You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
cuda.core: carry the cuda-bindings floors through minimal-sysctk12/13 extras - #3460
Moves the cuda-bindings floors from the cu12/cu13 extras into new minimal-sysctk12/minimal-sysctk13 extras, which cu12/cu13 include by self-reference. A package that depends on cuda-core[minimal-sysctk12] (or [cu12]) now passes the floor to pip and uv.
The floor readers (floors_from_extras and the CI wheel-floor reader) still read the floor of each cu<N> extra; they now follow a cuda-core[...] self-reference and skip an unversioned cuda-bindings[all]. The existing tests of the cu<N> form are unchanged, and new tests cover the self-reference.
Two additions beyond the issue:
The check-cuda-core-bindings-floor hook also checks that each cu<N> extra lists cuda-core[minimal-sysctk<N>], so the floor is written only in minimal-sysctk<N>.
[tool.uv].conflicts declares the new extras as conflicting across majors; without it uv lock fails.
A CUDA 12 wheel built from this branch carries the floors under extra == "minimal-sysctk12"/"minimal-sysctk13", and pip install --dry-run "cuda-core[cu12] @ <wheel>" cuda-bindings==12.9.3 (likewise [minimal-sysctk12]) is now rejected at resolve time instead of failing at import.
The floors lived only in the cu12/cu13 extras, so a package that depends
on plain cuda-core never passed them to pip or uv, and an older
cuda-bindings failed only at import (NVIDIA#3459).
Declare the floors in new minimal-sysctk12/13 extras, include them from
cu12/cu13, and depend on cuda-bindings without a version. The floor
parser, the CI wheel-floor reader, the pre-commit hook and the docs read
the new extras.
Drop the unversioned cuda-bindings from the base dependencies, so a
plain cuda-core install keeps asking for an extra.
Keep the floor readers on the cu<N> extras, as before, and teach them
to follow a cuda-core[...] self-reference and to skip an unversioned
cuda-bindings[all]. The cu<N> layout then yields the floor written in
minimal-sysctk<N>, and the existing tests of the cu<N> form stay as
they are. New tests cover the self-reference.
The reason will be displayed to describe this comment to others. Learn more.
Not reviewing either way (I think @Andy-Jost should probably weigh in) -- just sharing one note my agent found:
Stale wording in build_hooks.py. The docstrings near lines 136 and 256 still say the floor is "the cu extra in pyproject.toml". The issue listed these as optional, so the code works unchanged.
No actionable comments were generated in the recent review.
Summary
Summary by CodeRabbit
New Features
Added minimal-sysctk12 and minimal-sysctk13 installation options for using CUDA bindings with a system-installed toolkit. The cu12 and cu13 options now include the corresponding bindings version requirements alongside CUDA Toolkit wheels.
Documentation
Updated installation and support guidance on selecting compatible bindings, including for packages that depend on cuda-core, and clarified that installing cuda-core without a CUDA extra does not communicate the bindings version requirement to package installers.
Added guidance about the resolver fallback risk associated with the yanked cuda-core 1.3.0 release.
The PR moves CUDA binding version floors into minimal-sysctk12 and minimal-sysctk13 extras. The CUDA extras include those floor declarations. Floor readers, consistency checks, installation instructions, and maintainer guidance now reflect this structure.
Changes: CUDA binding floor extras
Layer / File(s)
Summary
Declare extras and resolve floors cuda_core/pyproject.toml, cuda_core/cuda/core/_bindings_floor.py, cuda_core/tests/test_bindings_floor.py
The minimal extras declare the CUDA binding floors, and cu12 and cu13 include their matching minimal extras. Floor resolution follows self-references and tests cover cycles and invalid requirements.
Update floor readers and consistency checks ci/tools/cuda_core_bindings_floor.py, ci/tools/tests/test_cuda_core_bindings_floor.py, toolshed/check_cuda_core_bindings_floor.py, cuda_core/tests/test_bindings_floor.py
The CI metadata reader and repository checker follow and validate extra references. Tests cover floor extraction and missing or invalid declarations.
Document installation and floor maintenance cuda_core/AGENTS.md, cuda_core/DESCRIPTION.rst, cuda_core/docs/source/install.rst, cuda_core/docs/source/release/1.3.1-notes.rst, cuda_core/docs/source/support.rst
The documentation describes the minimal extras, their relationship to the CUDA extras, and the floor limitations of plain cuda-core requirements.
