Summary: | sci-libs/vtk-9.2.6 failed to emerge with USE=cuda - Invalid CUDA architecture given | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Leonid Kopylov <leonchik1976> |
Component: | Current packages | Assignee: | Gentoo Science Related Packages <sci> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | negril.nx+gentoo, proxy-maint, waebbl-gentoo |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | AMD64 | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Attachments: |
build.log
build.log.xz build.log.xz |
Description
Leonid Kopylov
2023-06-29 16:29:13 UTC
Created attachment 864830 [details]
build.log
From your build.log
> Please properly set the VTK_CUDA_ARCH environment variable to
> one of: kepler, maxwell, pascal, volta, turing, ampere, all
Could you set VTK_CUDA_ARCH and try again?
(In reply to Paul Zander from comment #2) > From your build.log > > Please properly set the VTK_CUDA_ARCH environment variable to > > one of: kepler, maxwell, pascal, volta, turing, ampere, all > > Could you set VTK_CUDA_ARCH and try again? To which value I should set it ? If you know your graphics card architecture set it to one of `kepler, maxwell, pascal, volta, turing or ampere`. If it's an `ada` card set it to `ampere`. If you do not know set it to `all`. (In reply to Paul Zander from comment #4) > If you know your graphics card architecture set it to one of `kepler, > maxwell, pascal, volta, turing or ampere`. If it's an `ada` card set it to > `ampere`. > > If you do not know set it to `all`. ok, after some googling, it's 'ampere' card. will update shortly on result... Created attachment 864831 [details]
build.log.xz
Still fail (with another error) Seemingly caused by this: https://forums.developer.nvidia.com/t/including-cub-header-breakes-compilation-with-gcc-12-and-sse2-or-better/255018 https://github.com/NVIDIA/libcudacxx/issues/484 You could try to add "-mno-avx512f" to CFLAGS/CXXFLAGS for vtk as a stop gap and see if that helps. I can't reproduce this due to lack of AVX512 support. (In reply to Paul Zander from comment #8) > Seemingly caused by this: > > https://forums.developer.nvidia.com/t/including-cub-header-breakes- > compilation-with-gcc-12-and-sse2-or-better/255018 > > https://github.com/NVIDIA/libcudacxx/issues/484 > > You could try to add "-mno-avx512f" to CFLAGS/CXXFLAGS for vtk as a stop gap > and see if that helps. I can't reproduce this due to lack of AVX512 support. # cpuid2cpuflags CPU_FLAGS_X86: aes avx avx2 f16c fma3 mmx mmxext pclmul popcnt rdrand sha sse sse2 sse3 sse4_1 sse4_2 ssse3 i don't think i have it neither You have -march=native in your compiler flags which should enable AVX512F on your i7 12700. Created attachment 864834 [details]
build.log.xz
(In reply to Paul Zander from comment #8) > Seemingly caused by this: > > https://forums.developer.nvidia.com/t/including-cub-header-breakes- > compilation-with-gcc-12-and-sse2-or-better/255018 > > https://github.com/NVIDIA/libcudacxx/issues/484 > > You could try to add "-mno-avx512f" to CFLAGS/CXXFLAGS for vtk as a stop gap > and see if that helps. I can't reproduce this due to lack of AVX512 support. added -mno-avx512f - still fails Don't think this is something that can be solved on the gentoo side. There is a minimal test at https://github.com/fwyzard/nvidia_bug_4139266 Maybe upgrading to dev-util/nvidia-cuda-toolkit-12.2.0 does fix it and you can use the minimal test to verify that. (In reply to Paul Zander from comment #13) > Don't think this is something that can be solved on the gentoo side. > > There is a minimal test at https://github.com/fwyzard/nvidia_bug_4139266 > Maybe upgrading to dev-util/nvidia-cuda-toolkit-12.2.0 does fix it and you > can use the minimal test to verify that. test also works for me, so what does it mean? Could you try disabling USE-flags? You seem to have enabled all of them. So pick the ones you really need. A likely culprit is tbb. So start with -tbb and see if that already fixes the build error. (In reply to Paul Zander from comment #15) > Could you try disabling USE-flags? You seem to have enabled all of them. So > pick the ones you really need. > > A likely culprit is tbb. So start with -tbb and see if that already fixes > the build error. yes disabling tbb helped Can we close this? The TBB issue deserves a separate bug I think. seems similar to https://bugs.gentoo.org/908874 with USE="cuda tbb" same for sci-libs/vtk-9.2.6-r1, fails with USE="cuda tbb" The bug has been closed via the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=0167fc77186716079471d895b32b2283ca8b7fc4 commit 0167fc77186716079471d895b32b2283ca8b7fc4 Author: Paul Zander <negril.nx+gentoo@gmail.com> AuthorDate: 2023-07-06 14:26:47 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2024-01-10 12:41:01 +0000 sci-libs/vtk: add 9.3.0 Closes: https://bugs.gentoo.org/918898 Closes: https://bugs.gentoo.org/909520 Closes: https://bugs.gentoo.org/909380 Closes: https://bugs.gentoo.org/908874 Closes: https://bugs.gentoo.org/906264 Closes: https://bugs.gentoo.org/906075 Closes: https://bugs.gentoo.org/887505 Closes: https://bugs.gentoo.org/840776 Closes: https://bugs.gentoo.org/835659 Signed-off-by: Paul Zander <negril.nx+gentoo@gmail.com> Closes: https://github.com/gentoo/gentoo/pull/34432 Signed-off-by: Sam James <sam@gentoo.org> sci-libs/vtk/Manifest | 6 + sci-libs/vtk/files/vtk-9.2.5-pegtl-3.x.patch | 136 +++ ...ix-Java-error-with-integer-enum-overloads.patch | 51 ++ sci-libs/vtk/files/vtk-9.3.0-java.patch | 11 + sci-libs/vtk/files/vtk-9.3.0-opencascade.patch | 17 + sci-libs/vtk/files/vtk-9.3.0-pdal-2.6.0.patch | 77 ++ .../vtk-9.3.0-too-many-initializer-values.patch | 20 + sci-libs/vtk/metadata.xml | 1 + sci-libs/vtk/vtk-9.3.0.ebuild | 909 +++++++++++++++++++++ 9 files changed, 1228 insertions(+) |