Merge remote-tracking branch 'origin/master' into better-nbtable-layout

GridTools · Jun 18, 2024 · 2764b2f · 2764b2f
2 parents 458055f + 3b6c32a
commit 2764b2f
Show file tree

Hide file tree

Showing 15 changed files with 87,403 additions and 86,085 deletions.
diff --git a/.github/workflows/cmake-configure.yml b/.github/workflows/cmake-configure.yml
@@ -12,7 +12,7 @@ jobs:
     container: ghcr.io/gridtools/gridtools-base:clang-${{ matrix.version }}
     strategy:
         matrix:
-            version: [9, 10]
+            version: [12, 16, 17, 18]
             cuda-mode: [AUTO, Clang-CUDA, NVCC-CUDA]
     steps:
     - uses: actions/checkout@v2
@@ -38,10 +38,11 @@ jobs:
 
   clang-cuda:
     runs-on: ubuntu-latest
-    container: ghcr.io/gridtools/gridtools-base:clang-${{ matrix.version }}-cuda-11
+    container: ghcr.io/gridtools/gridtools-base:clang-${{ matrix.clang_version }}-cuda-${{ matrix.cuda_version }}
     strategy:
         matrix:
-            version: [11, 12]
+            clang_version: [17]
+            cuda_version: [12.5]
             cuda-mode: [AUTO, Clang-CUDA, NVCC-CUDA]
     steps:
     - uses: actions/checkout@v2
@@ -62,7 +63,7 @@ jobs:
     container: ghcr.io/gridtools/gridtools-base:gcc-${{ matrix.version }}
     strategy:
         matrix:
-            version: [8, 9, 10, 11]
+            version: [8, 11, 12, 13]
     steps:
     - uses: actions/checkout@v2
     - name: CMake version
@@ -78,7 +79,18 @@ jobs:
 
   gcc-nvcc:
     runs-on: ubuntu-latest
-    container: ghcr.io/gridtools/gridtools-base:gcc-10-cuda-11
+    container: ghcr.io/gridtools/gridtools-base:gcc-${{ matrix.gcc_version }}-cuda-${{ matrix.cuda_version }}
+    strategy:
+        matrix:
+            gcc_version: [10, 11, 12]
+            cuda_version: [11.8, "12.0", 12.2]
+            exclude:
+              - gcc_version: 12
+                cuda_version: 11.8
+              - gcc_version: 11
+                cuda_version: 11.8
+              - gcc_version: 11
+                cuda_version: 12.2
     steps:
     - uses: actions/checkout@v2
     - name: CMake version

diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
@@ -12,7 +12,7 @@ jobs:
     container: ghcr.io/gridtools/gridtools-base:${{ matrix.compiler }}
     strategy:
         matrix:
-            compiler: [gcc-8, gcc-9, gcc-10, gcc-11, gcc-12, gcc-13, clang-11, clang-12, clang-13, clang-14, clang-15, clang-16, clang-14-cuda-11, gcc-10-cuda-11.8, gcc-11-cuda-12.0, gcc-12-cuda-12.1, gcc-12-cuda-12.2, gcc-12-cuda-12.3, base-hip, gcc-10-hpx, nvhpc-23.3, nvhpc-23.9]
+            compiler: [gcc-8, gcc-9, gcc-10, gcc-11, gcc-12, gcc-13, clang-11, clang-12, clang-13, clang-14, clang-15, clang-16, clang-14-cuda-11, gcc-10-cuda-11.8, gcc-11-cuda-12.0, gcc-12-cuda-12.3, gcc-12-cuda-12.4, base-hip, gcc-10-hpx, nvhpc-23.3, nvhpc-23.9]
             build_type: [debug, release]
             exclude:
               - compiler: gcc-8

diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -55,7 +55,7 @@ else()
     endif()
 endif()
 
-set(REQUIRED_BOOST_VERSION 1.65.1)
+set(REQUIRED_BOOST_VERSION 1.73)
 find_package(Boost ${REQUIRED_BOOST_VERSION} REQUIRED)
 
