[WIP] C++20 is now the minimum required version

.github/workflows/gcc.yml

@@ -20,15 +20,15 @@ jobs:
   # Build and install libamrex as AMReX CMake project
   # Note: this is an intentional "minimal" build that does not enable (many) options
   library:
-    name: GNU@14 Release [lib]
+    name: GNU@11 Release [lib]
     runs-on: ubuntu-24.04
     needs: check_changes
     if: needs.check_changes.outputs.has_non_docs_changes == 'true'
     steps:
     - uses: actions/checkout@v6
     - name: Dependencies
       run: |
-        .github/workflows/dependencies/dependencies_gcc.sh 14
+        .github/workflows/dependencies/dependencies_gcc.sh 11
         .github/workflows/dependencies/dependencies_clang-tidy-apt-llvm.sh 21
         .github/workflows/dependencies/dependencies_ccache.sh
     - name: Set Up Cache
@@ -57,9 +57,9 @@ jobs:
             -DAMReX_PLOTFILE_TOOLS=ON             \
             -DCMAKE_VERBOSE_MAKEFILE=ON           \
             -DCMAKE_INSTALL_PREFIX=/tmp/my-amrex  \
-            -DCMAKE_C_COMPILER=$(which gcc-14)     \
-            -DCMAKE_CXX_COMPILER=$(which g++-14)   \
-            -DCMAKE_Fortran_COMPILER=$(which gfortran-14) \
+            -DCMAKE_C_COMPILER=$(which gcc-11)     \
+            -DCMAKE_CXX_COMPILER=$(which g++-11)   \
+            -DCMAKE_Fortran_COMPILER=$(which gfortran-11) \
             -DCMAKE_CXX_COMPILER_LAUNCHER=ccache
         make -j 4
         make install

.github/workflows/hypre.yml

@@ -25,8 +25,8 @@ jobs:
     - uses: actions/checkout@v6
     - name: Dependencies
       run: |
-        .github/workflows/dependencies/dependencies_nvcc.sh 12.0
-        sudo apt-get install -y libcublas-dev-12.0
+        .github/workflows/dependencies/dependencies_nvcc.sh 12.2
+        sudo apt-get install -y libcublas-dev-12.2
         .github/workflows/dependencies/dependencies_ccache.sh
     - name: Set Up Cache
       uses: actions/cache@v5

.github/workflows/sundials.yml

@@ -89,7 +89,7 @@ jobs:
     - uses: actions/checkout@v6
     - name: Dependencies
       run: |
-        .github/workflows/dependencies/dependencies_nvcc.sh
+        .github/workflows/dependencies/dependencies_nvcc.sh 12.2
         .github/workflows/dependencies/dependencies_ccache.sh
     - name: Set Up Cache
       uses: actions/cache@v5
@@ -103,7 +103,7 @@ jobs:
         ccache -z
 
         # sundials requirement
-        sudo apt-get install -y libcusolver-dev-12-0 libcusparse-dev-12-0 libcublas-dev-12-0
+        sudo apt-get install -y libcusolver-dev-12-2 libcusparse-dev-12-2 libcublas-dev-12-2
         export PATH=/usr/local/nvidia/bin:/usr/local/cuda/bin:${PATH}
         export LD_LIBRARY_PATH=/usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64:${LD_LIBRARY_PATH}
         which nvcc || echo "nvcc not in PATH!"

CMakeLists.txt

@@ -128,8 +128,10 @@ endif ()
 #
 # Check compiler version
 #
-set_mininum_compiler_version(CXX GNU 5.1)
-set_mininum_compiler_version(CXX MSVC 19.23)
+set_mininum_compiler_version(CXX GNU 11.0)
+set_mininum_compiler_version(CXX Clang 14.0)
+set_mininum_compiler_version(CXX AppleClang 14.0)
+set_mininum_compiler_version(CXX MSVC 19.34)
 
 #
 # Set CMAKE_<LANG>_FLAGS_<CONFIG> if not already defined

Docs/sphinx_documentation/source/BuildingAMReX.rst

@@ -36,8 +36,8 @@ list of important variables.
    | COMP            | gnu, cray, ibm, intel, intel-llvm,  |                    |
    |                 | intel-classic, llvm, or pgi         | none               |
    +-----------------+-------------------------------------+--------------------+
-   | CXXSTD          | C++ standard (``c++17``, ``c++20``) | compiler default,  |
-   |                 |                                     | at least ``c++17`` |
+   | CXXSTD          | C++ standard (e.g., ``c++20``)      | compiler default,  |
+   |                 |                                     | at least ``c++20`` |
    +-----------------+-------------------------------------+--------------------+
    | DEBUG           | TRUE or FALSE                       | FALSE              |
    +-----------------+-------------------------------------+--------------------+
@@ -123,7 +123,6 @@ MPI with support for concurrent MPI calls from multiple threads.
 Variables ``USE_CUDA``, ``USE_HIP`` and ``USE_SYCL`` are used for
 targeting Nvidia, AMD and Intel GPUs, respectively.  At most one of
 the three can be TRUE.
-For HIP and SYCL builds, we currently test only against C++17 builds.
 
