Attachment 10801 Details for Bug 24411 – nvidia-cuda 10.1 spec file

nvidia-cuda 10.1 spec file

nvidia-cuda-toolkit10.1.spec (text/x-matlab), 11.96 KB, created by Giuseppe Ghibò on 2019-02-27 18:06:12 CET

(hide)

Description:

Filename:

MIME Type:

Creator: Giuseppe Ghibò

Created: 2019-02-27 18:06:12 CET

Size: 11.96 KB

patch

obsolete

>%define		__jar_repack %nil
>%define		driver_ver 418.39
>%define		cuda_major 10.1
>
>%define		_enable_debug_packages %{nil}
>%define		debug_package %{nil}
>
>Summary:	NVIDIA CUDA runtime libraries
>Name:		nvidia-cuda-toolkit
>Version:	10.1.105
>Release:	%mkrel 1
>Source0:	https://developer.nvidia.com/compute/cuda/%{cuda_major}/Prod/local_installers/cuda_%{version}_%{driver_ver}_linux.run
>Source2:	nvidia
>Source10:	nvvp.desktop
>Source11:	nsight.desktop
>Patch0:		cuda-nsightcompute-fix-smem.patch
>License:	Freeware
>Group:		Development/Other
>Url:		http://www.nvidia.com/cuda/
>Recommends:	nvidia >= %{driver_ver}
>BuildRequires:	imagemagick
>BuildRequires:	jpackage-utils
>ExclusiveArch:	x86_64
>
># We don't require installation of the NVIDIA graphics drivers so that 
># folks can do CUDA development on systems without NVIDIA hardware.
>
># A library, libcudainj.so, was introduced in CUDA 4.1, which depends
># on libcuda.so. It is not needed to compile CUDA programs, though.
># python(abi) auto-require is triggered by runant.py script somewhere
># inside Eclipse plugins used by NVVP.
>%global __requires_exclude libcuda.so.1|devel\$libcuda\$|devel\$libcuda\\(64bit\$\\)|python\$abi\$
>%global __provides_exclude libcairo.so.2
>
>%description
>NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
>that leverages the parallel compute engine in NVIDIA graphics
>processing units (GPUs) to solve many complex computational problems
>in a fraction of the time required on a CPU. It includes the CUDA
>Instruction Set Architecture (ISA) and the parallel compute engine in
>the GPU. To program to the CUDAâ¢ architecture, developers can, today,
>use C++, one of the most widely used high-level programming languages,
>which can then be run at great performance on a CUDAâ¢ enabled
>processor. Support for other languages, like FORTRAN, Python or Java,
>is available from third parties.
>
>This package contains the libraries and attendant files needed to run
>programs that make use of CUDA.
>
>%package devel
>Summary:	NVIDIA CUDA Toolkit development files
>Group:		Development/Other
>Requires:	%{name} = %{version}-%{release}
>Recommends:	nvidia-devel >= %{driver_ver}
>Recommends:	gcc-c++
>
>%description devel
>NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
>that leverages the parallel compute engine in NVIDIA graphics
>processing units (GPUs) to solve many complex computational problems
>in a fraction of the time required on a CPU. It includes the CUDA
>Instruction Set Architecture (ISA) and the parallel compute engine in
>the GPU. To program to the CUDAâ¢ architecture, developers can, today,
>use C++, one of the most widely used high-level programming languages,
>which can then be run at great performance on a CUDAâ¢ enabled
>processor. Support for other languages, like FORTRAN, Python or Java,
>is available from third parties.
