/archive/may2022/DirectXMath-may2022.tar.gz
+    URL_HASH
+    SHA256=b2c5b419ca2c567860f7c204c9c0890573e8a58c8d877473e4f3ba6b851ca4ce
     DOWNLOAD_DIR "${OPEN3D_THIRD_PARTY_DOWNLOAD_DIR}/uvatlas"
     UPDATE_COMMAND ""
     CMAKE_ARGS
@@ -63,4 +63,4 @@ if(NOT WIN32)
     list(APPEND UVATLAS_INCLUDE_DIRS "${INSTALL_DIR}/include/wsl/stubs/")
 endif()
 set(UVATLAS_LIB_DIR ${INSTALL_DIR}/${Open3D_INSTALL_LIB_DIR})
-set(UVATLAS_LIBRARIES UVAtlas)
+set(UVATLAS_LIBRARIES UVAtlas)

CHANGELOG.md

@@ -63,7 +63,13 @@
 -   Exposed `get_plotly_fig` and modified `draw_plotly` to return the `Figure` it creates. (PR #7258)
 -   Fix build with librealsense v2.44.0 and upcoming VS 2022 17.13 (PR #7074)
 -   Fix `deprecated-declarations` warnings when compiling code with C++20 standard (PR #7303)
+-   Fix advanced indexing bug with sliced boolean masks on CUDA devices (PR #7340)
+-   Fix logic for adding -allow-unsupported-compiler to nvcc (PR #7337)
+-   Fix linker error "library limit of 65535 objects exceeded" with Ninja generator on MSVC (PR #7335)
+-   Download tarballs instead of Git repos for "3rdparty/uvatlas" (PR #7371)
 -   macOS x86_64 not longer supported, only macOS arm64 is supported.
+
+
 ## 0.13
 
 -   CUDA support 10.1 -> 11.0. Tensorflow 2.3.1 -> 2.4.1. PyTorch 1.6.0 -> 1.7.1 (PR #3049). This requires a custom PyTorch wheel from <https://github.com/isl-org/open3d_downloads/releases/tag/torch1.7.1> due to PyTorch issue #52663

CMakeLists.txt

@@ -125,9 +125,9 @@ option(USE_SYSTEM_VTK             "Use system pre-installed VTK"             OFF
 option(USE_SYSTEM_ZEROMQ          "Use system pre-installed ZeroMQ"          OFF)
 if(LINUX_AARCH64 OR APPLE_AARCH64)
     option(BUILD_VTK_FROM_SOURCE      "Build VTK from source"                ON )
-else()
-    option(BUILD_VTK_FROM_SOURCE      "Build VTK from source"                OFF)
-endif()
+ else()
+     option(BUILD_VTK_FROM_SOURCE      "Build VTK from source"                OFF)
+ endif()
 if(LINUX_AARCH64)
     option(BUILD_FILAMENT_FROM_SOURCE "Build filament from source"           ON )
 else()
@@ -395,14 +395,21 @@ cmake_language(EVAL CODE "cmake_language(DEFER CALL open3d_patch_findthreads_mod
 
 # Build CUDA module by default if CUDA is available
 if(BUILD_CUDA_MODULE)
-    # Suppress nvcc unsupported compiler error for MSVC 2022 with CUDA 11.7 to 12.4
-    # https://forums.developer.nvidia.com/t/problems-with-latest-vs2022-update/294150/12
-    if (MSVC AND MSVC_VERSION VERSION_LESS_EQUAL "1949")
-        # Set this before any CUDA checks
-        add_compile_definitions(_ALLOW_COMPILER_AND_STL_VERSION_MISMATCH)
-        set(CMAKE_CUDA_FLAGS "--allow-unsupported-compiler" CACHE STRING "Additional flags for nvcc" FORCE)
-        message(WARNING "Using --allow-unsupported-compiler flag for nvcc with MSVC 2022. "
-        "Set $Env:NVCC_PREPEND_FLAGS='--allow-unsupported-compiler' if nvcc still fails.")
+    if(MSVC)
+        # Handle/suppress nvcc unsupported compiler error for MSVC>=1940 with CUDA 11.7 to 12.4:
+        # (https://forums.developer.nvidia.com/t/problems-with-latest-vs2022-update/294150/12)
+        # Find CUDAToolkit first to get CUDAToolkit_VERSION:
+        find_package(CUDAToolkit REQUIRED)
+        if (CUDAToolkit_VERSION VERSION_LESS "12.5" AND MSVC_VERSION GREATER_EQUAL 1940)
+            # Set required nvcc flags before enable_language(CUDA).
+            # Note: CMake >=3.29.4 might be needed for CMAKE_CUDA_FLAGS to be passed correctly to
+            # CMake's try_compile environment.
+            set(CMAKE_CUDA_FLAGS
+                  "${CMAKE_CUDA_FLAGS} -allow-unsupported-compiler -D_ALLOW_COMPILER_AND_STL_VERSION_MISMATCH"
+                  CACHE STRING "Flags for NVCC" FORCE)
+            message(WARNING "Using --allow-unsupported-compiler flag and -D_ALLOW_COMPILER_AND_STL_VERSION_MISMATCH for nvcc<=12.4 with MSVC>=1940. "
+            "Set $Env:NVCC_PREPEND_FLAGS='--allow-unsupported-compiler' if nvcc still fails.")
+        endif()
     endif()
     if (CMAKE_CUDA_ARCHITECTURES)
         message(STATUS "Building with user-provided CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")
@@ -498,6 +505,10 @@ if(WIN32)
         #
         # See: https://github.com/tensorflow/tensorflow/pull/10962
         add_compile_options("$<$<COMPILE_LANGUAGE:CXX>:/bigobj;/Ob2>")
+        # MSBuild adds /Zc:inline by default, but it needs to be added explicitly for other generators (e.g. Ninja),
+        # for Open3D to compile (the linker fails otherwise with a "fatal error LNK1189: library limit of 65535
+        # objects exceeded"). /Zc:inline will remove unreferenced data and functions from object files.
+        add_compile_options($<$<COMPILE_LANGUAGE:CXX>:/Zc:inline>)
     endif()
     if (STATIC_WINDOWS_RUNTIME)
         set(CMAKE_MSVC_RUNTIME_LIBRARY "MultiThreaded$<$<CONFIG:Debug>:Debug>")

