CUDA Blackwell (sm_100/sm_120) & CUDA 13 Support [bug]

Open Beginner friendly
#245 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
84/100
Issue type
Bug
Clarity
Clearly specified
Activity status
Active
Tech stack
cmake, cpp
Domain
build-system

Research direction

Start in cmake/ChooseCudaCC.cmake, where CUDA_MAX_CC is capped at 90 and the CUDA 12.1 architecture list is documented. Review the CMake configuration path for selecting compute capabilities, then configure CCTag with CUDA 13 targeting sm_100 or sm_120. Done means configuration no longer caps these Blackwell targets and accepts the newer CUDA toolchain.

Written by the indexing model from the issue text.

Description

Describe the bug
CMake configuration fails to support CUDA 13 and newer NVIDIA Blackwell architecture targets (sm_100 / sm_120). cmake/ChooseCudaCC.cmake currently caps CUDA_MAX_CC at 90 (Ada Lovelace / Hopper), preventing proper architecture targeting for newer GPUs.

To Reproduce
Steps to reproduce the behavior:

  1. Attempt to build CCTag on a CUDA 13 toolchain targeting a Blackwell GPU (sm_120).
  2. Run CMake configure step.
  3. CMake caps CUDA compute capabilities at CC 90 because CUDA_MAX_CC is hardcoded to 90.

Expected behavior
CMake should recognize CUDA 13 and support setting compute capabilities for Blackwell architectures (sm_100 and sm_120).

Screenshots
N/A

Log
N/A

Desktop (please complete the following and other pertinent information):

  • OS: Arch Linux (rolling), kernel 6.18.45-2-lts, x86_64
  • Cuda version: 13.3.1 (nvcc release 13.3, V13.3.73); driver 610.57.04; GPU RTX 5060 Ti, compute capability 12.0 (sm_120)
  • Other dependencies version: OpenCV 5.0.0, Boost 1.92.0, Eigen 5.0.1, GCC 16.2.1, CMake 4.4.2
  • CCTag version: release version v1.0.4
    • Binary version: 1.0.4
    • Commit reference: N/A (built from v1.0.4 source tarball, sha256 f4ced6138419f4ad09d62d4373738c172586c91068faff8bf08c2caa5e76d516)

Additional context
In cmake/ChooseCudaCC.cmake, a comment indicates the architecture list was last updated for CUDA 12.1. Needs updating for the CUDA 13 branch and corresponding compute capabilities.

Dominant language
C++
Stars
417
Forks
101
PR merge metrics
No merged PRs in 30d

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from alicevision/CCTag

All issues in alicevision/CCTag

Similar issues

More C++ issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.