The change moves the CUDA binding floors into new minimal-sysctk12 and minimal-sysctk13 extras. The install guidance may still let pip resolve an older cuda-core that has neither the extra nor the floor, so import fails. The release notes also claim a yank that was not visible on PyPI. Fix the guidance and confirm the yank before merging.
Security architecture risk: Minimal · up to 22349
The inspected change preserves the existing binding-version floors and major-version bounds while making them available through minimal installation extras. No introduced or worsened security, failure-containment, rollback, or ownership risk was established.
Retained concerns
No architecture-level concerns identified.
Security review details
Security Blast Radius
inferred — The supported exposure is dependency selection for consumers requesting these extras and floor interpretation during project validation or CI wheel inspection. The inspected self-reference traversal remains within the supplied configuration or metadata rather than fetching or executing referenced requirements.
Trust Boundaries and Controls
observed — The project reader still rejects missing or duplicate versioned binding requirements, malformed floor constraints, and mismatched CUDA-major bounds. The consistency hook additionally requires each cu extra to name its matching minimal extra.
Resilience and Maintainability Implications
observed — Both recursive readers track visited extras along each traversal path, preventing cycles from recursing indefinitely. Their traversal state is local to the invocation; the inspected expansion logic does not update shared persistent state.
- [ ] Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts
Reviewing files that changed from the base of the PR and between c41e079 and 223499a.
Files selected for processing: 4
ci/tools/cuda_core_bindings_floor.py
cuda_core/cuda/core/_bindings_floor.py
cuda_core/docs/source/release/1.3.1-notes.rst
cuda_core/tests/test_bindings_floor.py
Files skipped from review as they are similar to previous changes: 1
cuda_core/docs/source/release/1.3.1-notes.rst
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 11 remain after this review.
suggestion: Remove the loud-validation-failure test.
This case checks that malformed metadata raises SystemExit. It does not cover a silently incorrect floor. Remove this case and retain the successful floor-extraction coverage.
As per coding guidelines, “Keep a test only when wrong behavior would be silent.”
Reviewing files that changed from the base of the PR and between 7de298b and c41e079.
Files selected for processing: 11
ci/tools/cuda_core_bindings_floor.py
ci/tools/tests/test_cuda_core_bindings_floor.py
cuda_core/AGENTS.md
cuda_core/DESCRIPTION.rst
cuda_core/cuda/core/_bindings_floor.py
cuda_core/docs/source/install.rst
cuda_core/docs/source/release/1.3.1-notes.rst
cuda_core/docs/source/support.rst
cuda_core/pyproject.toml
cuda_core/tests/test_bindings_floor.py
toolshed/check_cuda_core_bindings_floor.py
Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 10 remain after this review.
The reason will be displayed to describe this comment to others. Learn more.
Major · Data Integrity & Integration · Quick win
important: Require a version that provides the minimal extras.
With cuda-bindings==12.9.3, pip can select cuda-core 1.3.0 instead of 1.3.1. That release lacks minimal-sysctk12, so pip warns but installs it without the floor; import then fails. (raw.githubusercontent.com)
Recommend cuda-core[minimal-sysctk12]>=1.3.1 and the equivalent CUDA 13 requirement.
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
important: Confirm the yank before publishing this statement.
As of November 17, 2026, PyPI lists 1.3.0 without a yank notice. (pypi.org) Readers could assume resolvers will avoid that release. Yank it before publishing these notes, or update the statement.
The reason will be displayed to describe this comment to others. Learn more.
Minor · Functional Correctness · Quick win
important: Detect missing CUDA extras independently of parsed floors.
If cu13 is deleted while cu12 and both minimal extras remain, read_floors() returns only major 12. This loop never checks cu13, so check() can pass despite the missing CUDA 13 extra.
Derive the expected majors from the minimal extras, then validate each matching cu<N> extra. Add coverage through check(). The current None test injects major 13 directly and misses this failure.
… extras (NVIDIA#3460)
* cuda.core: carry the cuda-bindings floors through minimal-sysctk extras
The floors lived only in the cu12/cu13 extras, so a package that depends
on plain cuda-core never passed them to pip or uv, and an older
cuda-bindings failed only at import (NVIDIA#3459).
