update to C++20

.clang-tidy

@@ -25,6 +25,7 @@ Checks: >
        -misc-redundant-expression,
      modernize-*,
        -modernize-avoid-c-arrays,
+       -modernize-use-designated-initializers,
        -modernize-use-trailing-return-type,
      performance-*,
        -performance-avoid-endl,

.github/workflows/dependencies/dependencies_nvcc.sh

@@ -25,8 +25,8 @@ sudo apt-get install -y \
     wget
 
 VERSION_DOTTED=${1-12.0} && VERSION_DASHED=$(sed 's/\./-/' <<< $VERSION_DOTTED)
-curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.0-1_all.deb
-sudo dpkg -i cuda-keyring_1.0-1_all.deb
+curl -O https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2404/x86_64/cuda-keyring_1.1-1_all.deb
+sudo dpkg -i cuda-keyring_1.1-1_all.deb
 sudo apt-get update
 sudo apt-get install -y \
     cuda-command-line-tools-$VERSION_DASHED \
@@ -39,4 +39,7 @@ sudo apt-get install -y \
     libcufft-dev-$VERSION_DASHED            \
     libcurand-dev-$VERSION_DASHED           \
     libcusparse-dev-$VERSION_DASHED
+
+sudo apt-get install -y --no-install-recommends libnvjitlink-dev-$VERSION_DASHED || true
+
 sudo ln -s cuda-$VERSION_DOTTED /usr/local/cuda

.github/workflows/gpu_action.yml

@@ -8,7 +8,7 @@ concurrency:
 
 jobs:
   gpu-compilation:
-    runs-on: ubuntu-22.04
+    runs-on: ubuntu-24.04
     steps:
       - uses: actions/checkout@v5
         with:
@@ -29,14 +29,14 @@ jobs:
           sudo apt-get -qq -y install curl cmake jq clang g++>=9.3.0
 
       - name: Install CUDA
-        run: .github/workflows/dependencies/dependencies_nvcc.sh 11.8
+        run: .github/workflows/dependencies/dependencies_nvcc.sh 12.9
 
       - name: Install hypre
         run: |
-          wget -q https://github.com/hypre-space/hypre/archive/refs/tags/v2.28.0.tar.gz
-          tar xfz v2.28.0.tar.gz
-          cd hypre-2.28.0/src
-          CUDA_HOME=/usr/local/cuda HYPRE_CUDA_SM=60 ./configure --with-cxxstandard=17 --with-cuda --enable-unified-memory --without-MPI
+          wget -q https://github.com/hypre-space/hypre/archive/refs/tags/v2.33.0.tar.gz
+          tar xfz v2.33.0.tar.gz
+          cd hypre-2.33.0/src
+          CUDA_HOME=/usr/local/cuda HYPRE_CUDA_SM=60 ./configure --with-cxxstandard=20 --with-cuda --enable-unified-memory --without-MPI
           make -j 4
           make install
           cd ../../
@@ -56,5 +56,5 @@ jobs:
       - name: Compile problems for GPU
         run: |
           export PATH=$PATH:/usr/local/cuda/bin
-          export AMREX_HYPRE_HOME=${PWD}/hypre-2.28.0/src/hypre
+          export AMREX_HYPRE_HOME=${PWD}/hypre-2.33.0/src/hypre
           python3 external/cpp-linter-action/run_on_changed_files.py ${{ github.event.pull_request.base.sha }} ${{ github.event.pull_request.head.sha }} -header-filter=Castro -ignore-files="amrex|Microphysics" -gpu

Docs/source/getting_started.rst

@@ -14,7 +14,7 @@ Getting Started
 The compilation process is managed by AMReX and its build system.  The
 general requirements to build Castro are:
 
- * A C++17 (or later) compiler (for GCC, we need >= 9.0 for CUDA compilation)
+ * A C++20 (or later) compiler (for GCC, we need >= 13.1)
 
  * python (>= 3.10)
 
@@ -24,7 +24,7 @@ GCC is the main compiler suite used by the developers.
 
