PR CI: make cugraph builds depend on pylibcugraph builds

[jameslamb]

Proposes the following change to the wheel jobs for PR CI:

---
title: Current
---
flowchart LR
    A[build-pylibcugraph] --> B[test-pylibcugraph]
    B --> C[build-cugraph]
    C --> D[test-cugraph]

Loading

---
title: Proposed
---
flowchart LR
    A[build-pylibcugraph] --> B[test-pylibcugraph]
    A --> C[build-cugraph]
    C --> D[test-cugraph]

Loading

Notes for Reviewers

I think reducing the end-to-end time for changes here is even more important now that we have new downstream repos (https://github.com/rapidsai/nx-cugraph and https://github.com/rapidsai/cugraph-gnn) that depend on changes made here.

Benefits of this change

• shorter end-to-end time for CI runs
  • by parallelizing more work
  • by doing cugraph-cu{11,12} builds more frequently, which should mean more frequent population of the sccache cache
• faster feedback about cugraph-cu{11,12} build issues
• consistency with the rest of RAPIDS (every other RAPIDS project has builds depend on builds, not tests, as far as I know)

Costs of this change

• more at-one-moment load on GPU CI runners as a result of cugraph PRs (wheel tests for cugraph could now occupy 4 GPU runners at once instead of just 2)
  • pylibcugraph test jobs tend to take around 6-10 minutes once scheduled onto a runner, so this shouldn't be too bad

[bdice]

Nice catch. This feels like it was always the intended design, comparing to other RAPIDS repos.

[jameslamb]

All Python test jobs are failing because of this deprecation warning, treated as an error:

E   DeprecationWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.

full stack trace (click me)

ImportError while loading conftest '/__w/cugraph/cugraph/python/cugraph/cugraph/tests/conftest.py'.
tests/conftest.py:15: in <module>
    from cugraph.testing.mg_utils import (
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/__init__.py:14: in <module>
    from cugraph.community import (
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/community/__init__.py:14: in <module>
    from cugraph.community.louvain import louvain
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/community/louvain.py:15: in <module>
    from cugraph.structure import Graph
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/structure/__init__.py:14: in <module>
    from cugraph.structure.graph_classes import (
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/structure/graph_classes.py:15: in <module>
    from .graph_implementation import (
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/structure/graph_implementation/__init__.py:14: in <module>
    from .simpleGraph import simpleGraphImpl
/opt/conda/envs/test/lib/python3.12/site-packages/cugraph/structure/graph_implementation/simpleGraph.py:14: in <module>
    from cugraph.structure import graph_primtypes_wrapper
graph_primtypes_wrapper.pyx:1: in init cugraph.structure.graph_primtypes_wrapper
    ???
/opt/conda/envs/test/lib/python3.12/site-packages/pylibraft/common/__init__.py:19: in <module>
    from .device_ndarray import device_ndarray
/opt/conda/envs/test/lib/python3.12/site-packages/pylibraft/common/device_ndarray.py:18: in <module>
    import rmm
/opt/conda/envs/test/lib/python3.12/site-packages/rmm/__init__.py:17: in <module>
    from rmm import mr
/opt/conda/envs/test/lib/python3.12/site-packages/rmm/mr.py:14: in <module>
    from rmm.pylibrmm.memory_resource import (
/opt/conda/envs/test/lib/python3.12/site-packages/rmm/pylibrmm/__init__.py:15: in <module>
    from .device_buffer import DeviceBuffer
device_buffer.pyx:1: in init rmm.pylibrmm.device_buffer
    ???
memory_resource.pyx:31: in init rmm.pylibrmm.memory_resource
    ???
cuda/cudart.pyx:13: in init cuda.cudart
    ???
E   DeprecationWarning: The cuda.cudart module is deprecated and will be removed in a future release, please switch to use the cuda.bindings.runtime module instead.

(build link)

Just pushed 1a2590a ignoring that warning, as we've done in some other places.

If we accept that workaround for now, I'll put up an issue here about resolving those warnings so they don't become errors when those deprecations are eventually enforced in future cuda-python versions.

[jameslamb]

/merge

[alexbarghi-nv]

👍