 # setup main gridtools target

diff --git a/README.md b/README.md
@@ -36,7 +36,7 @@ CUDAHOSTCXX=`which g++` # full path to the C++ compiler to be used as CUDA host
 
 ##### Requirements
 - C++17 compiler (see also list of tested compilers)
-- Boost headers (1.77 or later)
+- Boost headers (1.73 or later)
 - CMake (3.18.1 or later)
 - CUDA Toolkit (11.0 or later, optional)
 - MPI (optional, CUDA-aware MPI for the GPU communication module `gcl_gpu`)

diff --git a/include/gridtools/common/defs.hpp b/include/gridtools/common/defs.hpp
@@ -52,8 +52,8 @@ namespace gridtools {
 #define GT_NVCC_DIAG_POP_SUPPRESS(x)
 #endif
 
-#if defined(__NVCC__) && (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ >= 1 && __CUDACC_VER_MINOR__ <= 3)
-// enables workaround for CTAD/constexpr issues in CUDA 12.1, 12.2, 12.3 (https://github.com/GridTools/gridtools/issues/1766)
+#if defined(__NVCC__) && (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ >= 1 && __CUDACC_VER_MINOR__ <= 4)
+// enables workaround for CTAD/constexpr issues in CUDA 12.1, 12.2, 12.3, 12.4 (https://github.com/GridTools/gridtools/issues/1766)
 #define GT_NVCC_WORKAROUND_1766 1
 #else
 #define GT_NVCC_WORKAROUND_1766 0

diff --git a/include/gridtools/storage/adapter/nanobind_adapter.hpp b/include/gridtools/storage/adapter/nanobind_adapter.hpp
@@ -25,24 +25,24 @@
 namespace gridtools {
     namespace nanobind_sid_adapter_impl_ {
 
-        // Use nanobind::any for dynamic stride, use an integral value for static stride.
-        template <std::size_t... Values>
-        using stride_spec = std::index_sequence<Values...>;
+        // Use `-1` for dynamic stride, use an integral value for static stride.
+        template <nanobind::ssize_t... Values>
+        using stride_spec = std::integer_sequence<nanobind::ssize_t, Values...>;
 
         template <class IndexSequence>
         struct dynamic_strides_helper;
 
         template <std::size_t... Indices>
         struct dynamic_strides_helper<std::index_sequence<Indices...>> {
-            using type = stride_spec<(void(Indices), nanobind::any)...>;
+            using type = stride_spec<(void(Indices), -1)...>;
         };
 
         template <std::size_t N>
         using fully_dynamic_strides = typename dynamic_strides_helper<std::make_index_sequence<N>>::type;
 
-        template <std::size_t SpecValue>
+        template <nanobind::ssize_t SpecValue>
         auto select_static_stride_value(std::size_t dyn_value) {
-            if constexpr (SpecValue == nanobind::any) {
+            if constexpr (SpecValue == -1) {
                 return dyn_value;
             } else {
                 if (SpecValue != dyn_value) {
@@ -52,20 +52,20 @@ namespace gridtools {
             }
         }
 
-        template <std::size_t... SpecValues, std::size_t... IndexValues>
+        template <nanobind::ssize_t... SpecValues, std::size_t... IndexValues>
         auto select_static_strides_helper(
             stride_spec<SpecValues...>, const std::size_t *dyn_values, std::index_sequence<IndexValues...>) {
 
             return gridtools::tuple{select_static_stride_value<SpecValues>(dyn_values[IndexValues])...};
         }
 
-        template <std::size_t... SpecValues>
+        template <nanobind::ssize_t... SpecValues>
         auto select_static_strides(stride_spec<SpecValues...> spec, const std::size_t *dyn_values) {
             return select_static_strides_helper(spec, dyn_values, std::make_index_sequence<sizeof...(SpecValues)>{});
         }
 
         template <class T,
-            std::size_t... Sizes,
+            nanobind::ssize_t... Sizes,
             class... Args,
             class Strides = fully_dynamic_strides<sizeof...(Sizes)>,
             class StridesKind = sid::unknown_kind>