>
>This package contains the development files needed to build programs
>that make use of CUDA.
>
>%package samples
>Summary:	NVIDIA CUDA Toolkit samples
>Recommends:	%{name}-devel = %{version}-%{release}
>
>%description samples
>NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
>that leverages the parallel compute engine in NVIDIA graphics
>processing units (GPUs) to solve many complex computational problems
>in a fraction of the time required on a CPU. It includes the CUDA
>Instruction Set Architecture (ISA) and the parallel compute engine in
>the GPU. To program to the CUDAâ¢ architecture, developers can, today,
>use C++, one of the most widely used high-level programming languages,
>which can then be run at great performance on a CUDAâ¢ enabled
>processor. Support for other languages, like FORTRAN, Python or Java,
>is available from third parties.
>
>This package contains numerous CUDA code samples (formerly CUDA SDK).
>
>%package -n nvidia-visual-profiler
>Summary:	NVIDIA Visual Profiler
>Group:		Development/Other
>Requires:	java
>Obsoletes:	nvidia-cuda-profiler < 4.0, nvidia-opencl-profiler < 4.0, nvidia-compute-profiler < 5.0
>Recommends:	nvidia-devel >= %{driver_ver}
>Recommends:	%{name} = %{version}-%{release}
>
># We don't strictly require NVIDIA CUDA Toolkit, because the profiler
># could be used to analyze CSV profile logs obtained elsewhere.
>
>%description -n nvidia-visual-profiler
>NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
>that leverages the parallel compute engine in NVIDIA graphics
>processing units (GPUs) to solve many complex computational problems
>in a fraction of the time required on a CPU. It includes the CUDA
>Instruction Set Architecture (ISA) and the parallel compute engine in
>the GPU. To program to the CUDAâ¢ architecture, developers can, today,
>use C++, one of the most widely used high-level programming languages,
>which can then be run at great performance on a CUDAâ¢ enabled
>processor. Support for other languages, like FORTRAN, Python or Java,
>is available from third parties.
>
>This package contains NVIDIA Visual Profiler for CUDA and OpenCL.
>
>%package -n nvidia-nsight
>Summary:	NVIDIA Nsight IDE
>Group:		Development/Other
>Requires:	java
>Recommends:	nvidia-devel >= %{driver_ver}
>
># We don't strictly require NVIDIA CUDA Toolkit, because Nsight IDE
># could be used to develop CUDA programs on a remote node.
>
>%description -n nvidia-nsight
>NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
>that leverages the parallel compute engine in NVIDIA graphics
>processing units (GPUs) to solve many complex computational problems
>in a fraction of the time required on a CPU. It includes the CUDA
>Instruction Set Architecture (ISA) and the parallel compute engine in
>the GPU. To program to the CUDAâ¢ architecture, developers can, today,
>use C++, one of the most widely used high-level programming languages,
>which can then be run at great performance on a CUDAâ¢ enabled
>processor. Support for other languages, like FORTRAN, Python or Java,
>is available from third parties.
>
>This package contains Nsight Eclipse Edition, a full-featured CUDA IDE.
>
>%prep
>%setup -q -T -c %{name}-%{version}
