[stdgpu] add new port (#53264)

Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>

ports/stdgpu/fix-msvc-preprocessor.patch

@@ -0,0 +1,13 @@
+diff --git a/src/stdgpu/CMakeLists.txt b/src/stdgpu/CMakeLists.txt
+index 937b60d..3a49dd2 100644
+--- a/src/stdgpu/CMakeLists.txt
++++ b/src/stdgpu/CMakeLists.txt
+@@ -145,6 +145,8 @@ target_compile_features(stdgpu PUBLIC cxx_std_17)
+ target_compile_options(stdgpu PRIVATE ${STDGPU_DEVICE_FLAGS}
+                                       ${STDGPU_HOST_FLAGS})
+ 
++target_compile_options(stdgpu PUBLIC $<$<COMPILE_LANG_AND_ID:CXX,MSVC>:/Zc:preprocessor>)
++
+ target_compile_definitions(stdgpu PRIVATE $<$<PLATFORM_ID:Windows>:NOMINMAX>)
+ 
+ target_link_libraries(stdgpu PUBLIC thrust::thrust)

ports/stdgpu/portfile.cmake

@@ -0,0 +1,55 @@
+# stdgpu's shared build on Windows exports no symbols and does not install its
+# DLL, so only static linkage is usable there.
+if(VCPKG_TARGET_IS_WINDOWS)
+    vcpkg_check_linkage(ONLY_STATIC_LIBRARY)
+endif()
+
+vcpkg_from_github(
+    OUT_SOURCE_PATH SOURCE_PATH
+    REPO stotko/stdgpu
+    # No release newer than 1.3.0 (2020-06); pin the tested master snapshot.
+    REF 22599cfdcc185c2ed721d78c792e3193cf84cc01
+    SHA512 de16413c81c7d95e48f2a5fbff018748e629a2316bc6c76c42f881b6922b337a12a1ec20eb732465ff244a04e22afe1846fdbacc9e4f2a72aa0204a0c11d397e
+    HEAD_REF master
+    # Current CCCL headers require the conforming MSVC preprocessor.
+    PATCHES
+        fix-msvc-preprocessor.patch
+)
+
+vcpkg_find_cuda(OUT_CUDA_TOOLKIT_ROOT CUDA_TOOLKIT_ROOT)
+
+string(COMPARE EQUAL "${VCPKG_LIBRARY_LINKAGE}" "dynamic" STDGPU_BUILD_SHARED_LIBS)
+
+vcpkg_cmake_configure(
+    SOURCE_PATH "${SOURCE_PATH}"
+    OPTIONS
+        -DSTDGPU_BACKEND=STDGPU_BACKEND_CUDA
+        -DSTDGPU_BUILD_SHARED_LIBS=${STDGPU_BUILD_SHARED_LIBS}
+        -DSTDGPU_SETUP_COMPILER_FLAGS=OFF
+        -DSTDGPU_COMPILE_WARNING_AS_ERROR=OFF
+        -DSTDGPU_BUILD_EXAMPLES=OFF
+        -DSTDGPU_BUILD_BENCHMARKS=OFF
+        -DSTDGPU_BUILD_TESTS=OFF
+        -DSTDGPU_BUILD_DOCUMENTATION=OFF
+        -DSTDGPU_ENABLE_CONTRACT_CHECKS=OFF
+        -DCMAKE_CUDA_COMPILER=${NVCC}
+        -DCUDAToolkit_ROOT=${CUDA_TOOLKIT_ROOT}
+        # Consistent with other CUDA ports (ginkgo, colmap); override via
+        # VCPKG_CMAKE_CONFIGURE_OPTIONS if needed.
+        -DCMAKE_CUDA_ARCHITECTURES=native
+)
+
+vcpkg_cmake_install()
+
+vcpkg_cmake_config_fixup(CONFIG_PATH lib/cmake/stdgpu)
+
+vcpkg_copy_pdbs()
+
+file(REMOVE_RECURSE
+    "${CURRENT_PACKAGES_DIR}/debug/include"
+    "${CURRENT_PACKAGES_DIR}/debug/share"
+)
+
+file(INSTALL "${CMAKE_CURRENT_LIST_DIR}/usage" DESTINATION "${CURRENT_PACKAGES_DIR}/share/${PORT}")
+
+vcpkg_install_copyright(FILE_LIST "${SOURCE_PATH}/LICENSE")

ports/stdgpu/usage

@@ -0,0 +1,6 @@
+stdgpu provides CMake targets:
+
+    find_package(stdgpu CONFIG REQUIRED)
+    target_link_libraries(main PRIVATE stdgpu::stdgpu)
+
+This port builds stdgpu with the CUDA backend.

ports/stdgpu/vcpkg.json

@@ -0,0 +1,19 @@
+{
+  "name": "stdgpu",
+  "version-date": "2026-07-08",
+  "description": "Open-source library providing generic GPU data structures for fast and reliable data management",
+  "homepage": "https://github.com/stotko/stdgpu",
+  "license": "Apache-2.0",
+  "supports": "(windows & x64 & !uwp & !xbox) | (linux & x64)",
+  "dependencies": [
+    "cuda",
+    {
+      "name": "vcpkg-cmake",
+      "host": true
+    },
+    {
+      "name": "vcpkg-cmake-config",
+      "host": true
+    }
+  ]
+}

versions/baseline.json

@@ -9812,6 +9812,10 @@
       "baseline": "2026-05-25",
       "port-version": 0
     },
+    "stdgpu": {
+      "baseline": "2026-07-08",
+      "port-version": 0
+    },
     "stduuid": {
       "baseline": "1.2.3",
       "port-version": 0

versions/s-/stdgpu.json

@@ -0,0 +1,9 @@
+{
+  "versions": [
+    {
+      "git-tree": "601763dfa5708121970135c199eca4f769644593",
+      "version-date": "2026-07-08",
+      "port-version": 0
+    }
+  ]
+}