add: cuda-nvcc (#3025)

2026-05-31 09:01:55 +00:00 · 2025-01-17 05:13:48 +07:00
parent 70e7b459fe
commit d6965718bb
4 changed files with 203 additions and 0 deletions
@@ -0,0 +1,5 @@
+project "pkg" {
+    rpm {
+        spec = "cuda-nvcc.spec"
+    }
+}
@@ -0,0 +1,189 @@
+%global real_name cuda_nvcc
+
+%global debug_package %{nil}
+%global __strip /bin/true
+%global _missing_build_ids_terminate_build 0
+%global _build_id_links none
+%global major_package_version 12-6
+
+Name:           %(echo %real_name | tr '_' '-')
+Epoch:          1
+Version:        12.6.85
+Release:        1%{?dist}
+Summary:        CUDA Compiler (NVCC)
+License:        CUDA Toolkit
+URL:            https://developer.nvidia.com/cuda-toolkit
+ExclusiveArch:  x86_64 aarch64
+
+Source0:        https://developer.download.nvidia.com/compute/cuda/redist/%{real_name}/linux-x86_64/%{real_name}-linux-x86_64-%{version}-archive.tar.xz
+Source1:        https://developer.download.nvidia.com/compute/cuda/redist/%{real_name}/linux-sbsa/%{real_name}-linux-sbsa-%{version}-archive.tar.xz
+Source3:        nvcc.profile
+
+Requires(post): ldconfig
+Conflicts:      %{name}-%{major_package_version} < %{?epoch:%{epoch}:}%{version}-%{release}
+
+# CUDA 12.4 does not support GCC 14+:
+%if 0%{?fedora} >= 40
+Requires:       cuda-gcc
+%endif
+
+%description
+The compilation trajectory involves several splitting, compilation,
+preprocessing, and merging steps for each CUDA source file. It is the purpose of
+nvcc, the CUDA compiler driver, to hide the intricate details of CUDA
+compilation from developers. It accepts a range of conventional compiler
+options, such as for defining macros and include/library paths, and for steering
+the compilation process. All non-CUDA compilation steps are forwarded to a C++
+host compiler that is supported by nvcc, and nvcc translates its options to
+appropriate host compiler command line options.
+
+NVVM IR is a compiler IR (intermediate representation) based on the LLVM IR.
+The NVVM IR is designed to represent GPU compute kernels (for example, CUDA
+kernels). High-level language front-ends, like the CUDA C compiler front-end,
+can generate NVVM IR. The NVVM compiler (which is based on LLVM) generates PTX
+code from NVVM IR.
+
+%prep
+%ifarch x86_64
+%setup -q -n %{real_name}-linux-x86_64-%{version}-archive
+%endif
+
+%ifarch aarch64
+%setup -q -T -b 1 -n %{real_name}-linux-sbsa-%{version}-archive
+%endif
+
+%install
+mkdir -p %{buildroot}%{_bindir}
+
+mkdir -p %{buildroot}%{_includedir}
+mkdir -p %{buildroot}%{_libdir}
+mkdir -p %{buildroot}%{_datadir}/nvcc/
+
+cp -fr bin/* nvvm/bin/* %{buildroot}%{_bindir}/
+cp -fr include/* nvvm/include/* %{buildroot}%{_includedir}/
+cp -fr lib/* nvvm/lib64/* %{buildroot}%{_libdir}/
+cp -fr nvvm/libdevice/* %{buildroot}%{_datadir}/nvcc/
+
+cp -f %{SOURCE3} %{buildroot}%{_bindir}/
+
+# Set proper variables
+sed -i \
+    -e 's|LIBDIR|%{_libdir}|g' \
+    -e 's|INCLUDE_DIR|%{_includedir}/cuda|g' \
+    %{buildroot}/%{_bindir}/nvcc.profile
+
+%{?ldconfig_scriptlets}
+
+%files
+%license LICENSE
+%{_bindir}/bin2c
+%{_bindir}/cicc
+%dir %{_bindir}/crt/
+%{_bindir}/crt/link.stub
+%{_bindir}/crt/prelink.stub
+%{_bindir}/cudafe++
+%{_bindir}/fatbinary
+%{_bindir}/nvcc
+%{_bindir}/__nvcc_device_query
+%{_bindir}/nvcc.profile
+%{_bindir}/nvlink
+%{_bindir}/ptxas
+%dir %{_datadir}/nvcc/
+%{_datadir}/nvcc/libdevice.10.bc
+%dir %{_includedir}/crt/
+%{_includedir}/crt/common_functions.h
+%{_includedir}/crt/cudacc_ext.h
+%{_includedir}/crt/device_double_functions.h
+%{_includedir}/crt/device_double_functions.hpp
+%{_includedir}/crt/device_functions.h
+%{_includedir}/crt/device_functions.hpp
+%{_includedir}/crt/func_macro.h
+%{_includedir}/crt/host_config.h
+%{_includedir}/crt/host_defines.h
+%{_includedir}/crt/host_runtime.h
+%{_includedir}/crt/math_functions.h
+%{_includedir}/crt/math_functions.hpp
+%{_includedir}/crt/mma.h
+%{_includedir}/crt/mma.hpp
+%{_includedir}/crt/nvfunctional
+%{_includedir}/crt/sm_70_rt.h
+%{_includedir}/crt/sm_70_rt.hpp
+%{_includedir}/crt/sm_80_rt.h
+%{_includedir}/crt/sm_80_rt.hpp
+%{_includedir}/crt/sm_90_rt.h
+%{_includedir}/crt/sm_90_rt.hpp
+%{_includedir}/crt/storage_class.h
+%{_includedir}/fatbinary_section.h
+%{_includedir}/nvPTXCompiler.h
+%{_includedir}/nvvm.h
+%{_libdir}/libnvptxcompiler_static.a
+%{_libdir}/libnvvm.so
+%{_libdir}/libnvvm.so.4
+%{_libdir}/libnvvm.so.4.0.0
+
+%changelog
+* Fri Dec 13 2024 Simone Caronni <negativo17@gmail.com> - 1:12.6.85-1
+- Update to 12.6.85.