>
>%build
># Nothing to do
>
>%install
>%__install -d -m 755 %{buildroot}%{_usr}
>%__install -d -m 755 %{buildroot}%{_bindir}
>%__install -d -m 755 %{buildroot}%{_datadir}/%{name}
>%__install -d -m 755 %{buildroot}%{_datadir}/%{name}/samples
>%__install -d -m 755 %{buildroot}%{_datadir}/applications
>%__install -d -m 755 %{buildroot}%{_docdir}/%{name}-devel
>
>bash %SOURCE0 --tar xf -C .
>
>pushd builds/cublas
>cp -pr include %{buildroot}%{_includedir}
>cp -pr lib64 %{buildroot}%{_libdir}
>popd
>
>pushd builds/cuda-toolkit
>cp -pr targets/x86_64-linux/include/* %{buildroot}%{_includedir}/
>cp -pr targets/x86_64-linux/lib/*     %{buildroot}%{_libdir}/
>cp -pr bin/*                          %{buildroot}%{_bindir}/
>cp -pr doc                            %{buildroot}%{_usr}/
>cp -pr extras                         %{buildroot}%{_usr}/
>cp -pr jre                            %{buildroot}%{_usr}/
>cp -pr libnvvp                        %{buildroot}%{_usr}/
>cp -pr libnsight                      %{buildroot}%{_usr}/
>cp -pr nsightee_plugins               %{buildroot}%{_usr}/
>cp -pr nvml                           %{buildroot}%{_usr}/
>cp -pr nvvm                           %{buildroot}%{_usr}/
>cp -pr NsightCompute-2019.1           %{buildroot}%{_usr}/
>cp -pr NsightSystems-2018.3           %{buildroot}%{_usr}/
>cp -pr share                          %{buildroot}%{_usr}/
>cp -pr src                            %{buildroot}%{_usr}/
>cp -pr tools                          %{buildroot}%{_usr}/
>cp -pr EULA.txt                       %{buildroot}%{_usr}/
>cp -pr version.txt                    %{buildroot}%{_usr}/
>popd
>
>
>#%__rm -rf %{buildroot}%{_usr}/lib
>%__rm -rf %{buildroot}%{_usr}/extras/CUPTI/lib
># Fix nvcc.profile for current path
>sed -i 's/lib:/lib64:/g' %{buildroot}%{_bindir}/nvcc.profile
>sed -i 's/lib\//lib64\//g' %{buildroot}%{_bindir}/nvcc.profile
># (tmb) restore libdevice
>sed -i 's/lib64device/libdevice/g' %{buildroot}%{_bindir}/nvcc.profile
>
># Fix wrong permissions
>find %{buildroot} -type f -name '*.h' -print0 | xargs -0 chmod 644
>find %{buildroot} -type f -name '*.hpp' -print0 | xargs -0 chmod 644
>
>
># Move compiler components from /usr/nvvm to libdir
>%__mv %{buildroot}%{_usr}/nvvm %{buildroot}%{_libdir}/
># Fix nvcc.profile to reflect the move
>sed -i 's/nvvm/%{_lib}\/nvvm/g' %{buildroot}%{_bindir}/nvcc.profile
>
># move version.txt to devel doc
>%__mv %{buildroot}%{_usr}/version.txt %{buildroot}%{_docdir}/%{name}-devel/
>
># Copy samples (SDK)
>pushd builds/cuda-samples
>cp -pr . %{buildroot}%{_datadir}/%{name}/samples
>popd
>
># move nvml examples files to samples
>%__mv %{buildroot}%{_usr}/nvml %{buildroot}%{_datadir}/%{name}/samples/
>
># Fix samples path
>find %{buildroot}%{_datadir}/%{name}/samples -type f -name Makefile -exec sed -i 's@/usr/local/cuda@%_usr@g' {} \; -exec echo "Fix path in " {} \;
>
># Remove duplicates (these will be shipped with -devel)
>%__rm -rf %{buildroot}%{_datadir}/%{name}/sdk/{doc,tools}
>%__rm -rf %{buildroot}%{_datadir}/%{name}/sdk/Documentation.html
>%__rm -rf %{buildroot}%{_usr}/InstallUtils.pm
>
>%__rm -f  %{buildroot}%{_datadir}/%{name}/install-sdk-linux.pl
>%__rm -rf %{buildroot}%{_usr}/install-linux.pl