cpp/benchmarks/t/geometry/PointCloud.cpp

@@ -189,8 +189,8 @@ void EstimateNormals(benchmark::State& state,
                      const core::Device& device,
                      const core::Dtype& dtype,
                      const double voxel_size,
-                     const utility::optional<int> max_nn,
-                     const utility::optional<double> radius) {
+                     const std::optional<int> max_nn,
+                     const std::optional<double> radius) {
     t::geometry::PointCloud pcd;
     t::io::ReadPointCloud(path, pcd, {"auto", false, false, false});
 
@@ -496,30 +496,30 @@ BENCHMARK_CAPTURE(EstimateNormals,
                   core::Float32,
                   0.02,
                   30,
-                  utility::nullopt)
+                  std::nullopt)
         ->Unit(benchmark::kMillisecond);
 BENCHMARK_CAPTURE(EstimateNormals,
                   CPU F64 KNN[0.02 | 30],
                   core::Device("CPU:0"),
                   core::Float64,
                   0.02,
                   30,
-                  utility::nullopt)
+                  std::nullopt)
         ->Unit(benchmark::kMillisecond);
 BENCHMARK_CAPTURE(EstimateNormals,
                   CPU F32 Radius[0.02 | 0.06],
                   core::Device("CPU:0"),
                   core::Float32,
                   0.02,
-                  utility::nullopt,
+                  std::nullopt,
                   0.06)
         ->Unit(benchmark::kMillisecond);
 BENCHMARK_CAPTURE(EstimateNormals,
                   CPU F64 Radius[0.02 | 0.06],
                   core::Device("CPU:0"),
                   core::Float64,
                   0.02,
-                  utility::nullopt,
+                  std::nullopt,
                   0.06)
         ->Unit(benchmark::kMillisecond);
 #ifdef BUILD_CUDA_MODULE
@@ -545,30 +545,30 @@ BENCHMARK_CAPTURE(EstimateNormals,
                   core::Float32,
                   0.02,
                   30,
-                  utility::nullopt)
+                  std::nullopt)
         ->Unit(benchmark::kMillisecond);
 BENCHMARK_CAPTURE(EstimateNormals,
                   CUDA F64 KNN[0.02 | 30],
                   core::Device("CUDA:0"),
                   core::Float64,
                   0.02,
                   30,
-                  utility::nullopt)
+                  std::nullopt)
         ->Unit(benchmark::kMillisecond);
 BENCHMARK_CAPTURE(EstimateNormals,
                   CUDA F32 Radius[0.02 | 0.06],
                   core::Device("CUDA:0"),
                   core::Float32,
                   0.02,
-                  utility::nullopt,
+                  std::nullopt,
                   0.06)
         ->Unit(benchmark::kMillisecond);
 BENCHMARK_CAPTURE(EstimateNormals,
                   CUDA F64 Radius[0.02 | 0.06],
                   core::Device("CUDA:0"),
                   core::Float64,
                   0.02,
-                  utility::nullopt,
+                  std::nullopt,
                   0.06)
         ->Unit(benchmark::kMillisecond);
 #endif