search

cdcseacave / openMVS

open Multi-View Stereo reconstruction library
http://cdcseacave.github.io
GNU Affero General Public License v3.0
3.18k stars 891 forks source link

RTX 4060 TI - MapSMtoCores for SM 8.9 is undefined #1130

Open jsnetz opened 2 months ago

jsnetz commented 2 months ago

Describe the bug i can't build Openmvs for my new graphic card

To Reproduce I am an absolutely Newbie on cmake, VS 2019 and these things. also on Win 10.

  1. Installed all the dependencies , Cuda 12.4.1 Visual Studio 2019 as in build.md

  2. in cmake-gui changed the CMAKE_CUDA_ARCHITECTURES to 89 and start configure. The log shows a red "Compiling with C++20" and the red lined "... policy CMP0146" FINDCUDA module is removed. After generate and open project in VS 2019 in the Project settings is no change from compute_52 sm_52 to compute_89 sm_89. Build Debug and Release runs without any error but the MapSMtoCores 8.9 is undefined.

  3. I'm tested it with selected max cuda comatibility - same Problem

  4. set compute_89 sm_89 by hand in VS 2019 - same Problem

  5. in cmake i have 4 DIR-NOTFOUND: BREAKPAD, CUDA_SDK_ROOT_DIR, Tiff_DIR and Z_VCPKG_PWSH_PATH Breakpad is not installed, Tiff_DIR is set to C:\src\vcpkg\installed\x64-windows\share\tiff Z_VCPKG_PWSH_PATH is set to C:\Windows\System32\WindowsPowerShell\v1.0 CUDA_SDK_ROOT_DIR I'm don't know, set it for testing to C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4

  6. See error

Expected behavior A clear and concise description of what you expected to happen.

Screenshots If applicable, add screenshots to help explain your problem.

Desktop (please complete the following information):

Thanks

ElHouas commented 1 month ago

Hey @Newbie-omvs ,

I am facing the same problem. Have you been able to find a solution?

Thanks!

jsnetz commented 1 month ago

Hello ElHouas No i can‘t find a solution. But, when I‘m understanding the infos right, the only part using cuda is RefineMesh. All other modules using cpu. In this case the Question for me: is it nessesary for speeding up only on part or wasting time.

Grettings

cdcseacave commented 1 month ago

in fact the part that benefits most on using the GPU is the densification step feel free to use the docker image with GPU support if your system does not allow you to build with GPU support