 The variable ``USE_RPATH`` controls the link mechanism to dependent libraries.
 If enabled, the library path at link time will be saved as a
@@ -435,15 +434,15 @@ The list of available options is reported in the :ref:`table <tab:cmakevar>` bel
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
    | CMAKE_CXX_FLAGS              |  User-defined C++ flags                         |                         | user-defined          |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
-   | CMAKE_CXX_STANDARD           |  C++ standard                                   | compiler/17             | 17, 20                |
+   | CMAKE_CXX_STANDARD           |  C++ standard                                   | compiler/20             | 20, 23                |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
    | AMReX_SPACEDIM               |  Dimension of AMReX build                       | 3 ``;``-separated list  | "1;2;3"               |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
    | USE_XSDK_DEFAULTS            |  Use xSDK defaults settings                     | NO                      | YES, NO               |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
    | AMReX_BUILD_SHARED_LIBS      |  Build as shared C++ library                    | NO (unless xSDK)        | YES, NO               |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
-   | AMReX_FASTMATH               |  Enable fast-math optimizations                 | NO (CUDA is ON)          | YES, NO              |
+   | AMReX_FASTMATH               |  Enable fast-math optimizations                 | NO (CUDA is ON)         |  YES, NO              |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
    | AMReX_FORTRAN                |  Enable Fortran language                        | NO                      | YES, NO               |
    +------------------------------+-------------------------------------------------+-------------------------+-----------------------+
@@ -781,7 +780,7 @@ The AMReX team does development on Linux machines, from laptops to supercomputer
 We do not officially support AMReX on Windows, and many of us do not have access to any Windows
 machines.  However, we believe there are no fundamental issues for it to work on Windows.
 
-(1) AMReX mostly uses standard C++17.
+(1) AMReX mostly uses standard C++20.
 We run continuous integration tests on Windows with MSVC and Clang compilers.
 
 (2) We use POSIX signal handling when floating-point exceptions, segmentation faults, etc. happen.

Docs/sphinx_documentation/source/BuildingAMReX_Chapter.rst

@@ -18,13 +18,22 @@ an application code then uses its own build system and links to AMReX as an exte
 
 Finally, AMReX can also be built with CMake, as detailed in the section on :ref:`sec:build:cmake`.
 
-AMReX requires a C++ compiler that supports the C++17 standard, a
+AMReX requires a C++ compiler that supports the C++20 standard, a
 Fortran compiler that supports the Fortran 2003 standard, and a C
 compiler that supports the C99 standard.  Prerequisites for building
 with GNU Make include Python (>= 2.7, including 3) and standard tools
 available in any Unix-like environment (e.g., Perl and sed).  For
 building with CMake, the minimal requirement is version 3.18.
 
+The minimum toolchain versions we target for C++20 builds are:
+
+* GCC 11 or newer (for both host builds and CUDA host compilers).
+* LLVM Clang 14 or newer, including AppleClang 14 on macOS.
+* Microsoft Visual Studio 2022 (MSVC 19.34 / 17.4) or newer.
+* NVIDIA CUDA Toolkit 12.2 or newer.
+* AMD ROCm/HIP 6.0 or newer.
+* Intel oneAPI DPC++ 2025.2 or newer.
+
 Please note that we fully support AMReX for Linux systems in general and on the
 DOE supercomputers (e.g., Perlmutter, Frontier) in particular.  Many of our users do build
 and use AMReX on Macs but we do not have the resources to fully support Mac users.

Docs/sphinx_documentation/source/GPU.rst

@@ -299,7 +299,7 @@ If autodetection fails, a list of "common" architectures is assumed.
 Building for multiple CUDA architectures will generally result in a larger library and longer build times.
 
 **Note that AMReX supports NVIDIA GPU architectures with compute capability 6.0 or higher and
-CUDA Toolkit version 11.0 or higher.**
+CUDA Toolkit version 12.2 or higher.**
 
 In order to import the CUDA-enabled AMReX library into your CMake project, you need to include
 the following code into the appropriate CMakeLists.txt file:

Docs/sphinx_documentation/source/GPU_Chapter.rst

@@ -8,8 +8,8 @@ NVIDIA, AMD, and Intel GPUs using their native vendor languages and therefore
 requires CUDA, HIP/ROCm, and SYCL for NVIDIA, AMD, and Intel GPUs, respectively.
 Users can also use OpenMP and/or OpenACC in their applications if desired.
 