>%__rm -rf %{buildroot}%{_usr}/uninstall_cuda.pl
>
>%__mv %{buildroot}%{_usr}/doc/* %{buildroot}%{_docdir}/%{name}-devel/
>%__rm -r %{buildroot}%{_usr}/doc
>%__mv %{buildroot}%{_usr}/src %{buildroot}%{_datadir}/%{name}
>%__mv %{buildroot}%{_usr}/libnvvp %{buildroot}%{_libdir}/nvvp
>%__mv %{buildroot}%{_usr}/libnsight %{buildroot}%{_libdir}/nsight
>%__mv %{buildroot}%{_usr}/EULA.txt %{buildroot}%{_docdir}/%{name}-devel/
>%__mv %{buildroot}%{_usr}/{extras,tools} %{buildroot}%{_datadir}/%{name}
>
># Move nsightee_plugins from /usr/nsightee_plugins to libdir
>%__mv %{buildroot}%{_usr}/nsightee_plugins %{buildroot}%{_libdir}
>
># Move /usr/NsightCompute-2019.1 and /usr/NsightSystems-2018.3 to libdir
>%__mv %{buildroot}%{_usr}/NsightCompute-2019.1 %{buildroot}%{_libdir}
>%__mv %{buildroot}%{_usr}/NsightSystems-2018.3 %{buildroot}%{_libdir}
>
># dont ship gdb files
>%__rm -rf %{buildroot}%_datadir/gdb
>
># Remove bundled JRE and fix paths for Java
>%__rm -rf %{buildroot}%{_usr}/jre
>sed -i 's|\.\./jre/bin/java|%{_jvmdir}/jre/bin/java|g' %{buildroot}%{_libdir}/nvvp/nvvp.ini
>sed -i 's|\.\./jre/bin/java|%{_jvmdir}/jre/bin/java|g' %{buildroot}%{_libdir}/nsight/nsight.ini
>
># dont ship OpenCL files, we already ship them as separate packages
>%__rm -rf %{buildroot}/%{_includedir}/CL
>%__rm -rf %{buildroot}/%{_libdir}/libOpenCL.so*
>
>sed -i 's|$CUDA_BIN/../libnsight/nsight|%{_libdir}/nsight/nsight|g' %{buildroot}%{_bindir}/nsight
>sed -i 's|$CUDA_BIN/../libnvvp/nvvp|%{_libdir}/nvvp/nvvp|g' %{buildroot}%{_bindir}/nvvp
>
>for S in 16 24 32 48 64 128 192 256; do
> %__install -d -m755 %{buildroot}%{_iconsdir}/hicolor/$S\x$S/apps
> convert -scale $S\x$S %{buildroot}/%{_libdir}/nvvp/icon.xpm %{buildroot}%{_iconsdir}/hicolor/$S\x$S/apps/nvvp.png
> convert -scale $S\x$S %{buildroot}/%{_libdir}/nsight/icon.xpm %{buildroot}%{_iconsdir}/hicolor/$S\x$S/apps/nsight.png
>done
>
>%__install -m644 %{SOURCE10} %{buildroot}%{_datadir}/applications/
>%__install -m644 %{SOURCE11} %{buildroot}%{_datadir}/applications/
>%__install -D -m 755 %SOURCE2 %{buildroot}%{_sysconfdir}/init.d/nvidia
>
># Don't prevent the use of gcc 5.4
>#sed -i 's|__GNUC__ > 4|__GNUC__ > 5|g' %{buildroot}%{_includedir}/crt/host_config.h
>#sed -i 's|__GNUC__ == 4|__GNUC__ == 5|g' %{buildroot}%{_includedir}/crt/host_config.h
>#sed -i 's|__GNUC_MINOR__ > 8|__GNUC_MINOR__ > 4|g' %{buildroot}%{_includedir}/crt/host_config.h
>
>%files
>%_libdir/*.so.*
>%_sysconfdir/init.d/*
>
>%files devel
>%doc %{_docdir}/%{name}-devel/*
>%_bindir/*
>%exclude %_bindir/nvvp
>%exclude %_bindir/nsight
>%_libdir/*.so
>%_libdir/*.a
>%_includedir/*
>%_libdir/nvvm/*
>%_libdir/stubs/*
>%_datadir/%{name}/*
>%exclude %_datadir/%{name}/samples
>
>%files samples
>%_datadir/%{name}/samples
>
>%files -n nvidia-visual-profiler
>%_bindir/nvvp
>%_libdir/nvvp/.eclipseproduct
>%_libdir/nvvp/*
>%_datadir/applications/nvvp.desktop
>%_iconsdir/hicolor/*/apps/nvvp.png
>
>%files -n nvidia-nsight
>%_bindir/nsight
>%_libdir/nsight/.eclipseproduct
>%_libdir/nsight/*
>%_datadir/applications/nsight.desktop
>%_iconsdir/hicolor/*/apps/nsight.png
>%_libdir/nsightee_plugins
>%_libdir/NsightCompute-2019.1
>%_libdir/NsightSystems-2018.3

