[WIP] Windows compilation #1298
Draft
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
karthikeyann
changed the title
|
Note that RMM is structurally incompatible with Windows as a header-only library as the functionality of RMM would need to build a DLL that holds the symbols relevant for |
harrism
added
feature request
|
Do we have Windows CI? |
Not yet. We're working on it as CCCL needs it, but it's still many months away. |
Replaces rapidsai#1308 Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Mark Harris (https://github.com/harrism) - AJ Schmidt (https://github.com/ajschmidt8)
Branch 23.10 merge 23.08
This PR migrates RMM to use `fetch_rapids.cmake` like most RAPIDS repos. This makes it easier to define a single source if the upstream branch of rapids-cmake needs to change for testing, like in rapidsai#1247. Authors: - Bradley Dice (https://github.com/bdice) Approvers: - Ray Douglass (https://github.com/raydouglass) - Robert Maynard (https://github.com/robertmaynard) URL: rapidsai#1319
This PR contains the minimal set of changes to compile using Cython 3 without warnings. Future PRs can be made to take advantage of new or improved features. The specific changes are: - Ensuring `nogil` always comes after `except`. `except * nogil` is a compile-time error in Cython 3 - Adding `noexcept` or `except *` to any `cdef ` functions missing them. In Cython 0.29 these would default to `noexcept`, which meant that exceptions would not be properly propagated. In Cython 3.0.0, these default to `except *`, which incurs a performance penalty for reacquiring the GIL to check the exception value even for `nogil` functions. Being explicit here is important. There are a large number of outstanding warnings due to NVIDIA/cuda-python#44. cuda-python for CUDA 12 has the necessary fix, but we will need a cuda-python 11.8.* bugfix with a backport to make those warnings go away. Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Ashwin Srinath (https://github.com/shwina) - Lawrence Mitchell (https://github.com/wence-) - Ray Douglass (https://github.com/raydouglass) URL: rapidsai#1313
This will allow us to more easily audit the quality of our docs in CI going forward. Authors: - Vyas Ramasubramani (https://github.com/vyasr) Approvers: - Bradley Dice (https://github.com/bdice) - Lawrence Mitchell (https://github.com/wence-) - AJ Schmidt (https://github.com/ajschmidt8) URL: rapidsai#1316
Forward-merge branch-23.08 to branch-23.10
The latest versions `4.1` and `4.0` of `sphinxcontrib-jquery` have an issue with jQuery installed currently in sphinx. This is causing the search functionality to fail so I've pinned the version to `3.0.0`, this resolves the issue. Authors: - Jake Awe (https://github.com/AyodeAwe) Approvers: - https://github.com/jakirkham - AJ Schmidt (https://github.com/ajschmidt8)
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
The changes in this PR allow windows compilation to succeed for librmm and their tests to pass (some fail TBD).
Windows Setup:
cl.exe Microsoft (R) C/C++ Optimizing Compiler Version 19.36.32535 for x64
Environment Setup:
conda env create --name rmm_dev --file conda/environments/all_cuda-120_arch-x86_64.yamlvcpkg install pthread:x64-windowsconda install -c conda-forge dlfcn-win32Compilation:
conda activate rmm_devmkdir build; cd build,mkdir ..\..\installcmake .. -G "Visual Studio 17 2022" -DCMAKE_INSTALL_PREFIX=..\..\install -DCMAKE_TOOLCHAIN_FILE=C:/Users/karthikeyann/Desktop/vcpkg/scripts/buildsystems/vcpkg.cmakecmake --build .Changes after cmake configure attempt:
rapids-cmake comment out compilation code in
C:\Users\karthikeyann\Desktop\rmm\build\_deps\rapids-cmake-src\rapids-cmake\test\generate_resource_spec.cmakeand update the cuda compute capability minor, major version in same file.
Renamed
C:\Users\karthikeyann\Desktop\rmm\build\_deps\thrust-src\cubsymlink to cub2.msvc does not support symlink within pre-processor.
so, copied folder
C:\Users\karthikeyann\Desktop\rmm\build\_deps\thrust-src\dependencies\cub\cubto
C:\Users\karthikeyann\Desktop\rmm\build\_deps\thrust-src\cubList of code changes:
All
*/CUDA_Asynctests are skipped, as this driver version is unsupported (TBD).Checklist