Update CUDA to version 12.9.0 and cuDNN to version 9.9.0

cuda.spec

@@ -1,8 +1,8 @@
-### RPM external cuda 12.8.1
+### RPM external cuda 12.9.0
 ## INITENV +PATH LD_LIBRARY_PATH %i/lib64
 
 %define runpath_opts -m compute-sanitizer -m drivers -m nvvm
-%define driversversion 570.124.06
+%define driversversion 575.51.03
 
 %ifarch x86_64
 Source0: https://developer.download.nvidia.com/compute/cuda/%{realversion}/local_installers/%{n}_%{realversion}_%{driversversion}_linux.run
@@ -28,18 +28,13 @@ cd %_builddir/
 # create target directory structure
 mkdir -p %{i}/include
 mkdir -p %{i}/lib64
-mkdir -p %{i}/lib64/stubs
 
 # package only the runtime static libraries
 mv %_builddir/build/lib64/libcudadevrt.a %{i}/lib64/
 mv %_builddir/build/lib64/libcudart_static.a %{i}/lib64/
 rm -f %_builddir/build/lib64/lib*.a
 
-# package only the CUDA driver and NVML library stubs
-mv %_builddir/build/lib64/stubs/libcuda.so      %{i}/lib64/stubs/libcuda.so
-ln -sf libcuda.so                               %{i}/lib64/stubs/libcuda.so.1
-mv %_builddir/build/lib64/stubs/libnvidia-ml.so %{i}/lib64/stubs/libnvidia-ml.so
-ln -sf libnvidia-ml.so                          %{i}/lib64/stubs/libnvidia-ml.so.1
+# the stub libraries will be replaced by the redistributable driver libraries
 rm -rf %_builddir/build/lib64/stubs/
 
 # do not package the OpenCL libraries
@@ -85,23 +80,32 @@ ln -s ../compute-sanitizer/compute-sanitizer %{i}/bin/compute-sanitizer
 # package the NVVM compiler (cicc), library (libnvvm.so), device library (libdevice.10.bc) and samples
 mv %_builddir/build/nvvm %{i}/
 
-# extract and repackage the NVIDIA libraries needed by the CUDA runtime
+# extract and repackage the redistributable NVIDIA driver libraries needed by the CUDA runtime
 /bin/sh %_builddir/pkg/builds/NVIDIA-Linux-%{_arch}-%{driversversion}.run --silent --extract-only --tmpdir %_builddir/tmp --target %_builddir/build/drivers
 
 mkdir -p %{i}/drivers
-mv %_builddir/build/drivers/libcuda.so.%{driversversion}                    %{i}/drivers/
-ln -sf libcuda.so.%{driversversion}                                         %{i}/drivers/libcuda.so.1
-ln -sf libcuda.so.1                                                         %{i}/drivers/libcuda.so
-mv %_builddir/build/drivers/libcudadebugger.so.%{driversversion}            %{i}/drivers/
-ln -sf libcudadebugger.so.%{driversversion}                                 %{i}/drivers/libcudadebugger.so.1
-ln -sf libcudadebugger.so.1                                                 %{i}/drivers/libcudadebugger.so
-mv %_builddir/build/drivers/libnvidia-ptxjitcompiler.so.%{driversversion}   %{i}/drivers/
-ln -sf libnvidia-ptxjitcompiler.so.%{driversversion}                        %{i}/drivers/libnvidia-ptxjitcompiler.so.1
-ln -sf libnvidia-ptxjitcompiler.so.1                                        %{i}/drivers/libnvidia-ptxjitcompiler.so
-mv %_builddir/build/drivers/libnvidia-nvvm.so.%{driversversion}             %{i}/drivers/
-ln -sf libnvidia-nvvm.so.%{driversversion}                                  %{i}/drivers/libnvidia-nvvm.so.4
-ln -sf libnvidia-nvvm.so.4                                                  %{i}/drivers/libnvidia-nvvm.so
-mv %_builddir/build/drivers/nvidia-smi                                      %{i}/drivers/
+cp -p %_builddir/build/drivers/libcuda.so.%{driversversion}                     %{i}/drivers/
+ln -sf libcuda.so.%{driversversion}                                             %{i}/drivers/libcuda.so.1
+ln -sf libcuda.so.1                                                             %{i}/drivers/libcuda.so
+cp -p %_builddir/build/drivers/libcudadebugger.so.%{driversversion}             %{i}/drivers/
+ln -sf libcudadebugger.so.%{driversversion}                                     %{i}/drivers/libcudadebugger.so.1
+ln -sf libcudadebugger.so.1                                                     %{i}/drivers/libcudadebugger.so
+cp -p %_builddir/build/drivers/libnvidia-ptxjitcompiler.so.%{driversversion}    %{i}/drivers/
+ln -sf libnvidia-ptxjitcompiler.so.%{driversversion}                            %{i}/drivers/libnvidia-ptxjitcompiler.so.1
+ln -sf libnvidia-ptxjitcompiler.so.1                                            %{i}/drivers/libnvidia-ptxjitcompiler.so
+cp -p %_builddir/build/drivers/libnvidia-nvvm.so.%{driversversion}              %{i}/drivers/
+ln -sf libnvidia-nvvm.so.%{driversversion}                                      %{i}/drivers/libnvidia-nvvm.so.4
+ln -sf libnvidia-nvvm.so.4                                                      %{i}/drivers/libnvidia-nvvm.so
+cp -p %_builddir/build/drivers/nvidia-smi                                       %{i}/drivers/
+
+# reuse the redistributable CUDA driver library and NVML library in place of the CUDA stub libraries
+mkdir -p %{i}/lib64/stubs
+cp -p %_builddir/build/drivers/libcuda.so.%{driversversion}                     %{i}/lib64/stubs/
+ln -sf libcuda.so.%{driversversion}                                             %{i}/lib64/stubs/libcuda.so.1
+ln -sf libcuda.so.1                                                             %{i}/lib64/stubs/libcuda.so
+cp -p %_builddir/build/drivers/libnvidia-ml.so.%{driversversion}                %{i}/lib64/stubs/
+ln -sf libnvidia-ml.so.%{driversversion}                                        %{i}/lib64/stubs/libnvidia-ml.so.1
+ln -sf libnvidia-ml.so.1                                                        %{i}/lib64/stubs/libnvidia-ml.so
 