%define		__jar_repack %nil
%define		driver_ver 418.39
%define		cuda_major 10.1

%define		_enable_debug_packages %{nil}
%define		debug_package %{nil}

Summary:	NVIDIA CUDA runtime libraries
Name:		nvidia-cuda-toolkit
Version:	10.1.105
Release:	%mkrel 1
Source0:	https://developer.nvidia.com/compute/cuda/%{cuda_major}/Prod/local_installers/cuda_%{version}_%{driver_ver}_linux.run
Source2:	nvidia
Source10:	nvvp.desktop
Source11:	nsight.desktop
Patch0:		cuda-nsightcompute-fix-smem.patch
License:	Freeware
Group:		Development/Other
Url:		http://www.nvidia.com/cuda/
Recommends:	nvidia >= %{driver_ver}
BuildRequires:	imagemagick
BuildRequires:	jpackage-utils
ExclusiveArch:	x86_64

# We don't require installation of the NVIDIA graphics drivers so that 
# folks can do CUDA development on systems without NVIDIA hardware.

# A library, libcudainj.so, was introduced in CUDA 4.1, which depends
# on libcuda.so. It is not needed to compile CUDA programs, though.
# python(abi) auto-require is triggered by runant.py script somewhere
# inside Eclipse plugins used by NVVP.
%global __requires_exclude libcuda.so.1|devel\\(libcuda\\)|devel\\(libcuda\\(64bit\\)\\)|python\\(abi\\)
%global __provides_exclude libcairo.so.2

%description
NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
that leverages the parallel compute engine in NVIDIA graphics
processing units (GPUs) to solve many complex computational problems
in a fraction of the time required on a CPU. It includes the CUDA
Instruction Set Architecture (ISA) and the parallel compute engine in
the GPU. To program to the CUDAâ¢ architecture, developers can, today,
use C++, one of the most widely used high-level programming languages,
which can then be run at great performance on a CUDAâ¢ enabled
processor. Support for other languages, like FORTRAN, Python or Java,
is available from third parties.

This package contains the libraries and attendant files needed to run
programs that make use of CUDA.

%package devel
Summary:	NVIDIA CUDA Toolkit development files
Group:		Development/Other
Requires:	%{name} = %{version}-%{release}
Recommends:	nvidia-devel >= %{driver_ver}
Recommends:	gcc-c++

%description devel
NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
that leverages the parallel compute engine in NVIDIA graphics
processing units (GPUs) to solve many complex computational problems
in a fraction of the time required on a CPU. It includes the CUDA
Instruction Set Architecture (ISA) and the parallel compute engine in
the GPU. To program to the CUDAâ¢ architecture, developers can, today,
use C++, one of the most widely used high-level programming languages,
which can then be run at great performance on a CUDAâ¢ enabled
processor. Support for other languages, like FORTRAN, Python or Java,
is available from third parties.

This package contains the development files needed to build programs
that make use of CUDA.

%package samples
Summary:	NVIDIA CUDA Toolkit samples
Recommends:	%{name}-devel = %{version}-%{release}

%description samples
NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
that leverages the parallel compute engine in NVIDIA graphics
processing units (GPUs) to solve many complex computational problems
in a fraction of the time required on a CPU. It includes the CUDA
Instruction Set Architecture (ISA) and the parallel compute engine in
the GPU. To program to the CUDAâ¢ architecture, developers can, today,
use C++, one of the most widely used high-level programming languages,
which can then be run at great performance on a CUDAâ¢ enabled
processor. Support for other languages, like FORTRAN, Python or Java,
is available from third parties.

