Skip to content

[nvtx3] Create a new port with 3.2.2#46800

Closed
luncliff wants to merge 2 commits intomicrosoft:masterfrom
luncliff:port/nvtx3
Closed

[nvtx3] Create a new port with 3.2.2#46800
luncliff wants to merge 2 commits intomicrosoft:masterfrom
luncliff:port/nvtx3

Conversation

@luncliff
Copy link
Contributor

@luncliff luncliff commented Aug 7, 2025

Create a new port from https://github.com/NVIDIA/NVTX/releases/tag/v3.2.2.

It's a header-only library, and the recent version supports find_package(nvtx3)

The following packages are already installed:
    nvtx3:[email protected]
Total install time: 546 us
nvtx3 provides CMake targets:

  # this is heuristically generated, and may not be correct
  find_package(nvtx3 CONFIG REQUIRED)
  target_link_libraries(main PRIVATE nvtx3::nvtx3-c nvtx3::nvtx3-cpp)

Target CUDA::nvToolsExt and CUDA::nvtx3

Some times, CUDA configuration may fail with creating nvToolsExt target.
Using the port will prevent the header search failures.

CMake Error at cmake/public/cuda.cmake:186 (set_property):
  The link interface of target "torch::nvtoolsext" contains:

    CUDA::nvToolsExt

References

Checklist

  • Changes comply with the maintainer guide.
  • The name of the port matches an existing name for this component on https://repology.org/ if possible, and/or is strongly associated with that component on search engines.
  • Optional dependencies are resolved in exactly one way. For example, if the component is built with CMake, all find_package calls are REQUIRED, are satisfied by vcpkg.json's declared dependencies, or disabled with CMAKE_DISABLE_FIND_PACKAGE_Xxx.
  • The versioning scheme in vcpkg.json matches what upstream says.
  • The license declaration in vcpkg.json matches what upstream says.
  • The installed as the "copyright" file matches what upstream says.
  • The source code of the component installed comes from an authoritative source.
  • The generated "usage text" is accurate. See adding-usage for context.
  • The version database is fixed by rerunning ./vcpkg x-add-version --all and committing the result.
  • Only one version is in the new port's versions file.
  • Only one version is added to each modified port's versions file.

@luncliff luncliff mentioned this pull request Aug 7, 2025
10 tasks
@BillyONeal
Copy link
Member

Related: #46694

We were not 100% positive if the right thing is to get this from the CUDA installer or try to build it separately. Do you know if there are statement(s) from NVIDIA recommending one or the other?

@BillyONeal
Copy link
Member

I will check the guide documents. Honestly, I haven't read them for years, so it may take some days. 🤯

I mean, if you don't have something off the top of your head we're going to assume that the presence of this thing in the CUDA installer suggests that's where NVIDIA wants people to get it. (That was the resolution of the situation in #46694)

@BillyONeal
Copy link
Member

As of #46911 NVTX should be on the machines.

@luncliff
Copy link
Contributor Author

Thank you. I will rebase the #46700

@luncliff luncliff closed this Sep 11, 2025
@luncliff luncliff deleted the port/nvtx3 branch December 3, 2025 16:39
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants