feat(rust/sedona-spatial-join): Add GPU-accelerated spatial join support

.github/workflows/rust-gpu.yml

@@ -27,15 +27,13 @@ on:
       - main
     paths:
       - 'c/sedona-libgpuspatial/**'
-      - 'rust/sedona-spatial-join-gpu/**'
       - '.github/workflows/rust-gpu.yml'
 
   push:
     branches:
       - main
     paths:
       - 'c/sedona-libgpuspatial/**'
-      - 'rust/sedona-spatial-join-gpu/**'
       - '.github/workflows/rust-gpu.yml'
 
 concurrency:
@@ -66,7 +64,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        name: [ "clippy", "docs", "test", "build" ]
+        name: [ "build_tests", "build_lib", "build_package" ]
 
     name: "${{ matrix.name }}"
     runs-on: ubuntu-latest
@@ -181,15 +179,6 @@ jobs:
           # Bump the number at the end of this line to force a new dependency build
           key: vcpkg-installed-${{ runner.os }}-${{ runner.arch }}-${{ env.VCPKG_REF }}-3
 
-      # Install vcpkg dependencies from vcpkg.json manifest
-      - name: Install vcpkg dependencies
-        if: steps.cache-vcpkg.outputs.cache-hit != 'true'
-        run: |
-          ./vcpkg/vcpkg install abseil openssl
-          # Clean up vcpkg buildtrees and downloads to save space
-          rm -rf vcpkg/buildtrees
-          rm -rf vcpkg/downloads
-
       - name: Use stable Rust
         id: rust
         run: |
@@ -200,10 +189,20 @@ jobs:
         with:
           prefix-key: "rust-gpu-v4"
 
+      - name: Build libgpuspatial Tests
+        if: matrix.name == 'build_tests'
+        run: |
+          echo "=== Building libgpuspatial tests ==="
+          cd c/sedona-libgpuspatial/libgpuspatial
+          mkdir build
+          cmake --preset=default-with-tests -S . -B build
+          cmake --build build --target all
+
       # Build WITH GPU feature to compile CUDA code
       # CUDA compilation (nvcc) works without GPU hardware
       # Only GPU runtime execution requires actual GPU
       - name: Build libgpuspatial (with CUDA compilation)
+        if: matrix.name == 'build_lib'
         run: |
           echo "=== Building libgpuspatial WITH GPU feature ==="
           echo "Compiling CUDA code using nvcc (no GPU hardware needed for compilation)"
@@ -215,10 +214,7 @@ jobs:
           # --lib builds only the library, not test binaries
           cargo build --locked --package sedona-libgpuspatial --lib --features gpu --verbose
 
-      - name: Build libgpuspatial Tests
+      - name: Build GPU Spatial Join Package
+        if: matrix.name == 'build_package'
         run: |
-          echo "=== Building libgpuspatial tests ==="
-          cd c/sedona-libgpuspatial/libgpuspatial
-          mkdir build
-          cmake --preset=default-with-tests -S . -B build
-          cmake --build build --target all
+          cargo build --workspace --package sedona-spatial-join --features gpu --verbose

.github/workflows/rust.yml

@@ -151,7 +151,7 @@ jobs:
       - name: Test
         if: matrix.name == 'test'
         run: |
-          cargo test --workspace --all-targets --all-features
+          cargo test --workspace --all-targets # Test all default features but GPU
           # Clean up intermediate build artifacts to free disk space aggressively
           cargo clean -p sedona-s2geography
           rm -rf target/debug/deps

.gitignore

@@ -51,3 +51,4 @@ __pycache__
 dev/release/.env
 
 /.luarc.json
+venv/

Cargo.toml

@@ -151,6 +151,7 @@ sedona-testing = { version = "0.3.0", path = "rust/sedona-testing" }
 # C wrapper crates
 sedona-geoarrow-c = { version = "0.3.0", path = "c/sedona-geoarrow-c" }
 sedona-geos = { version = "0.3.0", path = "c/sedona-geos" }
+sedona-libgpuspatial = { version = "0.3.0", path = "c/sedona-libgpuspatial" }
 sedona-proj = { version = "0.3.0", path = "c/sedona-proj", default-features = false }
 sedona-s2geography = { version = "0.3.0", path = "c/sedona-s2geography" }
 sedona-tg = { version = "0.3.0", path = "c/sedona-tg" }

c/sedona-geoarrow-c/build.rs

@@ -27,6 +27,7 @@ fn main() {
         .include("src/")
         .flag("-DGEOARROW_NAMESPACE=SedonaDB")
         .flag("-DNANOARROW_NAMESPACE=SedonaDB")
+        .flag("-Wno-type-limits")
         .compile("geoarrow");
 
     cc::Build::new()

c/sedona-libgpuspatial/Cargo.toml

@@ -40,8 +40,11 @@ which = "8.0"
 arrow-array = { workspace = true, features = ["ffi"] }
 arrow-schema = { workspace = true }
 thiserror = { workspace = true }
+geo = { workspace = true }
+wkt = { workspace = true }
 log = "0.4"
 sedona-schema = { path = "../../rust/sedona-schema" }
+nvml-wrapper = "0.10.0"
 