This package contains numerous CUDA code samples (formerly CUDA SDK).

%package -n nvidia-visual-profiler
Summary:	NVIDIA Visual Profiler
Group:		Development/Other
Requires:	java
Obsoletes:	nvidia-cuda-profiler < 4.0, nvidia-opencl-profiler < 4.0, nvidia-compute-profiler < 5.0
Recommends:	nvidia-devel >= %{driver_ver}
Recommends:	%{name} = %{version}-%{release}

# We don't strictly require NVIDIA CUDA Toolkit, because the profiler
# could be used to analyze CSV profile logs obtained elsewhere.

%description -n nvidia-visual-profiler
NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
that leverages the parallel compute engine in NVIDIA graphics
processing units (GPUs) to solve many complex computational problems
in a fraction of the time required on a CPU. It includes the CUDA
Instruction Set Architecture (ISA) and the parallel compute engine in
the GPU. To program to the CUDAâ¢ architecture, developers can, today,
use C++, one of the most widely used high-level programming languages,
which can then be run at great performance on a CUDAâ¢ enabled
processor. Support for other languages, like FORTRAN, Python or Java,
is available from third parties.

This package contains NVIDIA Visual Profiler for CUDA and OpenCL.

%package -n nvidia-nsight
Summary:	NVIDIA Nsight IDE
Group:		Development/Other
Requires:	java
Recommends:	nvidia-devel >= %{driver_ver}

# We don't strictly require NVIDIA CUDA Toolkit, because Nsight IDE
# could be used to develop CUDA programs on a remote node.

%description -n nvidia-nsight
NVIDIAÂ® CUDAâ¢ is a general purpose parallel computing architecture
that leverages the parallel compute engine in NVIDIA graphics
processing units (GPUs) to solve many complex computational problems
in a fraction of the time required on a CPU. It includes the CUDA
Instruction Set Architecture (ISA) and the parallel compute engine in
the GPU. To program to the CUDAâ¢ architecture, developers can, today,
use C++, one of the most widely used high-level programming languages,
which can then be run at great performance on a CUDAâ¢ enabled
processor. Support for other languages, like FORTRAN, Python or Java,
is available from third parties.

This package contains Nsight Eclipse Edition, a full-featured CUDA IDE.

%prep
%setup -q -T -c %{name}-%{version}

%build
# Nothing to do

%install
%__install -d -m 755 %{buildroot}%{_usr}
%__install -d -m 755 %{buildroot}%{_bindir}
%__install -d -m 755 %{buildroot}%{_datadir}/%{name}
%__install -d -m 755 %{buildroot}%{_datadir}/%{name}/samples
%__install -d -m 755 %{buildroot}%{_datadir}/applications
%__install -d -m 755 %{buildroot}%{_docdir}/%{name}-devel

bash %SOURCE0 --tar xf -C .

pushd builds/cublas
cp -pr include %{buildroot}%{_includedir}
cp -pr lib64 %{buildroot}%{_libdir}
popd

pushd builds/cuda-toolkit
cp -pr targets/x86_64-linux/include/* %{buildroot}%{_includedir}/
cp -pr targets/x86_64-linux/lib/*     %{buildroot}%{_libdir}/
cp -pr bin/*                          %{buildroot}%{_bindir}/
cp -pr doc                            %{buildroot}%{_usr}/
cp -pr extras                         %{buildroot}%{_usr}/
cp -pr jre                            %{buildroot}%{_usr}/
cp -pr libnvvp                        %{buildroot}%{_usr}/
cp -pr libnsight                      %{buildroot}%{_usr}/
cp -pr nsightee_plugins               %{buildroot}%{_usr}/
cp -pr nvml                           %{buildroot}%{_usr}/
cp -pr nvvm                           %{buildroot}%{_usr}/
cp -pr NsightCompute-2019.1           %{buildroot}%{_usr}/
cp -pr NsightSystems-2018.3           %{buildroot}%{_usr}/
cp -pr share                          %{buildroot}%{_usr}/
cp -pr src                            %{buildroot}%{_usr}/
cp -pr tools                          %{buildroot}%{_usr}/
cp -pr EULA.txt                       %{buildroot}%{_usr}/
cp -pr version.txt                    %{buildroot}%{_usr}/
popd