Declare the floors in new minimal-sysctk12/13 extras, include them from
cu12/cu13, and depend on cuda-bindings without a version. The floor
parser, the CI wheel-floor reader, the pre-commit hook and the docs read
the new extras.
* cuda.core: read cu<N> floors through the minimal-sysctk self-reference
Drop the unversioned cuda-bindings from the base dependencies, so a
plain cuda-core install keeps asking for an extra.
Keep the floor readers on the cu<N> extras, as before, and teach them
to follow a cuda-core[...] self-reference and to skip an unversioned
cuda-bindings[all]. The cu<N> layout then yields the floor written in
minimal-sysctk<N>, and the existing tests of the cu<N> form stay as
they are. New tests cover the self-reference.
---------
Co-authored-by: Andy Jost <ajost@nvidia.com>
(cherry picked from commit ad38141)
* [doc-only] docs(cuda.core): fix the floor-move wording; advise against pinning cuda-bindings (#3441)
* docs(cuda.core): drop the Breaking Changes section name from the floor policy
The release notes no longer have a Breaking Changes section for a floor
move; 1.3.0 lists it under Highlights. Say only that the release notes
list every move.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(cuda.core): advise against pinning cuda-bindings next to cuda-core
Users who pin cuda-bindings themselves can conflict with the floor of a
later cuda-core release. Tell them to depend on cuda-core with the cu12
or cu13 extra and let cuda-core choose the cuda-bindings version.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 83a3422)
* Mark cuda-gdb source-line test thread_unsafe (#3433)
Signed-off-by: Ralf Juengling <rjuengling@nvidia.com>
(cherry picked from commit 1a22eec)
* fix(cuda.core): serialize pinned-memory kernel-access test on Windows (#3436)
* fix(cuda.core): serialize pinned-memory kernel-access test on Windows
closes#3426
Add @thread_unsafe_on_windows to the pinned-memory kernel-access test.
The test accesses host memory while parallel workers may execute GPU
work; synchronizing its own stream does not coordinate those workers.
This follows the guard used by similar tests on Windows and WSL.
Windows CI is needed to confirm it resolves the reported crashes.
The separate OOM failures remain under investigation.
Validation: Ruff and git diff --check pass.
* fix(cuda.core): report CUDA synchronization errors during test teardown
Check the cuCtxSynchronize result in init_cuda so asynchronous errors
surface during the triggering test's teardown instead of going unnoticed
until a later test. Always attempt context cleanup via finally.
Related to #3426.
* fix(cuda.core): synchronize the bound context during test teardown
Use device.sync() because the test wrapper may pop the current context
before init_cuda teardown. This avoids spurious INVALID_CONTEXT errors
while preserving synchronization error reporting and context cleanup.
Add regression coverage for a missing current context and cleanup after
synchronization failures.
Related to #3426.
* fix(cuda.core): serialize memory-pool watermark reset tests
Mark test_mempool_watermark_reset thread_unsafe, matching test_mempool_trim.
Both exercise the pool teardown and synchronization sequence associated
with the suspected driver race in #3350.
* fix(cuda.core): bypass access revocation when destroying host pools
Skip peer-access revocation for locally created HOST and HOST_NUMA pools
to investigate pinned-pool teardown OOMs and crashes. Continue destroying
the pools and preserve cleanup for device, managed, and imported pools.
This is a diagnostic workaround pending Windows CI validation.
Related to #3426.
* fix(cuda.core): preserve the first exception during fixture cleanup
Keep setup or synchronization failures primary and attach subsequent
cleanup failures as exception notes. Always attempt to pop the context.
Add regression coverage for combined failures, including interrupts.
* fix(cuda.core): limit host-pool cleanup workaround to pinned allocations
HOST and HOST_NUMA also describe managed-pool preferences. Check the allocation type so the pinned-memory workaround preserves managed-pool peer-access revocation.
