| Summary: | sci-chemistry/gromacs-4.6.* with dev-util/nvidia-cuda-toolkit-4.2.9 - nvcc fatal : redefinition of argument 'compiler-bindir' | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | Thomas Haschka <raketenzwerg83> |
| Component: | Current packages | Assignee: | Justin Lecher (RETIRED) <jlec> |
| Status: | RESOLVED OBSOLETE | ||
| Severity: | normal | CC: | rose, sci-chemistry |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | All | ||
| OS: | Linux | ||
| See Also: | https://bugs.gentoo.org/show_bug.cgi?id=496632 | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Bug Depends on: | 486578 | ||
| Bug Blocks: | 472124 | ||
| Attachments: | build.log | ||
|
Description
Thomas Haschka
2013-05-28 08:06:02 UTC
Created attachment 349422 [details]
build.log
The problem is probably found here..
-DCUDA_HOST_COMPILER_OPTIONS=-O2 -v --compiler-bindir=\"/usr/x86_64-pc-linux-gnu/gcc-bin/4.6.3\"
or in the ebuild where it says
-DCUDA_HOST_COMPILER_OPTIONS="${NVCCFLAGS}" (line 185)
which by default seems to produce the line above on my machine.. which doesn't work.
But trying to set
export NVCCFLAGS="-O2"
or
export NVCCFLAGS=""
prior emerging did fail either..
( in both cases --compiler-bindir is set as shown above.. )
Actually it worked before those changes... i will try to get the old ebuild.. http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/sci-chemistry/gromacs/gromacs-4.6.1.ebuild?r1=1.1&r2=1.2 FAILED: cd /var/tmp/portage/sci-chemistry/gromacs-4.6.1/work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir && /usr/bin/cmake -E make_directory /var/tmp/portage/sci-chemistry/gromacs-4.6.1/work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//. && /usr/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=Gentoo -D generated_file:STRING=/var/tmp/portage/sci-chemistry/gromacs-4.6.1/work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//./gpu_utils_generated_gpu_utils.cu.o -D generated_cubin_file:STRING=/var/tmp/portage/sci-chemistry/gromacs-4.6.1/work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//./gpu_utils_generated_gpu_utils.cu.o.cubin.txt -P /var/tmp/portage/sci-chemistry/gromacs-4.6.1/work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//gpu_utils_generated_gpu_utils.cu.o.cmake nvcc fatal : redefinition of argument 'compiler-bindir' CMake Error at gpu_utils_generated_gpu_utils.cu.o.cmake:208 (message): Error generating /var/tmp/portage/sci-chemistry/gromacs-4.6.1/work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//./gpu_utils_generated_gpu_utils.cu.o FAILED: cd .../work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir && /usr/bin/cmake -E make_directory .../work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//. && /usr/bin/cmake -D verbose:BOOL=OFF -D build_configuration:STRING=Gentoo -D generated_file:STRING=.../gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//./gpu_utils_generated_gpu_utils.cu.o -D generated_cubin_file:STRING=.../work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//./gpu_utils_generated_gpu_utils.cu.o.cubin.txt -P .../work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//gpu_utils_generated_gpu_utils.cu.o.cmake nvcc fatal : redefinition of argument 'compiler-bindir' CMake Error at gpu_utils_generated_gpu_utils.cu.o.cmake:208 (message): Error generating .../work/gromacs-4.6.1_float/src/gmxlib/gpu_utils/CMakeFiles/gpu_utils.dir//./gpu_utils_generated_gpu_utils.cu.o I currently don't have access to my cuda system, so I leave it to Justin.
@justin: Gromacs' cuda detection is custom, so I guess the flags are duplicated in ${WORKDIR}/cmake/gmxManageNvccConfig.cmake !
Should be fixed in 4.6.2 Over here it works as expected in 4.6.2 - Thanks! *** Bug 472260 has been marked as a duplicate of this bug. *** This one isn't so easy. I will try to find a solution soonish. Is this fixed with nvidia-toolkit-5.5? *** Bug 486224 has been marked as a duplicate of this bug. *** With cuda 5.5 it should be fixed |
