Cuda CI windows

.github/workflows/build_upload_test.yaml

@@ -52,22 +52,44 @@ jobs:
         if: ${{ matrix.os_base == 'macos'}}
         run: sudo xcode-select --switch /Applications/Xcode_11.5.app/Contents/Developer
 
+      - name: Install CUDA (Windows)
+        if: ${{matrix.os_base == 'windows'}}
+        uses: Jimver/[email protected]
+        with:
+          cuda: '11.2.2'
+
       - name: Build SplinePSF (UNIX)
         if: ${{ matrix.os_base == 'ubuntu' || matrix.os_base == 'macos'}}
         run: |
           cd ./dist_tools/conda
 
           conda-build -c conda-forge spline
           conda-build -c conda-forge spline --output > $FILENAMES
-
+      
       - name: Build SplinePSF (Windows)
         if: ${{ matrix.os_base == 'windows' }}
         run: |
           Set-Location -Path .\dist_tools\conda
-
+          $env:VSCMD_DEBUG=3
           conda-build -c conda-forge spline
           conda-build -c conda-forge spline --output > $Env:FILENAMES
 
+      - name: Upload CMake error log
+        uses: actions/upload-artifact@v2
+        with:
+          name: CMakeError.log
+          path: |
+            C:/Miniconda/envs/test/conda-bld/*/work/python/build/temp.win-amd64-3.*/Release/CMakeFiles/CMakeError.log
+            C:/Miniconda/envs/test/conda-bld/*/work/python/build/temp.win-amd64-3.*/Release/CMakeFiles/CMakeOutput.log
+
+      - name: Upload other CMake error log
+        uses: actions/upload-artifact@v2
+        with:
+          name: Other_CmakeError.log
+          path: |
+            **/python/build/**/Release/CMakeFiles/CMakeError.log
+            **/python/build/**/Release/CMakeFiles/CMakeOutput.log
+
       - name: Change anaconda label on release
         if: github.event.release
         run: |

cpp_cuda_c/CMakeLists.txt

@@ -1,4 +1,4 @@
-cmake_minimum_required(VERSION 3.15 FATAL_ERROR)
+cmake_minimum_required(VERSION 3.17 FATAL_ERROR)
 #cmake_policy(SET CMP0094 NEW)
 
 project(SplinePSF LANGUAGES CXX C)
@@ -8,9 +8,9 @@ check_language(CUDA)
 
 find_package(Python COMPONENTS Interpreter Development)
 find_package(pybind11 REQUIRED)
-find_package(CUDA)
+find_package(CUDAToolkit)
 
-if(CMAKE_CUDA_COMPILER AND NOT CUDA_DISABLED)  # the latter is a user definable variable
+if(CUDAToolkit_FOUND AND NOT CUDA_DISABLED)  # the latter is a user definable variable
    enable_language(CUDA)
    add_compile_definitions(CUDA_ENABLED)
 
@@ -28,14 +28,14 @@ target_include_directories(spline_psf_cpu_impl PUBLIC include)
 target_compile_features(spline_psf_cpu_impl PUBLIC c_std_99)
 set_property(TARGET spline_psf_cpu_impl PROPERTY POSITION_INDEPENDENT_CODE ON)
 
-if(CMAKE_CUDA_COMPILER AND NOT CUDA_DISABLED)
+if(CUDAToolkit_FOUND AND NOT CUDA_DISABLED)
    add_library(spline_psf_cu_impl STATIC src/spline_psf_gpu.cu)
    target_include_directories(spline_psf_cu_impl PUBLIC include)
    target_compile_features(spline_psf_cu_impl PUBLIC cxx_std_11)
    set_target_properties(spline_psf_cu_impl PROPERTIES CUDA_SEPARABLE_COMPILATION ON POSITION_INDEPENDENT_CODE ON)
 
-   link_directories(/usr/local/cuda/lib64)
-   include_directories("${CUDA_INCLUDE_DIRS}")
+   # link_directories(/usr/local/cuda/lib64)
+   # include_directories("${CMAKE_CUDA_TOOLKIT_INCLUDE_DIRECTORIES}")
 
    pybind11_add_module(spline src/pybind_spline.cpp)
    target_link_libraries(spline PRIVATE spline_psf_cu_impl spline_psf_cpu_impl)

cpp_cuda_c/src/spline_psf_gpu.cu

@@ -522,17 +522,17 @@ auto kernel_roi(spline *sp, float *rois, const int npx, const int npy, const flo
     float phot = phot_[r];
 
     /* Compute delta. Will be the same for all following px */
-    x0 = (int)floor(xc);
+    x0 = (int)floorf(xc);
     x_delta = xc - x0;
 
-    y0 = (int)floor(yc);
+    y0 = (int)floorf(yc);
     y_delta = yc - y0;
 
-    z0 = (int)floor(zc);
+    z0 = (int)floorf(zc);
     z_delta = zc - z0;
 
     int n_threads = min(1024, npx * npy);  // max number of threads per block
-    int n_blocks = ceil(static_cast<float>(npx * npy) / static_cast<float>(n_threads));
+    int n_blocks = ceilf(static_cast<float>(npx * npy) / static_cast<float>(n_threads));
 
     fAt3Dj<<<n_blocks, n_threads>>>(sp, rois, r, npx, npy, x0, y0, z0, phot, x_delta, y_delta, z_delta);
 
@@ -555,17 +555,17 @@ auto kernel_derivative_roi(spline *sp, float *rois, float *drv_rois, const int n
     float bg = bg_[r];
 
     /* Compute delta. Will be the same for all following px */
-    x0 = (int)floor(xc);
+    x0 = (int)floorf(xc);
     x_delta = xc - x0;
 
-    y0 = (int)floor(yc);
+    y0 = (int)floorf(yc);
     y_delta = yc - y0;
 
-    z0 = (int)floor(zc);
+    z0 = (int)floorf(zc);
     z_delta = zc - z0;
 
     int n_threads = min(1024, npx * npy);  // max number of threads per block
-    int n_blocks = ceil(static_cast<float>(npx * npy) / static_cast<float>(n_threads));
+    int n_blocks = ceilf(static_cast<float>(npx * npy) / static_cast<float>(n_threads));
 
     kernel_derivative<<<n_blocks, n_threads>>>(sp, rois, drv_rois, r, npx, npy, x0, y0, z0, phot, bg, x_delta, y_delta, z_delta, add_bg);
 

dist_tools/conda/conda_build_config.yaml

@@ -2,11 +2,11 @@
 c_compiler:
     - clang                            # [osx]
     - gcc                              # [linux]
-    - vs2017                           # [win]
+    - vs2019                           # [win]
 cxx_compiler:
     - clangxx                          # [osx]
     - gxx                              # [linux]
-    - vs2017                           # [win]
+    - vs2019                           # [win]
 target_platform:
     - osx-64                           # [osx]
     - linux-64                         # [linux]

dist_tools/conda/spline/meta.yaml

@@ -13,7 +13,7 @@ requirements:
     - {{ compiler('c') }}
     - {{ compiler('cxx') }}
     - python
-    - cmake =3.15
+    - cmake =3.17
     - ninja
     - setuptools
 

python/environment.yml

@@ -4,7 +4,7 @@ channels:
 dependencies:
   - bump2version
   - python=3.7
-  - cmake=3.15
+  - cmake=3.17
   - ninja
   - setuptools
   - numpy