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* test(cuda.core): expose pinned allocation OOM and retain its traceback
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* ci: diagnose PR3436 pinned pool teardown on Windows
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* ci: move watermark investigation out of PR3436
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* test(cuda.core): restore expected pinned allocation OOM handling
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* revert(cuda.core): defer pinned host pool cleanup workaround
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
---------
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
(cherry picked from commit b715078)
* [no-ci] ci: exclude unavailable historical GitHub references (#3471)
* ci: exclude unavailable historical GitHub references
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* ci: require manual copy-pr-bot approval
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
---------
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
(cherry picked from commit 3991a69)
* cuda.core: map ml_dtypes bfloat16 to kDLBfloat in DLPack export (#3359)
_smv_dtype_numpy_to_dlpack dispatched on numpy dtype kind, but
ml_dtypes.bfloat16 has kind "V", so bf16 views fell through to the
unsupported-dtype BufferError and could not be re-exported, even though
the DLPack import path already accepts kDLBfloat and _tensor_map.pyx
maps bfloat16. Detect bfloat16 up front and export it as kDLBfloat
(16 bits).
The new round-trip test goes through StridedMemoryView.from_dlpack so
it cannot skip on stacks where NumPy's own DLPack support for extended
dtypes is unavailable, unlike the existing parametrized round-trip
test's NumPy-consumer probe.
Signed-off-by: Sainath-Reddy7 <180950590+Sainath-Reddy7@users.noreply.github.com>
Co-authored-by: Sainath-Reddy7 <180950590+Sainath-Reddy7@users.noreply.github.com>
Co-authored-by: Sebastian Berg <sebastianb@nvidia.com>
(cherry picked from commit d1d1457)
* fix(cuda.core): make IPC imports wait for a pending release of the same descriptor (#3450)
* fix(cuda.core): make IPC imports wait for a pending release of the same descriptor
deviceptr_import_ipc() shares one handle per descriptor because the
driver's first free of an imported pointer unmaps every import of it
(nvbug 5570902). The handle's deleter took ipc_import_mutex only after
the last reference was gone. In that gap an import found the cache entry
expired, re-imported the still-mapped pointer and registered a new entry.
The deleter then erased that new entry and freed the pointer, which
unmapped it under the new Buffer.
The cache entry now outlives the last reference and only its own deleter
erases it, under the mutex, together with the free. An import that finds
an expired entry waits on a condition variable until that release is
done, then imports fresh.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* Use CTAD for the ipc_import_mutex locks
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 2f1185e)
* test(cuda.core): skip HOST graph allocation before CUDA 13 (#3446)
(cherry picked from commit 11cc666)
* test(cuda.core): query green stream contexts with v2 (#3448)
Signed-off-by: Rui Luo <ruluo@nvidia.com>
(cherry picked from commit 7076a67)
* test(cuda.core): skip the VMM RDMA policy test on WSL (#3478)
GPUDirect RDMA is not supported on WSL, but some drivers report the
capability there, and the RDMA-enabled VMM allocation in
test_vmm_allocator_policy_configuration then fails. Skip the test on
WSL; the capability checks stay as they are.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit 36bc530)
* docs(cuda.core): add 1.3.0 release note for the IPC import/release race fix (#3479)
* docs(cuda.core): add 1.3.0 release note for the IPC import/release race fix
Follow-up to #3450, which merged without a release note.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* docs(cuda.core): scope the IPC race release note to memory pool buffers
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
(cherry picked from commit de4c60d)
* Fix conditional graph lifetime tests to use supported nodes (#3469)
* Fix conditional graph lifetime tests to use supported nodes
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
* ci: require manual copy-pr-bot approval
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
---------
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
Co-authored-by: Andy Jost <ajost@nvidia.com>
(cherry picked from commit b80866b)
* fix(cuda.core): run cuMemPoolSetAccess in relaxed mode in PinnedMemoryResource (#3475)
Since #3405, PinnedMemoryResource.__init__ without options calls
cuMemPoolSetAccess to grant the current device read/write access to
the driver's default pool. The driver refuses that call, and
invalidates the capture, while a global-mode capture is active on any
thread. Users of PyTorch's torch.cuda.graph (which defaults to global
mode) therefore see CUDA_ERROR_STREAM_CAPTURE_UNSUPPORTED on
construction and CUDA_ERROR_STREAM_CAPTURE_INVALIDATED on end_building.
Fix: introduce MP_set_access in _memory_pool.pyx, wrapping
cuMemPoolSetAccess in the existing _enter_relaxed_capture_mode /
_exit_relaxed_capture_mode helpers, following the identical pattern
used by pool attribute reads and writes and trim() since #2834.