 %post
 # let nvcc find its components when invoked from the command line

cudnn.spec

@@ -1,4 +1,4 @@
-### RPM external cudnn 9.6.0.74
+### RPM external cudnn 9.9.0.52
 ## INITENV +PATH LD_LIBRARY_PATH %i/lib64
 
 %define cudaver 12

pytorch-cluster.spec

@@ -10,7 +10,7 @@
 Source: git+https://github.com/%{github_user}/pytorch_cluster.git?obj=%{branch}/%{tag}&export=%{n}-%{realversion}&submodules=1&output=/%{n}-%{realversion}.tgz
 
 BuildRequires: cmake
-Requires: pytorch
+Requires: pytorch %{!?without_cuda:cuda}
 %define build_flags -Wall -Wextra -pedantic %{?arch_build_flags}
 %define cuda_arch_float $(echo %{cuda_arch} | tr ' ' '\\n' | sed -E 's|([0-9])$|.\\1|' | tr '\\n' ' ')
 
@@ -20,15 +20,15 @@ Requires: pytorch
 grep -q 'CMAKE_CXX_STANDARD  *14' CMakeLists.txt
 sed -i -e 's|CMAKE_CXX_STANDARD  *14|CMAKE_CXX_STANDARD %{cms_cxx_standard}|' CMakeLists.txt
 
+%build
+
 USE_CUDA=OFF
 %if 0%{!?without_cuda:1}
 if [ "%{cuda_gcc_support}" = "true" ] ; then
 USE_CUDA=ON
 fi
 %endif
 
-%build
-
 rm -rf ../build && mkdir ../build && cd ../build
 
 cmake ../%{n}-%{realversion} \
@@ -38,17 +38,14 @@ cmake ../%{n}-%{realversion} \
     -DCMAKE_PREFIX_PATH="%{cmake_prefix_path}" \
     -DCMAKE_CXX_STANDARD=%{cms_cxx_standard} \
     -DCMAKE_CXX_FLAGS="%{build_flags}" \
-    -DBUILD_TEST=OFF \
     -DWITH_PYTHON=OFF \
-    -DWITH_CUDA=OFF \
-    -DBUILD_TEST=OFF \
 %if 0%{!?without_cuda:1}
-    -DUSE_CUDA=${USE_CUDA} \
+    -DWITH_CUDA=${USE_CUDA} \
     -DTORCH_CUDA_ARCH_LIST="%{cuda_arch_float}" \
+    -Dnvtx3_dir=${CUDA_ROOT}/include \
 %endif
     -DBUILD_SHARED_LIBS=ON
 
-
 make %{makeprocesses} VERBOSE=1
 
 %install

pytorch-scatter.spec

@@ -2,14 +2,15 @@
 ## INCLUDE compilation_flags
 ## INCLUDE cpp-standard
 ## INCLUDE cuda-flags
+
 %define tag c095c62e4334fcd05e4ac3c4bb09d285960d6be6
 %define branch master
 %define github_user rusty1s
 
 Source: git+https://github.com/%{github_user}/pytorch_scatter.git?obj=%{branch}/%{tag}&export=%{n}-%{realversion}&submodules=1&output=/%{n}-%{realversion}.tgz
 
 BuildRequires: cmake
-Requires: pytorch
+Requires: pytorch %{!?without_cuda:cuda}
 %define build_flags -Wall -Wextra -pedantic %{?arch_build_flags}
 %define cuda_arch_float $(echo %{cuda_arch} | tr ' ' '\\n' | sed -E 's|([0-9])$|.\\1|' | tr '\\n' ' ')
 
@@ -19,18 +20,17 @@ Requires: pytorch
 grep -q 'CMAKE_CXX_STANDARD  *14' CMakeLists.txt
 sed -i -e 's|CMAKE_CXX_STANDARD  *14|CMAKE_CXX_STANDARD %{cms_cxx_standard}|' CMakeLists.txt
 