 For running in parallel, an MPI library is required.  For running on GPUs:
 
-* CUDA 11 or later is required for NVIDIA GPUs
+* CUDA 12 or later is required for NVIDIA GPUs
 
 * ROCM 6.3.1 or later is required for AMD GPUs (earlier versions have a register allocation bug)
 

Exec/Make.Castro

@@ -52,8 +52,8 @@ ifeq ("$(wildcard $(AMREX_HOME)/Tools/GNUMake/Make.defs)","")
   $(error AMReX has not been downloaded. Please run "git submodule update --init" from the top level of the code)
 endif
 
-# Require C++17
-CXXSTD := c++17
+# Require C++20
+CXXSTD := c++20
 
 # Use Lazy QueueReduction for the timing outputs
 LAZY := TRUE

Source/driver/main.cpp

@@ -2,6 +2,7 @@
 #include <new>
 #include <cstdio>
 #include <cstring>
+#include <format>
 #include <iostream>
 #include <iomanip>
 
@@ -255,14 +256,8 @@ main (int   argc,
         //
         // It's actually the high water mark of heap space required by FABs.
         //
-        char buf[256];
-
-        sprintf(buf,
-                "CPU(%d): Heap Space (bytes) used by Coalescing FAB Arena: %ld",
-                ParallelDescriptor::MyProc(),
-                arena->heap_space_used());
-
-        std::cout << buf << std::endl;
+        std::cout << std::format("CPU({}): Heap Space (bytes) used by Coalescing FAB Arena: {}\n",
+                                 ParallelDescriptor::MyProc(), arena->heap_space_used());
     }
 
     BL_PROFILE_VAR_STOP(pmain);

Source/gravity/Gravity.cpp

@@ -1771,7 +1771,7 @@ Gravity::init_multipole_grav() const
         maxWidth = amrex::max(maxWidth, probhi[2] - problo[2]);
     }
 
-    multipole::rmax = 0.5_rt * maxWidth * std::sqrt(static_cast<Real>(AMREX_SPACEDIM));
+    multipole::rmax = 0.5_rt * maxWidth * std::sqrt(static_cast<Real>(AMREX_SPACEDIM));  // NOLINT(modernize-use-std-numbers)
 }
 
 void

Source/radiation/Radiation.cpp

@@ -7,14 +7,13 @@
 
 #include <opacity.H>
 
+#include <format>
 #include <iostream>
 
 #ifdef _OPENMP
 #include <omp.h>
 #endif
 
-#include <sstream>
-
 using namespace amrex;
 
 Radiation::Solver_Type Radiation::SolverType = Radiation::InvalidSolver;
@@ -647,27 +646,23 @@ void Radiation::checkPoint(int level,
   //
   if (ParallelDescriptor::IOProcessor()) {
     int oldprec = os.precision(20);
-    sprintf(buf, "delta_e_rat_level[%d]= ", level);
-    std::string DeltaString = buf;
+    auto DeltaString = std::format("delta_e_rat_level[{}]= ", level);
     os << DeltaString << delta_e_rat_level[level] << '\n';
-    sprintf(buf, "delta_T_rat_level[%d]= ", level);
-    DeltaString = buf;
+    DeltaString = std::format("delta_T_rat_level[{}]= ", level);
     os << DeltaString << delta_T_rat_level[level] << '\n';
     os.precision(oldprec);
   }
 
   // Path name construction stolen from AmrLevel::checkPoint
 
-  sprintf(buf, "Level_%d", level);
-  std::string Level = buf;
+  std::string Level = std::format("Level_{}", level);
 
   //
   // Write name of conservation flux register to header.
   //
-  sprintf(buf, "/RadFlux");
 
   std::string PathNameInHeader = Level;
-  PathNameInHeader += buf;
+  PathNameInHeader += "/RadFlux";
   if (ParallelDescriptor::IOProcessor()) {
     os << PathNameInHeader;
   }
@@ -691,7 +686,7 @@ void Radiation::checkPoint(int level,
       FullPathName += '/';
     }
     FullPathName += Level;
-    FullPathName += buf;
+    FullPathName += "/RadFlux";
     //
     // Output conservation flux register.
     //

Util/model_parser/model_parser.H

@@ -5,6 +5,8 @@
 #include <fstream>
 #include <vector>
 #include <algorithm>
+#include <ranges>
+
 #include <network.H>
 #include <model_parser_data.H>
 #include <AMReX_Print.H>
@@ -60,20 +62,16 @@ namespace model_string
 {
     inline std::string& ltrim(std::string& s)
     {
-        auto it = std::find_if(s.begin(), s.end(),
-                               [](int c) {
-                                   return !std::isspace(c);
-                               });
+        auto it = std::ranges::find_if(s,
+                                       [] (int c) {return !std::isspace(c);});
         s.erase(s.begin(), it);
         return s;
     }
 
     inline std::string& rtrim(std::string& s)
     {
-        auto it = std::find_if(s.rbegin(), s.rend(),
-                               [](int c) {
-                                   return !std::isspace(c);
-                               });
+        auto it = std::ranges::find_if(std::ranges::reverse_view(s),
+                                       [] (int c) {return !std::isspace(c);});
         s.erase(it.base(), s.end());
         return s;
     }