+- Require new cuda-gcc profile package.
+
+* Thu Sep 19 2024 Simone Caronni <negativo17@gmail.com> - 1:12.6.68-1
+- Update to 12.6.68.
+
+* Thu Jul 11 2024 Simone Caronni <negativo17@gmail.com> - 1:12.5.82-1
+- Update to 12.5.82.
+
+* Thu Jul 11 2024 Simone Caronni <negativo17@gmail.com> - 1:12.4.99-1
+- Update to 12.4.99.
+
+* Fri Mar 22 2024 Simone Caronni <negativo17@gmail.com> - 1:12.4.99-2
+- Correct GCC requirements.
+
+* Tue Mar 12 2024 Simone Caronni <negativo17@gmail.com> - 1:12.4.99-1
+- Update to 12.4.99.
+- Drop ppc64le.
+
+* Sat Jan 06 2024 Simone Caronni <negativo17@gmail.com> - 1:12.3.107-1
+- Update to 12.3.107.
+
+* Tue Nov 28 2023 Simone Caronni <negativo17@gmail.com> - 1:12.3.103-1
+- Update to 12.3.103.
+
+* Thu Sep 28 2023 Simone Caronni <negativo17@gmail.com> - 1:12.2.140-1
+- Update to 12.2.140.
+
+* Tue Jul 11 2023 Simone Caronni <negativo17@gmail.com> - 1:12.2.91-1
+- Update to 12.2.91.
+
+* Thu Jun 08 2023 Simone Caronni <negativo17@gmail.com> - 1:12.1.105-1
+- Update to 12.1.105.
+
+* Tue Apr 11 2023 Simone Caronni <negativo17@gmail.com> - 1:12.1.66-1
+- Update to 12.1.66.
+
+* Mon Mar 13 2023 Simone Caronni <negativo17@gmail.com> - 1:12.0.140-2
+- Add requirement on cuda-gcc for Fedora 38+.
+
+* Sat Feb 25 2023 Simone Caronni <negativo17@gmail.com> - 1:12.0.140-1
+- Update to 12.0.140.
+
+* Tue Dec 13 2022 Simone Caronni <negativo17@gmail.com> - 1:12.0.76-1
+- Update to 12.0.76.
+
+* Fri Nov 11 2022 Simone Caronni <negativo17@gmail.com> - 1:11.8.89-1
+- Update to 11.8.89.
+- Use aarch64 archive in place of sbsa.
+
+* Sun Sep 04 2022 Simone Caronni <negativo17@gmail.com> - 1:11.7.99-1
+- Update to 11.7.99.
+
+* Thu Jun 23 2022 Simone Caronni <negativo17@gmail.com> - 1:11.7.64-1
+- Update to 11.7.64.
+
+* Thu Mar 31 2022 Simone Caronni <negativo17@gmail.com> - 1:11.6.124-1
+- Update to 11.6.124 (CUDA 11.6.2).
+
+* Tue Mar 08 2022 Simone Caronni <negativo17@gmail.com> - 1:11.6.112-1
+- Update to 11.6.112 (CUDA 11.6.1).
+
+* Fri Jan 28 2022 Simone Caronni <negativo17@gmail.com> - 1:11.6.55-1
+- First build with the new tarball components.
@@ -0,0 +1,6 @@
+NVVMIR_LIBRARY_DIR = /usr/share/cuda
+PATH += /usr/libexec/cuda/open64/bin:
+INCLUDES += "-IINCLUDE_DIR"
+LIBRARIES =+ "-LLIBDIR"
+CUDAFE_FLAGS +=
+PTXAS_FLAGS +=
@@ -0,0 +1,3 @@
+import "andax/nvidia.rhai" as nvidia;
+
+rpm.version(nvidia::nvidia_component_version("cuda_nvcc"));