#%__rm -rf %{buildroot}%{_usr}/lib
%__rm -rf %{buildroot}%{_usr}/extras/CUPTI/lib
# Fix nvcc.profile for current path
sed -i 's/lib:/lib64:/g' %{buildroot}%{_bindir}/nvcc.profile
sed -i 's/lib\//lib64\//g' %{buildroot}%{_bindir}/nvcc.profile
# (tmb) restore libdevice
sed -i 's/lib64device/libdevice/g' %{buildroot}%{_bindir}/nvcc.profile

# Fix wrong permissions
find %{buildroot} -type f -name '*.h' -print0 | xargs -0 chmod 644
find %{buildroot} -type f -name '*.hpp' -print0 | xargs -0 chmod 644


# Move compiler components from /usr/nvvm to libdir
%__mv %{buildroot}%{_usr}/nvvm %{buildroot}%{_libdir}/
# Fix nvcc.profile to reflect the move
sed -i 's/nvvm/%{_lib}\/nvvm/g' %{buildroot}%{_bindir}/nvcc.profile

# move version.txt to devel doc
%__mv %{buildroot}%{_usr}/version.txt %{buildroot}%{_docdir}/%{name}-devel/

# Copy samples (SDK)
pushd builds/cuda-samples
cp -pr . %{buildroot}%{_datadir}/%{name}/samples
popd

# move nvml examples files to samples
%__mv %{buildroot}%{_usr}/nvml %{buildroot}%{_datadir}/%{name}/samples/

# Fix samples path
find %{buildroot}%{_datadir}/%{name}/samples -type f -name Makefile -exec sed -i 's@/usr/local/cuda@%_usr@g' {} \; -exec echo "Fix path in " {} \;

# Remove duplicates (these will be shipped with -devel)
%__rm -rf %{buildroot}%{_datadir}/%{name}/sdk/{doc,tools}
%__rm -rf %{buildroot}%{_datadir}/%{name}/sdk/Documentation.html
%__rm -rf %{buildroot}%{_usr}/InstallUtils.pm

%__rm -f  %{buildroot}%{_datadir}/%{name}/install-sdk-linux.pl
%__rm -rf %{buildroot}%{_usr}/install-linux.pl
%__rm -rf %{buildroot}%{_usr}/uninstall_cuda.pl

%__mv %{buildroot}%{_usr}/doc/* %{buildroot}%{_docdir}/%{name}-devel/
%__rm -r %{buildroot}%{_usr}/doc
%__mv %{buildroot}%{_usr}/src %{buildroot}%{_datadir}/%{name}
%__mv %{buildroot}%{_usr}/libnvvp %{buildroot}%{_libdir}/nvvp
%__mv %{buildroot}%{_usr}/libnsight %{buildroot}%{_libdir}/nsight
%__mv %{buildroot}%{_usr}/EULA.txt %{buildroot}%{_docdir}/%{name}-devel/
%__mv %{buildroot}%{_usr}/{extras,tools} %{buildroot}%{_datadir}/%{name}

# Move nsightee_plugins from /usr/nsightee_plugins to libdir
%__mv %{buildroot}%{_usr}/nsightee_plugins %{buildroot}%{_libdir}

# Move /usr/NsightCompute-2019.1 and /usr/NsightSystems-2018.3 to libdir
%__mv %{buildroot}%{_usr}/NsightCompute-2019.1 %{buildroot}%{_libdir}
%__mv %{buildroot}%{_usr}/NsightSystems-2018.3 %{buildroot}%{_libdir}