-AMReX supports NVIDIA GPUs with compute capability >= 6 and CUDA >= 11, and
-AMD GPUs with ROCm >= 5. While SYCL compilers are still under development in
+AMReX supports NVIDIA GPUs with compute capability >= 6 and CUDA >= 12.2, and
+AMD GPUs with ROCm >= 6. While SYCL compilers are still under development in
 preparation for Aurora, AMReX officially supports only the latest publicly
 released version of the oneAPI compiler.
 

INSTALL

@@ -10,7 +10,7 @@ There are three ways to use AMReX.
     Fortran modules via `./configure` followed by `make` and `make
     install`.  Type `./configure -h` to show help message.  An
     application code uses its build system to compile and link to the
-    AMReX library.  Because AMReX uses C++17 and Fortran, the linker
+    AMReX library.  Because AMReX uses C++20 and Fortran, the linker
     needs to link the libraries.  See
     `Tutorials/Basic/Build_with_libamrex` for an example of this
     approach.  Note that this approach relies the make system in
@@ -20,4 +20,3 @@ There are three ways to use AMReX.
 (3) CMake. See the "Building with CMake" page in the AMReX documentation
     page at the link below for more details.
     https://amrex-codes.github.io/amrex/docs_html/BuildingAMReX.html#building-with-cmake
-

Tests/CTOParFor/main.cpp

@@ -6,7 +6,6 @@ using namespace amrex;
 int main (int argc, char* argv[])
 {
     amrex::Initialize(argc,argv);
-#if (__cplusplus >= 201703L)
     {
         enum A_options: int {
             A0 = 0, A1
@@ -57,8 +56,5 @@ int main (int argc, char* argv[])
             }
         }
     }
-#else
-    amrex::Print() << "This test requires C++17." << '\n';
-#endif
     amrex::Finalize();
 }

Tools/CMake/AMReXParallelBackends.cmake

@@ -91,7 +91,7 @@ if (  AMReX_GPU_BACKEND STREQUAL "CUDA"
    endforeach()
 
    # Check cuda compiler and host compiler
-   set_mininum_compiler_version(CUDA NVIDIA 9.0)
+   set_mininum_compiler_version(CUDA NVIDIA 12.2)
    check_cuda_host_compiler()
 
    # Required CUDA flags
@@ -366,8 +366,8 @@ if (AMReX_HIP)
        # 
        target_compile_options(amrex_${D}d PUBLIC $<$<COMPILE_LANGUAGE:CXX>:-munsafe-fp-atomics>)
 
-       # Ensure ROCm builds enable at least C++17 without overriding higher standards
-       target_compile_features(amrex_${D}d PUBLIC cxx_std_17)
+       # Ensure ROCm builds enable at least C++20 without overriding higher standards
+       target_compile_features(amrex_${D}d PUBLIC cxx_std_20)
    endforeach()
 
    # Equivalently, relocatable-device-code (RDC) flags are needed for `extern`

Tools/CMake/AMReXSYCL.cmake

@@ -13,7 +13,7 @@ set(_cxx_sycl "$<AND:$<COMPILE_LANGUAGE:CXX>,${_cxx_sycl}>")
 add_library(SYCL INTERFACE)
 add_library(AMReX::SYCL ALIAS SYCL)
 
-target_compile_features(SYCL INTERFACE cxx_std_17)
+target_compile_features(SYCL INTERFACE cxx_std_20)
 
 
 #

Tools/CMake/AMReX_Config.cmake

@@ -39,24 +39,18 @@ function (configure_amrex AMREX_TARGET)
    #
    # Setup compilers
    #
-   # Set C++ standard and disable compiler-specific extensions, like "-std=gnu++17" for GNU
+   # Set C++ standard and disable compiler-specific extensions, like "-std=gnu++20" for GNU
    # This will also enforce the same standard with the CUDA compiler
    # Moreover, it will also enforce such standard on all the consuming targets
    #
    set_target_properties(${AMREX_TARGET} PROPERTIES CXX_EXTENSIONS OFF)
-   # minimum: C++17
-   target_compile_features(${AMREX_TARGET} PUBLIC cxx_std_17)
-
-   # vir::cvt
-   # https://github.com/mattkretz/vir-simd/issues/45
-   if (AMReX_SIMD)
-       target_compile_features(${AMREX_TARGET} PUBLIC cxx_std_20)
-   endif()
+   # minimum: C++20
+   target_compile_features(${AMREX_TARGET} PUBLIC cxx_std_20)
 
    if (AMReX_CUDA)
       set_target_properties(${AMREX_TARGET} PROPERTIES CUDA_EXTENSIONS OFF)
-      # minimum: C++17
-      target_compile_features(${AMREX_TARGET} PUBLIC cuda_std_17)
+      # minimum: C++20
+      target_compile_features(${AMREX_TARGET} PUBLIC cuda_std_20)
    endif()
 