 [dev-dependencies]
 sedona-expr = { path = "../../rust/sedona-expr" }

c/sedona-libgpuspatial/build.rs

@@ -119,6 +119,13 @@ fn main() {
                 println!("cargo:warning=CMAKE_CUDA_ARCHITECTURES environment variable not set. Defaulting to '86;89'.");
                 "86;89".to_string()
             });
+        // Determine the build profile to match Cargo's debug/release mode
+        let profile_mode = if cfg!(debug_assertions) {
+            "Debug"
+        } else {
+            "Release"
+        };
+
         let dst = cmake::Config::new("./libgpuspatial")
             .define("CMAKE_CUDA_ARCHITECTURES", cuda_architectures)
             .define("CMAKE_POLICY_VERSION_MINIMUM", "3.5") // Allow older CMake versions
@@ -157,6 +164,17 @@ fn main() {
         println!("cargo:rustc-link-lib=static=gpuspatial");
         println!("cargo:rustc-link-lib=static=rmm");
         println!("cargo:rustc-link-lib=static=rapids_logger");
+        // Use the 'd' suffix for the debug build of spdlog (libspdlogd.a)
+        let spdlog_lib_name = if cfg!(debug_assertions) {
+            "spdlogd"
+        } else {
+            "spdlog"
+        };
+        println!(
+            "cargo:warning=Linking spdlog in {} mode: lib{}.a",
+            profile_mode, spdlog_lib_name
+        );
+        println!("cargo:rustc-link-lib=static={}", spdlog_lib_name);
         println!("cargo:rustc-link-lib=static=geoarrow");
         println!("cargo:rustc-link-lib=static=nanoarrow");
         println!("cargo:rustc-link-lib=stdc++");

c/sedona-libgpuspatial/libgpuspatial/CMakeLists.txt

@@ -132,8 +132,12 @@ config_shaders(PTX_FILES)
 
 message("-- Config shader PTX files ${PTX_FILES}")
 
-add_library(gpuspatial src/rt/rt_engine.cpp src/relate_engine.cu src/spatial_joiner.cu
-                       ${PTX_FILES})
+add_library(gpuspatial
+            src/rt/rt_engine.cpp
+            src/relate_engine.cu
+            src/rt_spatial_index.cu
+            src/rt_spatial_refiner.cu
+            ${PTX_FILES})
 
 # Link libraries
 target_link_libraries(gpuspatial
@@ -142,8 +146,7 @@ target_link_libraries(gpuspatial
                              cuda
                              rmm::rmm
                              rapids_logger::rapids_logger
-                             OptiX
-                      PRIVATE zstd)
+                             OptiX)
 
 # Set include directories
 target_include_directories(gpuspatial

c/sedona-libgpuspatial/libgpuspatial/CMakePresets.json

@@ -31,7 +31,7 @@
             "name": "default",
             "configurePreset": "default-with-tests",
             "environment": {
-                "GPUSPATIAL_TEST_DIR": "${sourceDir}/test_data"
+                "GPUSPATIAL_TEST_DIR": "${sourceDir}/test/data"
             }
         }
     ]

c/sedona-libgpuspatial/libgpuspatial/cmake/thirdparty/get_geoarrow.cmake

@@ -47,6 +47,7 @@ function(find_and_configure_geoarrow)
                   "BUILD_SHARED_LIBS OFF"
                   ${_exclude_from_all})
   set_target_properties(geoarrow PROPERTIES POSITION_INDEPENDENT_CODE ON)
+  target_compile_options(geoarrow PRIVATE -Wno-conversion)
   rapids_export_find_package_root(BUILD
                                   geoarrow
                                   "${geoarrow_BINARY_DIR}"

c/sedona-libgpuspatial/libgpuspatial/cmake/thirdparty/get_nanoarrow.cmake

@@ -48,6 +48,10 @@ function(find_and_configure_nanoarrow)
                   "NANOARROW_NAMESPACE gpuspatial"
                   ${_exclude_from_all})
   set_target_properties(nanoarrow PROPERTIES POSITION_INDEPENDENT_CODE ON)
+  if(TARGET nanoarrow_ipc) # Tests need this
+    target_compile_options(nanoarrow_ipc PRIVATE -Wno-conversion)
+  endif()
+  target_compile_options(nanoarrow PRIVATE -Wno-conversion)
   rapids_export_find_package_root(BUILD
                                   nanoarrow
                                   "${nanoarrow_BINARY_DIR}"