cuda-python / cuda_bindings_12
AGENTS.md3.4k starsChanged 3 months agoApache-2.0View on GitHub
This root maintains CUDA 12.9 bindings under the cuda.bindings namespace. Follow the repository-wide instructions, including Changes across bindings majors. Inspect the CUDA 13 counterpart during implementation and review; preserve intentional API, layout, and build differences. - Files carrying a CYTHON-BINDINGS-GENERATED seal are generated inputs. Change their generation source and regenerate against CUDA 12.9 inputs. Record generator revision, toolkit inputs, and affected files. Do not edit sealed files or recalculate their seals to conceal manual changes. - Handwritten utilities, build helpers,…
Trending
More talk lately.about the whole repository
Who's talking about NVIDIA/cuda-python
- rapidsai/rmm“Replace use of custom CUDA bindings with CUDA-Python”GitHub · Dec 3, 2021 · 12 reactions
- jungwoo-ha/WeeklyArxivTalk“[20210815] Weekly AI ArXiv 만담”GitHub · Aug 14, 2021 · 12 reactions
- A Hacker News story“CUDA Python Low-Level Bindings”Hacker News · Aug 21, 2021 · 4 points
- apache/arrow“POC: [Python] try out stubgen-pyx for auto-generated typing stubs”GitHub · Oct 1, 2026 · 3 reactions
- pytorch/pytorch“RFC: The State of Custom CUDA extensions in PyTorch”GitHub · Apr 23, 2025 · 3 reactions
The biggest 5 of 30 mentions linking to the repository that our crawl found. Each line is their own headline.
See all 30
Talked about on GitHub, Stack Overflow, Hacker News and Bluesky (30 times).
On GitHub · 26
- Dec 3, 2021Replace use of custom CUDA bindings with CUDA-Python — rapidsai/rmm12 reactions · 16 comments
- Aug 14, 2021[20210815] Weekly AI ArXiv 만담 — jungwoo-ha/WeeklyArxivTalk12 reactions · 3 comments
- Oct 1, 2026POC: [Python] try out stubgen-pyx for auto-generated typing stubs — apache/arrow3 reactions · 20 comments
- Apr 23, 2025RFC: The State of Custom CUDA extensions in PyTorch — pytorch/pytorch3 reactions · 29 comments
- Oct 6, 2026Refactor and expose publicly the Driver API layer in CCCL — NVIDIA/cccl2 reactions
21 more on GitHub
- Apr 11, 2026Allow getting `AtenTensorHandle` from the tensor object via AOTI stable C API — pytorch/pytorch2 reactions · 7 comments
- Oct 5, 2026deps: bump the python-minor group with 8 updates — pcalnon/juniper-cascor1 reaction · 1 comment
- Sep 4, 2026DiskKernelCacheBackend.load() lacks the PermissionError tolerance that save() has, causing crashes under concurrent access on Windows — cupy/cupy1 reaction · 2 comments
- Aug 10, 2026Bump the python-security-and-patches group across 1 directory with 74 updates — canfieldjuan/ATLAS1 reaction · 3 comments
- Jul 19, 2026chore(deps): bump cuda-bindings from 13.2.0 to 13.3.1 — grisuno/LazyOwn1 reaction · 1 comment
- Mar 2, 2026CuPy fails to auto-detect CUDA when CUDA_PATH is unset in containers (RawKernel / JIT compile) — cupy/cupy1 reaction · 1 comment
- Feb 26, 2026Unpin nvjitlink from ctk version — NVIDIA/numba-cuda1 reaction · 4 comments
- Feb 10, 2026[FEA] Migrate RAPIDS to CCCL `cuda::stream_ref` / `cuda::stream` in C++ and `cuda.core` streams in Python — rapidsai/rmm1 reaction · 1 comment
- Jan 7, 2026CI: Add CUDA 13.1 testing support — NVIDIA/numba-cudasays it tested it1 reaction · 23 comments
- Dec 23, 2025CI: Review/align conda test strategy — NVIDIA/numba-cuda1 reaction · 8 comments
- Sep 19, 2025RFC: Support CUDA Stream Protocol — rapidsai/rmm1 reaction · 6 comments
- Sep 19, 2025RFC: Support CUDA Stream Protocol — pytorch/pytorch1 reaction · 2 comments
- Sep 19, 2025RFC: Support CUDA Stream Protocol — cupy/cupy1 reaction · 4 comments
- Aug 21, 2025[FEA] Add SPDX checker pre-commit hook — NVIDIA/numba-cuda1 reaction
- Aug 15, 2025Add workflow to build and cleanup per PR docs previews — NVIDIA/cccl1 reaction · 21 comments
- Aug 5, 2025Add CUDA 13.0 Support across all of RAPIDS — rapidsai/build-planning1 reaction · 12 comments
- Jun 19, 2025CUB and compile_kernel composition — pytorch/pytorch1 reaction · 8 comments
