Build and test with CUDA 13.0.0 by jameslamb · Pull Request #5236 · rapidsai/cugraph

jameslamb · 2025-08-22T04:45:49Z

Contributes to rapidsai/build-planning#208

uses CUDA 13.0.0 to build and test
adds CUDA 13 devcontainers
moves some dependency pins:
- cuda-python: >=12.9.2 (CUDA 12), >=13.0.1 (CUDA 13)
- cupy: >=13.6.0

Contributes to rapidsai/build-planning#68

updates to CUDA 13 dependencies in fallback entries in dependencies.yaml matrices (i.e., the ones that get written to pyproject.toml in source control)

Notes for Reviewers

This switches GitHub Actions workflows to the cuda13.0 branch from here: rapidsai/shared-workflows#413

A future round of PRs will revert that back to branch-25.10, once all of RAPIDS supports CUDA 13.

This project no longer has any dependency on `cugraph-pyg` or `pylibwholegraph`. This removes some lingering configuration for them in `dependencies.yaml` and `update-version.sh`. ## Notes for Reviewers Discovered while working on #5236. Getting this merged soon would make the diff there a bit smaller. Found like this: ```shell git grep -i pyg git grep -i wholegraph ``` Authors: - James Lamb (https://github.com/jameslamb) Approvers: - Brad Rees (https://github.com/BradReesWork) - Kyle Edwards (https://github.com/KyleFromNVIDIA) - Bradley Dice (https://github.com/bdice) URL: #5237

jameslamb · 2025-08-25T14:36:55Z

/ok to test

jameslamb · 2025-08-25T17:53:26Z

Wheel builds all succeeded on the most recent run... good evidence that there aren't any significant C++ issues that'll prevent us from building against CUDA 13 here.

https://github.com/rapidsai/cugraph/actions/runs/17212019556/job/48826027332?pr=5236

conda C++ builds are failing because libcugraph_etl (which we don't build wheels for) needs libucdf. Most other jobs are also failing because we need cudf / libcudf / pylibcudf packages.

Will update this once rapidsai/cudf#19768 is done.

jameslamb · 2025-08-28T17:05:36Z

/ok to test

jameslamb · 2025-08-28T20:59:46Z

Everything for wheels is looking good! But libcugraph_etl is not compiling on CUDA 13.