# dont ship gdb files
%__rm -rf %{buildroot}%_datadir/gdb

# Remove bundled JRE and fix paths for Java
%__rm -rf %{buildroot}%{_usr}/jre
sed -i 's|\.\./jre/bin/java|%{_jvmdir}/jre/bin/java|g' %{buildroot}%{_libdir}/nvvp/nvvp.ini
sed -i 's|\.\./jre/bin/java|%{_jvmdir}/jre/bin/java|g' %{buildroot}%{_libdir}/nsight/nsight.ini

# dont ship OpenCL files, we already ship them as separate packages
%__rm -rf %{buildroot}/%{_includedir}/CL
%__rm -rf %{buildroot}/%{_libdir}/libOpenCL.so*

sed -i 's|$CUDA_BIN/../libnsight/nsight|%{_libdir}/nsight/nsight|g' %{buildroot}%{_bindir}/nsight
sed -i 's|$CUDA_BIN/../libnvvp/nvvp|%{_libdir}/nvvp/nvvp|g' %{buildroot}%{_bindir}/nvvp

for S in 16 24 32 48 64 128 192 256; do
 %__install -d -m755 %{buildroot}%{_iconsdir}/hicolor/$S\x$S/apps
 convert -scale $S\x$S %{buildroot}/%{_libdir}/nvvp/icon.xpm %{buildroot}%{_iconsdir}/hicolor/$S\x$S/apps/nvvp.png
 convert -scale $S\x$S %{buildroot}/%{_libdir}/nsight/icon.xpm %{buildroot}%{_iconsdir}/hicolor/$S\x$S/apps/nsight.png
done

%__install -m644 %{SOURCE10} %{buildroot}%{_datadir}/applications/
%__install -m644 %{SOURCE11} %{buildroot}%{_datadir}/applications/
%__install -D -m 755 %SOURCE2 %{buildroot}%{_sysconfdir}/init.d/nvidia

# Don't prevent the use of gcc 5.4
#sed -i 's|__GNUC__ > 4|__GNUC__ > 5|g' %{buildroot}%{_includedir}/crt/host_config.h
#sed -i 's|__GNUC__ == 4|__GNUC__ == 5|g' %{buildroot}%{_includedir}/crt/host_config.h
#sed -i 's|__GNUC_MINOR__ > 8|__GNUC_MINOR__ > 4|g' %{buildroot}%{_includedir}/crt/host_config.h

%files
%_libdir/*.so.*
%_sysconfdir/init.d/*

%files devel
%doc %{_docdir}/%{name}-devel/*
%_bindir/*
%exclude %_bindir/nvvp
%exclude %_bindir/nsight
%_libdir/*.so
%_libdir/*.a
%_includedir/*
%_libdir/nvvm/*
%_libdir/stubs/*
%_datadir/%{name}/*
%exclude %_datadir/%{name}/samples

%files samples
%_datadir/%{name}/samples

%files -n nvidia-visual-profiler
%_bindir/nvvp
%_libdir/nvvp/.eclipseproduct
%_libdir/nvvp/*
%_datadir/applications/nvvp.desktop
%_iconsdir/hicolor/*/apps/nvvp.png

%files -n nvidia-nsight
%_bindir/nsight
%_libdir/nsight/.eclipseproduct
%_libdir/nsight/*
%_datadir/applications/nsight.desktop
%_iconsdir/hicolor/*/apps/nsight.png
%_libdir/nsightee_plugins
%_libdir/NsightCompute-2019.1
%_libdir/NsightSystems-2018.3

Actions: View

Attachments on bug 24411: 10794 | 10795 | 10796 | 10801 | 10804 | 10805 | 10806 | 10811 | 10812 | 10813 | 10814 | 10815 | 10817 | 10818 | 10819 | 10822 | 10823 | 10824 | 10825 | 10826 | 10827 | 10828 | 10829 | 10831 | 10832 | 10833 | 10835 | 10837 | 10839 | 10869 | 10870 | 10871 | 10872