Call it from _PMR_init instead of the bare with-nogil block.
The with-options path (cuMemPoolCreate) is unaffected; that call is
still refused under global capture as documented in the 1.3.0 notes.
Add tests for the three capture combinations (global+same_thread,
global+other_thread, thread_local+same_thread) and a capture-mode
restore check modelled on test_dmr_init_restores_capture_mode.
Closes#3451
Co-authored-by: Andy Jost <ajost@nvidia.com>
(cherry picked from commit d94f19c)
* cuda.core: carry the cuda-bindings floors through minimal-sysctk12/13 extras (#3460)
* cuda.core: carry the cuda-bindings floors through minimal-sysctk extras
The floors lived only in the cu12/cu13 extras, so a package that depends
on plain cuda-core never passed them to pip or uv, and an older
cuda-bindings failed only at import (#3459).
Declare the floors in new minimal-sysctk12/13 extras, include them from
cu12/cu13, and depend on cuda-bindings without a version. The floor
parser, the CI wheel-floor reader, the pre-commit hook and the docs read
the new extras.
* cuda.core: read cu<N> floors through the minimal-sysctk self-reference
Drop the unversioned cuda-bindings from the base dependencies, so a
plain cuda-core install keeps asking for an extra.
Keep the floor readers on the cu<N> extras, as before, and teach them
to follow a cuda-core[...] self-reference and to skip an unversioned
cuda-bindings[all]. The cu<N> layout then yields the floor written in
minimal-sysctk<N>, and the existing tests of the cu<N> form stay as
they are. New tests cover the self-reference.
---------
Co-authored-by: Andy Jost <ajost@nvidia.com>
(cherry picked from commit ad38141)
---------
Signed-off-by: Ralf Juengling <rjuengling@nvidia.com>
Signed-off-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
Signed-off-by: Sainath-Reddy7 <180950590+Sainath-Reddy7@users.noreply.github.com>
Signed-off-by: Rui Luo <ruluo@nvidia.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: Ralf Juengling <rjuengling@nvidia.com>
Co-authored-by: jpascucci-nv <jpascucci@nvidia.com>
Co-authored-by: Ralf W. Grosse-Kunstleve <rgrossekunst@nvidia.com>
Co-authored-by: Ralf W. Grosse-Kunstleve <rwgkio@gmail.com>
Co-authored-by: B Sainath Reddy <bsainathreddy4@gmail.com>
Co-authored-by: Sainath-Reddy7 <180950590+Sainath-Reddy7@users.noreply.github.com>
Co-authored-by: Sebastian Berg <sebastianb@nvidia.com>
Co-authored-by: Rob Parolin <rparolin@users.noreply.github.com>
Co-authored-by: Rui Luo <ruluo@nvidia.com>
Co-authored-by: Leo Fang <leof@nvidia.com>
This file contains hidden or 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
CI/CDCI/CD infrastructurecuda.coreEverything related to the cuda.core modulepackagingAnything related to wheels or Conda packagesto-be-backportedTrigger the bot to raise a backport PR upon merge
3 participants
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
Part of #3459.
Moves the
cuda-bindingsfloors from thecu12/cu13extras into newminimal-sysctk12/minimal-sysctk13extras, whichcu12/cu13include by self-reference. A package that depends oncuda-core[minimal-sysctk12](or[cu12]) now passes the floor to pip and uv.The floor readers (
floors_from_extrasand the CI wheel-floor reader) still read the floor of eachcu<N>extra; they now follow acuda-core[...]self-reference and skip an unversionedcuda-bindings[all]. The existing tests of thecu<N>form are unchanged, and new tests cover the self-reference.Two additions beyond the issue:
check-cuda-core-bindings-floorhook also checks that eachcu<N>extra listscuda-core[minimal-sysctk<N>], so the floor is written only inminimal-sysctk<N>.[tool.uv].conflictsdeclares the new extras as conflicting across majors; without ituv lockfails.A CUDA 12 wheel built from this branch carries the floors under
extra == "minimal-sysctk12"/"minimal-sysctk13", andpip install --dry-run "cuda-core[cu12] @ <wheel>" cuda-bindings==12.9.3(likewise[minimal-sysctk12]) is now rejected at resolve time instead of failing at import.Checklist
-- Leo's bot