    #

Tools/CMake/AMReX_SetupCUDA.cmake

@@ -17,7 +17,7 @@ endif ()
 # Check CUDA compiler and host compiler
 #
 include(AMReXUtils)
-set_mininum_compiler_version(CUDA NVIDIA 10.0)
+set_mininum_compiler_version(CUDA NVIDIA 12.2)
 check_cuda_host_compiler()
 
 #

Tools/GNUMake/comps/armclang.mak

@@ -57,14 +57,14 @@ ifeq ($(WARN_ERROR),TRUE)
 endif
 
 # disable some warnings
-CXXFLAGS += -Wno-c++17-extensions
+# CXXFLAGS += -Wno-c++20-extensions # Do we need it?
 
 ########################################################################
 
 ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
 else
-  CXXSTD := c++17
+  CXXSTD := c++20
 endif
 
 CXXFLAGS += -std=$(CXXSTD)

Tools/GNUMake/comps/cray.mak

@@ -73,7 +73,7 @@ endif
 ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
 else
-  CXXSTD := c++17
+  CXXSTD := c++20
 endif
 
 ifeq ($(CRAY_IS_CLANG_BASED),TRUE)

Tools/GNUMake/comps/dpcpp.mak

@@ -79,7 +79,7 @@ endif
 ifdef CXXSTD
   CXXFLAGS += -std=$(strip $(CXXSTD))
 else
-  CXXFLAGS += -std=c++17
+  CXXFLAGS += -std=c++20
 endif
 
 CXXFLAGS += -fsycl

Tools/GNUMake/comps/gnu.mak

@@ -47,8 +47,8 @@ gcc_major_ge_12 = $(shell expr $(gcc_major_version) \>= 12)
 INLINE_LIMIT ?= 43210
 
 ifneq ($(NO_CONFIG_CHECKING),TRUE)
-ifneq ($(gcc_major_ge_8),1)
-  $(error GCC < 8 not supported)
+ifneq ($(gcc_major_ge_11),1)
+  $(error GCC < 11 not supported)
 endif
 endif
 
@@ -148,7 +148,7 @@ ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
   CXXFLAGS += -std=$(CXXSTD)
 else
-  CXXFLAGS += -std=c++17
+  CXXFLAGS += -std=c++20
 endif
 
 CFLAGS   += -std=c11

Tools/GNUMake/comps/hip.mak

@@ -19,7 +19,7 @@ F90 = gfortran
 ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
 else
-  CXXSTD := c++17
+  CXXSTD := c++20
 endif
 
 # Generic flags, always used

Tools/GNUMake/comps/intel-classic.mak

@@ -45,7 +45,7 @@ ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
   CXXFLAGS += -std=$(CXXSTD)
 else
-  CXXFLAGS += -std=c++17
+  CXXFLAGS += -std=c++20
 endif
 
 CFLAGS   += -std=c11

Tools/GNUMake/comps/intel-llvm.mak

@@ -71,7 +71,7 @@ ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
   CXXFLAGS += -std=$(CXXSTD)
 else
-  CXXFLAGS += -std=c++17
+  CXXFLAGS += -std=c++20
 endif
 
 CFLAGS   += -std=c11

Tools/GNUMake/comps/llvm-flang.mak

@@ -43,7 +43,7 @@ endif
 ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
 else
-  CXXSTD := c++17
+  CXXSTD := c++20
 endif
 
 CXXFLAGS += -std=$(CXXSTD)

Tools/GNUMake/comps/llvm.mak

@@ -65,14 +65,14 @@ ifeq ($(USE_COMPILE_PIC),TRUE)
 endif
 
 # disable some warnings
-CXXFLAGS += -Wno-c++17-extensions
+# CXXFLAGS += -Wno-c++20-extensions # Do we need it?
 
 ########################################################################
 
 ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
 else
-  CXXSTD := c++17
+  CXXSTD := c++20
 endif
 
 CXXFLAGS += -std=$(CXXSTD)

Tools/GNUMake/comps/nag.mak

@@ -54,7 +54,7 @@ ifdef CXXSTD
   CXXSTD := $(strip $(CXXSTD))
   CXXFLAGS += -std=$(CXXSTD)
 else
-  CXXFLAGS += -std=c++17
+  CXXFLAGS += -std=c++20
 endif
 
 CFLAGS   += -std=c11