- Apr 16, 2025Some PyTorch tensor functions silently change the default locale encoding — pytorch/pytorch1 reaction · 2 comments
- Jan 18, 2025[FEA]: Replace `__del__` with `weakref.finalize()` in cuda/parallel/experimental/algorithms/ — NVIDIA/cccl1 reaction · 1 comment
- Oct 25, 2024Prepare for CUDA Python new documentation layout — rapidsai/rmm1 reaction · 2 comments
- Sep 11, 2024Adopt cuda-python new module layout — cupy/cupy1 reaction · 1 comment
On Stack Overflow · 2
- May 19, 2026What is the best way to write code to my GPU's CUDA cores using python?0 votes
- Jan 12, 2024cannot run CUDA Python examples0 votes · 1 answer
On Hacker News · 1
- Aug 21, 2021CUDA Python Low-Level Bindings4 points
On Bluesky · 1
- Apr 9, 2025📦 NVIDIA / cuda-python ⭐ 1,537 (+50) 🗒 Python CUDA Python: Performance meets Productivity — @github-trending.bsky.social2 likes and reposts
Where files are headingNew AGENTS.md files have more rules about keeping secrets out of the code, and fewer about code style.
Of the AGENTS.md files we read, the 31,439 first seen in the last six months, against the older ones: the kind of rule that gained most share, and the kind that lost most. About files like this in general, not this one, and an estimate, since a kind is a pattern in the words.
In this repository9 other files it gives its agents.
- AGENTS.mdAGENTS.md
- benchmarks/cuda_bindings/AGENTS.mdAGENTS.md
- benchmarks/cuda_core/AGENTS.mdAGENTS.md
- cuda_bindings_13/AGENTS.mdAGENTS.md
- cuda_core/AGENTS.mdAGENTS.md
- cuda_core/tests/AGENTS.mdAGENTS.md
- cuda_pathfinder/AGENTS.mdAGENTS.md
- cuda_python/AGENTS.mdAGENTS.md
- create-cuda-python-pull-requestSkill
The file
We split this AGENTS.md into the rules it gives an agent
4 rules. About 0 are common, 3 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈3
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 4
Follow the repository-wide instructions, including Changes across bindings majors.
- Distinctive: line 24
Use CUDA 12.9 headers and a separate Python environment from CUDA 13.
- Very specific: line 27
Use the package's developer commands.
- Distinctive: line 30
Run the SCM preview bootstrap only in a disposable build checkout.
≈ “Common” is an estimate: at least 1 in 5 of the 40,246 AGENTS.md files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
CUDA 12 Maintenance Bindings
This root maintains CUDA 12.9 bindings under the cuda.bindings namespace.
Follow the repository-wide instructions, including
Changes across bindings majors.
Inspect the CUDA 13 counterpart during implementation and review; preserve
intentional API, layout, and build differences.
Generated and handwritten sources
- Files carrying a
CYTHON-BINDINGS-GENERATEDseal are generated inputs. Change their generation source and regenerate against CUDA 12.9 inputs. Record generator revision, toolkit inputs, and affected files. Do not edit sealed files or recalculate their seals to conceal manual changes. - Handwritten utilities, build helpers, tests, and documentation can be edited directly. Inspect headers and provenance; a filename extension alone does not identify generated code.
- Source builds parse CUDA headers and expand
.intemplates before Cython. Regenerate these outputs from their templates and headers. Preserve the CUDA 12 parser, legacy APIs, and namespace redirector.
Build and test guidance
- Use CUDA 12.9 headers and a separate Python environment from CUDA 13.
_build_shared.pylinks to the canonical helper incuda_bindings_13/. Preserve its symlink or byte-for-byte identity with the shared copies.- Use the package's developer commands. Build the Cython test
extensions before running
tests/cython/. - Legacy samples remain under
examples/; CUDA 13 uses a different layout. - Run the SCM preview bootstrap only in a disposable build checkout. Never push its synthetic tags or run it in a shared development repository.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.