Problem 1: `libcugraph_etl` fails to compile on CUDA 13

 │ │ sccache $BUILD_PREFIX/bin/nvcc -forward-unknown-to-host-compiler -DCCCL_DISABLE_PDL -DCUB_DISABLE_NAMESPACE_MAGIC -DCUB_IGNORE_NAMESPACE_MAGIC_ERROR -DCUDA_API_PER_THREAD_DEFAULT_STREAM -DCUTLASS_NAMESPACE=raft_cutlass -DLIBCUDACXX_ENABLE_EXPERIMENTAL_MEMORY_RESOURCE -DRAFT_LOG_ACTIVE_LEVEL=RAPIDS_LOGGER_LOG_LEVEL_INFO -DRAFT_SYSTEM_LITTLE_ENDIAN=1 -DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_CUDA -DTHRUST_DISABLE_ABI_NAMESPACE -DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -DTHRUST_IGNORE_ABI_NAMESPACE_ERROR -Dcugraph_etl_EXPORTS -I$SRC_DIR/cpp/libcugraph_etl/src -I$SRC_DIR/cpp/libcugraph_etl/include -I$SRC_DIR/cpp/build/_deps/cccl-src/lib/cmake/thrust/../../../thrust -I$SRC_DIR/cpp/build/_deps/cccl-src/lib/cmake/libcudacxx/../../../libcudacxx/include -I$SRC_DIR/cpp/build/_deps/cccl-src/lib/cmake/cub/../../../cub -isystem $BUILD_PREFIX/targets/x86_64-linux/include -isystem $BUILD_PREFIX/targets/x86_64-linux/include/cccl -isystem $SRC_DIR/cpp/include -isystem $PREFIX/include -isystem $SRC_DIR/cpp/build/_deps/cuco-src/include -O3 -DNDEBUG -std=c++17 "--generate-code=arch=compute_75,code=[sm_75]" "--generate-code=arch=compute_80,code=[sm_80]" "--generate-code=arch=compute_86,code=[sm_86]" "--generate-code=arch=compute_90a,code=[sm_90a]" "--generate-code=arch=compute_100f,code=[sm_100f]" "--generate-code=arch=compute_120a,code=[sm_120a]" "--generate-code=arch=compute_120,code=[compute_120,sm_120]" -Xcompiler=-fPIC --expt-extended-lambda --expt-relaxed-constexpr -Werror=cross-execution-space-call -Wno-deprecated-declarations -Xptxas=--disable-warnings -Xcompiler=-Wall,-Wno-error=sign-compare,-Wno-error=unused-but-set-variable -Xfatbin=-compress-all --compress-mode=size -MD -MT CMakeFiles/cugraph_etl.dir/src/renumbering.cu.o -MF CMakeFiles/cugraph_etl.dir/src/renumbering.cu.o.d -x cu -c $SRC_DIR/cpp/libcugraph_etl/src/renumbering.cu -o CMakeFiles/cugraph_etl.dir/src/renumbering.cu.o
 │ │ $PREFIX/include/cudf/utilities/span.hpp(82): warning #20012-D: __host__ annotation is ignored on a non-virtual function("span_base") that is explicitly defaulted on its first declaration
 │ │    __attribute__((host)) __attribute__((device)) 
...
 │ │ $SRC_DIR/cpp/libcugraph_etl/include/hash/concurrent_unordered_map.cuh(548): error: no suitable constructor exists to convert from "int" to "cudaMemLocation"
 │ │          cudaMemPrefetchAsync_ptsz( m_hashtbl_values, m_capacity * sizeof(value_type), dev_id, stream.value())
...
 │ │           detected during instantiation of "concurrent_unordered_map<Key, Element, Hasher, Equality, Allocator>::concurrent_unordered_map(concurrent_unordered_map<Key, Element, Hasher, Equality, Allocator>::size_type, concurrent_unordered_map<Key, Element, Hasher, Equality, Allocator>::mapped_type, concurrent_unordered_map<Key, Element, Hasher, Equality, Allocator>::key_type, const Hasher &, const Equality &, const concurrent_unordered_map<Key, Element, Hasher, Equality, Allocator>::allocator_type &, rmm::cuda_stream_view) [with Key=uint32_t, Element=cugraph::etl::str_hash_value, Hasher=cudf::hashing::detail::MurmurHash3_x86_32<uint32_t>, Equality=equal_to<uint32_t>, Allocator=rmm::mr::polymorphic_allocator<cuda::std::__4::pair<uint32_t, cugraph::etl::str_hash_value>>]" at line 802 of $SRC_DIR/cpp/libcugraph_etl/src/renumbering.cu
 │ │ $SRC_DIR/cpp/libcugraph_etl/include/hash/concurrent_unordered_map.cuh(548): error: argument of type "cudaStream_t" (aka "CUstream_st *") is incompatible with parameter of type "unsigned int"
 │ │          cudaMemPrefetchAsync_ptsz( m_hashtbl_values, m_capacity * sizeof(value_type), dev_id, stream.value())

(conda-cpp-build link)

robertmaynard · 2025-08-28T23:38:00Z

/okay to test

…nces

robertmaynard · 2025-08-28T23:50:48Z

/okay to test

dependencies.yaml

nv-rliu

Thanks

jameslamb · 2025-08-29T18:45:15Z

Thanks everyone!!!

jameslamb · 2025-08-29T18:45:26Z

/merge

Follow-up to #5236 Starting with rapidsai/dask-cuda#1536, `dask-cuda` wheels now have extras like `[cu12]` and `[cu13]` to ensure a consistent set of CUDA-major-version-specific dependencies are installed. This proposes using those extras in this project's wheel dependencies on `dask-cuda`. Authors: - James Lamb (https://github.com/jameslamb) Approvers: - Brad Rees (https://github.com/BradReesWork) - Jake Awe (https://github.com/AyodeAwe) URL: #5243

Follow-up to #5236 There, I intended to introduce a dependency on `torch` nightly wheels for CUDA 13, but made a typo in the `dependencies.yaml` filter. This fixes that, so that wheels for the correct `torch` variant + version will be installed in wheel-testing CI jobs here. Authors: - James Lamb (https://github.com/jameslamb) Approvers: - Gil Forsyth (https://github.com/gforsyth) URL: #5244

Contributes to rapidsai/build-planning#208 Follow-up to #5236 Similar to rapidsai/cugraph-gnn#355 PyTorch now has official releases with CUDA 13 support. This switches from nightlies to those releases for wheels in CI here. Authors: - James Lamb (https://github.com/jameslamb) Approvers: - Bradley Dice (https://github.com/bdice) URL: #5356

Build and test with CUDA 13.0.0

055ca2a

jameslamb added non-breaking Non-breaking change improvement Improvement / enhancement to an existing function labels Aug 22, 2025

This comment was marked as resolved.