+%build
+
 USE_CUDA=OFF
-%if "%{cmsos}" != "slc7_aarch64"
+%if 0%{!?without_cuda:1}
 if [ "%{cuda_gcc_support}" = "true" ] ; then
-USE_CUDA=%{!?without_cuda:ON}
+USE_CUDA=ON
 fi
 %endif
 
-%build
-
 rm -rf ../build && mkdir ../build && cd ../build
 
-
 cmake ../%{n}-%{realversion} \
     -DCMAKE_BUILD_TYPE=Release \
     -DCMAKE_INSTALL_PREFIX=%{i} \
@@ -40,15 +40,13 @@ cmake ../%{n}-%{realversion} \
     -DCMAKE_CXX_FLAGS="%{build_flags}" \
     -DBUILD_TEST=OFF \
     -DWITH_PYTHON=OFF \
-    -DWITH_CUDA=OFF \
-    -DBUILD_TEST=OFF \
 %if 0%{!?without_cuda:1}
-    -DUSE_CUDA=${USE_CUDA} \
+    -DWITH_CUDA=${USE_CUDA} \
     -DTORCH_CUDA_ARCH_LIST="%{cuda_arch_float}" \
+    -Dnvtx3_dir=${CUDA_ROOT}/include \
 %endif
     -DBUILD_SHARED_LIBS=ON
 
-
 make %{makeprocesses} VERBOSE=1
 
 %install

pytorch-sparse.spec

@@ -10,7 +10,7 @@
 Source: git+https://github.com/%{github_user}/pytorch_sparse.git?obj=%{branch}/%{tag}&export=%{n}-%{realversion}&submodules=1&output=/%{n}-%{realversion}.tgz
 
 BuildRequires: cmake
-Requires: pytorch
+Requires: pytorch %{!?without_cuda:cuda}
 %define build_flags -Wall -Wextra -pedantic %{?arch_build_flags}
 %define cuda_arch_float $(echo %{cuda_arch} | tr ' ' '\\n' | sed -E 's|([0-9])$|.\\1|' | tr '\\n' ' ')
 
@@ -20,18 +20,17 @@ Requires: pytorch
 grep -q 'CMAKE_CXX_STANDARD  *14' CMakeLists.txt
 sed -i -e 's|CMAKE_CXX_STANDARD  *14|CMAKE_CXX_STANDARD %{cms_cxx_standard}|' CMakeLists.txt
 
+%build
+
 USE_CUDA=OFF
 %if 0%{!?without_cuda:1}
 if [ "%{cuda_gcc_support}" = "true" ] ; then
 USE_CUDA=ON
 fi
 %endif
 
-%build
-
 rm -rf ../build && mkdir ../build && cd ../build
 
-
 cmake ../%{n}-%{realversion} \
     -DCMAKE_BUILD_TYPE=Release \
     -DCMAKE_INSTALL_PREFIX=%{i} \
@@ -41,15 +40,13 @@ cmake ../%{n}-%{realversion} \
     -DCMAKE_CXX_FLAGS="%{build_flags}" \
     -DBUILD_TEST=OFF \
     -DWITH_PYTHON=OFF \
-    -DWITH_CUDA=OFF \
-    -DBUILD_TEST=OFF \
 %if 0%{!?without_cuda:1}
-    -DUSE_CUDA=${USE_CUDA} \
+    -DWITH_CUDA=${USE_CUDA} \
     -DTORCH_CUDA_ARCH_LIST="%{cuda_arch_float}" \
+    -Dnvtx3_dir=${CUDA_ROOT}/include \
 %endif
     -DBUILD_SHARED_LIBS=ON
 
-
 make %{makeprocesses} VERBOSE=1
 
 %install

pytorch.spec

@@ -43,6 +43,7 @@ cmake ../%{n}-%{realversion} \
 %if 0%{!?without_cuda:1}
     -DUSE_CUDA=${USE_CUDA} \
     -DTORCH_CUDA_ARCH_LIST="%{cuda_arch_float}" \
+    -Dnvtx3_dir=${CUDA_ROOT}/include \
     -DCUDNN_INCLUDE_DIR=${CUDNN_ROOT}/include \
     -DCUDNN_LIBRARY=${CUDNN_ROOT}/lib64/libcudnn.so \
 %endif

scram-tools.file/tools/cuda/cuda.xml

@@ -1,10 +1,9 @@
-<tool name="cuda" version="@TOOL_VERSION@" revision="3">
+<tool name="cuda" version="@TOOL_VERSION@" revision="4">
   <info url="https://docs.nvidia.com/cuda/index.html"/>
+  <use name="cuda-interface"/>
   <use name="cuda-stubs"/>
   <lib name="cudart"/>
   <lib name="cudadevrt"/>
-  <lib name="nvToolsExt"/>
-  <use name="cuda-interface"/>
   <client>
     <environment name="CUDA_BASE" default="@TOOL_ROOT@"/>
     <environment name="NVCC"      default="$CUDA_BASE/bin/nvcc"/>