Sign in to view

This was referenced Aug 22, 2025

Add CUDA 13.0 Support across all of RAPIDS rapidsai/build-planning#208

Closed

Build and test with CUDA 13.0.0 rapidsai/nx-cugraph#182

Merged

update dependencies.yaml and docs

76b313d

This was referenced Aug 22, 2025

remove cugraph-pyg and wholegraph configuration #5237

Merged

Build and test with CUDA 13.0.0 rapidsai/cuxfilter#718

Merged

feature request: detect unused dependency lists rapidsai/dependency-file-generator#162

Closed

merge branch-25.10

9326993

This was referenced Aug 26, 2025

Build and test with CUDA 13.0.0 rapidsai/cugraph-gnn#286

Merged

build and test with CUDA 13.0.0 rapidsai/cugraph-docs#154

Merged

Build and test with CUDA 13.0.0 rapidsai/integration#795

Merged

update pytorch

97773bd

Use rmm prefetc API so we don't have to handle CUDA 12/13 API differe…

0c8b5d6

…nces

jameslamb changed the title ~~WIP: Build and test with CUDA 13.0.0~~ Build and test with CUDA 13.0.0 Aug 29, 2025

jameslamb requested review from alexbarghi-nv and nv-rliu August 29, 2025 03:20

jameslamb marked this pull request as ready for review August 29, 2025 03:20

jameslamb requested review from a team as code owners August 29, 2025 03:20

jameslamb requested a review from bdice August 29, 2025 03:20

jameslamb commented Aug 29, 2025

View reviewed changes

dependencies.yaml Show resolved Hide resolved

ChuckHastings approved these changes Aug 29, 2025

View reviewed changes

nv-rliu approved these changes Aug 29, 2025

View reviewed changes

vyasr approved these changes Aug 29, 2025

View reviewed changes

rapids-bot bot merged commit 9874c70 into rapidsai:branch-25.10 Aug 29, 2025
196 of 198 checks passed

jameslamb deleted the cuda-13.0.0 branch August 29, 2025 18:45

jameslamb mentioned this pull request Aug 29, 2025

use dask-cuda[cu12, cu13] extras for wheel dependencies #5243

Merged

jameslamb mentioned this pull request Sep 2, 2025

fix pytorch requirement for CUDA 13 builds #5244

Merged

jameslamb mentioned this pull request Nov 26, 2025

CUDA 13: use releases, not nightlies, for PyTorch wheels #5356

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Build and test with CUDA 13.0.0#5236

Build and test with CUDA 13.0.0#5236
rapids-bot[bot] merged 5 commits intorapidsai:branch-25.10from
jameslamb:cuda-13.0.0

jameslamb commented Aug 22, 2025 •

edited

Loading

Uh oh!

This comment was marked as resolved.

jameslamb commented Aug 25, 2025

Uh oh!

jameslamb commented Aug 25, 2025

Uh oh!

jameslamb commented Aug 28, 2025

Uh oh!

jameslamb commented Aug 28, 2025

Uh oh!

robertmaynard commented Aug 28, 2025

Uh oh!

robertmaynard commented Aug 28, 2025

Uh oh!

Uh oh!

nv-rliu left a comment

Uh oh!

jameslamb commented Aug 29, 2025

Uh oh!

jameslamb commented Aug 29, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Conversation

jameslamb commented Aug 22, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Notes for Reviewers

Uh oh!

This comment was marked as resolved.

jameslamb commented Aug 25, 2025

Uh oh!

jameslamb commented Aug 25, 2025

Uh oh!

jameslamb commented Aug 28, 2025

Uh oh!

jameslamb commented Aug 28, 2025

Problem 1: libcugraph_etl fails to compile on CUDA 13

Uh oh!

robertmaynard commented Aug 28, 2025

Uh oh!

robertmaynard commented Aug 28, 2025

Uh oh!

Uh oh!

nv-rliu left a comment

Choose a reason for hiding this comment

Uh oh!

jameslamb commented Aug 29, 2025

Uh oh!

jameslamb commented Aug 29, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

jameslamb commented Aug 22, 2025 •

edited

Loading

Problem 1: `libcugraph_etl` fails to compile on CUDA 13